# THE IMPORTANCE OF IRON IN PATHOPHYSIOLOGIC CONDITIONS

EDITED BY: Raffaella Gozzelino and Paolo Arosio PUBLISHED IN: Frontiers in Pharmacology

#### *Frontiers Copyright Statement*

*© Copyright 2007-2015 Frontiers Media SA. All rights reserved. All content included on this site, such as text, graphics, logos, button icons, images, video/audio clips, downloads, data compilations and software, is the property of or is licensed to Frontiers Media SA ("Frontiers") or its licensees and/or subcontractors. The copyright in the text of individual articles is the property of their respective authors, subject to a license granted to Frontiers.*

*The compilation of articles constituting this e-book, wherever published, as well as the compilation of all other content on this site, is the exclusive property of Frontiers. For the conditions for downloading and copying of e-books from Frontiers' website, please see the Terms for Website Use. If purchasing Frontiers e-books from other websites or sources, the conditions of the website concerned apply.*

*Images and graphics not forming part of user-contributed materials may not be downloaded or copied without permission.*

*Individual articles may be downloaded and reproduced in accordance with the principles of the CC-BY licence subject to any copyright or other notices. They may not be re-sold as an e-book.*

*As author or other contributor you grant a CC-BY licence to others to reproduce your articles, including any graphics and third-party materials supplied by you, in accordance with the Conditions for Website Use and subject to any copyright notices which you include in connection with your articles and materials.*

> *All copyright, and all rights therein, are protected by national and international copyright laws.*

*The above represents a summary only. For the full conditions see the Conditions for Authors and the Conditions for Website Use.*

ISSN 1664-8714 ISBN 978-2-88919-524-4 DOI 10.3389/978-2-88919-524-4

## About Frontiers

Frontiers is more than just an open-access publisher of scholarly articles: it is a pioneering approach to the world of academia, radically improving the way scholarly research is managed. The grand vision of Frontiers is a world where all people have an equal opportunity to seek, share and generate knowledge. Frontiers provides immediate and permanent online open access to all its publications, but this alone is not enough to realize our grand goals.

## Frontiers Journal Series

The Frontiers Journal Series is a multi-tier and interdisciplinary set of open-access, online journals, promising a paradigm shift from the current review, selection and dissemination processes in academic publishing. All Frontiers journals are driven by researchers for researchers; therefore, they constitute a service to the scholarly community. At the same time, the Frontiers Journal Series operates on a revolutionary invention, the tiered publishing system, initially addressing specific communities of scholars, and gradually climbing up to broader public understanding, thus serving the interests of the lay society, too.

## Dedication to Quality

Each Frontiers article is a landmark of the highest quality, thanks to genuinely collaborative interactions between authors and review editors, who include some of the world's best academicians. Research must be certified by peers before entering a stream of knowledge that may eventually reach the public - and shape society; therefore, Frontiers only applies the most rigorous and unbiased reviews.

Frontiers revolutionizes research publishing by freely delivering the most outstanding research, evaluated with no bias from both the academic and social point of view. By applying the most advanced information technologies, Frontiers is catapulting scholarly publishing into a new generation.

## What are Frontiers Research Topics?

Frontiers Research Topics are very popular trademarks of the Frontiers Journals Series: they are collections of at least ten articles, all centered on a particular subject. With their unique mix of varied contributions from Original Research to Review Articles, Frontiers Research Topics unify the most influential researchers, the latest key findings and historical advances in a hot research area! Find out more on how to host your own Frontiers Research Topic or contribute to one as an author by contacting the Frontiers Editorial Office: researchtopics@frontiersin.org

## **THE IMPORTANCE OF IRON IN PATHOPHYSIOLOGIC CONDITIONS**

Topic Editors:

**Raffaella Gozzelino,** Chronic Diseases Research Center (CEDOC)/Faculty of Medical Sciences, NOVA University of Lisbon, Portugal **Paolo Arosio,** University of Brescia, Italy

The iron element (Fe) is strictly required for the survival of most forms of life, including bacteria, plants and humans. Fine-tuned regulatory mechanisms for Fe absorption, mobilization and recycling operate to maintain Fe homeostasis, the disruption of which leads to Fe overload or Fe depletion.

Whereas the deleterious effect of Fe deficiency relies on reduced oxygen transport and diminished activity of Fe-dependent enzymes, the cytotoxicity induced by Fe overload is due to the ability of this metal to act as a pro-oxidant and catalyze the formation of highly reactive hydroxyl radicals via the Fenton chemistry. This results in unfettered oxidative stress generation that, by inducing protein, lipid and DNA oxidation, leads to Fe-mediated programmed cell death and organ dysfunction. Major and systemic Fe overloads occurring in hemochromatosis and Fe-loading anemias have been extensively studied. However, localized tissue Fe overload was recently associated to a variety of pathologies, such as infection, inflammation, cancer, cardiovascular and neurodegenerative disorders. In keeping with the existence of cross-regulatory interactions between Fe homeostasis and the pathophysiology of these diseases, further investigations on the mechanisms that provide cellular and systemic adaptation to tissue Fe overload are instrumental for future therapeutic approaches.

Thus, we encourage our colleagues to submit original research papers, reviews, perspectives, methods and technology reports to contribute their findings to a current state of the art on a comprehensive overview of the importance of iron metabolism in pathophysiologic conditions.

**Citation:** Gozzelino, R., Arosio, P., eds. (2015). The Importance of Iron in Pathophysiologic Conditions. Lausanne: Frontiers Media. doi: 10.3389/978-2-88919-524-4

# Table of Contents

*06 The importance of iron in pathophysiologic conditions* Raffaella Gozzelino and Paolo Arosio

## **Iron metabolism**


De-Liang Zhang, Manik C. Ghosh and Tracey A. Rouault


Christal A. Worthen and Caroline A. Enns


Esther G. Meyron-Holtz, Lyora A. Cohen, Lulu Fahoum, Yael Haimovich, Lena Lifshitz, Inbar Magid-Gold, Tanja Stuemler and Marianna Truman-Rosentsvit


Zvi Ioav Cabantchik

## **Heme-Iron**

*132 Heme in pathophysiology: a matter of scavenging, metabolism and trafficking across cell membranes*

Deborah Chiabrando, Francesca Vinchi, Veronica Fiorito, Sonia Mercurio and Emanuela Tolosano

*156 Like iron in the blood of the people: the requirement for heme trafficking in iron metabolism*

Tamara Korolnek and Iqbal Hamza

*169 Expression of ABCG2 (BCRP) in mouse models with enhanced erythropoiesis* Gladys O. Latunde-Dada, Abas H. Laftah, Patarabutr Masaratana, Andrew T. McKie and Robert J. Simpson

## **Genetic disorders**


## **Iron deficiency and anemia**


Amanda B. Core, Susanna Canali and Jodie L. Babitt

*211 Hepcidin antagonists for potential treatments of disorders with hepcidin excess*

Maura Poli, Michela Asperti, Paola Ruzzenenti, Maria Regoni and Paolo Arosio

## **Inflammation**


Carlos Penha-Gonçalves, Raffaella Gozzelino and Luciana V. de Moraes

*304 Behavioral decline and premature lethality upon pan-neuronal ferritin overexpression in* **Drosophila** *infected with a virulent form of* **Wolbachia** Stylianos Kosmidis, Fanis Missirlis, Jose A. Botella, Stephan Schneuwly, Tracey A. Rouault and Efthimios M. C. Skoulakis

## **Cardiotoxicity**

## *312 The role of iron in anthracycline cardiotoxicity*

Elena Gammella, Federica Maccarinelli, Paolo Buratti, Stefania Recalcati and Gaetano Cairo

*318 Epidemiological associations between iron and cardiovascular disease and diabetes*

Debargha Basuli, Richard G. Stevens, Frank M.Torti and Suzy V. Torti


## **Neurodegeneration**

*368 The iron regulatory capability of the major protein participants in prevalent neurodegenerative disorders*

Bruce X. Wong and James A. Duce

*378 Neurodegeneration with brain iron accumulation: update on pathogenic mechanisms*

Sonia Levi and Dario Finazzi

*398 The interplay between iron accumulation, mitochondrial dysfunction, and inflammation during the execution step of neurodegenerative disorders*

Pamela J. Urrutia, Natalia P. Mena and Marco T. Núñez


Grazia Isaya


Fatima Ali-Rahman, Cara-Lynne Schengrund and James R. Connor

*454 The role of iron in neurodegenerative disorders: insights and opportunities with synchrotron light*

Joanna F. Collingwood and Mark R. Davidson

*473 Regional siderosis: a new challenge for iron chelation therapy* Zvi Ioav Cabantchik, Arnold Munnich, Moussa B. Youdim and David Devos

## The importance of iron in pathophysiologic conditions

#### *Raffaella Gozzelino1 \* and Paolo Arosio2 \**

*<sup>1</sup> Inflammation Laboratory, Instituto Gulbenkian de Ciência, Oeiras, Portugal*

*<sup>2</sup> Department of Molecular and Translational Medicine, University of Brescia, Brescia, Italy*

*\*Correspondence: rgozzelino@igc.gulbenkian.pt; paolo.arosio@unibs.it*

#### *Edited and reviewed by:*

*Jaime Kapitulnik, The Hebrew University of Jerusalem, Israel*

**Keywords: iron, iron metabolism, iron and genetic disorders, iron deficiency and anemia, iron and inflammation, iron and cardiotoxicity, iron and neurodegeneration, heme iron**

Biological iron is necessary for vital functions and also potentially toxic to the organisms. This dual effect raised the interest of many investigators to study the mechanisms controlling its homeostasis that are altered in many pathologic conditions. Recently the understanding of iron metabolism significantly improved with the discovery of genes responsible for genetic disorders, such as hemochromatosis, the IRE/IRPs machinery and the hepcidinferroportin axis, which allowed to elucidate the basis of cellular and systemic iron homeostasis. In addition, these advances disclosed a causal link between deregulation of iron homeostasis, inflammation and oxidative stress, often induced by the iron accumulation that is commonly observed in many pathologic conditions.

Hence, believing this was time to provide a current state-ofthe-art on the importance of iron in pathophysiologic conditions, we thought to promote a Research Topic with the contribution of top-leading scientists who studied the effects of iron homeostasis disruption on the outcome of genetic, inflammatory, infectious, cardiovascular, and neurodegenerative diseases. Encountering an interest even larger than our ambitions, we successfully collected 42 manuscripts, which cover the major aspects of iron metabolism, from the essential role of iron for cell survival to its contribution in the pathogenesis of various disorders. They have been organized in an e-book in 7 sections.

The first section of the Research Topic includes 11 papers that cover the importance of iron in cellular proliferation, differentiation and functioning, and its crucial role in essential processes such as oxygen transport, DNA synthesis, metabolic energy and cellular respiration. They describe the expression and regulation of the main players involved in the mechanisms of iron absorption, recycling, and mobilization (Zhang et al., 2014), the cooperation among different cellular compartments that facilitates iron mobilization/storage and prevents the deleterious effects induced by its accumulation, the role of iron in the Fenton chemistry and its effects on oxidative stress and programmed cell death. Of interest is the study of the different types of circulating iron and the strategies more commonly used for its detection (Cabantchik, 2014). The papers also present updates on iron metabolism in zebrafish, in *C. elegans*, the role of iron in the skin and the regulatory mechanisms devoted to iron uptake, recycling and mobilization. Altogether they provide the information for a better understanding of the iron involvement in pathophysiologic conditions.

The second section includes three papers dealing with the role of heme inside cells and its cytotoxicity when it is released from hemoproteins. In fact, most body iron is contained within the protoporphyrin ring that acts as prosthetic group in many hemoproteins essential to cellular functions. The different aspects related to heme synthesis, intracellular trafficking, scavenging and catabolism are reviewed together with the protective mechanisms that cooperate to prevent the deleterious effects induced by heme accumulation and the pathological conditions in which heme plays a dominant role (Chiabrando et al., 2014). The expression and regulation of the main heme scavengers and transporters identified are also reviewed, together with the notion that maintenance of heme homeostasis is essential to prevent the deleterious effect induced by its overload (Korolnek and Hamza, 2014). The following sections are devoted to describe how disruption of iron homeostasis is associated with a series of syndromes and dictates the outcomes and severity of these disorders.

The third section deals with hemochromatosis, the genetic disease that has a fundamental importance for identifying the actors responsible for systemic iron regulation. It includes two reviews on the origin and genetic mutations characterizing this pathology as well as the incidence and different types of hemochromatosis identified so far (Vujic, 2014). The symptoms and manifestations that characterize these disorders and the alteration of the responsible proteins are also described (Silvestri et al., 2014).

The forth section deals with anemia and iron deficiency, problems that are common worldwide, and includes four papers. The distribution of iron deficiency in the population and in particular during aging are reviewed (Busti et al., 2014). The etiology of anemia, caused by genetic disorders, inflammation, infections, bleeding due to the development of ulcers, drug administration or cancers are covered. The roles of TMPRSS6 and of its substrate, hemojuvelin, in the regulation of BMP signaling and hepcidin expression are reviewed (Core et al., 2014; Wang et al., 2014). Finally, the increasing number of therapeutic approaches targeting the various steps involved in hepcidin regulation are summarized together with the promising results capable to correct altered hematological parameters in animal models (Poli et al., 2014).

The fifth section deals with one of the hottest topics of the field, which is the connection between iron and inflammation, largely mediated by hepcidin expression. Seven papers are included. One is an updated review on the known battle between host and pathogen for access to the iron necessary for proliferation, particular reference to the very complex malaria infection (Spottiswoode et al., 2014). The molecular mechanisms leading to disruption of iron homeostasis upon infections caused by parasites (Penha-Goncalves et al., 2014), intracellular or extracellular pathogens are also detailed (Nairz et al., 2014). Of importance are the effects of iron supplementation therapy to individuals suffering from infectious diseases (Clark et al., 2014) and the role of proteins that restricting iron availability to microbes may modify the outcome of the infection.

The sixth section discusses the role of iron overload in cardiovascular diseases, which includes six papers. One is a review on the long-standing and obscure relationship between iron availability and anthracycline cardiotoxicity, stressing the role of chelating agents and ferritins as agents protecting against the pro-oxidant activity of the drug (Gammella et al., 2014). A detailed overview on the pathways leading to the disruption of iron homeostasis and impairing heart functioning is described (Basuli et al., 2014). The toxicity of iron was addressed in different cell types, emphasizing in particular the effects exerted on macrophages for the development and progression of atherosclerotic plaque. In this section, a special attention is also given to the occurrence of cardiovascular abnormalities and death in hemochromatosis patients, thus further confirming the role of iron in the pathogenesis of these diseases (Vinchi et al., 2014).

Finally, the seventh section discusses the role of iron in neurodegenerative diseases and includes nine papers. Although the regulation of iron homeostasis in the brain remains rather obscure, its alteration has been observed in a variety of brain disorders, including Parkinson's, Alzheimer's, Huntington's, Prion and neurodegeneration with brain iron accumulation (NBIA). This stimulated many studies to verify if the local iron overload is one, or the main contributor to neuronal death (Wong and Duce, 2014). The pathogenic mechanisms leading to neurodegeneration that are associated to gene mutations of NBIAs are reviewed and they pose another tight connection between iron deregulation and oxidative damage (Levi and Finazzi, 2014), The involvement of inflammation in the establishment and progression of these pathologic conditions and its correlation to the disruption of iron homeostasis was also addressed (Urrutia et al., 2014). The beneficial effect of an oral iron chelator able to pass through the blood-brain barrier, the deferiprone, in scavenging excess iron from regional foci of siderosis is reviewed together with the ongoing clinical trials. The chelator is claimed to be able to relocate efficiently iron and to replenish iron-deprived regions, thus ameliorating the symptoms of iron maldistribution and suppressing the deleterious effects of its overload (Cabantchik et al., 2013).

In summary, this research topic enjoys the contribution of top-leading scientists aimed at providing a current state of the art on the importance of iron metabolism and its contribution in a variety of human disorders.

## **ACKNOWLEDGMENTS**

We would like to heartily acknowledge all the authors for the valuable sharing of their findings, knowledge, and opinions.

## **REFERENCES**


Zhang, D. L., Ghosh, M. C., and Rouault, T. A. (2014). The physiological functions of iron regulatory proteins in iron homeostasis - an update. *Front. Pharmacol.* 5:124. doi: 10.3389/fphar.2014.00124

**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

*Received: 23 December 2014; accepted: 30 January 2015; published online: 24 February 2015.*

*Citation: Gozzelino R and Arosio P (2015) The importance of iron in pathophysiologic conditions. Front. Pharmacol. 6:26. doi: 10.3389/fphar.2015.00026*

*This article was submitted to Drug Metabolism and Transport, a section of the journal Frontiers in Pharmacology.*

*Copyright © 2015 Gozzelino and Arosio. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*

## Zebrafish in the sea of mineral (iron, zinc, and copper) metabolism

## *Lu Zhao1,2, Zhidan Xia1,2 and Fudi Wang1,2\**

*<sup>1</sup> Department of Nutrition, Center for Nutrition and Health, School of Public Health, School of Medicine, Zhejiang University, Hangzhou, China <sup>2</sup> Institute of Nutrition and Food Safety, Zhejiang University, Hangzhou, China*

#### *Edited by:*

*Raffaella Gozzelino, Instituto Gulbenkian de Ciência, Portugal*

#### *Reviewed by:*

*Andrei Adrian Tica, University Of Medicine Craiova, Romania Constantin Ion Mircioiu, Carol Davila "University of Medicine and Pharmacy", Romania*

#### *\*Correspondence:*

*Fudi Wang, Department of Nutrition, Center for Nutrition and Health, School of Public Health, School of Medicine, Zhejiang University, 866 Yuhangtang Road, Hangzhou 310058, China e-mail: fwang@zju.edu.cn; fudiwang.lab@gmail.com*

Iron, copper, zinc, and eight other minerals are classified as essential trace elements because they present in minute *in vivo* quantities and are essential for life. Because either excess or insufficient levels of trace elements can be detrimental to life (causing human diseases such as iron-deficiency anemia, hemochromatosis, Menkes syndrome and Wilson's disease), the endogenous levels of trace minerals must be tightly regulated. Many studies have demonstrated the existence of systems that maintain trace element homeostasis, and these systems are highly conserved in multiple species ranging from yeast to mice. As a model for studying trace mineral metabolism, the zebrafish is indispensable to researchers. Several large-scale mutagenesis screens have been performed in zebrafish, and these screens led to the identification of a series of metal transporters and the generation of several mutagenesis lines, providing an in-depth functional analysis at the system level. Moreover, because of their developmental advantages, zebrafish have also been used in mineral metabolism-related chemical screens and toxicology studies. Here, we systematically review the major findings of trace element homeostasis studies using the zebrafish model, with a focus on iron, zinc, copper, selenium, manganese, and iodine. We also provide a homology analysis of trace mineral transporters in fish, mice and humans. Finally, we discuss the evidence that zebrafish is an ideal experimental tool for uncovering novel mechanisms of trace mineral metabolism and for improving approaches to treat mineral imbalance-related diseases.

**Keywords: zebrafish, trace elements, minerals, iron, copper, zinc, metabolism**

## **INTRODUCTION**

As children of the Earth, humans are intimately connected to our surroundings in many ways, and the relationship between humans and minerals is perhaps the most enigmatic. Based on their necessity for life and their limited quantities with the human body, 11 elements are classified as trace minerals, including iron (Fe), zinc (Zn), copper (Cu), selenium(Se), manganese(Mn), iodine(I), molybdenum(Mo), fluorine (F), cobalt (Co), chromium (Cr), and vanadium (V) (Fraga, 2005).

Metal trace minerals are biologically active primarily as metalloproteins formed by conjugating or binding with various protein partners. Metalloproteins account for approximately half of all proteins and perform a wide range of biological functions as enzymes, transporters and signal transducers. In metalloproteins, metal trace minerals are essential components, acting at the enzyme's active site or by stabilizing the protein's structure. Trace mineral deficiencies can cause a number of diseases that can be mild, severe, or even fatal. Conversely, excess levels of trace minerals can be toxic. For example, excess iron or copper produces reactive oxygen species (ROS) via the Fenton reaction, resulting in lipid peroxidation, DNA damage, altered calcium homeostasis, and cell death (Stohs and Bagchi, 1995). Excess levels of redoxinactive metals such as zinc are also harmful; the accumulation of zinc triggers neuronal death in the brain and induces copper deficiency. Thus, the balance of endogenous trace minerals must be tightly regulated.

Organisms have evolved comprehensive systems for maintaining trace element homeostasis; these systems are composed primarily of transport proteins, storage proteins, and some hormones. Our current knowledge regarding these regulatory factors has come primarily from studies using model organisms ranging from yeast to mice. Among these species, the zebrafish (*Danio rerio*) has been a valuable vertebrate system with several unique advantages. First, their small size, high fertility rate, and rapid development make zebrafish an ideal model for large-scale genetic screens. Secondly, because they are fertilized *ex vivo* and are optically transparent, zebrafish embryos are ideally suited for experimental techniques such as gene knockout/knockdown and overexpression. Because the embryos develop *ex utero*, zebrafish are also an excellent model for studying pharmacology and toxicology in early developmental stages. Finally, the zebrafish is a vertebrate species, and many of its genes and metabolic systems are highly conserved with humans; indeed, 80% of genes and expressed sequence tags (ESTs) are present in conserved synteny groups between fish and humans (Barbazuk et al., 2000).

Here, we systematically review the major findings obtained from zebrafish studies of trace element homeostasis, with a focus on iron, zinc, copper, selenium, manganese, and iodine. We also performed a homology analysis of trace mineral transporters in fish, mice and humans, and we summarize the available zebrafish mutant models in the field. This review demonstrates that zebrafish are an ideal experimental tool for investigating novel mechanisms of trace mineral metabolism and for improving therapeutic approaches for treating mineral imbalance-related diseases.

## **ZEBRAFISH AND IRON METABOLISM**

## **OVERVIEW OF IRON METABOLISM**

Iron is present in nearly all living organisms. As an essential component of heme and iron-sulfur cluster–containing proteins, iron plays a central role in many biological activities, including oxygen transport, cellular respiration, and DNA synthesis (Muckenthaler and Lill, 2012). Of all the trace elements, the iron homeostasis system is one of the best characterized, primarily because of iron's role in erythropoiesis and its causative relationships with irondeficiency anemia and hematochromatosis. To date, many major proteins involved in the uptake, transport, storage and release of iron have been identified (**Figure 1**).

Under normal conditions, dietary iron is absorbed by enterocytes through Divalent Metal Transporter 1 (DMT1); from there, it is exported to the circulation through Ferroportin 1 (Fpn1). In the blood, iron is transported in the form of Transferrin (Tf)-Fe3+, which is taken up by endocytosis into cells with surface Transferrin receptors (TfRs). Iron in the endosomes is then released to the cytoplasm and delivered to the mitochondria, where it is used to make iron-sulfur (Fe-S) clusters, to synthesize heme, or to be stored as Ferritin. Most of the iron used for producing hemoglobin in erythrocytes is obtained from the recycling iron pool released from senescent red blood cells that are phagocytized by macrophages. Aside from transport and storage proteins, Hepcidin—a peptide hormone released by the liver—plays an important role in regulating iron levels by binding to Fpn1 and promoting its internalization. Other factors such as oxidoreductases [e.g., Duodenal Cytochrome b (Dytb), Ceruloplasmin (Cp), Hephaestin (Heph), and STEAP3) and modulatory proteins (e.g., Hemochromatosis (HFE), Hemojuvelin (HJV), Iron Regulatory Protein (IRP) 1/2, and Transmembrane Serine Protease 6 (TMPRSS6)] also play an active role in iron metabolism (Muckenthaler and Lill, 2012; Srai and Sharp, 2012). An overview of the protein homology and expression patterns among fish and mammalian iron-regulating proteins is provided in **Table 1**.

**FIGURE 1 | Generalized overview of iron metabolism in vertebrate cells.** Dietary iron is absorbed by enterocytes through the concerted activity of the reductase DCYTB and the transporter DMT1. Iron is then oxidized by HEPH and exits the enterocytes through the iron exporter FPN1. Iron is transferred as a complex with Transferrin (TF) in the bloodstream and is delivered to target cells that express Transferrin receptors (TFRs) on their plasma membrane. TF-Iron-TFR complexes are then endocytosed. In the endosome, iron is released from TF by STEAP3 and then transported out of the endosome through DMT1. The cytoplasmic iron then enters the labile iron pool and is delivered by MFRN and siderophores to the mitochondria to be used for the synthesis of heme and Fe-S clusters. Excess iron is stored in Ferritin. Iron leaves the cell through FPN1, the plasma expression of which is negatively regulated by Hepcidin. Proteins for which zebrafish knockout and/or knockdown models are available are written in red.



#### **Table 1 | Continued**


*efth1a.*

*<sup>f</sup> fth1b.*

*H, human; M, mouse; Z, zebrafish; N.D., Not determined; YSL, yolk syncytial layer; KC, Kupffer cells; CNS, central nervous system.*

Zebrafish absorb waterborne iron via the gastrointestinal tracts and the gills. The fish branchial iron uptake has high- and low-affinity components, with Km of 5.9 nmol/l Fe and Vmax of 2.1 pmol/g·h at low Fe concentration (*<*40 nmol/1), and a linear manner increase of the uptake rate at higher Fe concentration (40–200 nmol/1) (Bury and Grosell, 2003). Zebrafish branchial iron transport can be inhibited by high level of Cd, but not by other divalent metals such as Zn, Cu, and Mn (Bury and Grosell, 2003). Moreover, low iron diet fed zebrafish exhibited a significant increase in tissue Cd accumulation, suggesting an interaction between Fe and Cd assimilation in fish (Cooper et al., 2006).

## **ZEBRAFISH MODELS OF IRON METABOLISM IDENTIFIED USING FORWARD GENETIC SCREENS**

The Nüsslein-Volhard lab performed two large-scale *N*-ethyl-*N*nitrosourea (ENU) mutagenesis screens in zebrafish and identified several thousand mutations that affect various aspects of early development (Haffter et al., 1996). Using this strategy with an anemic phenotype as the screening output, several mutant zebrafish lines with defects in iron metabolism have been established; these lines are summarized in **Table 2**.

## *Ferroportin 1*

Although researches had long suspected that an iron exporter is present in enterocytes, this was only confirmed in 2000 by studies performed in the zebrafish mutant *weissherbst* (*weh*) (Donovan et al., 2000). The *weh* mutant line was originally isolated from the 1996 Tübingen screen (Haffter et al., 1996) and was subsequently found to have severe hypochromic anemia phenotypes, including decreased hemoglobin levels, blocked erythroid maturation, and reduced numbers of erythrocytes (Ransom et al., 1996; Donovan et al., 2000). Interestingly, mutant embryos have significantly lower iron levels in their erythroid cells compared with wild-type fish, suggesting a circulatory iron deficiency. Indeed, the reduced


*\*Knockdown (KD) model.*

hemoglobin level in *weh* mutants can be rescued by intravenous iron-dextran injections, demonstrating that their hypochromia is caused by inadequate iron in the blood (Donovan et al., 2000). To identify the precise location of the gene mutation, chromosomal walking was performed and revealed a premature stop codon in a novel gene named *ferroportin1* (*fpn1*), consistent with a loss-offunction mutation. Importantly, overexpressing *fpn1* in mutant embryos rescued the hypochromia phenotype, suggesting that the *fpn1* gene is causally linked to the disease. *Fpn1* transcripts are present specifically in the zebrafish yolk syncytial layer (YSL), between the developing hematopoietic cells in the intermediate cell mass and the yolk, which contains iron and other nutrients essential for early embryonic development. This specific expression pattern of *fpn1*, together with the iron-deficiency phenotypes observed in *weh* mutants, suggests that the function of the fpn1 protein is to export iron from the yolk into the embryonic circulation. This hypothesis was confirmed by performing an iron efflux assay in a *Xenopus* oocyte expression system, in which oocytes expressing *fpn1* had increased iron efflux (Donovan et al., 2000). Moreover, both mice and humans have homologs of *FPN1* that are highly conserved with the fish *fpn1*, and mammalian *FPN1* is robustly expressed in the placenta, duodenum, and liver, all of which are major sites of iron transport. At the protein level, human FPN1 is concentrated at the basal surface of the syncytiotrophoblasts in the placenta, an organ that is functionally similar to the zebrafish YSL, indicating that human FPN1 plays a role in maternal-fetal iron export. In mice, *Fpn1* is expressed at the basolateral surface of enterocytes, suggesting a role as an intestinal iron transporter (Donovan et al., 2000). This study serves as a prime example as how genetic screens in zebrafish can lead to the identification of essential novel genes. Nevertheless, Fpn1 remains the only iron exporter that has been identified in all eukaryotic organisms.

Shortly after these findings in the zebrafish *weh* mutants were reported (Donovan et al., 2000), two groups independently cloned *Fpn1* (also called *Ireg1* or *Mtp1*) from mouse duodenal epithelial cells and a mouse mRNA library (Abboud and Haile, 2000; McKie et al., 2000). Both studies confirmed the essential role of Fpn1 in iron export and the regulation of iron homeostasis. Mutations in *SLC11A3*, the human homolog of *FPN1*, were later identified as causing autosomal dominant hemochromatosis, a disorder characterized by iron overload and multiple organ damage (Montosi et al., 2001; Njajou et al., 2001).

The effect of Fpn1 deficiency on the adult zebrafish system was examined in further detail. Although *weh* homozygotes are embryonic lethal and usually die 7–14 days post-fertilization (dpf), repeated intravenous injections of iron-dextran enables the mutants to reach adulthood (Donovan et al., 2000; Fraenkel et al., 2005). These rescued fish are normal until 6 months of age, but develop hypochromic blood by 12 months. Compared with ironinjected wild-type fish, the rescued mutants had increased iron staining in the kidney macrophages at 12 months of age, as well as increased staining in the intestinal villi at 6 and 12 months, suggesting that the *fpn1* mutation impairs iron export in these tissues. The iron-rescued *weh* mutants also has hepatic iron overload, with particularly high iron levels in the liver Kupffer cells (Fraenkel et al., 2005). The role of Fpn1 in iron mobilization from enterocytes, hepatocytes, and macrophages was confirmed by studies using adult tissue-specific *Fpn1* knockout mice (Zhang et al., 2011, 2012). Hepcidin, a peptide hormone secreted by the liver, was found to regulate iron by triggering the internalization of Fpn1 and inhibiting iron efflux (Nemeth et al., 2004). Hepcidin is also conserved in fish, and injecting zebrafish embryos with iron causes a significant increase in endogenous *hepcidin* expression. Moreover, a similar iron-stimulated increase in *hepcidin* expression occurs in *weh* mutant embryos, suggesting that *hepcidin* expression is independent of Fpn1's normal function as an iron exporter (Fraenkel et al., 2005).

## *Dmt1*

The protein DMT1 (also called DCT1 and Nramp2), which contains 12 transmembrane domains, was originally isolated in the rat duodenum as a divalent ion transporter that is upregulated by dietary iron deficiency (Gunshin et al., 1997). Shortly after its discovery, two mammalian hypochromic anemia models the *mk*/*mk* mouse and the Belgrade rat—were found to carry DMT1 mutations (Fleming et al., 1997, 1998). The *chardonnay* (*cdy*) zebrafish mutant revealed a conserved role of DMT1 in zebrafish iron metabolism (Donovan et al., 2002). The *cdy* mutant is a hypochromic, microcytic anemia model with reduced hemoglobin expression and delayed erythrocyte maturation. The *cdy* mutant has a premature stop codon in the zebrafish homolog of *dmt1*, resulting in a severely truncated protein. The zebrafish homolog of DMT1 is 73% identical to human and mouse DMT1 homologs, and its transcripts are concentrated both in erythroid cells and in the intestine. Direct evidence of the role of zebrafish DMT1 in iron transport came from experiments using a mammalian cell line; cells that overexpressed wild-type zebrafish *dmt1* took up nearly 10 times the amount of iron as non-transfected control cells, whereas the truncated protein produced by the *cdy* mutation was non-functional. Interestingly, unlike the *weh* mutants, which die during early development (Donovan et al., 2000), the anemic *cdy* homozygotes survive and reach adulthood. The viability of *cdy* fish may be attributed to additional pathways for iron absorption (Donovan et al., 2002). Of clinical relevance, the first identified human mutation in *DMT1* was reported to cause symptoms that include severe hypochromic microcytic anemia and iron overload (Mims et al., 2005). One possible explanation for the excess iron in these patients is the presence of an alternate iron absorption route in the duodenum, thereby bypassing DMT1 (Mims et al., 2005).

## *Tfr1*

Transferrin receptor 1 (Tfr1) is a membrane-bound protein that facilitates iron uptake by binding to the iron carrier Transferrin. Tfr1 was identified as being essential for erythropoiesis and embryonic development in a *Tfr1*-knockout mouse model (Levy et al., 1999). *Tfr1*−*/*<sup>−</sup> mice develop anemia, have retarded growth and neurological defects, and die during embryogenesis (Levy et al., 1999). Four different zebrafish *chianti* (*cia*) mutants were identified with various degrees of hypochromic anemia and defective erythroid differentiation (Haffter et al., 1996; Wingert et al., 2004), and positional cloning revealed that *cia* alleles are missense and splicing mutations. During early development, *tfr1a* transcripts are expressed specifically in erythrocytes. Importantly, cytoplasmic delivery of iron by microinjection at the 1-cell stage—but not intravenous iron injections—can rescue the hypochromia phenotypes of *cia* mutants, indicating that the *tfr1a* mutation prevents erythrocytes from taking up and utilizing circulating iron (Wingert et al., 2004). Interestingly, while cloning *tfr1a*, a second *tfr1*-like gene, *tfr1b*, was also identified (Wingert et al., 2004). This gene duplication phenomenon in zebrafish is believed to have occurred as an evolutionary genetic event in teleosts (Amores et al., 1998; Postlethwait et al., 1998). The *tfr1b* gene is expressed ubiquitously throughout embryogenesis. Notably, although overexpressing *tfr1b* partially rescues the anemic phenotypes of *cia* mutants, *tfr1b* morphants (animals in which the gene has been knocked down by injecting morpholino antisense oligonucleotides) have normal hemoglobinization. Nevertheless, *tfr1b* morphants have retarded growth and develop brain necrosis, a phenotype that is similar to the neurologic defects observed in the mouse model (Levy et al., 1999), indicating that *tfr1b* may be involved in iron uptake through non-erythroid tissues (Wingert et al., 2004). Thus, the combined phenotypes of *tfr1a* and *tfr1b* deficient zebrafish embryos appear to recapitulate the entire phenotypic spectrum of *Tfr1*−*/*<sup>−</sup> mice. Therefore, the *cia* mutant zebrafish is an ideal model for studying the function of *tfr1* in erythropoiesis without the complication of other developmental abnormalities.

## *Grx5*

The *shiraz* (*sir*) zebrafish mutants were originally isolated from the Tübingen 2000 screen consortium; these mutants were later identified as a typical hypochromic anemia disease model with a deletion in the *glutaredoxin 5* (*grx5*) gene which encodes an antioxidant protein (Wingert et al., 2005). Functional studies of *grx5* in *sir* zebrafish revealed a novel connection between heme biosynthesis and iron-sulfur (Fe-S) cluster formation, two primary functions of iron that were previously believed to be independent processes in vertebrates.

Studies in yeast revealed that *GRX5* is required for the mitochondrial synthesis of Fe-S clusters (Rodriguez-Manzaneque et al., 2002). Similar to the yeast GRX5, zebrafish grx5 is also localized primarily in the mitochondria, and expression of the zebrafish *grx5* gene can rescue a *GRX5*-deficient yeast strain, suggesting that the function of *grx5* is evolutionarily conserved. However, the *sir* zebrafish mutants have a hypochromic anemia phenotype, with no changes in their mitochondrial iron content or oxidative stress level (Wingert et al., 2005). With respect to iron metabolism, one key difference between yeast and higher eukaryotes is that in the latter, iron regulatory proteins 1 and 2 (IRP1/2) control intracellular iron levels by binding to Iron Response Elements (IREs) in the 5- -UTR of target gene transcripts, thereby blocking their translation. Importantly, the IRE-binding capacity of IRP1 is negatively regulated by Fe-S clusters. Thus, a possible explanation for the anemic phenotype in *sir* zebrafish mutants is that the *grx5* mutation reduces Fe-S assembly, inappropriately triggering IRP1 activity, which then inhibits the expression of select target genes that are critical for heme biosynthesis. In support of this hypothesis, red blood cells in *sir* zebrafish mutants lack aminolevulinate synthase 2 (ALAS2), the first enzyme in the heme biosynthesis pathway. Moreover, overexpressing an *ALAS2* gene in which the IRE is deleted rescues hemoglobin production in *sir* mutants; in contrast, overexpressing the wild-type *ALAS2* gene does not rescue hemoglobin production. Interestingly, knocking down the expression of IRP1 also rescues the *sir* embryonic phenotype. These compelling results strongly suggest that heme synthesis in vertebrates is regulated via Fe-S cluster levels (Wingert et al., 2005). A conserved role for GRX5 in regulating heme synthesis was additionally confirmed in human patients (Camaschella et al., 2007). The findings obtained from *sir* mutants serve to highlight the advantages of using zebrafish as a vertebrate model system for discovering mechanisms that are not necessarily conserved in lower organisms.

## *Mitoferrin*

The role of Mitoferrin (Mfrn) in mitochondrial iron uptake was originally discovered in yeast studies (Foury and Roganti, 2002). MRS3 and MRS4, the yeast homologs of Mfrn, increase the efficiencies of both heme formation and Fe-S biosynthesis, the two key mitochondrial processes that utilize iron (Muhlenhoff et al., 2003). Studies of *frascati* (*frs*) zebrafish mutants further support Mfrn's role as a principal mitochondrial iron importer in vertebrate erythroblasts. *Frs* mutants develop defects such as hypochromic anemia and erythroid maturation arrest (Shaw et al., 2006), and positional cloning identified missense mutations in the *mfrn* gene in all mutant lines. Importantly, overexpressing *mfrn* in *frs* mutants rescued the erythropoiesis deficiency in half of the injected animals, and *mfrn* knockdown morphants mimicked the mutant phenotype, suggesting that *mfrn* is the diseasecausing gene in *frs* mutants. Expression array analysis revealed that *mfrn* is highly expressed in the intermediate cell mass (the tissue in which erythropoiesis occurs for early embryos), further supporting *mfrn*'s role in erythroid heme synthesis. Similar to yeast *MRS3/4*, when overexpressed in transfected mammalian cell lines, zebrafish *mfrn* localizes to the mitochondria (Shaw et al., 2006). To investigate the function of mammalian *Mfrn* further, an *Mfrn*−*/*<sup>−</sup> mouse hematopoietic cell line was established. These cells have impaired terminal erythroid maturation and an inability to incorporate iron into heme proteins. Furthermore, mouse *Mfrn* rescues the phenotype in zebrafish *frs* mutants, and fish *mfrn* restores the activity of an *MRS3/4*-deficient yeast strain (Shaw et al., 2006). Taken together, these results suggest that Mfrn's role in mitochondrial iron uptake is evolutionarily conserved among eukaryotes.

## *Transferrin-a*

Zebrafish *gav* mutant strains carry mutations in their *transferrina* (*tf-a*) gene, which encodes the principal iron carrier in all vertebrate organisms (Fraenkel et al., 2009). *Gav* mutants develop severe hypochromic anemia. Importantly, this phenotype can be phenocopied by injecting *tf-a* morpholinos into embryos, and it can be rescued by overexpressing *tf-a*. Together, these findings confirm that *tf-a* is the disease-causing gene. Homozygous *gav* mutants are generally embryonic lethal and die at approximately 14 dpf (Fraenkel et al., 2009). In humans, genetic mutations in transferrin cause congenital hypotransferrinemia, a rare disease with features that are strikingly similar to the phenotype in *gav* fish, including hypochromic anemia and premature death (Hayashi et al., 1993; Goldwurm et al., 2000). Thus, the zebrafish *gav* mutant serves as an ideal model for studying human hypotransferrinemia. In addition, Fraenkel and colleagues examined *hepcidin* expression in the *gav* mutant, as well as several previously established zebrafish mutants with iron metabolism defects. In 2-dpf zebrafish embryos (the stage in which endogenous *hepcidin* expression is least affected by environmental stimuli), the number of *hepcidin* transcripts was measured in various mutants and morphants either with or without iron injection. The results suggest that both *tf-a* and *tfR2* are required for hepcidin expression, whereas *tfR1a* and *dmt1* are required for increasing *hepcidin* expression in response to iron loading (Fraenkel et al., 2009). The primary role of Tf in driving *hepcidin* expression is further supported by studies performed using a mouse model of hypotransferrinemia (Bartnikas et al., 2011).

## **ZEBRAFISH AS A REVERSE GENETICS TOOL IN IRON METABOLISM STUDIES**

In addition to helping identify novel iron transporters and metabolic mechanisms via forward genetic screens, zebrafish have also been used as a reverse genetics tool for increasing our understanding of the iron homeostasis system.

## *Gene knockdown/knockout*

Gene knockdown/knockout techniques are used as a primary step in investigating the unknown biological functions of a given gene. In this approach, antisense morpholino oligonucleotides bind to the targeted gene transcript, thereby inhibiting the gene's expression by blocking the initiation of translation or by modifying pre-mRNA splicing. Because zebrafish develop *ex utero*, microinjection-mediated gene suppression is used widely among zebrafish researchers.

*Hd.* Studies performed using hd zebrafish morphants revealed a novel role for Huntingtin (Htt, a protein linked to Huntington's disease) in the utilization of iron by erythrocytes (Lumsden et al., 2007). Although it has been known for more than two decades that Huntington's patients carry an expanded CAG repeat in the coding region of the HD gene (Huntington's Disease Collaborative Research Group, 1993), the normal function of Htt remains unclear. In order to explore the biological roles of HD, hd knockdown morphant zebrafish were created (Lumsden et al., 2007). Interestingly, in addition to neurological defects such as brain necrosis, the hd knockdown fish also develop blood hypochromia (characterized by reduced red pigments in the erythrocytes) and decreased hemoglobin staining. Nevertheless, Prussian blue staining revealed that hd morphants have normal iron levels in their red blood cells, suggesting that Tf-Tfr–mediated iron transport and endocytosis are intact in the erythrocytes. Intriguingly, Prussian blue only stains ferric iron that is bound to Tf, but does not stain ferrous iron in hemoglobin. Thus, the hypochromia observed in the hd morphants is likely due to defects that are downstream of Tf-Tfr endocytosis. This hypothesis was tested by injecting iron-dextran directly into the cytoplasm at the 1-cell stage, which circumvents the Tf-Tfr–mediated iron cycle. Using this approach, the hypochromia defects in hd morphants were largely rescued (Lumsden et al., 2007). This important study revealed an interesting role for Htt in making endocytosed iron available for use by the cell; however, the detailed mechanism by which Htt regulates iron release and/or downstream utilization must be explored further.

*Bdh2.* Although the vast majority of intracellular iron is bound by proteins, a small amount of cytoplasmic iron is bound to low-molecular-weight carriers called siderophores, forming a labile iron pool (Breuer et al., 2008). Enterobactin is a classic bacterial siderophore that binds to Lipocalin 24p3, an irontrafficking protein that functions as the iron-chelating moiety (Yang et al., 2002). In mouse cell cultures, 2,5-dihydroxybenzoic acid (2,5-DHBA) is the iron-binding moiety of the 24p-associated mammalian siderophore, the synthesis of which is catalyzed by 3-hydroxybutyrate dehydrogenase type 2 (Bdh2), a dehydrogenase/reductase family member (Devireddy et al., 2010). In Bdh2-knockdown mouse cells, the intracellular siderophore was depleted, and the cells accumulated abnormally high amounts of free cytoplasmic iron, resulting in elevated levels of ROS. Notably, these cells were also deficient in mitochondrial iron, suggesting that siderophores also participate in the transport of iron from the cytoplasm to the mitochondria. Importantly, bdh2 zebrafish morphants develop hypochromic blood and have reduced hemoglobin levels—but normal globin expression—confirming a defect in mitochondrial heme synthesis (Devireddy et al., 2010). These findings demonstrate that the function of siderophores in regulating intracellular iron homeostasis is conserved from bacteria to vertebrates. In this respect, zebrafish are a convenient model for confirming and extending findings obtained from studying mammalian cell cultures.

*Arhgef3.* Genome-wide association and meta-analysis studies have identified more than 100 independent genetic loci associated with erythrocytes and platelets (Ganesh et al., 2009; Soranzo et al., 2009). Because of its advantages with respect to reverse genetics, the zebrafish model was used to investigate the biological functions of several candidates identified from the meta-analysis. This screen revealed that Rho guanine nucleotide exchange factor 3 (Arhgef3) plays an unexpected role in regulating iron uptake and driving erythroid cell maturation (Serbanovic-Canic et al., 2011). Silencing arhgef3 expression in zebrafish disrupts erythroid differentiation and causes hypochromic erythrocytes, which are indicative of iron-deficiency anemia. Indeed, cytoplasmic iron supplementation significantly rescues the hemoglobinization phenotype in arhgef3 morphants. Moreover, disrupting the arhgef3 target RhoA produces a phenotype that is similar to arhgef3 morphants, and this can also be rescued by cytoplasmic iron injection. The concerted roles of Arhgef3 and RhoA in regulating the internalization of membrane-bound Tf was supported by studies in a human cell line (Serbanovic-Canic et al., 2011).

*In vivo validation of cis-regulatory elements in mfrn1 using transgenic fish.* In erythroblasts, Mfrn1, and Mfrn2 are solute carriers that import cytoplasmic iron into the mitochondria for heme and Fe-S cluster biogenesis (Shaw et al., 2006; Paradkar et al., 2009). In mice, both the Mfrn1 and Mfrn2 genes contain CpG-rich promoter regions. The cis-regulatory modules (CRMs) in Mfrn1 form a chromatin immunoprecipitation dataset for GATA-1, the primary erythroid transcription factor (Cantor and Orkin, 2002), suggesting that Mfrn1 is transcriptionally regulated by GATA-1 via binding at CRM regions. Though quite compelling, these bioinformatics results still needed to be validated functionally at the systemic level, and transgenic fish were a convenient tool for this purpose. Mfrn1 transcripts are concentrated primarily in hematopoietic tissues, whereas mfrn2 transcripts are expressed throughout the central nervous system and in somites (Shaw et al., 2006). Transgenic zebrafish expressing GFP-tagged mouse Mfrn1 or Mfrn2 promoter sequences were generated to study the expression of these genes. When expressed in fish, the Mfrn2 promoter is expressed in a pattern similar to the endogenous mfrn2 expression pattern, suggesting that the promoter functions appropriately. However, the tagged Mfrn1 promoter failed to drive detectable GFP expression in fish, suggesting the need for other transcriptional regulatory elements (Amigo et al., 2011). Interestingly, injecting two of the three predicted mouse Mfrn1 CRMs yielded transient transgenic fish that expressed GFP in the same tissues as endogenous mfrn1. The specific expression pattern was refined further in transgenic fish carrying a construct that contains a CRM linked with the CpG-rich promoter. Thus, the critical role of CRMs in regulating Mfrn1 expression has been demonstrated in vivo (Amigo et al., 2011). Moreover, the critical role for GATA-1 in regulating Mfrn1 transcription was confirmed by the finding that the Mfrn1-specific expression pattern was abolished after the GATA-1 core binding sites were mutated in the CRMs of the transgenic fish. Finally, the expression of endogenous mfrn1 was also markedly reduced in GATA-1 zebrafish morphants (Amigo et al., 2011). This seminal study supports the high value of using zebrafish transgenics to complement and validate findings obtained from in silico analyses.

## *Gene overexpression*

The *ex utero* development of zebrafish also enables researchers to drive gene overexpression through microinjection. Importantly, the biological mechanisms that underlie dominant-negative gene mutations in mammals can be examined readily in zebrafish using gene overexpression techniques. In humans, *FPN1* mutations have been linked to a form of autosomal dominant hemochromatosis (Pietrangelo, 2004). Interestingly, two distinct clinical phenotypes have been characterized: one phenotype includes iron accumulation in macrophages, low transferrin saturation, and iron-limited erythropoiesis, whereas the other phenotype includes iron accumulation in hepatocytes and high transferrin saturation. The diversity of these clinical traits can be explained by the different natures of the underlying *FPN1* mutations (De Domenico et al., 2006). Mutations that cause a defect in FPN1's cell-surface localization or iron export capacity cause iron loading in macrophages, whereas mutations that impair FPN1's sensitivity to Hepcidin (thus impeding FPN1 incorporation) cause iron accumulation in hepatocytes (De Domenico et al., 2005, 2006; Schimanski et al., 2005). Nevertheless, the clinical complexity of the long-term disease process makes it difficult to determine the precise nature of a given *FPN1* mutation. Overexpressing mutant *FPN1* alleles in zebrafish is a high-throughput approach for identifying the functional effects of many mutations (De Domenico et al., 2007). The cDNA of wild-type mouse *Fpn1* or previously identified human and mice *Fpn1* mutants was injected into 1 cell stage zebrafish embryos. Expressing either wild-type *Fpn1* or the N144H *Fpn1* mutant (a Hepcidin-irresponsive protein) had no effect on endogenous hemoglobinization. However, overexpressing the H32R *Fpn1* mutant (which has defective Fpn1 membrane localization) or the N174I *Fpn1* mutant (which is transport-deficient) led to severe defects in hemoglobin synthesis, and these hemoglobinization defects were rescued by intravenous iron injections, suggesting the existence of iron-deficiency erythropoiesis in these mutant-expressing embryos (De Domenico et al., 2007). This study demonstrates nicely that the functional consequences of mammalian *Fpn1* mutations can be studied rapidly and effectively, and it supports the important role of zebrafish as a valuable vertebrate model in functional studies of dominant-negative mutations.

## **ZEBRAFISH AND ZINC METABOLISM**

## **OVERVIEW OF ZINC METABOLISM**

After iron, zinc is the second-most abundant trace mineral in humans. Zinc is essential for the activity of more than 300 enzymes and for maintaining the structural integrity of nearly 2000 transcription factors. Thus, zinc plays a critical role in cellular homeostasis, the immune response, oxidative stress, apoptosis, and aging (Stefanidou et al., 2006; Prasad, 2012). Zinc deficiency can cause a wide range of clinical defects, including growth retardation, hypogonadism, rough skin, weakened immunity, and neurosensory and cognitive disorders (Prasad, 2012). Although zinc is a redox-inactive metal, it can be toxic (albeit less toxic than iron and copper), and both acute and chronic forms of zinc poisoning have been reported to cause hematopoietic abnormalities, altered lipoprotein metabolism, and impaired immune function (Stefanidou et al., 2006).

*In vivo*, zinc homeostasis relies primarily on the zinc transporter family, which in mammals contains 10 Zinc Transporter proteins (ZnTs, or SLC30) and 14 Zrt- and Irt-like proteins (ZIPs, or SLC39) (Huang and Tepaamorndech, 2013; Jeong and Eide, 2013). ZnTs are zinc exporters that facilitate the efflux of zinc from cells and/or into intracellular vesicles, whereas ZIPs increase intracellular zinc concentration by driving the uptake of extracellular zinc and/or the release of vesicular zinc into the cytoplasm. The concerted actions of ZnTs and ZIPs maintain the balance of intracellular zinc and deliver zinc to its appropriate protein partners (**Figure 2**). Metallothioneins (MTs) are a group of lowmolecular-weight metal-binding proteins that also have a high affinity for binding zinc. MTs play an important regulatory role in zinc metabolism, possibly by competing with—or supplying zinc to—a variety of transporter proteins (Vasak and Hasler, 2000; Chasapis et al., 2012).

## **ZINC HOMEOSTASIS IS HIGHLY CONSERVED BETWEEN ZEBRAFISH AND MAMMALS**

The system that regulates zinc metabolism is highly conserved between zebrafish and mammals. The zebrafish *zip1* gene was cloned from the zebrafish gill, an ion-transporting epithelium that absorbs minerals from the surrounding water (Qiu et al., 2005). The Zip1 protein is conserved both structurally and functionally with its mammalian homologs. *Zip1* transcripts are expressed ubiquitously in zebrafish embryos, with the highest expression in the ovaries. As with human *ZIP1* (Gaither and Eide, 2001), overexpressing zebrafish *zip1* significantly increases zinc uptake. Interestingly, because Zip1 does not increase zinc influx at high extracellular zinc concentrations, it has no effect on the maximum endogenous rate of zinc uptake (Qiu et al., 2005). This ceiling effect might be due to the fact that zinc regulates the cellular localization of Zip1. Studies in mice suggest that the extracellular zinc concentration mediates the cellular localization of Zip1 through an endocytosis-mediated pathway (Wang et al., 2004). A second zinc importer in teleosts, *zip2*, was cloned from the gill of a pufferfish species (*Takifugu rubripes*) and plays a role in mediating zinc uptake (Qiu and Hogstrand, 2005).

A systematic bioinformatics data-mining approach identified the zebrafish zinc transporter genes from two previously released zebrafish databases (Zv4 Ensembl 31 and Zv5 Ensembl 34), and these genes were phylogenetically assigned to mammalian orthologs (Feeney et al., 2005). To date, eight ZnT members (ZnT1, ZnT2, and ZnT4–9) and 11 ZIP members (ZIP1, ZIP3, ZIP4, ZIP6–11, ZIP13, and ZIP14) have been identified in zebrafish. Interestingly, the teleost ZIP8 sequence differs from mammalian ZIP8 orthologs to a larger degree than other ZIP genes. Studies of gene expression patterns revealed that the ovaries and intestine—the two organs that have the most dynamic nutrient metabolism—have the highest expression of zinc transporters (Feeney et al., 2005). The expression level of each zinc transporter has been examined during zebrafish embryogenesis under normal maternal zinc conditions (Ho et al., 2012). The results showed that despite a relatively constant level of endogenous zinc during embryonic development from fertilization through 120 h post-fertilization (hpf), zinc transporters are differentially expressed throughout this period. Nearly all zinc

transporters have their highest expression at 120 hpf, with the exception of ZNT8, which peaks at 48 hpf (Ho et al., 2012). The release of the latest zebrafish database (Zv9 Ensembl 73) has further increased our knowledge of zebrafish genome, confirming the existence of zebrafish ZNT10 and ZIP5 in zebrafish. The homology of zinc transporters among zebrafish, mice and humans is summarized in **Table 3**. An influence of Zn on the uptake and circulatory influx of Cd has been reported in fish, suggesting that the uptake of Zn and Cd occurs through common pathways (Wicklund Glynn, 2001).

granules (ZnT8), and synaptic vesicles (ZnT3). In addition, ZnT9 can

## **ZEBRAFISH MODELS OF ZINC METABOLISM**

The current zebrafish and mouse models available for studying zinc metabolism are listed in **Table 4**.

## *Zip6*

Zip6 is a member of the LIV-1 subfamily of ZIP zinc transporters, which in humans consists of nine ZIP members that contain a highly conserved metalloprotease motif. Importantly, LIV-1 is regulated by estrogen and has been implicated in metastatic breast cancer (Taylor, 2000), although how LIV-1 mediates cancer metastasis is unclear. Studies using Zip6 (LIV1) zebrafish morphants have been instrumental in addressing this problem (Yamashita et al., 2004); the zebrafish *zip6* cDNA was cloned by subtraction screening. Interestingly, the expression pattern of endogenous *zip6* mimics the expression of *stat3*, an important player in the epithelial-mesenchymal transition (EMT) during gastrulation, organogenesis, wound-healing, and cancer progression (Sano et al., 1999; Yamashita et al., 2002). Notably, the expression of *zip6* is abolished in *stat3* zebrafish morphants, suggesting that *zip6* is downstream of *stat3*. The ability of *stat3* to transcriptionally regulate *zip6* was confirmed in studies using mouse and human cell lines (Yamashita et al., 2004). The role of *zip6* in early embryonic development was assessed further in *zip6*-depleted zebrafish embryos. By the end of gastrulation, these *zip6* morphants have malpositioned heads and a shortened anterior-posterior axis, although early cell-fate specification was not affected, suggesting that *zip6* plays a critical role in cell migration during gastrulation. Cell tracing and cell transplantation assays further support the cell-autonomous role of Zip6 in the migration of mesendodermal cells (Yamashita et al., 2004). Phenotypic analyses of *zip6* morphants revealed that cellcell adhesion was not downregulated as occurs normally, thus resulting in severe perturbations in cell migration. These same defects were also observed in zebrafish morphants in which the zinc-finger protein Snail, a master regulator of EMT, is knocked down (Batlle et al., 2000; Cano et al., 2000). Moreover, the *in vivo* activity of Zip6 is dependent on Snail; Zip6 regulates the nuclear translocation of Snail (Yamashita et al., 2004). This was the first study to use multiple zebrafish knockdown models to establish a molecular link between Stat3, Zip6, and Snail during EMT.

zebrafish knockout/knockdown models are available are written in red.

## *Zip7*

Zip7 is also a member of the LIV-1 subfamily of zinc transporters. Studies using mammalian cell lines suggest that human ZIP7 plays a role in elevating cytoplasmic zinc concentrations

## **Table 3 | Zinc metabolism–related proteins in zebrafish, mice, and humans.**


*(Continued)*

**Table 3 | Continued**


*aslc30a1a.*

*bslc30a1b.*

*H, human; M, mouse; Z, zebrafish; N.D., not determined; CNS, central nervous system.*


**Table 4 | Zinc metabolism–related mouse and zebrafish knockout/knockdown models and their phenotypes.**

*\*Knockdown (KD) model.*

by transporting zinc from the Golgi apparatus to the cytoplasm (Huang et al., 2005b). The systemic expression and function of Zip7 was examined in zebrafish (Yan et al., 2012), and endogenous Zip7 was found to be expressed ubiquitously in early stages of somitogenesis, but becomes concentrated around the retina after 24 hpf. In adult fish, *zip7* is also highly expressed in the eyes and the brain. *Zip7* zebrafish morphants have developmental defects that include a curved notochord and small eyes. Moreover, co-injecting *zip7* mRNA or supplementing the surrounding water with zinc significantly rescues the phenotypic defects in *zip7* morphants, suggesting that the developmental defects are caused specifically by *zip7* knockdown and are closely related to zinc deficiency (Yan et al., 2012). The distribution pattern of zinc was also compared between wild-type embryos and *zip7* morphants; the analysis revealed a significant loss of zinc in the eyes of the *zip7* morphants, and this was rescued by the addition of exogenous zinc. These results suggest that Zip7 plays a critical role in maintaining intracellular zinc levels in the eyes and demonstrate that exogenous zinc supplementation can compensate for Zip7 deletion, possibly through the activity of other zinc-importing pathways. Indeed, in the *zip7* morphants, the expression levels of several zinc transporters are altered, including *zip3*, *zip6, znt2*, *znt5*, and *znt6* (Yan et al., 2012). This study revealed the tissue-specific function of Zip7 and nicely illustrates the dynamic interaction between environmental nutrient levels and endogenous transcriptional regulation.

#### **ZINC-REGULATED GENE EXPRESSION IN ZEBRAFISH**

Zinc is required for the function of thousands of transcription factors. Fluctuations in environmental zinc levels actively influence an organism's various biological activities through transcriptional regulation. Zebrafish is a convenient model for studying the effect of fluctuating exogenous nutrients on endogenous gene expression, as the nutrient concentration in the surrounding water can be easily manipulated. In fish, the gill is a unique structure comprised of polarized epithelial cells; this configuration is essential for the fish's ability to extract zinc and other minerals directly from the water. Importantly, the zinc transport system in gills is highly conserved with the transport system in mammals.

Recently, two related studies examined the dynamic transcriptome profiles of gills in zebrafish that were subjected to either zinc depletion or zinc supplementation (Zheng et al., 2010a,b). Juvenile zebrafish were exposed for 2 weeks to water that was either zinc-enriched (4µM), zinc-normal (0.25µM), or zincdeficient (0.04µM). From 14 days of treatment, fish gill samples were collected from each group at multiple time points and processed through microarray analysis in order to measure changes in the transcriptome. In the group that received zinc supplementation, most of the changes in the transcriptome were associated with "transcription factors," "steroid hormone receptors," and "development." Additional data mining suggested that these detected changes in the transcriptome were likely to be induced by only a few key transcription factors, including Mtf1 (the principal regulator of zinc-driven metallothionein expression), Jun, Stat1, Ppara and Gata3, reflecting a process similar to hedgehog and bone morphogenic protein signaling. Moreover, the transcriptional changes tended to slow after seven days of treatment, suggesting that the fish gradually became acclimated to the elevated zinc in the water (Zheng et al., 2010b). In the zincdeficient group, the most significant transcriptional changes were found in genes associated with "developmental processes," which account for up to 26% of all regulated genes. The expression levels of genes correlated with diabetes and bone/cartilage development were also significant changed, which is consistent with previously reported biological roles of zinc (Huang and Tepaamorndech, 2013). Several transcription factors were identified as key coordinators of the homeostatic response to zinc depletion, including Hnf4a, Foxl1, Wt1, Nr5a1, and Nr6a1 (Zheng et al., 2010a). Taken together, these two complementary studies present a systemic, longitudinal overview of the complicated changes that occur in the transcriptome under abnormal environmental zinc levels.

The effect of changing environmental zinc levels on the regulation of zinc transporter expression was also studied in zebrafish by examining the expression patterns of zinc transporters in various tissues under zinc-enriched and zinc-deficient conditions (Feeney et al., 2005). The fish's gills and intestine—two major sites of zinc exchange in fish—had the largest differences in zinc transporter expression. In contrast, the expression of zinc transporters in the muscle and liver was affected the least (Feeney et al., 2005). Two studies examined the transcriptional changes of zinc transporters in fish gills in various zinc concentrations, and these studies reported different sets of genes with altered expression. However, both studies observed increased expression of*znt5*,*zip3,* and *zip10* in the gills under zinc-deficient conditions and reduced expression of *zip10* under zinc-enriched conditions (Feeney et al., 2005; Zheng et al., 2008). Moreover, the inverse relationship between zinc concentration and *zip10* expression may be controlled by metal-responsive clusters in two distinct promoters in the *zip10* gene that have opposing regulatory roles in response to zinc availability; this process is potentially mediated by Mtf-1 (Zheng et al., 2008). Interestingly, studies using mice have also suggested an essential role for Mtf-1 in regulating *Zip10* expression (Wimmer et al., 2005; Lichten et al., 2011).

## **ZEBRAFISH AND COPPER METABOLISM OVERVIEW OF COPPER METABOLISM**

Copper is an essential nutrient that is present in nearly all living organisms. Similar to iron, copper is a redox-active metal. Copper functions as a key catalytic cofactor in a wide range of enzymes and is therefore essential for many fundamental biological processes, including cellular respiration, free radical detoxification, connective tissue formation, and melanin production. Despite its essential role in biology, excess copper is highly toxic due to its high redox potential. Copper overload leads to the production of ROS such as hydroxyl radicals, and the accumulation of these radicals can cause devastating damage to cellular components, ultimately causing cell death (Pena et al., 1999). In humans, Menkes syndrome and Wilson's disease are genetic diseases that are caused by copper deficiency and copper overload, respectively (Ala et al., 2007; Tumer and Moller, 2010).

The endogenous copper metabolism system can be divided into three major steps (**Figure 3**): (i) copper uptake, (ii) intracellular distribution of copper, and (iii) and copper export. The high-affinity copper transporter 1 (CTR1) is the primary player in the uptake of extracellular copper, whereas the low-affinity copper transporter 2 (CTR2) is primarily intracellular and may function to release copper from vesicles. Upon entry into the cell, copper binds to a variety of cytosolic copper chaperones and is then transported to specific subcellular destinations. The three major copper chaperones are Copper chaperone for superoxide dismutase (CCS), Cytochrome c oxidase assembly protein 17 (COX17), and Antioxidant copper chaperone 1 (ATOX1). CCS delivers copper to cytosolic superoxide dismutase 1 (SOD1) to activate its function in mediating superoxide protection. COX17 transports copper to the mitochondria and facilitates its incorporation into Cytochrome c oxidase (CCO), the final enzyme in the respiratory electron transport chain. ATOX1 carries copper to copper-ATPases in the Golgi apparatus, from which copper is then transferred to various cuproenzymes via secretory pathways. The secretion of copper is also dependent on copper-ATPases. ATP7A (the disease gene linked to Menkes syndrome) and ATP7B (the gene that is defective in patients with Wilson's disease) encode two major types of copper exporters. When intracellular copper levels are high, ATP7A and ATP7B are expressed in close proximity to the basolateral and apical membranes, respectively, where they export copper via vesicle-mediated fusion (Lutsenko, 2010). Approximately 95% of copper in the plasma is bound to ceruloplasmin (CP), the principal circulating copper carrier. CP is also a ferroxidase, serving as a molecular link between copper and iron metabolism. The copper metabolism-related proteins in fish and mammals are summarized in **Table 5**, and the currently available fish and mammalian models are summarized in **Table 6**.

## **ZEBRAFISH MODELS OF COPPER METABOLISM** *Identifying copper-deficient phenotypes in zebrafish*

Unlike iron, which is directly related to anemia and hemochromatosis, many trace minerals do not cause specific phenotypes when they are deficient or in excess. The identification of copperrelated phenotypes is essential for identifying animal models to study mineral imbalance. A spectrum of distinct developmental abnormalities was linked to copper deficiency through a chemical genetic screen (Mendelsohn et al., 2006). Copper has been proposed to play a role in melanin formation through the activity of tyrosinase, a copper-containing oxidase (Rawls et al., 2001). Using copper-induced reversible depigmentation as a screening output, a library of small molecules was evaluated for their role in interfering with copper metabolism. Notably, in addition to pigment loss, a specific combination of other abnormalities was observed in all of the molecule-treated embryos, including a wavy notochord, impaired cartilage and vascular development, lack of hematopoiesis, and defective neurogenesis. Adding copper—but not any other trace mineral—to the water of molecule-treated embryos rescued the phenotypes (Mendelsohn et al., 2006). Establishing these copper deficiency– induced phenotypes in zebrafish greatly facilitated the discovery of zebrafish with mutations linked to copper metabolism. The

role of copper in notochord development was suggested to be related to lysyl oxidase, a cuproenzyme that may be important for maintaining notochord sheath integrity (Csiszar, 2001). Lysyl oxidase was further implicated in notochord development through studies of zebrafish knockdown models, in which lysyl oxidase morphants develop a notochord distortion that is similar to copper-deficient fish (Gansner et al., 2007). It is also notable that adding exogenous copper at an early stage of development rescues all defects, suggesting that copper—and likely other metals as well—can enter the embryo from the surrounding water, presumably via transport through the cell membrane.

chaperones and transported to various proteins in the following

Using the copper deficiency–linked zebrafish phenotypes as a screening standard (Mendelsohn et al., 2006), a more recent study examined nearly 3000 small molecules and identified a novel panel of copper inhibitors (Ishizaki et al., 2010). Interestingly, the authors combined the zebrafish phenotype screen with a yeast chemical-genetics screen. The molecules that were identified from the zebrafish screen were used to treat a genome-wide library of mutant yeast strains in order to identify novel genetic pathways involved in copper metabolism. Select copper-related genes identified from the yeast screen were then verified using zebrafish knockdown models (Ishizaki et al., 2010). This flexible and powerful combination of zebrafish and yeast chemicalgenetics screening approaches will likely be useful in other studies of diseases with identifiable phenotypes.

## *Atp7a*

models are available are written in red.

ATP7A plays important roles in exporting excess cytosolic copper from the cell and in the delivery of copper to cuproenzymes via secretory pathways. In humans, mutations in *ATP7A* cause Menkes syndrome, which has a broad spectrum of clinical disorders that are related to copper deficiency, including progressive neurodegeneration, connective tissue abnormalities, and kinky, colorless hair (Tumer and Moller, 2010). After they identified the copper-deficient phenotypes in zebrafish, the same group performed an *N*-ethyl-*N*-nitrosourea (ENU) mutagenesis screen to search for mutants that mimic the chemically induced copper-deficient phenotypes, particularly the pigment loss and wavy notochord. This screen identified the *calamity* mutant, which has the same set of developmental defects. Positional cloning revealed that *calamity* mutants contain a splice variant of the zebrafish homolog of *atp7a*; this alternatively spliced product causes a frame-shift and introduces a premature stop codon (Mendelsohn et al., 2006). In fish, the expression pattern of *atp7a* correlates with the phenotypic defects, with strong expression in the developing notochord. Moreover, the phenotype of the *calamity* mutants can be rescued by overexpressing human *ATP7A*, suggesting conservation of function (Mendelsohn et al., 2006). In addition, *atp7a* zebrafish morphants have the same hypopigmentation and defective notochord phenotype as *calamity* mutants. Finally, a role for Atp7a in modulating the expression of *sp1* and *sod1* has been



*H, human; M, mouse; Z, zebrafish; N.D., not determined; CNS, central nervous system; RPE, retinal pigment epithelium.*

suggested based on studies of *atp7a* morphants (Chen et al., 2011).

Although the *mottled* mouse is a well-characterized animal model of Menkes syndrome and has been studied for nearly 40 years (Hunt, 1974; Levinson et al., 1994), thanks to its unique properties, the recently developed zebrafish Menkes model has significantly increased our understanding of the underlying disease mechanism. The rapid and *ex utero* development of zebrafish embryos has greatly facilitated the feasibility of performing experiments during early embryogenesis. Embryonic transplantation assays have revealed that transplanted wild-type cells can develop melanin normally in *calamity* mutants, indicating that Atp7a functions cell-autonomously. Furthermore, a combination of copper-suppressing treatment and *atp7a* morpholino injections revealed that the gene dosage of Atp7a determines the animal's sensitivity to copper deficiency (Mendelsohn et al., 2006). These two novel findings obtained using the zebrafish *atp7a* model suggest new therapeutic strategies that focus on tissue-specific gene replacement for treating patients with Menkes syndrome. The zebrafish Menkes model is also a powerful tool for identifying and screening potential compounds that can restore cuproenzyme function in the Atp7a mutant background using chemical screens.

A rescue assay of two zebrafish *atp7a* mutants was conducted via morpholino injection (Madsen et al., 2008). The same ENU screen described above revealed a second allele of *calamity*, and animals bearing this mutation have a phenotype that is similar to animals with the first allele (Mendelsohn et al., 2006; Madsen



*\*Knockdown model.*

et al., 2008). Noticing that both *calamity* mutants cause splicing defects, the researchers attempted to rescue the mutants by overexpressing antisense morpholino oligonucleotides designed to specifically target the splice-site junctions in the two mutations. Remarkably, the morpholino injections fully rescued the copper-deficient defects in the *calamity* mutants and permitted the production of wild-type Atp7a protein in all rescued embryos. Nevertheless, although the rescued morphants had decreased amounts of mutant mRNA, they did not have a significant increase in wild-type mRNA, suggesting the presence of competitive translational regulation (Madsen et al., 2008). This study made available promising therapeutic options for using gene correction therapy to treat patients with Menkes syndrome, although the feasibility of such an approach is currently limited by several factors, including the nature of the mutations and the delivery of morpholinos.

Small-molecule copper suppressors used in combination with ENU mutagenesis screens identified zebrafish mutants that are sensitive to limited copper availability (Madsen and Gitlin, 2008). After treating with suboptimal doses of copper chelators (which has no effect in wild-type embryos), developmental defects (including pigment loss and an undulating notochord) were detected in certain mutants that are phenotypically silent with adequate copper supply (Madsen and Gitlin, 2008). Two separate copper-sensitive mutants were identified: a hypomorphic allele of *calamity* and a novel mutant called *catastrophe* (*cto*). Interestingly, in the hypomorphic *calamity* mutants, homozygous embryos from homozygous mothers have reduced pigmentation; this phenomenon does not occur in homozygous embryos from heterozygous mothers, suggesting that the mutant's copper sensitivity is related to deficient loading of maternal copper into the egg. The *cto* mutation is homozygous lethal, with homozygotes developing small, punctate melanocytes; moreover, these animals lose all pigments when treated with copper chelators. *Cto* mutant have a mutation in Atp6, a vacuolar ATPase that is predicted to play a role in proton translocation. Embryonic transplantation of wild-type melanocytes restores normal melanin levels in *cto* mutants, suggesting that the mutants have an intact copper uptake system. However, a suboptimal dose of Atp7a morpholinos completely blocks the production of melanin in the Atp6 mutants, suggesting a novel relationship between the proton transport and copper secretory pathways (Madsen and Gitlin, 2008). This study revealed the mechanisms that underlie copper metabolism by studying mutants that were identifiable only with suboptimal copper supply, and it illustrates the value of using the zebrafish model system to study gene-nutrient interactions.

## *Ctr1*

CTR1 is a high-affinity copper importer that was originally identified in yeast (Dancis et al., 1994). Subsequent studies identified human CTR1 as a potent copper transporter that is located primarily in the plasma membrane (Lee et al., 2002). The role of CTR1 in the absorption of dietary copper was further demonstrated using a tissue-specific *Ctr1*-knockout mouse model (Nose et al., 2006). The zebrafish Ctr1 protein shares approximately 70% identity with the human homolog, particularly among the amino acid residues that are essential for copper transport (Mackenzie et al., 2004). Early in development, *ctrl1* expression is generalized, but then becomes concentrated in the intestine, which is consistent with its role in the uptake of dietary copper. *Ctr1*-knockdown fish are embryonic lethal, with massive cell death throughout the neural system, including the brain and spinal cord (Mackenzie et al., 2004). However, the mechanism by which Ctrl1 transports copper in zebrafish, and the protein's role in embryogenesis, requires further study.

## **ZEBRAFISH AND THE METABOLISM OF OTHER TRACE ELEMENTS SELENIUM**

Selenium plays essential biological functions in the body, primarily in the form of selenoproteins, which contain selenocysteine amino acid residues. These functions include antioxidant defense, thyroid hormone production, and cancer prevention, The mechanisms by which selenium is absorbed and excreted vary based on its chemical form in foods, and these mechanisms are poorly understood (Mehdi et al., 2013).

Several zebrafish selenoproteins have been identified and have a function that is conserved with human homologs, making fish a powerful model for studying selenium metabolism. The expression patterns of more than 20 selenoprotein genes were analyzed in zebrafish embryos, and all of the examined genes have tissue-specific expression patterns, many of which reflect their known functions in mammals (Thisse et al., 2003). Mutations in the human Selenoprotein N-encoding gene *SEPN1* cause various forms of congenital muscular diseases called *SEPN1*-related myopathies, which are characterized by earlyonset hypotonia and weakness (Lescure et al., 2009). During somitogenesis, the zebrafish *sepn1* gene is expressed specifically in the somites and notochord, which are the precursors of skeletal muscle and vertebrae, respectively (Thisse et al., 2003; Deniziak et al., 2007). *Seph1*-knockdown embryos have reduced motility, poorly coordinated movement, and poorly delimited myotome boundaries. At the ultrastructural level, morphants exhibit pathological muscle changes, including defects in sarcomere organization and myofiber attachment, as well as altered myoseptum integrity (Deniziak et al., 2007). Thus, the functions of zebrafish *seph1* and mammalian *SEPN1* are strikingly similar, and studies of the zebrafish *seph1* model have yielded new insights into the pathological changes that occur in human *SEPN1*-related myopathy and may serve as an ideal model for future studies of disease mechanisms and treatments.

On the other hand, the zebrafish selenium metabolism has non-mammalian properties as well. Gene duplication is an interesting genetic phenomenon in zebrafish selenium metabolism. Two distinct genes that encode the selenocysteine tRNA[Ser]Sec have been identified (Xu et al., 1999). Selenocysteine tRNA[Ser]Sec, the principal component in selenoprotein biosynthesis, is encoded by a single-copy gene in mammals and many other classes. In zebrafish, the two tRNA[Ser]Sec genes have identical coding sequences, and their flanking regions (several hundred bases in length in both directions) are highly homologous, which likely reflects evolutionary gene duplication in teleosts (Xu et al., 1999). Similarly, multiple selenoprotein genes have been detected in fish (Kryukov and Gladyshev, 2000). Two of the zebrafish homologs of the human selenoproteins SEPT—glutathione peroxidase 1 and glutathione peroxidase 4—each have two genes in the zebrafish genome. In addition, the zebrafish *sepp* gene contains duplicated Sec insertion sequence elements and encodes a protein containing 17 Sec residues, which is the largest number of Sec residues in any known protein (Kryukov and Gladyshev, 2000; Tujebajeva et al., 2000). Finally, a novel family of selenoproteins called Sepu was identified in fish, chicken, and many other non-mammalian species, suggesting the divergent evolutionary distribution of selenoproteins in eukaryotes (Castellano et al., 2004). Thus, although the zebrafish is a convenient and powerful model for studying selenium metabolism, researchers must be aware of key genetic differences between fish and mammals.

## **MANGANESE**

Manganese (Mn) is a key trace element associated with bone development, superoxide elimination, and the metabolism of amino acids, lipids and carbohydrates. Biologically, manganese functions primarily as a cofactor of various enzymes, including Mn superoxide dismutase (Mn-SOD), glutamine synthetase, and arginase. Mn is transported through the body by transferrin, macroglobulins, and albumin (Fraga, 2005). However, the mechanism of manganese metabolism in animals is poorly understood. The zebrafish Mn-SOD has been cloned and shares 85% sequence identity with the human homolog, including high conservation of the amino acids located in the Mn-binding sites. The fish Mn-SOD gene is highly expressed during the early cleavage stage and has been suggested to be maternally distributed to the eggs, indicating an essential role in embryonic development (Lin et al., 2009). In addition, a Mn-dependent enzyme, ADPribose/CDP-alcohol diphosphatase (ADPRibase-Mn), has been functionally characterized in zebrafish (Rodrigues et al., 2012). ADPRibase-Mn family members may play a role in the immune system in vertebrates, as suggested by their expression patterns in the rat (Canales et al., 2008). Similar to the rodent homolog, the zebrafish ADPRibase-Mn is also Mn-dependent and catalyzes the hydrolysis of ADP-ribose and CDP-alcohol. However, the enzyme's cyclic-ADP-ribose hydrolysis activity, which is robust in rat ADPRibase-Mn, is negligible in the fish homolog, possibly due to the lack of any known cyclic ADPR synthesis pathways in fish (Rodrigues et al., 2012).

## **IODINE**

Iodine is also a non-metal trace mineral. The primary biological role of iodine is a constituent of thyroid hormones, which are essential regulators of body growth, cell metabolism, and body temperature maintenance. Zebrafish are an important model system for studying thyroid development and function. Unlike humans, zebrafish lack a compact thyroid gland. Nevertheless, zebrafish thyroid tissue expresses genes that are critical for its patterning and development, and these genes are conserved with mammals (Porazzi et al., 2009). In zebrafish, thyroid hormones play a role in regulating the differentiation of the pectoral fins and determining the transition from the larval stage to the juvenile stage (Brown, 1997). Thyroid hormones are also essential for the normal function of several physiological systems in fish, including the cardiovascular system, the skeletomuscular system, and the digestive system (Power et al., 2001). Therefore, zebrafish can be an ideal model for studying thyroid diseases and iodine metabolism.

## **CONCLUSIONS**

Due to the availability of powerful genetic tools and its developmental advantages, the zebrafish has become an invaluable model system for studying mineral metabolism. The generation of various zebrafish knockdown and knockout models has greatly facilitated the identification of novel genes and mechanisms that underlie mineral metabolism, particularly with respect to iron and copper, which produce characteristic phenotypic changes when their concentrations are altered. Moreover, the flexible combination of mutagenesis screening and metal chelation treatments, for example in studies of copper metabolism (Mendelsohn et al., 2006; Madsen and Gitlin, 2008), may be useful for identifying representative deficiencies that are related to the metabolism of other metals. In addition, because the zebrafish is an aquatic organism that develops *ex utero*, the endogenous mineral levels in zebrafish can be readily altered by changing the concentration of nutrients in the surrounding water, even as early as embryonic day 0. Thus, the effect of dynamically changing trace mineral levels on early development can be studied quite conveniently in zebrafish (Ho et al., 2012). This advantage also makes zebrafish an ideal tool for studying geneticnutrient interactions (Madsen and Gitlin, 2008). Furthermore, the zebrafish is a vertebrate organism that is highly conserved with humans, but is small in size and has rapid development, making the zebrafish an ideal tool for studying the functional consequences of gene mutations that have been identified in mammals (De Domenico et al., 2007), for expanding cell-based findings to the systemic/organism level (Devireddy et al., 2010), and for confirming findings obtained from studying lower organisms (Ishizaki et al., 2010). Finally, the results obtained using zebrafish disease models provide important directions for treating human patients (Madsen et al., 2008).

The recent breakthroughs in reverse gene-editing technologies—such as transcription activator-like effector nucleases (TALENs) and the clustered regularly interspaced short palindromic repeats (CRISPR)/Cas9 system—have greatly facilitated the generation of loss-of-function fish mutants, which in turn have enormously accelerated our ability to examine the biological functions of many genes that had a previously elusive relationship with trace mineral metabolism. Future studies such as the construction of zebrafish mutant libraries of mineral transporter proteins will allow researchers to gain important insights into the field of mineral metabolism.

## **ACKNOWLEDGMENTS**

This work was supported by research grants from the Chinese National Natural Science Foundation grants (numbers 31225013, 31330036, and 31030039 to Fudi Wang). This work was also supported by the Distinguished Professorship Program from Zhejiang University (to Fudi Wang). We are also grateful to the members of the Wang laboratory for their encouragement and helpful comments.

## **REFERENCES**


lacking a zinc transporter gene, Znt5. *Hum. Mol. Genet* 11, 1775–1784. doi: 10.1093/hmg/11.15.1775


embryonic development. *Transgenic. Res.* 12, 131–133. doi: 10.1023/A:1022118 627058


of fish. *Comp. Biochem. Physiol. C Toxicol. Pharmacol.* 130, 447–459. doi: 10.1016/S1532-0456(01)00271-X


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

*Received: 24 January 2014; paper pending published: 12 February 2014; accepted: 17 February 2014; published online: 06 March 2014.*

*Citation: Zhao L, Xia Z and Wang F (2014) Zebrafish in the sea of mineral (iron, zinc, and copper) metabolism. Front. Pharmacol. 5:33. doi: 10.3389/fphar.2014.00033*

*This article was submitted to Drug Metabolism and Transport, a section of the journal Frontiers in Pharmacology.*

*Copyright © 2014 Zhao, Xia and Wang. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*

## The physiological functions of iron regulatory proteins in iron homeostasis – an update

## *De-Liang Zhang, Manik C. Ghosh and Tracey A. Rouault\**

Molecular Medicine Program, Eunice Kennedy Shriver National Institute of Child Health and Human Development, National Institute of Health, Bethesda, MD, USA

#### *Edited by:*

Raffaella Gozzelino, Instituto Gulbenkian de Ciência, Portugal

#### *Reviewed by:*

Janis Lynne Abkowitz, University of Washington, USA Wolff Mayer Kirsch, Loma Linda University Medical Center, USA

#### *\*Correspondence:*

Tracey A. Rouault, Molecular Medicine Program, Eunice Kennedy Shriver National Institute of Child Health and Human Development, National Institute of Health, Bethesda, MD 20892, USA e-mail: rouault@mail.nih.gov

Iron regulatory proteins (IRPs) regulate the expression of genes involved in iron metabolism by binding to RNA stem-loop structures known as iron responsive elements (IREs) in target mRNAs. IRP binding inhibits the translation of mRNAs that contain an IRE in the 5- untranslated region of the transcripts, and increases the stability of mRNAs that contain IREs in the 3- untranslated region of transcripts. By these mechanisms, IRPs increase cellular iron absorption and decrease storage and export of iron to maintain an optimal intracellular iron balance. There are two members of the mammalian IRP protein family, IRP1 and IRP2, and they have redundant functions as evidenced by the embryonic lethality of the mice that completely lack IRP expression (Irp1−/−/Irp2−/<sup>−</sup> mice), which contrasts with the fact that Irp1−/<sup>−</sup> and Irp2−/<sup>−</sup> mice are viable. In addition, Irp2−/<sup>−</sup> mice also display neurodegenerative symptoms and microcytic hypochromic anemia, suggesting that IRP2 function predominates in the nervous system and erythropoietic homeostasis. Though the physiological significance of IRP1 had been unclear since Irp1−/<sup>−</sup> animals were first assessed in the early 1990s, recent studies indicate that IRP1 plays an essential function in orchestrating the balance between erythropoiesis and bodily iron homeostasis. Additionally, Irp1−/<sup>−</sup> mice develop pulmonary hypertension, and they experience sudden death when maintained on an iron-deficient diet, indicating that IRP1 has a critical role in the pulmonary and cardiovascular systems. This review summarizes recent progress that has been made in understanding the physiological roles of IRP1 and IRP2, and further discusses the implications for clinical research on patients with idiopathic polycythemia, pulmonary hypertension, and neurodegeneration.

**Keywords: iron regulatory protein, iron responsive element, erythropoiesis, polycythemia, pulmonary hypertension, iron metabolism**

## **INTRODUCTION**

Iron is an indispensable element for all living organisms. Healthy adults contain 4–5 g of iron, about 65% of which is contained in hemoglobin where it participates in oxygen transport, 30–35% is stored in liver, primarily in the storage protein, ferritin, and 1–2% is found in the form of iron–sulfur clusters or heme in the catalytic centers of numerous essential enzymes and multiprotein complexes such as the mitochondrial respiratory chain complexes, which contain twelve iron–sulfur clusters and seven hemes (Hentze et al., 2004; Darshan et al., 2010; Ganz and Nemeth, 2012b; Rouault, 2013). Due to the essential role of iron *in vivo*, iron deficiency can retard early development and impair cognitive ability of children, and iron deficiency anemia is a common nutrient deficiency disease worldwide. Conversely, because of the chemical reactivity of iron and its ability through Fenton chemistry to generate reactive hydroxyl radicals, which can then oxidize lipids, proteins and DNA, iron overload can damage cells and tissues, and lead to adverse consequences, such as those seen in hemochromatosis and hemolytic anemias. Therefore iron concentration has to be tightly regulated in the tissues and cells of organisms *in vivo*.

Mammals have developed sophisticated mechanisms to maintain appropriate iron concentrations *in vivo.* Thanks to the application of genetic screens and transgenic technology in biomedical research, our understanding of iron homeostasis regulation has advanced significantly in the last 15 years. Iron homeostasis in mammals is mainly regulated by a set of interlocking regulatory systems, including: (i) Hepcidin– ferroportin (FPN1) mediated regulation of serum iron levels, (ii) iron regulatory proteins (IRPs)/iron responsive element (IRE) mediated regulation of intracellular iron homeostasis, (iii) hypoxia inducible factor-2α (HIF2α) mediated transcriptional regulation. These mechanisms regulate iron homeostasis at different levels, and the interaction and cooperation of these mechanisms fine-tunes iron levels *in vivo*. The IRP/IRE machinery post-transcriptionally regulates the expression of target genes according to cellular iron status, providing the fundamental regulation of iron homeostasis at the cellular level. Recently, studies in animal models have also shown that IRPs contribute significantly to systemic iron homeostasis and regulation of erythropoiesis. This review will begin with an overview of mammalian iron homeostasis, and will then focus on the more recent progress made in understanding the roles of IRP1 and IRP2 in cellular and systemic iron homeostasis.

## **OVERVIEW OF SYSTEMIC AND CELLULAR IRON METABOLISM**

For a healthy adult, about 25–30 mg of iron is needed daily for protein synthesis and cellular regeneration. To meet these requirements, about 90% of iron is acquired from the recycle of senescent red blood cells (RBC) by splenic macrophages, whereas the remaining 10% is absorbed from the diet to compensate for iron loss caused by bleeding, urinary excretion, and sloughing of epithelial and mucous cells (Rigby, 1971). More than 90% of daily iron consumption is used for RBC production in erythropoietic tissues, and the most prominent manifestation of iron deficiency is microcytic anemia (Rigby, 1971). The systemic iron homeostasis is mainly maintained by coordinating iron absorption through the duodenum, iron recycling through splenic macrophages, iron utilization in bone marrow by erythropoiesis, and iron storage in the liver. Because there is no known regulated iron excretion pathway, regulation of intestinal iron absorption plays an important role in maintenance of systemic iron homeostasis.

Mammals can absorb both heme iron and non-heme iron. Iron derived from heme, especially in the people of western societies, is estimated to contribute two thirds of the daily dietary iron absorption (West and Oates, 2008). However, the mechanism for heme absorption is not yet clear. In the last decade, several heme transporters have been identified, including heme carrier protein-1 (HCP1; Qiu et al., 2006; Laftah et al., 2009), HRG-1 (Rajagopal et al., 2008; White et al., 2013), and FLVCR1 and 2 (Quigley et al., 2004; Keel et al., 2008; Duffy et al., 2010), but their significance in intestinal iron absorption remains to be elucidated. For non-heme iron absorption, ferric iron [Fe(III)] in the diet must be reduced by a ferrireductase duodenal cytochrome b561 (Dcytb) to ferrous iron [Fe(II)] before the divalent metal transporter 1 (DMT1, also known as DCT1 or NRAMP2) can transport iron across the apical membrane into the cytosol of duodenal epithelial cells (socalled enterocytes; Gunshin et al., 1997; McKie et al., 2001). Once inside the cells, part of the newly absorbed iron is exported across the basolateral membrane by the iron export protein, ferroportin (FPN1, also known as MTP1 or IREG1) into circulating blood, where Fe(II) is converted to Fe(III) by a membrane-bound ferroxidase, hephaestin (HEPH), before binding to transferrin and circulating in blood to tissues and cells where iron is needed (Abboud and Haile, 2000; Donovan et al., 2000; McKie et al., 2000); enterocytes may use part of the iron absorbed for their own metabolic needs, whereas excess iron can be sequestered in ferritin for detoxification and storage, and sloughing of these cells at the end of their 3–4 days life span results in excretion of excess duodenal iron (Potten, 1998).

Because of the essential role of duodenal iron absorption in systemic iron homeostasis, duodenal uptake and transfer of iron is regulated by multiple mechanisms. HIF2α has been shown to transcriptionally regulate the expression of iron transporters in iron deficiency and anemia conditions (Majmundar et al., 2010; Shah and Xie, 2013). There are hypoxia responsive elements (HREs) in the promoters of *FPN1*, *DMT1,* and *DCYTB* (also known as *CYBRD1*) and the physiological importance of these HRE–HIF interactions has been demonstrated in the duodenum of *Hif2*α conditional knockout mice, where expression of FPN1, DMT1, and DCYTB does not increase in response to iron deficiency, suggesting that HIF2α has an important physiological role in transcriptional regulation of iron homeostasis (Mastrogiannaki et al., 2009; Shah et al., 2009). Secondly, hepcidin regulates the expression of FPN1 on the basolateral membrane of enterocytes and thereby adjusts iron export from enterocytes into blood (Ganz and Nemeth, 2011, 2012a). Hepcidin is a systemic iron regulatory hormone that is secreted mainly by the liver. Circulating hepcidin can bind FPN1 on the plasma membrane and induce its ubiquitination, internalization and degradation, and thereby reduce iron influx into blood in a feedback manner (Nemeth et al., 2004). Dysregulation of the hepcidin-FPN1 interaction causes the systemic iron overload disease, hemochromatosis, which highlights its significance in systemic iron homeostasis. Thirdly, the intracellular IRP/IRE machinery also regulates duodenal iron absorption, and IRP-related regulation will be discussed later in this review.

Unlike intestinal epithelial cells, other cells *in vivo* use holotransferrin (transferrin bearing two ferric iron) as the major iron source, and cells absorb iron through the transferrin (Tf)/transferrin receptor (TfR1) cycle. Holo-transferrin in the circulation binds to TfR1 on plasma membranes to form TfR1/Tf/Fe complexes which then internalize to endosomes, whereupon acidification of endosomes induces the release of Fe(III)from Tf. Fe(III) undergoes reduction to Fe(II) by the endosomal ferrireductase STEAP3 before being transported by DMT1 across endosomal membranes into cytosol (Fleming et al., 1998; Ohgami et al., 2005). DMT1 may also directly transport non-transferrin-bound iron (NTBI) into cells *in vivo,* especially in conditions including hemochromatosis and hemolytic anemia when serum iron concentrations exceed the binding ability of transferrin, and therefore NTBI accumulates (Sarkar, 1970; Chua et al., 2004). Once inside the cytosol, part of the iron is taken up by mitochondria and used for heme and iron–sulfur cluster synthesis; excess iron can be exported out of the cells by the iron exporter FPN1, and much extra iron is sequestered in ferritin for detoxification and storage. Ferritin can store up to 4500 iron atoms in a spherical structure formed by 24 subunits of H- and L- ferritin, which self-organize in different ratios, depending on the tissue (Theil, 2012). The IRP/IRE machinery coordinates cellular iron absorption, export, utilization and storage, and thereby regulates intracellular iron homeostasis

## **REGULATION OF INTRACELLULAR IRON HOMEOSTASIS BY THE IRP/IRE MACHINERY**

The IRP/IRE machinery registers intracellular iron levels, and coordinates iron absorption, export, utilization and storage, providing the fundamental machinery for regulation of intracellular iron metabolism (Rouault, 2006, 2013; Wallander et al., 2006; Muckenthaler et al., 2008). The IRE is a conserved stem-loop structure in the untranslated region (UTR) of target mRNAs. A typical loop has six nucleotides with the sequence of CAGUGN, in which the first C and the fifth G are believed to form a base-pair that stabilizes the structure (Sanchez et al., 2011). The six-nucleotide loop is connected to a stem that is separated into an upper and lower part by an unpaired bulge C residue that divides the stem in the middle. IRE sequences are highly

conserved, and mutations of the IRE can cause iron dysregulation and diseases, suggesting a significant role of IRP/IRE regulation in iron homeostasis (Kato et al., 2001; Ismail et al., 2006). IREs are recognized and bound by IRP proteins, but the effect of IRP binding depends on the position of the IRE in the mRNA targets. If an IRE is located in the 5- UTR of target mRNAs, IRP binding can inhibit the translation of such target mRNAs, including L- and H-ferritin (iron storage protein) (Theil, 1990), FPN1 (iron export protein; Abboud and Haile, 2000; Donovan et al., 2000; McKie et al., 2000), erythroid 5-aminolevulinate synthase (eALAS or ALAS2, the first enzyme for heme synthesis; Dandekar et al., 1991), mitochondrial aconitase (ACO2, energy production; Kim et al., 1996; Schalinske et al., 1998), HIF2α (erythropoiesis and hypoxia response; Sanchez et al., 2007), and *Drosophila* succinate dehydrogenase (SDH, citric acid cycle and mitochondrial electron transport chain; Kohler et al., 1995; Melefors, 1996). Conversely, if IREs are present in the 3- UTR of target mRNAs, IRP binding can increase their expression by stabilizing the mRNAs that include TfR1 and DMT1 (iron import proteins; Garrick et al., 2003). In summary, when activated by iron deficiency, IRPs bind the IREs of target mRNAs to increase iron absorption and decrease iron export, iron utilization and iron storage, thereby maintaining appropriate intracellular iron concentrations.

## **IRON REGULATORY PROTEINS**

Iron regulatory proteins are soluble cytosolic proteins that alter their activities according to intracellular iron levels. There are two IRP proteins in mammalian cells, IRP1 and IRP2, which share 56% sequence identity. In addition, IRP2 has a cysteine-rich 73 amino acid insertion in its N-terminal, but the function of the insertion is not clear yet (Pantopoulos, 2004). Both IRP1 and IRP2 are ubiquitously expressed, with IRP1 highly expressed in the kidneys, liver and brown fat, and IRP2 highly expressed in the central nervous system (Meyron-Holtz et al., 2004). IRP1 and IRP2 are regulated by different mechanisms. IRP1 is a bifunctional enzyme. In iron-replete conditions, IRP1 acquires a [4Fe–4S] cluster in its active cleft and displays cytosolic aconitase activity that catalyzes the conversion of citrate and isocitrate in the cytosol, which probably enhances NADPH generation and lipid synthesis (Tong and Rouault, 2007). In iron-deficient conditions, IRP1 loses its iron–sulfur cluster and acquires IRE-binding activity. The iron–sulfur cluster functions as the iron sensor of IRP1 that endows the protein with the ability to register intracellular iron concentration and adjust its activity accordingly (Rouault, 2006; Medenbach et al., 2011). In contrast to the bifunctional enzyme activity of IRP1, IRP2 does not have an iron–sulfur cluster and lacks aconitase activity, and its activity is regulated by ubiquitination and proteasomal degradation (Rouault, 2009; Salahudeen et al., 2009; Vashisht et al., 2009). Evidence from two independent groups demonstrated that IRP2 is targeted for proteasomal degradation by an E3 ubiquitin ligase complex that contains an F-box protein, FBXL5 (Salahudeen et al., 2009; Vashisht et al., 2009). FBXL5 has a hemerythrin domain that likely binds iron and oxygen, enabling it to function as a regulatory switch that determines the stability of FBXL5, and consequently regulates E3 ubiquitin ligase activity (Chollangi et al., 2012). In brief, iron

deficiency and hypoxia destabilize FBXL5 protein, decrease the activity of the E3 ubiquitin ligase, and thereby increase IRP2 activity. *Fbxl*5−/<sup>−</sup> mice were embryonic lethal, and the lethality could be rescued by deletion of *Irp2*, suggesting that the lethality is likely caused by augmented expression of IRP2 protein, a result that underscores the essential role of FBXL5-E3 ubiquitin ligase in regulation of IRP2 expression (Moroishi et al., 2011). IRP1 is also a target of the FBXL5-E3 ubiquitin ligase complex, and after it loses its [4Fe-4S] cluster by mutation of three cystine residues, IRP1 is down-regulated likely by the FBXL5-mediated proteasomal degradation pathway (Salahudeen et al., 2009; Vashisht et al., 2009). Because IRP1 is relatively stable in the cytosolic aconitase form at iron-replete conditions when FBXL5-E3 ubiquitin ligase is active, the physiological significance of the FBXL5-mediated proteasomal degradation pathway on IRP1 expression is still elusive (Recalcati et al., 2006).

## **PHYSIOLOGICAL SIGNIFICANCE OF IRP1 AND IRP2**

The IRP/IRE machinery maintains intracellular iron homeostasis and plays a crucial role in development and normal physiology. Animals bred to lack both alleles of *Irp1* and *Irp2* are not viable, and further analyses have shown that embryos at the blastocyst stage display brown color and abnormal morphologies, supporting an essential role of IRPs in early development, before implantation of the embryo (Smith et al., 2006). In contrast, mice with either Irp1 or Irp2 deficiency are viable and fertile, suggesting that Irp proteins can compensate for the loss of one another and are functionally redundant (Meyron-Holtz et al., 2004). The essential role of Irp proteins is also highlighted by conditional knockout experiments which have shown that lack of Irp1 and Irp2 in the intestine of mice results in early death at around 4 weeks of age likely due to intestinal malabsorption and dehydration, and lack of Irp1 and Irp2 in hepatocytes causes liver failure and death of animals within 12 days postpartum (Galy et al., 2008, 2010). Conditional deletion of both *Irp1* and *Irp2* in hepatocytes compromises iron–sulfur cluster and heme synthesis, and impairs mitochondrial functions, suggesting an essential role of Irps in supplying iron to mitochondria to maintain respiration. Though adult mice with ligand-induced deletion of both *Irp1* and *Irp2* in duodenal enterocytes responded well to iron loading and erythropoietic stimulation, these mice displayed reduced iron absorption and iron accumulation in duodenal enterocytes, suggesting that Irps play an important role in adjusting duodenal iron absorption by regulating ferritin expression to create a ferritin dependent "mucosal block" (Galy et al., 2013).

## **PHYSIOLOGICAL SIGNIFICANCE OF IRP2**

Though deletion of both *Irp1* and *Irp2* is embryonic lethal, mice with deletion of either *Irp1* or *Irp2* are viable and fertile (Meyron-Holtz et al., 2004). Irp2 deficiency causes iron misregulation in the duodenum, central nervous system, and most prominently in motor neurons of spinal cord; misregulation is characterized by increased expression of ferritin, decreased expression of TfR1, and significant iron accumulation in the duodenal mucosa and neurons throughout the brain (LaVaute et al., 2001). Iron misregulation correlates with axonal degeneration and neuronal death in

the brain and spinal cord, and animals in late stages of adulthood display a movement disorder characterized by abnormal gait, tremor, and hind-limb paralysis (LaVaute et al., 2001; Jeong et al., 2011). Compared with *Irp2*−/<sup>−</sup> mice, loss of one more copy of *Irp1* in *Irp1*+/−*Irp2*−/<sup>−</sup> mice exacerbates the neurodegenerative symptoms as evidenced by the increased myelin dense bodies in the ventral spinal cord in the region where motor neurons are found (a hallmark of neurodegeneration), increased stress markers, increased macrophage infiltration, and decreased diameters of motor neuronal axon bundles, suggesting that there is a dosage effect of Irp1 and Irp2 deficiency and confirming that neurodegenerative symptoms of *Irp2*−/<sup>−</sup> mice are caused by iron dysregulation. Deficiency of Irp*2* increases the expression of the iron storage protein ferritin and decreases expression of the iron importer TfR1, leading to functional iron deficiency (lack of biologically available iron) in conjunction with apparent ferric iron overload caused by sequestration of iron in ferritin; and the notion that there is functional iron deficiency is also supported by deficiency of mitochondrial complex I/II activity (**Figure 1**; Jeong et al., 2011). The neurodegenerative symptoms of *Irp2*−/<sup>−</sup> mice were improved by activation of IRP1 activity with oral treatment by Tempol (4-hydroxy-2,2,6,6-tetramethylpiperidin-1 oxyl), a scavenger of reactive oxygen species, which was shown to destabilize the iron–sulfur cluster ligand of IRP1 and restore its IRE-binding activity (Ghosh et al., 2008; Wilcox and Pearlman, 2008). The neurodegenerative symptoms of *Irp2*−/<sup>−</sup> mice were also improved by deletion of one allele of ferritin-H chain, which limited iron sequestration and increased intracellular iron availability. These studies support that lack of IRE-binding activity of IRP2 and subsequent derepression of ferritin translation cause the neurodegenerative diseases of Irps deficient animals (Ghosh et al., 2008; Jeong et al., 2011). The neurodegenerative symptoms of *Irp2*−/<sup>−</sup> mice were also assessed by other two groups in independent *Irp2*−/<sup>−</sup> mouse colonies. Compared with the neurodegenerative symptoms characterized by ataxia, tremor and motor impairment associated with iron deposit in the white matter and nuclei throughout the brain in the *Irp2*−/<sup>−</sup> mice of our group (LaVaute et al., 2001; Jeong et al., 2011), one group found that while their*Irp2*−/<sup>−</sup> mice displayed discrete impairment of balance and/or motor coordination, they did not find iron deposit with Perl's-DAB iron staining and did not find evidence of neuronal degeneration in the brain (Galy et al., 2006); the other group found that their*Irp2*−/<sup>−</sup> mice had significant locomotor dysfunction and increased iron deposits in the cortex, mid-brain and cerebellum of *Irp2*−/<sup>−</sup> mice, but they did not find cellular degeneration with Fluoro-Jade staining (Zumbrennen et al., 2009). The locomotor dysfunction displayed by these three *Irp2*−/<sup>−</sup> mouse colonies confirms the significant role of Irp2 in mouse neuronal system and also suggests that the discrepancy is likely from either different mouse genetic backgrounds or differences in experimental assessement, but not due to off-target effects as was suggested before (Galy et al., 2006). Since IRP1 and IRP2 have redundant functions, deletion of one copy of *Irp1* in *Irp2*−/<sup>−</sup> mice probably could exacerbate the neurodegenerative symptoms and provide more information of IRPs function in the other two *Irp2*−/<sup>−</sup> mouse colonies. Further, conditional knockout of both *Irp1* and *Irp2* in neurons as in hepatocytes and enterocytes could shed light on the role of

IRPs in neuronal system, although complete loss of Irps has been lethal in several settings (Smith et al., 2006; Galy et al., 2008, 2010, 2013).

*Irp2*−/<sup>−</sup> mice also display microcytic hypochromic anemia (Cooperman et al., 2005; Galy et al., 2005). Erythroblasts of *Irp2*−/<sup>−</sup> mice have decreased TfR1 expression and iron levels, and increased protoporphyrin IX levels, supporting an iron-limited erythropoiesis model. Irp2 deficiency destabilizes TfR1 mRNA, and lack of TfR1 protein causes iron deficiency and eventually leads to microcytic anemia. The microcytic anemia of *Irp2*−/<sup>−</sup> mice suggests that Irp2 plays a dominant role in regulating iron homeostasis in erythroid cells (**Figure 1**). While the neurodegenerative symptoms of *Irp2*−/<sup>−</sup> are corrected by Tempol treatment, the treatment did not improve the microcytic anemia (Ghosh et al., 2008). Analogously, while Tempol treatment significantly converted Irp1 fraction from cytosolic aconitase form to IRE-binding form in forebrain lysate, which likely compensated the loss of IREbinding activity in the brain of *Irp2*−/<sup>−</sup> mice, Tempol treatment did not increase the IRE-binding activity of Irp1 in erythroblasts, likely because Irp1 is already mainly in the IRE-binding form in erythroblasts (Ghosh et al., 2008).

In addition to the microcytic anemia and neurodegenerative symptoms, *Irp2*−/<sup>−</sup> mice displayed iron overload in duodenum and liver, but interestingly, these animals also displayed iron deficiency in the spleen and bone marrow (Cooperman et al., 2005; Galy et al., 2005). Conditional deletion of *Irp2* in mouse duodenal enterocytes and liver hepatocytes repeats the iron overload phenotype of duodenum and liver, suggesting that iron overload in these two tissues is likely due to cell-autonomous functions of Irp2 deficiency in enterocytes and hepatocytes (Ferring-Appel et al., 2009). In contrast, conditional deletion of *Irp2* in mouse splenic macrophages did not produce the splenic iron deficiency seen in *Irp2*−/<sup>−</sup> mice, suggesting that the splenic iron deficiency of *Irp2*−/<sup>−</sup> mice is likely secondary to iron misregulation in other cell types. Considering that (i) both spleen and bone marrow macrophages of *Irp2*−/<sup>−</sup> mice display iron deficiency, (ii) both spleen and bone marrow are erythropoietic tissues, (iii) splenic macrophages play an essential role in recycling iron from senescent RBC, the iron deficiency of splenic macrophages probably results from reduced acquisition of iron from red cell turnover in anemic *Irp2*−/<sup>−</sup> mice. Conditional deletion of *Irp2* in erythroblasts might shed light on the pathogenesis of splenic iron deficiency.

## **PHYSIOLOGICAL SIGNIFICANCE OF IRP1**

Though both IRP1 and IRP2 are ubiquitously expressed and show similar binding affinities for target IREs in *in vitro* experiments, *Irp1*−/<sup>−</sup> and *Irp2*−/<sup>−</sup> mice display different phenotypes (Henderson et al.,1993;Kim et al.,1995). Because the protein levels and IRE binding activities of IRP1 in electrophoresis mobility shift assays are much higher than those of IRP2 in cultured cells and in certain tissues such as liver and kidneys, the physiological significance of IRP1 has been relatively elusive. In contrast to the neurodegeneration and anemia symptoms of *Irp2*−/<sup>−</sup> mice, *Irp1*−/<sup>−</sup> mice initially appeared asymptomatic and the physiological significance of IRP1 in iron metabolism was unclear. However, evidence from *Irp1*−/<sup>−</sup> mice has recently emerged, which indicates that IRP1

plays an essential role in regulation of systemic iron homeostasis and erythropoiesis.

expression of the iron storage protein, ferritin, the iron exporter FPN1, and some iron utilization-related genes including ACO2 and eALAS, as well as other potential target genes, to maintain optimal intracellular

## **PHYSIOLOGICAL SIGNIFICANCE OF IRP1 IN ERYTHROPOIESIS AND SYSTEMIC IRON HOMEOSTASIS**

In contrast to *Irp2*−/<sup>−</sup> mice that develop microcytic hypochromic anemia with hematocrits of about 36%, adult *Irp1*−/<sup>−</sup> mice produce more RBCs than wild type animals (hematocrit ∼50 vs. <sup>∼</sup>45%,*Irp1*−/<sup>−</sup> vs. wild type; Cooperman et al., 2005; Ghosh et al., 2013). High hematocrits were also observed in 4–6 weeks old mice of a different colony where two research groupsfound that*Irp1*−/<sup>−</sup> mice had hematocrits of more than 70% (Pawlus and Hu, 2013; Wilkinson and Pantopoulos, 2013). However, these researchers found that hematocrits of *Irp1*−/<sup>−</sup> mice decreased with age, and the difference of hematocrits between *Irp1*−/<sup>−</sup> and wild type mice disappeared after 8 weeks of age. We also checked the hematocrits of *Irp1*−/<sup>−</sup> mice in different ages (from 4 weeks to 14 months old), but did not find a significant decrease of the hematocrits of

mice older than 8 weeks of age (Zhang et al., unpublished data), and the reason for the inconsistent observations is not yet clear. More striking results are obtained from *Irp1*−/<sup>−</sup> mice that are maintained on an iron-deficient diet (Ghosh et al., 2013). In contrast to the general observation that iron deficiency causes iron deficiency anemia, maintenance on an iron-deficient diet significantly increased the hematocrits of *Irp1*−/<sup>−</sup> mice from about 50% to more than 60%. Serum erythropoietin (EPO) levels of *Irp1*−/<sup>−</sup> mice were more than seven-fold higher than that in wild type mice maintained on an iron-deficient diet, and animals with these high EPO levels developed splenomegaly and increased splenic erythropoiesis, supporting a model that there is EPO-dependent extracellular erythropoiesis in *Irp1*−/<sup>−</sup> mice. HIF2α is the master transcription factor *in vivo* that regulates EPO levels and subsequent RBC production according to both hypoxia and iron status (Semenza, 2009; Haase, 2013). HIF2α mRNA has an IRE in its 5- UTR with an initially unknown significance (Sanchez et al., 2007). In the renal carcinoma cell line,

accumulation of iron by ferritin depletes biologically available iron from the cytosol and leads to functional iron deficiency, mitochondrial

dysfunction and neuronal degradation.

786-O cells, HIF2α expression is increased by *Irp1* knockdown but not *Irp2* knockdown, suggesting that Irp1 probably plays a major function in repressing HIF2α translation by binding to its 5- IRE in these particular cells (Zimmer et al., 2008). Irp1 but not Irp2 deficiency significantly increased the percentage of HIF2α mRNA found in the polysomal fractions of mouse kidneys and liver, proving that Irp1 but not Irp2 represses HIF2α mRNA translation in cells of these two tissues *in vivo* (Anderson et al., 2013; Wilkinson and Pantopoulos, 2013). Thus, in *Irp1*−/<sup>−</sup> mice, Irp1 deficiency derepresses HIF2α translation, which transcriptionally increases EPO expression and subsequently drives red blood production, leading to polycythemia as a consequence (Anderson et al., 2013; Ghosh et al., 2013; Wilkinson and Pantopoulos, 2013). Notably, *Irp1*−/<sup>−</sup> mice have severe iron deficiency as evidenced by low serum iron levels and reduced stainable tissue iron compared with wild type animals, which is likely caused by increased

erythropoiesis that channels iron into the production of red cells, and consequently depletes Tf bound iron in blood and tissue iron stores.

The polycythemia of *Irp1*−/<sup>−</sup> mice and its exacerbation by iron deficiency suggest a crucial role of Irp1 in regulating systemic iron levels and erythropoiesis (**Figure 2**). HIF2α is regulated at the posttranslational level by prolyl hydroxylases (PHDs) that use oxygen and iron as substrates to hydroxylate HIF2α at two conserved proline residues. Following hydroxylation, HIF2α undergoes ubiquitination by the von Hippel–Lindau E3 ubiquitin ligase and is subsequently degraded through the proteasomal pathway (Majmundar et al., 2010; Lee and Percy, 2011; Yoon et al., 2011; Haase, 2013). Hence, HIF2α can sense hypoxia and iron deficiency, and then increases EPO expression and drives red blood production, a process that consumes large amounts of iron. The regulation of HIF2α translation by Irp1 provides a safeguard that prevents

**FIGURE 2 |The scheme of physiological significance of IRP1** *in vivo.* IRP1 is the predominant IRP protein in renal interstitial fibroblasts and pulmonary endothelial cells. In iron-replete conditions, IRP1 ligates an iron–sulfur cluster and displays cytosolic aconitase activity; in iron-depleted conditions, IRP1 loses its iron–sulfur cluster and binds to IREs of target mRNAs to regulate expression of iron metabolism related genes. By binding to the 5- IRE of HIF2α mRNA, IRP1 regulates HIF2α expression according to iron and oxygen status, and thereby fine-tunes the levels of HIF2α protein in cooperation with prolyl hydroxylases and the Von Hippel–Lindau mediated proteasomal degradation pathway. In hypoxia or iron deficiency conditions, HIF2α protein is stabilized due to inactivation of prolyl hydroxylases, and then translocates to nucleus and transcriptionally increase erythropoietin (EPO) expression. Circulating through blood, EPO binds to EPO receptors on erythroblasts and

stimulates erythroblasts to produce red blood cells (RBC), a process that consumes large amount of iron. When too much iron is consumed and systemic iron levels are low, IRP1 will be activated to reduce HIF2α expression and restrict RBC production. By this feedback mechanism, IRP1 regulates the balance between systemic iron homeostasis and erythropoiesis. In pulmonary endothelial cells, IRP1 regulates HIF2α translation and subsequently regulates the expression of endothelin-1 (ET-1), a peptide hormone that regulates pulmonary vascular contraction and the proliferation of smooth muscle cells, cardiomyocytes, and fibroblasts. Though the exact mechanism remains to be elucidated, IRP1 deficiency causes pulmonary hypertension and cardiovascular diseases in mice likely by derepression of HIF2α expression, increase of ET-1 levels and most probably other HIF2α targets also.

erythropoiesis from consuming too much iron to deplete systemic iron. While HIF2α senses hypoxia and stimulates EPO expression and RBC production, Irp1 fine-tunes HIF2α expression to ensure that there is enough iron available for iron–sulfur cluster synthesis; if intracellular iron levels are low, iron–sulfur cluster synthesis is impaired and IRP1 will be converted to the IRE-binding form, which represses HIF2α translation, and thereby decreases RBC production to restore systemic iron balance (**Figure 2**). *Irp1*−/<sup>−</sup> mice on an iron-deficient diet have increased polycythemia, severe iron deficiency, and sudden death due to peritoneal hemorrhage, which emphasizes the crucial role of Irp1 in systemic iron homeostasis and erythropoiesis (Ghosh et al., 2013). Maintenance of *Irp1*−/<sup>−</sup> animals in 10% oxygen for 3 weeks can increase the hematocrits of *Irp1*−/<sup>−</sup> mice to as high as 80%, compared to 65% in wild type mice, highlighting the significance of IRP1 in systemic iron homeostasis and erythropoiesis under hypoxia (Zhang et al., unpublished data). Mice that inducibly express a constitutively active IRP1 mutant (IRP1∗) develop macrocytic anemia, probably due to impaired erythropoiesis as displayed by increased erythroid progenitor and decreased numbers of mature cells (Casarrubea et al., 2013). Erythroblasts of IRP1∗ mice have higher TfR1 expression compared to wild type animals, which could potentially cause iron overload and impair normal erythropoiesis. In addition, high levels of Irp1 are expected to repress HIF2α translation and subsequently reduce EPO levels and RBC production. The macrocytic anemia of IRP1∗ highlights the importance of IRP/IRE balance in iron and erythropoiesis homeostasis.

In addition to its function in the regulation of erythropoiesis and systemic iron homeostasis, IRP1 probably also plays a role in coordinating intracellular Fe-S cluster and heme synthesis in erythroid cells (Wingert et al., 2005; Ye and Rouault, 2010; Chung et al., 2014). Because IRP1 has a [4Fe-4S] cluster as the sensor to adjust its IRE-binding activity, IRP1 can sense Fe-S cluster deficiency and potentially repress the translation of ALAS2 (eALAS) by binding to a IRE in its 5- UTR. As ALAS2 is the first enzyme of the heme synthesis pathway, translational repression by IRP proteins can coordinate the synthesis of Fe-S clusters and heme (Dandekar et al., 1991; Wingert et al., 2005; Chung et al., 2014). Deficiency of glutaredoxin 5 (GLRX5), a scaffold protein required for mitochondrial Fe-S cluster synthesis, activated IRP1, which inhibited ALAS2 translation and subsequently led to anemia in zebrafish, suggesting a role of IRP1 in coordinating Fe-S cluster and heme synthesis (Wingert et al., 2005). Consistent with the zebrafish studies, RNAi knockdown of GLRX5 in K562 cells markedly reduced ALAS2 expression, and deficiency of GLRX5 in a patient caused by an intronic mutation that caused missplicing significantly increased the IRE-binding activity of IRP1, and caused sideroblastic anemia (Camaschella et al., 2007; Ye et al., 2010). IRP1 deficiency significantly increased protoporphyrin levels in erythroid cells with heterozygous deficiency of mitoferrin1, the major mitochondrial iron importer in erythroid cells, also suggesting that IRP1 is an important link between Fe-S cluster and heme synthesis (Shaw et al., 2006; Chung et al., 2014). The evidence suggests that IRP1 likely plays an important role in orchestrating the heme and Fe-S cluster synthesis in erythroid precursors; however, the role of IRP2 was not really tested, considering that: (1) Zebrafish do not have an IRP2 homolog (Wingert et al., 2005); (2) The effect of

IRP2 deficiency on protoporphyrin levels in erythroid precursors with heterozygous mitoferrin1 deficiency was not checked (Chung et al., 2014); (3) IRP2 expression was also significantly increased in GLRX5 deficient fibroblasts (Ye et al., 2010); (4) *Irp2*−/<sup>−</sup> mice*,* but not *Irp1*−/<sup>−</sup> mice, have increased protoporphyrin IX levels (Cooperman et al.,2005). Thus, the roles of IRP1 and IRP2 in coordinating heme and Fe-S cluster synthesis in erythroid precursors may require further investigation.

## **PHYSIOLOGICAL SIGNIFICANCE OF IRP1 IN PULMONARY AND CARDIOVASCULAR SYSTEM**

In addition to the essential role of IRP1 in maintaining erythropoiesis and systemic iron homeostasis, IRP1 also plays an important role in pulmonary and cardiovascular system (**Figure 2**). *Irp1*−/<sup>−</sup> mice displayed cardiac hypertrophy and pulmonary hypertension, two severe human diseases with unclear pathogenesis (Ghosh et al., 2013). HIF2α has been previously implicated in the pathogenesis of pulmonary hypertension. Heterozygous deficiency of HIF2α protects mice against developing pulmonary hypertension and right ventricular dysfunction during prolonged hypoxia, suggesting that HIF2α is involved in pathogenesis of pulmonary hypertension (Brusselmans et al., 2003). Chuvash polycythemia is a hereditary disease caused byVHLR200W mutation that disrupts the degradation pathway of HIFα and consequently increases HIF2α protein to high levels (Hickey et al., 2007). Chuvash polycythemia patients and a corresponding mouse model with a VHLR200W mutation develop polycythemia and pulmonary hypertension, and the pulmonary hypertension of the mouse model is improved by deletion of one allele of *Hif2*α, suggesting that there is a pathogenic role of HIF2α in this disease (Bushuev et al., 2006; Hickey et al., 2010; Formenti et al., 2011). Endothelin-1, a potent vasoconstrictor and HIF target, is significantly increased in Chuvash mice and could be a downstream effector molecule in the pathogenesis of pulmonary hypertension (Bushuev et al., 2006; Hickey et al., 2010; Thorin and Clozel, 2010; Shao et al., 2011). HIF2α expression is significantly increased in cultured primary endothelial cells of *Irp1*−/<sup>−</sup> mice, and endothelin-1 expression is also significantly increased in lung tissues of *Irp1*−/<sup>−</sup> mice, supporting a pathogenic role of these molecules in the pulmonary hypertension (Ghosh et al., 2013). However, preliminary results of *Irp1*−/<sup>−</sup> mice did not reveal significant pulmonary vascular remodeling, in contrast to the Chuvash and the prolonged-hypoxia mouse models, and thus the mechanism underlying the pulmonary hypertension of *Irp1*−/<sup>−</sup> mice is not very clear yet. Low iron treatment significantly increases EPO expression and exacerbates the polycythemia of *Irp1*−/<sup>−</sup> mice, which could be attributed to the stabilization of HIF2α by iron deficiency; in contrast, the iron-deficient diet did not exacerbate the pulmonary hypertension of *Irp1*−/<sup>−</sup> mice, and the endothelin-1 levels were not altered either, suggesting that there are mechanistic differences in the pathophysiology of polycythemia and pulmonary hypertension. Analysis of the underlying molecular pathophysiology could shed light on the pathogenesis of these diseases. The IRP1/HIF2α interaction in kidneys represents a mechanism that protects systemic iron homeostasis during erythropoiesis by preventing red cells production from depleting systemic iron, whereas the evolutionary rationale for the

IRP1/HIF2α interaction in pulmonary vascular function is not yet clear (**Figure 2**).

Maintenance on an iron-deficient diet also significantly decreases life span of *Irp1*−/<sup>−</sup> mice. From the age of 3 months old, *Irp1*−/<sup>−</sup> mice on the iron-deficient diet are prone to sudden death in which apparently healthy mice die without warning (Ghosh et al., 2013). Pathological analyses of mouse carcasses revealed that these mice died of abdominal hemorrhage around the perinephric area, suggesting that there is an essential role of IRP1 in the vascular system. Considering that the iron-deficient diet exacerbates the polycythemia of *Irp1*−/<sup>−</sup> mice by stimulating HIF2α expression, HIF2α probably also plays a role in the peritoneal hemorrhage. Notably, Chuvash polycythemia patients, who have high HIF2α expression, also have increased incidence of thrombotic events, major bleeding episodes, and premature mortality, suggesting that there are probably similar mechanisms operating in Chuvash polycythemia patients and the *Irp1*−/<sup>−</sup> mouse models (Gordeuk and Prchal, 2006). Coagulation factor VIII (FVIII) is an essential blood-clotting protein, and high levels of FVIII are reported to associate with increased risk of deep vein thrombosis and pulmonary embolism (Jenkins et al., 2012). An alternative transcript variant 2 of FVIII has a potential 5- IRE that, if functional, could derepress FVIII expression in IRP1 deficient animals or patients (Livesey et al., 2012). The expression and role of FVIII in the phenotypes associated with the *Irp1*−/<sup>−</sup> mouse model warrant investigation.

## **DIFFERENT PHENOTYPES OF** *Irp1***−***/***<sup>−</sup> AND** *Irp2***−***/***<sup>−</sup> MICE ARE LIKELY CAUSED BY CELL-SPECIFIC EXPRESSION**

The relative functions of IRP1 and IRP2 have been a subject of debate since they were identified two decades ago. On one hand, IRP1 and IRP2 are highly conserved in sequence and displayed similar binding affinity to their target in *in vitro* experiments (Henderson et al., 1993; Kim et al., 1995); on the other hand, IRP1 and IRP2 have different regulatory mechanisms with IRP1 functioning as a bifunctional protein and IRP2 being degraded through the proteasomal pathway. The polycythemia and pulmonary hypertension phenotype of *Irp1*−/<sup>−</sup> mice and the anemia and neurodegeneration phenotype of *Irp2*−/<sup>−</sup> mice likely support a unique functions of IRP1 and IRP2 in erythropoiesis/cardiovascular regulation and erythroblasts/nervous system, respectively. However, a recent study, which analyzed the endogenous transcripts bound *in vitro* by overexpressed IRP1 and IRP2 proteins with microarray, showed that IRP1 and IRP2 shared 44 transcripts including the transcripts that had been confirmed in literature, i.e., FTL, FTH, TfR1, DMT1, FPN1, ACO2, eALAS, and HIF2α (Sanchez et al., 2011). Derepression of the shared target HIF2α most likely causes the polycythemia and pulmonary hypertension of *Irp1*−/<sup>−</sup> mice (Anderson et al., 2013; Ghosh et al., 2013; Wilkinson and Pantopoulos, 2013), and dysregulation of the shared targets ferritin, TfR1, and eALAS likely lead to the diseases of *Irp2*−/<sup>−</sup> mice (LaVaute et al., 2001; Cooperman et al., 2005; Galy et al., 2005; Ferring-Appel et al., 2009; Jeong et al., 2011; Ghosh et al., 2013). IRP1 expression is very high relative to IRP2 in the kidneys and lung, whereas IRP2 expression is very high in the brain, which is consistent with the symptoms of *Irp1*−/<sup>−</sup> and *Irp2*−/<sup>−</sup> mice in affected tissues, suggesting that the differences

in the phenotypes of *Irp1*−/<sup>−</sup> and *Irp2*−/<sup>−</sup> mice are most likely caused by differences in the cell-specific expression of these two proteins (**Figures 1** and **2**; Meyron-Holtz et al., 2004; Ghosh et al., 2008, 2013). Nevertheless, in addition to the 44 targets shared by both IRP1 and IRP2, microarray analysis of IRP-binding transcripts also identified 101 potential IRP1-specific targets and 113 potential IRP2-specific targets. The contribution of these targets in the pathogenesis of *Irp1*−/<sup>−</sup> and *Irp2*−/<sup>−</sup> mice is not clear yet, and characterization of these IRP1- or IRP2- specific targets could shed new light on the unique functions of IRP1 and IRP2 in iron metabolism and development. In addition to the different expression profiles of IRP1 and IRP2 in cells and tissues affected by IRP1 and IRP2 deficiency, we have to keep in mind that the expression profiles of other genes including IRP targets as well as other iron metabolism related genes are also unique in each of these cell types and tissues. Since IRP targets share the same IRE-binding protein pool, changes in mRNA expression of each of these IRP targets will create a unique gene expression context and thereby inevitably affect the binding of IRPs to other targets, which eventually leads to the diseases of Irp1 and Irp2 deficiency animals. Since there have been very few attempts to understand iron metabolism in gene expression context (Hower et al., 2009), application of systemic biology methodologies probably could elucidate a better picture of the molecular pathophysiology of disease in the *Irp1*−/<sup>−</sup> and *Irp2*−/<sup>−</sup> mice.

## **CLINICAL IMPLICATIONS**

The IRP/IRE system plays an essential role in iron homeostasis, and dysregulation of IRP/IRE system has been reported to cause many diseases. Mutations in the IRE element of human L-ferritin disrupt the regulation of IRPs on L-ferritin translation and result in high ferritin expression and early onset cataracts, causing hereditary hyperferritinemia cataract syndrome (Ismail et al., 2006). A mutation in the IRE of human H-ferritin causes autosomal dominant iron overload (Kato et al., 2001). Disruption of IRE elements in the mouse FPN1 promoter alters erythropoiesis and iron homeostasis, and induces age-dependent loss of photoreceptors of the retina (Mok et al., 2004a,b; Iacovelli et al., 2009). These diseases and phenotypes highlight the essential role of IRP/IRE machinery in iron metabolism and development. The neurodegeneration and anemia of *Irp2*−/<sup>−</sup> mice and the polycythemia and pulmonary hypertension of *Irp1*−/<sup>−</sup> mice underline the essential role of IRP/IRE machinery in regulating cellular and systemic iron homeostasis, and also suggest that mutations of IRP1 and IRP2 could underlie some human diseases. Though patients with diseases attributable to IRP2 mutations have not been identified, such patients could be treated by activating IRP1 with stable nitroxide, Tempol, or other nitric oxide sources, which could compensate for the loss of IRP2 and ameliorate disease, as demonstrated by alleviation of symptoms in Tempoltreated *Irp2*−/<sup>−</sup> mice (Lipinski et al., 2005; Ghosh et al., 2008; Jeong et al., 2011).

Polycythemia is a severe disease that stresses the cardiovascular system and endangers the lives of patients. A routine treatment for polycythemia is phlebotomy to remove excess blood from patient. The essential role of IRP1 in regulating HIF2α translation, as displayed by *Irp1*−/<sup>−</sup> mice, suggests that activation of IRP1 by Tempol or nitric oxide could repress HIF2α translation and thereby decrease EPO expression and reduce RBC production. Activation of IRP1 could be a therapeutic strategy to treat Chuvash polycythemia patients. Erythropoiesis-stimulating agents have been widely used to treat anemia of chronic diseases, and the significant role of IRP1 in repressing HIF2α and EPO production suggests that inhibiting the interaction between IRP1 and HIF2α with small molecules could increase HIF2α translation and upregulate endogenous EPO levels (Zimmer et al., 2008). Phenotypes of *Irp1*−/<sup>−</sup> mice also suggest that IRP1 mutations could cause idiopathic polycythemia and pulmonary hypertension in some patients, and IRP1 should be screened as a candidate disease gene in these patients. *Irp1*−/<sup>−</sup> mice provide a novel mouse model for pulmonary hypertension, cardiac hypertrophy and aneurysm, and investigations of this model will provide insights into the molecular mechanisms of these diseases. Iron deficiency anemia is one of the most common diseases worldwide. The translational derepression of HIF2α and subsequent increase of EPO levels in *Irp1*−/<sup>−</sup> mice prove that IRP1 plays a critical role in balancing erythropoiesis and systemic iron homeostasis during iron deficiency; these observations explain why the first manifestation of iron deficiency is usually anemia.

## **FUTURE DIRECTIONS**

The phenotypes of *Irp1*−/<sup>−</sup> and *Irp2*−/<sup>−</sup> mice provide compelling evidence for the essential role of the IRP/IRE system in systemic iron homeostasis and physiology, suggesting that IRP1 or IRP2 mutations could cause human diseases similar to those discovered in animal models. Screening human patients and identifying IRP1 or IRP2 mutations could deepen our understanding of their roles in human iron metabolism. *Irp1*−/<sup>−</sup> mice on low iron diet died from peritoneal hemorrhage, and the physiological function of IRP1 in the pulmonary and cardiovascular system is not clear yet. Low iron treatment exacerbates the polycythemia but not pulmonary hypertension, and the molecular mechanism remains to be elucidated. Pharmacological activation of IRP1 could be a therapeutic strategy to treat Chuvash polycythemia and pulmonary hypertension, and inhibition of the interaction between IRP1 and HIF2α could increase EPO production and potentially treat anemia, and thus modulations of IRP1 activity should be investigated further. As IRP1 is a bifunctional enzyme, *Irp1*−/<sup>−</sup> mice lose both the IRE-binding activity and cytosolic aconitase activity, and the contribution and significance of cytosolic aconitase activity of IRP1 *in vivo* merits further investigation. Microarray analysis of IRP1- and IRP2- associated mRNAs identified 35 novel mRNAs that can bind both IRP1 and IRP2, as well as mRNAs bind exclusively to either IRP1 or IRP2, and experimental analysis and characterization of those mRNAs could advance our understanding of the function of IRP/IRE system in systemic iron metabolism (Sanchez et al., 2011).

## **ACKNOWLEDGMENT**

This work was supported by the intramural programs of NICHD.

## **REFERENCES**

Abboud, S., and Haile, D. J. (2000). A novel mammalian iron-regulated protein involved in intracellular iron metabolism. *J. Biol. Chem.* 275, 19906–19912. doi: 10.1074/jbc.M000713200


regulatory protein 2 (IRP2). *Blood* 106, 2580–2589. doi: 10.1182/blood-2005-04- 1365


polycythaemia mice. *Brain Res.* 1289, 85–95. doi: 10.1016/j.brainres.2009. 06.098


transfer of iron to the circulation. *Mol. Cell* 5, 299–309. doi: 10.1016/S1097- 2765(00)80425-6


Zumbrennen, K., Holter, S., Becker, L., Rathkolb, B., Rodansky, E., and Leibold, E. (2009). A novel Irp2-/- mouse model displays locomotor dysfunction and neuronal iron accumulation. International BioIron Society Meeting. *Am. J. Hematol*. 84, E257.

**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

*Received: 04 April 2014; paper pending published: 16 April 2014; accepted: 10 May 2014; published online: 13 June 2014.*

*Citation: Zhang D-L, Ghosh MC and Rouault TA (2014) The physiological functions of iron regulatory proteins in iron homeostasis - an update. Front. Pharmacol. 5:124. doi: 10.3389/fphar.2014.00124*

*This article was submitted to Drug Metabolism and Transport, a section of the journal Frontiers in Pharmacology.*

*Copyright © 2014 Zhang, Ghosh and Rouault. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*

## Iron, hepcidin, and the metal connection

## *Olivier Loréal1,2,3 \*,Thibault Cavey1,2,4 , Edouard Bardou-Jacquet1,2,3 , Pascal Guggenbuhl 1,2,5 , Martine Ropert1,3,4 and Pierre Brissot1,2,3*

<sup>1</sup> INSERM UMR 991, Iron and the Liver Team, Rennes, France


#### *Edited by:*

Paolo Arosio, University of Brescia, Italy

#### *Reviewed by:*

Stanislav Yanev, Institute of Neurobiology – Bulgarian Academy of Sciences, Bulgaria Donatella Barisani, University of Milano Bicocca, Italy

#### *\*Correspondence:*

Olivier Loréal, INSERM UMR 991, Iron and the Liver Team and CHU Pontchaillou, French Reference Centre for Rare Iron Overload Diseases of Genetic Origin, University Hospital-Rennes, 35033 Rennes Cedex, France e-mail: olivier.loreal@inserm.fr

Identification of new players in iron metabolism, such as hepcidin, which regulates ferroportin and divalent metal transporter 1 expression, has improved our knowledge of iron metabolism and iron-related diseases. However, from both experimental data and clinical findings, "iron-related proteins" appear to also be involved in the metabolism of other metals, especially divalent cations. Reports have demonstrated that some metals may affect, directly or indirectly, the expression of proteins involved in iron metabolism. Throughout their lives, individuals are exposed to various metals during personal and/or occupational activities. Therefore, better knowledge of the connections between iron and other metals could improve our understanding of iron-related diseases, especially the variability in phenotypic expression, as well as a variety of diseases in which iron metabolism is secondarily affected. Controlling the metabolism of other metals could represent a promising innovative therapeutic approach.

**Keywords: iron, metal, metabolism, disease, ferroportin, DMT1, transferrin**

## **INTRODUCTION**

Iron is a metalfound at very high levels in Earth's crust that is essential for cell life due to its major role in most biological systems and metabolic pathways through its involvement in oxygen transport and delivery, and participation in a large number of enzymatic processes (Crichton, 2001a). In addition, iron may participate in the genesis of reactive oxygen species (ROS) through the Fenton and Haber–Weiss reactions (Wardman and Candeias, 1996). Production of ROS alters proteins, lipids, and DNA during oxidative stress favoring the development of cellular alterations that lead to the development of lesions in organs (Meneghini, 1997; Crichton, 2001b).

During the beginning of the third millenary, knowledge of iron metabolism has consistently progressed with the identification of numerous genes encoding iron metabolism proteins, including hepcidin, and a better understanding has been gained of systemic iron metabolism. These discoveries led to the identification of a strong potential relationship between iron and other metals, contributing to an understanding of previous unexplained findings described primarily during the 20th century.

Here, our objective is to provide evidence of true metal interactions that likely represent a way of better understanding metal metabolism and metal-related diseases in the future.

## **IRON METABOLISM PLAYERS**

Iron is a transition metal with a metabolism characterized by the following: (i) a major role of iron found in plasma (1–2 mg), which is delivered to every cell in the body; (ii) constant recycling of iron associated with hemoglobin within erythrocytes, followed by iron export toward plasma (20 mg/day); (iii) the absence of an active, iron excretory process, and iron losses (1–2 mg/day) are uncontrolled; and (iv) a tuning of digestive absorption that must strictly compensate for iron losses (Andrews, 1999, p. 677). Thus, most of the iron required by the organism daily (20 mg/day) is provided by the erythrophagocytosis process that occurs within macrophages.

In plasma (Andrews, 1999), iron is linked in its ferric form to transferrin, a β-globin protein of hepatocytic origin. The iron saturation of transferrin is 30–45%. Iron-transferrin is taken up by cells through transferrin receptor 1 via an endocytic process. An additional step occurs in the endocytic vesicle and involves divalent metal protein 1 (DMT1; Fleming et al., 1997, 1998; Gunshin et al., 1997), which is encoded by the SLC22a1 gene. After reduction of ferric iron by the STEAP 3 protein (Lambe et al., 2009), the iron is transferred to the cytosol, where it becomes available for cells. Plasma iron is directed mainly to erythroblasts, where it associates with heme for hemoglobin synthesis (Andrews, 1999, p. 677). In some cases, mainly during ironoverload diseases, non-transferrin-bound iron may appear in the plasma (Hershko and Peto, 1987; Brissot et al., 2012). Nontransferrin-bound iron associates mainly with low-molecularweight molecules. This form of iron can be avidly taken up by parenchymal cells, especially hepatocytes (Brissot et al., 1985). Proteins expressed at the hepatocyte membrane level, such as ZIP14 (Liuzzi et al., 2006), could participate in this process, conferring to hepatocytes, and more globally the liver, a major

role in the control of excess iron in the plasma by offering an iron-storage site.

Macrophages ensure recycling of senescent erythrocytes (Andrews, 1999). The last phase of this process involves the ferroportin protein (SLC40a1 gene), which transfers ferrous iron toward the plasma (Donovan et al., 2000; McKie et al., 2000). Next, iron is oxidized by the ceruloplasmin protein, a multicopper oxidase secreted by the liver (Osaki et al., 1966; Miyajima et al., 1987; Harris, 1995). Iron is subsequently taken in charge by transferrin (Andrews, 1999).

Iron losses are mostly related to cutaneous and digestive cell peeling, urine leakage, and regular (menstruations) or occasional bleeding.

Digestive iron absorption occurs mostly in the duodenum and involves enterocytes (Andrews, 1999). Non-heme ferric iron is taken up from the digestive lumen by DMT1 (Fleming et al., 1997, 1998; Gunshin et al., 1997) after an oxidation process that could involve Dcytb, a ferrireductase protein (McKie et al., 2001). Heme iron could be taken up from nutrients by a specific transporter. Heme carrier protein 1 is a candidate (Shayeghi et al., 2005), allowing further heme processing by heme-oxygenase 1 within enterocytes, and then iron reaches the cytoplasmic iron pool (Fleming et al., 1997, 1998; Gunshin et al., 1997). In both cases, the transfer of iron in the plasma as ferrous iron is ensured by the ferroportin protein (Donovan et al., 2000; McKie et al., 2000). Iron is then oxidized in ferric iron by hephaestin, another multicopper oxidase anchored in the cell membrane at the basolateral level (Vulpe et al., 1999; Anderson et al., 2002), and/or ceruloplasmin, allowing iron to interact with plasma transferrin and its delivery to other cells (Andrews, 1999).

### **SYSTEMIC REGULATORS OF IRON METABOLISM**

Regulation of systemic iron metabolism involves hepcidin, a cysteine-rich, 25 amino acid, iron-inducible peptide secreted by the liver (Nicolas et al., 2001; Park et al., 2001; Pigeon et al., 2001). Schematically, serum hepcidin may interact with ferroportin protein localized on the cell membrane of macrophages and enterocytes, inducing its internalization and degradation (Nemeth et al., 2004). The global effect is a control of iron egress from cells, avoiding increased transferrin saturation, which would expose to the appearance of non-transferrin-bound iron and to the development of parenchymal iron overload, as observed during genetic hemochromatosis (Loréal et al., 2005).

Iron is then sequestered in macrophages or enterocytes within ferritin, the iron-storage protein (Munro and Linder, 1978; Theil, 1987). In macrophages, iron is mobilized toward the plasma as required based on the plasma hepcidin level. Thus, iron can be available immediately after a decrease in hepcidin. In enterocytes, stored iron (corresponding to the classical "mucosal block") is lost during cell peeling. This mechanism is related to the low ratio between iron originating every day from enterocytes compared to macrophages and likely occurs to control iron storage over a longer period (Andrews, 1999).

Hepcidin regulators play a critical role in the maintenance of adequate serum hepcidin levels and, therefore, in the control of serum iron and systemic iron homeostasis (Muckenthaler, 2008). Inducers of hepcidin expression by hepatocytes include

excess iron (Pigeon et al., 2001) and inflammation (Nemeth et al., 2003). Increased hepcidin gene transcription related to iron involves the hemojuvelin/bone morphogenetic/SMAD (HJV/BMP/SMAD) pathway, which is activated following BMP6 over-expression (Andriopoulos et al., 2009; Meynard et al., 2009). In addition, HFE and TFR2 proteins, which are both expressed on the hepatocyte cell membrane, participate in the induction of hepcidin expression under conditions of excess iron. Although a role of increased serum transferrin saturation has been proposed (Goswami and Andrews, 2006; Muckenthaler, 2008), the molecular mechanism involved and the interaction with the HJV/BMP/SMAD pathway are not fully understood. Such regulation contributes to limiting iron excess. The control of hepcidin levels is deficient during genetic hemochromatosis related to *HFE, TFR2*, *HJV*, or *HAMP* mutation in humans, leading to iron-overload diseases (Brissot et al., 2011). Inflammation strongly increases hepcidin expression through the IL6/STAT3 pathway (Wrighting and Andrews, 2006; Pietrangelo et al., 2007; Verga Falzacappa et al., 2007), contributing to limited iron bioavailability for either growing pathogenic agents during infectious disease or oxidative stress during chronic inflammation.

Factors reducing hepcidin expression include hypoxia. Such impact of hypoxia increases serum iron and transferrin saturation, allowing intense erythropoiesis to compensate for tissue hypoxia (Muckenthaler, 2008). Whether the impact of hypoxia on *HAMP* transcription involves a stimulation of hypoxia inducible factor (HIF; Peyssonnaux et al., 2007) remains to be definitively determined. Recent arguments suggest that factors secreted during erythropoiesis could impact hepatocytes and limit hepcidin expression (Muckenthaler, 2008; Zhao et al., 2013).

Taken together, these findings put the focus on the hepcidin and ferroportin duo.

#### **REGULATION OF INTRACELLULAR IRON METABOLISM**

Within cells, an integrated system ensures the control of total iron content and distribution. The iron responsive element (IRE) and iron regulatory protein (IRP) together control the expression of proteins encoded by mRNA exhibiting an IRE nucleotide motif localized in the 5- UTR or 3- UTR (Hentze et al., 1987, 1988). When IRPs are active, they interact with the IRE, limiting the expression of proteins, such as ferritin, which has an IRE in its 5- UTR, and stabilizing mRNA, such as transferrin receptor 1 mRNA, which displays IRE motifs in its 3- UTR. This interaction is promoted by intracellular iron deficiency, the global effect being the promotion of iron ingress into the cell and associating with proteins that require this metal to reach their full activities. Conversely, in the presence of iron excess, the decrease in IRP activities leads to increased ferritin protein expression, favoring the storage of iron in a chemically inactive form, and decreased cellular iron entry due to a strong decrease in transferrin receptor 1 mRNA expression. Together, these processes avoid the production of ROS by limiting the amount of iron available for their production.

There are other iron-related proteins coded by mRNAs containing an IRE in their 5- UTR or 3- UTR, including DMT1 and ferroportin (Gunshin et al., 1997; McKie et al., 2000).

## **IRON-RELATED DISEASES CAN BE ASSOCIATED WITH ALTERATIONS IN THE METABOLISM OF OTHER METALS**

Iron homeostasis is lost during iron-related diseases. Schematically, three conditions can be found: (i) true iron deficiency related to insufficient intake, malabsorption, or excessive losses (Miller, 2013); (ii) iron misdistribution linked to systemic inflammation or cell-specific processes (Ganz and Nemeth, 2009); and (iii) systemic iron excess associated with genetic iron overload (Brissot et al., 2011) or anemia related to hematological cause (Gardenghi et al., 2010), with or without transfusions. Clinical, biological, and genetic characterization of these conditions pinpoints the links between iron and other metals.

#### **IRON DEFICIENCY**

Iron deficiency has been associated with abnormal absorption of metals from the digestive lumen. Pollack et al. (1965) reported a significant impact of iron deficiency, which was different according to etiology. In the rat, anemia consecutive to bleeding induced the absorption of manganese, cobalt, and iron. During iron deficiency related to poor iron intake, zinc absorption was increased (Pollack et al., 1965). Notably, a biological marker of chronic iron deficiency is an increase in zinc-protoporphyrin, which reflects the substitution of iron by zinc as a substrate for ferrochelatase during the last step of heme synthesis (Labbe et al., 1999). Conversely, absorption of calcium, magnesium, mercury, and copper was not significantly affected. Interestingly, supplementation of the diet with iron did not modify cobalt and manganese hyperabsorption. More recently, it was reported (Nam and Knutson, 2012) that iron status, especially iron deficiency, may increase the expression of ZIP 5, a zinc transporter, and conversely decreases the hepatic expression of ZIPs 6, 7, and 10. Moreover, iron-deficient rats had higher hepatic copper concentrations. The authors underlined that zinc transporters could play a role in hepatic iron/metal homeostasis during iron deficiency.

In particular, the relationship between iron deficiency and non-iron metals has been investigated in the brain. Iron deficiency may increase zinc concentration in the midbrain and hippocampus, whereas copper concentrations have been reported to be increased in the cerebral cortex and corpus striatus (Shukla et al., 1989). In addition, Erickson (Erikson et al., 2004) showed that iron deficiency induced an increase in manganese concentration in the putamen, globus pallidus, and substantia nigra. Zinc concentrations were also increased. The authors suggested that an increase in DMT1 expression related to iron deficiency could be involved. In the same way, iron depletion and loading increased brain manganese concentrations in young rats; this impact remained significant for 9 weeks. Moreover, the uptake of manganese by the brain, liver, kidneys, and bones was significantly increased by excess iron in younger rats. Manganese supplementation increased radioactive iron uptake by the brain, liver, and kidneys of rats receiving control and Fedeficient diets compared to rats supplemented with dietary iron (Chua and Morgan, 1996). DMT1 could be involved. Indeed, in Belgrade rats, the DMT1 mutation similarly affected manganese and iron metabolism, suggesting that they share similar transport mechanisms in the cells of erythroid tissue, duodenal

Another argument for an impact of iron deficiency on the metabolism of other metals is that iron deficiency confers a susceptibility to tissue accumulation of heavy, potentially toxic, metals, such as cadmium, nickel, and lead (Six and Goyer, 1972; Flanagan et al., 1978; Tandon et al., 1993). Mechanisms may include increased digestive absorption and metal accumulation in tissues, such as has been reported for cadmium (Flanagan et al., 1978). For nickel, a dynamic study in rats argued for increased absorption and decreased excretion (Tallkvist and Tjalve, 1997).

#### **IRON MISDISTRIBUTION**

The most frequent cause of iron misdistribution is the inflammatory process, which is the second most common etiology of anemia worldwide through the anemia of chronic disease (ACD; Weiss and Goodnough, 2005). Regarding iron metabolism, ACD is characterized by macrophagic iron sequestration and decreased iron absorption. The underlying mechanism involves increased plasma hepcidin levels, which limit the expression of ferroportin on the cell membrane (Ganz, 2011). Such situations have been associated with alterations in the metabolism of other metals. Thus, zinc and copper serum concentrations were reported to be increased in an acute model of inflammation (Milanino et al., 1986). Notably, zinc plays a major role in inflammation and the immune response, and zinc supplementation may improve innate immunity during inflammatory/infectious processes in acute, septic models (Knoell et al., 2009; Bao et al., 2010). Recently, zinc deficiency was reported to up-regulate the JAK/STAT3 pathway and could contribute to the severity of inflammation (Liu et al., 2014). The concentrations of calcium, strontium, and iron are increased in neutrophil granules, but the manganese increase in leukocytes was not localized to the granules (Hallgren et al., 1989). Whether the impact of inflammation on non-iron metals is related to decreased iron bioavailability and/or due to other independent mechanisms, including cytokine production, is not known.

Regarding heavy metals, a recent report described a role for ZIP14, which is involved in the cellular uptake of non-transferrinbound iron and cadmium accumulation during inflammation (Min et al., 2013).

## **SYSTEMIC IRON OVERLOAD**

Systemic iron-overload diseases include genetic iron overload – involving mutations in iron-related genes –, and secondary iron overload associated with hematologic diseases, and iron excess associated with liver diseases.

Regarding genetic iron-overload diseases, reports emphasize disturbances in other metals. Thus, during HFE-related hemochromatosis, in which low hepcidin levels lead to an abnormal increase in both digestive iron absorption and macrophagic iron release, an increase in hepatic zinc concentration has been reported (Adams et al., 1991), whereas plasma zinc concentration was normal (Brissot et al., 1978).

Abnormalities in manganese metabolism have been reported, in addition to those reported by Chua and Morgan (1997). In a mouse model of genetic hemochromatosis, Jouihan et al. (2008) showed that mitochondrial manganese uptake was altered, leading to mitochondrial dysfunction. Moreover, Kim et al. (2013) reported that the digestive absorption of manganese was strongly increased in *Hfe*−/<sup>−</sup> mouse model further emphasizing the relationship between iron and manganese metabolism.

A metal for which strong interactions with iron have been reported is cobalt. Cobalt may mimic iron deficiency by stabilizing HIF and, in turn, induce a large number of genes related to hypoxia and iron metabolism (Schuster et al., 1989; Yuan et al., 2003; Karovic et al., 2007). Cobalt may also reduce hepcidin expression by hepatocytes without involvement of the transcriptional factor HIF-1 (Braliou et al., 2008). Digestive absorption of cobalt was increased in hemochromatotic patients or patients exhibiting liver cirrhosis complicated by iron overload (Valberg et al., 1969; Olatunbosun et al., 1970). In patients with hepatic steatosis or cirrhosis and normal iron status, digestive absorption of cobalt and iron was not affected compared to controls. Conversely, in patients with cirrhosis and iron deficiency, both cobalt and iron absorption were increased to similar levels as a group of patients exhibiting iron deficiency alone.

Blood lead concentration was found to be increased during genetic hemochromatosis, in contrast with transfusional iron overload (Barton et al., 1994). Iron depletive treatment performed by phlebotomies in genetic hemochromatotic patients induced an increase in cadmium uptake (Akesson et al., 2000).

## **ALTERATIONS OF NON-IRON METALS MAY IMPACT IRON METABOLISM**

Numerous reports have studied the impact of modulations of noniron metals on iron metabolism. Here, we will focus on copper, zinc, cobalt, manganese, and lead.

## **COPPER**

Alterations in copper metabolism may strongly affect iron metabolism. Thus, the discovery of mutations in the ceruloplasmin gene provided an explanation for the peculiar phenotype of systemic iron overload involving the brain, which contrasts low plasma iron levels and concomitant anemia (Miyajima et al., 1996; Gitlin, 1998; Loréal et al., 2002). No significant alterations in copper metabolism were found. These effects are related to the role of ceruloplasmin, a multicopper oxidase, in oxidizing ferrous iron before its transferrin linkage in plasma. This role of ferroxidase activity has been reported for many years (Osaki et al., 1966). Notably, copper metabolism is a therapeutic target during Wilson disease, which is characterized by a toxic accumulation of copper due to a defect in the ATP7B gene (Tanzi et al., 1993; Huster, 2010). Thus, zinc oral supplementation has been used to limit copper absorption. The mechanisms involved are competition between zinc and copper, as well as an induction of enterocyte metallothioneins by zinc, as they can link copper as well as zinc (Schilsky et al., 1989). However, whether zinc is associated with copper chelating therapy during the active or maintenance phase of treatment remains to be discussed (Schilsky, 2009). Recently, the impact of zinc on copper metabolism was reinforced by the description of a new form of systemic iron overload (Videt-Gibou et al., 2009) related to secondary aceruloplasminemia resulting from excessive zinc intake (Nations et al., 2008). The defect is corrected by copper supplementation.

### **ZINC**

Modulations in zinc metabolism may also affect iron metabolism. Thus, in swine, zinc supplementation induces liver iron depletion without modulation of hepatic copper content (Cox and Hale, 1962). In rats, zinc supplementation decreased growth and favored anemia (Cox and Harris, 1960). O'Neil-Cutting et al. (1981) demonstrated that, in rats receiving low amounts of copper, a zinc-enriched diet induced anemia and low hepatic copper concentrations. This effect was not observed in animals with a balanced zinc diet, supporting the potential impact of zinc on erythropoiesis when appropriate cofactors exist. Another study suggested that zinc supplementation during gestation and lactation could have a differential effect on liver iron content, whereas copper content is not affected (Ketcheson et al., 1969). Taken together, these reports support that the potential effects of zinc on other metals should be evaluated, despite the presence of studies supporting recommendations for zinc supplementation (Hess and King, 2009). Moreover, despite numerous publications on the impact of zinc on the brain, especially during neurodegenerative diseases in which abnormal excess levels of iron have been reported, the effect of zinc appears to be ambiguous (review in Kawahara et al., 2014). Zinc has been proposed as an antioxidant molecule to improve neurodegenerative disease. However, zinc has also been suspected to favor Alzheimer's disease and neuronal death. Knowing the potential impact of zinc on iron metabolism, and whether this effect in neurodegenerative disorders is partly related to abnormal iron metabolism in the brain, warrants further investigation.

## **COBALT**

In trace amounts, cobalt is essential, as it is an integral part of the vitamin B12 complex and has a physiological impact on iron metabolism by contributing to erythropoiesis (Simonsen et al., 2012). In addition, cobalt supplementation may facilitate tolerance to hypobaric hypoxia (Shrivastava et al., 2008), which could be related to the modulation of HIF-regulated genes in order to promote oxygenation. Cobalt chloride supplementation has been evoked as a potential doping strategy in athletes (Lippi et al., 2005). To date, cobalt is not explicitly prohibited by world anti-doping agencies, despite its potential toxicity in the case of abnormal exposure. An increase in hemoglobin/hematocrit levels and polycythemia has been recorded in humans, rats, and dogs exposed to cobalt, demonstrating a strong impact on iron metabolism (see the US Agency for Toxic Substances and Disease Registry: http://www.atsdr.cdc.gov/toxprofiles/tp33.pdf).

## **MANGANESE**

Manganese, an essential component of metalloenzymes, is also essential for cell life. One of the main manganese-requiring enzymes is manganese superoxide dismutase, which plays a major role in counteracting oxidative stress, especially with iron, by detoxifying the superoxide radicals (Martinez-Finley et al., 2013). Some published data support an impact of manganese on iron metabolism. In cattle, oral manganese supplementation in animals receiving a low copper diet leads to decreased DMT1 expression in enterocytes. In addition, the down-regulation of hepcidin and ferroportin mRNA was found in the liver of animals receiving a copper-deficient diet alone (Hansen et al., 2010). In rats, during the neonatal period in animals receiving an iron-deficient diet, manganese supplementation of dams was reported to increase brain levels of manganese, chromium, zinc, cobalt, aluminum, molybdenium, and vanadium in the pups. In addition, iron decreased and copper increased in the brain (Garcia et al., 2007).

## **LEAD**

For many years, lead exposure was reported to strongly modulate iron metabolism. Multiple mechanisms are likely involved in this interaction. As known for a long time, excess lead inhibits δ-aminolevulinic acid dehydratase and ferrochelatase activities, in particular, induces zinc protoporphyrin accumulation in erythrocytes, and favors the occurrence of microcytic hypochromic anemia (Ku et al., 1990; Braun, 1999). In addition, lead has been reported to limit the transfer of iron from endosomes toward the cytoplasm (Qian et al., 1997). More recently, lead exposure in rats was reported to decrease serum iron and transferrin saturation levels (Moshtaghie et al., 2013). In workers exposed to lead, copper, and ceruloplasmin serum concentrations are increased, but no significant alteration in iron and zinc serum levels has been found (Kasperczyk et al., 2012). Potential interactions between lead and ceruloplasmin protein may explain a decrease in ceruloplasmin-linked ferroxidase activity (Leelakunakorn et al., 2005). *In vitro*, lead decreased transferrin synthesis in a human hepatic cell line (Barnum-Huckins et al., 1997). However, such an impact has not been reported in lead-exposed workers (Kasperczyk et al., 2012).

## **MOLECULAR EVENTS AND PATHWAYS LINKING IRON METABOLISM TO NON-IRON METAL METABOLISM A ROLE FOR PROTEINS INVOLVED IN IRON TRANSPORT**

There are proteins playing a role in the processes contributing to maintain iron homeostasis that have been associated to non-iron metal metabolisms (**Figure 1**).

DMT1 is a potential major link between iron and other divalent cations. The first description of DMT1 was associated with iron metabolism, especially in enterocytes and erythroblasts (Gunshin et al., 1997). However, DMT1 was also demonstrated to be able to take in charge other metals (Gunshin et al., 1997). Such a potential role of DMT1 with metals other than iron was reinforced by recent data showing that the efficacy of metal transport was very important for Cd2, Fe2+, Co2+, and Mn2+, and lesser for Zn2+, Ni2+, and Vo2+. However, the authors found that DMT1 expression did not stimulate the transport of Cr2+, Cr3+, Cu+, Cu2+, Fe3+, Ga3+, Hg2+, or VO<sup>+</sup> (Illing et al., 2012). In addition, iron processing by DMT1 was competitively inhibited by Co2<sup>+</sup> and Mn2+.

Ferroportin is also a candidate for interactions between iron and other metals. Ferroportin was initially reported as the iron exporter. However, recent data suggest that manganese, zinc, and cobalt could also be taken in charge by ferroportin (Troadec et al., 2010;Yin et al., 2010). This finding was confirmed by data showing that ferroportin expression in *Xenopus* oocytes enhance the efflux of 65Zn and 57Co but not 64Cu, 109Cd, or 54Mn. In addition, iron, zinc, and cobalt egress were inhibited by oocyte exposure to hepcidin (Mitchell et al., 2014).

Transferrin is the major plasma protein involved in iron delivery to cells. However, some data suggest that cobalt competes with iron on transferrin (Harrington, 1992). Thus, biochemical determination of non-transferrin-bound iron levels can be altered in the presence of cobalt (Gosriwatana et al., 1999). Others have suggested that transferrin may interact with manganese (Aschner and Gannon, 1994), indium, bismuth (Li et al., 1996; Zhang et al., 2004), copper (O'Neil-Cutting et al., 1981), and chromium (Harris, 1977). Transferrin has also been reported to interact with lead when present in excess (Leelakunakorn et al., 2005). In addition, when saturated by non-iron metals, transferrin could compete with iron-transferrin during its interaction with transferrin receptor 1 (Ha-Duong et al., 2008; El Hage Chahine et al., 2012). Taken together, these data suggest that metals other than iron could, by using transferrin as a Trojan horse, play a role in metal distribution within the body. This could also theoretically modulate the kinetics of iron ingress into cells. However, the impact on iron uptake is likely moderate due to the ratios of concentrations of different metals and iron in plasma and to a lower affinity of these non-iron metals for transferrin receptor 1 (El Hage Chahine et al., 2012). The low amount of non-iron metal linked to transferrin has been recently confirmed in mouse serum (Herrera et al., 2014). The same authors studying wild-type mice and transferrin-deficient mice also showed that transferrin does not play a major role in the delivery of manganese, copper, or zinc to tissues. Moreover, they suggest that an increase of tissue manganese found in transferrin deficient mice is linked to an indirect effect of transferrin deficiency on hepcidin expression or iron metabolism (Herrera et al., 2014). The interaction of transferrin with In3<sup>+</sup> and Cu2<sup>+</sup> has been shown to induce conformational changes similar to Fe3+; in addition, Al3<sup>+</sup> causes a conformational change of a somewhat smaller magnitude, whereas Hf4<sup>+</sup> (hafnium) does not induce significant conformational changes (Grossmann et al., 1993). Gallium transferrin is also taken in charge by the transferrin receptor (Chikh et al., 2007).

Taken together, these findings suggest that physiological or pathological modulation of the expression or activity of iron proteins, such as DMT1 or ferroportin, as well as transferrin, could modify the metabolism of other metals.

## **A ROLE FOR HEPCIDIN IN THE METABOLIC CONTROL OF NON-IRON METALS**

Hepcidin, the key regulator of iron bioavailability in plasma, exerts its role by controlling the expression of ferroportin protein, its main target, on the cell membrane (**Figure 1**).

Hepcidin may also modulate DMT1 expression. DMT1 expression and activity were decreased *in vitro* at the apex of Caco-2 cells after exposure of the basolateral part of the cells to hepcidin (Brasse-Lagnel et al., 2011). Notably, during genetic hemochromatosis, DMT1 and ferroportin are highly expressed in enterocytes, as observed during iron deficiency, and the inverse correlation between serum iron parameters and protein expression found in enterocytes of iron-deficient patients disappeared in genetic hemochromatosis patients (Zoller et al., 2001). Whether hepcidin plays a role on the modulation of DMT1 in addition to its impact on ferroportin has not been addressed. Molecular mechanisms could involve the proteasome (Brasse-Lagnel et al., 2011).

These findings suggest that physiological and pathological modulation of hepcidin levels in plasma could also strongly modulate metal fluxes in the body. Low levels of hepcidin, as observed during iron deficiency, genetic hemochromatosis, or liver diseases, favor the uptake of iron and, likely, non-iron metals that can be processed by DMT1 and/or ferroportin. Such mechanisms could contribute to the clinical findings reported in older (aforementioned) studies, especially those regarding cobalt (Valberg et al., 1969; Olatunbosun et al., 1970).

In addition, hepcidin expression could be regulated by noniron metals at the transcriptional level. Cadmium, copper, and zinc could modulate hepcidin expression through interactions between the metal-responsive element located in the hepcidin promoter with metal transcription factor 1 (MTF1; Balesaria et al., 2010). In addition, the authors showed that hepcidin expression was increased by cadmium, copper, and zinc salts in HuH7 hepatoma cells and decreased by cobalt and iron salts. In addition, methallothionein-1 expression followed hepcidin modulation by non-iron metals, but not by iron *per se*.

Animals developing hemochromatosis related to the *Hfe*−/<sup>−</sup> genotype in association with abnormally low hepcidin levels were recently reported to exhibit significant modulation of their digestive microbiota. Whether this observation is related to iron excess only or whether it involves a molecular defect consecutive to HFE dysfunction is not known. However, as bacteria present with differential equipment, including siderophores, for acquiring metals from the diet and different requirements regarding other metals, modulation of microbiota during iron overload could have an impact on the absorption of iron or other metals.

Finally, hepcidin has been reported to physically interact in plasma with transition metals, including copper, nickel, and zinc (Tselepis et al., 2010). Whether this observation has significant biological relevance is not currently known.

## **INTERACTIONS BETWEEN IRON AND HEPCIDIN AND THE METAL CONNECTION IN IRON-RELATED DISEASES**

Until now, one of the major problems in iron-related diseases has been understanding the variability of clinical and biological expression.

This is especially true in genetic hemochromatosis related to the p.Cys282Tyr mutation. The homozygous mutation is found in three subjects per thousand in the Caucasian population. However, an increase in serum transferrin saturation, which is the earliest biochemical expression of the disease, only occurs in 50% of homozygous subjects. The clinical manifestations that impact life expectancy or welfare are even rarer (Brissot et al., 2011). These discrepancies led to a search for putative genetic cofactors associated with the p.Cys282Tyr mutation. Some explanations have been proposed, including digenism or polymorphisms in other genes directly or indirectly involved in iron metabolism (Merryweather-Clarke et al., 2003; Island et al., 2009). In addition, an impact of hormonal status (Latour et al., 2014) and environmental factors related to metabolic syndrome (Desgrippes et al., 2013) has been found.

The identification of interconnections between non-iron metals and their potential impact on iron metabolism provides a new way to explore the differential expression of iron-related diseases, especially in a world where every individual may be exposed to various metals during personal and occupational life, as well as after prosthesis implantation.

### **ACKNOWLEDGMENTS**

This work was supported by INSERM (Institut National de la Santé et de la Recherche Médicale), University of Rennes 1 and ANR IRONREG.

## **REFERENCES**


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

*Received: 05 March 2014; accepted: 13 May 2014; published online: 04 June 2014. Citation: Loréal O, Cavey T, Bardou-Jacquet E, Guggenbuhl P, Ropert M and Brissot P (2014) Iron, hepcidin, and the metal connection. Front. Pharmacol. 5:128. doi: 10.3389/fphar.2014.00128*

*This article was submitted to Drug Metabolism and Transport, a section of the journal Frontiers in Pharmacology.*

*Copyright © 2014 Loréal, Cavey, Bardou-Jacquet, Guggenbuhl, Ropert and Brissot. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*

## The IRP/IRE system *in vivo*: insights from mouse models

## *Nicole Wilkinson and Kostas Pantopoulos\**

*Lady Davis Institute for Medical Research, Jewish General Hospital, and Department of Medicine, McGill University, Montreal, QC, Canada*

#### *Edited by:*

*Paolo Arosio, University of Brescia, Italy*

#### *Reviewed by:*

*Mayka Sanchez, Josep Carreras Leukaemia Research Institute (IJC), Spain Carole Beaumont, INSERM, France*

#### *\*Correspondence:*

*Kostas Pantopoulos, Lady Davis Institute for Medical Research, Jewish General Hospital, and Department of Medicine, 3755 Cote Ste-Catherine Rd., Montreal, QC H3T 1E2, Canada e-mail: kostas.pantopoulos@ mcgill.ca*

Iron regulatory proteins 1 and 2 (IRP1 and IRP2) post-transcriptionally control the expression of several mRNAs encoding proteins of iron, oxygen and energy metabolism. The mechanism involves their binding to iron responsive elements (IREs) in the untranslated regions of target mRNAs, thereby controlling mRNA translation or stability. Whereas IRP2 functions solely as an RNA-binding protein, IRP1 operates as either an RNA-binding protein or a cytosolic aconitase. Early experiments in cultured cells established a crucial role of IRPs in regulation of cellular iron metabolism. More recently, studies in mouse models with global or localized Irp1 and/or Irp2 deficiencies uncovered new physiological functions of IRPs in the context of systemic iron homeostasis. Thus, IRP1 emerged as a key regulator of erythropoiesis and iron absorption by controlling hypoxia inducible factor 2α (HIF2α) mRNA translation, while IRP2 appears to dominate the control of iron uptake and heme biosynthesis in erythroid progenitor cells by regulating the expression of transferrin receptor 1 (TfR1) and 5-aminolevulinic acid synthase 2 (ALAS2) mRNAs, respectively. Targeted disruption of either Irp1 or Irp2 in mice is associated with distinct phenotypic abnormalities. Thus, Irp1−*/*<sup>−</sup> mice develop polycythemia and pulmonary hypertension, while Irp2−*/*<sup>−</sup> mice present with microcytic anemia, iron overload in the intestine and the liver, and neurologic defects. Combined disruption of both Irp1 and Irp2 is incombatible with life and leads to early embryonic lethality. Mice with intestinal- or liver-specific disruption of both Irps are viable at birth but die later on due to malabsorption or liver failure, respectively. Adult mice lacking both Irps in the intestine exhibit a profound defect in dietary iron absorption due to a "mucosal block" that is caused by the de-repression of ferritin mRNA translation. Herein, we discuss the physiological function of the IRE/IRP regulatory system.

**Keywords: iron metabolism, ferritin, ferroportin, aconitase, transferrin receptor, HIF2α, DMT1, hepcidin**

## **PRINCIPLES OF MAMMALIAN IRON METABOLISM**

Iron is a vital micronutrient and constituent of several metalloproteins (Aisen et al., 2001). Even though it may acquire many oxidation states (from −2 to +6), within cells iron commonly alternates between the *ferrous* Fe(II) and *ferric* Fe(III) forms. Because of its abilities to coordinate with proteins and to act as both an electron donor and an electron acceptor, iron has been utilized throughout evolution by almost all living cells and organisms for metabolic purposes. Thus, iron is an integral component in: oxygen transport, through the heme moiety of hemoglobin; cellular respiration, as part of heme-containing cytochromes and Fe-S cluster-containing proteins of the *electron transport chain* (ETC); DNA synthesis and cellular growth, as part of the M2 subunit of ribonucleotide reductase.

As essential as iron is to survival, it can also be toxic. "Free" iron engages in Fenton chemistry to catalyze the production of hydroxyl radicals from superoxide and hydrogen peroxide (Papanikolaou and Pantopoulos, 2005). These are extremely reactive species that can damage lipid membranes, proteins and nucleic acids within a cell. To avoid the deleterious effects of the Fenton reaction, iron has to be shielded. Circulating iron is oxidized to Fe(III) and tightly binds to transferrin (Tf), which maintains it in a redox-inert state and delivers it to tissues (Gkouvatsos et al., 2012). Cellular iron uptake involves the binding of iron-loaded Tf to transferrin receptor 1 (TfR1) and the internalization of the complex by endocytosis (Aisen, 2004). Within endosomes, Fe(III) is released from Tf following acidification, and undergoes reduction to Fe(II). Subsequently, Fe(II) crosses the endosomal membrane via the divalent metal transporter 1 (DMT1). Internalized iron is mostly utilized in mitochondria for synthesis of heme and Fe-S clusters, and in the cytosol for incorporation into metalloproteins, while excess of iron is stored and detoxified within ferritin (Arosio et al., 2009). Ferritin is ubiquitously expressed in the cytosol of cells. It is comprised of 24 H- and L-polypeptide subunits that form a shell-like nanocage for iron storage, with a diameter of 7–8 nm. This can accommodate up to 4500 iron atoms, which are safely stored in the form of ferric oxyhydroxide phosphate, following oxidation of Fe(II) to Fe(III) at the ferroxidase center of the H-ferritin subunit. In addition, a distinct ferritin isoform (M-ferritin) is expressed in mitochondria of some cell types, such as testicular Leydig cells, neuronal cells and pancreatic islets of Langherans (Levi and Arosio, 2004).

The iron content of the adult human body ranges between 3 and 5 g (Gkouvatsos et al., 2012). Most of it (∼70%) is utilized in hemoglobin of red blood cells (RBCs) in the bloodstream and of erythroid progenitor cells in the bone marrow, and is recycled by tissue macrophages at a rate of 25–30 mg/day (**Figure 1**). A substantial amount of body iron (up to 1 g) is stored within ferritin in the liver. Muscles contain ∼300 mg of iron (mostly in myoglobin), and all other tissues (excluding the duodenum) merely ∼8 mg. Circulating Tf-bound iron represents a small (∼3 mg) but dynamic fraction that turns over ∼10 times per day (Cavill, 2002). The Tf iron pool is primarily replenished by iron from senescent RBCs that is recycled via macrophages, and to a smaller extent by iron absorbed from the diet via duodenal enterocytes. In adults, dietary iron absorption (1–2 mg/day) serves to compensate for non-specific iron losses (by desquamation or bleeding).

Iron metabolism is regulated at both the systemic and cellular levels. Systemic iron homeostasis is controlled through hepcidin, a liver-derived peptide hormone (Ganz, 2013). Iron and other stimuli (inflammation, endoplasmic reticulum stress) positively regulate hepcidin transcription, while erythropoietic drive suppresses it. Hepcidin binds to the cellular iron exporter ferroportin expressed on enterocytes, macrophages, and hepatocytes causing ferroportin's subsequent internalization and degradation (Nemeth et al., 2004). Ergo hepcidin functions to control iron absorption within the enterocytes and the efflux of iron into the bloodstream from macrophages and other iron exporting cells (**Figure 2**). Misregulation of hepcidin is associated with disease (hemochromatosis, when iron-regulation of hepcidin is blunted, and anemia, when hepcidin is induced by

inflammatory pathways or by genetic inactivation of its inhibitor *TMPRSS6*) (Sebastiani and Pantopoulos, 2011). Cellular iron metabolism is controlled through the IRE/IRP system (Wang and Pantopoulos, 2011; Joshi et al., 2012), which is the focus of this review.

## **COORDINATE POST-TRANSCRIPTIONAL REGULATION OF CELLULAR IRON METABOLISM VIA IRE/IRP INTERACTIONS**

The IRE/IRP regulatory system was first described in the late 1980s with the discovery of iron responsive elements (IREs) in the untranslated regions (UTRs) of the mRNAs encoding ferritin (both H- and L-subunits) and TfR1 (Hentze et al., 1987; Casey et al., 1988; Müllner and Kühn, 1988); see **Figure 3**. IREs are highly conserved hairpin structures of 25–30 nucleotides (Piccinelli and Samuelsson, 2007). They contain a stem that is stabilized by base pairing, and a loop with the sequence 5 - CAGUGH-3 (H denotes A, C or U). The stem is interrupted to an upper and lower part by an unpaired C residue or an asymmetric UGC/C bulge/loop. H- and L-ferritin mRNAs contain a single IRE in their 5 UTR, which is located relatively close to the cap structure at the 5 end. TfR1 mRNA contains five IREs in its 3 UTR.

The IREs constitute binding sites of two cytoplasmic iron regulatory proteins, IRP1 and IRP2 (Rouault, 2006). Under conditions of iron deficiency, IRP1 and IRP2 bind with high affinity to the IRE in H- and L-ferritin mRNAs, and thereby inhibit their translation by a steric hindrance mechanism. Likewise, they bind to the IREs in TfR1 mRNA and thereby protect it against endonucleolytic degradation. This homeostatic response mediates increased cellular iron uptake from Tf and prevents storage of the metal. By contrast, in iron-replete cells the IRE-binding activities of IRP1 and IRP2 are diminished, allowing TfR1 mRNA degradation and ferritin mRNA translation. This inhibits further iron uptake and stimulates storage of excessive intracellular iron within ferritin. A scheme with the coordinate post-transcriptional regulation of ferritin and TfR1 by IRE/IRP interactions is shown in **Figure 4**.

Additional IRE-containing mRNAs (**Figure 5**) have been discovered by computational and biochemical approaches, as well as high throughput screens (Dandekar and Hentze, 1995; Gunshin et al., 1997; McKie et al., 2000; Sanchez et al., 2006, 2007, 2011). These include the mRNAs encoding the iron transporters DMT1 and ferroportin, the enzyme 5-aminolevulinic acid synthase 2 (ALAS2) that catalyzes the first reaction for heme biosynthesis in erythroid progenitor cells, the enzyme of the citric acid cycle mitochondrial aconitase, the cell cycle regulator CDC14A, and hypoxia inducible factor 2 alpha (HIF2α), a transcription factor that orchestrates molecular responses to hypoxia. All these transcripts harbor a single IRE. The IRE of ferroportin, ALAS2, mitochondrial aconitase and HIF2α mRNAs is localized in the 5 UTR and functions as a translational control element, analogous to ferritin IRE. The IRE of DMT1 and CDC14A mRNAs is localized in the 3 UTR and appears to function as an mRNA stability element. Post-transcriptional regulation of DMT1 expression by the IRE/IRP system is cell-specific (Gunshin et al., 2001) and depends, at least partially, on an alternatively transcribed upstream exon at the 5 end of DMT1 mRNA

(Hubert and Hentze, 2002). Notably, earlier experiments showed that the presence of at least three 3 UTR IREs is essential for iron-dependent regulation of TfR1 mRNA (Casey et al., 1989), which remains the only transcript containing multiple IREs. It should also be noted that DMT1, ferroportin and CDC14A mRNAs include additional non-IRE-containing isoforms, which are generated by alternative splicing and exhibit differential tissue distribution (Hubert and Hentze, 2002; Sanchez et al., 2006; Zhang et al., 2009).

Another level of complexity is added by the fact that iron regulation of HIF2α, ferroportin and DMT1 integrates multiple and often opposing signals (Mastrogiannaki et al., 2013; Shah and Xie, 2014). Thus, while iron stimulates HIF2α and ferroportin mRNA translation, it also promotes the degradation of HIF2α via iron- and oxygen-dependent prolyl hydroxylases (PHDs) and the von Hippel Lindau E3 ubiquitin ligase (pVHL) (Schofield and Ratcliffe, 2004), and the degradation of ferroportin via hepcidin (Ganz, 2013). Conversely, iron deficiency inhibits translation of HIF2α and ferroportin mRNAs but promotes stabilization of the proteins. On the other hand, iron deficiency leads to HIF2α-mediated transcriptional activation of duodenal DMT1 and ferroportin, the apical and basolateral transporters of iron in enterocytes (see below).

## **SENSING OF INTRACELLULAR IRON BY IRP1 AND IRP2**

IRP1 and IRP2 are homologous to mitochondrial aconitase (Rouault, 2006; Wang and Pantopoulos, 2011). A characteristic feature of this enzyme is the presence of a 4Fe-4S cluster within its active site, which is indispensable for catalytic isomerization of citrate to iso-citrate. IRP1 assembles an aconitase-type 4Fe-4S cluster in response to increased intracellular iron levels. The complex pathway involves several co-factors, such as Nfs1 (ISCS), frataxin, ISCU, glutaredoxin 5 (*GLRX5*) and others. Assembly of the 4Fe-4S cluster alters the conformation of IRP1 and precludes IRE-binding (Dupuy et al., 2006; Walden et al., 2006). At the same time holo-IRP1 (containing the 4Fe-4S cluster) acquires enzymatic activity as cytosolic aconitase, which

is comparable to that of its mitochondrial counterpart. Iron starvation, as well as other signals, such as nitric oxide (NO) or hydrogen peroxide (H2O2), promotes the loss of the 4Fe-4S cluster and conversion of apo-IRP1 to an IRE-binding protein (Haile et al., 1992; Drapier et al., 1993; Weiss et al., 1993; Pantopoulos and Hentze, 1995). Thus, IRP1 is a bifunctional protein that is regulated by an unusual 4Fe-4S cluster switch. Holo-IRP1 is stabilized by hypoxia and at oxygen concentrations ranging between 3 and 6%, which reflect physiological tissue oxygenation (Meyron-Holtz et al., 2004b). Consistently with this notion, IRP1 predominates in the cytosolic aconitase form within tissues, but a fraction of apo-IRP1 is readily detectable by virtue of its IRE-binding activity (Meyron-Holtz et al., 2004a).

Contrary to IRP1, IRP2 does not bind an iron-sulfur cluster and is regulated at the level of protein stability. IRP2 accumulates in iron-starved and/or hypoxic cells, and undergoes proteasomal degradation in iron-replete oxygenated cells. Mechanistically, this involves ubiquitination of IRP2 by an E3 ubiquitin ligase complex consisting of the F-box protein FBXL5 and the auxiliary proteins SKP1, CUL1, and RBX1 (Salahudeen et al., 2009; Vashisht et al., 2009). FBXL5 exhibits genuine iron-sensing properties, as it is stabilized in the presence of iron and oxygen by forming a Fe-O-Fe center within its N-terminal hemerythrinlike domain. Loss of this center in iron-starved and/or hypoxic cells exposes a degron via a conformational change (Chollangi et al., 2012; Thompson et al., 2012). This allows proteasomal degradation of FBXL5, which leads to concomitant accumulation of IRP2. Interestingly, FBXL5 can also promote the ubiquitination and degradation of IRP1 mutants that cannot form a 4Fe-4S cluster (Salahudeen et al., 2009; Vashisht et al., 2009). Proteasomal degradation of wild type apo-IRP1 under conditions of impairment of the iron-sulfur cluster assembly pathway has been proposed to operate as a reserve mechanism to control the IRE-binding activity of IRP1 (Clarke et al., 2006; Wang et al., 2007).

## **HUMAN DISEASES LINKED TO THE IRE/IRP SYSTEM**

Mutations in the IRE of L-ferritin mRNA that abrogate IRP binding are causatively linked to the hereditary hyperferritinemia-cataract syndrome (HHCS) (Beaumont et al., 1995). The hallmark of this disorder is overexpression of serum ferritin (up to 20-fold) in the absence of systemic iron overload or inflammation (Yin et al., 2014). Patients also have an increased tendency to develop bilateral cataract, which is possibly caused by accumulation of non-functional L-ferritin homopolymers in the lens (Levi et al., 1998). HHCS is transmitted in an autosomal dominant manner. Several HHCS-associated mutations in L-ferritin IRE have been described (Luscieti et al., 2013). Mutations affecting the loop structure or the C bulge result in the highest levels of serum ferritin and increased severity of cataract when compared with mutations in the stem structure of the IRE (Cazzola et al., 1997). Overall, the severity of the clinical phenotype appears to correlate with the degree of inhibition in IRP binding (Allerson et al., 1999). Nevertheless, the involvement of additional factors (genetic, environmental, inflammatory) has not been excluded (Roetto et al., 2002).

A point mutation (A49U) in the loop of H-ferritin IRE has been associated with an autosomal dominant iron overload disorder in a Japanese family (Kato et al., 2001). Iron deposits were documented primarily in hepatocytes, but also in some Kupffer cells. The A49U mutation increased the affinity of IRPs for Hferritin IRE and suppressed H-ferritin expression in cultured cells (Kato et al., 2001). Nevertheless, it remains unclear how this response promotes iron overload. The lack of follow-up studies to support the validity of these findings should also be noted.

Disruption of the zebrafish *glrx5* gene, encoding glutaredoxin 5, leads to constitutive activation of IRP1 for IRE-binding, due


to defective assembly of its 4Fe-4S cluster (Wingert et al., 2005). In a case report, a patient with *GLRX5* deficiency developed a sideroblastic-like anemia with microcytosis and systemic iron overload (Camaschella et al., 2007). This disease is caused by IRP1-mediated suppression of ALAS2 mRNA translation, which results in impaired heme biosynthesis. A secondary effect is the depletion of cytosolic iron and the development of mitochondrial iron overload (Ye et al., 2010).

The above examples represent the only documented human disorders that are causatively linked to defects in the IRE/IRP system (**Table 1**). Genome-wide association studies (GWAS) showed that the IRP2 gene (*IREB2*) confers susceptibility to chronic obstructive pulmonary disease (COPD) (Demeo et al., 2009; Chappell et al., 2011; Zhou et al., 2012) and lies within a lung cancer-associated locus (Hansen et al., 2010; Cho et al., 2012; Fehringer et al., 2012). Apart from the identification of single nucleotide polymorphisms (SNPs) in *IREB2*, it was shown that COPD patients exhibit increased IRP2 mRNA and protein expression (Demeo et al., 2009) and it was speculated that this may contribute to iron accumulation in the lungs. *IREB2* polymorphisms have also been associated with Alzheimer's disease (Coon et al., 2006). Interestingly, biochemical studies showed that IRP1 selectively regulates translation of amyloid precursor protein (APP) mRNA by binding to an atypical putative IRE motif (Cho et al., 2010), which may provide another connection of the IRE/IRP system with Alzheimer's disease. The IRP1 gene (*ACO1*) has been associated with cutaneous malignant melanoma (Yang et al., 2010) and neuropathic pain in HIV-infected patients (Kallianpur et al., 2014), while the IRE-binding activity of IRP1 was reported to be increased in Friedreichs' ataxia (Lobmayr et al., 2005) and in Parkinson's disease (Faucheux et al., 2002). Finally, polymorphisms in both *ACO1* and *IREB2* have been linked to agerelated macular degeneration (Synowiec et al., 2012). GWAS data with *ACO1* and *IREB2* are summarized in **Table 2**. The molecular mechanisms by which IRPs may contribute to pathogenesis of the above disorders remain to be established.

## **MOUSE MODELS WITH UBIQUITOUS ABERRATION IN THE IRE/IRP SYSTEM**

Mice with ubiquitous ablation of both Irp1 and Irp2 cannot be generated because the embryos do not survive the blastocyst stage, possibly due to misregulation of iron metabolism (Smith et al., 2006). These data highlight the physiological significance of the IRE/IRP system in early development. On the other hand, single Irp1−*/*<sup>−</sup> or Irp2−*/*<sup>−</sup> mice are viable. At first glance, this suggests that Irp1 and Irp2 share overlapping functions *in vivo*. Nonetheless, Irp1−*/*<sup>−</sup> and Irp2−*/*<sup>−</sup> mice manifest distinct phenotypes, indicating variable target specificity. The phenotypic features of current mouse models with global or tissue-specific "loss" or "gain" of Irp1 and/or Irp2 functions are summarized in **Table 3**.

Irp1−*/*<sup>−</sup> mice were initially reported to lack any overt abnormalities (Meyron-Holtz et al., 2004a; Galy et al., 2005a). Because they merely misregulated ferritin and TfR1 expression in the kidney and the brown fat, tissues with the highest abundance of IRP1, it was proposed that Irp2 dominates cellular iron metabolism in tissues, with Irp1 having an auxiliary physiological function (Meyron-Holtz et al., 2004a). In addition, Irp1−*/*<sup>−</sup> (and Irp2−*/*−) mice did not exhibit any defects when challenged with turpentine to induce an inflammatory response (Viatte et al., 2009), despite the fact that IRPs are modulated in cultured cells by inflammatory reactive species such as H2O2 and NO (Weiss et al., 1993; Pantopoulos and Hentze, 1995; Wang et al., 2005; Hausmann et al., 2011).

Recently, Irp1−*/*<sup>−</sup> mice were documented to develop polycythemia and pathological iron metabolism due to stress erythropoiesis, as well as pulmonary hypertension and cardiac hypertrophy and fibrosis (Anderson et al., 2013; Ghosh et al., 2013; Wilkinson and Pantopoulos, 2013). These phenotypes are attributed to relief of translational suppression of Hif2α mRNA, which leads to transcriptional activation of the downstream Hif2α targets erythropoietin and endothelin-1. Thus, murine Irp1 operates as specific regulator of the Hif2α IRE, in agreement with previous *in vitro* data (Zimmer et al., 2008). Irp1−*/*<sup>−</sup> mice exhibit

#### **Table 1 | Human disorders that are causatively linked to defects in the IRE/IRP system.**


#### **Table 2 | Genome-wide association studies (GWAS) with the IRPI and IRP2 genes,** *ACO1* **and** *IREB2***, respectively.**


high mortality when placed on an iron-deficient diet, which is known to stabilize Hif2α, due to abdominal hemorrhages (Ghosh et al., 2013). When Irp1−*/*<sup>−</sup> mice are fed with a standard diet, polycythemia attenuates after the 10th week of age (Wilkinson and Pantopoulos, 2013), possibly due to enhanced Hif2α degradation by the pVHL pathway. This may explain why their pathology escaped attention in the past.

Irp2−*/*<sup>−</sup> mice develop microcytic hypochromic anemia and erythropoietic protoporphyria, associated with relatively mild duodenal and hepatic iron overload and splenic iron deficiency (Cooperman et al., 2005; Galy et al., 2005b). They exhibit a low Tfr1 content and express high levels of protoporphyrin IX in erythroid precursor cells. A closer look on the role of Irp1 and Irp2 in erythropoiesis, duodenal iron absorption and systemic iron metabolism will be provided in the next sections.

Global Irp2−*/*<sup>−</sup> deficiency has also been associated with progressive neurodegeneration, loss of Purkinje neurons and iron overload in white matter areas of the brain (LaVaute et al., 2001; Ghosh et al., 2006), with more severe presentation in Irp1 haploinsufficient Irp1+*/*−Irp2−*/*<sup>−</sup> mice (Smith et al., 2004). The neuronal pathology of Irp2−*/*<sup>−</sup> mice can be partially rescued by pharmacological activation of endogenous Irp1 for IRE-binding (Ghosh et al., 2008). Mechanistically, the pathology may be caused by functional iron deficiency in neuronal cells due to derepression of ferritin mRNA translation, which could result in enhanced iron storage and reduced availability for metabolic purposes (Jeong et al., 2011). This scenario is reminiscent of neuroferritinopathy, a neurodegenerative disease caused by expression of mutant L-ferritin (Burn and Chinnery, 2006). Suppression of Tfr1 and enhancement of ferroportin expression in Irp2−*/*<sup>−</sup> neurons could also contribute to pathology. Nevertheless, isogenic Irp2−*/*<sup>−</sup> mice generated by another strategy did not develop any signs of neurodegeneration but rather manifested minor performance deficits in specific neurological tests (motor coordination and balance) (Galy et al., 2006). A third independent Irp2−*/*<sup>−</sup> mouse strain was recently generated and analyzed, corroborating the abnormalities in erythropoiesis and brain iron metabolism of these animals; the latter were linked to mild neurological and


#### **Table 3 | Phenotypic features of mouse models with global or tissue-specific "loss" or "gain" of Irp1 and/or Irp2 functions.**

*(Continued)*

#### **Table 3 | Continued**


behavioral defects, as well as nociception (Zumbrennen-Bullough et al., 2014). The discrepancies in the phenotypes of the three Irp2−*/*<sup>−</sup> mouse lines may be related to the targeting strategies and their impact on flanking genomic sequences, genetic factors, and possibly also environmental factors.

Ubiquitous Irp1 "gain of function" mice were generated by the transgenic expression of a constitutively active IRP1 mutant from the Rosa26 locus (Casarrubea et al., 2013). Expression of the transgene was relatively low and the mice developed erythropoietic abnormalities (macrocytic erythropenia due to impaired erythroid differentiation).

The targeted disruption of Fbxl5 could yield ubiquitous Irp2 "gain of function" mice. Nevertheless, Fbxl5−*/*<sup>−</sup> mice are not viable and die during embryogenesis (Moroishi et al., 2011; Ruiz et al., 2013). This phenotype is rescued by concomitant disruption of Irp2, but not Irp1, indicating that the major physiological function of Fbxl5 is to regulate Irp2.

## **THE IRE/IRP SYSTEM IN ERYTHROPOIESIS**

Most iron in the body is utilized in erythropoiesis, a process that occurs mainly in the bone marrow of humans and in the bone marrow and spleen of mice (Cavill, 2002). Erythroid precursor cells express high levels of TfR1 and consume the majority of Tf-bound iron for the production of heme, the oxygen-binding moiety of hemoglobin (Ponka, 1997). An adult human produces about 2 million RBCs per second. Increased erythropoietic drive is known to stimulate iron absorption (Finch, 1994). This is mediated by suppression of hepcidin expression, which allows increased iron efflux to the bloodstream from duodenal enterocytes and tissue macrophages (Ganz, 2013). During iron deficiency, erythroid precursor cells will have priority for iron utilization, for the production of RBCs, over cells in other tissues (Finch, 1994). Maturation of erythroid cell is induced by erythropoeitin (EPO), a circulating glycoprotein hormone that prevents apoptosis of erythroid progenitor cells (Koury and Bondurant, 1990). EPO is primarily synthesized in peritubular interstitial fibroblasts of the kidney, and to a smaller extent in hepatocytes of the liver. Murine Epo is transcriptionally activated during hypoxemia by Hif2α (Rankin et al., 2007; Kapitsinou et al., 2010), while Hif2α indirectly suppresses hepcidin expression by stimulating erythropoiesis via Epo (Liu et al., 2012; Mastrogiannaki et al., 2012). There are several points where the IRE/IRP system regulates erythropoiesis.

First, the phenotype of Irp1−*/*<sup>−</sup> mice demonstrates that Irp1 operates as a key upstream regulator of Hif2α expression at the level of mRNA translation (Anderson et al., 2013; Ghosh et al., 2013; Wilkinson and Pantopoulos, 2013). Essentially, these animals exhibit features of Hif2α overexpression, which are also manifested in patients with Chuvash polycythemia (Ang et al., 2002), and other forms of familial polycythemia caused by "gain of function" HIF2α mutations (Percy et al., 2008), or inactivating mutations in pVHL, a negative regulator of HIF2α stability (Percy et al., 2006). One of them, is hyperproduction of Epo, which causes splenomegaly due to extramedullary erythropoiesis, promotes expansion of late stage basophilic erythroblasts, as well as polychromatic and orthochromatic erythroblasts, and finally leads to reticulocytosis and polycythemia. Importantly, overexpression of hepatic Epo is rescued by intercrossing Irp1−*/*<sup>−</sup> and liver-specific Hif2α−*/*−mice (Wilkinson and Pantopoulos, 2013). On the other hand, the defects in erythroid differentiation documented in the constitutive IRP1 transgenic mice can be attributed to reduced Hif2α expression (Casarrubea et al., 2013). Hence, the translational regulation of HIF2α mRNA by IRP1 links iron metabolism with erythropoiesis via EPO. Physiologically, this response probably serves to contain EPO expression and subsequent stimulation of erythropoiesis under conditions of iron deficiency. In addition, it is tempting to speculate that stress activation of IRP1 (by H2O2 or NO) (Weiss et al., 1993; Pantopoulos and Hentze, 1995) may contribute to the impairment of EPO production that is frequently observed under chronic inflammatory conditions or in chronic kidney disease (Weiss and Goodnough, 2005), via suppression of HIF2α mRNA translation.

The above data suggest that IRP1 functions as an iron and oxygen sensor (**Figure 6**). According to this model, in normoxic renal interstitial fibroblasts the small steady-state fraction of apo-IRP1 inhibits HIF2α mRNA translation and thereby limits EPO production to physiological levels. This fraction is expected to increase in normoxic iron deficiency due to 4Fe-4S cluster disassembly, further suppressing HIF2α and EPO expression as a homeostatic response to reduced iron availability, consistently with iron-restricted erythropoiesis. Conversely, increased iron supply is expected to increase the abundance of holo-IRP1, allowing de-repression of HIF2α mRNA translation. This response is also favored by hypoxia, which is known to stabilize holo-IRP1. Thus, in hypoxic and/or iron-replete renal interstitial fibroblasts, unimpeded HIF2α mRNA translation leads to increased generation of EPO and thereby stimulates erythropoiesis, as a homeostatic adaptation to the scarcity of oxygen and/or the abundance of iron.

IRP2 is not involved in the regulation of HIF2α mRNA *in vivo*. This makes physiological sense, since IRP2 is highly active under hypoxia, when HIF2α synthesis is required. Nevertheless, the phenotype of Irp2−*/*<sup>−</sup> mice suggests that IRP2 is a critical regulator of other erythropoietic pathways. These animals present with hypochromic microcytic anemia with reduced iron availability for erythropoiesis (Cooperman et al., 2005; Galy et al., 2005b; Zumbrennen-Bullough et al., 2014). This is not a result of global iron deficiency but rather misregulation of cellular iron traffic. Irp2−*/*<sup>−</sup> erythroid cells exhibit a reduced Tfr1 content, apparently due to lack of Tfr1 mRNA stabilization in the absence of Irp2. As a result, these cells fail to acquire sufficient amounts of iron for erythropoiesis, in spite of the physiological iron supply, which is reflected in the normal Tf saturation and total ironbinding capacity (TIBC) in serum. Therefore, Irp2 appears to act as major activator of Tfr1 in erythroid cells. It is conceivable that this function is more pronounced at earlier stages of erythroid development, considering that during terminal erythroid differentiation, Tfr1 mRNA stability remains unresponsive to iron, bypassing the IRE/IRP checkpoint (Schranzhofer et al., 2006).

Irp2−*/*<sup>−</sup> mice also manifest very high levels of free and zinc protoporphyrin IX in RBCs (∼200-fold increase for free protoporphyrin IX), consistently with erythropoietic protoporphyria (Cooperman et al., 2005). This is very likely caused by enhanced heme biosynthetic activity due to de-repression of Alas2 mRNA translation. In the absence of adequate iron, it results in accumulation of free protoporphyrin IX and in incorporation of zinc in the protoporphyrin IX ring. Therefore, IRP2 also acts as a major regulator of ALAS2 mRNA translation via the IRE in its 5 UTR. The interpretation that IRP2 regulates erythroid iron uptake and utilization by controlling expression of the IRE-containing TfR1 and ALAS2 mRNAs, respectively, is also supported by data with tissue-specific Irp2−*/*<sup>−</sup> mice. Thus, liveror intestinal-specific ablation of Irp2 recapitulates the iron overload phenotype of these tissues in ubiquitous Irp2−*/*<sup>−</sup> mice, but fails to promote microcytic anemia (Ferring-Appel et al., 2009).

## **THE IRE/IRP SYSTEM IN DIETARY IRON ABSORPTION**

Dietary iron absorption takes place in brush border enterocytes of the duodenum, especially in the upper tract. The pathway is highly regulated at different levels. Hormonal regulation via hepcidin is well established. Hepcidin is produced in the liver in response to high serum or hepatic iron (Corradini et al., 2011; Ramos et al., 2011) and limits further iron absorption by promoting the degradation of duodenal ferroportin (Nemeth et al., 2004; Chung et al., 2009), the basolateral iron transporter. Hepcidin may also promote degradation of DMT1, the apical iron transporter, by the proteasome (Chung et al., 2009; Brasse-Lagnel et al., 2011). Both ferroportin (Taylor et al., 2011) and DMT1 (Mastrogiannaki et al., 2009; Shah et al., 2009) are transcriptionally activated in iron-deficient enterocytes by HIF2α. In addition, ferroportin and DMT1 could be directly regulated by IRPs at the levels of mRNA translation or stability, respectively. Duodenal enterocytes also express ferritin, which limits excessive iron transfer to the circulation (Vanoaica et al., 2010). TfR1 expression is restricted to non-absorptive enterocytes in the intestinal crypts (Waheed et al., 1999).

Iron deficiency elicits multiple and antagonistic signals on duodenal ferroportin: transcriptional activation of its gene via HIF2α (which is negatively regulated by IRP1), translational repression of its mRNA via IRPs, and stabilization of the protein via downregulation of hepcidin. The net result is induction of ferroportin, which is considered a physiologic response to increase iron supply to the bloodstream (McKie et al., 2000). It has been proposed that the IRP blockade can be bypassed via enriched expression of an alternatively spliced non-IRE containing ferroportin transcript (Zhang et al., 2009). Nevertheless, in other reports the canonical IRE-containing ferroportin mRNA was found to be predominant in the duodenum of rats and mice (Darshan et al., 2011; Galy et al., 2013). Another possibility is that protein stabilization due to suppression of hepcidin is dominant.

Likewise, iron deficiency elicits multiple and antagonistic signals on duodenal DMT1: transcriptional activation of its gene via HIF2α (which is negatively regulated by IRP1), possible stabilization of its IRE-containing transcripts by IRPs, and possible stabilization of the protein via downregulation of hepcidin. Again, the net result is induction of DMT1 (Gunshin et al., 1997), which serves to increase dietary iron absorption from the intestinal lumen. Four different DMT1 mRNA isoforms are generated

by the combination of alternative promoter usage and alternative splicing, and two of them harbor an IRE (Hubert and Hentze, 2002). Interestingly, the IRE-containing DMT1 mRNAs predominate in the duodenum and their expression is induced in iron deficiency (Canonne-Hergaux et al., 1999; Hubert and Hentze, 2002). Their transcription is selectively activated by HIF2α (Mastrogiannaki et al., 2009; Shah et al., 2009).

Experiments with mice bearing global or enterocyte-specific Irp deficiencies provided insights on the role of the IRE/IRP system in dietary iron absorption. Irp1−*/*<sup>−</sup> mice express high levels of duodenal ferroportin and IRE-Dmt1 mRNAs, most likely as a result of de-repression of Hif2α mRNA translation and enhanced Hif2α transcriptional activity in this tissue (Anderson et al., 2013). This interpretation is consistent with the high expression of further Hif2α target genes. Ferroportin and Dmt1 expression was not altered in duodena of Irp2−*/*<sup>−</sup> mice, which exhibit iron overload and express high levels of ferritin (Galy et al., 2005b). High levels of duodenal ferritin were also documented in mice with enterocyte-specific disruption of Irp2 (Ferring-Appel et al., 2009). On the other hand, duodenal ferritin was not suppressed in mice expressing the constitutive IRP1 transgene (Casarrubea et al., 2013), in agreement with previous cell culture experiments (Wang and Pantopoulos, 2002), possibly due to alternative ferritin mRNA translation via internal initiation (Daba et al., 2012). Thus, IRP2 alone appears to function as an important regulator of ferritin but not ferroportin and DMT1 in the duodenum.

Mouse pups with enterocyte-specific deletion of both Irp1 and Irp2 are viable at birth but die within 4 weeks due to malabsorption and dehydration, associated with abnormalities in intestinal architecture (Galy et al., 2008). Their enterocytes manifested highly increased expression of ferritin and ferroportin, and reduced levels of Tfr1 and Dmt1. The downregulation of ferritin and ferroportin was not associated with alterations in their mRNA levels, reinforcing the role of IRPs in the regulation of duodenal ferritin synthesis, but also demonstrating a prominent function of the IRE/IRP system in translational control of duodenal ferroportin mRNA. Expression of the major IRE-Dmt1 mRNA isoform was slightly reduced in the mutant mice, providing first *in vivo* evidence for a role of IRPs in the control of IRE-DMT1 mRNA stability.

Adult mice with enterocyte-specific ablation of both Irp1 and Irp2 were generated by using a tamoxifen-inducible Cre-deleter strain under the control of the villin promoter (Galy et al., 2013). These animals are viable and the ablation of Irps does not significantly alter intestinal architecture. As expected, the expression of ferritin and ferroportin was very high, while Tfr1 levels were low. Dmt1 was also upregulated, in contrast to the data obtained in pups with enterocyte-specific deletion of both Irps (Galy et al., 2008). The induction of Dmt1 was associated with an increase in steady-state levels of the IRE-Dmt1 mRNA isoforms, which was attributed to transcriptional stimulation by Hif2α. Together, the above data suggest that Irps regulate the IRE-Dmt1 mRNA only in newborn mice but not in adult animals. This is consistent with the notion that duodenal iron absorption in suckling pups is enhanced to satisfy metabolic needs of the rapidly growing organism, and is not subjected to negative regulation by hepcidin (Darshan et al., 2011). Stabilization of the IRE-Dmt1 mRNA isoforms by Irps could contribute to the increased iron absorption capacity before weaning. Surprisingly, adult mice with enterocyte-specific ablation of both Irps exhibited impaired iron absorption, in spite of the profound induction of the apical and basolateral iron transporters. This was attributed to a "mucosal block" imposed by the overexpression of ferritin, which stores internalized iron and prevents its delivery to the bloodstream. Therefore, alleviation of the "mucosal block" by limiting the expression of ferritin emerges as a crucial function of IRPs, possibly with a major contribution of IRP2, in enterocytes (**Figure 7**).

## **THE IRE/IRP SYSTEM IN SYSTEMIC IRON METABOLISM**

Systemic iron metabolism is regulated by hepcidin, a peptide hormone synthesized by liver hepatocytes, which targets ferroportin in duodenal enterocytes, tissue macrophages and other ironexporting cells (Ganz, 2013). The IRE/IRP system appears to intersect with the hepcidin ferroportin axis. Young Irp1−*/*<sup>−</sup> mice exhibit a marked suppression of hepcidin mRNA, accompanied by accumulation of ferroportin in splenic macrophages (Wilkinson and Pantopoulos, 2013). This response is caused by the increased erythropoietic activity of these animals due to induction of Hif2α and Epo, which is an established and potent inhibitor of hepcidin expression (Ganz, 2013). Irp2−*/*<sup>−</sup> mice were found to have physiological hepcidin mRNA levels, in spite of hepatic iron overload (Cooperman et al., 2005; Galy et al., 2005b). However, at the age of 4–6 weeks, these animals showed a significant induction of hepcidin mRNA expression (Wilkinson and Pantopoulos, 2013). Conceivably, these seemingly discrepant findings are related to the antagonistic signals of erythropoietic drive and iron overload, which act as negative or positive

**synthesis in duodenal enterocytes.** Regulated expression of ferritin is essential to prevent a "mucosal block" following iron intake from the intestinal lumen. In addition, IRPs control the expression of ferroportin and DMT1 mRNAs, the latter only in the period after birth and before weaning. transcriptionally induces, among other targets, the iron transporters ferroportin and DMT1, and the ferrireductase Dcytb. Ferroportin and possibly also DMT1 are negatively regulated by hepcidin at the level of protein stability.

regulators of hepcidin, respectively (Ganz, 2013). These stimuli do not operate in a strictly hierarchical manner, and their dominance depends on signal strength (Huang et al., 2009). Hepatic iron overload may be dominant in young animals, and later neutralized by the erythropoietic drive. Nevertheless, 8–10-week-old liver-specific Irp2−*/*<sup>−</sup> mice, which develop hepatic iron overload without microcytic anemia, exhibit physiological hepcidin expression (Ferring-Appel et al., 2009). This may indicate that Irp2 expression is required for iron-dependent hepcidin induction in mice older than 4–6 weeks, but the underlying mechanism is unclear.

The liver is a major site for excessive iron storage within ferritin, but also the central regulator of systemic iron balance via hepcidin (Meynard et al., 2014). The hepatic iron overload phenotype that was observed in mice with global Irp2 disruption persisted in liver-specific Irp2−*/*<sup>−</sup> counterparts, which is indicative of a cell autonomous function (Ferring-Appel et al., 2009). Liverspecific disruption of both Irp1 and Irp2 resulted in early lethality within 1–2 weeks after birth due to liver failure (Galy et al., 2010). This was associated with mitochondrial iron deficiency, as well as histological and functional defects in this organelle. Thus, the IRE/IRP system is essential for liver function. Along similar lines, unregulated overexpression of endogenous Irp2 in liver-specific Fbxl5−*/*<sup>−</sup> mice impaired hepatic and systemic iron homeostasis (Moroishi et al., 2011). The animals developed hepatic iron overload and steatohepatitis, and exhibited inappropriately low hepcidin mRNA expression. Moreover, they succumbed after feeding with an iron-enriched diet due to lethal liver failure.

### **THE IRE/IRP SYSTEM IN CANCER**

Considering that HIF2α may function either as a tumor promoter or suppressor (Keith et al., 2012), the regulation of HIF2α mRNA translation by IRP1 provides a link between the IRE/IRP system and cancer biology. Other links were previously provided by tumor xenograft experiments. Thus, overexpression of IRP1 or a constitutive IRP1 mutant in human H1299 lung cancer cells impaired tumor xenograft growth in nude mice (Chen et al., 2007). Contrary, the overexpression of IRP2 elicited the opposite phenotype, which required the presence of an IRP2-specific 73 amino acids domain (Maffettone et al., 2010). IRE-containing mRNAs were not differentially expressed in IRP1 or IRP2-overexpressing xenografts, which exhibited distinct gene expression profiles (Maffettone et al., 2010). These data raise the possibility for a role of IRPs in modulating cancer growth independently of their IRE-binding activities, which remains to be further explored. Nevertheless, IRP2 was recently shown to be overexpressed in breast cancer cells and to promote tumor growth by modulating iron metabolism (Wang et al., 2014). This finding is consistent with reprogramming of iron metabolism in cancer cells (Torti and Torti, 2013). Notably, in earlier experiments IRP2 was reported to be transcriptionally induced by the proto-oncogene c-myc and to promote cell transformation by suppressing ferritin and by regulating cellular iron availability (Wu et al., 1999). On the other hand, the tumor suppressor p53 upregulated ferritin by reducing IRP activities (Zhang et al., 2008). Development of cancer models with Irp1−*/*<sup>−</sup> and Irp2−*/*<sup>−</sup> mice is expected to better define the roles of IRPs in cancer and to elucidate the molecular basis underlying the association of IRP2 (*IREB2*) genomic locus with susceptibility to lung cancer (Hansen et al., 2010; Cho et al., 2012; Fehringer et al., 2012).

## **CONCLUSIONS**

The discovery of the IRE/IRP regulatory system provided a framework to understand the coordinate regulation of cellular iron uptake via TfR1, and storage within ferritin. Misregulation of Lferritin expression due to "loss of function" mutations in the IRE of its mRNA is clinically relevant and underlies the molecular basis of the hyperferritinemia-caratact syndrome. The expansion of IRE-containing mRNAs raised the possibility that IRPs may control further biochemical pathways. This was firmly established by the analysis of mouse models with global or tissue-specific Irp1 and/or Irp2 deficiency. The animal studies highlighted a key role of the IRE/IRP system in regulation of erythropoiesis, dietary iron absorption, hepatic iron metabolism and body iron homeostasis via crosstalk with the hepcidin/ferroportin axis. These findings may be highly relevant to human medical conditions.

## **ACKNOWLEDGMENT**

This work was supported by a grant from the Canadian Institutes for Health Research (CIHR; MOP-86514).

## **REFERENCES**


protein (IRP)-1 function. *J. Mol. Med.* 91, 871–881. doi: 10.1007/s00109-013- 1008-2


binding protein: disassembly of the cubane iron-sulfur cluster results in highaffinity RNA binding. *Proc. Natl. Acad. Sci. U.S.A.* 89, 11735–11739. doi: 10.1073/pnas.89.24.11735


element in the 3 -untranslated region of human cell division cycle 14A mRNA by a refined microarray-based screening strategy. *J. Biol. Chem.* 281, 22865–22874. doi: 10.1074/jbc.M603876200


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

*Received: 05 May 2014; accepted: 07 July 2014; published online: 28 July 2014. Citation: Wilkinson N and Pantopoulos K (2014) The IRP/IRE system in vivo: insights from mouse models. Front. Pharmacol. 5:176. doi: 10.3389/fphar.2014.00176 This article was submitted to Drug Metabolism and Transport, a section of the journal Frontiers in Pharmacology.*

*Copyright © 2014 Wilkinson and Pantopoulos. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*

## The role of hepatic transferrin receptor 2 in the regulation of iron homeostasis in the body

## *Christal A.Worthen and Caroline A. Enns\**

Department of Cell and Developmental Biology, Oregon Health and Science University, Portland, OR, USA

#### *Edited by:*

Paolo Arosio, University of Brescia, Italy

#### *Reviewed by:*

Peng Hsiao, Seattle Genetics Inc, USA Laura Silvestri, Vita-Salute San Raffaele University, Italy

#### *\*Correspondence:*

Caroline A. Enns, Department of Cell and Developmental Biology, Oregon Health and Science University, 3181 SW Sam Jackson Park Road, L215 Portland, OR 97239, USA e-mail: ennsca@ohsu.edu

Fine-tuning of body iron is required to prevent diseases such as iron-overload and anemia. The putative iron sensor, transferrin receptor 2 (TfR2), is expressed in the liver and mutations in this protein result in the iron-overload disease Type III hereditary hemochromatosis (HH).With the loss of functionalTfR2, the liver produces about 2-fold less of the peptide hormone hepcidin, which is responsible for negatively regulating iron uptake from the diet. This reduction in hepcidin expression leads to the slow accumulation of iron in the liver, heart, joints, and pancreas and subsequent cirrhosis, heart disease, arthritis, and diabetes.TfR2 can bind iron-loaded transferrin (Tf) in the bloodstream, and hepatocytes treated with Tf respond with a 2-fold increase in hepcidin expression through stimulation of the bone morphogenetic protein (BMP)-signaling pathway. Loss of functional TfR2 or its binding partner, the original HH protein, results in a loss of this transferrin-sensitivity. While much is known about the trafficking and regulation of TfR2, the mechanism of its transferrin-sensitivity through the BMP-signaling pathway is still not known.

**Keywords: transferrin receptor 2, hereditary hemochromatosis, iron homeostasis, hepcidin, liver**

## **INTRODUCTION**

Iron is a necessary element for organisms, playing a role in vital processes such as the electron transport chain, the distribution of oxygen throughout the body by hemoglobin, and as a cofactor in numerous enzymatic reactions. Despite its importance, excess iron can be very toxic to the cell. Its participation in the Fenton reaction results in the formation of free radicals, which can wreak havoc by oxidizing lipids, cleaving proteins, and damaging DNA and RNA. Because of this duality, cells and organisms have evolved exquisite control mechanisms to ensure that the proper amount of iron is present, that excess iron is stored in non-toxic forms, and that within the body iron is chaperoned both outside and within cells.

The body needs 20–30 mg of iron per day for erythropoiesis, the vast majority of this iron is acquired through the efficient recycling of red blood cells by macrophages (Cook et al., 1973). The remainder is met by the dietary absorption of ∼2 mg of iron per day (Cook et al., 1973). The iron importer, divalent metal transporter 1(DMT1) is located on the apical side of enterocytes in the small intestine and transports dietary iron, which has been reduced to ferrous iron by the ferrireductase, Dcytb, into the cell (Gunshin et al., 1997; Andrews, 1999; McKie et al., 2001). Once in the cell, iron that is not exported can be stored in the iron storage protein, ferritin, which is a 24-subunit protein with a hollow core that can oxidize and store up to 4500 atoms of ferric iron (Koorts and Viljoen, 2007). Enterocytes are quickly turned over by sloughing into the lumen of the intestine. Thus, enterocyte iron stored in ferritin is lost if not mobilized beforehand. Iron transport into the bloodstream is accomplished through the basolateral iron exporter, ferroportin (FPN; Abboud and Haile, 2000; Donovan et al., 2000; McKie et al., 2000), which facilitates the transport of ferrous iron and its subsequent oxidation by

hephaestin and loading into transferrin (Tf) the iron transport protein in plasma (Vulpe et al., 1999; Nemeth et al., 2004). This process is at the crux of body iron regulation; amount of FPN on the basolateral membrane determines whether iron is transported into the body and FPN levels are regulated by the peptide hormone, hepcidin.

Hepcidin is regulated in and secreted by hepatocytes in the liver. This 25 amino acid peptide is the result of two cleavages of the 87 amino acid precursor, a prepropeptide cleavage of the signal sequence, and a propeptide cleavage by furin (Valore and Ganz, 2008). Serum and urine levels correspond to mRNA levels, indicating that in many circumstances hepcidin is regulated at the level of transcription (Kemna et al., 2008). Serum hepcidin binds FPN and stimulates the internalization and subsequent degradation of FPN, thus making it a negative regulator of total body iron influx (Nemeth et al., 2004).

In addition to its role as a master regulator of dietary iron absorption, hepcidin plays a role in immunity. Macrophages can also secrete hepcidin in response to inflammation through the TLR-4 signaling pathway (Liu et al., 2005; Peyssonnaux et al., 2006). Increased hepcidin prevents both absorption of iron from the gut, and the release of iron from iron-recycling macrophages, restricting the use of iron by invading pathogens. High levels of hepcidin in diseases of chronic inflammation cause the anemia of chronic disease (ACD). In contrast, abnormally low levels of hepcidin expression leads to iron-overload. Genetic mutations that reduce hepcidin expression result in the disease, HH.

Of the four types of HH, three are the result of low levels of hepcidin expression; Type I, (HFE mutation), Type II a and b [hemojuvelin (HJV) and hepcidin mutations respectively], and Type III (TfR2 mutation). Type IV is a result of mutations in FPN itself, making it insensitive to hepcidin regulation. Type I and III

mutations cause a slow accumulation of iron over the individual's lifetime and a disease phenotype starting in adulthood, while Type II a and b mutations have the more severe phenotype of Juvenile Hemochromatosis. Left untreated, HH leads to iron accumulation in the liver, heart, pancreas, and joints leading to cirrhosis, arrhythmias, diabetes, and arthritis (Weber, 1931; Schumacher, 1964; Cecchetti et al., 1991). The slow onset of adult HH clearly shows that fine-tuning of the daily influx of iron is necessary for normal iron homeostasis.

TfR2 is the putative "iron sensor" that fine-tunes this system, based on its ability to bind and be stabilized by iron-bound Tf (Johnson and Enns, 2004; Robb and Wessling-Resnick, 2004; Johnson et al., 2007). In addition, its mutation leads to low hepcidin expression (Kawabata et al., 2005; Nemeth et al., 2005), as well as an inability to respond to acute iron-loading (Girelli et al., 2011).

## **BINDING PARTNERS, REGULATION, AND TRAFFICKING OF TfR2**

The structure of TfR2 and its known binding partners provide clues to its function. The interesting features of TfR2 start in its structural similarity to the originally characterized transferrin receptor 1 (TfR1). They are both type II membrane proteins that function as a disulfide-linked dimer and share 66% homology in their ectodomain (Kawabata et al., 1999). Both TfR1 and TfR2 bind Tf and many of the amino acids involved in the binding of TfR1 to Tf are conserved in TfR2 (Giannetti et al., 2003). TfR1 binds iron-loaded Tf (holo-Tf) with a KD of 1.1 nM and is responsible for the endocytosis of holo-Tf into acidic compartments. This iron is then released from Tf, reduced, and transported across the endosomal membrane by the metal transporters DMT1 and Zip14 (Dautry-Varsat et al., 1983; Rao et al., 1983; West et al., 2000; Zhao et al., 2010). While TfR2 is capable of iron uptake, its binding affinity for Tf is 25-fold less than that of TfR1 (Kawabata et al., 1999; West et al., 2000). This difference in affinity may enhance the role of TfR2 as an iron sensor, allowing it to be sensitive to changes in Tf saturation in the blood. The lower affinity of TfR2 does not seem to diminish its ability to endocytose iron. In TRVb cells lacking both TfR1 and TfR2, transfection of TfR2 increased Tf- mediated 55Fe uptake to similar levels as transfected TfR1(Kawabata et al., 1999). TfR1 is expressed in many tissues whereas *TfR2* expression is limited to the liver and erythropoietic progenitors (Sposi et al., 2000). The limited expression of *TfR2* may explain why deletion of TfR1 is embryonic lethal (Levy et al., 1999). While both TfR1 and TfR2 bind and endocytose Tf, their different affinity for Tf and different expression patterns suggest different functions.

Other differences exist which explain the inability of TfR2 to replace TfR1. TfR1 and TfR2 are differentially regulated by iron and holo-Tf. Iron response elements (IRE's) on the 3 TfR1 mRNA account for the rapid turnover of TfR1 mRNA under high iron conditions, which functions to reduce iron import (Owen and Kuhn, 1987). While TfR1 mRNA levels respond quickly to iron levels it is a relatively stable protein with a turnover of ∼24 h. Therefore, the response of cells to high intracellular iron by downregulation of TfR1 is relatively slow. In contrast, *TfR2* lacks the IRE's for the regulation of its mRNA by intracellular iron and at the protein level, turns over much faster. The binding of Tf to TfR2

regulates both its stability and its trafficking within cells (Johnson and Enns, 2004; Johnson et al., 2007). In the presence of holo-Tf, TfR2 levels are increased by redirection of TfR2 to the recycling endosomes, which increases its stability (Johnson and Enns, 2004; Robb and Wessling-Resnick, 2004; Chen et al., 2009). These differences are the result of very distinct cytoplasmic domains. The TfR1 and TfR2 cytoplasmic domains both have a YXX--based endocytic motif for clathrin-mediated endocytosis, but share little else. In addition to the YXX motif, TfR2 also has a phosphofurin acidic cluster sorting-1 (PACS-1) motif and coprecipitates with the PACS-1 protein (Chen et al., 2009). This motif is most likely responsible for the Tf-dependent recycling of TfR2 from endosomes to the cell surface (Chen et al., 2009). Human TfR2 is glycosylated at three sites: 240, 339, and 754. This glycosylation is necessary for the Tf-induced stabilization of TfR2, but does not affect its ability to bind Tf or its trafficking to the cell surface (Zhao and Enns, 2013). Despite their structural similarity and ability to bind Tf, the differences in Tf-induced stability and the cytoplasmic domains of TfR1 and TfR2 indicate that they both handle and are affected by Tf differently.

In addition to functional differences in Tf handling, TfR1 and TfR2 appear to interact with the original hereditary hemochromatosis protein (HFE) through alternate domains. TfR1 and HFE interact through the helical domain of TfR1 and the α1 and α2 domains of HFE (Bennett et al., 2000). Tf and HFE compete with each otherfor binding to TfR1 because they have overlapping binding sites (Giannetti et al., 2003; Giannetti and Bjorkman, 2004). TfR2 and HFE interact through the TfR2 stalk region between residues 104 and 250 and the HFE α3 domain (Chen et al., 2007; D'Alessio et al., 2012). The binding sites of HFE and Tf do not appear to overlap in TfR2 (Chen et al., 2007). This lends itself to the hypothesis that Tf-binding to TfR1 releases HFE, making it available to functionally interact with TfR2. Coprecipitation studies indicate that TfR2 and HFE interact readily, however, TfR2/HFE interaction remains controversial as coprecipitation of endogenous Tfr2 from liver lysates expressing myc-tagged Hfe did not yield positive results (Chen et al., 2007; Schmidt and Fleming, 2012). However, in terms of functionality, it appears that both TfR2 and HFE are needed for Tf-sensing (Gao et al., 2010). In addition to the binding of HFE and Tf, a recent report has found an interaction between TfR2 and the BMP co-receptor, HJV, which is an interesting link between the TfR2/HFE complex and BMP-signaling (D'Alessio et al., 2012). The ability of HFE, TfR2, and HJV to form a complex *in vitro*, coupled with the fact that mutations in any one of these proteins causes HH suggests a role for this complex in the regulation of hepcidin. This is consistent with the different roles of TfR1 and TfR2. TfR1 regulates cellular iron uptake and TfR2 senses iron levels and regulates body iron uptake.

## **DISEASE-CAUSING MUTATIONS IN TfR2**

TfR2 mutations result in the disease HH. Unlike HFE HH, which seems to have for the most part risen from a single amino acid HFE mutation and spread throughout Europe, TfR2 HH is far rarer and is the result of various mutations. The first reported TfR2 mutation, the truncation mutant *Y250X*, was found in two unrelated Sicilian families (Camaschella et al., 2000). Since then, a variety of other TfR2 mutations have been found in Italian patients (Majore et al., 2006; Biasiotto et al., 2008; Gerolami et al., 2008; Radio et al., 2014). Because the most common mutation in HFE is not present in the Japanese population, Japanese patients with HH most frequently have mutations in TfR2 (Hattori et al., 2003; Koyama et al., 2005; Hayashi et al., 2006). Many mutations in TfR2 identified to date fail to give insight into TfR2 function because most mutations result in misfolded proteins that remain in the endoplasmic reticulum (ER; Wallace et al., 2008), where they are presumably degraded by the ER quality control pathway. Two interesting point mutations, Q890P and M172K (predicted to disrupt Tf binding and HFE binding respectively; Mattman et al., 2002; Roetto et al., 2010), fail to reach the cell surface in analogous mouse mutations (Q685P and M167K; Wallace et al., 2008). TfR2 mutations continue to be found in patients around the world (Le Gac et al., 2004; Hsiao et al., 2007; Zamani et al., 2012)

## **HEPCIDIN REGULATION**

Hepcidin is a primary regulator of total body iron homeostasis and disease-causing TfR2 mutations cause HH through a reduction in hepcidin transcription. The transcription of hepcidin is regulated by iron, bone morphogenetic proteins (BMPs), inflammation, hypoxia, and erythropoietic activity. The hepcidin promoter has two BMP response elements (REs), the distal BMP RE2 and the proximal RE1 (Verga Falzacappa et al., 2008; Truksa et al., 2009). Within the proximal BMP RE1 lies a signal transducer and activator of transcription3 (STAT3) binding site that is responsible for upregulation of hepcidin in response to inflammation (Verga Falzacappa et al., 2008). Interestingly, response of hepcidin to inflammation requires the BMP-binding element in RE1 be intact as well as the STAT binding site, indicating that BMP signaling may be required to keep the chromatin open for STAT binding (Wang et al., 2005; Casanovas et al., 2009) Within the distal BMP RE2, lies a hepatocyte-specific hepatocyte nuclear factor 4 alpha (HNF4α) binding site as well as bZIP (basic leucine zipper domain) and COUP (chicken ovalbumin upstream promoter transcription factor) motifs indicating that hepcidin transcription relies on a set of transcriptionfactors, including ones that are tissue specific (Truksa et al., 2009). Stimulation of hepcidin in response to BMP signaling and HJV expression requires that both BMP RE1 and BMP RE2 be intact. The distal BMP RE2 is required for hepcidin response to iron levels (Truksa et al., 2007, 2009). While hepcidin can respond to inflammation and hypoxia, it seems that regulation of hepcidin expression requires BMP-signaling.

While BMP's 2, 4, and 6 are all expressed in the liver and are capable of stimulating hepcidin expression in hepatocytes and hepatoma cell lines, BMP-6 is the only one that is positively regulated by iron (Kautz et al., 2008), and it is expressed mainly in the endothelial cells of the liver (Knittel et al., 1997; Zhang et al., 2011; Enns et al., 2013). Deletion of BMP-6 results in severe ironoverload, confirming its role in iron homeostasis (Andriopoulos et al., 2009; Meynard et al., 2009). BMP-6 also interacts with the BMP co-receptor HJV (Andriopoulos et al., 2009). This indicates that while other BMP's may have an effect on basal hepcidin expression, it is BMP-6 that is regulated by iron, and functions as a regulator of iron homeostasis. As of yet, how BMP-6 is regulated by iron is not known. Deletion of BMP-6 has little effect

on bone formation, with only a slight delay in sternal ossification (Solloway et al., 1998), indicating that BMP-6 may function primarily as a regulator of iron homeostasis.

BMP-signaling requires binding of BMP's to BMP-receptors. BMP-receptors are arranged as a tetramer of two type I and type II receptors. The type II receptors phosphorylate the type I receptors upon ligand binding, and the phosphorylated receptor can then phosphorylate and activate intracellular receptor-associated SMADs (R-SMADs), which then bind the co-SMAD, SMAD4 to then enter the nucleus and regulate transcription (Wrana et al., 1992; Lagna et al., 1996; Macias-Silva et al., 1996; Souchelnytskyi et al., 1996). In the liver, HJV binds to the BMP type II receptor ActRIIa to enhance BMP-signaling and hepcidin expression (Xia et al., 2008). Two BMP type I receptors are involved in hepcidin regulation, Alk2 and Alk3 (Babitt et al., 2006; Xia et al., 2008; Steinbicker et al., 2011). The Alk3 receptor is necessary for basal hepcidin expression in mice and deletion of Alk3 has a more severe iron-overload phenotype than Alk2, however, Alk2 seems to be necessaryfor the response of hepcidin to iron and HJV (Steinbicker et al., 2011). Therefore, the ligand BMP-6, the BMP co-receptor HJV, and BMP receptors ActRIIa, Alk2, and Alk3 all make up the liver BMP-signaling pathway.

The BMP-signaling pathway can also be modulated by the inhibitory SMADs (iSMADs), SMAD6 and SMAD7. These iSMADs are part of a negative feedback loop and are induced by BMP signaling. They inhibit BMP-signaling by binding to BMP receptors (which inhibits SMAD phosphorylation), recruiting ubiquitin ligases (to induce degradation of the receptors), or they can enter the nucleus and disrupt binding of phosphorylated SMADs to target genes (Hayashi et al., 1997; Kavsak et al., 2000; Zhang et al., 2007). In keeping with their role as negative-feedback loop inhibitors of BMP-signaling, SMAD6 and SMAD7 are coregulated with hepcidin and SMAD7 is upregulated in response to iron (Kautz et al., 2008; Vujic Spasic et al., 2013). In addition, SMAD7 can modulate signaling by directly binding the promoter region and inhibiting hepcidin expression (Mleczko-Sanecka et al., 2010).

Deletion of HFE, HJV, or TfR2 results in a reduction of phosphorylated SMADs 1/5/8, indicating that reduced hepcidin expression is mediated through the BMP-signaling pathway (Babitt et al., 2006; Corradini et al., 2009, 2011). Presumably, HJV regulates BMP-signaling through enhancing binding of BMP ligand to BMP-receptors and promoting the assembly of the BMP-signaling complex (Babitt et al., 2006). The severe phenotype of HJV knockout mice and juvenile hemochromatosis patients is in keeping with the important role of HJV as a BMP co-receptor and with the importance of the BMP-signaling pathway in basal hepcidin transcription. HFE and TfR2 mutations are far less severe, indicating that they are involved in fine-tuning of iron levels. How HFE and TfR2 modulate hepcidin expression through the BMP-signaling pathway is not understood.

## **PHYSIOLOGICAL FUNCTION OF TfR2**

TfR2 has been hypothesized to be the Tf-sensor since the discovery of its disease-causing mutations (Camaschella et al., 2000; Fleming et al., 2000; Roetto et al., 2001). Wild type mouse primary hepatocytes, when treated with holo-Tf, will respond within 24 h by a 2-fold upregulation of hepcidin expression and wild type mice injected with iron will also see this increase in hepcidin levels (Kawabata et al., 2005; Lin et al., 2007; Ramey et al., 2009; Ramos et al., 2011). This mirrors the ∼2-fold increase in urinary hepcidin seen in humans who were challenged with iron and had a corresponding increase in Tf-saturation (Lin et al., 2007; Girelli et al., 2011). In contrast, Tfr2 mutant mouse primary hepatocytes do not respond to treatment, indicating a role of Tfr2 in Tf-sensitivity (Gao et al., 2009). In addition, deletion of the Tfr2 binding partner, Hfe, also results in loss of Tf- sensitivity, indicating that it may be the TfR2/HFE complex that is involved in iron-sensing (Gao et al., 2009). In human patients with TfR2 HH, urinary hepcidin levels do not respond to iron challenge, and HFE HH patients have a blunted hepcidin response, indicating that both molecules are needed in order to modulate iron uptake in response to dietary iron (Girelli et al., 2011). The 2-fold hepcidin response to holo-Tf in primary hepatocytes is physiological, as HFE HH patients only have a 2-fold difference in hepcidin levels and the disease results in the slow accumulation of iron over the lifetime of the individual (Van Dijk et al., 2008).

While mutations in TfR2 and HFE both result in a slow disease progression, loss of TfR2 appears to be more severe than loss of HFE. There is a reported case of juvenile hemochromatosis resulting from TfR2 mutation and serum hepcidin levels are lower in TfR2 HH patients (Nemeth et al., 2005; Pietrangelo et al., 2005; Girelli et al., 2011). Because of the scarcity of TfR2 HH patients in contrast to HFE HH patients, it is hard to compare severity of HFE and TfR2 mutations. Tfr2 mutant mice of the same genetic background as Hfe−/<sup>−</sup> mice have higher iron accumulation than Hfe−/<sup>−</sup> mice (Wallace et al., 2009). Mice and humans lacking both Tfr2 and Hfe have a more severe phenotype than either single mutation (Pietrangelo et al., 2005; Wallace et al., 2009; Corradini et al., 2011), indicating that either one or both proteins may have alternate functions, or that the complex may be able to partially function with one member missing. Transfection of HFE into cell lines that do and do not express TfR2 decreases iron uptake indicating that HFE almost certainly has another function than that of the TfR2/HFE complex (Roy et al., 1999; Wang et al., 2003; Carlson et al., 2005). Other responses that can be attributed to TfR2 remain unknown.

The existence and role of the Tfr2/Hfe complex is not without controversy. While Tfr2 and Hfe immunoprecipitate readily in transfected cells, one report was unable to confirm interaction with endogenous Tfr2 in primary hepatocytes expressing myctagged Hfe transgene (Schmidt and Fleming, 2012). In addition, reports differ as to whether Hfe overexpression in Tfr2 mutant mice can increase hepcidin levels and reduce iron accumulation in mice (Gao et al., 2010; Schmidt and Fleming, 2012). These results are further complicated by the ability of Hfe to affect iron uptake, as chronic higher iron stores lead to increased BMP-6 expression independently of either HFE or TfR2. BMP-6 expression is not however, dependent on Tf-saturation. Tf-deficient mice have high iron stores, despite being anemic (Trenor et al., 2000). In keeping with their high tissue iron levels, these mice have increased BMP-6 levels while hepcidin levels are still below normal (Bartnikas et al., 2011; Bartnikas and Fleming, 2012), indicating that Tf is a necessary part of the BMP-signaling pathway that

leads to hepcidin expression. Tf-deficient mice that are treated with Tf increase hepcidin expression, however, this increase in hepcidin expression is attenuated when Hjv is also deleted (Bartnikas and Fleming, 2012), indicating that the hepcidin response to Tf requires HJV. Experiments in isolated primary hepatocytes get around the complication of BMP-6 expression, because BMP-6 is expressed in the endothelial cells. In primary hepatocytes, both Hfe and Tfr2 are needed for a hepcidin response to holo-Tf, indicating that, at least in regards to blood iron-sensing, both proteins are required.

While the mechanism by which TfR2 affects hepcidin expression through the BMP-signaling pathway remains nebulous, its ability to bind Tf and its requirement for the hepcidin response to holo-Tf makes it likely that TfR2 is the Tf- sensor. That it also interacts with HFE, and that HFE is also required for the response of hepcidin to holo-Tf provides a strong indication that TfR2 senses iron as part of a TfR2/HFE complex, and that this complex formation is important to body iron homeostasis. The further binding of both HFE and TfR2 to the BMP co-receptor, HJV, provides a possible link between the TfR2/HFE Tf-sensing complex and the BMP-signaling complex, and further research is needed to ascertain the functionality of this complex.

## **CURRENT TfR2 MODELS**

The controversy regarding the TfR2/HFE complex, coupled with the new report of a TfR2/HFE/HJV complex lends itself to three possible models for the Tf-sensitive regulation of hepcidin by TfR2 (**Figure 1**). First, if TfR2 and HFE do not functionally interact, then Tf/TfR2, HFE, and the BMP-signaling complex affect pSMAD levels independently of one another. Support for this model lies in the increased severity of the Tfr2-Hfe double knockout mouse (Wallace et al., 2009) and the failed interaction of the myc-tagged Hfe transgene with endogenous Tfr2 (Schmidt and Fleming, 2012). Second, if reports of TfR2/HFE and TfR2/HFE/HJV interactions are functionally significant, then the TfR2/HFE complex could interact with the BMP-signaling complex upon Tf-binding, thereby affecting pSMAD levels. Support for this model lies in the reports of TfR2/HFE interaction (Goswami and Andrews, 2006; Chen et al., 2007; D'Alessio et al., 2012), the requirement of both Tfr2 and Hfe for Tf-sensitivity (Gao et al., 2009, 2010), and the recent report of a TfR2/HFE/HJV complex (D'Alessio et al., 2012). The third model proposes that HJV interacts with both the TfR2/HFE complex and the BMP-signaling complex, and both of these complexes affect pSMAD levels independently. While the functional significance of the TfR2/HFE complex or the TfR2/HFE/HJV complex may still be up for debate, TfR2 plays an important role in regulating hepcidin levels in response to holo-Tf through the BMP-signaling pathway.

## **SUMMARY**

TfR2 plays an important role in the fine-tuning of body iron uptake and loss of TfR2 function leads to Type III HH. TfR2 senses changes in blood-iron levels through its interaction with holo-Tf. While it is similar in structure to the iron-endocytosis protein, TfR1, it has a lower affinity for Tf, an alternate binding site for HFE, and is differentially trafficked and regulated. These differences, along with the tissue expression pattern of TfR2, indicate

enhance pSMAD.

that the function of TfR1 is to bind and endocytose iron for cellular purposes, while the function of TfR2 is to sense blood iron levels. TfR2 is able to regulate body iron uptake in response to blood iron levels by modulating hepcidin expression through the BMPsignaling pathway. The formation and functional significance of the TfR2/HFE complex remains controversial, but both proteins are necessary for Tf-sensitivity. The interaction of TfR2/HFE with the BMP co-receptor, HJV,may provide an interesting link between TfR2, HFE, and the BMP-signaling pathway.

induction of hepcidin transcription. **(B)** Tf induces the formation of a large

## **ACKNOWLEDGMENTS**

Funding for this work was provided in part by the Ruth L. Kirschstein T32 training grant (5T32GM071338-08) to Christal A. Worthen and NIH RO1 DK054488 to Caroline A. Enns. We would like to thank Drs. An-Sheng Zhang and Ningning Zhao for carefully reading this manuscript.

## **REFERENCES**


Cook, J. D., Barry,W. E., Hershko, C., Fillet, G., and Finch, C. A. (1973). Iron kinetics with emphasis on iron overload. *Am. J. Pathol.* 72, 337–344.


type 3 hemochromatosis. *J. Hepatol.* 47, 303–306. doi: 10.1016/j.jhep.2007. 04.014


absorption of dietary iron. *Science* 291, 1755–1759. doi: 10.1126/science. 1057206


normal hematopoiesis. *Eur. J. Biochem.* 267, 6762–6774. doi: 10.1046/j.1432- 1033.2000.01769.x


characteristics, laboratory data and gene mutations. *Med. Sci. Monit.* 18, CR622–CR629. doi: 10.12659/MSM.883489


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

*Received: 14 January 2014; paper pending published: 04 February 2014; accepted: 18 February 2014; published online: 06 March 2014.*

*Citation: Worthen CA and Enns CA (2014) The role of hepatic transferrin receptor 2 in the regulation of iron homeostasis in the body. Front. Pharmacol. 5:34. doi: 10.3389/fphar.2014.00034*

*This article was submitted to Drug Metabolism and Transport, a section of the journal Frontiers in Pharmacology.*

*Copyright © 2014 Worthen and Enns. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*

**REVIEW ARTICLE** published: 22 April 2014 doi: 10.3389/fphar.2014.00082

## Pathophysiology of the Belgrade rat

## *Tania Veuthey and MarianneWessling-Resnick\**

Department of Genetics and Complex Diseases, Harvard School of Public Health, Boston, MA, USA

#### *Edited by:*

Raffaella Gozzelino, Instituto Gulbenkian de Ciência, Portugal

#### *Reviewed by:*

Sharad Kumar, SA Pathology, Australia Elizabeth A. Leibold, University of Utah, USA

*\*Correspondence:* Marianne Wessling-Resnick, Department of Genetics and Complex Diseases, Harvard School of Public Health, 665 Huntington Avenue, Boston, MA 02115, USA e-mail: wessling@hsph.harvard.edu

The Belgrade rat is an animal model of divalent metal transporter 1 (DMT1) deficiency. This strain originates from an X-irradiation experiment first reported in 1966. Since then, the Belgrade rat's pathophysiology has helped to reveal the importance of iron balance and the role of DMT1. This review discusses our current understanding of iron transport homeostasis and summarizes molecular details of DMT1 function. We describe how studies of the Belgrade rat have revealed key roles for DMT1 in iron distribution to red blood cells as well as duodenal iron absorption. The Belgrade rat's pathology has extended our knowledge of hepatic iron handling, pulmonary and olfactory iron transport as well as brain iron uptake and renal iron handling. For example, relationships between iron and manganese metabolism have been discerned since both are essential metals transported by DMT1. Pathophysiologic features of the Belgrade rat provide us with a unique and interesting animal model to understand iron homeostasis.

**Keywords: SLC11A2, DMT1, iron, manganese, Belgrade rat**

## **OVERVIEW OF IRON HOMEOSTASIS**

Iron is present in hemoproteins, such as the important oxygen carriers hemoglobin and myoglobin, and it is a component of non-heme proteins that carry out other key functions in cellular metabolism, such as mitochondrial aconitase and ribonucleotide reductase. The relevance of iron lies in its ability to cycle reversibly between the ferrous (Fe2+) and the ferric (Fe3+) oxidation states (Wessling-Resnick, 1999). Free ferrous iron is a potent catalyst for lipid peroxidation and protein and DNA oxidation since it is able to react with molecular oxygen and generate reactive oxygen species (ROS) through Fenton chemistry (Hentze et al., 2004) Given the biological importance of iron and the potential for its toxicity, iron homeostasis is tightly regulated.

Dietary iron is absorbed through the duodenum. Regulated pathways of iron excretion do not appear to exist, and therefore regulation of iron absorption is critical to maintain iron balance. Iron reaches the liver by portal circulation, where it can be stored until needed or delivered through systemic circulation to peripheral tissues. A significant amount of iron is transported to the bone marrow where erythropoiesis takes place. Senescent red blood cells (RBCs) are phagocytosed by reticuloendothelial macrophages, which catabolize iron from heme to be recycled and used for new RBC synthesis. This recovery is highly efficient so that most of the body iron is usually found in circulating hemoglobin within erythrocytes (Wessling-Resnick, 2000; Hentze et al., 2004).

Duodenal absorption of iron begins at the apical membrane of enterocytes, where the reduction of Fe3<sup>+</sup> to Fe2<sup>+</sup> is carried out by DcytB and possibly other ferrireductases (McKie et al., 2001; Mackenzie and Garrick, 2005). Then, iron is transported into the cell through its primary importer DMT1 [divalent metal transporter 1; SLC11A2 (solute carrier family 11, member 2); DCT1 (divalent cation transporter 1)]. Once iron enters the cell it may be stored in ferritin, a protein complex constituted by heavy and

light chains that is able to store up to 4500 iron atoms (Harrison and Arosio, 1996). Iron can also be immediately exported from the cell by ferroportin [Fpn; SLC40A1; MTP1 (metal transporter protein 1)], which is located in the basolateral membrane of enterocytes (Abboud and Haile, 2000; Donovan et al., 2000; McKie et al., 2000). The ferroxidase hephaestin mediates its conversion to the ferric state before it is released and bound by serum transferrin (Tf; Wessling-Resnick, 2006). **Figure 1** summarizes the elements involved in this process.

At the site of liver iron storage, hepatocytes take up iron through two distinct pathways. Tf-bound iron is taken by the Tf cycle (**Figure 2**). In this pathway, iron-loaded Tf binds to its receptor on the cell surface at neutral pH. After binding, the complex is internalized by receptor-mediated endocytosis, entering an endosomal compartment that is acidified by a proton pump. This acidification enables release of iron from Tf and its subsequent transport out of the endosome through DMT1. Subsequently, the apo-Tf–Tf receptor (TfR) complex returns to the cell surface to dissociate at neutral pH (Dautry-Varsat et al., 1983). There are two TfRs that have been characterized: TfR1 is ubiquitously expressed while TfR2 is only expressed in certain tissues, including the liver. Hepatocytes are also capable of importing non-Tf-bound iron (NTBI), which accumulates in circulation during iron loading conditions such as hereditary hemochromatosis. Although DMT1 might be one route for NTBI uptake (Shindo et al., 2006), a member of the SLC39 ZIP family called Zip14 may play a more important role in this process (Liuzzi et al., 2006; Nam et al., 2013; Wang and Knutson, 2013). Once inside of the cell, the iron can be stored as ferritin or released back into the circulation in response to deficiency conditions. The latter process is thought to involve iron export by ferroportin (Wessling-Resnick, 2006).

Because most of the body's iron is found in hemoglobin, erythrocytes are especially important in iron handling. Erythroid

precursors take up iron by the Tf cycle in a highly efficient manner; they probably lack an iron export mechanism since all the iron is retained by the cell, intended for hemoglobin synthesis (Andrews, 2000). The release of iron from the endocytic compartment requires functional DMT1 in these cells (Fleming et al., 1998). A ferrireductase, Steap3, acts to reduce Fe3<sup>+</sup> to Fe2+, enabling transport to the cytosol (Ohgami et al., 2005). Iron entering the erythron goes to the mitochondria where the major proportion of intracellular iron metabolism takes place (Garrick and Garrick, 2009).

Although iron homeostasis mainly relies on the control of iron efflux from duodenal enterocytes and iron reutilization by macrophages balanced by the demands of heme synthesis by RBCs, increasing data suggest that the kidney could also be involved in systemic iron handling (Moulouel et al., 2013; Veuthey et al., 2013). DMT1 is found at the apical membrane of renal tubular cells (Ferguson et al., 2001; Canonne-Hergaux and Gros, 2002; Veuthey et al., 2008). Wareing et al. (2000) demonstrated that a significant amount of iron is filtered at the glomerulus and the majority is reabsorbed along the nephron. There is evidence to suggest that Tf-bound iron may be filtered and reabsorbed along the proximal tubule following endocytosis via the megalin–cubilin pathway, a possible novel TfR, and/or the TfR1 (Kozyraki et al., 2001; Smith and Thevenod, 2009). These data support the hypothesis that the kidney, in addition to maintaining its self-supply, may also undertake iron reabsorption and thus contribute to general

iron homeostasis (**Figure 3**). Furthermore, renal Fpn expression in proximal tubule-specific ferritin heavy chain-knockout mice is decreased, both in mRNA and protein levels (Zarjou et al., 2013). Under normal conditions Fpn was predominantly expressed in the apical membrane of proximal tubules, while after acute kidney injury Fpn was redistributed to the cytosol and basolateral membrane. This recent finding reinforces the hypothesis that kidney may be more extensively involved in iron metabolism than initially proposed.

## **DIVALENT METAL TRANSPORTER 1**

DMT1, also known as Nramp2, DCT1, and SLC11A2, was identified in 1997 by two groups using different approaches. Gunshin et al. (1997) uncovered the iron transporter by functional expression cloning and Fleming et al. (1997) determined that defects in its gene were responsible for the microcytic anemia phenotype of *mk* mice. As outlined above, it is the major point of iron entry into the body. DMT1 has 12 putative transmembrane (TM) domains. Predicted glycosylation sites in the fourth extracellular loop and a consensus transport motif in the fourth intracellular loop were defined, with both N- and C-termini determined to be topologically situated within the cytoplasm (Gruenheid et al., 1995; Gunshin et al., 1997). Three negatively charged and highly conserved residues in TM 1, 4, and 7 of DMT1 are suggested to be essential for cation transport. In addition, two histidine residues in the TM domain 6 appear critical to normal function (Lam-Yuk-Tseung et al., 2003). DMT1 selectively imports iron in a pH-dependent manner and it has been described to operate as a H+/divalent cation cotransporter (symporter) but also as a H+ uniporter. Another important feature is the voltage-dependent activity of DMT1 gradient (Gunshin et al., 1997).

Transcription of the SLC11A2 genes encoding DMT1 originates four different mRNA transcripts, called 1A/+IRE, 1A/−IRE, 1B/+IRE, and 1B/−IRE. Alternate promoters determine whether the 5 end of the mRNA will be exon 1A or exon 1B (Hubert and Hentze, 2002). Moreover, variable 3 processing give rise to two transcripts that differ in the 3- -translated and untranslated regions (UTRs). One of the transcripts is referred as +IRE because it contains an iron-responsive element (IRE) in its 3- -UTR, while the other is called −IRE because it lacks that element (Lee et al., 1998). The four transcripts encode related but distinct proteins. Thus, +IRE and −IRE encode two isoforms that differ in the Cterminus: an 18 amino acid residue found in the +IRE isoforms while −IRE transcripts have a 25 amino acid residue.

The key difference between +IRE and −IRE transcripts is the fact that the presence of the IRE confers post-transcriptional regulation by iron status. Similar to TfR regulation (Galy et al., 2013), decreased iron supply promotes the interactions of the +IRE DMT1 transcript with iron regulatory proteins (IRPs), resulting in the stabilization of the mRNA (Gunshin et al., 2001; Pantopoulos, 2004). This is one mechanism that accounts for the increased duodenal expression of DMT1 seen under iron deficiency, thus promoting iron uptake (Canonne-Hergaux et al., 1999). Another mechanism is conferred by 5 regions flanking exon 1A,which contain hypoxia response elements (HREs). In the intestine, HREs are recognized by HIF-2α to modulate DMT1 expression under hypoxic conditions (Mastrogiannaki et al., 2009; Shah et al., 2009).

The isoforms also differ in the sites of expression. Whereas the 1B isoform is ubiquitous, the 1A isoform appears to be tissuespecific. Expression of DMT1 1A isoform in duodenal enterocytes has been reported by several authors, with a gradient of expression from the proximal to distal small intestine (Gunshin et al., 1997; Canonne-Hergaux et al., 1999). The kidney is the tissue with second highest expression of the 1A isoform mRNA of DMT1 (Hubert and Hentze, 2002). On the other hand, both +IRE and −IRE isoforms are expressed in many tissues such as liver, lung, brain, and thymus (Gunshin et al., 1997; Hubert and Hentze, 2002).

As the name implies, DMT1 is able to transport a broad spectrum of divalent metals, although its affinity for Fe2<sup>+</sup> is much higher than for other metals (Mackenzie and Hediger, 2004). Manganese, in particular, possesses many physicochemical properties similar to iron, and both metals have been shown to compete in membrane transport processes such as intestinal absorption (Thomson et al., 1971; Rossander-Hulten et al., 1991) and uptake by erythroid cells (Morgan, 1988; Chua et al., 1996). Like iron, manganese is an essential nutrient, and exogenous expression studies (Conrad et al., 2000; Garrick and Dolan, 2002), and more recent molecular studies (Roth and Garrick, 2003; Thompson et al., 2007b; Illing et al., 2012; Kim et al., 2012) have documented a role for DMT1 in manganese uptake as well as iron transport. Unlike iron deficiency, hypomanganesemia is rare. However, manganese loading or manganism can be caused by excess metal

exposure resulting in a Parkinson-like disorder. Loading of this metal is not typically associated with ingestion since hepatic firstpass elimination provides an important protective mechanism against potential toxicity. Since intake of airborne Mn2<sup>+</sup> bypasses the biliary excretion route, distribution of metal directly to the brain can occur, promoting neurotoxicity. Several studies have demonstrated a strong relation between high levels of manganese and impaired behavior (Yamada et al., 1986; Tran et al., 2002). The role of DMT1 not only in iron pathophysiology but also in manganese toxicity will be discussed below.

## **THE BELGRADE RAT**

Belgrade (*b*) rats were described for first time in 1966 as offspring of an X-irradiated albino rat in Belgrade, Yugoslavia (Sladic-Simic et al., 1966). No differences were observed in litters of F1 generation. When the female was bred with a normal male, 11 apparently normal rats were born, four of which died young. Of the seven remaining, two pairs were mated, producing anemic rats in the F2 generation. In subsequent filial generations anemic rats were also seen, suggesting that offspring were heterozygous for a recessive mutation that caused anemia.

The newborn anemic rats were pale, with growth retardation compared to normal littermates. Peripheral blood smears showed marked microcytosis, anisocytosis, and poikilocytosis that increased with age. The mean value for erythrocytes in the peripheral blood was significantly decreased and a progressive decline in

**FIGURE 3 | Hypothetical model of renal iron handling.** Filtered Tf–Fe is reabsorbed in proximal tubule cells by receptor-mediated endocytosis. Potential receptors include cubilin–megalin complex, a possible novel Tf receptor (TfR), and TfR1. Reabsorption of iron bound to a chelator (LCN2) has also been described, although the receptor involved is less clear. Alternatively, ferrous iron is directly imported into the cell by DMT1.

Although several features of the model are not completely known, it has been postulated that iron complexes could be endocytosed. Once in the endosome, iron would be reduced and released into the cytosol by endosomal DMT1. Finally, ferrous iron is exported by FPN located in the basolateral membrane; an apical distribution has been noted under conditions of renal failure.

**FIGURE 4 | Belgrade rat.** A b/b rat pup (foreground) and heterozygous +/b littermate are shown at post-natal day 10. Image courtesy of Dr. Xuming Jia.

hemoglobin values occurred with age (Sladic-Simic et al., 1966). Adult *b/b* rats are distinguished by lower body weight than +/b rats (Thompson et al., 2007a), and characterized by pinkish retinal reflex and white ears, in contrast with the bright red of +*/b* animals (Ivanovic, 1997). Anemia was also accompanied by a decrease in platelets count and leukocytosis (Ivanovic, 1997).

In the spring of 1967, a colony of these rats was brought to the U.S. and established in New York (Sladic-Simic et al., 1969). It was maintained by crossing Belgrade males with Wistar female rats. Then, the F1 heterozygous females were back-crossed to homozygous anemic males; some rats from the F2 generation exhibited an anemic condition similar to that described in Belgrade rats. The appearance of the phenotype in F2 generations was compatible with a single autosomal recessive character, as previously demonstrated in Belgrade (Sladic-Simic et al., 1969). This genetic factor was designated as *b* and the homozygous anemic Belgrade rats were called *b/b* (**Figure 4**).

It was nearly 30 years after the discovery of these rats that detailed genetic characteristics of Belgrade rats were ascertained. A glycine-to-arginine substitution (G185R) in the fourth TM domain of DMT1 was reported, resulting in loss of activity of the transporter (Fleming et al., 1998). Remarkably, it was the exact same mutation reported for *mk* mice that led to the identification of DMT1 (Fleming et al., 1997). Despite the fact that increased transcript levels are induced by iron deficiency, relative amounts of protein are reduced (Oates et al., 2000; Yeh et al., 2000; Ferguson et al., 2003), and the function of residual DMT1 may be altered (Yeh et al., 2000; Touret et al., 2004; Xu et al., 2004). Work on the Belgrade rat defined a key role for DMT1 not only in intestinal iron absorption, but also in the Tf iron delivery cycle. In fact, Belgrade rats can display high serum iron due to ineffective erythropoiesis (Thompson et al., 2006).

Since the iron imbalance of *b/b* rats became known, many attempts were made to relieve the anemia, extend lifespan and improve husbandry of the Belgrade rat. Iron supplementation by iron-dextran injections was first tested, but although the anemic state was improved, tissue iron deposition was observed (Garrick et al., 1997). Iron supplementation by diet turned out to be more beneficial since it improved anemia without alteration in red cell morphology. Although in both cases iron treatment raised hemoglobin values, levels remained lower than control (Sladic-Simic et al., 1966). Later on, it was reported that the critical nutritional factor would be the source of iron, given that ferrous

iron was more bioavailable than ferric iron (Sladic-Simic et al., 1969; Garrick et al., 1997).

### **CHARACTERISTICS OF BELGRADE RAT ERYTHROID CELLS**

The complexity of iron metabolism in Belgrade rats triggered many questions about the mechanism underlying the impaired utilization of iron. Since the anemia of *b/b* rats resembled thalassemia in humans, first thoughts were focused on possible differences of hemoglobin characteristics. It was demonstrated that reticulocyte globin synthesis was diminished in *b/b* rats although no major imbalance between α- and β-chain production was observed (Edwards et al., 1978). A lack of sideroblasts was also observed in *b/b* rats, indicating that iron was not accumulated within erythroid cells (Edwards et al., 1978). This finding suggested a defect in iron transport into the erythroid cells rather than a defect in intracellular utilization. To consider the possibility that Belgrade rats harbored a defect in heme synthesis, iron incorporation into heme was inhibited. Iron uptake by reticulocytes became diminished with no evidence of intra-erythrocyte iron accumulation in *b/b* rats compared to controls (Edwards et al., 1978). These data suggested that Belgrade anemia was probably not due to a defect in heme synthesis. The same authors were ultimately able to demonstrate that the main defect related to iron utilization in *b/b* rats lies in a significantly lower reticulocyte iron uptake compared to +*/b* controls.

Each step in the iron uptake mechanism by reticulocytes was studied in detail to decipher the underlying cause of Belgrade anemia (Bowen and Morgan, 1987). When receptor interactions were evaluated, Tf binding affinities between *b/b* and control reticulocytes were found to be similar. The Tf molecule itself had the same molecular weight and net charge compared to Wistar rat Tf (Farcich and Morgan, 1992b). Studies on Tf endocytosis showed a slower mechanism in Belgrade reticulocytes, although the relative decrease was not as great as the defect in iron uptake. In addition, Tf exocytosis rate was similar between Belgrade and control reticulocytes, although *b/b* cells released more iron with Tf than control cells (Bowen and Morgan, 1987). In fact, a significant proportion of the iron taken up by *b/b* reticulocytes returned to the extracellular medium, concordant with reduced iron accumulation. These early studies clearly indicated that Belgrade reticulocytes were defective in iron release after uptake by Tf within endocytic vesicles such that if iron was released, it was unable to pass through the endocytic vesicle membrane to the cytoplasm (Bowen and Morgan, 1987; **Figure 2**).

Garrick et al. (1993b) focused on the Tf cycle to more definitively demonstrate that *b/b* reticulocytes retained only half of iron borne by incoming Tf, while around 90% of iron that entered to normal cells remained intracellular. This evidence confirmed that ineffective iron utilization by *b/b* reticulocytes contributes to the Belgrade defect. Furthermore, when heme synthesis was inhibited, iron from Tf failed to accumulate in the stromal (mitochondrial) fraction or in the non-heme cytosolic fraction (Garrick et al., 1993a). These studies reinforced the idea that the Belgrade defect was related to iron release from Tf or its transport from endocytic vesicles. The significantly lower uptake of iron by erythroid precursors most likely contributes to the increased Tf saturation, serum total iron-binding capacity (TIBC) and serum iron levels

reported in *b/b* rats (Thompson et al., 2006). It was also reported that Belgrade reticulocytes have only about half the amount of globin mRNA compared to normal cells, suggesting some sort of a translational defect (Chu et al., 1978). Given that heme regulates mRNA translation in reticulocytes, the failure in this process could be explained because *b/b* reticulocytes contain about 40% of "free" heme compared to +*/b* cells (Garrick et al., 1999b).

Non-Tf-bound iron uptake by erythroid cells was also characterized in Belgrade rats (Garrick et al., 1999a). NTBI uptake differed from Tf–Fe uptake in the pattern of iron distribution between subcellular fractions. Moreover, *b/b* cells only incorporated 20% of the NTBI compared to +*/b* cells, a fraction similar to the residual levels found for Tf–Fe utilization. The results strongly suggested that DMT1 could also play a key role in NTBI acquisition into heme, as well as in iron transport out of endosomes in the reticulocyte Tf cycle. Since the relationship between iron and manganese transport was recognized many years ago, the abnormal iron metabolism seen in erythroid cells of *b/b* rats elicited interest to determine if manganese transport was also affected. Three manganese transport mechanisms have been identified in reticulocytes, one for Mn–Tf and two for the unbound divalent cation Mn2+, one of low and the other of high affinity (Chua et al., 1996). Impaired manganese uptake from Tf and defective import via the high affinity Mn2<sup>+</sup> transport were both observed in Belgrade reticulocytes. These findings strongly supported the idea that these two pathways utilize the same transporter (Chua and Morgan, 1997).

Several studies sought to gain deeper insight into hematopoietic function in Belgrade rats. Homozygous rats display decreased general cellularity in bone marrow than normal rats. For example, early and late erythroid progenitors (BFU-E and CFU-E; Pavlovic-Kentera et al., 1989), granulocyte–monocyte progenitors (Stojanovic et al., 1990), and megakaryocyte progenitors (Rolovic et al., 1991) are significantly diminished. In addition, intensive splenic hematopoiesis has been described in *b/b* rats, indicated by increased iron uptake by spleen, expression of erythroid differentiation markers and elevated erythropoietin serum levels (Pavlovic-Kentera et al., 1989; Biljanovic-Paunovic et al., 1992; Ivanovic, 1997). Indeed, while 42-day-old +*/b* rats have a spleen weight fraction of 0.32 ± 0.1, age- and diet-matched *b/b* rats have a spleen weight fraction of 1.50 ± 0.13 (both values are % body weight; *n* = 3–4; *P* < 0.05; J. Kim and M. Wessling-Resnick, personal observations).

### **DUODENAL ABSORPTION OF IRON AND MANGANESE**

Early studies by Morgan's group (Farcich and Morgan, 1992a) demonstrated reduced intestinal iron absorption by Belgrade rats, a key finding supported by later work from several groups under different experimental conditions (Knopfel et al., 2005b; Thompson et al., 2007a). Since duodenal iron absorption is highly regulated by iron status, Morgan and colleague also characterized absorption by Belgrade rats fed diets of normal, low, and high iron content (Oates and Morgan, 1996). While duodenal iron uptake in control +*/b* rats varied inversely with iron intake, *b/b* rats failed to show changes under iron loading or iron deficiency. It was later shown that a bolus of dietary iron could induce a rapid decrease in intestinal DMT1 mRNA in +*/b* and *b/b* rats (Yeh et al., 2000).

In addition, this study showed that *b/b* rats had generally higher basal protein levels of DMT1. It was also observed that *b/b* rats have lower duodenal DMT1 protein than expected based on the higher mRNA levels, possibly due to impaired release of the mutant protein from its site of synthesis or accelerated degradation (Morgan and Oates, 2002). Finally, TfR gene expression and Tf-bound iron uptake by duodenal enterocytes also has been investigated in the Belgrade rat. TfR1 mRNA expression in *b/b* epithelial cells of the crypt region and crypt–villus junction and 125I-labeled Tf uptake in *b/b* villus cells were both similar to Wistar rats, while uptake was significantly greater in *b/b* crypt cells (Oates et al., 2000). These observations suggest that uptake of Tf by enterocytes is largely independent of DMT1's activity, while it remains unknown if iron delivery by Tf to these intestinal cells might be affected in the Belgrade rat.

Molecular studies on the G185R mutant of DMT1 have shown that transfected cells will target the protein product to the plasma membrane and endosomes although levels are lower than wildtype at the cell surface (Su et al., 1998; Touret et al., 2004). These studies have suggested the mutant has reduced residual activity. Whether the Belgrade rat survives on this reduced activity, or if some other mechanism compensates for iron absorption has been studied (Yeh et al., 2011). Gene expression studies showed HIF-2α expression was increased in *b/b* rats compared with +/+ rats with greater expression in the villus compared to crypt, a response that correlated with the presence of hypoxic protein adducts. Under hypoxic conditions, compounds as nitroimidazole or some of its derivatives, enter viable cells and interact with thiol groups of intracellular proteins thus forming the adducts. Under normal oxygen levels the compounds are reoxidized and diffuse out of the cells.

Moreover, most of the genes whose protein products are responsible for duodenal iron transport were up-regulated. One factor that did not increase was ZIP14, suggesting that it is unlikely to compensate for iron absorption (Yeh et al., 2011). There may be other parallel uptake pathways for iron, however. This idea is supported by studies of neonatal iron assimilation in Belgrade pups. After administration of 59Fe to lactating foster dams, total levels of assimilated 59Fe between suckling *b/b* and <sup>+</sup>*/b* pups were not different. However, the examination of blood compartments in *b/b* pups showed elevated iron levels in serum and reduced levels in RBCs compared to +*/b* siblings. Tissue iron distribution was significantly higher in heart, kidney, liver, spleen, and intestine in homozygous pups (Thompson et al., 2007a). Thus, during lactation iron absorption occurs normally, but delivery to red cells is impaired in *b/b* rats causing apparent iron loading in other peripheral tissues. Later on, iron absorption is impaired in adult Belgrade rats as discussed above. Thus, DMT1 does not appear to play a significant role in iron assimilation during lactation, but a developmental transition to DMT1-mediated iron uptake occurs as food intake begins. What factors mediate iron absorption in early development remain to be better characterized, but may provide some clues to additional mechanisms that could compensate for the intestinal uptake defect in the Belgrade rat.

Belgrade rats have also been used to study duodenal transport of manganese and the role of DMT1. Chua and Morgan (1997) evaluated manganese absorption from the intestine using closed *in situ* loops of duodenum and distinguishing between uptake, transfer, and absorption. Manganese uptake was decreased in *b/b* rats compared to +*/b* rats. Although manganese transfer from the duodenum into the carcass was similar in both genotypes, the percentage of absorption in homozygous rats was significantly low. Hence, the primary defect appears to be the absorption step responsible for the uptake of manganese from the gut lumen implicating DMT1 in this process. Indeed, studies from our own laboratory have indicated that Belgrade rats have lower blood manganese levels, supporting the idea that they have manganese as well as iron deficiency. Levels of manganese are 14.6 ± 2.3 versus 8.1 ± 1.2 ng/g blood in +*/b* versus *b/b* rats (±SEM, *n* = 7, *P* = 0.0029; J. Kim and M. Wessling-Resnick, personal observations).

#### **HEPATIC HANDLING OF IRON AND MANGANESE**

Iron accumulation in tissues of *b/b* rats after parenteral iron administration was reported early on in the characterization of the Belgrade phenotype (Sladic-Simic et al., 1969). In fact, this feature defect resembled thalassemia, an iron-loading anemia. When the first human mutations in DMT1 were discovered (Mims et al., 2005; Beaumont et al., 2006; Lam-Yuk-Tseung et al., 2006), patients were found to suffer not only from anemia but also from hepatic iron overload. It was suggested that rodent models of DMT1 deficiency did not display iron overload since heme iron was not present in chow, and that the human condition might be caused up-regulation of heme absorption. Our laboratory investigated this issue in the Belgrade rat model (Thompson et al., 2006). First, female +*/b* rats crossed with male *b/b* rats were fed an iron-supplemented diet (500 ppm) to support pregnancy. After birth, litters were cross-fostered to F344 Fischer dams fed a standard diet, and upon weaning both +*/b* and *b/b* pups were fed an iron-supplemented diet 3 weeks. To control for the anemic status of the Belgrade rats, +*/b* rats are fed a low iron diet (5 ppm) under our laboratory's husbandry protocol (**Figure 5**). Despite the anemic state of *b/b* rats, liver non-heme iron content was greater compared with age-matched (and diet-matched) +*/b* sibling controls. Perl's Prussian blue staining showed iron deposition was evident in both periportal and centrilobular zones. In contrast, no iron staining was observed in age-matched +*/b* rats (Thompson et al., 2006). This pattern of liver iron deposition suggested that the primary defect in erythron iron utilization seen in homozygous rats leads to liver iron loading. Consistent with this idea, the Belgrade rats also display higher serum iron levels (Kim et al., 2013). These data argue that the liver can acquire iron independent of DMT1, in agreement with previous studies that showed iron-loading in DMT1 knockout mice (Gunshin et al., 2005). The high hepatic expression of Zip14 – which is up-regulated by high iron (Nam et al., 2013) – raises the likelihood that this transporter is responsible. Interestingly, quantitative real-time RT-PCR analysis showed hepcidin expression was threefold higher in *b/b* compared to +*/b* littermates (Thompson et al., 2006), consistent with studies showing that hepcidin expression increases with liver iron loading despite severe anemia (Vokurka et al., 2006). Yeh et al. (2011) have also reported hepcidin expression increases in iron-fed *b/b* rats. Thus, the Belgrade rat's inability to take up adequate iron might be compounded

by hepcidin's down-regulation of ferroportin, the basolateral iron exporter. Recent studies in mice with the Dmt1 gene selectively inactivated in hepatocytes reinforce the idea that hepatic DMT1 is dispensable for NTBI uptake, although they also showed unaffected hepatic iron levels in these mice, suggesting that DMT1 is also not essential for hepatic iron accumulation (Wang and Knutson, 2013).

Manganese as well as iron content is altered in liver of Belgrade rats. It has been reported that the concentration and total content of manganese in liver is significantly less in *b/b* rats compared to +*/b* rats (Chua and Morgan, 1997). As discussed above, intestinal manganese absorption is diminished in Belgrade rats. Interestingly, a significantly higher uptake of 54Mn into liver was observed when it is administered as Mn–Tf or Mn–serum (Chua and Morgan, 1997). The difference between iron and manganese metabolism is that while iron is retained, and therefore might accumulate due to ineffective erythroid uptake, excess manganese is rapidly cleared by the liver (Papavasiliou et al., 1966). The fact that hepatic uptake is increased suggests the liver import mechanisms of these two metals once again overlap: up-regulation of Zip14 or another transporter might induce this effect.

## **RENAL IRON HANDLING**

Despite the high renal expression of DMT1, we only have a minimal understanding of its function in this tissue. Studies of the renal physiology of Belgrade rats have helped to shed some light. Analysis of renal DMT1 mRNA showed the same transcript size in *b/b* and +*/b* rats, without changes in levels of mRNA expression between genotypes (Ferguson et al., 2003). Expression of DMT1 protein is reduced in the kidneys, but this might be expected since protein levels do not generally reflect transcript levels. Reduced immunostaining for DMT1 was not specific for a region, since it is observed in proximal, distal, and collecting tubules. Once again, these observations indicate that the mutation may accelerate DMT1 degradation or cause defective posttranslational processing (Ferguson et al., 2003).

General tissue disorganization and abnormal morphology of cortical tubules of the Belgrade rat kidney was first noted by Ferguson et al. (2003). A later study from our group described glomerulosclerosis and interstitial sclerosis in *b/b* rats upon aging, with fibrosis in glomeruli and areas of tubulointerstitial fibrosis (Veuthey et al., 2013). Tubular dilation with flattened epithelium in some cortical tubules and occlusion of the luminal space in other cases was observed. Although serum creatinine appears normal in *b/b* rats, creatinine clearance was significantly reduced suggesting a decrease in the glomerular filtration rate. Moreover, elevated urinary albumin in *b/b* compared to +*/b* rats has been reported by us (Veuthey et al., 2013), suggesting damage in the glomerular membrane (Satchell and Tooke, 2008). Since albumin is normally reabsorbed in the tubular system, the evidence suggests that the increased tubular burden of albumin could be associated with progressive interstitial fibrosis and tubular damage (Eddy, 1994; Jerums et al., 1997; Wang and Hirschberg, 2000).

A detailed measurement of several electrolytes has been carried out in feces, serum, and urine of Belgrade rats (Ferguson et al., 2003). Fecal excretion of almost all tested ions was similar between *b/b* and <sup>+</sup>*/b* animals, except for Fe2+, which was higher in *b/b* rats. In addition, serum levels of electrolytes showed high Mg2<sup>+</sup> and decreased K<sup>+</sup> in *b/b* compared to +*/b* rats. Urinary analysis revealed higher Ca2<sup>+</sup> levels but without changes in Fe levels. This last finding differs from studies of the Belgrade rat in our lab that show higher urinary iron output; the age of rats that were studied could be one factor accounting for this difference since kidney function worsens with age in Belgrade rats (Jia et al., 2013;Veuthey et al., 2013). The fact that more iron is excreted in *b/b* rat urine could result from the altered glomerular membrane function, but also from loss of tubular reabsorption. Evidence suggests that both pathways could be affected in *b/b* rats (Veuthey et al., 2013). Since tubular uptake of Fe3<sup>+</sup> bound to Tf seems to be mediated by cubilin, the high urinary Tf seen in *b/b* together with the absence of changes in cubilin suggest that tubular reabsorption of Tf is not affected.

Another interesting finding related to the renal pathology was the premature death of homozygous Belgrade rats. We observed that several early urinary biomarkers of renal injury were altered *b/b* rats, pointing to a kidney defect (Veuthey et al., 2013). The evaluation of early renal development in *b/b* pups by the radial glomerular count (RGC) method supports the idea that limited iron supply during early life could affect renal development in adults leading to injury and even death from renal failure (Drake et al., 2009; Veuthey et al., 2013). Maternal iron restriction during pregnancy has been previously documented to induce altered renal morphology in adult offspring (Lisle et al., 2003). Nephron number is set early in life and does not increase (Al-Awqati and Preisig, 1999). Our RGC study of nephrogenesis indicated that *b/b* pups have decreased nephron allotment. Brenner et al. (1996) have put forward the hyperfiltration hypothesis that these early development defects explain compromised renal metabolism observed in adulthood. Adaptive activity of remnant nephrons would need to maintain glomerular filtration. Over time, increased glomerular pressure promotes fibrosis and sclerosis to produce glomerular injury. Such injury leads to further nephron loss, thereby continuing a vicious cycle that finally decreases the glomerular filtration, ending in renal damage and poor kidney function.

#### **RESPIRATORY AND OLFACTORY UPTAKE**

Due to the nature of its transport and metabolism, airborne manganese promotes neurotoxicity upon its distribution to the brain. Inhalation exposures to manganese pose a significant occupational health risk to welders, for example, who are exposed to fumes containing iron, chromium, manganese, aluminum, nickel, and cadmium. Within these mixtures are multiple transport substrates for DMT1. Our laboratory was interested in the potential role DMT1 played in this process due to the fact this pathway would be up-regulated during iron deficiency. "Iron-responsive manganese uptake" would exacerbate the neurotoxicity of airborne metal. Here, the Belgrade rat was used as a model system, with cohorts of homozygous, heterozygous, and anemic heterozygous controls (**Figure 5**). Uptake of intranasally instilled 54Mn was markedly reduced in *b/b* rats compared to iron-replete +*/b* rats (Thompson et al., 2007b). Enhanced 54Mn absorption was observed in irondeficient +*/b* controls relative to both *b/b* and iron-replete +/b rats. In contrast, 54Mn clearance from blood to peripheral tissues showed the same pharmacokinetics for intravenously injected *b/b*, +*/b* and iron-deficient +*/b* rats. The sum of this pharmacokinetic data supports a functional role for uptake of DMT1 in manganese uptake by the olfactory pathway. Immunohistochemistry revealed that DMT1 was associated with the microvilli of the olfactory epithelium and the endfeet of olfactory epithelial sustentacular cells. Most importantly, DMT1 levels in the olfactory epithelium were significantly greater in iron-deficient rats. Finally, the fact that total levels of Mn in brain of intranasally instilled rats were 10-fold higher than in intravenous injected animals directly demonstrates that inhalation promotes greater brain manganese uptake. Thus, the apparent function of DMT1 in olfactory Mn absorption suggests that manganese neurotoxicity can be modified by iron status due to the iron-responsive regulation of DMT1. Subsequent studies of the Belgrade rat also showed that lack of DMT1 can affect the absorption of iron from the nasal cavity to the blood and finally

to the brain (Ruvin Kumara and Wessling-Resnick, 2012), further localizing iron-regulated DMT1 expression in the olfactory bulb, too.

The Belgrade rat also has been used to investigate whether DMT1 plays a role in uptake across the pulmonary epithelium of the lungs. Previous investigation by others showed the Belgrade rat has reduced clearance of iron (Wang et al., 2002). In our intratracheal instillation experiments, the transport of 54Mn to the blood was unaltered (Brain et al., 2006). However, studies of the Belgrade rat do suggest that DMT1 is involved in pulmonary inflammation (Wang et al., 2005; Kim et al., 2011) as well as metalinduced injury (Ghio et al.,2005,2007), emphasizing an important detoxification pathway that can transport and sequester metals in the lungs after inhalation exposures. Using the Belgrade rat model, Ghio and colleagues demonstrated that ozone-induced lung injury was dependent on DMT1 and that the transporter could modify oxidative stress responses. Lipopolysaccharide (LPS) also has been shown to induce DMT1 (Nguyen et al., 2006), supporting that idea that under infection and inflammation, it functions to take up and sequester metals as a protective host response. The net effect of such alterations would be to reduce levels of manganese available in the lung to limit survival of pathogens like *S. pneumonia* and to help protect the lungs from inflammation due to air pollution and other irritants.

#### **THE BLOOD–BRAIN BARRIER**

Brain metal homeostasis is critical to cognitive development, behavior, and motor control. Excess iron and manganese are also associated with neurodegeneration, for example, in Alzheimer's and Parkinson's disease patients. Belgrade rats have been used to study transport of iron and manganese across the blood– brain barrier. Iron content in brain seems to vary depending on the age of *b/b* rats. Similar levels were reported for *b/b* and +*/b* pups, while significantly lower levels were seen in 21-dayold *b/b* rats compared to +*/b* age-matched controls (Thompson et al., 2007a). Burdo et al. (1999) performed a detailed analysis of iron distribution in brain of Belgrade rats. Cortical gray matter showed iron-positive astrocytes in brain of both *b/b* and +*/b* rats, although fewer cells were stained in the Belgrade rat cohort. Iron was also observed in pyramidal neurons, but they were fewer in number and less intensely stained than in +*/b* rats. In white matter of +*/b* rats, iron was present in patches of intensely iron-stained oligodendrocytes and myelin, while the same pattern of expression was seen but with dramatically less intensity in homozygous rats. In addition, the stained oligodendrocytes were associated with blood vessels. The general decrease in iron staining observed in *b/b* rats compared to +*/b* is consistent with the decreased Tf and iron uptake into the brain previously reported for Belgrade rats (Farcich and Morgan, 1992a). Loss of neuronal iron staining in brain of *b/b* agrees with the finding that DMT1 mRNA is mainly expressed in these cells (Gunshin et al., 1997).

Early in 1992, impaired iron uptake by *b/b* rat brain was reported, although at that time the details of this mechanism were not clear (Farcich and Morgan, 1992a). Later on, it was established brain uptake of Tf-bound iron was decreased in young and adult *b/b* rats compared to +*/b* control rats, while Tf uptake was similar between both cohorts (Moos and Morgan, 2004). These authors also reported significantly lower iron content in brain of *b/b* rats together with higher expression of neuronal TfR1, confirming the iron-deficient stage of homozygous rats. By looking at the brain capillary endothelial cells (BCECs) involved in the transport trough the blood–brain barrier, TfR expression appeared to be identical. DMT1 was mainly detected in the cytoplasm of neurons and choroid plexus epithelial cells, but was not detected in BCECs (Moos and Morgan, 2004).

Based in that study, it has been proposed that neurons could acquire iron by receptor-mediated endocytosis of Tf, followed by iron transport out of endosomes mediated by DMT1. In Belgrade rats, the mutation on DMT1 could explain the low cerebral iron uptake, suggesting that it is due to a reduced neuronal uptake rather than an impaired transport trough the blood–brain barrier. For manganese, however, the story is much less clear. Uptake of intravenously injected 54Mn into the brain was similar for *b/b* and +*/b* rats, while iron-deficient +*/b* controls took up less. Since presumably both TfR and DMT1 would be up-regulated in the latter cohort, these data suggest the pathway for manganese may be different than iron uptake into the brain. This evidence agrees with previous reports that DMT1 is not involved in the blood–brain transport of manganese (Crossgrove and Yokel, 2004).

## **LIPID METABOLISM**

Belgrade rats also exhibit pathological changes in lipid metabolism. Homozygous *b/b* rats display hypertriglyceridemia and elevated free fatty acids compared to +*/b* rats, without significant changes in cholesterol levels (Kim et al., 2013). Lipoprotein triglycerides (TGs) are associated with very low density lipoprotein (VLDL). Since hepatic TG levels were similar in *b/b* and +*/b* rats, higher serum TG levels in *b/b* rats appear to be unrelated to increased production, an idea supported by the fact that hepatic lipogenic gene expression is not affected. Instead, all of the evidence points to a block in TG uptake, which can be explained by the fact that serum lipoprotein lipase (LPL) activity is significantly reduced in *b/b* rats compared to +*/b* age-matched controls (Kim et al., 2013). LPL is the key enzyme responsible for release of TG lipids for uptake after VLDL binds to its receptor. The fact that VLDL receptor levels in muscle [and lowdensity lipoprotein (LDL) receptor levels in liver] are similar in *b/b* and +*/b* rats supports the idea that LPL inhibition accounts for hypertriglyceridemia. The unusual iron-loading anemia of the Belgrade rat suggested an interaction with high serum iron. Indeed, serum LPL activity is also reduced in rats with dietary iron loading, confirming the fact that increased serum iron is associated with decreased LPL activity. Based on these data, our laboratory studied how iron alters serum LPL *ex vivo* or recombinant LPL *in vitro*, finding that exogenously added iron inhibits this enzyme in a dose-dependent manner. These independent lines of evidence suggests that elevated serum iron levels of *b/b* rats promote reduced TG clearance due to LPL inhibition, resulting in higher levels of serum TG. Although the molecular basis for iron-mediated regulation of LPL activity is not clear, oxidative modification of the enzyme, its substrate, and/or reaction products could interfere with lipolysis (Kim et al., 2013). Clinically, reduction of iron loading in patients by phlebotomy (Casanova-Esteban et al., 2011), chelation (Cutler, 1989) and diet (Cooksey et al., 2010) can help to improve lipid disorders. We also found that reducing serum iron in Belgrade rats by treatment with the uptake inhibitor ferristatin II improved their TG levels, suggesting pharmacological interventions could be helpful (Kim et al., 2013).

#### **GLUCOSE METABOLISM**

Despite their hyperferremia, Belgrade rats display normal insulin and glucose tolerance (Jia et al., 2013). Comparable levels of insulin-induced Akt phosphorylation in *b/b* and +*/b* rats suggests that downstream insulin signaling is also unaffected. Moreover, insulin secretory capacity of the pancreas showed no significant differences between *b/b* and +*/b* rats, although pancreatic nonheme iron levels were >5-fold higher in Belgrade rats. Histological evaluation suggests an absence of tissue damage despite the known association between high serum iron and tissue damage (Inoue et al., 1997). The sum of these data support that loss of DMT1 protects pancreatic β-cells and helps to maintain insulin sensitivity despite iron overload (Jia et al., 2013). These results correlate very well with studies of the β-cell specific *Dmt1* knockout mouse, which is resistant to diabetes (Hansen et al., 2012).

An unusual characteristic of Belgrade rats is that they consume more food that heterozygous littermates but display lower body weight, suggesting an apparent imbalance in energy metabolism (Ferguson et al., 2003; Jia et al., 2013). That unexpected phenotype has been associated, at least in part, with increased urinary glucose excretion. Since the kidneys play an important role in glucose homeostasis through tubular reabsorption, the glycosuria observed could be explained by the abnormalities reported in kidney histology and physiology discussed above (Jia et al., 2013; Veuthey et al., 2013). Collectively, these surprising findings reveal that lack of DMT1 function could have an unexpected and significant role in energy balance.

## **TRANSPORT OF OTHER METALS**

As its name implies, DMT1 not only services iron and manganese metabolism, but it is also involved in uptake of other metals. Knopfel et al. (2005b) isolated brush border membrane (BBM) vesicles to study nickel transport in Belgrade rats. This group measured active nickel transport in BBM vesicles from +*/b* rats but not *b/b* rats. This result implies that the Belgrade mutation disables the transport capacity for this metal. The potential role of DMT1 in copper transport also has been evaluated using the same approach. When unenergized vesicles were utilized, transport of copper was disrupted in BBM vesicles isolated from *b/b* rats, while +*/b* vesicles showed normal copper transport (Knopfel et al., 2005a). Interestingly, when ATP-loaded or energized vesicles were studied, transport characteristics for this metal appeared to be identical in *b/b* and +*/b* rats. It was suggested that ATP-driven copper uptake is a principal copper transport mechanism, while DMT1 might act under conditions of copper excess. Since Belgrade rats are not copper deficient despite DMT1 mutation and iron deficiency, this evidence agrees with previous reports suggesting the existence of several copper transporters other than DMT1 in the intestinal BBM (Knopfel et al., 2005a).

However, the influence of copper in iron metabolism, and reciprocally, the influence of iron status on copper metabolism, presents an interesting and compelling relationship (Yokoi et al., 1991; Ece et al., 1997). Induction of the duodenal Menkes copper ATPase (Atp7a) and metallothionein (Mt1a) have been described in iron-deficient rats, together with high serum and hepatic copper levels and increased ceruloplasmin (Collins et al., 2005). Whether or not DMT1 could contribute to copper uptake remains controversial, with evidence for (Arredondo et al., 2003; Espinoza et al., 2012) and against (Illing et al., 2012; Shawki et al., 2012). Moreover, the valency of copper as a potential transport substrate (Cu+ versus Cu2+) must be considered. Jiang et al. (2013) addressed some of these questions using the Belgrade rat model by comparing *b/b* rats with diet-matched +*/b* and anemic +*/b* littermate controls (**Figure 5**, for example). Previous work by this group showed Belgrade rats did not display the expected changes in hepatic or serum copper upon iron deficiency, suggesting DMT1 might play a role in these processes (Jiang et al., 2011). To test this more directly, an everted gut sac transport assay was used to assess Cu<sup>+</sup> transport. Increased uptake in iron-deficient +*/b* control rats was observed while *b/b* and diet-matched +*/b* controls had similar levels of uptake. Thus, DMT1 could play a role in copper uptake under iron-deficiency conditions.

Interestingly, DMT1 demonstrates a stronger selectivity for Cd2<sup>+</sup> than Fe2<sup>+</sup> or other more physiologically relevant metals (Illing et al., 2012). This toxic metal is absorbed by the intestine and other tissues and can exert nephrotoxicity. The notion that DMT1 might play a role in Cd distribution and uptake remains to be tested – and the Belgrade rat may provide an excellent model system to explore this possibility.

## **CONCLUSION**

Although murine models are perhaps more often used to study mammalian metabolism, rat models have provided critical information about obesity (Zucker fatty rat), diabetes (Wistar fatty rat; Otsuka Long–Evans Tokushima fatty rats; Goto-Kakizaki rat); hypertension (Spontaneously hypertensive rat; Dahl salt-sensitive rat), copper metabolism (Long–Evans cinnamon rat), and renal metabolism (ZSF1 rat).We have outlined multiple examples where the Belgrade rat, as a model of iron deficiency, has been useful in characterizing not only the role of DMT1 in transport of this metal, but also its contribution to pathologies of intermediary metabolism, its protective role in detoxification of the lungs, its participation in neurotoxicity of airborne metal uptake by the olfactory pathway, in the development of the kidneys, in promoting altered renal function, in brain iron metabolism and in hepatic iron handling. While the Belgrade rat is the first known rat model of inherited anemia, recently Bartnikas et al. (2013) described a rat model of hereditary hemochromatosis. This group suggests that rats homozygous for a novel Ala679Gly allele of the TfR2 gene spontaneously load iron, thus recapitulating the human disease. Although many murine models of hemochromatosis exist, none exhibit fibrosis of the liver, a debilitating aspect in human patients with this disease. This opens the exciting possibility to explore new metabolic features associated with inherited iron overload with the potential to explore therapeutic avenues that will ameliorate the disease. Moreover, as genetic manipulation of the rat becomes more routinely available, future models may contribute a better understanding of the genes of iron metabolism – in both iron deficiency and overload – and their contributions to mammalian physiology and pathology.

## **ACKNOWLEDGMENTS**

We gratefully acknowledge the support and generous time and effort of lab members who contributed to our work on the Belgrade rat, including Khristy Thompson, Elizabeth Heilig, Jonghan Kim, Dorathy Vargas, and Xuming Jia. The Belgrade rats establishing our colony were kindly provided by Drs. Michael and Laura Garrick, who have shepherded the field and maintained this unusual animal model for so many years. We thank them for their helpful advice and guidance. Marianne Wessling-Resnick is supported by the NIEHS and NIDDK of the National Institutes of Health under awards R01ES014638 and R01DK064750.

## **REFERENCES**


*Physiol.* 178, 349–358. doi: 10.1002/(SICI)1097-4652(199903)178:3<349::AID-JCP9>3.0.CO;2-R


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

*Received: 18 February 2014; paper pending published: 10 March 2014; accepted: 02 April 2014; published online: 22 April 2014.*

*Citation: Veuthey T and Wessling-Resnick M (2014) Pathophysiology of the Belgrade rat. Front. Pharmacol. 5:82. doi: 10.3389/fphar.2014.00082*

*This article was submitted to Drug Metabolism and Transport, a section of the journal Frontiers in Pharmacology.*

*Copyright © 2014 Veuthey and Wessling-Resnick. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited andthatthe original publication inthis journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*

## Special delivery: distributing iron in the cytosol of mammalian cells

## *Caroline C. Philpott\* and Moon-Suhn Ryu*

Genetics and Metabolism Section, Liver Diseases Branch, National Institute of Diabetes and Digestive and Kidney Diseases, National Institutes of Health, Bethesda, MD, USA

#### *Edited by:*

Paolo Arosio, University of Brescia, Italy

#### *Reviewed by:*

Constantin Ion Mircioiu, Carol Davila "University of Medicine and Pharmacy", Romania Paolo Arosio, University of Brescia, Italy

#### *\*Correspondence:*

Caroline C. Philpott, Genetics and Metabolism Section, Liver Diseases Branch, National Institute of Diabetes and Digestive and Kidney Diseases, National Institutes of Health, Building 10, Room 9B-16, 10 Center Drive, Bethesda, MD 20892-1800, USA e-mail: carolinep@intra.niddk.nih.gov

Eukaryotic cells contain hundreds of proteins that require iron cofactors for activity. These iron enzymes are located in essentially every subcellular compartment; thus, iron cofactors must travel to every compartment in the cell. Iron cofactors exist in three basic forms: Heme, iron–sulfur clusters, and simple iron ions (also called non-heme iron). Iron ions taken up by the cell initially enter a kinetically labile, exchangeable pool that is referred to as the labile iron pool. The majority of the iron in this pool is delivered to mitochondria, where it is incorporated into heme and iron–sulfur clusters, as well as non-heme iron enzymes. These cofactors must then be distributed to nascent proteins in the mitochondria, cytosol, and membrane-bound organelles. Emerging evidence suggests that specific systems exist for the distribution of iron cofactors within the cell. These systems include membrane transporters, protein chaperones, specialized carriers, and small molecules. This review focuses on the distribution of iron ions in the cytosol and will highlight differences between the iron distribution systems of simple eukaryotes and mammalian cells.

**Keywords: non-heme iron, diiron, glutathione, metallochaperone, iron chaperone, glutaredoxin, labile iron pool**

One of the most important concepts to emerge from the field of eukaryotic cell biology is that the contents of the cell are highly organized and the movement of organelles and macromolecules within the cell does not occur through simple diffusion. Movement is highly regulated through packaging and assembly of complexes and through directed transport via components of the cytoskeleton. Cells express hundreds, perhaps thousands of proteins that require bound metal cofactors for activity and stability (Waldron et al., 2009). Iron and zinc are the most abundant metals in cells, followed by copper and manganese, and, to a lesser extent, cobalt, nickel, and molybdenum. Thus, it is not surprising to discover the existence of intracellular systems for the distribution of metals and metal cofactors. Why do cells need these metal delivery systems? It is because the cell faces several obstacles in achieving the two major goals of metal cofactor delivery, which are incorporation of the native cofactor and exclusion of non-native cofactors.

A major obstacle to acquisition of metal cofactors by apoenzymes is that the binding sitesfor these cofactorsfrequently lack the capacity to discriminate between different divalent metal cations and cofactors (Kasampalidis et al., 2007; Waldron et al., 2009). The sulfur, oxygen, and nitrogen ligands that coordinate metals in enzymes will frequently bind non-native cofactors with affinities equal to or greater than those of the native cofactor. A second obstacle is that redox-active metals, such as iron, copper, and manganese can engage in Fenton-type chemistry in the presence of oxygen and produce potentially damaging reactive oxygen species. Thus, cells must tightly regulate the uptake, storage, and distribution of metal ion species. A third obstacle is that zinc and copper ions exhibit the highest affinity for transition metal-binding sites and would occupy iron and

manganese sites if the metals were present in freely exchangeable pools of similar concentrations (Irving and Williams, 1953). Consequently, cells maintain pools of zinc and copper at exceedingly low levels (Rae et al., 1999; Outten and O'Halloran, 2001). This is accomplished through tightly regulated uptake and efflux and through the expression of metal-binding proteins, such as metallothioneins, that effectively sequester pools of zinc and copper. Iron, however, appears to be managed differently.

## **THE FATE OF INTRACELLULAR IRON**

In mammals, iron enters the cell through a variety of transport systems, including the endosomal transporter DMT1 (**Figure 1**) (reviewed in Shawki et al., 2012), which receives transferrinbound iron, and Zip14, which can take up both endosomal transferrin- and extracellular non-transferrin-bound iron (Liuzzi et al., 2006; Zhao et al., 2010). The majority of iron taken up through membrane transporters enters a metabolically active pool in the cytosol (Shvartsman and Ioav Cabantchik, 2012), although a small amount may be transferred directly from endosomes to mitochondria (Sheftel et al., 2007). The molecular nature of this cytosolic pool is largely unknown, but operationally it is defined as the freely exchangeable iron that is loosely coordinated by water, small molecules, and proteins, and has been termed the labile iron pool (LIP). Most of the iron entering the LIP is directed to mitochondria (Shvartsman and Ioav Cabantchik, 2012), where it can be incorporated into mononuclear or dinuclear iron centers, inserted into protoporphyrin IX to form heme, or used in the assembly of iron–sulfur clusters. Mitochondrial heme and iron– sulfur clusters are bound by enzymes within mitochondria, but heme and some product of the Fe–S cluster assembly machinery

and BolA2 coordinate [2Fe-2S] clusters with the help of GSH. These clusters may be used for the metallation of cytosolic Fe-S enzymes (via the cytosolic iron–sulfur cluster assembly machinery, CIA) and may have a role in mitochondrial iron delivery in erythroid tissues. PCBP1 and PCBP2

coordinated by GSH or iron coordinated by Grx3. PCBP2 may directly acquire iron from transporters such as DMT1 and may also deliver iron to ferroportin for efflux. Some intracellular organelles have been omitted for clarity. Red spheres indicate inorganic sulfide as part of an Fe–S cluster.

are also exported from mitochondria for utilization in the cytosol, nucleus, and other membrane-bound organelles. Cytosolic iron that is not directed to mitochondria may be used to metallate non-heme iron enzymes of the cytosol and nucleus, to assemble cytosolic Fe–S clusters, or be stored and sequestered in ferritin. The various fates of cytosolic iron are likely determined by the activity of iron transporters, such as the mitochondrial iron importers (Mitoferrin 1 and Mitoferrin 2, Shaw et al., 2006) and iron efflux pumps on the lysosomal or plasma membrane (e.g., ferroportin, Abboud and Haile, 2000; Donovan et al., 2000; McKie et al., 2000). Iron storage in ferritin is largely determined by the level of ferritin protein, which is tightly regulated according to cellular iron levels (reviewed in Anderson et al., 2012). The partitioning of iron to iron-binding small molecules and proteins of the LIP and proteins specialized for the delivery of iron cofactors to recipient apoenzymes play important roles in determining the fate of cytosolic iron.

## **COMPOSITION OF THE CYTOSOLIC LABILE IRON POOL**

From a conceptual standpoint, the cytosol must contain a kinetically labile, metabolically accessible pool of iron so that iron can

be distributed to its various sites of utilization, transport, or storage (reviewed in Hider and Kong, 2013). Technical hurdles have hampered the study of the LIP largely because of the need to study the pool in intact, living cells. Breaking open cells, diluting their contents, and exposing the contents to oxygen will greatly alter the character of the LIP. From a practical standpoint, the measurable LIP is defined as the pool of iron that is accessible to weak iron chelators introduced to the cytosol (Petrat et al.,2002). Fluorescent iron chelators are used to measure the LIP (Esposito et al., 2002). Chemically, these compounds are composed of a fluorescein moiety conjugated to an Fe(II)- or Fe(III)-binding moiety. These chelators are introduced to cells as a lipophilic, esterified precursor, which is then hydrolyzed and trapped in the cytosol. Iron binding quenches fluorescence; when a cell-permeable, strong iron chelator is then added, iron shifts from the fluorescent chelator to the strong chelator and fluorescence is regained. The ratio of fluorescence after and before addition of the strong iron chelator provides an estimate of the chelatable iron pool. Using this approach, the cytosolic LIP has been measured with calcein and found to be 0.3–1.6 μM, with different cell types exhibiting different baseline levels of the LIP and different responses to iron-supplemented or

iron-depleted medium. Measurements using Phen Green-SK suggest higher levels for the cytosolic LIP, which is consistent with the much higher affinity of Phen Green-SK for Fe (II). Measurements with more iron-specific fluorescent chelators also indicate the LIP to be in the low micromolar range. Intracellular iron species in whole cells have also been measured using spectroscopic methods. Human Jurkat cells grown in iron-supplemented medium were found to contain ∼30 μM high-spin Fe(II) in the cytosol, a portion of which likely represents the LIP, with the remainder coordinated by non-heme iron enzymes (Jhurry et al., 2013).

Attempts have been made to characterize the molecular composition of the cytosolic LIP and some progress has been made. Several approaches based upon the relative binding of iron species to Fe(II)- and Fe(III)-specific chelators before and after treatment with reductants or oxidants indicate that >80–90% of the cytosolic LIP is in the reduced state (Egyed and Saltman, 1984; Breuer et al., 1995). Using Mossbauer and electron paramagnetic resonance spectroscopy, the cytosolic iron pool of mammalian cells was found to contain high-spin Fe(III) signals typical of ferritin and Fe(II) signals in the form of heme and non-heme iron (Jhurry et al., 2013). This predominance of Fe(II) is consistent with the presence of excess reduced glutathione and NADPH in the cytosol, both of which can act as physiologic iron reductants at the pH and ionic strength of the cytosol. Furthermore, all known metazoan iron uptake systems are specific for Fe(II). Thus, the iron initially entering the cytosol is also in the reduced state.

## **GLUTATHIONE AS A MAJOR CYTOSOLIC LIGAND FOR Fe(II)**

Because the cytosolic LIP is essentially present as Fe(II), the physiologic ligands for the cytosolic LIP must have affinity for Fe(II). Many small molecules proposed as physiologic ligands are unlikely to form complexes with cytosolic iron because they form stable complexes with Fe(III) and not Fe(II). These would include ATP/AMP, free amino acids, inositol triphosphates (Veiga et al., 2009), and 2,5 dihydroxybenzoic acid (Devireddy et al., 2010; Shvartsman and Ioav Cabantchik, 2012). Candidates for physiologic Fe(II) ligands include citrate, cysteine, and reduced glutathione (GSH). Analysis of these potential ligands *in vitro* suggests that Fe–GSH complexes are likely the only species that forms at significant levels in cells, largely because GSH is present at relatively high (2–10 mM) levels in the cytosol and exhibits moderate affinity (*K*d∼8 μM) for Fe(II) (Hider and Kong, 2011). Other ligands that demonstrate strong affinities for Fe(II) may not form complexes because they are present at low concentrations in the cytosol. Genetic studies in yeast support an important role of Fe-GSH complexes in the LIP. In yeast cells engineered to overexpress a GSH importer, GSH overload was associated with activation of the iron-sensing transcription factor Aft1 and impaired activity of a cytosolic iron–sulfur cluster enzyme (Kumar et al., 2011). Depletion of cellular GSH by deletion of the essential γ-glutamylcysteine synthetase also led to Aft1 activation and impaired cytosolic Fe–S cluster enzyme activity, which could be partially restored with exogenous iron (Sipos et al., 2002; Kumar et al., 2011); neither of these phenotypes was due to oxidative stress or damage. These studies point to Fe–GSH complexes as having a critical role in maintaining iron distribution

and homeostasis in eukaryotic cells. Additional studies indicate that Fe–GSH, in conjunction with monothiol glutaredoxins, has a direct role in the formation and transfer of cytosolic Fe–S clusters.

## **PROTEINS INVOLVED IN CYTOSOLIC IRON DISTRIBUTION: MONOTHIOL GLUTAREDOXINS**

In addition to its role as a ligand in the cytosolic LIP, GSH coordinates iron as part of the Fe–S carrier complex formed by cytosolic monothiol glutaredoxins (reviewed in Rouhier et al., 2010). The glutaredoxins are a ubiquitous class of enzyme with a thioredoxin fold and are generally thought to function as thioldisulfide oxidoreductases with roles in glutathione conjugation. These enzymes contain one or two cysteine residues in their active site that are critical for enzymatic activity. A subset of the glutaredoxin family does not exhibit oxidoreductase activity and contains a monothiol active site motif that consists of the tetrapeptide CGFS. These monothiol glutaredoxins instead function as Fe–S cluster carriers. Eukaryotic cells contain separate monothiol glutaredoxins that localize to either the mitochondria (Grx5) or the cytosol (Grx3/4). Grx5 plays an important role in the assembly and transfer of Fe–S clusters in mitochondria (Rodriguez-Manzaneque et al., 2002; Wingert et al., 2005; Ye et al., 2010). Yeast, fish, and human cells lacking Grx5 exhibit profound defects in the assembly of both mitochondrial and cytosolic Fe–S clusters. Recent evidence suggests that cytosolic Grx3-type glutaredoxins may have a broader role in cytosolic iron distribution.

Structurally, Grx3-type glutaredoxins have a single N-terminal thioredoxin domain and 1–3 C-terminal glutaredoxin domains (Rouhier et al., 2010). The functionally interchangeable Grx3 and Grx4 of yeast contain a single CGFS glutaredoxin domain while the sole human ortholog, Grx3 (also called Glrx3 and PICOT), contains two. *In vitro*, Grx3 forms homodimers that can accommodate a bridging [2Fe−2S] cluster, which is coordinated by the single active site cysteine present in each monomer and by the free sulfhydryl on each of two molecules of GSH. The cluster is labile, sensitive to both reductants and oxidants, and can be transferred to a recipient Fe–S scaffold protein (Li et al., 2009; Haunhorst et al., 2010).

In many species, including prokaryotes, plants, yeast, and metazoans, monothiol glutaredoxins may be found in complex with BolA-like proteins, with distinct complexes formed in the cytosol/nucleus, mitochondria, and plastids. *In vitro*, Grx3 domains can form heterooligomers with the yeast and human cytosolic BolA2-type proteins in a 1:1 stoichiometric ratio. These complexes also coordinate bridging [2Fe−2S] clusters, with BolA2 contributing a histidine ligand to coordinate the cluster(s) (Li et al., 2011). Clusters formed on Grx3–BolA2 complexes exhibit more stability than those formed on Grx3 homodimers, and Grx3 homodimers will spontaneously transfer a [2Fe−2S] cluster to Grx3–BolA2 *in vitro* (Li et al., 2009).

Genetic studies in yeast, fish, and human cells suggest that Grx3 orthologs have similar but distinct roles in cytosolic iron delivery. In yeast, assembly of Fe–S clusters on Grx4 requires the cytosolic Fe–S assembly machinery, but it is not dependent on the remaining cytosolic Fe–S cluster machinery. Although Grx3/4 is expressed in the cytosol and nucleus, their activities affect iron handling in both the cytosol and mitochondria. Strains lacking Grx3/4 exhibit defects in the incorporation of iron–sulfur clusters into mitochondrial Fe–S enzymes (fivefold decrease in aconitase), defects in the incorporation of iron into heme (fivefold decrease), and reduced accumulation of iron into mitochondria (2.3-fold decrease). These observations are all consistent with a defect in mitochondrial iron delivery. In the cytosol, Grx3/4-deficient strains also exhibit reduced metallation of Fe–S enzymes (four to tenfold decrease). Furthermore, a cytosolic dinuclear iron enzyme, ribonucleotide reductase, also appears to require Grx3/4 for acquisition of iron (Muhlenhoff et al., 2010). This study suggests that Grx3/4 has roles in the delivery of iron to mitochondria and to cytosolic non-heme enzymes as well as in the delivery or assembly of Fe–S clusters.

Grx3-type glutaredoxins from several species have been found to interact with BolA-like proteins *in vivo*. In yeast, Grx3/4 and Fra2, the cytosolic BolA ortholog, have been found together in a complex with Aft1, the major iron-dependent transcription factor (Kumanovics et al., 2008). Grx3/4−Fra2 complexes appear to have a specialized regulatory function in yeast, as both are required for the sensing of iron (Ojeda et al., 2006; Pujol-Carrion et al., 2006) and can mediate the transfer of a [2Fe−2S] cluster to the Aft1 paralog, Aft2 (Poor et al., 2014). Zebrafish contain a single Grx3 ortholog, which could be depleted in embryos through morpholino injection. Zebrafish embryos lacking Grx3 exhibited severe defects in heme synthesis in tissues devoted to embryonic erythropoiesis (38% of embryos exhibited hemoglobin staining), but mild defects in the activities of Fe–S cluster and heme enzymes (20% decrease). In human cells, knockdown of Grx3 was associated with loss of Fe–S cluster-dependent, cytosolic aconitase activity (60% decrease) and concomitant disruption of cellular iron homeostasis. In contrast, activities of mitochondrial heme enzymes and heme synthesis were only mildly diminished (20% decrease, Haunhorst et al., 2013). Thus, in vertebrates, Grx3 may be important for iron delivery to mitochondria primarily in erythropoietic tissues, which have very high requirements for iron. Grx3 may also be involved in cytosolic Fe–S cluster delivery or assembly. Although *in vitro* studies have demonstrated that Grx3–Fra2 complexes can donate a [2Fe−2S] cluster to Aft2, and genetic studies in yeast support this evidence of a direct Fe–S transfer to Aft1, studies have not yet made clear whether Grx3-type homodimers directly deliver Fe–S clusters or iron ions to targets in yeast or human cells. Similarly, the role of Fra2/BolA2 in human cells is largely unknown, although the mitochondrially targeted BolA3 is involved in Fe–S cluster acquisition and/or maintenance for some mitochondrial Fe–S enzymes, especially lipoate synthase (Cameron et al., 2011; Haack et al., 2013; Baker et al., 2014). The mechanism by which Grx3 acquires iron is also not yet known, but it appears not to require the activity of the cytosolic Fe–S cluster assembly system.

## **PROTEINS INVOLVED IN CYTOSOLIC IRON DISTRIBUTION: PCBPs**

The term metallochaperone has been used to describe proteins that directly deliver metals to target enzymes or transporters through metal-mediated protein–protein interactions (Lee et al., 1993; Culotta et al.,1997; Pufahl et al.,1997). Iron chaperone activity has been demonstrated for the poly C binding protein (PCBP) family of proteins (Shi et al., 2008). PCBP1 was initially identified as an iron chaperone for ferritin, the ubiquitous iron storage protein. Ferritin, composed of 24 subunits of H- and L-isoforms, functions as a cellular repository of surplus iron by accommodating up to 4500 iron atoms within its spherical core. Delivery of iron atoms into the hollow sphere of ferritin occurs via the hydrophilic channels formed by the carboxylate side chains along the threefold symmetry axes in the heteropolymer. Initially, ferritin accrues iron atoms in the ferrous form, which are then oxidized to ferric iron by the ferroxidase center of H-ferritin, located in the interior of the ferritin cavity (reviewed in Arosio and Levi, 2010). Even though the regulatory mechanisms of this gene product have been extensively studied under various physiological conditions, the cytosolic trafficking system directing iron to the metalloprotein had been elusive until the recent discovery of the iron chaperone activity of PCBP1.

The role of PCBP1 in the mineralization of ferritin was initially identified through functional screening of a human liver cDNA library in a eukaryotic model lacking endogenous ferritin expression, i.e., *Saccharomyces cerevisiae* (Shi et al., 2008). Coexpression of human ferritins and PCBP1 in yeast cells produced an iron deficiency response, implying sequestration of cellular iron into the exogenous iron storage protein. This was confirmed by measuring PCBP1-dependent increases in ferritin iron content. The requirement of endogenous PCBP1 in mammalian cells was shown by loss-of-function experiments using cultured human Huh7 cells. RNA interference of PCBP1 led to inefficient mineralization of ferritin (63% reduction) along with increases in iron-mediated degradation of iron-regulatory protein 2 (IRP2) and in the cellular LIP (67% increase). *In vitro* experiments with purified PCBP1 and ferritin supported the direct involvement of PCBP1 in the mineralization of ferritin by a dose-dependent enhancement in the incorporation of iron into apoferritin by the chaperone protein.

PCBP1 [also known as heterogeneous nuclear ribonucleoprotein (hnRNP) E1 or α-CP1] is a member of the poly(rC)-binding protein family, composed of four homologous proteins of which others are PCBP2, PCBP3, and PCBP4. PCBPs were initially characterized as RNA-binding proteins, modulating the stability or translation of cellular and viral RNA species containing Crich motifs (Makeyev and Liebhaber, 2002; Ostareck-Lederer and Ostareck, 2004; Chaudhury et al., 2010). Other regulatory functions include their involvement in transcriptional regulation of gene expression and protein–protein interactions. PCBPs contain three highly conserved hnRNP K homology (KH) domains, which are ancient, RNA-binding motifs broadly distributed in prokaryotes and eukaryotes. Sequences outside of the KH domains exhibit much less conservation. PCBP1 and PCBP2 are expressed at high levels in essentially all mammalian cell types, while PCBP3 and PCBP4 exhibit much lower levels of expression in a limited range of tissues. PCBP1 is specific to mammals, while genes orthologous to those of PCBP2, 3, and 4 are present in vertebrates, flies, and worms, and distantly related orthologs are present in yeast.

Recently, PCBP2 was also identified as an iron chaperone for ferritin, conferring increased iron storage when co-expressed in yeast and manifesting decreased iron storage when depleted from human Huh7 cells (Leidgens et al., 2013). PCBP3 could also induce iron deficiency responses in yeast by enhancing the sequestration of iron into exogenously expressed ferritin. Notably, the presence of KH domains alone was not sufficient for conferring iron delivery activity to PCBPs, as a splice variant of PCBP3 lacking 26 amino acids between KH domains 2 and 3 and an unrelated RNA-binding protein containing two KH domains (fragile X mental retardation 1) had essentially no iron activity in yeast cells. PCBP4 also activates the iron deficiency response in yeast, but it shows weak genetic and physical interactions with ferritin. PCBP4 could function as a buffer for labile iron or could be involved in organellar iron delivery in specialized cell types.

Purified recombinant PCBP1 and PCBP2 were found to bind ferrous iron *in vitro* using isothermal titration calorimetry (ITC). ITC characterizes the thermodynamic properties of a ligand binding reaction, and thus allows a quantitative measurement of both binding affinity and stoichiometry. Anaerobic titration of ferrous iron into purified PCBP1 revealed an ironbinding capacity of three ferrous iron atoms per molecule of PCBP1, with a dissociation constant of 0.9 ± 0.1 μM for the first and an average of 5.8 ± 0.3 μM for the remaining two (Shi et al., 2008). PCBP2 exhibits similar iron-binding characteristics (T. Stemmler, personal communication). Physical interactions between PCBP1, PCBP2, and ferritin were also measured using ITC (Leidgens et al., 2013). Purified PCBPs exhibited no significant interaction with ferritin in the absence of iron. However, PCBP1 and PCBP2, anaerobically loaded with ferrous iron, exhibited affinities for apoferritin that were 30- and 20 fold higher, respectively, than the affinity of free ferrous iron for ferritin. Approximately, nine Fe-PCBP1 molecules bind to a ferritin oligomer. As the number of putative iron delivery channels is eight per ferritin polymer, this stoichiometry supports a model of PCBP1 facilitating iron incorporation into ferritin via direct binding at pores formed by the threefold axes of symmetry. Fe-PCBP2 binds to ferritin in a stoichiometric ratio of 4:1 Fe-PCBP2:ferritin.

In human cells, PCBP1, PCBP2, and PCBP3 can be isolated in complexes with ferritin (Leidgens et al., 2013). PCBP1 and PCBP2 are also found to interact with each other, both in yeast and human cells. Two lines of evidence suggest that PCBP1 and PCBP2 may act together in iron delivery. First, Huh7 cells depleted of PCBP1 or PCBP2 contain wild-type levels of the other paralog, yet exhibit marked defects in ferritin iron loading. The simultaneous depletion of both PCBP1 and PCBP2 produces an iron storage defect similar in magnitude to the single deletion of either. These observations indicate that PCBP1 cannot substitute for PCBP2 in human cells and vice versa. Second, quantitative immunoprecipitation studies indicate that the interaction between ferritin and PCBP1 is diminished when PCBP2 is depleted; similarly, the coprecipitation of PCBP2 with ferritin is diminished in cells depleted of PCBP1. These observations suggest that PCBP1 and PCBP2 function cooperatively, perhaps as a complex, in the distribution of iron.

## **METALLATION OF NON-HEME IRON ENZYMES**

In addition to ferritin, PCBPs exhibit iron chaperone activity toward other cytosolic iron enzymes. The class of iron- and 2 oxoglutarate (2-OG)-dependent dioxygenases comprises a large family of non-heme iron enzymes that require the incorporation of a single iron atom into its active site for activation (Ozer and Bruick, 2007; Loenarz and Schofield, 2008). Recently, two of these mononuclear iron enzymes, prolyl hydroxylase 2 (PHD2) and the asparaginyl hydroxylase, factor inhibiting HIF1 (FIH1), have been identified as clients for iron delivery by PCBPs (Nandal et al., 2011).

Hypoxia-inducible factors (HIF) are heterodimeric transcription factors that mediate the cellular adaptation to hypoxia (Semenza, 2007; Kaelin and Ratcliffe, 2008). Conditions of reduced oxygen lead to the accumulation of the alpha subunits (HIFα), HIF1α and HIF2α, which assemble into the active form with its constitutively expressed counterpart, HIF1β. In normoxia or hyperoxia, PHD hydroxylates HIFα on proline residues within the highly conserved oxygen-dependent degradation domains. Upon recognition by the von Hippel-Lindau (VHL) tumor suppressor protein, hydroxylated HIFα undergo proteolysis via the ubiquitin-proteasome pathway. FIH1 also regulates HIFα via hydroxylation of a conserved amino acid, in this case, however, at asparagine. The hydroxylation at the asparaginyl residue blocks the interaction of the alpha subunits with coactivators and thus results in the inactivation of HIF. In the absence of oxygen, a co-substrate for both enzymes, HIFα subunits remain unmodified and can accumulate and assemble into active heterodimers.

The activities of the oxygen-sensing HIF hydroxylases are also responsive to the availability of iron. Treatment of cells with supplemental iron results in an increase in the HIF hydroxylase activity, while iron chelation produces the opposite effect (Kaelin and Ratcliffe, 2008). Moreover, induction of HIF2 signaling has been reported in the small intestine of mice subjected to dietary iron deficiency, which may be attributable to compromised HIF hydroxylase activity (Shah et al., 2009). While the requirement of iron for HIF hydroxylase activation has been known for several decades, the mechanism which mediates the incorporation of a mononuclear iron center into the apoenzyme remained unknown until recent studies implicated the involvement of PCBPs.

Cultured human cells lacking PCBP1 or PCBP2 exhibit nuclear accumulation of active HIF1α, which increases in cells made mildly or transiently iron deficient (Nandal et al., 2011). Loss of PCBP expression leads to prolonged HIF1α half-life without altering steady-state levels of HIF1α transcripts. The responses of HIF1α to PCBP1 or PCBP2 depletion are due to the inactivation of PHD2, which accounts for nearly all of the prolyl hydroxylation of HIF1α in cells (Berra et al., 2003). Cells depleted of PCBP1, PCBP2, or both exhibit reduced incorporation of the iron cofactor into PHD2 (20–40% of control) and complete loss of enzyme activity in cell lysates. In lysates, the effects of PCBP depletion on PHD activity can be reversed by exogenous addition of excess iron. As for the case of ferritin, PCBP1 can form complexes with PHD2 in human cells. Thus, the PCBP-mediated delivery of iron into the active site of the mononuclear iron enzyme is likely accomplished via direct protein–protein interaction.

The activity of FIH1, also an iron- and oxygen-dependent HIF1α regulator, is dependent on PCBP1 as well (Nandal et al., 2011). Cells depleted of PCBP1 exhibit increased expression of a reporter construct repressed by FIH1, and FIH1 was detected in a complex with PCBP1 in iron-treated cells. The loss of FIH1 activity and the physical interaction between cellular PCBP1 and FIH1 suggest that FIH1 is also a target for iron delivery by PCBP1. Whether all iron- and 2-OG-dependent enzymes require PCBPs for their metallation and activation remains undetermined.

Other classes of iron metalloenzymes may also depend on iron delivery by PCBPs. Of particular interest are the dinuclear, nonheme iron enymes, some of which contain iron-binding active sites structurally related the ferroxidase site of ferritin, a known target of PCBPs. Examples of this class of enzymes are the small subunit of ribonucleotide reductase (Stubbe and Cotruvo, 2011) and deoxyhypusine hydroxylase (DOHH, Vu et al., 2009). DOHH is a dinuclear iron enzyme that is required for the posttranslational modification of a single lysine residue on eukaryotic initiation factor 5A (eIF5A, Park et al., 2010). EIF5A and the conversion of this conserved lysine to hypusine are essential in all eukaryotes, as it enables the translation of peptides containing polyproline sequences (Gutierrez et al., 2013). Cells lacking PCBP1 or PCBP2 exhibit loss of DOHH activity, as measured by the accumulation of partially modified deoxyhypusine and reduced levels of fully modified hypusine (20% of control) in living cells. DOHH from PCBP-depleted cells also exhibits loss of the iron cofactor when the cells are exposed to mild iron deficiency. As for other PCBP targets, DOHH is detected in a complex with PCBP1, indicating their direct association during the assembly of the diiron active site. PCBPs do not appear to be required for iron delivery to mitochondria or for the *de novo* assembly of iron–sulfur clusters in the cytosol (Frey et al., 2014). Collectively, PCBPs can regulate the cellular distribution of labile iron by depositing it into the cellular iron reservoir ferritin and/or by directing it to the site of utilization as an inorganic cofactor for cytosolic non-heme iron enzymes.

## **ROLE OF PCBPs IN REGULATION OF IRON HOMEOSTASIS**

The iron chaperone activities of PCBPs do not appear to be regulated in response to changing cellular iron availability. That is, the expression level and iron-binding activity do not appear to change under different iron conditions (Frey et al., 2014). Although interactions with client enzymes do appear, in some cases, to be lessened when iron is scarce, this effect could be mediated solely by the lessened stability of the chaperone–client interaction in the absence of bound iron. PCBPs may, however, influence cellular iron homeostasis through their iron chaperone activities. Huh7 cells depleted of PCBPs store less cytosolic iron in ferritin, thereby expanding the LIP (Shi et al., 2008). This expansion of the LIP is reflected in lowered levels of iron regulatory protein 2 (IRP2), which undergoes ubiquitin-mediated degradation in the presence of increased labile iron (Guo et al., 1995). Whether the degradation associated with PCBP depletion is dependent on the recently described ubiquitin ligase FBXL5 has not been determined (Salahudeen et al., 2009; Vashisht et al., 2009). HEK 293T cells depleted of PCBPs exhibit reduced activity of cytosolic (c-) aconitase (Frey et al., 2014), an enzyme that carries a labile [4Fe-4S] cluster in its active site. C-aconitase, also called iron regulatory protein 1 (IRP1), is

a bifunctional enzyme because it acquires RNA-binding activity when the [4Fe-4S] cluster is absent (reviewed in Anderson et al., 2012). Together, IRP1 and IRP2 regulate cellular iron uptake and storage by altering the translation and stability of mRNA transcripts that contain iron-responsive elements (IREs) in the 5 or 3 UTR. These transcripts encode proteins of iron uptake, e.g., transferrin receptor and divalent metal transporter 1 (DMT1), and iron storage, e.g., ferritin. Although PCBP depletion leads to loss of caconitase activity, it does not appear to lead to a reciprocal increase in IRE-binding activity, suggesting that PCBP depletion does not result in the complete disassembly of the Fe–S cluster in IRP1. In fact, PCBP depletion was associated with loss of IRP1-mediated IRE binding activity. Thus, loss of cellular PCBPs results in loss of IRP1- and IRP2-mediated IRE-binding activity, which likely produce misregulation of proteins that maintain the iron balance of the cytosol.

Tissues may also exhibit specific requirements for PCBPs in the delivery of iron. In mice, dietary iron deficiency leads to increased HIF2α in intestinal epithelial cells, where it promotes the transcription of DMT1 and other genes involved in the absorption of dietary iron (Mastrogiannaki et al., 2009; Shah et al., 2009). Intestinal HIF2α is likely regulated by prolyl and asparagyl hydroxylases, which receive iron cofactors from PCBP1 and PCBP2. Thus, PCBPs may be important for the response to iron deficiency in gut cells. Recently, H-ferritin was determined to be an indispensible factor for accurately controlling intestinal iron absorption, primarily by retaining iron in absorptive enterocytes and preventing excessive iron transfer from enterocytes to the systemic circulation (Vanoaica et al., 2010). Loss of H-ferritin expression in the small intestine led to characteristic manifestations of hemochromatosis. The perturbation in systemic iron homeostasis was attributed to higher systemic absorption of iron due to enhanced efflux from enterocytes through the basolateral transporter, ferroportin 1. Dysfunction of the iron storage protein also affected cellular iron homeostasis in the enterocytes. It is not yet known whether PCBPs mediate ferritin iron loading in the intestinal enterocyte, but if enterocytes do rely on PCBPs, they may also serve as a limiting component for body iron accrual by directing and loading the metal to ferritin. In addition to these indirect effects on iron uptake, storage and efflux, PCBPs may exert direct effects on transport. A recent report indicates that PCBP2 directly interacts with both DMT1 and ferroportin in cells, enhancing the transport of iron via DMT1 (Yanatori et al., 2014). Other cell types, such as macrophages and erythropoietic cells have special requirements for iron handling, as they accommodate the large iron fluxes associated with the recycling of iron from senescent red blood cells and hemoglobin synthesis, respectively. Whether PCBPs play a role in these processes remains to be determined. An important consideration in these studies is the kinetics involved in the intracellular iron delivery systems. What is the time frame for delivery of iron and metallation of enzymes in the cell? New approaches will be required to address these types of questions. In summary, PCBPs may provide strategies for the efficient management of the newly acquired or mobilized iron atoms by efficiently delivering them to their appropriate destination, particularly in cell types that routinely encounter dynamic changes in the cytosolic LIP.

## **REFERENCES**


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

*Received: 18 April 2014; paper pending published: 18 June 2014; accepted: 04 July 2014; published online: 22 July 2014.*

*Citation: Philpott CC and Ryu M-S (2014) Special delivery: distributing iron in the cytosol of mammalian cells. Front. Pharmacol. 5:173. doi: 10.3389/fphar.2014.00173 This article was submitted to Drug Metabolism and Transport, a section of the journal Frontiers in Pharmacology.*

*Copyright © 2014 Philpott and Ryu. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*

## Ferritin polarization and iron transport across monolayer epithelial barriers in mammals

## *Esther G. Meyron-Holtz\*, Lyora A. Cohen, Lulu Fahoum, Yael Haimovich, Lena Lifshitz, Inbar Magid-Gold, Tanja Stuemler and Marianna Truman-Rosentsvit*

Laboratory for Molecular Nutrition, Faculty of Biotechnology and Food Engineering, Technion – Israel Institute of Technology, Technion City, Haifa, Israel

#### *Edited by:*

Paolo Arosio, University of Brescia, Italy

#### *Reviewed by:*

James Connor, Penn State Hershey Medical Center, USA Abolfazl Zarjou, University of Alabama at Birmingham, USA

#### *\*Correspondence:*

Esther G. Meyron-Holtz, Laboratory for Molecular Nutrition, Faculty of Biotechnology and Food Engineering, Technion – Israel Institute of Technology, Technion City, Haifa 32000, Israel e-mail: meyron@tx.technion.ac.il

Epithelial barriers are found in many tissues such as the intestine, kidney and brain where they separate the external environment from the body or a specific compartment from its periphery. Due to the tight junctions that connect epithelial barrier-cells (EBCs), the transport of compounds takes place nearly exclusively across the apical or basolateral membrane, the cell-body and the opposite membrane of the polarized EBC, and is regulated on numerous levels including barrier-specific adapted trafficking-machineries. Iron is an essential element but toxic at excess. Therefore, all iron-requiring organisms tightly regulate iron concentrations on systemic and cellular levels. In contrast to most cell types that control just their own iron homeostasis, EBCs also regulate homeostasis of the compartment they enclose or the body as a whole. Iron is transported across EBCs by specialized transporters such as the transferrin receptor and ferroportin. Recently, the iron storage protein ferritin was also attributed a role in the regulation of systemic iron homeostasis and we gathered evidence from the literature and original data that ferritin is polarized in EBC, suggesting also a role for ferritin in iron trafficking across EBCs.

**Keywords: iron transport, iron metabolism, epithelial barriers, tight junctions, ferritin polarization**

#### **SELECTED PROTEINS INVOLVED IN IRON TRAFFICKING**

In the blood stream, iron normally circulates bound to transferrin and is taken up by cells through binding of diferric transferrin to the transferrin receptor 1 (TfR1). This complex is internalized via clathrin-coated pits that form early endosomes. These are acidified, iron is released from transferrin, apo-transferrin is recycled to the plasma membrane (PM) and is released to the blood-stream (Trowbridge et al., 1993). The endosomally released iron undergoes reduction by the endosomal ferric-reductase Steap3 and is transported from the endosome to the cytosol by the divalent metal transporter 1 (DMT1), an H+/iron cotransporter (Picard et al., 2000; von Drygalski and Adamson, 2013). DMT1 is also found on the PM and in other subcellular locations where it can import iron into the cytosol. Imported iron may enter: (1) the cytosolic labile iron-pool, (2) different cellular compartments where it will be integrated to functional heme-, iron-sulfur cluster- or other ironcontaining proteins, (3) ferritin, the intracellular iron storage protein. Imported iron may also be released from the cell via ferroportin.

Due to the potential toxicity of iron, uptake, storage, and mobilization pathways are tightly regulated. iron regulatory proteins (IRP) 1 and 2 regulate iron uptake and storage by binding to mRNA structures called iron regulatory elements (IRE). Iron release from cells is regulated by the iron-regulated hormone hepcidin, which controls ferroportin levels. Serum hepcidin concentration is regulated by many signals including iron, oxygenation and inflammation (Napier et al., 2005; Richardson et al., 2010).

#### **EPITHELIAL BARRIERS**

To cross a monolayer of an epithelial barrier, a molecule or element must: (1) Reach the EBC, (2) Enter it, (3) Get across, and (4) Be exported on the other side. EBCs are connected with tight junctions that separate between the apical and basolateral membrane of the EBC and thus EBCs create a living cellular barrier. The apical and basolateral poles face completely different environments. EBCs are able to sense the two environments and transport nutrients and other molecules across according to the received signals. This depends on close interaction and crosstalk between EBCs and their neighboring cells. The direction of transport is dictated in part by the expression of transporters and receptors on the respective membrane. Most of these carriers have a default membrane to which they are trafficked, but in specific epithelial barriers the trafficking machinery has adapted to enable a change of direction. TfR1 has been used extensively as a marker for recycling-endosome trafficking and information on its localization is available in various cell-types.

## **IRON TRAFFICKING PROTEINS IN SELECTED EPITHELIAL MONOLAYERS**

Examining the location of TfR1, DMT1 and ferroportin in different epithelial monolayers revealed that each of these barriers has adapted iron trafficking to its environment and specific function. Interestingly, we found evidence that ferritin distribution is also affected by cell polarization (**Figures 1** and **2B**). Accumulation of ferritin at one end of a polarized cell can be due to trafficking of ferritin itself or a subcellular compartment in which ferritin is contained. In addition, ferritin may be taken up and endocytosed

**FIGURE 1 | Ferritin polarization. (A)** Confocal microscopy images of Caco2 cells stained with anti L-subunit ferritin antibodies (green) and nuclei are stained with DAPI (blue). Ferritin levels were higher between the 8–12 μm sections showing the apical side of the cells. Optical section-size is indicated as distance from glass-bottom in red. Laser power, voltage and offset were identical between different sections for each fluorophore. **(B)** Quantification of ferritin-fluorescence in MSC-1 and

by one of the recently described ferritin receptors and endosomal ferritin may not be distributed evenly throughout the cell.

#### **INTESTINE**

The duodenum is responsible for digestion and absorption of most nutrients including iron, while the jejunum and ileum mainly absorb nutrients that were not absorbed earlier. Dietary iron reaches the duodenal enterocytes either as ferric- or heme-iron. Ferric iron is reduced to ferrous iron by dietary components, such as amino acids, amines and ascorbic acid, or by ferric reductases of the brush border prior to absorption. Ferric iron can also be absorbed after chelation by mucins, which maintain the iron in the ferric state. Around 25 to 50% of dietary heme-iron is absorbed compared to only 1 to 10% of ionic iron and the two forms do not compete (Roy and Enns, 2000; Shayeghi et al., 2005; Ma et al., 2006; Sharp and Srai, 2007; MacKenzie et al., 2008; Le Blanc et al., 2012).

Ferrous- and heme-iron enter the enterocyte apically via DMT1 and possibly the heme-carrier-protein1 (HCP1), respectively (Gunshin et al., 1997; Le Blanc et al., 2012). Following absorption, heme is detectable in membrane bound vesicles within the cytoplasm. Heme-oxygenase1 (HO1) removes the iron from the protoporphyrin ring and the ferrous iron joins the intracellular pool along with non-heme-iron. The mechanism by which iron is translocated within the enterocyte, i.e., from the apical to the basolateral membrane, has not yet been elucidated.

During translocation, iron maintains its solubility and low reactivity possibly by binding to protein chaperones or to ferritin (Ma Caco2 cells is expressed as percent of total fluorescence of the image area. Each confocal slice is numbered. Number 1 indicates the coverslip side. The percent area occupied by the ferritin signal in each Z slice was calculated from the sum of fluorescence in all stacks and plotted as a function of confocal slice. An increase of intensity is clearly detected on the basolateral side of the MSC-1 and the apical side of the Caco-2 cells.

et al., 2006). Efflux of iron across the basolateral membrane is mediated by ferroportin, which is regulated systemically by hepcidin and locally by ferritin (Vanoaica et al., 2010). Following export, iron is oxidized by hephaestin and loaded onto transferrin for transport in the blood circulation (MacKenzie et al., 2008). If systemic signaling causes downregulation of ferroportin, enterocytes will fill up with ferritin and eventually slough into the lumen (Sharp and Srai, 2007). Interestingly, ferritin distribution in enterocytes is not even throughout the cytosol, but appears in a punctate pattern that is more concentrated near the apical pole (Figure 2B from Vanoaica et al., 2010) and **Figure 1**. TfR1 is expressed basolaterally in enterocytes, securing a continuous iron supply for these fast-dividing cells.

## **KIDNEY**

One of the many functions of the kidney is the protein- and nutrient-reabsorption from the primary urine. This takes place in the nephron, the basic unit of the kidney, which is composed of four main areas: glomerular capsule, proximal convoluted tubule (PCT), loop of Henle, and the distal convoluted tubule (DCT). Most reabsorption occurs in the PCT where epithelial cells apically express microvilli. In patients suffering from proteinuria, transferrin is abundant in the urine, suggesting that transferrin passes the glomerulus into the primary filtrate (Kozyraki et al., 2001). As the final urine contains neither protein nor iron, there must be a mechanism by which both transferrin and the bound iron are reabsorbed. In most tissues, transferrin-iron uptake is mediated by TfR1 and after intracellular iron release, apo-transferrin is recycled and released extracellularly (Fuller and Simons, 1986). A

barriers found in intestinal, renal, testicular, retinal and choroid plexus (CP) cells. Iron transporters are localized on the relevant membrane (N, nucleus; E, endosome; LE, late endosome; L, lysosome; P, phagosome; FPN, ferroportin; \*represents suggested ferroportin localizations by the literature, arrows represent proposed direction of iron transport). **(B)** Intracellular ferritin polarization in epithelial barriers published in the literature. (1) Intestinal villi paraffin sections stained with rabbit anti-ferritin H antibody (from Vanoaica et al., 2010) and reveal that ferritin is not evenly distributed throughout the cytosol, but appears in a punctate pattern that is more concentrated near the apical pole (arrows). (2) Kidney cortex sections stained with rabbit anti-ferritin H antibody (green). Nuclei were stained with DAPI (blue), (from Cohen et al., 2010) and reveals an uneven ferritin

polyclonal GFP antibody (green), (from Leichtmann-Bardoogo et al., 2012), and reveals that ferritin is enriched in the SC cytosol close to the basolateral pole (white), especially around the early primary spermatocyte. (4) Retinal samples stained with rabbit anti-ferritin L antibody (red). Nuclei were stained with DAPI (blue; unpublished data, Joshua Dunaief, personal communication). This image reveals that ferritin is distributed in a polarized manner and was localized to the basal RPE (arrow) in a WT C57BL6/J retina. (5) Choroid plexus samples stained with anti-ferritin antibody (green), (from Rouault et al., 2009) shows expression of ferritin in 24-months old mouse-choroid plexus, but subcellular location cannot be determined. Permissions have been obtained for use of copyrighted material from all these sources.

variety of studies addressed the question of how the transferrin is reabsorbed in the kidney, regardless of its iron-loading status.

Cubilin is a 460 kDa receptor, known to bind and internalize many ligands such as the vitamin B12-intrinsic factor complex in the intestine and apo-lipoprotein in the kidney, where it is located apically on the PCT membrane. Cubilin lacks a transmembrane domain, and it builds a complex with amnionless and megalin and depends on megalin for proper localization and internalization (Christensen and Birn, 2002; Verroust and Christensen, 2002). Transferrin was found to be one of cubilin's ligands (Kozyraki et al., 2001), suggesting that cubilin might be responsible for the reabsorption of apo- and holo-transferrin from the primary urine. In both cubilin-defective dogs and megalin-deficient mice, transferrin was detected in the urine, suggesting that, although the high affinity of transferrin is related to cubilin, megalin is probably also essential for the cubilin-mediated transferrin internalization.

To evaluate if TfR1 may play a role in transferrin reabsorption, kidney tissues where analyzed. In murine kidney sections, TfR1 was detected in the apical membrane of mouse-PCT (Zhang et al., 2007). PCT epithelium was recently shown to lack the clathrin adaptor AP-1B and therefore TfR1 is redirected to the apical rather than the basolateral membrane (Perez Bay et al., 2013). This suggests that TfR1 may also be involved in transferrin uptake and reabsorption from primary urine, but does not answer the question of the fate of the TfR1 bound transferrin. Usually this transferrin would only release its iron and recycle to the primary urine, thus further studies are needed to clarify the pathway of transferrin after its internalization. Nevertheless, reuptake of transferrin from primary urine is probably not performed exclusively by cubilin but rather in conjunction with TfR1 (Kozyraki et al., 2001).

The megalin/cubilin-complex is trafficked to the lysosome after internalization where iron is released and possibly transported to the cytosol by DMT1. In kidney-epithelium, DMT1 is mostly found intracellularly and only in distal tubules was it found on the apical membrane as well (Wareing et al., 2003). Intracellular DMT1 is likely localized to the endo/lysosomal membrane where it can mediate iron import to the cytosol after transferrin internalization by cubilin or TfR1. Research on ion transport modulation, in renal epithelial cells specifically (Welling and Weisz, 2010), can shed additional light on iron transport.

Following entry to the cytosol, iron needs to be exported to the blood through the basolateral membrane, where ferroportin is located, and may export iron to the renal interstitium. However, a recent study in mice localized ferroportin apically and suggested a role for ferroportin in iron absorption (Wolff et al., 2011; Zarjou et al., 2013).

In addition, ferritin may also play an important role in the kidney. Deletion of the H-ferritin subunit in renal PCT in mice with acute kidney injury worsened their condition (Zarjou et al., 2013). Moreover, an uneven ferritin distribution was detected in renal proximal tubule cells where ferritin was co-localized with villin and was enriched near the apical pole of the PCT-cells, a region enriched with lysosomes (Cohen et al., 2010).

## **TESTIS**

The testis is divided into two major compartments: (1) the looped seminiferous tubule (SFT), where spermatogenesis occurs, and (2) the interstitium composed of androgen secreting Leydig cells, blood vessels, macrophages, lymphocytes, lymphatic vessels, and connective tissue (Mital et al., 2011; Goldstein and Schlegel, 2013). In the SFT spermatocytes differentiate to mature sperm in close interaction with the Sertoli cells (SCs). SCs are connected to each other by tight junctions, forming the blood-testis barrier (BTB). The BTB plays an important role in protecting the developing spermatocytes from immune mediators (leukocytes and antibodies), toxins, pathogens, and nutritional fluctuations (Dym and Fawcett, 1970; Mital et al., 2011) and provides one of the mechanisms that maintain testis as an immune-privileged site (Arck et al., 2014).

Sertoli cells function as "nurse cells" to the developing spermatocytes (Sylvester and Griswold, 1994), surround the germ cells and thus are able to supply them with nutrients and regulatory factors while functioning as a scaffold on which the germ cells move unidirectionally toward the SFT lumen. The SFT is lined with peritubular myoid cells (PTM) that form an additional barrier and are responsible for the contraction of the tubule and the subsequent transport of maturing sperm cells and testicular fluid from the SFT lumen to the epididymis where they further mature (Maekawa et al., 1996; Goldstein and Schlegel, 2013).

The BTB is not completely impervious, as it allows the passage of developing spermatocytes from the basolateral to the adluminal compartment. This process is highly regulated, ensuring a continuing presence of the BTB and protection of the delicate developing spermatocytes (Smith and Braun, 2012).

Iron is needed by spermatocytes mainly for DNA synthesis and mitochondriogenesis. Recently, we suggested a novel model for an autonomous iron cycle within the SFT (Leichtmann-Bardoogo et al., 2012) that renders testes resistant to fluctuations of peripheral iron and ensures a constant supply of iron for maturing spermatocytes. The findings that ferritin levels were high in early spermatocytes near the basal membrane of the SFT, but mRNA levels, especially of the H- subunit were much lower in these cells than in the neighboring SCs and PTM, raised the possibility that ferritin may be imported to spermatogonia (http://public.wsu.edu/∼griswold/microarray). As the spermatocytes develop, their ferritin levels decreased toward the SFT lumen, suggesting intracellular iron redistribution into functional compartments such as the mitochondria of developing meiotic germ cells. During spermatid-maturation the elongating spermatids shed residual bodies containing cytosol and mitochondria that are phagocytosed by SCs near the SFT lumen, thus recycling the iron back to the SCs. DMT1 did not colocalize with TfR1 and was found in elongating spermatids and near the apical pole of SCs, signifying that DMT1 is involved in TfR1-independent iron transport in the SCs. SCs synthesize ferritin and we suggested that they traffic ferritin to the basolateral pole where ferritin is secreted and passed on to the early spermatocytes in a regulated manner (Leichtmann-Bardoogo et al., 2012). Supporting this concept is the polarized distribution of ferritin with increased ferritin concentration near the basolateral pole detectable in the MSC1 mouse Sertoli cell-line (**Figure 1**).

Ferroportin is mainly expressed on the PTM surrounding the SFT and possibly on the basal membrane of SCs. Although its role is unknown, one possibility is that it could link the SFT iron cycle to the iron rich interstitium by exporting excess toxic iron from the SFT.

## **RETINA**

The retinal pigment epithelium (RPE) is a monolayer of pigmented cells that form the blood–retina barrier (BRB) together with the neuro-retinal vasculature. These two cell layers, epithelial and endothelial, respectively, define the retinal compartment and separate it from the periphery. The RPE apical membrane and processes face the retinal photoreceptors' outer segments, enabling close interaction and exchange between these two celltypes, and the RPE basal membrane faces the Bruch's membrane that separates between RPE and the chorio-capillaries (Strauss, 2005).

TfR1 on the basal surface of RPE cells binds and takes up transferrin from the choroidal circulation into endosomal compartments. However, a full iron transport mechanism from the basal to the apical surface has not been elucidated, and TfR1 was also found on the apical membrane of RPE where it can bind and internalize retinal transferrin that is synthesized and secreted from RPE (Hunt et al., 1989; Yefimova et al., 2000). These findings suggest that the direction of iron flow through RPE may be adapted to retinal needs. *Trans*-cytosis of transferrin, release of elemental iron that originates from transferrin and release of ferritin-like molecules have all been implicated in RPE iron trafficking (Hunt et al., 1989; Hunt and Davis, 1992; Burdo et al., 2003). Elemental iron is exported from cells by ferroportin, which was detected in mouse retina in several cell types including RPE. Though in RPE it was localized near the basal surface, suggesting ferroportin involvement, along with ceruloplasmin and hephaestin, in iron export to the choroidal vasculature (He et al., 2007).

Furthermore, ferritin is distributed in a polarized manner in the RPE and was more concentrated near the basolateral pole (Hahn et al., 2004 and **Figure 2B** Joshua Dunaief, personal communication). In addition, the role of DMT1 in the retina and its localization in RPE remains unclear (He et al., 2007). Taken together, the physiology of iron transport across the RPE barrier awaits interesting research.

## **CHOROID PLEXUS**

Transport of molecules into the brain is strictly regulated by two major barrier systems: the blood–brain barrier (BBB) and the blood–cerebrospinal fluid barrier (BCB). The BBB is formed by the tight junctions of the endothelial cells, and separates the blood circulation from the brain interstitial fluid (ISF). The BCB is formed by the tight junctions of the choroid plexus (CP) epithelium and arachnoid membrane, and separates the blood from the cerebrospinal fluid (CSF; Brightman, 1977; Caplan et al., 1997; Cipolla, 2009).

Iron is essential for neuronal function, however, iron deficiency and excess are known to underlie the patho-etiology of several neurodegenerative disorders (Rouault, 2013). The uptake of iron at the BBB is well documented (Rothenberger et al., 1996; Richardson and Ponka, 1997; Fillebeen et al., 1999; Morgan and Moos, 2002; Moos and Rosengren Nielsen, 2006). In recent years, however, an iron uptake pathway through the CP was proposed to be of importance, (Marques et al., 2009; Rouault et al., 2009; Connor et al., 2011; Mesquita et al., 2012) mainly due to the combination of high surface area (Speake and Brown, 2004) and high blood supply (Maktabi et al., 1990) of the CP cells. The basolateral side of these polarized epithelial cells faces the blood, whereas the apical side contains microvilli that are in direct contact with the CSF (Abbott et al., 2006). Thus, iron transport across CP cells is an example of an epithelial barrier-transport.

CP seems to be the central expression site for the majority of iron metabolism key proteins within the brain (Rouault et al., 2009). TfR1 was expressed on the basolateral side of CP cells facing the capillaries (Moos, 2002). In another report, TfR1 was localized perinuclear while DMT1 and ferroportin were located apically (Wang et al., 2008). The ferroportin localization suggests a direction of iron flux into the CSF (Wu et al., 2004). The iron reductase duodenal cytochrome B (dcytb) was found mainly apically and ferritin distribution was strikingly different from dcytb, but could not be confirmed to show basolateral accumulation (Rouault et al., 2009). Additional studies must be performed to clarify the role of the CP in iron metabolism**.**

## **CONCLUSION**

Transferrin receptor 1 and divalent metal transporter 1 are so far the most studied iron importers and ferroportin is the main confirmed iron exporter. Information on their localization and trafficking in epithelial barriers accumulates and is suggestive for directions of iron transport across these barriers.

Still little is known about the intracellular pathway of iron across these polarized cells. In our view, it is intriguing that ferritin is not evenly distributed throughout the cytosol, appears often punctate and accumulates near specific poles of these barriercells (**Figure 2A**). It is possible that polarized ferritin resides in the endo-lysosomal system, and that the punctate accumulation of ferritin is within these subcellular compartments. In the two epithelial monolayers that separate the outside from the inside of the body, namely intestinal and renal epithelium, ferritin distribution is polarized in such a way, that it is closer to the apical side and the microvilli. Conversely, in epithelial monolayers that separate a compartment from the periphery, such as Sertoli cells and RPE, ferritin is close to the basolateral membrane (LaVaute et al., 2001; Hahn et al., 2004; Rouault et al., 2009; Cohen et al., 2010; Vanoaica et al., 2010; Leichtmann-Bardoogo et al., 2012; and demonstrated in **Figures 1** and **2**). In Sertoli cells, polarized ferritin may be secreted and may function as an iron exporter (Leichtmann-Bardoogo et al., 2012), which could be one of the roles of polarized ferritin. Nevertheless, the function of ferritin polarization still needs to be elucidated.

## **ACKNOWLEDGMENTS**

This format of Perspective is limited to 3000 words and we apologize to all researchers whose original work had to be cited through reviews. We thank Prof. Kenneth Tung for critical reading of the manuscript and Joshua Dunaief for sharing his pictures from the retina. This work was funded by the US–Israel Binational Science Foundation, Grant no. 2007466 to Esther G. Meyron-Holtz and Tracey A. Rouault.

## **REFERENCES**


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

*Received: 18 May 2014; accepted: 02 August 2014; published online: 25 August 2014. Citation: Meyron-Holtz EG, Cohen LA, Fahoum L, Haimovich Y, Lifshitz L, Magid-Gold I, Stuemler T and Truman-Rosentsvit M (2014) Ferritin polarization and iron transport across monolayer epithelial barriers in mammals. Front. Pharmacol. 5:194. doi: 10.3389/fphar.2014.00194*

*This article was submitted to Drug Metabolism and Transport, a section of the journal Frontiers in Pharmacology.*

*Copyright © 2014 Meyron-Holtz, Cohen, Fahoum, Haimovich, Lifshitz, Magid-Gold, Stuemler and Truman-Rosentsvit. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*

**REVIEW ARTICLE** published: 10 July 2014 doi: 10.3389/fphar.2014.00156

## The role of iron in the skin and cutaneous wound healing

## *Josephine A. Wright1\*, Toby Richards1 and Surjit K. S. Srai <sup>2</sup>*

<sup>1</sup> Division of Surgery and Interventional Science, University College London, University College & Royal Free Hospitals, London, UK <sup>2</sup> Department of Structural and Molecular Biology, Division of Biosciences, University College London, London, UK

#### *Edited by:*

Raffaella Gozzelino, Instituto Gulbenkian de Ciência, Portugal

#### *Reviewed by:*

Edward Pelle, Estée Lauder Inc., USA Maureane Hoffman, Duke University Medical Center, USA

#### *\*Correspondence:*

Josephine A. Wright, Division of Surgery and Interventional Science, University College London, 4th Floor, 74 Huntley Street, London WC1E 6AU, UK e-mail: josephine.wright@nhs.net

In this review article we discuss current knowledge about iron in the skin and the cutaneous wound healing process. Iron plays a key role in both oxidative stress and photo-induced skin damage. The main causes of oxidative stress in the skin include reactive oxygen species (ROS) generated in the skin by ultraviolet (UVA) 320–400 nm portion of the UVA spectrum and biologically available iron. We also discuss the relationships between iron deficiency, anemia and cutaneous wound healing. Studies looking at this fall into two distinct groups. Early studies investigated the effect of anemia on wound healing using a variety of experimental methodology to establish anemia or iron deficiency and focused on wound-strength rather than effect on macroscopic healing or re-epithelialization. More recent animal studies have investigated novel treatments aimed at correcting the effects of systemic iron deficiency and localized iron overload. Iron overload is associated with local cutaneous iron deposition, which has numerous deleterious effects in chronic venous disease and hereditary hemochromatosis. Iron plays a key role in chronic ulceration and conditions such as rheumatoid arthritis (RA) and Lupus Erythematosus are associated with both anemia of chronic disease and dysregulation of local cutaneous iron hemostasis. Iron is a potential therapeutic target in the skin by application of topical iron chelators and novel pharmacological agents, and in delayed cutaneous wound healing by treatment of iron deficiency or underlying systemic inflammation.

**Keywords: iron, skin, wound-healing, ultraviolet, iron chelating agents**

## **INTRODUCTION**

Iron is a vital co-factor for proteins and enzymes involved in energy metabolism, respiration, DNA synthesis, cell cycle arrest and apoptosis. Over the past 10 years, major advances have been made in understanding the genetics of iron metabolism and this has led to identification of a number of new proteins, including hepcidin, an acute phase protein that is the master regulator of iron absorption and utilization, often activated in chronic diseases (Weiss, 2009; Finberg, 2013).

Historically, it has long been known that iron is essential for healthy skin, mucous membranes, hair and nails. Clinical features of iron deficiency include skin pallor, pruritus, and predisposition to skin infection (impetigo, boils and candidiasis), angular chelitis, swollen tongue, fragile nails, kolionychia, and dry brittle hair.

## **ROLE OF IRON IN THE SKIN**

## **NORMAL PHYSIOLOGY OF IRON**

The normal physiology of iron in the skin is complex and not clearly understood. It is known that iron levels in normal epidermis are thought to vary over a wide range (Molin andWester, 1976; Kurz et al., 1987). Within normal dermis, iron levels also vary and are thought to increase during the aging process (Leveque et al., 2003). Furthermore, iron-containing proteins have specific function such as the metabolism of collagen by procollagen-proline dioxygenase (Richardson et al.,1996; Polefka et al.,2012; **Figure 1**). Iron is not actively excreted from the body, however the skin is a key organ in iron hemostasis as iron is lost through the skin by desquamation (**Figure 2**). Current theories regarding the

underlying mechanisms of desquamation include active dissolution of desmosomes involved in keratinocyte cell–cell adhesion, by hydrolytic protease digestion (Milstone, 2004). Desquamation of keratinocytes is thought to account for 20–25% of absorbed iron that is lost (Jacob et al., 1981). Yet overall, the daily loss of iron by desquamation is approximately 25% that of daily urinary iron excretion (Molin and Wester, 1976). Evidence is emerging from genetic model mouse studies by Milstone et al. (2012) that both loss of iron by desquamation and local changes in epidermal iron metabolism have some role in systemic iron metabolism (these studies investigated three groups of mice: firstly mice overexpressing of HPV16 E7 gene, which causes a threefold increase in epidermal turnover, secondly mice overexpressing the transferrin receptor which causes a three tofourfold increase of epidermal iron in a skin model, and finally a systemic hemochromatosis knockout model crossed with the epidermal iron sink model). Additionally, gender-related differences in iron status may be responsible for the increased longevity of women as compared to men. The relative difference in cell iron levels between the sexes may be of importance both physiologically and in setting of pathophysiological conditions (Pouillot and Polla, 2013).

## **IRON, OXIDATIVE STRESS, AND PHOTO-INDUCED DAMAGE**

The main causes of oxidative stress in the skin are reactive oxygen species (ROS) generated in the skin by ultraviolet (UVA) 320–400 nm portion of the UVA spectrum. Iron plays a key role in oxidative stress processes, as it is a transition metal, which exists in two stable states, Fe2<sup>+</sup> (electron

donor) and Fe3<sup>+</sup> (electron acceptor). Intracellular labile iron can undergo redox cycling between its most stable oxidation states (Fe2+/Fe3+) and react with ROS such as superoxide anion, hydrogen peroxide, giving rise to hydroxyl radicals via the Fenton

reaction or superoxide-driven Fenton chemistry (Pelle et al., 2011).

Exposure of skin fibroblasts to UVA can generate ROS that promote oxidative damage in lysosomal, mitrochondrial, nuclear, and plasma membranes. Ultimately loss of plasma membrane integrity together with mitrochondial ATP depletion results in necrotic cell death (Aroun et al., 2012). It is thought that compared with skin fibroblasts, keratinocytes are more resistant to UVA mediated membrane damage and cytotoxicity. In vitro studies have shown that although UVA starts lysosomal damage, ferritin degradation and cytosolic labile iron release in keratinocytes, the absolute level of UVA induced labile iron release is several fold lower than in fibroblasts, suggesting a link between labile iron release and keratinocyte resistance to UVA mediated damage (Zhong et al., 2004).

## **ANEMIA, IRON DEFICIENCY, AND CUTANEOUS WOUND HEALING**

Wound healing is a dynamic and highly regulated process consisting of cellular, humoral and molecular mechanisms (Reinke and Sorg, 2012). The normal cutaneous wound healing process is a temporal process involving a complex series of overlapping events, which can be divided into key stages including: hemorrhage and fibrin-clotformation, inflammatory response, re-epithelialization, granulation tissue formation, angiogenic response, connective tissue contraction, and remodeling, see **Figure 3**.

## **EXPERIMENTAL STUDIES – IRON, ANEMIA, AND WOUND HEALING**

In current literature, animal studies fall into two distinct groups. Early studies investigating the effect of anemia on wound healing used a variety of experimental methodology to establish anemia or iron deficiency. They focused on wound strength studies rather than initial macroscopic healing or histological studies of re-epithelialization. More recent studies have investigated novel treatments aimed at correcting the effect of systemic iron deficiency and topical application of iron-chelators to reduce iron at the specific site of inflammation and in particular their effect on pro inflammatory macrophages.

## **IN VIVO STUDIES – THE EFFECT OF IRON DEFICIENCY ON WOUND HEALING**

Early initial experimental rodent studies used powdered milk diet to establish iron deficiency. Jacobson and Vanprohaska (1965) found that chow-fed control mice showed significantly higher wound breaking-strength than anemic mice that were fed on an iron-free powdered milk diet. Bains et al. (1966) found that young rats fed low iron (powdered milk) diet and subjected to repeated bleeding to produce chronic anemia had weaker wound tensile strength. However, later studies undertaken by Macon and Pories (1971) had contrary findings; iron-deficiency anemia (IDA) had no effect on wound breaking strength. This may reflect the methodological issues when using powdered milk to establish iron deficiency as Waterman et al. (1952) showed that control and anemic rats fed powdered milk had slower wound contraction and reduced wound breaking strength, when compared with animals fed normal chow.

Investigation into the impact of anemia and blood volume on wound healing strength by Sandberg and Zederfeldt (1960) found that replacing blood volume with dextran restored normal wound healing, following acute hemorrhage in a rabbit model. Heughan et al. (1974) found that there was no significant change in woundfluid oxygen tension (PO2) in rabbits made anemic by bleeding and re-transfusing plasma. Additionally, connective tissue weight was greater at lower packed cell volumes, an initial finding that suggested a deleterious effect of hypoxia on collagen synthesis.

Oliveira Sampaio et al. (2013) investigated the effect of delivery of an iron free diet for 15 days in Wistar rats, though histological study of excisional wounds at 7, 14, and 21 days posthealing. LED light caused a significant positive bio-modulation of fibroblastic proliferation in anemic animals, with laser being more effective on increasing proliferation in non-anemic animals. This study did not describe the effect of iron deficiency on the early stages of wound healing (re-epithelialization) or later resolution.

There are various mechanisms by which iron deficiency may impair wound healing. Current evidence favors a key role played by hypoxia. Hypoxia-inducible factor-1 (HIF-1) contributes to all stages of wound healing (through its role in cell migration, cell survival under hypoxic conditions, cell division, growth factor release, and matrix synthesis) and positive regulators of HIF-1, such as prolyl-4-hydroxylase inhibitors, have been shown to be beneficial in enhancing diabetic healing (Hong et al., 2014). Further studies are required to directly answer this question.

Of note, the functional role of iron in the wound healing process has not undergone detailed *in vitro* study. Recent interest in lactoferrin, an iron-binding glycoprotein secreted from glandular epithelial cells, has focused on its role in promoting cutaneous wound healing by enhancing the initial inflammatory phase, and cell proliferation and migration. Takayama and Aoki (2012) found using an in vitro model of wound contraction, lactoferrin promoted fibroblast-mediated collagen gel contraction.

## *IN VIVO* **STUDIES – THE EFFECT OF IRON CHELATORS ON CUTANEOUS WOUND HEALING**

There is considerable variability in iron chelator structure, mechanism of action and their consequent applications. The most widely used iron chelator used to treat iron over-load is deferoxamine, see **Table 1**. Different iron chelators have been applied in studies using a variety of wound healing models.

Early studies of porcine flap necrosis found that intramuscular injection of deferoxamine decreased the percentage of flap necrosis (Weinstein et al., 1989). This study provided some indirect evidence suggesting that iron chelators have a positive effect on wound healing. Mohammadpour et al. (2013) carried out a wound healing study in Wistar rats, primarily assessing macroscopic wound area calculations at days 4, 8, and 12. They found that topical deferiprone treatment accelerated macroscopic wound healing more than Kojic acid, and on the basis of further DPPH scavenging assay suggested that this was due to its higher antioxidant and iron chelation abilities.

Iron chelation results in increased VEGF and HIF 1-α and positive effect on angiogenesis The effect of iron chelation on granulation tissue formation and angiogenesis has not been demonstrated in cutaneous wound healing studies, although there have been some studies of bone tissue in the context of fracture healing (Phelps et al., 1986; Farberg et al., 2012; Donneys et al., 2013). Localized Deferoxamine injection has been shown to both reverse radiation induced hypovascularity and augment vascularity in pathologic fracture healing.

The incorporation of iron chelators in novel wound dressing for human chronic wound treatment has also been described. Wenk et al. (2001) suggested that in human chronic wounds, wound fluid iron-levels are elevated compared with acute wounds. They developed a novel wound dressing based on deferoxamine coupled cellulose and in vitro assays suggested that this dressing may target iron-driven induction of matrix-degrading metalloproteinase-1 and lipid peroxidation. Taylor et al. (2005) also developed and described successful biomechanical testing of deferoxamine coupled polyurethane net substrates.

## **CLINICAL STUDIES – ROLE OF IRON IN HUMAN CUTANEOUS WOUND HEALING**

Human studies in patients with anemia have focused on wound strength. These studies have involved small case-series of patients with a variety of acute surgical conditions. Jonsson et al. (1991) performed a study of 33 patients undergoing subcutaneous implantation of ePTFE graft, collagen deposition was directly proportional to wound oxygen tension and measures of perfusion, although the anemia seen in these patients was not fully described. Pavlidis et al. (2001) carried out a retrospective analysis of 89 patients and found that anemia was not associated with laparotomy wound dehiscence. In a study of 35 normovolaemic anemic patients undergoing skin grafting, Agarwal et al. (2003) found there was no difference in wound healing as assessed by mean split-thickness skin graft take.

To date, human studies have not demonstrated the specific effects of iron deficiency and anemia on the histological stages of chronic wound healing. Clinical studies by our group have found an association between diabetic foot ulceration (DFU) severity and hemoglobin (Hb) decline. DFU is a complex condition, characterized by poor wound healing. Over half all severe DFU patients have IDA (Khanbhai et al., 2012). Clinically the anemia is difficult to characterize; a significant proportion of patients have a functional iron deficiency (FID) caused by chronic inflammation and disruption of the normal Hepcidin mediated iron absorption pathways.

## **DYSREGULATION OF LOCAL CUTANEOUS IRON HOMEOSTASIS IN CHRONIC LEG ULCERATION**

Chronic inflammatory conditions such as rheumatoid arthritis (RA) and Lupus Erythematosus are associated with dysregulation of local cutaneous iron hemostasis.

RA is a progressive inflammatory autoimmune disease, with joint articular and systemic effects including development of ulceration and poor wound healing. The release of cytokines, especially TNF-α, IL-6, and IL-1, causes synovial inflammation. Pro-inflammatory cytokines also promote the development of systemic effects, including production of acute-phase proteins (such as CRP) which in turn may contribute to development dysregulation of iron homeostasis and anemia (Choy, 2012). Indeed, clinical studies of RA patients have reported both iron deficiency anemia and anemia of chronic disease (Bari et al., 2013). Inflammation upregulates the expression of iron-related proteins in the duodenum and monocytes of RA patients (Sukumaran et al., 2013). Evidence for a role for IL-6 signaling in RA is emerging. Isaacs et al. (2013) found that tocilizumab results in an improvement in anemia, reduction in hepcidin/haptoglobin and increase in

#### **Table 1 | Summary of iron chelators.**


As can be seen, iron chelators may have antibacterial/antifungal, anti-inflammatory and skin lightening effects (Porter, 2009). These mixed effects limit the testing of experimental hypothesis surrounding role of iron and iron deficiency in cutaneous wound-healing.

iron-binding capacity. Approximately 10% of patients with RA develop leg ulceration.

Clinically, RA leg ulcers are typically associated with venous insufficiency, trauma, arterial insufficiency and rarely vasculitis (for review, see Rayner et al., 2009). Further work is needed to look at the effect of IAD in patients with RA and leg ulceration.

Lupus Erythematosus is an autoimmune disorder with diverse clinical manifestation ranging from mild cutaneous disorder to a life-threatening systemic illness (SLE). Some patients suffer from a skin-limited form (with a variety of manifestations including oral ulceration), while in others it evolves into SLE, although this process is not fully understood. A key exogenous trigger to the onset of cutaneous disease activity is exposure to UV radiation. It has been shown that photosensitive patients with cutaneous lesions express anti-Ro/SSA autoantibodies. In vivo studies have demonstrated up-regulation of antigens such as Ro52 in keratinocytes (Oke and Wahren-Herlenius, 2013). This is of some interest; it is possible that iron release in

response to UV radiation impairs the function of these antigens, which appear to play a role in negative feedback in response to inflammation.

## **DELETERIOUS EFFECTS OF LOCAL CUTANEOUS IRON DEPOSITION**

There has been some interest in the role of excess iron stored in the skin as hemosiderin, in the pathophysiology of chronic venous disease (CVD). It is now thought that the severe skin changes (such as lipodermatosclerosis) and leg ulceration associated with CVD happen after iron overload occurs. The mechanisms underlying the deleterious effects of local cutaneous iron deposition in CVD are shown in **Figure 4**.

Recent studies by Sindrilaru et al. (2011) and Sindrilaru (2013) have identified a subset of iron-overloaded inflammatory M1 like macrophages, which are implicated in the pathogenesis of CVD. Hb release from extravasated erythrocytes results in a serum haptoglobin/hemoglobin complex that is then taken up by the macrophages, upon upregulation of the hemoglobin–haptoglobin

receptor CD163. In CD163high macrophages, continuous uptake of Hb is thought to be the cause high intracellular concentrations of heme-iron which induce an unrestrained pro-inflammatory macrophage activation. Macrophage iron can be further increased during inflammation by virtue of increased systemic or local hepcidin expression, which leads to reduction in ferroportin, an iron efflux protein, resulting in intracellular iron accumulation. Individuals may also be predisposed to CVD disease through a genetic inability to counteract the skin iron overload (Caggiati et al., 2010). Studies have shown that common hemochromatosis gene mutations such as the C282Y mutation significantly increase the risk of ulcer in CVD by almost seven times (Zamboni et al., 2005).

In hereditary hemochromatosis, plasma iron content increases beyond the iron binding capacity of transferrin, although normal erythropoiesis is occurring. Studies using quantitative nuclear microscopy measurements of iron concentration in the epidermis (which is a readily accessible tissue) have shown that skin iron levels reflect the liver iron overload. Interestingly, this technique has been proposed as a clinical tool to enable better informed decisions on when to initiate, change or stop phlebotomy therapy. In both CVD and hereditary hemochromatosis, parenchymal iron deposition leads to activation of metalloproteinases and subsequently fibrosis. Hereditary hemochromotosis has also been used to study the effects of iron on the aging process. Relative iron overload may also have a deleterious effect on normal skin aging, as iron chelators assist "successful" normal skin aging when applied topically (Polla et al., 2003).

Leg ulceration represents one of the main causes of morbidity in sickle cell anemia (SCA). Known risk factors for leg ulcer development in SCA include Hb (≤6 g/dL), lower levels of fetal Hb, hemolysis, raised lactate dehydrogenase (Lobe et al., 1992), infections and inflammation (Cumming et al., 2008). It is likely that in SCA and other disorders causing hemolytic anemia (such as hereditary spherocytosis, thalassemias, and other hemoglobinopathies) the deleterious effects of excessive local cutaneous or macrophage iron deposition play key roles in poor wound healing.

## **CONCLUSION**

Over recent years there has been some advancement in knowledge about iron in the skin and iron deficiency in cutaneous wound healing. It is clear from studies on pathology of CVD that high iron in macrophages can induce unrestrained proinflammatory macrophage activation. Furthermore in cases of iron deficiency/anemia of inflammation, when serum hepcidin levels are elevated, hepcidin/ferroportin interaction can lead to increased iron concentration in cells particularly macrophage and this could also have a detrimental effect on wound healing. Iron deficiency without inflammation is likely to affect one of the later stages of wound healing such as remodeling. Additional in-depth scientific study of both the underlying pathophysiological mechanisms and role of local cutaneous iron in conditions associated with iron overload and iron deficiency is a priority. Iron is a potential therapeutic target in the skin by application of topical iron chelators and other novel pharmacological agents, and in delayed cutaneous wound healing by treatment of iron deficiency.

## **REFERENCES**


*J. Pediatr. Surg*. 27, 1054–1059; discussion 9–60. doi: 10.1016/0022-3468(92) 90559-P


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

*Received: 04 March 2014; accepted: 16 June 2014; published online: 10 July 2014. Citation: Wright JA, Richards T and Srai SKS (2014) The role of iron in the skin and cutaneous wound healing. Front. Pharmacol. 5:156. doi: 10.3389/fphar.2014.00156 This article was submitted to Drug Metabolism and Transport, a section of the journal Frontiers in Pharmacology.*

*Copyright © 2014 Wright, Richards and Srai. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*

## Mechanisms of iron metabolism in Caenorhabditis elegans

## *Cole P. Anderson and Elizabeth A. Leibold\**

Department of Medicine, Division of Hematology and Hematologic Malignancies and Department of Oncological Sciences, University of Utah, Salt Lake City, UT, USA

#### *Edited by:*

Paolo Arosio, University of Brescia, Italy

#### *Reviewed by:*

Deliang Zhang, National Institute of Child Health and Human Development – National Institute of Health, USA Deborah Chiabrando, University of Torino, Italy

#### *\*Correspondence:*

Elizabeth A. Leibold, Department of Medicine, Division of Hematology and Hematologic Malignancies and Department of Oncological Sciences, University of Utah, 15 N 203 E, Room 3240, Salt Lake City, UT 84112, USA e-mail: betty.leibold@genetics. utah.edu

Iron is involved in many biological processes essential for sustaining life. In excess, iron is toxic due to its ability to catalyze the formation of free radicals that damage macromolecules. Organisms have developed specialized mechanisms to tightly regulate iron uptake, storage and efflux. Over the past decades, vertebrate model organisms have led to the identification of key genes and pathways that regulate systemic and cellular iron metabolism. This review provides an overview of iron metabolism in the roundworm Caenorhabditis elegans and highlights recent studies on the role of hypoxia and insulin signaling in the regulation of iron metabolism. Given that iron, hypoxia and insulin signaling pathways are evolutionarily conserved, C. elegans provides a genetic model organism that promises to provide new insights into mechanisms regulating mammalian iron metabolism.

**Keywords: ferritin, DMT1, SMF-3, iron deficiency, hypoxia, hypoxia-inducible factor, insulin signaling,** *C. elegans*

## **INTRODUCTION**

Iron is essential due to its presence in proteins involved in key metabolic processes such as DNA synthesis, mitochondrial respiration, and oxygen transport. Regulation of cellular iron content is crucial as excess iron catalyzes the generation of reactive oxygen species that damage DNA and proteins, while cellular iron deficiency causes cell cycle arrest and cell death. Disruption of iron metabolism, by iron excess or iron deficiency, leads to common hematological, neurodegenerative, and metabolic diseases (Fleming and Ponka,2012). As a consequence, organisms have developed strategies to sense, transport and store this metal.

Our understanding of the mechanisms that regulate iron metabolism has advanced through the use of model organisms. Physiological and genetic studies in transgenic mice have revealed the mechanism regulating systemic iron metabolism by the ferroportin–hepcidin axis. *Saccharomyces cerevisiae* have been used to unravel the complex pathways involved in Fe-S cluster synthesis (Lill and Muhlenhoff, 2008), while zebrafish have been critical in the identification of genes involved in hematopoiesis (Shafizadeh and Paw, 2004). More recently, the soil nematode *Caenorhabditis elegans* has emerged as a model of iron metabolism. The advantages of *C. elegans* include a short generation time and life span, the feasibility of genetic screens and the opportunity to study physiological processes in a whole organism context*. C. elegans* orthologs have been identified for many human genes (Shaye and Greenwald, 2011) and many of the key genes and pathways regulating mammalian iron metabolism are conserved in *C. elegans.* The genetic tractability of *C. elegans* can provide a complementary approach to mammalian systems to identify novel genes and unravel complex pathways involved in iron metabolism. This review provides an overview of our current understanding of iron metabolism in *C. elegans*, how iron metabolism integrates with oxygen and insulin signaling, and how this genetic model can provide insights in mammalian iron metabolism.

## **CONSERVATION OF IRON METABOLISM IN** *C. ELEGANS*

All organisms must maintain cellular iron content within a narrow range to avoid the adverse consequences of iron depletion or excess. This is accomplished in vertebrates by precise mechanisms that regulate iron uptake, storage and efflux (Andrews and Schmidt, 2007; Zhang and Enns, 2009; **Figure 1**). Mammals acquire iron solely from the diet. Dietary non-heme iron is reduced by membrane bound ferrireductases (e.g. DCYTB, also known as CYBRD) and transported across the apical membrane of intestinal enterocytes by divalent-metal transporter 1 (DMT1, also known as NRAMP2, SLC11A2 and DCT1; Mackenzie and Garrick, 2005; Shawki et al., 2012). Iron is released into a cellular labile iron pool thought to consist of low molecular weight iron complexes. This pool is kept small due to the ability of iron to catalyze the production of reactive oxygen species. Iron is utilized by the mitochondria for Fe-S cluster and heme biosynthesis, and by iron-containing proteins in the cytosol and nucleus. Iron is exported across the basolateral membrane into the circulation by ferroportin (FPN1, also known as SLC40A1, IREG1 and MTP1) in concert with its oxidation by the multicopper oxidase hephaestin (HEPH). Iron enters the circulation where it binds with high affinity to transferrin for delivery to cells expressing transferrin receptor 1 (TfR1, also known as TFRC). TfR1-transferrin-Fe(III) complexes are internalized by receptor mediated endocytosis. Iron is released from transferrin, reduced to Fe(II) by the ferrireductase STEAP3 and transported across the endosomal membrane to the cytoplasm by DMT1. Thus, DMT1 is essential in intestinal non-heme iron absorption as well as transport of endosomal iron released by transferrin into the cytoplasm. Although most cell types express TfR1, erythroid

Anderson and Leibold Iron metabolism in Caenorhabditis elegans

precursors are dependent on Tf-TfR1-DMT1 for iron uptake as disruption of *Tfrc* gene in mice (Levy et al., 1999) or mice with reduced transferrin (Trenor et al., 2000) developed severe anemia. DMT1 mutations in humans (Shawki et al., 2012), the *mk* mouse (Fleming et al., 1997), and the Belgrade rat (Fleming et al., 1998) also cause a severe microcytic hypochromic anemia, underscoring the importance of DMT1 in intestinal and erythroid iron acquisition.

Mammals can also acquire iron by the intestinal absorption of heme iron that comes primarily from animal sources. Although several heme importers have been identified (Yuan et al., 2013), the mechanism regulating intestinal heme import is not well understood. It is likely that heme oxygenase 1 releases iron from dietary heme, which is then exported by ferroportin into the circulation.

When body iron stores are high, cytosolic iron is not exported, and is instead sequestered in ferritin in an inert form unable to catalyze free radical formation (Harrison and Arosio, 1996; Torti and Torti, 2002; Theil, 2011). After 3 days, iron in ferritin is lost by enterocyte sloughing into the intestinal lumen. The regulation of intestinal ferritin is crucial as it serves as a cellular iron "sink" to limit efflux of iron into the circulation (Vanoaica et al., 2010; Galy et al., 2013). Because there is no regulated mechanism for iron excretion, precise regulation of intestinal iron uptake and storage is required. Given the fundamental nature of iron metabolism, it is not surprising that many proteins involved in intestinal iron uptake, storage and export are highly conserved between *C. elegans* and mammals. *C. elegans* express orthologs for DMT1 (SMF-3), ferritin (FTN-1, FTN-2), and ferroportin (FPN-1.1, FPN-1.2, FPN-1.3; **Figure 1**). The *C. elegans* genome also encodes potential orthologs for DCYTB ferrireductase and hephaestin multicopper oxidase. The intestinal anatomy in *C. elegans* is similar to vertebrates in that they contain an apical brush border facing the lumen and a basolateral membrane facing the interstitial space (circulation in mammals) (McGhee, 2013) (**Figure 1**). The intestine serves as the major site for absorption of dietary nutrients and a defense against xenobiotics and pathogens. *C. elegans* lack adipose tissue, liver, and pancreas and the intestine fulfills these functions by serving as a major site of lipid and glucose metabolism. Unlike mammals, *C. elegans* are heme auxotrophs and are dependent on acquiring heme from the environment (Rao et al., 2005; Hamza and Dailey, 2012; Yuan et al., 2013).

SMF-3 is the principal intestinal Fe(II) transporter in *C. elegans*. Consistent with its role in intestinal iron transport, SMF-3 is highly expressed at the apical membrane of intestinal epithelium (Au et al., 2009; Bandyopadhyay et al., 2009), transcriptionally activated during iron deficiency (Romney et al., 2011) and loss of SMF-3 expression leads to reduced iron content in *smf-3(ok1035)* null mutants (Romney et al., 2011). SMF-3 also transports Mn(II) as demonstrated by reduced Mn content in *smf-3(ok1035)* mutants (Romney et al., 2011), increased tolerance of *smf-3(ok1035)* mutants to Mn overload (Au et al., 2009) and Mn-mediated reduction in *smf-3* mRNA and SMF-3 protein in intestine (Au et al., 2009; Settivari et al., 2009). Like SMF-3, DMT1 transports Mn(II), which competes with Fe(II) uptake (Gunshin et al., 1997; Illing et al., 2012). The DMT1-deficient Belgrade rat

**FIGURE 1 | Conservation of intestinal iron metabolism in mammals and** *Caenorhabditis elegans***.** C. elegans anatomy is shown in the left panel. The body plan of C. elegans is made up of two concentric tubes separated by the interstitial space (pseudocoelum). The inner tube consists of the intestine and the outer tube consists of cuticle, hypodermis, muscle and nervous tissue. The digestive tract is an epithelial tube containing the mouth, pharynx (foregut) and intestine (midgut). Right panel, an intestinal epithelial cell is shown with an apical brush border membrane facing the lumen and a basolateral membrane facing the interstitial space or blood in mammals. Mammalian proteins (black) and C. elegans orthologs (red) are indicated. Dietary non-heme iron is reduced by ferrireductases (e.g. DCYTB1) and transported across the apical intestinal membrane by SMF-3/DMT1. Cytosolic iron is incorporated in iron-containing proteins and transported to mitochondria for Fe-S cluster biosynthesis and heme biosynthesis in mammals. C. elegans are heme auxotrophs and are dependent on acquiring heme from the environment (Rao et al., 2005). Iron is also exported across the basolateral membrane into the interstitial space (blood in mammals) by FPN-1.1, FPN-1.2, FPN-1.3/ferroportin. Ferroportin is the sole iron exporter in mammals, whereas C. elegans express three orthologs whose specific functions in iron export are not well understood. Iron export by ferroportin is coupled to the oxidation of iron by the multicopper oxidase hephaestin (HEPH). The C. elegans genome harbors putative DCYTB and HEPH homologs. Iron not utilized or exported is stored in FTN-1, FTN-2/ferritin. Iron deficiency stabilizes C. elegans HIF-1 and mammalian HIF-2α, leading to the transcriptional activation of smf-3/DMT1 to increase iron absorption. Mammalian DCYTB and ferroportin are also activated by HIF-2α during iron deficiency. HIF-1 regulation of C. elegans DCYTB and FPN-1-1, FPN1.2, and FPN-1.3 orthologs remains to be determined. Iron deficiency reduces ferritin abundance in C. elegans and mammals by different mechanisms: C. elegans lack the IRP-IRE network and ferritin is transcriptionally repressed by HIF-1, whereas mammalian ferritin is translationally repressed by IRPs.

displays impaired Mn uptake in intestine and erythroid precursors consistent with a physiological role for DMT1 in Mn uptake in mammals (Chua and Morgan, 1997). In excess, manganese is toxic, and in humans chronic occupational nasopulmonary exposure to Mn causes a neurological disease known as manganism (Roth and Garrick, 2003). Because Mn(II) and Fe(II) compete for DMT1 transport, this suggests that iron deficiency may be an important factor in the predisposition to Mn toxicity. Consistent with this are

studies showing that iron deficiency is associated with increased Mn content in the brain of rats (Chua and Morgan, 1996; Erikson et al., 2002), in the olfactory epithelium of the DMT1-deficient Belgrade rat (Thompson et al., 2007) and in serum of humans with anemia or an iron deficient diet (Davis et al., 1992; Rahman et al., 2013).

*Caenorhabditis elegans* also express DMT1-like proteins SMF-1 and SMF-2 that share about 55–58% amino acid identity with DMT1 (Settivari et al., 2009). SMF-1 is widely expressed, but showed high expression in the apical intestinal membrane (Au et al., 2009; Bandyopadhyay et al., 2009), whereas SMF-2 is mainly cytoplasmic with high expression in pharyngeal epithelium (Au et al., 2009). *smf-3* and *smf-1* are transcriptionally induced upon exposure to pathogenic *Staphylococcus aureus,* and *smf-3(ok1035)*, and *smf-1(ok1748)* mutants showed hypersensitivity to this pathogen, indicating a role for these proteins in innate immunity (Bandyopadhyay et al., 2009). Like *smf-3*, exposure to high Mn reduces *smf-1* and *smf-2* mRNA levels, suggesting that reduced expression of these transporters may be a mechanism to reduce Mn toxicity (Settivari et al., 2009). This is consistent with a study showing that SMF-1 expression in dopamine neurons contributes to Mn2+-mediated neuronal death (Settivari et al., 2009). The roles of SMF-1 and SMF-2 in iron metabolism are not well understood; however, unlike *smf-3* mutant worms, iron and manganese content were not significantly reduced in *smf-1* and *smf-2* mutants compared to wildtype worms consistent with a prominent role of SMF-3 in iron and manganese transport (Romney et al., 2011).

The mechanism regulating basolateral transfer of iron to the interstitial space and to tissues in *C. elegans* is not known. In mammals, ferroportin is the sole exporter of iron to the circulation. *C. elegans* express three ferroportin orthologs, FPN1.1, FPN-1.2, and FPN-1.3, but their specific roles in iron export remains to be determined.

*Caenorhabditis elegans* express genes orthologous to humanferritin heavy subunit (*FTH*) and ferritin light subunit (*FTL*) genes. Ferritin is a ubiquitously expressed protein that stores iron in a form that is unable to generate free radicals. Mammalian ferritin is composed of a mixture of 24 FTL and FTH subunits that form a shell containing up to 4500 iron atoms (Theil, 2013). FTH exhibits ferroxidase activity that facilitates oxidation of iron, while FTL participates with FTH in the nucleation of iron (Bou-Abdallah, 2010; Liu and Theil, 2005). *C. elegans* FTN-1 and FTN-2 are more similar to human FTH than to FTL and both FTN-1 and FTN-2 contain ferroxidase active-site residues (Gourley et al., 2003). *ftn-1* is highly expressed in intestine whereas *ftn-2* is expressed in many tissues such as pharynx, body-wall muscle, hypodermis and intestine (Gourley et al., 2003; Kim et al., 2004). *ftn-1,* and to a lesser extent *ftn-2*, are induced by high iron exposure (Gourley et al., 2003; Kim et al., 2004). Only *ftn-1* mutants are iron sensitive and have reduced lifespans when exposed to high iron (Kim et al., 2004; Valentini et al., 2012).

Iron induces ferritin expression in mammals and in *C. elegans*, but the mechanism regulating ferritin differs in these organisms. In mammals, ferritin is primarily regulated at the translational level by iron-regulatory proteins 1 and 2 (IRP1 and IRP2) (Hentze et al., 2010; Anderson et al., 2012). During iron deficiency, IRPs

bind to an RNA stem-loop known as the iron-responsive element (IRE) in the 5 untranslated regions of *FTH* and *FTL* mRNAs to repress ferritin synthesis. When cellular iron increases, IRP1 is converted to its Fe-S cluster aconitase form concomitant with loss of RNA-binding activity, while IRP2 is targeted for ubiquitination and proteasomal degradation causing ferritin synthesis to increase (Salahudeen et al., 2009; Vashisht et al., 2009). *C. elegans* lack the IRP-IRE system, but express a cytosolic aconitase (ACO-1; Gourley et al., 2003; Kim et al., 2004). ACO-1 is homologous to mammalian IRP1 and its aconitase activity is regulated by iron, but unlike IRP1, it lacks RNA-binding ability. Despite lacking IRP-IRE regulation, *C. elegans* have evolved unique mechanisms to regulate iron storage.

## **HIF-1 REGULATES IRON UPTAKE AND STORAGE DURING IRON DEFICIENCY**

In *C. elegans*, hypoxia signaling is the predominant mechanism for regulating iron metabolism (Romney et al., 2011; Ackerman and Gems, 2012). Hypoxia signaling is a highly conserved process that conditions organisms to low oxygen and iron environments by regulating diverse biologic processes, including glucose metabolism, angiogenesis and iron metabolism (Semenza, 2007; Kaelin and Ratcliffe, 2008). During iron deficiency in mammals, hypoxiainducible factor 2α (HIF-2α, also known as EPAS1) activates the transcription of *DMT1*, *FPN1* and *DCYTB* genes in the intestine to increase iron absorption (Taylor et al., 2011; Mastrogiannaki et al., 2009; Shah et al., 2009). Hypoxia-inducible factors (HIF-1 and HIF-2) are basic helix-loop-helix (bHLH) transcription factors that consist of oxygen-regulated α subunits (HIF-1α and HIF-2α) and a constitutively expressed β subunit (HIF-1β, also known as aryl hydrocarbon nuclear translocator or ARNT) (Semenza, 2007; Kaelin and Ratcliffe, 2008; Kaluz et al., 2008). Under normal conditions, in the presence of oxygen and iron, HIF-α subunits are hydroxylated by prolyl hydroxylase (PHD2, also known as EGLN1) whose activity is dependent upon oxygen and iron. Hydroxylated HIF-α is targeted for proteasomal degradation by the E3 ubiquitin ligase von Hippel Lindau tumor suppressor protein (VHL) (Ivan et al., 2001). During hypoxia or iron deficiency, PHDs are inactive, thus allowing HIF-α subunits to translocate to the nucleus, dimerize with HIF-1β and recruit coactivators to activate target gene expression in pathways such as erythropoiesis, iron metabolism, glucose metabolism and angiogenesis (Semenza, 2007; Kaelin and Ratcliffe, 2008; Kaluz et al., 2008). HIF-1α and HIF-2α regulate overlapping, but distinct sets of target genes (Kaluz et al., 2008). For example, only HIF-2α is responsible for the coordinate upregulation of *DMT1*, *DCYTB* and *FPN1* in intestine during iron deficiency (Mastrogiannaki et al., 2009; Shah et al., 2009; Taylor et al., 2011). HIF-2α regulation of intestinal iron metabolism during iron deficiency ensures that sufficient iron is absorbed and delivered to the bone marrow for production of red blood cells (Shah and Xie, 2014).

The HIF signaling pathway is conserved in *C. elegans*. *C. elegans* express HIF-1, AHA-1, VHL-1, and EGL-9, which are orthologs of HIF-1α/HIF-2α, HIF-1β, VHL and PHD, respectively, in vertebrates (Epstein et al., 2001; Jiang et al., 2001). Unlike mammals, *C. elegans* express a single *hif-1* gene that shares homology to *HIF1*α

and *HIF2*α (Jiang et al., 2001). HIF-1 functions in a variety of biological processes ranging from stress response, innate immunity, neuronal development, ageing and iron metabolism as discussed below (Shen et al., 2005; Chang and Bargmann, 2008; Pocock and Hobert, 2008; Luhachack et al., 2012; Jones et al., 2013).

During iron deficiency, *ftn-1* and *ftn-2* transcription is repressed and is dependent upon a *cis*-regulatory element termed the iron-dependent enhancer (IDE) located in the *ftn-1* and *ftn-2* promoters (Kim et al., 2004; Romney et al., 2008) (**Figure 2**). Basal expression of *ftn-1* and *ftn-2* is mediated by the intestinal GATA transcription factor ELT-2 that binds GATA sites located in ferritin IDEs (Romney et al., 2008). Further studies revealed that HIF-1 binds to hypoxia response elements (HREs) located in the IDEs of *ftn-1* and *ftn-2* to repress transcription during iron deficiency (Romney et al.,2011;Ackerman and Gems,2012). Intestinal iron uptake through SMF-3 is also regulated by HIF-1 during iron deficiency. Similar to *ftn-1* and *ftn-2* IDEs, *smf-3* contains an IDE in its promoter that contains HRE binding sites that confer HIF-1 dependent activation during iron deficiency (Romney et al., 2011) (**Figure 2**). Romney et al. (2011) also showed that *hif-1 (ia04)* mutants have reduced iron and manganese content and are developmentally delayed when grown in iron deficient conditions.

Notably, development of *hif-1(ia04)* mutants was restored when the cellular iron pool was increased by RNAi depletion of *ftn-1* and *ftn-2.* It is not known whether the ferroportin homologs *fpn-1.1, fpn-1.2* and *fpn-1.3* and *DCYTB* homologs are regulated by hypoxia. These studies show that regulation of iron uptake and storage by HIF-1 is crucial for ensuring proper growth and development during iron deficiency.

HIF-1 is well known as a transcriptional activator but less is known about its role as a transcriptional repressor. The question arises regarding the mechanism of HIF-1 transcriptional repression of *ftn-1* and *ftn-2.* Chromatin immunoprecipitation analysis and electrophoretic mobility gel assays showed direct HIF-1 binding to the *ftn-1* IDE (Romney et al., 2011; Ackerman and Gems, 2012). Another study showed that mutations of all three HREs in the *ftn-1* IDE abolished expression of a *pftn-1::gfp* transcriptional reporter, suggesting that an activator may bind the HREs during normal conditions (Romney et al., 2011; **Figure 2**). HREs resemble E-box elements and it is possible that this activator may be a member of the basic helix loop helix (bHLH) transcription factor family that can bind to non-canonical E-boxes (Kewley et al., 2004). A MAD-like transcription factor MDL-1 was identified in an RNAi screen as a

transcriptional activator of *ftn-1* expression (Ackerman and Gems, 2012). *mdl-1* encodes a bHLH transcription factor that bind E-box sequences as a dimer with MXL-1 to regulate target genes(Yuan et al., 1998). MDL-1 transcriptional regulation of *ftn-1* was shown to be iron independent (Ackerman and Gems, 2012), suggesting the possibility that MDL-1 may bind to the *ftn*-1 and *ftn-2* HREs when iron is sufficient, but is displaced by HIF-1 when iron is low. Alternatively, it is possible that during iron deficiency the displacement of ELT-2 from its GATA binding sites by HIF-1 results in decreased *ftn-1* and *ftn-2* transcription. Further work is required to define this mechanism. In mammals, ferritin has not been reported to be regulated by HIF-2α; however, hypoxia regulates ferritin expression by altering IRP1 RNA binding activity and IRP2 protein abundance (Schneider and Leibold, 2003; Meyron-Holtz et al., 2004; Salahudeen et al., 2009; Vashisht et al., 2009).

## **FERRITIN REGULATION BY THE INSULIN/INSULIN-LIKE GROWTH FACTOR SIGNALING PATHWAY**

Ferritin is regulated by the insulin/insulin-like (IIS) growth factor signaling pathway in *C. elegans*. The IIS pathway is a conserved pathway in vertebrates and *C. elegans* that coordinates nutrient availability with development, metabolism and stress responses (Accili and Arden, 2004; **Figure 3**). When nutrients are available, insulin and insulin-like growth (IGF) factors activate tyrosine kinase receptors DAF-2/IGFR1, triggering a kinase cascade that leads to the phosphorylation of the Forkhead box, Class O (FOXO) transcription factor DAF-16/FOXO and its cytoplasmic retention and inhibition. When IIS is reduced during nutrient deprivation, DAF-16/FOXO phosphorylation is reduced, promoting DAF-16/FOXO translocation to the nucleus where it regulates the expression of target genes involved in stress resistance, metabolism, and innate immunity (Murphy and Hu, 2013). A recent study showed that *ftn-1* expression was elevated in *daf-2* mutants compared to *daf-16;daf-2* mutants, indicating that DAF-16 activated *ftn-1* expression (Ackerman and Gems, 2012). Further genetic studies showed that *hif-1* and *daf-16* act in parallel pathways to regulate *ftn-1* and that DAF-16 regulation of *ftn-1* was not iron dependent (Ackerman and Gems, 2012). Less is known about the role of IIS in *smf-3* regulation. One study showed that glucose treatment induced the *smf-3* expression, suggesting a potential role for IIS and DAF-16 in *smf-3* downregulation (Lee et al., 2009). Reduced IIS leads to DAF-16 dependent upregulation and downregulation of a diverse set of genes, which are designated as class 1 and class II genes, respectively (Lee et al., 2003; McElwee et al., 2003; Oh et al., 2006). More recently, the transcription factor PQM-1 was discovered to regulate class II genes by binding to the DAF-16 associated element (DAE) located in the promoter of these genes, whereas DAF-16 regulates class 1 genes by binding to the DAF-16 binding element (DBE; Tepper et al., 2013). The *smf-3* promoter contains both DBE and DAE binding sites, but whether DAF-16 or PQM-1 regulates *smf-3* awaits future studies. Taken together, these studies suggest that DAF-16 activation of *ftn-1* during reduced IIS provides *C. elegans* with a mechanism to increase iron storage, thereby limiting iron toxicity during stress conditions (**Figure 3**). When IIS is stimulated, DAF-16 is inhibited and *ftn-1* transcription is reduced,

deficiency or hypoxia and results in HIF-1 dependent repression of ftn-1 and ftn-2 and activation of smf-3. This leads to reduced iron storage and increased iron uptake, ultimately increasing the cellular iron pool and promoting survival during iron limitation. Insulin/insulin growth factor-1 (IGF-1) signaling (IIS) pathway*:* the IIS pathway is a conserved pathway in worms, flies and in vertebrates that regulates the transcription factor DAF-16/FOXO. DAF-16/FOXO regulates genes in essential processes, such as metabolism, stress resistance, pathogen defense and lifespan extension. Activated IIS initiates a phosphorylation cascade that leads to the cytoplasmic retention of DAF-16/FOXO and inhibits its function. Nutrient deprivation reduces IIS leading to the nuclear localization of DAF-16 to activate ftn-1. Glucose upregulates smf-3 expression, but whether DAF-16 directly regulates smf-3 remains to be determined. This pathway provides a mechanism to increase cellular iron by reducing ftn-1 when nutrients are abundant to promote growth and to reduce cellular iron during stress by increasing ftn-1 to limit iron-catalyzed oxidative stress.

increasing the availability of iron required for development and growth

Insulin signaling and FOXO regulation of mammalian ferritin has not been reported. However, mammalian ferritin is transcriptionally activated by oxidative stress (Thimmulappa et al., 2002; Pietsch et al., 2003a,b; Hintze and Theil, 2005) and repressed by oncogenes, providing a mechanism to sequester iron during stress and to increase iron availability during cell proliferation (Tsuji et al., 1993, 1995; Wu et al., 1999). Similarly, several studies have shown that ferritin depletion stimulates cell proliferation by increasing available iron, whereas sequestration of iron by ferritin overexpression slows cell proliferation (Cozzi et al., 2000; Kakhlon et al., 2001; Cozzi et al., 2004; Baldi et al., 2005). Like *C. elegans*, changes in ferritin expression in response to environmental stimuli are essential for survival during stress and growth during normal conditions.

## **OTHER REGULATORS OF FERRITIN EXPRESSION**

*ftn-1* transcription has also been shown to be repressed by the REF-1-like protein HLH-29 (Quach et al., 2013) and UNC-62 a member of the TALE family of homeobox transcription factors (Catoire et al., 2011; Ackerman and Gems, 2012). HLH-29 is homologous to the HairyVEnhancer of Split (HES) transcription factors that regulate embryonic development through Notch-dependent and independent pathway (Fischer and Gessler, 2007). HLH-29 was recently shown to bind promoter sequences upstream of the *ftn-1* IDE and repress its transcriptional expression independent of the iron responsive HIF pathway (Quach et al., 2013). Additionally, *hlh-29* mutants have elevated levels of *ftn-1* and are resistant to peroxide stress. Further studies are needed to define the mechanism and significance of this regulation.

*unc-62* encodes the mammalian ortholog of MEIS1 that has a crucial role in normal development and in leukemia (Azcoitia et al., 2005; Argiropoulos et al., 2007). MEIS1 has also been identified as a Restless Leg Syndrome (RLS) predisposing gene (Winkelmann et al., 2007; Xiong et al., 2009; Spieler et al., 2014). RLS is a sensorimotor disorder that is associated with iron insufficiency in brain, but the role of iron in RLS is not well understood (Clardy et al., 2006; Allen and Earley, 2007; Catoire et al., 2011). It is of interest that *ftn-1* expression is significantly decreased in *C. elegans* treated with *unc-62* RNAi (Ackerman and Gems, 2012), suggesting that dysregulation of MEIS-1/MEIS can lead to altered iron metabolism.

Ferritin regulation spans beyond iron and nutrient stress. For instance, *ftn-2*, but not *ftn-1*, was shown to be necessary for proper innate immune response to pathogenic *S. aureus* (Simonsen et al., 2011). During infection, *ftn-2* was also transcriptionally upregulated along with several DAF-16 targets. It is likely that DAF-16 activates *ftn-2* to protect *C. elegans* from bacterial infection by limiting iron availability.

### **CONCLUDING REMARKS**

We have highlighted recent studies showing the potential of *C. elegans* as a useful genetic platform to explore mechanisms integrating iron and oxygen metabolism. Future genomic studies are needed to identify additional target genes of HIF-1 that are specific to hypoxia or iron deficiency and the unique HIF-1 partner proteins that coordinate these responses. A better understanding of how iron and insulin signaling are coordinated in *C. elegans* could provide new knowledge about the role of iron in glucose metabolism and in the pathogenesis of diabetes in humans. Finally, these studies have set a foundation for the development of genetic screens to identify novel regulators that are involved in iron sensing, uptake, storage and utilization. *C. elegans* holds promise as a system to decipher complex pathways regulating iron metabolism that can be followed up in mammals.

#### **ACKNOWLEDGMENTS**

This work was supported by the National Institute of Health grants R01GM45201 and R01DK068602 to Elizabeth A. Leibold. Cole P. Anderson was supported by National Institute of Health Research Training in Hematology grant T32DK007115. We thank Dr. Paul Rindler for critical comments on the manuscript and Ms. Diana Lim for help with illustrations.

#### **REFERENCES**

Accili, D., and Arden, K. C. (2004). FoxOs at the crossroads of cellular metabolism, differentiation, and transformation. *Cell* 117, 421–426. doi: 10.1016/S0092- 8674(04)00452-0


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

*Received: 18 March 2014; paper pending published: 05 April 2014; accepted: 28 April 2014; published online: 21 May 2014.*

*Citation: Anderson CP and Leibold EA (2014) Mechanisms of iron metabolism in Caenorhabditis elegans. Front. Pharmacol. 5:113. doi: 10.3389/fphar.2014.00113 This article was submitted to Drug Metabolism and Transport, a section of the journal*

*Frontiers in Pharmacology.*

*Copyright © 2014 Anderson and Leibold. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*

## Labile iron in cells and body fluids: physiology, pathology, and pharmacology

## *Zvi Ioav Cabantchik\**

Department of Biological Chemistry, Institute of Life Sciences, Hebrew University of Jerusalem, Jerusalem, Israel

#### *Edited by:*

Paolo Arosio, University of Brescia, Italy

#### *Reviewed by:*

Stanislav Yanev, Bulgarian Academy of Sciences, Bulgaria Dario Finazzi, University of Brescia, Italy

#### *\*Correspondence:*

Zvi Ioav Cabantchik, Department of Biological Chemistry, Institute of Life Sciences, Hebrew University of Jerusalem, Safra Campus at Givat Ram, Jerusalem 91904, Israel e-mail: ioav@cc.huji.ac.il

In living systems iron appears predominantly associated with proteins, but can also be detected in forms referred as labile iron, which denotes the combined redox properties of iron and its amenability to exchange between ligands, including chelators.The labile cell iron (LCI) composition varies with metal concentration and substances with chelating groups but also with pH and the medium redox potential. Although physiologically in the lower μM range, LCI plays a key role in cell iron economy as cross-roads of metabolic pathways. LCI levels are continually regulated by an iron-responsive machinery that balances iron uptake versus deposition into ferritin. However, LCI rises aberrantly in some cell types due to faulty cell utilization pathways or infiltration by pathological iron forms that are found in hemosiderotic plasma. As LCI attains pathological levels, it can catalyze reactive O species (ROS) formation that, at particular threshold, can surpass cellular anti-oxidant capacities and seriously damage its constituents. While in normal plasma and interstitial fluids, virtually all iron is securely carried by circulating transferrin (Tf; that renders iron essentially non-labile), in systemic iron overload (IO), the total plasma iron binding capacity is often surpassed by a massive iron influx from hyperabsorptive gut or from erythrocyte overburdened spleen and/or liver. As plasma Tf approaches iron saturation, labile plasma iron (LPI) emerges in forms that can infiltrate cells by unregulated routes and raise LCI to toxic levels. Despite the limited knowledge available on LPI speciation in different types and degrees of IO, LPI measurements can be and are in fact used for identifying systemic IO and for initiating/adjusting chelation regimens to attain full-day LPI protection. A recent application of labile iron assay is the detection of labile components in intravenous iron formulations per se as well as in plasma (LPI) following parenteral iron administration.

**Keywords: iron, iron metabolism, chelator, siderophore, mitochondria, iron overload, oxidative stress, fluorescence**

## **INTRODUCTION**

The various forms of iron present in biological fluids are largely determined by the chemical composition of the medium, particularly its reductive power (commonly dictated by the GSH-NADPH/NADH levels) and the repertoire of substances with metal complexing groups (e.g., carboxylates, phosphates, amides, thiolates, and hydroxylates). In extracellular fluids, essentially all the iron is safely carried by the protein transferrin (Tf) that shields the Fe(III) from the environment and renders it virtually redox-inactive as well as non-exchangeable with, or displaceable by, physiological substances or metals. Thus Tf-bound iron (TBI) is a tightly bound form of ferric iron that by definition is *non-labile* and it remains as such, until conformational changes triggered by binding of Tf to its cognitive Tf receptor (TfR), in conjunction with medium acidification, jointly lead to its release and ensuing reduction to ferrous iron (Crichton, 2001; Cairo and Recalcati, 2007). That complex series of reactions is mechanistically integrated in the physiological process of receptor mediated endocytosis (RME) of TBI that operates in all mammalian cells and serves as the key route of regulated iron uptake (**Figure 1**).

In cellular compartments almost all (>95%) of the iron is also protein-bound, either directly to protein residues or via ironcontaining prosthetic groups such as heme or iron sulfur clusters (ISC; Crichton, 2001). Although most of the iron-containing proteins are endowed with catalytic metal centers, for this exposé we consider the redox-active metal as stricto senso labile if it is also physiologically exchangeable and/or pharmacologically chelatable. This fits with the more traditional definition of labile cell iron (LCI) that was introduced in order to describe transitory (i.e., exchangeable) forms of cell iron that are important for cell iron metabolism and homeostasis (Jacobs, 1977; Crichton, 2001). We describe here LCI in physiological and pathophysiological states and its potential relevance to pharmacology and therapeutics. Methodological aspects related to the analysis of LCI and its counterpart in plasma, labile plasma iron (LPI), are the subject of other reviews (Esposito et al., 2002; Kakhlon and Cabantchik, 2002; Petrat et al., 2002; Kruszewski, 2003; Breuer et al., 2007; Brissot et al., 2012).

## **LABILE CELL IRON**

Many attempts made to chemically define LCI in extracts from lysed cells led biochemists to implicate various Fe3<sup>+</sup> complexes as potential LCI candidates [e.g., nucleotides and even glutathione (GSH); Weaver and Pollack, 1989; Hider and Kong, 2011]. However, those implications should be interpreted with caution as changes in environments (including pH and redox potential) that result from cell de-compartmentalization are likely to result in metal redistribution among potential metal binders. Therefore, in this review we regard LCI as a dynamic parameter that is relevant only for living cells and that prevails under defined conditions, spatial, temporal and environmental (Cabantchik et al., 2009).

## **PHYSIOLOGY**

The most compelling evidence that LCI prevails in physiological conditions in most mammalian cells, is the demonstrable ability of permeating iron chelators to inhibit the *in situ* (cell or organelle) catalytic contribution of labile iron to basal as well as peroxide-stimulated cell reactive O species (ROS) formation and to inhibit iron-dependent cell functions (Glickstein et al., 2005). LCI is *per se* a generic term used to describe labile iron in the cell as a whole, or in particular cell compartments. The cytosolic and organellar components of LCI are comprised of chelatable complexes of both Fe2+and Fe3+, whose relative proportions largely reflect (or are dictated by) the reductive properties and chemical composition of the compartment in question and those in turn depend on cell metabolism (Petrat et al., 2002; Breuer et al., 2007).

The cytosolic LCI, that is "strategically" located at the crossroads of cell iron metabolism, serves both as metabolic source of metal but also as indicator of cell iron levels. Thus, cells sense and regulate LCI by balancing the uptake of circulating TBI with the storage of unutilized (surplus?) cell iron in shells of ferritin molecules (**Figure 1**). As expected for a dynamic cell parameter, the various LCI pools are likely to vary over time in response to chemical or biological stimuli as well as to metabolic demands/responses. For example, in human K562

or murine erythroleukemia cells, exposure to increasing TBI concentration results in a commensurate rise in cytosolic and mitochondrial LI indicating increased iron uptake by RME and ensuing release into cytosol (and almost concurrent delivery to mitochondria; Shvartsman et al., 2007, 2010; Levi and Rovidac, 2009). Similar results are obtained by modulation of cell ferritin expression (Picard et al., 1998; Kakhlon et al., 2001). Conversely, LCI levels fall in iron starvation, following stable or transient overexpression of cytosolic or mitochondrial ferritin as well as in some mitochondrial disorders that involve aberrant mitochondrial iron accumulation (Rouault, 2006, 2012; Levi and Rovidac, 2009).

The search for putative LCI components have led also to complementary searches for LCI counterparts, namely putative iron chaperons or carriers that, in analogy with copper chaperons, might mediate "safe" iron traffic within cells via formation of non-labile or occluded metal complexes. Accordingly, PBP proteins were implicated as facilitators of iron incorporation into ferritin (Shi et al., 2008) and 2,4 dihydroxybenzoate (2,4-DHB) as universal cytosol-mitochondrial iron chaperon (Devireddy et al., 2010). It must be stressed however, that: (a) neither the putative presence of nM concentrations of 2,4-DHB, seemingly detected in mammalian cells, nor (b) the almost undetectable chelating ability of 2,4-DHB *per se* in buffered medium, let alone (c) in a cellular milieu that is comprised of mM concentrations of organophosphates and -carboxylates, could possibly confer upon DHB a putative role as cell iron chaperon, carrier, or the like (Shvartsman and Cabantchik, 2012). For hemoglobin synthesizing reticulocytes, it was hypothesized that iron derived from uptake of TBI by RME is delivered to mitochondria by a mechanism referred as "kiss and run" (Zhang et al., 2005; Sheftel et al., 2007). In that mechanism, vesicle-mitochondria interactions were proposed to provide bridges or channels for trans-vesicular transfer of Fe2<sup>+</sup> to cytosol.


**FIGURE 1 | Labile cell iron (LCI) and labile plasma iron (LPI).** Proposed routes of labile iron ingress into cells comprise voltage dependent Ca channels (VDCC; Oudit et al., 2006), the Zn transporter ZIP14 (Liuzzi et al., 2006) and others (Sohn et al., 2012).

In live-cell fluorescence studies done on human K562 erythroleukemia or murine erythroleukemia (MEL) cells, changes in cytosolic LCI levels occurred in response to increasing concentrations of TBI (but also non-TBI (NTBI) compounds), and almost in parallel to changes in mitochondrial LCI (Shvartsman et al., 2007, 2010). Those studies indicated the operation of various routes of cell iron traffic, most of which are manifested as transient changes in the levels of the cytosolic LCI pool. Observations describing a rapid delivery of iron from the point of entry into cells to mitochondria have also been observed in cells exposed to various NTBI substances (e.g., organic iron salts; Shvartsman et al., 2007). Moreover, in the intestinal CACO-2 cell model, iron trafficked intracellularly from mucosal to serosal cell phases appeared to be shielded in dimetal transporter 1 (DMT1)-containing endocytic vesicles (Núñez et al., 1994; Núñez et al., 1999; Linder et al., 2006). Hitherto, the demonstration of "cell iron passages" as "safe and efficient" routes of iron delivery mediated by cell chaperons or vesicles, awaits experimental support. However, their possible contribution to cell iron physiology does not exclude cellular iron traffic also (or primarily) via LCI pools, whose levels are dynamically monitored live with metalsensing or redox-sensing probes in response to physiological challenges.

## **PATHOLOGY**

It is generally accepted that a major and persistent rise in LCI levels can compromise cell integrity, since excessively accumulated labile iron is prone to engage in the catalytic generation of noxious ROS from reactive O intermediates (ROIs) and those can override the cell antioxidant defenses (**Figure 1**; Kruszewski, 2003). These reactions occur primarily in mitochondria, which

not only tend to accumulate excessive LI in various disorders but also suffer the consequences of local ROS formation and ensuing damage. In the various types of systemic siderosis (primary or transfusional), the etiopathology of iron overload (IO) is classically associated with the ability of components of the hemosiderotic plasma to infiltrate cells and raise the LCI levels in both cytosol and mitochondria (Kakhlon and Cabantchik, 2002; Cabantchik et al., 2009). However, a unique feature in inherited mitochondrial disorders caused by faulty biosynthesis/assembly of heme or ISC is the under-utilization of iron that results in mitochondrial iron accumulation/deposition and ensuing cytosolic iron deprivation (Rouault, 2006, 2012; Breuer and Cabantchik, 2009; Camaschella, 2009; Li-Hsuan Huang et al., 2011). That scenario of iron maldistribution (that might be causatively interrelated) is promoted by a vicious circle of increased cell iron uptake induced by a reduction in cytosolic LCI (Li-Hsuan Huang et al., 2011). The latter in turn induces TfR expression via activation of the iron-regulatory IRP-IRE system (Cairo and Recalcati, 2007), leading to increased TfFe uptake and ensuing iron deposition in mitochondria. The resulting phenotypes of regional siderosis are demonstrated in sideroblastic anemia (SA) caused by the mutated gene *alas2* (for XLSA), or *glrx5* (for SA) or *slc25a38* (for SA) but also in neuro-siderosis caused by the mutated gene fxn (for *FRDA*) or *abcb7* (for x-linked SA with ataxia, xlsa/a; Rouault, 2006, 2012; Camaschella, 2009; Li-Hsuan Huang et al., 2011).

## **PHARMACOLOGY**

To the extent that a persistently elevated LCI is a risk factor for cell survival (Li-Hsuan Huang et al., 2011; Fleming and Ponka, 2012), metal detoxification by chelation should be regarded as

the most direct mode of pharmacological intervention (Hershko et al., 2005; Pietrangelo, 2007; Porter, 2009; Ma et al., 2012; Camaschella, 2013). However, such an approach should be endowed with specificity for labile forms of iron and designed with an adequate regimen so as not to generate long term metal deficiency, local or systemic. In principle, the major goal of chelation is the neutralization or attenuation of iron propensity for catalyzing radical formation. In order for a chelator to meet that goal in an iron overloaded tissue, it must be permeant to cells and endowed with an effective binding affinity for the labile metal (to ensure specificity) and a mode of coordinating complexation that should render the metal essentially non-labile. Those are exemplified by: (a) the tridentate chelator deferasirox (DFR) that binds Fe3<sup>+</sup> with 2:1 stoichiometry and (b) the bidentate deferiprone (DFP) that binds it with 3:1 stoichiometry. As in systemic siderosis, the targets of chelation are both within cells and in extracellular fluids, the goal of chelation is twofold: (a) metal detoxification by means of reducing the iron burden in overloaded tissues and (b) prevention of tissue iron accumulation, by maintaining a low level of LPI, namely the labile components of plasma NTBI (**Figure 2**).

Unlike in systemic siderosis, in regional siderosis, iron is generally maldistributed among cells or cell organelles, often accumulating in some at the expense of iron-sufficient ones. Scavenging of accumulated iron by chelation might accomplish regional detoxification but also concurrently generate deprivation in iron-sufficient or iron-deficient regions (e.g., tissues, cells, or organelles; Breuer and Cabantchik, 2009). Thus, for regional siderosis, treatment should not be limited to scavenging of surplus iron but should be followed by its redeployment, either within or across cells, as depicted in **Figure 2**. Such mode of metal redistribution can be accomplished by chelators with siderophore properties, namely membrane permeant chelators that have a combined accessibility and iron affinity for scavenging LCI in cell and organelles but also the ability to transfer the chelated metal to cell acceptors or the plasma iron acceptor Tf (Kakhlon and Cabantchik, 2002; Breuer and Cabantchik, 2009). A paradigm for that modus operandi is given by DFP, that in addition of scavenging and redeploying iron between cells and extracellular Tf, it has also the ability to correct iron maldistribution *per se* (Sohn et al., 2008, 2011; Breuer and Cabantchik, 2009; Kakhlon et al., 2010; e.g., between mitochondria and cytosol, as found in cells rendered frataxin deficient by shRNA suppression technology) as well as most of the affected cell properties (Kakhlon et al., 2008, 2010).

#### **DETERMINATION OF LCI IN LIVING CELLS**

The determination of LCI in living cells has been based on spectroscopic probes (herewith referred as fluorescence metal sensors FMS) that detect labile iron *in situ* and in real time by one of the following mechanisms (Petrat et al., 2002; Glickstein et al., 2006; Ma et al., 2006; Breuer et al., 2007): (a) reversible quenching of fluorescence upon binding of iron to a probe endowed with a fluorescent-tagged chelating unit or (b) generation of fluorescence by metal-catalyzed oxidation of a fluorogenic probe (Glickstein et al., 2006; Ma et al., 2006). Changes in fluorescence signal are attributed to labile iron if they are prevented or reversed by a strong and permeant iron chelator that can swiftly gain

access to cell compartments and scavenge LCI as well as FMSchelated iron (!). FMSs like calcein blue (CALB as shown in **Figure 3**) or calcein green (CALG as shown in **Figure 4**) are comprised of one or two metal chelating arms (aminodiacetate or iminodiacetate) linked to fluorescent probes (e.g., fluorescein or methylumbelliferone) that undergo signal quenching upon binding of transition metals like Fe or Cu. The probes can be easily loaded into cells via membrane permeant precursors that are nonfluorescent and non-chelating, e.g., CALG-AM or CALB-AM; AM-acetomethoxy. These probes are hydrolyzed intracellularly,

CALG (d) (loaded into the cell cytosol via the CALG-AM precursor) and red RPA (f) (loaded into cell mitochondria by potentiometric distribution) (a) were monitored with time by fluorescence microscopy imaging. Following addition of 5 μM ferric-hydroxyquinoline (1:1) for 10 min (b) the fluorescence intensity (in relative units) in both cytosol and mitochondria dropped substantially (from a to b and a to b for CALG and RPA, respectively). The reduction in the fluorescence signal elicited by the

releasing the fluorescent, impermeant chelating CALG (binds iron 1:1) or CALB (binds iron 2:1) that interact (reversibly) with resident labile iron and undergo quenching commensurate with LCI concentrations. That mode of iron sensing is depicted schematically at the bottom of **Figures 3** and **4** for CALB and CALG, respectively, whereby iron added to the free CALB or CALG quenches the FMS and addition of a relatively high concentration of a strong iron chelator restores the quenched fluorescence. In CALB-laden human hepatic HepG2 cells (**Figure 3**), the FMS fills the cytosolic space and binds a fraction of LCI that can be revealed upon addition of excess permeant chelator. Thus addition of DFR elicits a rise in fluorescence -F that corresponds to Fe released form the quenched CALB-Fe the chelator (-F can be converted to actual concentration of LCI with the aid of appropriate calibration curves as demonstrated earlier for CALG (Epsztejn et al., 1997; Cabantchik et al., 2002; Petrat et al., 2002; Breuer et al., 2007).

The principle of LCI measurement demonstrated with the FMS CALB can also be used to assess dynamic changes in various LCI pools with organelle targeted FMS (Glickstein et al., 2005; Shvartsman et al., 2007, 2010; Sohn et al., 2008; Porter, 2009; Cabantchik et al., 2013). **Figure 4** depicts h9c2 cardiac cells double labeled with CALG (cytosol and nucleus) and RPA, a red rhodamine-phenanthroline iron-sensitive probe that targets potentiometrically (binds iron 3:1) to mitochondria. Addition of a permeant iron source reduces the fluorescent intensity of both CALG and RPA (indicating ingress of labile iron to the respective

pentaacetic acid (to chelate all extracellular Fe). Subsequent addition of 50 μM of the permeant chelator DFP, led to the recovery CALG fluorescence quenched by the added Fe as well the original LCI, denoted by -DFP. On the other hand, the quenched RPA signal was not recovered by addition of DFP, due to the relatively poor reversibility of phenanthroline-complexed metal in physiological conditions. g depicts the principle of CALG use as FMS.

compartment) and subsequent addition of the permeant chelator DFP restores the quenched fluorescence signal of CALG, but not that of RPA (that binds iron tightly and demands extreme conditions for scavenging the bound metal from the phenanthroline:metal 3:1 complex). Parallel time-dependent changes in cytosolic and mitochondrial LCI compartments can be monitored by flow cytometry or fluorescence microscopy as means to assess iron transport into the respective compartments in response to addition of TBI or NTBI (Shvartsman et al., 2007, 2010). As demonstrated in **Figure 5** for TBI, the relative timedependent changes in fluorescence intensity obtained for either CALG or RPA by flow cytometry match those obtained independently by fluorescence microscopy imaging. **Figure 5** depicts both the resident cytosolic LCI prior to and after addition of TBI (labeled LCI and LCI+, respectively) that are revealed as changes in CALG fluorescence after addition of the permeant tridentate chelator (salicylaldehyde isonicotinoyl hydrazone SIH). When time-dependent profiles are done on probe labeled cells exposed to different substrate concentrations, it is also possible to obtain the kinetic parameters of iron ingress into the respective LCI pools (cytosolic and mitochondrial), by converting ΔF to LCI concentrations, as described elsewhere (Shvartsman et al., 2007, 2010).

Various types of FMS have been applied to cells as sensors for monitoring changes in LCI by following stoichiometry quenching of fluorescence (Rauen et al., 2007; Prus and Fibach, 2008a,b; Ma et al., 2012). Among those are probes based on phenanthroline,

F obtained by addition

deferrioxamine, or hydroxypyridinones chelating moieties and others originally designed for monitoring intracellular Ca. The major shortcoming of most those probes is the difficulty in reversing the quenched signal in physiological conditions and thereby the difficulty in unequivocally assigning the quenched signal to *in situ* LCI.

fluorescent calG is shown; arrow indicated median F values). The right panel shows values of F at different time points (following addition of

Targeted FMSs have also been used to demonstrate intracellular redistribution of iron (scavengery and redeployment) by addition of the chelator siderophore DFP in a cell model of iron maldistribution (Kakhlon et al., 2008; Sohn et al., 2008, 2011). As depicted in **Figure 6**, h9c2 cells, were pre-labeled with the FMSs RPA (red, for mitochondria) and the fluorescence quenched histone-CALG-Fe (green, for the nucleus) and incubated with the bidentate chelator DFP. As schematized in the lower panel, an increase in nuclear fluorescence elicited by added DFP denotes iron scavengery from the CALG-Fe moiety, whereas a decrease in RPA fluorescence denoted DFP-mediated redeployment of Fe to mitochondria. As some DFP-Fe might also egress from cells into a medium containing apo-Tf, iron can redeployed to Tf, that in turn can also deliver the metal to other cells. This is depicted in the right panel, whereby the medium from the h9c2 cells treated with DFP offered to iron-deficient erythroid MEL cells supported hemoglobin Hb synthesis.

An additional mode of monitoring LCI in living cells is by assessing the ability of labile iron to catalyze oxidation of fluorogenic analogs of dihydrofluorescein or dihydrorhodamine loaded into cells. In the presence of ROIs (spontaneously generated in the cell or prompted by addition of oxidants like H2O2 or other peroxides) the labile metal converts *in situ* the non-fluorescent dihydro-probes into the respective fluorescent probes. As with the other FMS, the contribution of LCI to the time-dependent generation of fluorescence is assessed by the degree of inhibition attained by addition of a permeant chelator (Glickstein et al., 2006; Ma et al., 2006).

## **LPI AND NTBI**

## **PATHOLOGY AND PHARMACOLOGY**

CALG fluorescence intensity commensurately. The -

of SIH provides a measure of LCI.

In systemic IO, plasma is the compartment that harbors the pathophysiological source of uncontrolled iron ingress into cells and ensuing tissue IO (Hershko et al., 1978, 2005; Graham et al., 1979; Pietrangelo, 2007). The most implicated component of that source is LPI (Graham et al., 1979), the plasma counterpart of LCI that represents the labile fraction of plasma NTBI (Breuer et al., 2000; Esposito et al., 2003). As LPI comprises the redoxactive and exchangeable forms of iron in native plasma that are also direct pharmacological targets of chelation (Pootrakul et al., 2004), it represents a parameter with potential diagnostic as well as therapeutic (i.e., theragnostic) value (Cabantchik et al., 2013). The generic term NTBI was originally introduced to denote the nonphysiological, low molecular weightforms of iron that is not tightly associated with Tf and appears in plasma of IO patients (Hershko et al., 1978; Graham et al., 1979). Those forms were postulated on the basis that iron levels in plasma of IO patients often supersede total iron binding capacity (TIBC). Since a substantial fraction

of NTBI is apparently adsorbed to plasma proteins, its chemical detection has posed technical difficulties in native plasma/serum unless pretreated with metal mobilizing agents like the polycarboxylate nitrilotriacetate (NTA) or diethylenetriamine pentaacetic acid (DTPA; Hershko et al., 1978; Graham et al., 1979; Singh et al., 1990; Gosriwatana et al., 1999; Kolb et al., 2009). Although several tests have been performed to ascertain that the applied extraction measures do not generally lead to some mobilization of Tf iron, those claims have been challenged leading to the application of more gentle mobilization conditions (Kolb et al., 2009). However, in general, most mobilizing maneuvers applied to plasma/sera that contains any of the chelators used in clinical practice, can lead to mobilization from highly saturated Tf (Kolb et al., 2009). A more serious difficulty pertains to the use of the generic term NTBI to denote plasma iron forms that are unique to IO and presumed to be potentially toxic. Defining "*something that is by what it is not*" (classically known as an *apophasis, from Greek* α - πóϕασις*),* can be confusing and in practice oxymoronic. For example, plasma that does not contain pathological species contains in fact NTBI in the form of ferritin or iron-chelates generated during iron chelation therapy or polymeric iron that is given supplementary via parenteral routes and detected in plasma by NTBI assays. On the other hand LPI is a parameter that denotes, operationally, the level of labile iron species in native

plasma/serum irrespective of treatment or medium composition, an asset that can also become a liability if plasma/serum is not properly stored or some components of serum that affect its redox properties undergo extreme modifications (Breuer et al., 2012).

Attempts were done to identify the membrane transport mechanisms involved in NTBI uptake into cells that cause IO and oxidative damage. However, few studies took into consideration : (a) the actual repertoire of iron permeant species (i.e., the iron complexes that comprise the iron overloaded plasma containing NTBI) and their relative concentrations at different degrees of IO, (b) the composition of the medium milieu (e.g., the presence of high protein that limits substrate availability and redox-active substances that affect substrate composition) and (c) the different transport agencies present in different cell types. Those agencies comprise a variety of transporters or carriers of metals or metalcomplexes, ionic channels or even the means for non-specific adsorptive endocytosis. As most NTBI transport studies have only been done with model simulating substrates in protein-free media and with model cells, their relevance to the etiopathology of tissue iron accumulation in systemic IO remains to be experimentally demonstrated (**Figure 1**). Recent studies indicated dimeric and oligomeric iron-citrate complexes as potential NTBI candidates (Evans et al., 2008), although the chemical speciation is likely to

vary with the level of IO and the natural history of the disease in a given patient (Silva and Hider, 2009). Likewise, NTBI association with plasma proteins is exacerbated in oxidative conditions, such as those that prevail under inflammation or in chronic diabetes (Arezes et al., 2013). A single study with T lymphocytes demonstrated that oligomeric ferric citrate is taken up by T lymphocytes (Arezes et al., 2013) and another study indicated that a substantial fraction of iron that accumulated from exposure of cells to NTBIcontaining human plasma was by adsorptive endocytosis (Sohn et al., 2012).

## **DETERMINATION OF LPI/PLASMA NTBI**

The measurement of a few μM of plasma NTBI in the presence of 70 to >100% saturated TBI (30–50 μM) poses some methodological problems. Two different approaches have tried to overcome them (schematically depicted in **Figure 7**): (1) by extraction of plasma NTBI with iron mobilizing/extraction/chelating agents like NTA or DTPA followed by size filtration (to separate NTBI from TBI) and detection of the iron by iron complex formation and direct measurement of colored complexes or following HPLC analysis (Singh et al., 1990; Gosriwatana et al., 1999; Kolb et al., 2009) and (2) by the detection of LPI in native plasma/serum by: (a) prompting resident LPI (with physiological concentrations of ascorbate) to either catalytically convert the non-fluorescent probe dihydrorhodamine (DHR) into the fluorescent rhodamine (R; Esposito et al., 2003) or co-catalyze with bleomycin chemical changes in chromophoric substrates (Halliwell et al., 1988; Evans and Halliwell, 1994) or (b) binding the labile iron with a

fluorescent-chelator that undergoes a commensurate quenching of fluorescence, as exemplified in **Figure 7** with fluoresceinated DFO (FDFO; Breuer et al., 2001). For either modality (a or b) of method 2, a specific iron chelator (DFO or DFP in excess) is added to a parallel sample in order to assess the specific contribution of labile iron to the observed change in fluorescence elicited in the test sample.

The LPI/NTBI assays have been used extensively to monitor chelation efficacy in transfusional siderotic patients treated with different chelators (Cabantchik et al., 2005; Daar et al., 2009; Zanninelli et al., 2009; Greenberg et al., 2010; Danjou et al., 2014) or hemochromatosis patients following phleboptomy (Le Lan et al., 2005), particularly for establishing: (a) the attainment of sufficiently low LPI (<0.1 μM) at trough levels of chelator as a goal of a particular therapeutic regimen (Cabantchik et al., 2005; Zanninelli et al., 2009) an the appearance of LPI as a criterion for initiating chelation therapy in polytransfused children (Danjou et al., 2014). An attempt has also been done to compare different assays for assessing NTBI (not LPI; Jacobs et al., 2005) but a newer completed one that included also LPI should provide more comprehensive and updated information regarding the usefulness of the LPI/NTBI assays in the clinical setting.

#### **CONCLUSION AND PERSPECTIVES**

With presently available methodologies based on FMS, it has become possible to obtain real time measures of LCI and their compartments in both physiological and pathological conditions (i.e., systemic and regional siderosis) and of LPI/NTBI Cabantchik Labile iron in cells and body fluids

in body fluids as early marker of systemic IO and treatment efficacy. Of particular importance is the ability to assess biological responses to pharmacological measures (e.g., intravenous iron supplementation or iron chelation) and correlate them with clinical outcomes. That area still needs further strengthening, particularly in establishing the thresholds of changes in LPI and LCI that define a given pathological state, either overt or impending.

Unfortunately, hitherto, it has been difficult to define the chemical components of LCI and LPI in biological systems, largely because of their heterogeneous character and variable nature of the media they dwell in. The same pertains to hypothesized iron chaperons that might be operative in mammalian cells and play some physiological role in intracellular iron traffic and metabolism. Although LCI is the parameter closely associated with tissue iron accumulation and ensuing iron-dependent damage, from the clinical perspective, only LPI has thus far proven diagnostically therapeutically relevant for assessing impending/emerging IO and therapeutically, for initiating chelation and monitoring its short term and long term efficacy.

## **ACKNOWLEDGMENTS**

Zvi Ioav Cabantchik was supported by the A&M Della Pergola Chair in Life Sciences, served on the Scientific Advisory Board of Novartis ESB, received in the past research grants from NIH, EEC (5 and 6 framework), Israel Science Foundation, AFIRNE (Paris), Novartis, Shire and Apopharma (also lecture honoraria from the last two) and is presently a consultant for Aferrix Ltd and Hinoman Ltd, Israel.

## **REFERENCES**


in murine erythroleukemia cells. *J. Biol. Chem.* 273, 15382–15386. doi: 10.1074/jbc.273.25.15382


**Conflict of Interest Statement:** The author declares that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

*Received: 12 January 2014; paper pending published: 19 February 2014; accepted: 26 February 2014; published online: 13 March 2014.*

*Citation: Cabantchik ZI (2014) Labile iron in cells and body fluids: physiology, pathology, and pharmacology. Front. Pharmacol. 5:45. doi: 10.3389/fphar.2014.00045 This article was submitted to Drug Metabolism and Transport, a section of the journal Frontiers in Pharmacology.*

*Copyright © 2014 Cabantchik. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*

## Heme in pathophysiology: a matter of scavenging, metabolism and trafficking across cell membranes

## *Deborah Chiabrando‡, Francesca Vinchi † ‡, Veronica Fiorito‡, Sonia Mercurio and Emanuela Tolosano\**

*Department of Molecular Biotechnology and Health Sciences, Molecular Biotechnology Center, University of Turin, Turin, Italy*

#### *Edited by:*

*Raffaella Gozzelino, Instituto Gulbenkian de Ciência, Portugal*

#### *Reviewed by:*

*Dominik J. Schaer, University of Zurich, Switzerland Iqbal Hamza, University of Maryland, USA*

#### *\*Correspondence:*

*Emanuela Tolosano, Department of Molecular Biotechnology and Health Sciences, Molecular Biotechnology Center, University of Turin, Via Nizza 52, 10126 Turin, Italy e-mail: emanuela.tolosano@unito.it*

#### *†Present address:*

*Francesca Vinchi, Molecular Medicine Partnership Unit, Heidelberg University Hospital and EMBL, Heidelberg, Germany*

*‡These authors have contributed equally to this work.*

Heme (iron-protoporphyrin IX) is an essential co-factor involved in multiple biological processes: oxygen transport and storage, electron transfer, drug and steroid metabolism, signal transduction, and micro RNA processing. However, excess free-heme is highly toxic due to its ability to promote oxidative stress and lipid peroxidation, thus leading to membrane injury and, ultimately, apoptosis. Thus, heme metabolism needs to be finely regulated. Intracellular heme amount is controlled at multiple levels: synthesis, utilization by hemoproteins, degradation and both intracellular and intercellular trafficking. This review focuses on recent findings highlighting the importance of controlling intracellular heme levels to counteract heme-induced oxidative stress. The contributions of heme scavenging from the extracellular environment, heme synthesis and incorporation into hemoproteins, heme catabolism and heme transport in maintaining adequate intracellular heme content are discussed. Particular attention is put on the recently described mechanisms of heme trafficking through the plasma membrane mediated by specific heme importers and exporters. Finally, the involvement of genes orchestrating heme metabolism in several pathological conditions is illustrated and new therapeutic approaches aimed at controlling heme metabolism are discussed.

**Keywords: hemopexin, FLVCR1, FLVCR2, ABCG2, HCP1/PCFT, HO-1**

## **THE "GOOD" AND THE "BAD" FACE OF HEME**

#### **HEME AS A PROSTETIC GROUP IN HEMOPROTEINS**

Heme is an iron-containing porphyrin of vital importance since it constitutes the prosthetic moiety of several hemoproteins. It can interact with different inactive apo-heme proteins giving rise to active hemoproteins, whose function is ultimately determined by the properties of the polypeptide bound to it (Dawson, 1988). Thanks to its ability to coordinate an iron atom inside its structure, heme is engaged in controlled oxido-reducing reactions that allow hemoproteins to work properly. For this reason, heme is involved in a multitude of crucial biological functions. In hemoglobin and myoglobin, it is used for oxygen transport and storage, respectively, whereas in cytochromes it is involved in electron transport, energy generation, and chemical transformation. In catalases and peroxidases, heme functions in hydrogen peroxide inactivation or activation, respectively, and in tryptophan pyrrolase, it catalyzes the oxidation of tryptophan (Kumar and Bandyopadhyay, 2005). Furthermore, heme is indispensable for a wide array of other important enzyme systems, such as cyclooxygenase and nitric-oxide synthase (Seed and Willoughby, 1997).

## **HEME AS A MODULATOR OF GENE EXPRESSION AND CELL PROLIFERATION/DIFFERENTIATION**

Other than acting as a prostetic group in hemoproteins, heme itself may influence the expression of many genes. In non erythroid cells, heme regulates its own production by down-regulating heme biosynthesis at the level of the rate-limiting enzyme 5-aminolevulinic acid (ALA) synthase 1 (ALAS1) and by up-regulating its metabolism (Yamamoto et al., 1982). Conversely, in erythroid cells, heme acts as a positive feedback regulator for its synthesis and inhibits its degradation (Sassa, 1976; Rutherford and Harrison, 1979). Heme may control gene expression at almost all levels by regulating transcription, mRNA stability, splicing, protein synthesis, and post-translational modification (Ponka, 1999; Zhu et al., 1999). Genes coding for globins, heme biosynthetic enzymes, heme-oxygenase (HO)-1, ferroportin, cytochromes, myeloperoxidase, and transferrin receptor are all regulated by heme. Most of these genes are regulated via heme response elements (HREs) and the mammalian transcription repressor, Bach1 (Ogawa et al., 2001). Heme responsive elements (HREs) are located in enhancer regions of genes induced by heme itself. The minimal heme response element was identified as an extended activator protein 1 (AP-1, c-Fos/c-Jun family) binding site, TGCTGAGTCAT/C. In addition to the AP-1 heptad (TGAGTCA), this element also contains an interdigitated antioxidant response element (ARE), GCnnnGTCA. This motif resembles the binding site of the transcription factors of the sMaf family (Maf recognition element, MARE; consensus sequence: TGCTGAC(G)TCAGCA), which contains an ARE core for transcription factors of the NF-E2 family (NF-E2, Nrf1, Nrf2) (Inamdar et al., 1996). The transcription factor Bach1 interacts with proteins of the Maf-related family and the resulting heterodimer binds HRE elements, thus repressing gene transcription. Under conditions of intracellular heme accumulation, heme binds to Bach1, thus leading to a conformational change, a decrease in DNA binding activity and finally its removal from HREs (Marro et al., 2010). This allows Maf-Maf, Nrf2-Maf, and other activating heterodimers (Nrf2-AP-1) to occupy the HRE sites in gene promoters and leads to increased transcription of target genes.

In addition, heme regulates differentiation and proliferation of various cell types. It stimulates neuronal differentiation of mouse neuroblastoma cells (Ishii and Maniatis, 1978), erythroid differentiation of erythroleukemia cells (Granick and Sassa, 1978), formation of erythroid colonies in mouse as well as in human bone marrow cultures (Partanen et al., 1988; Abraham et al., 1991), differentiation of 3T3 fibroblasts into adipocytes (Chen and London, 1981), and it stimulates cell growth of cultured fibroblasts (Verger et al., 1983).

All together, these functions can be regarded as the "good face" of heme, without which many critical biological processes would not occur.

## **FREE HEME TOXICITY**

In contrast to the positive functions of heme, free heme excess can cause cell damage and tissue injury since heme catalyzes the formation of reactive oxygen species (ROS), resulting in oxidative stress. Heme that is not bound to proteins is considered the labile heme pool; this portion of heme is derived from newly synthesized heme that has not yet been incorporated into hemoproteins, or heme that has been released from hemoproteins under oxidative conditions. "Free heme" is an abundant source of redox-active iron that can participate in the Fenton reaction to produce toxic free hydroxyl radicals. ROS damage lipid membranes, proteins and nucleic acids, activate cell signaling pathways and oxidant-sensitive, proinflammatory transcription factors, alter protein expression, and perturb membrane channels (Vercellotti et al., 1994; Jeney et al., 2002). Heme toxicity is further exacerbated by its ability to intercalate into lipid membranes. Due to its lipophilic nature, heme may initially lodge within the hydrophobic phospholipid bilayer. Within this highly oxidizable matrix, iron catalyzes the oxidation of cell membrane and promotes the formation of cytotoxic lipid peroxide, which enhances membrane permeability, thus promoting cell lysis and death (Balla et al., 1991; Ryter and Tyrrell, 2000; Kumar and Bandyopadhyay, 2005; Tolosano et al., 2010). Additionally, heme is a potent hemolytic agent. It affects erythrocyte membrane stability as a result of ROS formation and oxidative membrane damage. Finally, heme is strongly pro-inflammatory since it induces the recruitment of leukocytes, platelets and red blood cells to the vascular endothelium, it oxidizes low-density lipoproteins and it consumes nitric oxide, thus impairing vascular function (**Figure 1**).

When intracellular heme accumulation occurs, heme is able to exert its pro-oxidant and cytotoxic action. The cellular free heme pool may increase after extracellular heme overload, increased heme synthesis, accelerated hemoprotein breakdown, impaired incorporation into apo-hemoproteins, or impaired HO activity, resulting in ROS formation, oxidative damage and cell injury. Several pathological conditions are associated with hemolysis or myolysis, and tissues can subsequently be exposed to large amounts of free heme (sickle cell anemia, thalassemia, malaria, paroxysomal nocturnal hemoglobinuria, etc.) (Gozzelino et al., 2010).

In summary, heme is a double-faced molecule: physiological amounts of heme act in gene regulation or as the functional group of hemoproteins, providing essential cellular functions, whereas excessive free heme levels result in oxidative stress and tissue injury. Therefore, the amount of free heme must be tightly controlled to maintain cellular homeostasis and avoid pathological conditions. To this purpose, mammals evolved several defense mechanisms to specifically counteract free heme-mediated oxidative stress and inflammation (Wagener et al., 2003; Gozzelino et al., 2010).

Here the mechanisms involved in the maintenance of heme homeostasis and in the control of heme levels are reviewed: the regulation of both extracellular and intracellular heme content is described, with particular emphasis on the emerging role of heme transporters.

## **REGULATION OF EXTRACELLULAR HEME CONTENT**

Mammals are equipped with various systems able to prevent extracellular heme toxicity. Among them, a key function is covered by the soluble scavengers of free hemoglobin and heme, Haptoglobin (Hp) (Schaer and Buehler, 2013; Schaer et al., 2013a) and Hemopexin (Hx) (Tolosano et al., 2010), respectively.

During several pathological conditions, red blood cells undergo hemolysis and hemoglobin and heme are released into the circulation. Once released, free hemoglobin is captured by its carrier Hp and transported to the macrophages of the reticuloendothelial system, where the complex is bound by the scavenger receptor CD163. When the buffering capacity of plasma Hp is exceeded, hemoglobin is quickly oxidized to methemoglobin, which releases free heme (Ascenzi et al., 2005). Ferriheme is then bound by Hx, by virtue of its high affinity (**Figure 2**). Hx is a 57-kDa acute phase plasma glycoprotein able to bind an equimolar amount of heme and to transport it into the circulation. Hx is expressed mainly in the liver, but also in the brain and retina (Tolosano et al., 1996, 2010). Hx is an acute phase response protein. The acute phase response is a complex systemic early-defence system activated by trauma, infection, stress, neoplasia, and inflammation. Most of these stimuli, in particular hemolytic stress and inflammatory stimuli, induce Hx synthesis (Tolosano and Altruda, 2002). Hx functions as a heme scavenger, maintaining lipophilic heme in a soluble state in aqueous environment and is essential in the re-utilization of heme-bound iron and prevention of heme-induced oxidative damage and cell death (Eskew et al., 1999). Hx acts as an antioxidant thanks to the ability to tightly bind heme, thus effectively reducing heme toxicity by 80–90% (Grinberg et al., 1999). Hx has the specific function to deliver heme to hepatocytes where the heme-Hx complex is internalized by receptor-mediated endocytosis. To date, the only known Hx receptor is the LDL receptor-related protein 1 (LRP 1), a multi-ligand scavenger receptor present on the surface of many cell types. Some studies have suggested that Hx can be recycled as an intact molecule to the extracellular milieu. However, Hvidberg et al. have shown that most Hx is degraded in lysosomes

**FIGURE 1 | Free heme toxicity.** Free heme has potentially toxic properties due to the catalytic active iron atom it coordinates. Here, toxic effects of heme are depicted. Heme causes cellular oxidative damage **(1)** by promoting ROS formation, lipid peroxidation, DNA and protein damage. Additionally, heme is a source of iron. Therefore, heme overload leads to intracellular accumulation of iron, with further ROS generation. Heme is a hemolytic agent **(2)**, since it intercalates red blood cell membrane, thus favoring cell

rupture and further amplifying the hemolytic process. Heme promotes inflammation **(3)**, by stimulating inflammatory cell activation and cytokine production. Finally, heme causes endothelial dysfunction **(4)** by several mechanisms: increasing adhesion molecule expression and endothelial activation, promoting inflammatory cell recruitment and platelet aggregation, causing nitric oxide (NO) oxidative consumption and vasoconstriction, oxidizing low-density lipoprotein (LDL).

(Hvidberg et al., 2005). The binding of Hx to free heme limits the amounts of heme available as a catalyst of radical formation, makes the essential iron unavailable to invasive microorganisms and contributes to the recycling of iron, as heme iron enters the intracellular iron pool.

Once the scavenging capacity of plasma Hx is exhausted, free heme can still be scavenged by plasma albumin (∼66 kDa) (**Figure 2**). Whether heme/albumin complexes are recognized by specific receptors allowing heme degradation via the HO system has not been established. Thus, the protective role of albumin against heme toxicity remains uncertain (Tolosano et al., 2010).

## **REGULATION OF INTRACELLULAR HEME CONTENT**

The control of intracellular heme content occurs at multiple levels. Here we focus on the regulation of heme synthesis, degradation and plasma membrane heme trafficking, which ensure the maintenance of appropriate intracellular heme concentration. Animal models discussed in this section are reported in **Table 1**.

**FIGURE 2 | Control steps in heme metabolism.** The main mechanisms involved in the control of heme levels outside, inside and across the cell are illustrated. **(1)** Heme scavenging: Circulating free heme toxicity is avoided thanks to the action of the scavenging proteins Hx and Albumin. **(2)** Heme Import: Heme might be imported inside the cell via the putative heme importers HCP1/PCFT and FLVCR2. **(3)** Heme Synthesis: in the mitochondrion and cytosol, the heme biosynthetic enzymes, starting from succinyl-CoA and glycine, give rise to heme. After synthesis, heme is exported out of the mitochondrion to the cytosol by the mitochondrial heme exporter FLVCR1b. **(4)** Heme Incorporation in Hemoproteins: once released in the cytosol, heme is inserted in apo-hemoproteins to allow the formation of functional hemoproteins. **(5)** Heme Degradation: in the endoplasmic reticulum, the heme

## **CONTROL OF HEME SYNTHESIS**

The first level of regulation of cellular heme content occurs at the level of heme synthesis control. Heme is synthesized through a series of eight enzymatic reactions (**Figure 2**). Work in the past decade has shown that heme synthesis is an almost ubiquitous process. The biosynthesis of heme starts in the mitochondrial matrix with the condensation of succinyl-CoA and glycine to form ALA, a reaction catalyzed by the enzyme ALAS. There are two isoforms of ALAS: *Alas1* gene is located on chromosome 3 in humans and codes for an ubiquitously expressed protein whereas *Alas2* gene is on the X chromosome and codes for an erythroid-specific protein (Bishop et al., 1990). The two isoforms of ALAS mainly degrading enzyme HO is responsible for heme degradation into iron (Fe), carbon monoxide and biliverdin. **(6)** Heme Export: the heme exporters FLVCR1a and ABCG2 regulate heme export out of the cell across the plasma membrane. ALAS, aminolevulinic acid synthase; SLC25A38, solute carrier family 25 member 38; ABCB10, ATP-binding cassette sub-family B member 10; ALAD, amino levulinic acid dehydratase; HMBS, hydroxymethylbilane synthase; UROS, uroporphyrinogen III synthase; UROD, uroporphyrinogen decarboxylase; ABCB6, ATP-binding cassette sub-family B member 6; CPOX, coproporphyrinogen oxidase; PPOX, protoporphyrinogen oxidase; FECH, ferrochelatase; MFRN, mitoferrin; FLVCR, feline leukemia virus subgroup C receptor; HCP1/PCFT heme carrier protein 1/proton-coupled folate transporter; ABCG2, ATP-binding cassette sub-family G member 2; HO, heme oxygenase; Hx, Hemopexin.

differ for their mode of regulation, as discussed later in this section.

ALA is exported in the cytosol soon after its synthesis. The precise molecular mechanism by which ALA is transported through the two mitochondrial membranes is not completely understood. Two mitochondrial inner membrane proteins, SLC25A38 (solute carrier family 25, member 38) and the ATP-binding cassette transporter ABCB10 have been proposed to play this role.

Yeast lacking YDL119c, the ortholog of SLC25A38, shows a defect in the biosynthesis of ALA (Guernsey et al., 2009). Thus, it has been suggested that SLC25A38 could facilitate the production of ALA by importing glycine into mitochondria or by exchanging glycine for ALA across the mitochondrial inner membrane. **Table 1 | Mouse models deficient for genes involved in the control of heme homeostasis.**


Recently, Bayeva et al. reported that the silencing of ABCB10 causes a decrease of cellular and mitochondrial heme levels. The administration of ALA fully restores heme level in ABCB10 downregulated cells whereas Alas2 overexpression fails to do this. Thus, it has been proposed that ABCB10 could facilitate mitochondrial ALA synthesis or its export from mitochondria (Bayeva et al., 2013).

Both proteins are located on the inner mitochondrial membrane; the ALA transporter on the outer mitochondrial membrane still remains to be identified.

In the cytosol, two molecules of ALA are condensed to form the monopyrrole porphobilinogen, a reaction catalyzed by aminolevulinate dehydratase (ALAD). Then, the enzyme hydroxymethylbilane synthase (HMBS) catalyzes the head-totail synthesis of four porphobilinogen molecules to form the linear tetrapyrrole hydroxymethylbilane which is converted to uroporphyrinogen III by uroporphyrinogen synthase (UROS). The last cytoplasmic step, the synthesis of coproporphyrinogen III (CPgenIII), is catalyzed by uroporphyrinogen decarboxylase (UROD) (Ajioka et al., 2006). All the remaining steps of heme biosynthesis take place inside mitochondria, thus CPgenIII needs to be transported in the mitochondrial intermembrane space. It has been initially proposed that the ATP-binding cassette transporter ABCB6 could play this role (Krishnamurthy et al., 2006). However, data concerning the localization and function of ABCB6 in mitochondria are controversial. ABCB6 was also found to be expressed on the plasma membrane, in the Golgi compartment and in lysosomes. Some works even fail to detect ABCB6 in mitochondria (Paterson et al., 2007; Tsuchida et al., 2008; Kiss et al., 2012). In addition, ABCB6 has been associated to other functions unrelated to porphyrin homeostasis: ABCB6 contributes to anticancer drug resistance (Kelter et al., 2007); it was identified as the genetic basis of the Lan blood group antigen expressed on red blood cells (Helias et al., 2012); defects in *Abcb6* cause an inherited developmental defect of the eye, with no known relationship with porphyrins accumulation (Wang et al., 2012). Recently, it has been reported that Abcb6−*/*<sup>−</sup> mice completely lack mitochondrial ATP-driven import of CPgenIII and shows the up-regulation of compensatory porphyrin and iron pathways. Abcb6−*/*<sup>−</sup> mice are phenotypically normal; increased mortality and reduced heme synthesis were observed following phenylhydrazine administration thus suggesting that Abcb6 is essential during conditions of high porphyrin demand (Ulrich et al., 2012).

In the mitochondrial intermembrane space, CPgenIII is converted to protoporphyrinogen IX by the enzyme coproporphyrinogen III oxidase (CPOX), a homodimer weakly associated with the outside of the inner mitochondrial membrane. The following oxidation of protoporphyrinogen IX to protoporphyrin IX (PPIX) is catalyzed by protoporphyrinogen oxidase (PPOX), located on the outer surface of the inner mitochondrial membrane. Finally, ferrous iron is incorporated into PPIX to form heme in the mitochondrial matrix, a reaction catalyzed by the enzyme ferrochelatase (FECH) (Ajioka et al., 2006). It has been reported that FECH is part of a multi-enzyme complex composed by the mitochondrial iron importer MITOFERRIN1 (MFRN1) and the ATP-binding cassette transporter ABCB10. The association of FECH with MFRN1 allows the coupling of mitochondrial iron import and the integration of iron into PPIX ring while ABCB10 stabilizes MFRN1 expression (Chen et al., 2009, 2010).

Once synthesized, heme must be exported through the two mitochondrial membranes for incorporation into hemoproteins. The only mitochondrial heme exporter identified to date is the mitochondrial isoform of *Flvcr1* (Feline Leukemia Virus subgroup C Receptor 1) gene. Two different isoforms of FLVCR1 exist, FLVCR1a and FLVCR1b, expressed on the plasma membrane and mitochondria respectively. FLVCR1b derives from an alternative transcription start site located in the first intron of the *Flvcr1* gene thus resulting in the production of a shorter protein (Chiabrando et al., 2012). *Flvcr1a* transcript codes for a protein with 12 transmembrane domains (Tailor et al., 1999; Quigley et al., 2000) while *Flvcr1b* mRNA codes for a protein with just six transmembrane domains (Chiabrando et al., 2012). The role of FLVCR1b as a mitochondrial heme exporter is suggested by *in vitro* data indicating that its overexpression promotes heme synthesis whereas its silencing causes detrimental heme accumulation in mitochondria. Moreover, FLVCR1b expression is upregulated following the stimulation of heme synthesis *in vitro* (Chiabrando et al., 2012). According to its role as a mitochondrial heme exporter, FLVCR1b is essential for erythroid differentiation both *in vitro* and *in vivo* (as discussed in section Control of Heme Export). The submitochondrial localization of FLVCR1b is still unknown as well as its ability to interact with other mitochondrial transporters. Further work is needed to definitively understand how heme is transported across the two mitochondrial membranes.

The importance of controlling the rate of heme synthesis is highlighted by the fact that mutations in many genes coding for enzymes involved in this pathway cause specific pathological conditions characterized by the accumulation of toxic heme precursors (as discussed in section Pathological Conditions Associated with Alterations of Heme Synthesis). The regulation of heme synthesis mainly occurs at the level of ALAS, the first and ratelimiting enzyme of the heme biosynthetic pathway. It has been demonstrated that heme plays an essential role in the regulation of its own synthesis by regulating the expression of ALAS.

In non-erythroid cells, heme synthesis is dependent on the activity of ALAS1. It has been reported that ALAS1 is directly regulated by heme levels through several mechanisms: heme negatively controls the transcription (Yamamoto et al., 1982), translation (Sassa and Granick, 1970; Yamamoto et al., 1983) and stability (Hamilton et al., 1991) of *Alas1* mRNA. In addition, it has been shown that ALAS1 contains three heme regulatory motifs (HRMs). Heme binding to two of these HRMs, located in the mitochondrial targeting sequence of ALAS1, inhibits the transport of ALAS1 precursor protein in mitochondria (Yamauchi et al., 1980). The regulation of ALAS1 by heme represents a crucial negative feedback mechanism to maintain appropriate intracellular heme levels in non-erythroid cells, thus avoiding heme-induced oxidative damage.

In erythroid cells, heme synthesis is exclusively dependent on the activity of ALAS2. Contrary to ALAS1, heme does not inhibit the expression of ALAS2, as high amount of heme are required for the differentiation of erythroid progenitors. The expression of ALAS2 is controlled at multiple levels. The transcription of *Alas2* is regulated by erythroid-specific transcription factors, like GATA1 (Surinya et al., 1998; Kaneko et al., 2013). At the posttranscriptional level, the expression of ALAS2 is regulated by iron availability. The 5 untranslated region of *Alas2* mRNA contains an iron responsive element (IRE) which interacts with iron regulatory protein (IRP) 1 and 2. The IRE binding activity of IRPs is regulated by cellular iron. In iron-deficient cells, the binding of IRPs to the 5 IRE inhibits the translation of *Alas2* mRNA. In iron-repleted cells, IRPs are degraded thus stimulating the translation of *Alas2* mRNA. This mechanism ensures the coordination of heme synthesis to the availability of iron thus avoiding the production of potentially toxic heme precursors when iron concentrations are limiting. The regulation of the IRE-binding activity of IRPs by cellular iron occurs through several mechanisms. The IRE binding activity of IRP1 is controlled by iron-sulfur clusters assembly while that of IRP2 by its oxidation, ubiquitination and degradation by the proteasome (Hentze et al., 2010). This latter process is dependent on iron but also on heme availability. It has been reported that IRP2 contains a HRM; heme binding to the HRM mediates the oxidation of IRP2 that triggers its ubiquitination and degradation (Yamanaka et al., 2003; Ishikawa et al., 2005). Thus, during the differentiation of erythroid progenitors, increased cellular iron level stimulates the translation of *Alas2* mRNA by inducing the degradation of IRPs. The following accumulation of heme contributes to the oxidation and degradation of IRP2, further enhancing heme synthesis. This positive feedback mechanism allows a sustained production of heme for hemoglobin synthesis in differentiating erythroid progenitors.

### **CONTROL OF HEME INCORPORATION INTO HEMOPROTEINS**

The rate of *de novo* heme synthesis has to be proportionate to its rate of incorporation into newly synthesized apo-hemoproteins. This is obtained at different levels via the control of heme synthesis as well as apo-hemoproteins synthesis. As reported in section Heme as a Modulator of Gene Expression and Cell Proliferation/Differentiation, heme itself can induce the transcription and the expression of several apo-hemoproteins, such as hemoglobin, myoglobin, and neuroglobin (Bruns and London, 1965; Tahara et al., 1978; Zhu et al., 1999, 2002; Correia et al., 2011), as well as cytochromes and many other heme-containing proteins. This evolutionary conserved strategy prevents intracellular heme accumulation, presumably limiting heme cytotoxicity (**Figure 2**). Additionally, in non-erythroid cells, heme is able to regulate its own synthesis and degradation via inhibition of ALAS1 expression/activity and induction of HO-1 expression/activity, thus properly balancing the amount of synthesized heme with that one incorporated into hemoproteins or catabolized (Zheng et al., 2008; Correia et al., 2011).

#### **CONTROL OF HEME DEGRADATION**

HO is the primary enzyme involved in heme degradation and plays an important role in the protection of cells from hemeinduced oxidative stress. It is a 32 kDa protein, mainly located on membranes of the smooth endoplasmic reticulum, able to break down the pro-oxidant heme into the antioxidant biliverdin, the vasodilator carbon monoxide (CO) and iron (Fe2+) (**Figure 2**) (Gozzelino et al., 2010; Tolosano et al., 2010). Biliverdin is then reduced to bilirubin by the enzyme biliverdin reductase. To date, three isoforms of HO have been identified: HO-1, HO-2, and HO-3.

HO-1 is highly inducible by a variety of stimuli including oxidative stress, heat shock, hypoxia, ischemia-reperfusion, lipopolysaccharide, heavy metals, cytokines and its substrate heme. Heme is the most potent physiologic inducer of HO-1. The enzyme activity was shown to increase in many tissues, including the liver, kidney, adrenals, ovaries, lung, skin, intestine, heart, and peritoneal macrophages. HO-2 is ubiquitously expressed and participates in the normal heme capturing and metabolism. The isoenzymes HO-1 and HO-2 are products of two different genes. The two forms show only 45% aminoacid homology but they share a region with 100% secondary structure homology, that corresponds to the catalytic site (Maines, 1988). HO-3 has poor heme-degrading capacity (Wagener et al., 2003).

Herein, we focus on the inducible isoform HO-1.

HO-1 plays a vital function in heme degradation and protects against heme-mediated oxidative injury. Overexpression of HO-1 is associated to the resolution of inflammation through the generation of beneficial molecules like CO, bilirubin, and ferritin resulting from catabolism of toxic heme (Wagener et al., 2001, 2003; Kapturczak et al., 2004). Bilirubin efficiently scavenges peroxyl radicals, thereby inhibiting lipid peroxidation, attenuating heme-induced oxidative stress, cell activation and death (Dore et al., 1999; Soares et al., 2004; Kawamura et al., 2005). CO controls the activity of several heme proteins and causes vasodilation. It also exerts anti-inflammatory effects by inhibiting the expression of pro-inflammatory cytokines (Ndisang et al., 2003; Beckman et al., 2009). Finally, ferritin, by sequestering toxic free iron, limits microrganism growth and ROS production.

HO-1 confers cytoprotection against different forms of programmed cell death, including apoptosis driven by heme and tumor necrosis factor (Gozzelino and Soares, 2011). This cytoprotective effect is driven by heme degradation *per se* as well as by its end products, CO and biliverdin/bilirubin (Yamashita et al., 2004; Gozzelino et al., 2010). CO can exert cytoprotective effects via the modulation of cellular signal pathways, including the p38 mitogen activating protein kinase (Ndisang et al., 2003). In addition, CO can bind Fe in the heme pockets of hemoproteins, inhibiting heme release and preventing its cytotoxic effects (Seixas et al., 2009; Ferreira et al., 2011). On the other hand, biliverdin has been described to participate in an antioxidant redox cycle in which, once produced by HO, biliverdin is reduced to bilirubin by biliverdin reductase. This is followed by the subsequent oxidation of bilirubin by ROS back to biliverdin, forming a catalytic antioxidant cycle that is driven by NADPH, the reducing cofactor of biliverdin reductase. This cycle has the ability to strongly suppress the oxidizing and toxic potential of hydrogen peroxide and other ROS, thus acting as one of the most powerful anti-oxidant physiological system.

Another detoxifying system is represented by ferritin, an evolutionarily conserved Fe sequestering protein that acts as the major intracellular depot of non-metabolic iron (Balla et al., 1992a,b; Berberat et al., 2003; Cozzi et al., 2004; Pham et al., 2004). Ferritin is a multimeric protein composed of 24 subunits of two types, the heavy chain (H-Ft) and the light chain (L-Ft) and has a very high capacity for storing iron (up to 4500 mol of iron per mol of ferritin). H-Ft manifests ferroxidase activity that catalyses the oxidation of ferrous iron to ferric iron, thus favoring its storage in L-Ft (Hentze et al., 2004) and limiting its participation in the production of free radicals via Fenton reaction (Pham et al., 2004). Together, HO and ferritin allow rapid shifting of iron from heme into ferritin core where it is less available to catalyze deleterious reactions. By increasing the expression of HO-1 and ferritin, cells can survive to lethal heme-induced oxidative stress (Balla et al., 2005; Gozzelino and Soares, 2013).

## **CONTROL OF HEME EXPORT**

Recent evidence demonstrated that also heme export out of the cell significantly contributes to the regulation of intracellular heme levels. Two heme exporters located at the plasma membrane have been identified to date: FLVCR1a and ABCG2 (ATP-Binding Cassette, subfamily G, member 2) (**Figure 2**).

## *The plasma heme exporter FLVCR1a*

FLVCR1a was originally identified and cloned as a cell-surface protein receptor for feline leukemia virus subgroup C, causing pure red blood cell aplasia in cats (Tailor et al., 1999; Quigley et al., 2000). The role of FLVCR1a as a plasma membrane heme exporter is suggested by several *in vitro* data. Indeed, the overexpression of FLVCR1a in NRK or HeLa cells causes a slight but significant decrease of heme content whereas its silencing in FEA or HeLa cells enhances heme level (Quigley et al., 2004; Chiabrando et al., 2012). In addition, the ability of FLVCR1a to export cytoplasmic heme has been demonstrated using zinc mesoporphyrin (ZnMP), a fluorescent heme analog, and 55Fe-heme (Quigley et al., 2004). The heme export activity of FLVCR1a is regulated by the presence of plasma proteins with high affinity for heme, like Hx (Yang et al., 2010).

The analysis of the role of FLVCR1a *in vivo* is limited by the complexity of the available mouse models of *Flvcr1* deficiency (**Table 1**). It was initially reported that FLVCR1a plays an essential role during erythropoiesis, by preventing the toxic accumulation of heme in erythroblasts (Keel et al., 2008). This hypothesis was suggested by the observation that mice lacking *Flvcr1* die *in utero* due to an impairment of erythroid differentiation at the proerythroblast stage. Similarly, post-natal mice lacking *Flvcr1* show a block of erythroid maturation leading to hyperchromic, macrocytic anemia and reticulocytopenia (Keel et al., 2008). Following the identification of FLVCR1b, it was realized that these mouse models were likely lacking both FLVCR1a and FLVCR1b, as they were generated by the deletion of the third exon of the *Flvcr1* gene, which is common to both *Flvcr1* isoforms. Thus, the described phenotype was due to FLVCR1a and/or FLVCR1b deficiency. The specific expression of the two FLVCR1 isoforms in these mouse models still remain to be experimentally verified. Recently, the generation and analysis of *Flvcr1a*−*/*<sup>−</sup> mice suggested that the previously described phenotype was mainly due to the absence of FLVCR1b, since mice lacking FLVCR1a but still expressing FLVCR1b show normal erythroid differentiation and die *in utero* due to severe hemorrhages, edema and skeletal malformations (**Figure 3**). Taken together, these data suggest that FLVCR1b, by exporting heme from mitochondria, is essential for fetal erythroid differentiation (Chiabrando et al., 2012). These data do not exclude a role for FLVCR1a during erythropoiesis; probably the two FLVCR1 isoforms cooperate to determine the appropriate heme level needed for erythroid differentiation (**Figure 4A**).

As FLVCR1a is ubiquitously expressed, it has been hypothesized that its heme export activity could be relevant in different tissues. Post-natal mice lacking *Flvcr1* show iron overload in hepatocytes, duodenal enterocytes and splenic macrophages (Keel et al., 2008). In this mouse model the observed phenotype could not be easily attributed to one of the two isoforms since both FLVCR1a and FLVCR1b were deleted. The specific contribution of FLVCR1a isoform to heme export became evident in conditional mice lacking FLVCR1a expression in hepatocytes. These

**FIGURE 3 | The loss of the heme exporter FLVCR1a in mice causes embryonic lethality, skeletal malformation and extended hemorrhages. (A)** Stereoscopic view of a wild-type (left) and a *Flvcr1a*−*/*<sup>−</sup> (right) embryo at the embryonic stage of E14,5. At this stage, the wild-type embryo shows normal skeletal structure, with fully formed limbs. The *Flvcr1a*−*/*<sup>−</sup> embryo shows extended hemorrhages and edema through the body, in particular in the limbs, back and head. *Flvcr1a*−*/*<sup>−</sup> embryos show skeletal malformations, as suggested by the absence of the lower jaw and properly formed digits. **(B)** An enlarged view of E15,5 wild-type and *Flvcr1a*−*/*<sup>−</sup> anterior limbs (marked with a broken line). In the *Flvcr1a*−*/*<sup>−</sup> embryo, the limbs show severe hemorrhage, leading to impairment in limb and toe formation.

mice accumulate heme and iron in the liver as a consequence of FLVCR1a suppression (Vinchi et al., 2014). Additionally, they show HO and ferritin upregulation, together with alteration of the hepatic oxidative status and induction of the antioxidant genes. These data suggest that heme is normally exported intact from these cells and that in the absence of FLVCR1a, the heme degrading and iron storage systems are upregulated as an attempt to compensate for the lack of heme export (**Figure 4B**).

The analysis of conditional *Flvcr1a*−*/*<sup>−</sup> mice in hepatocytes demonstrated a crucial role for FLVCR1a in the maintenance of hepatic heme homeostasis and hemoprotein function. Interestingly, FLVCR1a expression was found upregulated during cytochrome induction, suggesting that hepatic heme export activity of FLVCR1a was closely associated with heme biosynthesis required to sustain new cytochrome synthesis. Indeed, the lack of FLVCR1a in hepatocytes caused the expansion of the cytosolic heme pool that was responsible for the early inhibition of heme synthesis and increased degradation of heme. As a result, the expression as well as the activity of cytochromes P450 was reduced. These findings indicate that FLVCR1a-mediated heme export is crucial to control intracellular heme levels that in turn regulate heme synthesis, thus determining cytochrome function in the liver (Vinchi et al., 2014).

Additionally, the analysis of *Flvcr1a*−*/*<sup>−</sup> embryos highlighted a previously unrecognized role for FLVCR1a in the maintenance of endothelial integrity. *Flvcr1a*−*/*<sup>−</sup> embryos are characterized by reduced vasculature development and complexity. This is particularly evident in the limbs and tail, where vessels do not form properly and branching is severely compromised (Chiabrando et al., 2012) (**Figure 3**). The molecular mechanism leading to the observed phenotype is still unknown. Interestingly, FLVCR1a is regulated at the transcriptional level by hypoxia, which has a well-established role in angiogenesis and vasculogenesis (Fiorito et al., 2014). We hypothesize a role for FLVCR1a in preventing heme-induced oxidative stress in endothelial cells (**Figure 4C**). FLVCR1a mediated heme export could work in strong association with HO-1 to determine the appropriate amount of heme in endothelial cells. It is well established that HO-1 plays a pivotal role in the regulation of vascular biology (Belcher et al., 2006, 2010; Stocker and Perrella, 2006; Kim et al., 2011). For this reason, it will be interesting to investigate the role of FLVCR1a in hemolytic disorders characterized by enhanced heme-induced oxidative stress in endothelial cells.

The understanding of FLVCR1a subcellular localization, particularly in polarized cell types, will help to gain new insight in FLVCR1a role and function. To date, it has been reported that FLVCR1a localizes on the sinusoidal membrane of hepatocytes, likely exporting heme in the bloodstream (Vinchi et al., 2014). These results have been obtained by studies on the overexpressed protein and are limited by the lack of specific antibodies for this exporter that prevents any analysis of the localization of endogenous FLVCR1a. Future work is required to specifically address FLVCR1a localization in different cell types *in vivo*.

#### *The plasma heme exporter ABCG2*

ABCG2 (ATP-binding cassette, sub-family G, member 2; also known as BRCP, breast cancer resistance protein) is a member

#### **FIGURE 4 | The heme exporter FLVCR1a acts as a new heme detoxifying system. (A)** Erythroid progenitors are able to synthesize and handle high amount of heme, in view of their hemoglobin (Hb)-mediated

oxygen transport activity. FLVCR1b acts as a mitochondrial heme exporter to allow newly formed heme release from the mitochondrion to the cytosol, where it is incorporated into hemoproteins. FLVCR1a has been described as a system involved in the control of heme levels inside erythroid progenitors. By mediating heme export out of these cells, FLVCR1a regulates intracellular heme amount, thus limiting free heme toxicity and oxidative damage. **(B)** Hepatocytes have the highest rate of heme synthesis after the erythroid progenitors. Hepatic heme is mostly used for synthesis of P450 enzymes, which metabolize endogenous compounds and xenobiotics. FLVCR1a mediates heme export out of

of the ABC transporter family that was originally found to confer drug resistance in breast cancer cells (Doyle et al., 1998). The *Abcg2* gene is located on human chromosome 4q22 and it consists of 16 exons and 15 introns (Bailey-Dell et al., 2001). ABCG2 is a "half transporter" as it has only one ABC cassette in a single polypeptide chain of 70 kDa that consists of six transmembrane domains. Only the homodimer is functional (Kage et al., 2002). *Abcg2* expression is regulated by hypoxia through Hypoxia Inducible Factor (HIF) 1 (Krishnamurthy et al., 2004). ABCG2 is localized at the plasma membrane and it is expressed in several tissues including hepatic canalicular membranes, renal proximal tubules, intestinal epithelium and placenta (Doyle and Ross, 2003). In all these tissues, ABCG2 detoxifies drugs, toxins and metabolites (Bailey-Dell et al., 2001). Moreover, ABCG2 is expressed in a sub-population of hematopoietic stem cells. It prevents cytotoxicity of chemotherapics and confers resistance to hypoxic conditions (Krishnamurthy et al., 2004).

ABCG2 has a wide substrate specificity (Keppler and Konig, 2000). Recently, genetic studies in human disease established that ABCG2 functions as a urate transporter that promotes urate excretion in the kidney as discussed in section Hyperuricaemia and Gout (Qiu et al., 2006). The role of ABCG2 as a heme transporter was serendipitously discovered when *Abcg2*−*/*<sup>−</sup> mice fed a modified diet developed skin photosensitivity (Jonker et al., 2002). This was caused by the accumulation of pheophorbide, a degradation product of chlorophyll present in the diet, structurally similar to PPIX. *Abcg2*−*/*<sup>−</sup> mice were found to hepatocytes, thus maintaining hepatic heme homeostasis and controlling cell oxidative status. FLVCR1a export function allows the maintenance of a proper cytosolic heme pool that matches cell need for new hemoprotein generation (e.g., cytochrome P450). Block of heme export causes heme pool expansion leading to the inhibition of heme synthesis and the reduction of cytochrome activity. **(C)** A similar role for FLVCR1a was proposed to occur in endothelial cells. *Flvcr1a*−*/*<sup>−</sup> embryos show reduced vascular arborization, potentially due to altered endothelial integrity, suggesting that the lack of FLVCR1a leads to intracellular heme overload and oxidative stress. Endothelial cells are highly sensitive to heme overload and, in this context, FLVCR1a function could be of crucial importance to export heme excess, thus maintaining heme homeostasis and controlling heme-induced oxidative stress.

accumulate PPIX and other porphyrin-like compounds in erythrocytes and other cells implying that ABCG2 has a role in cellular efflux of these compounds (Jonker et al., 2002). In addition, it has been demonstrated that ABCG2 is precipitated by hemin-agarose and it exports the heme analog ZnMP in K562 cells transfected with the human *Abcg2* cDNA (Krishnamurthy et al., 2004; Desuzinges-Mandon et al., 2010). Nevertheless, the physiologic porphyrin substrate of ABCG2 seems to be PPIX as erythroid progenitors of *Abcg2*−*/*<sup>−</sup> mice accumulate this compound and erythroid cells overexpressing ABCG2 have reduced levels of PPIX (Krishnamurthy and Schuetz, 2005; Zhou et al., 2005). Interestingly, erythroid progenitors of *Abcg2*−*/*<sup>−</sup> mice are more sensitive to hypoxic conditions. Since *Abcg2* expression is induced by HIF1, ABCG2 is thought to confer a strong survival advantage to stem cells under hypoxic stress by reducing intracellular porphyrin content (Krishnamurthy et al., 2004).

Despite of porphyrin accumulation in erythroid progenitors, *Abcg2*−*/*<sup>−</sup> mice do not show porphyria or an overt anemic phenotype. This might indicate that pophyrin export does not significantly contribute to intracelluar porphyrin homeostasis and/or that compensatory mechanisms are activated when *Abcg2* is lacking. Alternatively, this could also indicate that ABCG2 is not a physiologic porphyrin exporter. Nevertheless, ABCG2 expression is induced in HeLa cells after the stimulation of heme synthesis (Chiabrando et al., 2012), supporting the conclusion that, in some way, it is involved in this process.

Based on its localization at the apical membrane of duodenal enterocytes, it has been hypothesized that ABCG2 could transport excess heme or porphyrin from the enterocyte to the lumen. Similarly, it could be involved in porphyrin/heme detoxification through the biliary system and/or the kidney (Latunde-Dada et al., 2006; Krishnamurthy et al., 2007). Consistently, ABCG2 was found to cooperate with HO-1 to protect kidney cells against heme-induced cytotoxicity (Wagener et al., 2013). Finally, *Abcg2* deficiency was found to increase oxidative stress and to exacerbate cognitive/memory deficit in a mouse model of Alzheimer's disease (Zeng et al., 2012).

Thus, ABCG2 is a plasma membrane transporter for a wide variety of substrates including urate porphyrin/heme, chemiotherapeutics, antibiotics, xenobiotics, and food metabolites. If, other than in urate transport and in drug metabolism, ABCG2 plays a role in heme export under physiologic or pathologic conditions remains to be elucidated.

## **CONTROL OF HEME IMPORT**

A further level of control of intracellular heme content is represented by the modulation of heme import inside the cells (**Figure 2**). To date the only known proteins with a wellestablished function as heme importers are HRGs (Rajagopal et al., 2008; White et al., 2013), reviewed by Hamza et al. in the present Research Topic. Two other putative heme importers have been described, the Feline leukemia virus subgroup C receptor 2 (FLVCR2) and the Heme carrier protein 1/Proton-coupled folate transporter (HCP1/PCFT).

## *The heme importer FLVCR2*

FLVCR2 is the second member of the SLC49 family of heme transporters, a family also including the founding member FLVCR1 (SLC49A1), and other two proteins, MFSD7 (SLC49A3) and DIRC2 (SLC49A4) (Khan and Quigley, 2013). The SLC49 family belongs to the Major Falicitator Superfamily (MFS) of secondary active permeases that transports small solutes across membranes in response to chemico-osmotic gradients, contributing to the maintenance of normal cell homeostasis (Pao et al., 1998).

The gene encoding FLVCR2 contains 10 exons, is located on chromosome 12D2 in mouse and 14q24 in human (Quigley et al., 2000) and is highly homologous to the *Flvcr1* gene. FLVCR2 protein shares about 60% amino acid sequence identity with FLVCR1. A variant transcript exists, encoding a shorter hypothetical protein that differs in the N-terminal and whose significance is unknown (Brown et al., 2006).

*Flvcr2* mRNA is found ubiquitously, with the highest transcript levels observed in human brain, placenta, lung, liver, kidney and hematopoietic tissues (Duffy et al., 2010). In mouse, its expression has been reported in brain, spinal cord (Lein et al., 2007) and in the columnar cells overlying the fetal blood vessels in the placental yolk sac at day E20 (Brasier et al., 2004).

The FLVCR2 protein is composed by 12 predicted transmembrane domains and six presumptive extracellular loops, it is not N-linked glycosylated and shows a molecular mass of about 55 kDa, while its truncated variant weights about 40 kDa (Brown et al., 2006). Unfortunately, to date a FLVCR2-specific antibody is not available, so protein expression in the different tissues and cell compartments has not been determined. Nevertheless, the FLVCR2 functions, described below, strongly indicate its localization on the cell plasma membrane.

Brasier et al. (2004) proposed a putative role for FLVCR2 as a calcium transporter, based on its expression in murine and human tissues characterized by rapid calcium exchange. Although this function has not been formally excluded by subsequent studies, alternative roles for FLVCR2 have emerged. FLVCR2 has been identified as the receptor for the FY981 variant of FeLV-C retrovirus, a variant that can also use FLVCR1 and the thiamine transporter 1 to infect cells (Shalev et al., 2009). Unlike FLVCR1, FLVCR2 was unable to export heme (Quigley et al., 2004). Conversely, FLVCR2 was postulated to be a heme importer since (i) the overexpressed protein coprecipitated with hemin-agarose, (ii) human cells overexpressing FLVCR2 showed an enhanced uptake of ZnMP, and (iii) *Xenopus* oocytes expressing human FLVCR2 showed increased uptake of [55Fe]hemin (Duffy et al., 2010). Nevertheless, it has to be noted that the binding of FLVCR2 to hemin-agarose was not very efficiently competed by free hemin and the transport assays showed only a two-fold increase of ZnMP or 55Fe-heme uptake when FLVCR2 was overexpressed. Furthermore, studies in yeast failed to demonstrate a heme transport function for FLVCR2 (Yuan et al., 2012). Finally, genetic studies in humans associated *Flvcr2* mutation to the Fowler syndrome, a disorder with no obvious link to heme metabolism (see section Fowler Syndrome). Thus, the assumption that FLVCR2 is a heme importer is not definitive and further studies are needed to fully address the substrate specificity of this transporter.

## *The heme importer HCP1/PCFT*

HCP1/PCFT belongs to the SLC46 family of the MFS. The gene encoding HCP1/PCFT spans five exons and four introns and is located on chromosome 11B5 in mouse and 17q11 in human. The HCP1/PCFT protein is highly conserved among different species, sharing about 90% similarity between human and mouse or rat (Qiu et al., 2006). Furthermore, it shows a significant homology to bacterial metal tetracycline transporters, likely due to similarities in the structures of the substrates specific for HCP1/PCFT and this kind of transporters (Shayeghi et al., 2005).

*Hcp1/Pcft* transcript is abundant in the mouse duodenum after weaning and in the adult rat liver (Shayeghi et al., 2005), but a weak expression can also be detected in the mouse kidney (Shayeghi et al., 2005) and in many other mouse tissues (Salojin et al., 2011). No expression was observed in mouse placenta and ileum (Shayeghi et al., 2005) and in mouse adipose tissue or total bone tissue including bone marrow (Salojin et al., 2011). In human tissues, two different *Hcp1/Pcft* transcripts were observed, the shorter one being dominant. Both transcripts were detected in kidney, liver, placenta, duodenum and spleen, and to a lesser extent in jejunum, ileum, cecum, colon, rectum, and testis. Very low *Hcp1/Pcft* mRNA levels were observed in human brain, lung, stomach, heart and muscle (Qiu et al., 2006). Furthermore, the *Hcp1/Pcft* transcript was found in human macrophages (Schaer et al., 2008), in retina and in retinal pigment epithelium (Sharma et al., 2007). Finally, *Hcp1/Pcft* mRNA was also observed in several cell lines, like hepatoma, neuroblastoma and macrophage cell line (Shayeghi et al., 2005), in primary rat astrocytes (Dang et al., 2010) and in the Caco2 human adenocarcinoma cell line (Qiu et al., 2006).

Duodenal *Hcp1/Pcft* mRNA levels are not influenced by systemic iron amount or by increased ineffective erythropoiesis and IREs were not observed in the 5 and 3 untranslated regions of *Hcp1/Pcft* transcript (Shayeghi et al., 2005). Conversely, *Hcp1/Pcft* mRNA levels strongly increase upon hypoxia, despite the absence of hypoxia responsive elements in the promoter sequence of the *Hcp1/Pcft* gene. Moreover, the anti-inflammatory agents glucocorticoids increase *Hcp1/Pcft* mRNA in macrophages (Schaer et al., 2008), while the lipopolysaccharide and Toll-like receptors agonists lead to suppression of *Hcp1/Pcft* transcript in these cells.

The HCP1/PCFT protein contains 459 amino acids, shows a molecular mass of about 50 kDa, is composed of 9 (Shayeghi et al., 2005) or 12 (Sharma et al., 2007) predicted transmembrane domains and is localized both in the plasma membrane (particularly the apical membrane of polarized cells like enterocytes) and in subcellular vesicles (Shayeghi et al., 2005; Yanatori et al., 2010). In non-polarized cells, it could also localize in the lysosome (Yanatori et al., 2010), and in human macrophages it has been detected in the endosome (Schaer et al., 2008). The localization of HCP1/PCFT is highly influenced by changes in cellular iron stores, accumulating at the brush border membrane in iron-deficiency, and in the cytoplasm in iron-loaded conditions (Shayeghi et al., 2005).

HCP1/PCFT was initially identified as a heme importer (Shayeghi et al., 2005). Nevertheless, Qiu et al. (2006) reported that HCP1/PCFT is a folate transporter (Qiu et al., 2006). Considering that the K*<sup>m</sup>* at pH 6.5 for 5-methyltetrahydrofolate was measured at 0.8μM whereas the K*<sup>m</sup>* at neutral pH for hemin was 125◦μM, it appears that HCP1/PCFT is a poor heme transporter. This conclusion is strengthened by the finding that the gene coding for HCP1/PCFT was found mutated in patients with hereditary folate malabsorption (Qiu et al., 2006) (see section Hereditary Folate Malabsorption Syndrome).

Here, we report some other considerations on the putative role of HCP1/PCFT as a heme importer that could be relevant only in some cell types or in particular physiologic or pathologic situations. The expression of HCP1/PCFT in the duodenum and liver indicate that this protein might be important both for the import of dietary heme from the gut lumen to the enterocytes and for the uptake of plasma heme by non-intestinal tissues (Shayeghi et al., 2005). Moreover, it has been proposed that it could be involved in the export of hemoglobin-derived heme from the endosome to the cytoplasm of macrophages (Schaer et al., 2008). In addition to free heme import, both heme-arginine and heme-bovine serum albumin are able to donate heme for uptake via HCP1/PCFT. Nevertheless, in primary rat astrocyte cultures the greatest amount of heme accumulation was measured in the absence of Hx/albumin-containing fetal calf serum (Dang et al., 2010). The transport of heme by HCP1/PCFT is a saturable and energy/temperature-dependent process. HCP1/PCFT is able to import other substrates containing a porphyrin ring, as ZnMP and protoporphyrin (Shayeghi et al., 2005; Qiu et al., 2006).

The energy source for HCP1/PCFT-mediated heme transport is unknown, as no adenosine triphosphate binding motifs were observed in its sequence. Nevertheless, the prevalent hypothesis is that, as for folates uptake, it could utilize the co-transport of protons along a favorable concentration gradient to drive the transport of heme (Shayeghi et al., 2005).

## **PATHOLOGICAL CONDITIONS ASSOCIATED WITH ALTERATIONS OF HEME METABOLISM PATHOLOGICAL CONDITIONS ASSOCIATED WITH ALTERATIONS OF**

**EXTRACELLULAR HEME SCAVENGING** Many pathological conditions are associated with hemolysis and extracellular heme release. Among them, sickle cell anemia, β-thalassemia, malaria, paroxysomal nocturnal hemoglobinuria are the most paradigmatic (Gozzelino et al., 2010). In all these conditions, the pool of circulating Hx is diminished and, consequently, the plasma heme scavenging capacity is strongly reduced.

In the last decade, the use of a knockout mouse model for Hx was valuable to elucidate its function as well as its potential use as a therapeutic molecule in the prevention of heme adverse effects (**Table 1**). Additionally, other functions of Hx not clearly related to its role as an heme scavenger, have also been described thanks to the use of knock-out mice (Fagoonee et al., 2008; Spiller et al., 2011; Rolla et al., 2013).

Works on Hx−*/*<sup>−</sup> mice have demonstrated that these animals have alterations in heme/iron homeostasis in the duodenum (Fiorito et al., 2013) and brain (Morello et al., 2009, 2011), are highly sensitive to acute hemolysis (Vinchi et al., 2008) and show a defective recovery after intravascular hemolysis, suffering from severe renal damage associated with iron loading and lipid peroxidation (Tolosano et al., 1999, 2002). After acute heme overload, Hx−*/*<sup>−</sup> mice develop severe hepatic red blood cell congestion associated to cell iron redistribution, lipid peroxidation and inflammation. This is likely due to the increased endothelial activation and vascular permeability occurring in Hx−*/*<sup>−</sup> mice compared to wild-type controls (Vinchi et al., 2008). Endothelial activation is a proinflammatory and procoagulant state of the endothelial cells lining the lumen of blood vessels. This state is mainly characterized by an increased expression of adhesion molecules on endothelial cell surface, which promotes the adherence of leukocytes as well as platelets and red blood cells, thus favoring inflammation, clot and eventually thrombus formation. According to its function as a free heme scavenger, Hx is expected to counteract heme toxicity on the vascular endothelium and increasing experimental evidence is supporting this concept.

Similarly, sickle cell disease and β-thalassemia mice show heme-driven endothelial activation, vaso-occlusion and cardiovascular dysfunction that could be efficiently recovered through the administration of an Hx-based therapy (Vinchi et al., 2013). This recent observation further strengthens the concept that heme triggers vascular inflammation and damage, and emphasizes the importance of Hx in counteracting heme-driven cardiovascular dysfunction associated with hemolytic conditions. This could have relevance, in the future, for therapeutic interventions against cardiovascular and endothelial dysfunctions in hemolytic patients (Vinchi and Tolosano, 2013).

Malaria is another well-known hemolytic condition, associated with the accumulation of high concentrations of free heme in plasma. The appearance of hemoglobin and heme in plasma has been linked to the development of cerebral malaria, which remains the most severe and difficult to treat complication of the infection (Ferreira et al., 2008; Seixas et al., 2009). The potential protective effect of both Hp and Hx in this pathology still needs to be elucidated.

Recently, hemolysis has been observed to occur after red blood cell transfusion, one of the most common therapeutic interventions in medicine. Over the last two decades, however, transfusion practices have been restricted, limiting unnecessary transfusions (Barr and Bailie, 2011). The adverse effects of transfusions seem to be mainly related to the storage period between blood donation and transfusion (Wang et al., 2012). Transfusions of old blood in animal models result in both intravascular and extravascular hemolysis and cause hypertension, acute renal failure, hemoglobinuria, and vascular injury. Conversely, old blood transfusion together with the hemoglobin scavenger Hp attenuated most of the transfusion-related adverse effects (Baek et al., 2012; Schaer et al., 2013b). Whether a similar protection, alone or in association with Hp, could be afforded by Hx still needs to be explored.

Not long ago, severe sepsis was found accompanied by hemolysis and Hx exhaustion in mice (Larsen et al., 2012). Larsen and coworkers showed that the administration of exogenous Hx was protective against organ injury and prevents the lethal outcome of severe sepsis in mice (Larsen et al., 2010). The protective effect in this model is related to the ability of Hx to counteract heme proinflammatory effects upon pathogen infection.

Furthermore, a neuroprotective effect for Hx was also described. Hx was found expressed by cortical neurons and present in mouse cerebellum, cortex, hippocampus, and striatum. Upon experimental ischemia, neurologic deficits as well as infarct volumes in the brain were increased in Hx deficient mice, indicating that Hx regulates extracellular free heme levels and the heme-Hx complexes protect primary neurons against the heme-induced toxicity (Li et al., 2009).

The rationale for the use of Hx as a therapeutic is based on the idea that it acts by scavenging circulating free heme, the ultimate mediator of hemoglobin toxicity. In particular, the crucial role of Hx in the protection from heme toxicity has relevance to pathological conditions associated with hemolysis that are often characterized by partial/total exhaustion of hemoglobin/heme scavengers. Therefore, replenishing the circulating stores of heme scavengers, thereby compensating for the loss of heme scavenging capacity in plasma, may be used as a therapeutic approach to target circulating free heme and prevent its deleterious effects (Schaer and Buehler, 2013; Schaer et al., 2013b; Vinchi and Tolosano, 2013). The recent demonstration of the effectiveness of an Hx therapy in mouse models of sickle cell anemia, β-thalassemia and sepsis opens new perspectives concerning the use of this molecule as a new therapeutic drug in hemolytic diseases.

To date, we are still far from the use of heme scavengers as therapeutics and no clinical trials are ongoing. Anyway, the possible use of these molecules for therapeutic purposes has elicited the interest of the research community in the field as well as of several pharmaceutical companies. Pre-clinical studies are ongoing with the aim of translating the protective effects of heme scavengers into clinical practice in the near future.

## **PATHOLOGICAL CONDITIONS ASSOCIATED WITH ALTERATIONS OF HEME SYNTHESIS**

## *Porphyrias*

Eight distinct types of inherited porphyrias have been described, each resulting from a partial deficiency of a specific enzyme of the heme biosynthetic pathway (**Figure 2** and **Table 2**). The porphyrias are characterized by an impairment of heme synthesis, leading to the accumulation of specific intermediates of the heme biosynthetic pathway in various tissues. Depending on the primary site of overproduction and accumulation of heme precursors, the porphyrias have been traditionally classified as "hepatic" or "erythropoietic." However, many types of porphyrias have overlapping features. The accumulation of porphyrins in different tissues leads to hepatic and hematopoietic alterations, neurological and/or cutaneous symptoms.

Three different types of "erythropoietic protoporphyria" have been described. The initial accumulation of the porphyrin precursors occurs primarily in bone marrow erythroid cells. Mutations in the *FECH* gene cause erythropoietic protoporphyria (EPP; OMIM: #177000) (Gouya et al., 2002); the partial deficiency of FECH leads to the accumulation of free-PPIX in bone marrow, erythrocytes, plasma and finally in the liver (Murphy, 2003; Lecha et al., 2009). A clinically indistinguishable form of erythropoietic protoporphyria, named X-linked erythropoietic protoporphyria (XLEPP; OMIM: #300752), is due to gain of function mutations in the *ALAS2* gene (Whatley et al., 2008). The overexpression of ALAS2 causes an increased PPIX production in spite of normal FECH activity; as iron became limiting, FECH uses its alternative metal substrate leading to the accumulation of Znprotoporphyrin in erythrocytes. The major phenotype of EPP and XLEPP is porphyrin-induced photosensitivity. Hypochromic, microcytic anemia is also common and the disorders may evolve to severe hepatobiliary disease and hepatic failure (Anstey and Hift, 2007; Whatley et al., 2008; Lecha et al., 2009). Mutations in the *UROS* gene cause the third form of erythropoietic protoporphyria, called congenital erythropoietic protoporphyria (CEP; OMIM: #263700); the partial loss of UROS activity leads to the incomplete metabolism of hydroxymethylbilane and the accumulation of non-physiologic porphyrin isomers in the bone marrow, erythrocytes, urine and other organs. CEP is mainly characterized by cutaneous photosensitivity and hemolytic anemia (Sassa and Kappas, 2000; Murphy, 2003; Bishop et al., 2006).

The "hepatic porphyrias" are characterized by the overproduction and initial accumulation of porphyrin precursors in the liver. Five different types of hepatic protoporphyria have been described. Mutations in the *ALAD*, *HMBS*, *CPOX*, *PPOX* genes cause different forms of acute hepatic porphyrias: Aladdeficiency porphyria (ADP; OMIM: #612740), acute intermittent porphyria (AIP; OMIM: #176000), hereditary coproporphyria (HCP; OMIM: #121300) and porphyria variegate (VP; OMIM: #176200), respectively. These disorders are characterized by the accumulation of ALA, porphobilinogen and/or porphyrins primarily in the liver. The major manifestations of these disorders are the life-threatening acute neurologic attacks and abdominal pain (Balwani and Desnick, 2012). Mutations in the *Urod* gene cause a cutaneous form of hepatic porphyria, called porphyria cutaneous tarda (PCT; OMIM: #176100); impaired *UROD* activity leads to **Table 2 | Disorders associated to mutations in genes involved in heme metabolism.**


the accumulation of uroporphyrin and other highly carboxylated porphyrins in the skin, liver and erythrocytes. PCT is characterized by blistering skin lesions that appear most commonly on the backs of the hands (increased cutaneous photosensitivity) and liver disease whereas neurologic features are usually absent (Frank and Poblete-Gutierrez, 2010; Balwani and Desnick, 2012).

The generation of animal models of porphyrias allowed a better understanding of the pathophysiological mechanisms involved in the distinct types of porphyria as well as the development of novel therapeutic strategies (Richard et al., 2008).

## *Sideroblastic anemias*

Mutations in the *ALAS2* gene are responsible for the most common form of inherited sideroblastic anemia, named X-linked sideroblastic anemia (XLSA; OMIM: #300751) (**Table 2**). Sideroblastic anemias are genetically and clinically heterogeneous disorders characterized by the pathological accumulation of iron in the mitochondria of erythroid precursors. XLSA is due to loss of function mutations in the *ALAS2* gene (Bergmann et al., 2010; Ducamp et al., 2011) or mutations in an enhancer of *ALAS2* gene that cause the disruption of a GATA binding site important for its transcriptional regulation (Kaneko et al., 2013). The reduced activity of ALAS2 causes an impairment of heme biosynthesis; as PPIX production is decreased, excess iron accumulates in perinuclear mitochondria of erythroblasts creating a ring-like appearance and thus the characteristic "ringed sideroblasts." XLSA patients are characterized by hypochromic, microcytic anemia of variable severity.

The role of ALAS2 in XLSA has been confirmed in several animal models of the disease. *Sauternes* (*sau*) is a zebrafish mutant characterized by a delay in erythroid differentiation, abnormal globin gene expression and heme deficiency. Using positional cloning strategies, it has been reported that the *sau* gene encodes for ALAS2, thus confirming that loss of ALAS2 leads to a microcytic, hypochromic anemia similar to XLSA (Brownlie et al., 1998). The absence of *Alas2* in mice causes embryonic lethality due to a severe block of erythroid differentiation. In contrast to human patients, ring sideroblasts are not present and iron deposition occurs in the cytoplasm instead of mitochondria (Nakajima et al., 1999; Harigae et al., 2003).

Interestingly, other forms of inherited sideroblastic anemias are due to mutations in genes indirectly involved in the heme biosynthetic pathway. Mutations in the gene coding for *SLC25A38*, the putative mitochondrial exporter of ALA, have been identified in patients with an autosomal recessive form of sideroblastic anemia (Guernsey et al., 2009). In other patients, mutations in the ATP-binding cassette transporter *ABCB7* gene or in *GLUTAREDOXIN5* (*GLRX5*) gene have been identified (Allikmets et al., 1999; Camaschella et al., 2007). Both ABCB7 and GLRX5 are involved in the assembly of iron-sulfur clusters (Lill and Muhlenhoff, 2006). Iron-sulfur clusters deficiency causes the activation of IRP1, mitochondria iron accumulation and cytosolic iron depletion that in turn activates IRP2. The activation of IRPs determines the translational repression of ALAS2 resulting in sideroblastic anemia (Allikmets et al., 1999; Wingert et al., 2005; Camaschella et al., 2007; Sheftel et al., 2009; Ye et al., 2010). Thus, a primary defect in iron-sulfur clusters biogenesis secondarily affects heme synthesis in erythroblasts, resulting in mitochondrial iron loading and the same pathophysiology of ALAS2 deficiency.

## **PATHOLOGICAL CONDITIONS ASSOCIATED WITH ALTERATIONS OF HEME DEGRADATION**

#### *Disorders associated with mutations in HO-1 gene*

Mice lacking functional HO-1 were generated by Poss and Tonegawa (1997) (**Table 1**). These mice were characterized by serum iron deficiency and pathologic tissue iron-loading, indicating that HO-1 is crucial for the expulsion of iron from tissue stores and for its reutilization. *HO-1*−*/*<sup>−</sup> mice were shown to accumulate, with age, hepatic and renal iron that contributed to oxidative damage, tissue injury and chronic inflammation (Yachie et al., 1999; Koizumi, 2007). These data demonstrated that, although HO-1 is a stress-induced protein, it is important under basal conditions to protect liver and kidney from oxidative damage and that it is an essential regulator of iron metabolism and homeostasis. Additionally, these mice suffered from delayed growth and progressive chronic inflammatory diseases as suggested by an enlarged spleen and lymph nodes, hepatic inflammatory cell infiltrates, vasculitis, and glomerulonephritis. Furthermore, they were found to be extremely sensitive to oxidative injury and prone to hepatic necrosis and death upon lipopolysaccharide administration.

Recent studies on *HO-1*−*/*<sup>−</sup> mice revealed that HO-1 deficiency causes the depletion of resident splenic and liver macrophages, due to their inability to catabolize hemoglobinderived heme during erythrophagocytosis (Kovtunovych et al., 2010). In the spleen, initial splenic enlargement was observed to progress to red pulp fibrosis, atrophy, and functional hyposplenism in older mice. Finally, the failure of tissue macrophages to remove senescent red blood cells led to intravascular hemolysis, circulating hemoglobin release, and iron redistribution to hepatocytes and kidney proximal tubules. Indeed, the lack of HO-1 strongly impairs macrophage function, thus causing iron redistribution and severe oxidative tissue injury.

HO-1 deficiency in humans was described for the first time in 1999 (OMIM: #614034) (Yachie et al., 1999; Kawashima et al., 2002) (**Table 2**). The sequence analysis of the *HO-1* gene revealed the complete loss of exon 2 on the maternal allele and a 2-nucleotide deletion in exon 3 on the paternal allele. The disease was reported in a 6-year-old boy, who suffered from severe growth retardation, asplenia, marked hepatomegaly, renal injury, tissue iron deposition and paradoxically elevated Hp levels. Moreover, he showed increased red blood cells fragility, chronic hemolysis, anemia, leukocytosis, thrombocytosis, disseminated intravascular coagulation, hyperlipidemia and mesangio-proliferative glomerular changes, likely resulting from endothelial injury and reticulo-endothelial dysfunction.

The serum level of Hp is usually reduced in hemolytic states; in this patient suffering from hemolytic anemia, however, the Hp level was rather increased. In the HO-1 deficient case, Hp production rather than its consumption could be increased due to a dominant effect of inflammation. In addition, reticuloendothelial dysfunction could delay the clearance of the Hp-hemoglobin complex. Contrarily to the *HO-1*−*/*<sup>−</sup> mouse, in human HO-1 deficiency fatty streaks and fibrous plaques were observed. Histologically, the fibrous plaques were characterized by the proliferation of smooth muscle cells and few foam macrophages. The enhanced proliferation of smooth muscle cells could be a consequence of HO-1 loss.

Recently, another case of HO-1 deficiency was described in a 15-year-old girl who presented massive hemolysis, inflammation, nephritis and congenital asplenia (Radhakrishnan et al., 2011). The key features that suggested HO-1 deficiency were marked hemolysis, generalized inflammation with evidence of endothelial injury and nephropathy with underlying asplenia. Mutation analysis showed the presence of homozygous missense mutations in exon 2 (R44X) on chromosome 22q12, which resulted in the absence of the functional HO-1.

HO-1 deficiency in humans is characterized by total asplenia and this is fully recapitulated by the mouse model of HO-1 deficiency suggesting that this enzyme has a key role in macrophage heme metabolism and heme-iron reutilization. On the other hand, compared with the knockout mouse model, the human cases of HO-1 deficiency were observed to involve more severely the endothelial cells. Both the reported cases of HO-1 deficiency presented endothelial dysfunction, systemic inflammation and hemolysis. In particular, in the last reported case of HO-1 deficiency, the initial trigger of cold antibody-mediated hemolytic anemia, with intravascular hemolysis, would have resulted in heme-induced endothelial damage. In both cases, signs of severe endothelial damage in the form of raised inflammatory markers, von Willebrand factor, and coagulopathy were found and these children died of intracranial hemorrhage. It was proposed that HO-1 deficiency results in a novel form of vasculitis or endothelial injury syndrome. How asplenia modulated the disease is unexplained. In view of the absence of spleen in both cases, it is possible that HO-1 gene expression may have a role in splenic arteriogenesis, other than in angiogenesis, as already proposed.

This very rare condition provided important insight into the functional role and importance of this enzyme. In both mice and human cases, a severe proinflammatory and pro-oxidant phenotype was noted, further highlighting the anti-oxidant, anti-inflammatory and cytoprotective function of this enzyme. Although it appears evident that HO-1 is essential to keep intracellular heme levels within a physiologic non-toxic range, it is not clear whether the protection against heme toxicity by this enzyme mainly relies on its primary heme degrading activity, on the biologic activities of its metabolic end products, namely CO and bilirubin, or both.

### **DISORDERS INFLUENCED BY HO-1 POLYMORPHISMS**

Humans differ quantitatively in their ability to build up an HO-1 response and variations in HO-1 expression dictate the pathologic outcome of a broad range of diseases. This differential response seems to be modulated by two polymorphisms in the HO-1 gene promoter region. Several studies demonstrated that the ability of a patient to respond strongly in terms of upregulating HO-1 may be an important endogenous protective factor.

A well described (GT)n microsatellite polymorphism in the HO-1 promoter is thought to regulate the extension of HO-1 induction as well as its response to many stimuli. Individuals with a lower number of (GT)n repeats retain the ability to more efficiently induce HO-1 than individuals with a higher number of repeats. Multiple studies demonstrate that individuals with fewer repeats are less prone to certain pathologies. It is interesting to observe that most of the pathologies affected by this polymorphism in humans overlap with those in which the outcome is exacerbated by HO-1 deletion (e.g., endotoxin shock, severe sepsis, atherosclerosis, myocardial infarction, ischemia reperfusion injury). Additionally, one study correlated the presence of the shorter (GT)n polymorphism with a longer life of the individual. Individuals who lived longer were more likely to have the "high transcriptional induction" genotype, associated with the short HO polymorphism (Kimpara et al., 1997; Hirai et al., 2003; Denschlag et al., 2004; Exner et al., 2004).

Studies comparing the outcome of several experimental pathologic conditions in *HO-1*−*/*<sup>−</sup> or wild-type mice show that when HO-1 is lacking, the pathologic conditions studied are exacerbated and result in high incidence of mortality. Pharmacological inhibition of HO activity mimics the phenotypes associated with HO-1 deletion, while, on the other hand, induction of HO-1 or administration of its reaction products, CO and biliverdin, is protective and usually ameliorates the pathologic conditions. The cytoprotective effects of HO-1 and/or of its final products, CO and biliverdin/bilirubin have been demonstrated in a variety of diseases, including immune-mediated inflammatory diseases (rejection of transplanted organs, autoimmune diseases, asthma, arthritis, colitis, pancreatitis, recurrent abortions), infectious diseases (severe malaria, sepsis, endotoxin shock), cardiovascular diseases (atherosclerosis, myocardial infarction, endothelial dysfunction, vaso-occlusion) and ischemia-reperfusion injury (Gozzelino et al., 2010).

Several molecules that exert a beneficial effect against different diseases are known to act via HO-1 induction. Some of these synthetic compounds include sialic acid, statins and rapamycin. In addition, several molecules produced under physiologic conditions might act in a similar manner, including the anti-inflammatory cytokine interleukin (IL)-10, some prostaglandins, VEGF (vascular endothelial growth factor), stromal cell-derived factor 1, nitric oxide (NO) and NGF (nerve growth factor) (Gozzelino et al., 2010). The mechanisms via which these molecules could induce the expression of HO-1 is the inhibition of Bach1 activity and the activation of transcription factors that promote HO-1 transcription, such as Nrf2. Alternatively, therapeutic induction of endogenous HO-1 might be obtained via the delivery of non-cytotoxic amount of heme or heme-containing proteins, the so-called preconditioning. In this manner, the administration of hemoglobin or heme was shown to improve survival in response to endotoxic shock (Otterbein et al., 1995), to reduce liver and kidney injury (Nath et al., 1992) and decrease vascular stasis (Belcher et al., 2006, 2010). Similarly, the preconditioning of Hx−*/*<sup>−</sup> mice with a small dose of hemin strongly decreases their susceptibility to acute heme overload, thus limiting heme-mediated tissue damage (Vinchi et al., 2008). Interestingly, a physiological paradigm of heme preconditioning is represented by sickle cell trait, in the heterozygous form. This mutation confers protection against severe forms of malaria due to the low concentrations of heme in the circulation that lead to induction of endogenous HO-1 expression and protection against *Plasmodium* infection (Ferreira et al., 2011). In this condition, chronic hemolysis associated to sickle cell disease induces a state of malaria tolerance by inducing HO-1 up-regulation, which directly supports heme degradation and drives the generation of anti-inflammatory and anti-oxidant metabolites. Additionally, the ability of CO to directly bind hemoglobin and inhibit its oxidation, thus avoiding heme release from oxidized hemoglobin, has been recently shown to prevent experimental cerebral malaria, thus highlighting a new and key mechanism of HO-mediated protection in pathologies associated with hemoglobin/heme release (Pamplona et al., 2007; Ferreira et al., 2008; Seixas et al., 2009).

Clinical trials based on the use of CO are currently ongoing to evaluate the potential of CO as a therapeutic agent in humans. CO is administered by inhalation for the treatment of pathologies such as pulmonary arterial hypertension, post-operative ileus and idiopathic pulmonary fibrosis (see http://www*.*clinicaltrials*.*gov website). Alternatively, in the last decade a great effort was made to design and study molecules able to bind and deliver CO in biological systems, known as CO-releasing molecules (CO-RMs) (Motterlini et al., 2005). In the next years the evaluation of the therapeutic potential of CO via inhalation or CO-RM administration will reveal whether this way may be pursued as a therapeutic strategy to counteract pathological outcome due to heme-driven toxicity.

## **PATHOLOGICAL CONDITIONS ASSOCIATED WITH MUTATIONS IN GENES CODING FOR HEME TRANSPORTERS POSTERIOR COLUMN ATAXIA AND RETINITIS PIGMENTOSA (PCARP)**

Recently, *FLVCR1* has been reported to be the causative gene for PCARP (OMIM: #609033) (Rajadhyaksha et al., 2010; Ishiura et al., 2011) (**Table 2**). PCARP is a childhood-onset, autosomalrecessive, neurodegenerative syndrome with the clinical features of sensory ataxia and retinitis pigmentosa. PCARP begins in infancy with areflexia and retinitis pigmentosa. During infancy, night blindness, peripheral visual loss with subsequent progressive constriction of the visual field and loss of central retinal function became apparent. The sensory ataxia caused by the degeneration of the posterior column of the spinal cord results in a loss of proprioceptive sensation. Scoliosis, camptodactyly, achalasia, gastrointestinal dysmothility and a sensory peripheral neuropathy are variable features of the disease. PCARP is considered a sensory ganglionopathy causing a degeneration of central and peripheral axons without evidence of primary demyelination (Higgins et al., 1997).

Four different homozygous mutations in *FLVCR1* gene have been reported to date. Interestingly, three of these mutations (c.361A*>*G, c.574T*>*C, and c.721G*>*A) (Rajadhyaksha et al., 2010) occur in the first exon of *FLVCR1* gene, thus probably affecting only FLVCR1a. However, we cannot exclude the possibility that these mutations could interfere with any still unknown regulatory sequences important for FLVCR1b expression. The other mutation (c1477G*>*C) (Ishiura et al., 2011) is in the tenth exon of *FLVCR1* gene, common to both *Flvcr1a* and *Flvcr1b* transcripts (Chiabrando et al., 2012), thus probably affecting both transporters. Each mutation corresponds to a specific aminoacid substitution (Asn121Asp, Cys192Arg, Ala241Thr, and Gly493Arg) within putative transmembrane domains of FLVCR1: 1, 3, 5, and 12 transmembrane domains, respectively. Yanatori et al. (2012) investigated the consequences of the reported *FLVCR1* mutations on the plasma membrane heme exporter FLVCR1a. *In vitro* studies indicate that all the four identified mutations affect the subcellular localization, half-life and hemeexport function of FLVCR1a (Yanatori et al., 2012). However, the consequences of all these mutations on the expression, localization and activity of FLVCR1b have not been investigated yet. It has been proposed that the mutant FLVCR1a proteins fail to fold properly in the endoplasmic reticulum and are rapidly degraded in the lysosomes; loss of FLVCR1a could lead to heme overload, oxidative stress and apoptosis of photoreceptors in the retina and sensory neurons in the posterior column of the spinal cord of PCARP patients (Yanatori et al., 2012). Actually, this is merely a hypothesis and evidences of heme overload, heme-induced toxicity, and apoptosis in PCARP patients or mouse models of the disease are still lacking. Of note, apoptosis is the final and common cause of photoreceptors degeneration in several forms of retinitis pigmentosa (Marigo, 2007).

The reason why the loss of FLVCR1a could affect two specific sensory modalities, vision and proprioception, is unknown. The expression levels of *Flvcr1* mRNA have been confirmed in the mouse brain (neocortex, striatum, hippocampus, and cerebellum), posterior column of the spinal cord, retina and retinal pigment epithelium. The highest levels of *Flvcr1* mRNA have been found in the retina and spinal cord thus suggesting a correlation between the regional specificity of *Flvcr1* mRNA expression and the selective degenerative pathology of PCARP (Rajadhyaksha et al., 2010; Gnana-Prakasam et al., 2011). However, FLVCR1 is ubiquitously expressed (Quigley et al., 2004; Chiabrando et al., 2012) and the observation that mutations in the *FLVCR1* gene cause sensory ataxia and retinitis pigmentosa was completely unexpected. Two different mouse models of FLVCR1 deficiency have been generated; in both cases, the loss of specific FLVCR1 isoforms severely affects embryonic development (Keel et al., 2008; Chiabrando et al., 2012), a stronger phenotype compared to that of PCARP patients. Furthermore, the mouse models of *Flvcr1* deficiency suggest an essential role of FLVCR1 isoforms in erythroid differentiation and in the maintenance of endothelial integrity (Keel et al., 2008; Chiabrando et al., 2012). PCARP patients did not suffer from anemia (Rajadhyaksha et al., 2010; Yanatori et al., 2012). On the other hand, neurological defects have not been reported in mouse models of *Flvcr1* deficiency. The reasons of the discrepancies between the human disease and the mouse models need to be further investigated.

Further work is needed to understand how the loss of a ubiquitously expressed heme exporter could lead to the selective pathological features of PCARP.

#### **DIAMOND BLACKFAN ANEMIA (DBA)**

A role for *FLVCR1* in the pathogenesis of Diamond Blackfan Anemia (DBA; OMIM: #105650) has been proposed due to the observation that the feline leukemia virus subgroup C causes a pure cell aplasia in cats by interfering with the expression of FLVCR1 and that mice lacking both FLVCR1 isoforms phenocopy the human disease (Tailor et al., 1999; Quigley et al., 2000; Keel et al., 2008).

DBA is an autosomal dominant disorder with incomplete penetrance due to mutations in several genes coding for ribosomal proteins; the most common mutation is in ribosomal protein S19 gene (RPS19) (Campagnoli et al., 2008). DBA patients are characterized by a severe block of erythroid differentiation at the proerythroblast stage associated to congenital malformations and cancer predisposition (Flygare and Karlsson, 2007; Chiabrando and Tolosano, 2010; Narla and Ebert, 2010). Accordingly, FLVCR1 deficient mice are characterized by a block of erythroid differentiation as well as growth defects (Keel et al., 2008).

Mutations in *FLVCR1* gene have not been identified in DBA patients (Quigley et al., 2005). Rey et al. (2008) identified several alternatively spliced *Flvcr1* transcripts, encoding proteins with aberrant expression and function. Due to the experimental strategy, only *Flvcr1a* transcript was evaluated in this work. Interestingly, the aberrant alternative splicing of *Flvcr1a* is increased in immature erythroid cells of some DBA patients negative for RPS19 mutations, while the expression of the wild-type protein is decreased (Rey et al., 2008). The molecular mechanism leading to the aberrant alternative splicing of *Flvcr1a* in DBA is unknown. These data suggest that decreased FLVCR1a expression and alteration of heme metabolism could contribute to the pathogenesis of DBA.

The identification of FLVCR1b essential role in heme synthesis and erythroid differentiation, leads to the hypothesis that alteration of FLVCR1b expression could also concur to the pathogenesis of DBA (Chiabrando et al., 2012). Whether aberrant alternative splicing of *Flvcr1b* transcript also occurs in DBA immature erythroid cells, still needs to be addressed.

Mouse models of *Flvcr1* deficiency are also characterized by defective growth (Keel et al., 2008) and skeletal malformations (Chiabrando et al., 2012). Thus, FLVCR1a could also have a role in the development of the congenital malformations observed in DBA patients.

Although *FLVCR1* is far to be the causative gene for DBA, we could speculate that the aberrant expression of FLVCR1 isoforms could contribute to different pathological features of DBA.

Considering the essential role of FLVCR1 isoforms during erythroid differentiation, it will be interesting to analyze their expression also in other erythroid diseases. An alteration of the expression of FLVCR1 isoforms may play a role in the pathogenesis of disorders characterized by an imbalance between heme and globin synthesis too.

#### **PUTATIVE ROLE OF FLVCR1 IN DRUG METABOLISM**

The recent finding that liver conditional *Flvcr1a*−*/*<sup>−</sup> mice show a reduced expression and activity of cytochromes P450 suggests that heme export may have an impact on cytochrome function. About 50% of hepatic heme is used for synthesis of P450 enzymes, which metabolize exogenous and endogenous compounds, including hormones, xenobiotics, drugs, and carcinogens. It is well-known that a reduction in heme availability due to an enhanced heme degradation leads to the impairment of cytochrome function (Correia et al., 2011). On the other hand, the lack of FLVCR1a-mediated heme export in hepatocytes, although leading to an increase in the cytosolic heme pool, similarly causes a reduction in cytochrome activity. In hepatocytes, heme is likely formed in excess over its metabolic needs and heme homeostasis is ensured by a combination of synthetic, degradative, and export mechanisms. FLVCR1a, by exporting heme excess out of the cell, controls the size of the cytosolic heme pool, thus allowing proper heme synthesis and new cytochrome formation. FLVCR1a loss causes heme synthesis inhibition, due to the expansion of the cytosolic heme pool, ultimately leading to a decrease in cytochrome function (Vinchi et al., 2014). These observations have potential implications for hepatic metabolism of xenobiotics and drugs. In particular, deletion or mutations in *Flvcr1a* and other pathologic conditions that reduce its expression potentially cause a reduction in cytochrome activity, thus altering drug metabolism. Therefore, individuals that routinely assume drugs for therapeutic purposes may show a reduced ability to induce cytochromes P450 and to metabolize pharmaceuticals, thus being more susceptible to drug intoxication. In the near future, the mouse model of FLVCR1a deletion in hepatocytes could be applied for metabolic studies to address the importance of hepatic heme export upon drug administration.

## **HYPERURICAEMIA AND GOUT**

As stated above, ABCG2 was initially shown to transport a wide range of substrates including drugs, xenobiotics, food metabolites and heme. Only recently, genetic studies in humans have identified urate as a physiological substrate of ABCG2, perhaps the most relevant. Indeed, a genome-wide association study identified common alleles in *ABCG2* as associated with serum urate level and risk of gout (OMIM: #138900) (**Table 2**) (Dehghan et al., 2008). ABCG2 expression in the apical membrane of human proximal tubule cells (Doyle and Ross, 2003), the main site of urate handling in the kidney, is consistent with a role for ABCG2 in urate excretion. Gout is a common form of inflammatory arthritis and occurs when uric acid crystallizes in the form of monosodium urate, precipitating in joints, on tendons, and in the surrounding tissues. These crystals then trigger a local immune-mediated inflammatory reaction. Genetic studies identified a SNP in the exon 5 of *ABCG2* gene that leads to a glutamine-to-lysine amino acid substitution (Q141K) (Dehghan et al., 2008). The Q141K variant has been shown to have a significant reduced capacity to transport urate and, when expressed in mammalian cells, Q141K ABCG2 expression is significantly lower than that of wild-type protein (Imai et al., 2002; Mizuarai et al., 2004; Morisaki et al., 2005; Woodward et al., 2009). Another study on the Japanese population identified a different variant as associated to hyperuricemia and gout (Yamagishi et al., 2010).

Many additional SNPs in *ABCG2* gene have been identified and associated to altered drug resistance (Woodward et al., 2011). Future studies will have to address whether additional SNPs also affect serum urate concentration in human. Finally, how these polymorphisms affect heme transport remains to be determined.

## **FOWLER SYNDROME**

Mutations in the *FLVCR2* gene have been associated to the Fowler syndrome (OMIM: #225790) (**Table 2**), a rare lethal autosomalrecessive disorder of brain angiogenesis, first described in 1972 (Fowler et al., 1972). The Fowler syndrome is a cerebral proliferative glomeruloid vasculopathy resulting in abnormally thickened and aberrant perforating vessels, forming glomeruloid structures throughout the central nervous system. These defects lead to hydranencephaly and hydrocephaly and are restricted to the central nervous system parenchyme. In addition to cerebral proliferative glomeruloid vasculopathy, fetuses affected by the Fowler syndrome also show limb deformities.

Several kinds of mutations were identified in cases of Fowler syndrome (Meyer et al., 2010; Thomas et al., 2010), but it is not clear how these mutations could affect *FLVCR2* expression, functions or localization.

The mechanism by which *FLVCR2* mutations could lead to abnormal angiogenesis is also unknown. Some hypotheses have been put forward.

First, it has been postulated that the role of FLVCR2 in calcium trafficking, proposed by Brasier et al. (2004), could explain its involvement in Fowler syndrome. Indeed, the altered angiogenesis in Fowler syndrome has been supposed to be due to a deficit in pericytes, cells essential for capillary stabilization and remodeling during brain angiogenesis (Thomas et al., 2010). These cells interact with endothelial cells, whose proliferation and motility are calcium-dependent events. Moreover, the Fowler syndrome has been associated with varying degrees of calcification throughout the central nervous system (Harding et al., 1995).

Although possible, this hypothesis seems to be inadequate because the involvement of FLVCR2 in calcium transport has not been demonstrated. Moreover, it is not clear whether the effects on pericytes observed in fetuses affected by Fowler syndrome is the primary cause or an effect of the disease.

A second hypothesis to explain *FLVCR2* implication in the Fowler syndrome is based on its heme import activity. It has been observed in some cases of Fowler syndrome that there is a deficiency in the heme-dependent complexes IV of the mitochondrial electron transport chain, essential for the production of adenosine triphosphate by oxidative phosphorylation. Since young neurons that migrate to the neocortex are dependent on oxidative phosphorylation, it has been speculated that *Flvcr2* disruption could lead to dysfunction of the mitochondrial electron transport chain, thus causing the neurodegeneration and the developmental abnormalities found in Fowler syndrome (Duffy et al., 2010).

Finally, a third hypothesis is that alteration on FLVCR2 hemeimport activity could account for cellular iron overload, a condition known to be associated to neurodegeneration and bone abnormalities (Duffy et al., 2010).

Although all conceivable, the three hypotheses need experimental verification and it remains unclear why the damaged areas in Fowler syndrome are restricted to the central nervous system and whether the proliferative vasculopathy is primary or secondary to the neurodegeneration.

## **HEREDITARY FOLATE MALABSORPTION SYNDROME**

The only known disorder associated to mutations in the *HCP1/PCFT* gene relates to its role in the transport of folates, while the effects of the loss of its heme-import functions seem to be negligible, suggesting the instauration of alternative compensatory mechanisms to support cellular heme-iron uptake in the absence of HCP1/PCFT.

Mutations in the *HCP1/PCFT* gene have been associated to the hereditary folate malabsorption syndrome (HFM; OMIM: #229050), a rare autosomal recessive disorder first described in 1961 (Luhby et al., 1961) (**Table 2**).

Folates are essential cofactors required for key epigenetic and biosynthetic processes, such as de novo synthesis of purines and pyrimidines, methionine and deoxythymidylate monophosphate. Mammals do not synthesize folates, and body needs are met by duodenal and jejunal absorption of folates contained in the diet.

The basis for HFM is the impaired intestinal folate absorption, resulting in severe systemic and cerebrospinal fluid folate deficiency in a few months after birth.

Patients affected by HFM show anemia, diarrhea, pancytopenia, hypoimmunoglobulinemia, and frequent neurological dysfunctions like developmental delays, cognitive impairment, and epilepsy. Many kinds of mutations have been reported (Qiu et al., 2006; Zhao et al., 2007; Shin et al., 2011; Diop-Bove et al., 2013), occurring in different sites of *HCP1/PCFT* gene and with different effects on the protein encoded by the mutated alleles.

If untreated, the disease is fatal and, if treatment is delayed, the neurologic deficits can become permanent. Normalization of blood, and rarely cerebrospinal fluid, folate level could be achieved by administration of low doses of parenteral folates or of pharmacological doses of oral folates. In the latter case, folates are probably absorbed by the bidirectional Reduced Folate Carrier 1 (RFC1), a folate importer alternative to HCP1/PCFT, which is expressed in the upper small intestine. RFC1 shows a lower affinity for folates as compared with HCP1/PCFT, but can compensate for HCP1/PCFT loss in HFM patients treated with high doses of folates (Qiu et al., 2006; Zhao et al., 2007).

The *Hcp1/Pcft*−*/*<sup>−</sup> mouse represents the murine model for this HFM syndrome (Salojin et al., 2011) (**Table 1**). *Hcp1/Pcft*−*/*<sup>−</sup> mice die by 10–12 weeks of age and show a haematological profile which closely resembles that observed in human HFM patients.

## **CONCLUDING REMARKS**

The excess of free-heme is toxic due to its high reactivity and its ability to promote oxidative stress on lipids, proteins and nucleic acids, eventually leading to cell death.

The existence of multiple levels of heme control indicates that mammals specifically evolved systems able to limit heme amount outside and inside the cell to counteract heme potential toxicity.

As a primary defence mechanism, Hx scavenges free circulating heme, thereby avoiding it to exert its pro-oxidant effects. Indeed, Hx saturation observed in hemolytic disorders leads to heme-driven tissue damage and oxidative stress.

The control of heme synthesis and degradation has a key role in the regulation of intracellular heme levels and these two mechanisms are respectively inhibited and induced by heme itself. Alterations in these processes lead to pathological iron or heme accumulation. Mutations in *ALAS2* gene cause insufficient heme production, mitochondria iron overload, and oxidative stress, as observed in sideroblastic anemia (Sheftel et al., 2009). HO-1 deficiency leads as well to anemia and inability to efficiently reuse tissue heme-iron. Moreover, HO-1 has a cytoprotective action that resides not only in the degradation of toxic heme but also in the production of the antioxidant molecules CO and biliverdin.

The recent identification of plasma membrane heme transporters suggests that intracellular heme amount could be regulated not only at the level of its synthesis and catabolism but also at the level of heme export and import. Nevertheless, apart FLVCR1 for which studies in mouse models have provided compelling evidence that it exerts a physiologic role as a heme exporter, the experimental evidence that ABCG2, FLVCR2, and HCP1/PCFT are heme transporter with a physiologic role in heme/iron homeostasis are very poor. Thus, it would be more correct to refer to these molecules as secondary regulators of heme and/or porphyrin metabolism. This conclusion is corroborated by genetic studies in humans that failed in finding a clear association between mutations in genes coding for heme transporters and disorders of heme metabolism as was expected from studies on cellular and animal models. Mutations in *FLVCR1* or *FLVCR2* affect specific cell populations, sensory neurons and photoreceptors or vascular cells of the central nervous system, respectively. Polimorphisms in *ABCG2* result in defective urate excretion while mutations in *HCP1/PCFT* cause folate malabsorption. These discrepancies may be related to the different kind of mutations taken into considerations, loss-of-function mutations in mice and point mutations resulting in proteins with altered function/localization in humans. It is possible that null mutations can affect heme metabolism also in humans or that specific polimorphisms in genes coding for heme transporters can differentially alter the import/export of specific substrates. The full comprehension of the role of FLVCR1-mediated heme export in human physiology and pathology is a challenge for future studies as well as the definition of the role of FLVCR2, ABCG2, and HCP1/PCFT in heme metabolism. Of course, the definition of protein structure and the identification of residues critical for heme transport in heme transporters are very important issues. In addition, the description of the regulation of these transporters in different tissues and following different stimuli will be of relevance to fully understand the multi-level system of heme control.

In the future, the comprehension of these mechanisms could have important clinical implications. Targeting heme transporters could be therapeutically useful to prevent hemedriven tissue damage in pathologies characterized by enhanced oxidative-stress.

## **REFERENCES**


cause nonsyndromic autosomal recessive congenital sideroblastic anemia. *Nat. Genet.* 41, 651–653. doi: 10.1038/ng.359


Lecha, M., Puy, H., and Deybach, J. C. (2009). Erythropoietic protoporphyria. *Orphanet. J. Rare Dis.* 4:19. doi: 10.1186/1750-1172-4-19


linkage interval reveals FLVCR2 deletions and mutations in lethal cerebral vasculopathy. *Hum. Mutat.* 31, 1134–1141. doi: 10.1002/humu.21329


Zhu, Y., Lee, H. C., and Zhang, L. (2002). An examination of heme action in gene expression: heme and heme deficiency affect the expression of diverse genes in erythroid k562 and neuronal PC12 cells. *DNA Cell Biol.* 21, 333–346. doi: 10.1089/104454902753759744

**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

*Received: 15 January 2014; paper pending published: 16 February 2014; accepted: 18 March 2014; published online: 08 April 2014.*

*Citation: Chiabrando D, Vinchi F, Fiorito V, Mercurio S and Tolosano E (2014) Heme in pathophysiology: a matter of scavenging, metabolism and trafficking across cell membranes. Front. Pharmacol. 5:61. doi: 10.3389/fphar.2014.00061*

*This article was submitted to Drug Metabolism and Transport, a section of the journal Frontiers in Pharmacology.*

*Copyright © 2014 Chiabrando, Vinchi, Fiorito, Mercurio and Tolosano. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*

## Like iron in the blood of the people: the requirement for heme trafficking in iron metabolism

## *Tamara Korolnek1,2 and Iqbal Hamza1,2 \**

<sup>1</sup> Department of Animal & Avian Sciences, University of Maryland, College Park, MD, USA

<sup>2</sup> Department of Cell Biology & Molecular Genetics, University of Maryland, College Park, MD, USA

#### *Edited by:*

Raffaella Gozzelino, Instituto Gulbenkian de Ciência, Portugal

#### *Reviewed by:*

Caroline C. Philpott, National Institutes of Health, USA John G. Quigley, University of Illinois at Chicago, USA

#### *\*Correspondence:*

Iqbal Hamza, Department of Animal & Avian Sciences and Department of Cell Biology & Molecular Genetics, University of Maryland, 2413 ANSC, Building 142, College Park, Maryland 20742, USA e-mail: hamza@umd.edu

Heme is an iron-containing porphyrin ring that serves as a prosthetic group in proteins that function in diverse metabolic pathways. Heme is also a major source of bioavailable iron in the human diet. While the synthesis of heme has been well-characterized, the pathways for heme trafficking remain poorly understood. It is likely that heme transport across membranes is highly regulated, as free heme is toxic to cells. This review outlines the requirement for heme delivery to various subcellular compartments as well as possible mechanisms for the mobilization of heme to these compartments. We also discuss how these trafficking pathways might function during physiological events involving interand intra-cellular mobilization of heme, including erythropoiesis, erythrophagocytosis, heme absorption in the gut, as well as heme transport pathways supporting embryonic development. Lastly, we aim to question the current dogma that heme, in toto, is not mobilized from one cell or tissue to another, outlining the evidence for these pathways and drawing parallels to other well-accepted paradigms for copper, iron, and cholesterol homeostasis.

**Keywords: heme, iron, transport, hematology, porphyrins, anemia, erythropoiesis, nutrition disorders**

## **INTRODUCTION**

Heme is an iron-containing porphyrin that functions as a cofactor in a wide array of cellular processes. Heme is also a major source of bioavailable iron in the human diet. By the end of the 20th century, the pathways for heme biosynthesis had been well elucidated and the structures of many heme-containing proteins have been solved. Indeed, Max Perutz and John Kendrew were awarded the Nobel Prize in chemistry in 1962 for their crystal structures of hemoglobin and myoglobin (Green et al., 1954; Kendrew et al., 1958). While the synthesis of heme has been wellcharacterized, the pathways for inter- and intra-cellular heme transport remain poorly understood. This gap in our knowledge is largely due to the inability to uncouple the processes of heme biosynthesis and heme transport, as well as heme's ability to promiscuously bind to proteins. By contrast, great strides have been made in the fields of iron and copper trafficking, which have not suffered similar setbacks. Transport pathways for non-heme iron have been studied for decades and, all the while, knowledge of heme-iron trafficking has languished. The title of this review includes a verse from the poem "Kids Who Die" by the American poet Langston Hughes raising the possibility that heme too moves *like iron in the blood of the people*.

This review aims to (a) outline the biological requirement for heme trafficking pathways in eukaryotes, including the inability of heme to efficiently traverse lipid membranes and the presence of hemoproteins in almost all subcellular compartments, (b) summarize how these putative pathways may participate in various biological processes, including the synthesis and recycling of red blood cells (RBCs), intestinal absorption of dietary heme, embryogenesis, and (c) summarize the evidence for inter-cellular and inter-tissue transport of heme.

## **MOVEMENT OF HEME ACROSS MEMBRANES**

The terminal step of heme biosynthesis, which occurs in the mitochondrial matrix, is the insertion of iron into protoporphyrin IX. As will be discussed in depth in this article, heme is then targeted to both soluble and membrane-bound hemoproteins. Heme must be able to cross membranes, yet also be delivered to soluble proteins in the cytosol. Heme's amphipathic nature means that it is able to function in both milieus, but not without the assistance of partners. Due to the hydrophobicity of the porphyrin ring, heme is not readily soluble in aqueous solutions. When heme is at neutral pH and is in the ferrous state it has no net charge, as the negative charges of the propionate groups balance the positive charges of the chelated iron. While this form of heme readily inserts into and diffuses through lipid bilayers, the hydrophilic propionate head groups hinder the flipping of heme from one leaflet to another (Cannon et al., 1984; Rose et al., 1985; Light and Olson, 1990). Heme exit from the membrane is inefficient without the presence of soluble heme-binding proteins (HBPs; Yoda and Israels, 1972).

In addition to the difficulties posed by heme's dual chemical nature, heme is highly toxic due to its peroxidase activity and capacity to generate reactive oxygen species combined with its ability to intercalate into membranes and bind many proteins non-specifically. The toxic effects of heme have been reviewed extensively in mammalian cells and other systems (Vincent, 1989; Jeney et al., 2002; Oliveira et al., 2002; Kumar and Bandyopadhyay, 2005; Chow et al., 2008; Larsen et al., 2012). Thus, even though heme is able to diffuse into membranes and be extracted

by HBPs, it is likely that this process would be regulated, as peroxidation of membrane lipids would result in severe damage, especially during a process such as erythropoiesis, with each red cell processing the heme required for over 300 million hemoglobin molecules. Additionally, many of heme synthesis precursors and heme breakdown products are also toxic, and thus careful control of heme synthesis, trafficking, and degradation is a prerequisite.

Given these chemical and biochemical properties of heme, it is apparent that a system of transporters and chaperones would be necessary to efficiently and specifically distribute heme to hemoproteins found in various compartments in the cell, all the while preventing adventitious heme toxicity.

## **REQUIREMENTS AND MECHANISMS FOR HEME DELIVERY TO SUBCELLULAR COMPARTMENTS**

Hemoproteins can be found in virtually every subcellular compartment of eukaryotic cells. In this section we will outline the requirement for heme delivery from its site of synthesis in the matrix, to the cytosol, to the secretory compartment, as well as other possible direct shuttling routes between organelles (**Figure 1**).

#### **HEME REQUIREMENTS IN THE MITOCHONDRIA**

Most of the mitochondrial respiratory chain complexes rely on the redox capability of a heme prosthetic group (**Table 1**). Heme is present in these complexes in the *a*, *b*, or *c* forms. Heme *b* is the most common form of heme. It is the form synthesized by ferrochelatase, containing a C2 vinyl group and C8 methyl group, and is attached to proteins via non-covalent coordination of the heme iron with amino acid side chains (Gonzales and Neupert, 1990; Hederstedt, 2012). Heme *a* is modified from heme *b* and its porphyrins ring contains a C2 hydroxyethylfarnesyl and a C8 formyl side group (Hederstedt, 2012). Heme *c* is found covalently linked to proteins; the two vinyl side chains are converted to thioether linkages to cysteine residues in the apoprotein (Bowman and Bren, 2008).

The biogenesis and cofactor requirements of the mitochondrial respiratory complexes have recently been reviewed, but will be briefly discussed here with regard to heme transport and trafficking (Kim et al., 2012; Smith et al., 2012; Soto et al., 2012). The cytochrome *bc*<sup>1</sup> complex (also known as coenzyme Q-cytochrome *c* reductase or Complex III) is a multi-subunit protein located with the other electron transport complexes in the inner mitochondrial membrane (IMM). The cytochrome *bc*<sup>1</sup> complex includes two heme *b*-containing cytochromes as well as a heme *c*-containing cytochrome, and interacts with another hemoprotein, the electron carrier cytochrome *c*, for the purpose of generating a proton gradient along the IMM. Cytochrome *c* is soluble in the intermembrane space (IMS) of mitochondria and is found loosely associated with the IMM, as it carries an electron from the cytochrome *bc*<sup>1</sup> complex to cytochrome *c* oxidase (COX). COX (also known as Complex IV) includes two heme *a*-containing cytochromes, which also participate in electron transfer and proton translocation into the IMS. Succinate dehyrogenase (Complex II) also contains a heme *b* moiety, which is actually required for stability of the complex (Lemarie and Grimm, 2009). The heme

*b* in succinate dehydrogenase may also be involved in electron transport, but the exact function of this heme remains unclear (Kim et al., 2012).

In eukaryotes, the cytochrome *bc*<sup>1</sup> complex (Complex III) is composed of seven or eight nuclear-encoded subunits that function together with three catalytic core proteins encoded in the mitochondrial genome (Xia et al., 1997; Iwata et al., 1998; Lange and Hunte, 2002). Cytochrome *b* is a membrane-bound core protein that contains two heme groups, each of which has a unique redox potential (Zhang et al., 1998). Heme *b*H, the high potential heme, is bound in an accessible cavity on the matrix side of the protein. Heme *b*L, the low potential heme, is found within the IMS portion of cytochrome *b*. Heme *b*<sup>H</sup> can be inserted into cytochrome *b* without having to traverse a membrane; it is unknown whether or not a chaperone assists in this process (Brasseur et al., 1996). It is unclear if heme *b*<sup>L</sup> is acquired in the matrix, or if this heme is inserted in the IMS, thus necessitating movement of heme across the IMM (**Figure 1**, Pathway 1). Cytochrome *c*1, being nuclear-encoded, is targeted to the mitochondria after being translated in the cytosol. Its heme cofactor is covalently attached in the IMS by the heme lyases Cyc3 and Cyt2 in yeast or holocytochrome c synthase (HCCS) in mammals (Bernard et al., 2003). In addition to the heme in cytochrome *c*<sup>1</sup> and possibly the heme *b*<sup>H</sup> of cytochrome *b*, the soluble respiratory protein cytochrome *c* also acquires its heme in the IMS from HCCS (Bernard et al., 2003).

The biogenesis of COX (Complex IV) in highly conserved in eukaryotes (Soto et al., 2012). Heme *a* in COX is bound to the mitochondrially encoded COX1 subunit. Heme *a* is synthesized via the conversion of heme *b* to the intermediate heme *o* by the farnesylation of the C2 position by COX10 (heme *o* synthase), and the subsequent oxidation of the C2 methyl side chain to formyl by COX15 (heme *a* synthase; Hederstedt, 2012; Soto et al., 2012). Both COX10 and COX15 are integral membrane proteins located in the IMM. Studies have indicated that insertion of heme *a* into COX1 occurs post-translationally, as early intermediates of the COX assembly do not contain heme *a* and are able to form in the absence of heme *a* biosynthesis (Khalimonchuk et al., 2010). This is in contrast with later intermediates, which do not form without the presence of heme *a*. It remains unclear if heme chaperones exist to shuttle heme *b* from its site of synthesis to COX10, or to deliver heme *a* to the COX1 subunit. By contrast, multiple mitochondrial chaperones, including COX17, SCO1, and COX11, have been shown to act as chaperones for delivery of copper to the COX complex [Also reviewed in Soto et al. (2012)].

In addition to the respiratory chain complex proteins, another cytochrome, cytochrome P450 side chain cleavage enzyme (also known as CYP11A1), is also found in the mitochondrial matrix. This enzyme is expressed in steroid-synthesizing tissues such as the brain and adrenal glands and functions to convert cholesterol to pregnenolone, a precursor for mineralocorticoids, glucocorticoids, androgens, and estrogens (Simpson and Boyd, 1967). CYP11A1 catalyzes this first step in steroid-synthesis by initiating three sequential monooxygenase reactions of the cholesterol side chain: hydroxylation of C20 and C22 and then cleavage the C20–C22 bond to generate the steroid precursor pregnenolone and isocaproic aldehyde. While the crystal structures of both bovine

and human CYP11A1 have recently been solved, nothing is known regarding how the heme cofactor is inserted into this important hemoprotein (Mast et al., 2011; Strushkevich et al., 2011).

#### **HEME REQUIREMENTS IN THE CYTOSOL**

While heme moves from the mitochondrial matrix into the IMS for insertion into particular cytochromes, heme must also exit the mitochondria for insertion into hemoproteins which acquire heme in the cytosol (**Table 1**). The most well known cytosolic hemoproteins are the gas-binding globins, including hemoglobin, myoglobin, neuroglobin, and cytoglobin. Adult hemoglobin consists of two α and two β chains, each bound to a molecule of heme. Hemoglobin can be generated *in vitro* (for example, by heterologous expression in *Escherichia coli,* or by recombining purified apo-hemoglobin with heme) indicating that no specific chaperone is required for heme insertion. Because of its asymmetric side chains, each heme can be incorporated into hemoglobin in two orientations. The term "heme disorder" refers to mixtures of these two orientations within a subset of hemoglobin molecules (Brown, 1976). Interestingly, the amount of heme disorder differs between hemoglobin synthesized *in vivo* or*in vitro* (Santucci et al., 1990). This is a possible indication that stereoselective insertion of heme into hemoglobin occurs, although the mechanism for this phenomenon remains unclear.

Other categories of hemoproteins that acquire their heme in cytosol include nitric oxide (NO) synthases and the soluble guanylyl cyclases (a NO receptor), the nuclear HBPs Hap1p, DGCR8, mPer2, Bach1 and Rev-erb-α, as well as the soluble form of cytochrome *b*5. Even P450 cytochromes, the most well known hemoproteins in the ER, are anchored in the ER membrane with their heme-containing globular domain facing the cytosol (Black, 1992; Avadhani et al., 2011). Thus, cytochrome P450 proteins likely acquire their heme from the cytosolic face of the membrane.

## **HEME DELIVERY TO CYTOSOLIC PROTEINS**

Cytosolic hemoproteins can acquire their heme from two possible sources: heme synthesized *de novo* in the mitochondria, or heme imported from outside the cell. Heme synthesized in mitochrondria exits the active site pocket of ferrochelatase, which faces the matrix side of the IMM (Wu et al., 2001). Exactly how heme exits ferrochelatase remains unknown. It is possible that this heme immediately intercalates into the IMM or is actively transported across the IMM and then is removed into the IMS by a hemebinding chaperone and then moved across the OMM by a high affinity heme transporter (**Figure 1**, Pathways 1 and 2).

The sole mitochondrial heme exporter identified to date is the mitochondrial isoform of Flvcr1 (Quigley et al., 2004; Keel et al., 2008). Flvcr1, a member of the major facilitator superfamily

#### **Table 1 | List of selected eukaryotic hemoproteins.**


of transporters, was initially identified as a plasma membrane heme exporter. Cats infected with feline leukemia virus, subgroup C (FeLV-C) develop aplastic anemia due to failure of erythroid differentiation of burst-forming units-erythroid (BFU-E) to colony-forming units-erythroid (CFU-E). Viral infection of these erythroid progenitors interferes with and possibly downregulates cell surface expression of the viral receptor, Flvcr1 (Quigley et al., 2004). Early experiments showed that Flvcr1 could mediate cellular efflux of either exogenously added or endogenous heme (Quigley et al., 2004; Yang et al., 2010). Mice lacking Flvcr1 die either at E7.5 or between E14.5 and E16.5. Similar to Hmox1, Flvcr1 is expressed in the yolk sac, ectoplacental cone, and placenta at E7.5. Thus the early death of Flvcr1 null embryos is likely due to heme accumulation within cells, although it is also possible that Flvcr1 is required for delivery of maternal heme. Keel et al. (2008) conclude that Flvcr1 null embryos die at the later stage due to impaired erythropoiesis. Red cells in these mice are blocked at

the proerythroblast stage, in keeping with feline infection model as well as *in vitro* studies showing that K562 cells fail to differentiate when infected with FeLV-C (Keel et al., 2008). Interestingly, Flvcr1 mice exhibit cranial and limb deformities which are not usually associated with lack of definitive erythropoiesis, indicating additional functions for Flvcr1 in development.

Recently, Chiabrando et al. (2012) identified a mitochondrial isoform of Flvcr1, termed Flvcr1b. While Flvcr1a encodes a plasma membrane-localized transporter, Flvcr1b contains an alternative transcription start site, resulting in a shortened amino terminus containing a mitochondrial targeting signal. Depletion of Flvcr1b in HeLa cells results in accumulation of mitochondrial heme, indicating that Flvc1b plays a role in heme export from the mitochondria (Chiabrando et al., 2012). Specific deletion of Flvcr1a in mice resulted in death between E14.5 and birth due to hemorrhage, edema, and skeletal abnormalities; however, erythropoiesis in these mice appeared to be normal and fetal liver cells from these mutant mice were able to repopulate irradiated wild type mice (Chiabrando et al., 2012). The authors conclude that Flvcr1a null mice suffer mainly from the loss of a plasma membrane heme exporter. Heme overload in endothelial cells leads to a loss of vascular integrity resulting in the hemorrhage, edema, and skeletal defects observed in these mice. The authors attribute the lethality of the original Flvcr1 knockout mice to lack of heme export from the mitochondria by Flvcr1b, as mice lacking Flvcr1a but not mitochondrial Flvcr1b appear to have normal erythropoiesis.

Numerous questions regarding heme exit from the mitochondria remain unanswered. It is not yet known if Flvcr1b resides on the mitochondrial IMM or OMM, with the possibility of another distinct mitochondrial heme exporter yet to be found. It is also unclear whether Flvcr1b is actually located within the mitochondria or is associated with membranes that tether the mitochondria to other organelles. If mitochondrial heme export is indeed attenuated, how do *Flvcr1*−/<sup>−</sup> embryos even survive until E14.5? Does Flvcr1b function in specific cell types? Yeast do not appear to have an obvious FLVCR homolog, yet are able to export heme from the mitochondria; what alternate mechanisms exist for mitochondrial heme export? No specific heme chaperones have been found to bind heme in the IMS, or to deliver heme to cytosolic hemoproteins. A number of cytosolic HBPs have been identified, including 22- and 23-kDa HBPs and the HBP homolog, SOUL (Iwahara et al., 1995; Taketani et al., 1998; Zylka and Reppert, 1999). These proteins could serve as "sponges" for heme exiting the mitochondria, simultaneously protecting the cell from heme's cytotoxicity and also delivering heme to target hemoproteins. It remains unclear whether multiple non-specific porphyrin-binding molecules perform the role of chaperoning heme in the cytosol, or if there are heme- and/or hemoprotein-specific pathways.

Presumably, heme imported into the cell, rather than solely heme exported from the mitochondria, could be bound by cytosolic heme chaperones and delivered to newly forming hemoproteins (**Figure 1**, Pathway 8). As such, plasma membrane heme importers could also be sources of heme for these enzymes. A discussion of putative heme importers can be found in the accompanying review of heme in pathophysiology by Chiabrando et al. (2014). Here we will focus on the heme importer, Hrg1.

The heme-responsive paralogs, *hrg-1* and *hrg-4*, were initially discovered as transmembrane domain-containing permeases upregulated in response to heme deficiency in *Caenorhabditis elegans* (Rajagopal et al., 2008). HRG-4 is localized to the apical plasma membrane of the worm intestine and serves to import dietary heme into the animal (Rajagopal et al., 2008; Yuan et al., 2012). HRG-1 localizes to lysosome-like vesicles in the intestine, and serves to mobilize heme stored in these compartments (Rajagopal et al., 2008; Yuan et al., 2012). While *hrg-4* does not appear to have homologs in other species, *hrg-1* homologs, also called solute carrier 48 A1 (SLC48A1) proteins, have been found in diverse organisms, including *Leishmania* spp., zebrafish, and mammals (Rajagopal et al., 2008; Huynh et al., 2012; Yuan et al., 2012; White et al., 2013). Knockdown of *hrg-1* in zebrafish leads to anemia, although the genetic etiology for the anemia phenotype is not known. Transport of heme by *hrg-1* and *hrg-4* has been demonstrated directly using

electrophysiological currents in *Xenopus* oocytes, and indirectly using heme-dependent growth of *C. elegans* and *Saccharomyces cerevisiae* or uptake of heme analogs as indicators of heme import.

Mammalian Hrg1 has been shown to be expressed widely, with the highest expression in the brain, heart, kidney, and muscle, with some expression in the placenta and intestine (Rajagopal et al., 2008). Hrg1 has been linked to a possible role in cancer progression, as its interaction with V-type ATPases is associated with changes in endocytic trafficking, extracellular acidification, altered glucose metabolism, and matrix metalloprotease activity (O'Callaghan et al., 2010; Fogarty et al., 2013). Depletion of Hrg1 in MCF7 cells resulted in decreased invasiveness and migration capabilities in these cells. Hrg1 has also been found to be a target of nuclear factor (erythroid-derived 2)-like 2 (Nrf2), the antioxidant response transcription factor (Campbell et al., 2013). In independent studies, human Hrg1 colocalized with Lamp1 in lysosomes, and Hrg1 is found on the erythrophagolysosome in macrophages during RBC heme-recycling (Delaby et al., 2012; White et al., 2013). Hrg1 has been shown to mediate hemoglobinderived heme export from this compartment in macrophages and this process will be discussed in Section "Heme Transport during Erythropoiesis."

### **HEME REQUIREMENT IN THE SECRETORY PATHWAY**

A variety of hemoproteins function in multiple components of the secretory pathway (**Table 1**). Lysosomal, peroxisomal, plasma membrane-targeted and secreted hemoproteins are folded and processed within the ER and Golgi, and thus the cell must have a means of transferring heme from the mitochondria into various subcellular membrane-bound compartments. These hemoproteins include the well known prostaglandin synthases (COX1 and COX2) in the ER, myeloperoxidase (MPO) in lysosome-like azurophil granules, eosinophil peroxidase in eosinophil granules, catalase in peroxisomes, the plasma membrane proteins thyroperoxidase and ferric reductase, as well as numerous secreted proteins including lactoperoxidase, fungal ligninase and chloroperoxidase, *C. elegans hrg-3*, *Drosophila* peroxidasin, and RHBP (*Rhodnius* heme-binding protein) in the blood-feeding insect *Rhodnius prolixus*.

Almost nothing is known regarding how these proteins obtain their heme cofactors. Given that the secretory and endocytic pathways are somewhat contiguous, with protein and lipid components being shuttled from location to another, it is feasible that heme need only be transported across a single membrane into the lumen of one such compartment, and from there can be mobilized to its intended destination by carrier proteins (**Figure 1**, Pathways 3– 7; De Matteis and Luini, 2008). For example, heme exiting the mitochondria may be transported into the ER, and from there be inserted into its target hemoprotein during protein folding, following which the protein can then traffic to its target organelle. It is also possible that one or many heme chaperones exist in the secretory pathway for delivery of heme to hemoproteins in their destination compartments. The fact that no mechanism for targeting heme to the secretory pathway has been discovered suggests that multiple mechanisms exist and can compensate for loss of a single such constituent.

In contrast to many hemoproteins in the secretory pathway, the maturation process of MPO has been well-characterized. MPO is a microbicidal protein generated by myeloid cells; its ability to chlorinate substrates enables it to generate hypochlorous acid from hydrogen peroxide [Reviewed in Klebanoff et al. (2013)]. This hemoprotein is extensively processed as it moves through the ER and Golgi; a single 80 kDa apoproMPO peptide is converted to a glycosolated, heme-containing protein consisting of a 59 kD heavy subunit and a 15.5 kD light subunit that is targeted for secretion in its pro-form or for storage in azurophil granules as a mature protein [Reviewed in Hansson et al. (2006)]. Early studies showed that disruption of the Golgi stacks with brefeldin A treatment resulted in MPO that is able to acquire heme but was improperly processed and remained in the pro-form (Nauseef et al., 1992). Moreover, treatment of the cells with succinyl acetone to inhibit heme synthesis also resulted inMPO that was improperly processed, while the processing and trafficking of other lysosomal proteins remained unchanged. The authors concluded that cleavage of MPO into its mature heavy and light subunits occurred in a post-ER compartment, that insertion of heme occurred in the ER, and that heme insertion was required for further processing. Further support for these conclusions was provided by mutants, such as the R569W MPO mutant which is unable to acquire heme and remains trapped in an apo-form in the ER (a result suggesting that heme insertion is necessary for ER exit), and the Y173C MPO mutant, which does acquire heme but is unable to exit the ER (suggesting that heme insertion is mediated within the ER; Nauseef et al., 1994; DeLeo et al., 1998). The aberrant processing of these mutant forms of MPO is especially informative, as these results lack the confounding effects of brefeldin A treatment.

The ER chaperones calreticulin and calnexin, which perform quality control during the folding of glycoproteins, unsurprisingly bind MPO as it is being folded and processed in the ER. Interestingly, these two similar proteins are able to distinguish between processed and unprocessed MPO, in a manner that likely enables sequential interaction of MPO with both chaperones. Calreticulin appears to bind apoproMPO preferentially, while calnexin binds both apoproMPO and the heme-containing apoMPO (Nauseef et al., 1998). It is possible that calnexin plays a role in the insertion of heme into MPO as it binds to both the apopro- and heme-containing pro- versions of the protein. Interestingly, one study showed that overexpression of calnexin in the fungus *Aspergillus niger* resulted in increased production of a secreted hemoprotein, manganese peroxidase, but only in the absence of heme supplementation in the media. This suggested that calnexin can function as a protein chaperone capable of enhancing holo manganese peroxidase assembly (Conesa et al., 2002). Studies have shown that palmitoylation of calnexin increased its localization to mitochondrial-associated membranes (MAMs) of the ER (Lynes et al., 2012) – a location where heme could, in principle, enter the ER (as discussed in Section "Mechanisms for Heme Delivery to the Secretory Pathway"). It has very recently been shown that palmitoylation switches calnexin from an ERp47-interacting quality control molecule to a MAM-associated regulator of ER-mitochondrial crosstalk and two proteomic analyses of MAMs have independently detected the presence of calnexin

(Lynes et al., 2013). While it is tempting to speculate that calnexin could moonlight as a chaperone for heme entering the ER from the mitochondria, the matter remains a mystery until further studies are performed.

#### **MECHANISMS FOR HEME DELIVERY TO THE SECRETORY PATHWAY**

Theoretically, there are a number of ways heme could enter the lumen of the secretory pathway. Heme released from the mitochondria could pass through the cytosol to be actively transported across the membrane of any subcellular organelle, (although no such transporter has yet been identified; **Figure 1**, Pathway 5). Heme imported into the cytosol could be moved into the secretory pathway in this manner as well. Heme entering the cell through an endocytic process (i.e., heme bound to hemopexin internalized via the low-density lipoprotein receptor-related protein LRP/CD91) could be trafficking to such compartments, with the possible participation of a chaperone to sequester and guide heme to its destination (**Figure 1**, Pathway 7). One appealing possibility is that nascent heme synthesized in the mitochondria could be shielded by membranes using vesicular trafficking or by entering the secretory pathway through direct contacts between the ER and mitochondrial membranes (**Figure 1**, Pathways 3 and 4).

Contacts between ER and mitochondrial outer membranes were observed over 40 years ago (Ruby et al., 1969; Lewis and Tata, 1973). Jean Vance coined the term MAM when she identified a functional relationship between these organelles to exchange phospholipids (Vance, 1990). With better subcellular fractionation techniques, and the advent of fluorescent imaging, especially the recent development of superresolution imaging, dynamic microdomains where mitochondrial membranes are tethered to ER membranes have been observed. These include the ER-mitochondrion encounter structure (ERMES) in yeast and the previously mentioned MAMs, found in plants and animals (Kornmann et al., 2009, 2011). These structures are involved in facilitating the transport of ionic calcium into the mitochondria, regulation of autophagy and apoptosis, and, more relevant to this review, the trafficking of lipids (Michel and Kornmann, 2012). The transport of phosphatidylserine from the ER to mitochondria has been demonstrated and it is hypothesized that other lipids are mobilized between these two organelles in a similar manner. Interestingly, two independent proteomic analyses of MAMs have detected the presence of coproporphyrinogen III oxidase, ferrochelatase, HBP 1, and heme oxygenase 2 (Poston et al., 2011; Zhang et al., 2011). It is intriguing to postulate that these ER-mitochondria connections could be an axis for heme transport (**Figure 1**, Pathway 3).

Another possible mechanism for the trafficking of heme from the mitochondria to other organelles is the use of mitochondrialderived vesicles (MDVs), which have been shown to traffic to both peroxisomes and lysosomes (**Figure 1**, Pathway 4; Neuspiel et al., 2008; Schumann and Subramani, 2008). Initially, 70–100 nm vesicles were shown to deliver specific mitochondrial cargo proteins to peroxisomes. Other vesicles were later observed carrying cargo to lysosomes in response to increase in cellular oxidative stress (Soubannier et al., 2012; McLelland et al., 2014). As the processes of heme synthesis, mitochondrial respiration, and responding to oxidative stress are innately coupled by their shared metabolic

pathways and mitochondrial location in the cell, it is possible to speculate that heme may be mobilized to these organelles via this mechanism.

#### **HEME TRAFFICKING AND IRON METABOLISM**

#### **HEME TRANSPORT DURING ERYTHROPOIESIS**

During definitive erythropoiesis, hematopoietic progenitors differentiate into mature RBCs. In the fetal liver or bone marrow, hematopoietic stem cells develop into early BFU-E progenitors and then into CFU-E progenitors, which are both defined by their capacity to form cell clusters *in vitro* [Reviewed in Palis (2014)]. These progenitor cells progress into erythroid precursors, which develop from early proerythroblasts (ProE), to basophilic erythroblasts (Baso), to polychromatophilic erythroblasts (PolyE), and finally become orthochromatic erythroblasts (OrthoE). Orthochromatic erythroblasts enucleate to form reticulocytes which move out into circulation and after about 24 h become mature RBCs. The erythroid precursors develop from the ProE stage to the OrthoE stage in a unique niche called

the erythroblastic island. This niche consists of a central nurse macrophage surrounded by a ring of developing RBC precursors. It is at this stage where heme and hemoglobin synthesis rapidly increase, as the cells draw near their fate as RBCs (**Figure 2A**).

Nurse macrophages maintain physical contacts with erythroblasts as they differentiate. They phagocytose the nuclei extruded by red cells, and are thought to enhance differentiation in some capacity. This specialized niche has been shown to be dispensable for erythropoiesis under steady state conditions (Chow et al., 2013; Ramos et al., 2013). However, in response to anemia, erythropoiesis expands from the bone marrow to the spleen and liver, with increased iron mobilization and RBC formation. Recent studies have shown that nurse macrophages play a regulatory role during stress erythropoiesis. Depletion of macrophages does not alter the numbers or profiles of erythroid cells in normal mice. However, mice with phlebotomy-induced anemia showed a decreased ability to recover from this state in the absence of macrophages, as shown by decreased red cell formation compared to control mice. Interestingly, the loss of these macrophages ameliorated symptoms

of both polycythemia vera and β-thalassemia, as both diseases involve a pathological increase in RBC production (Hashimoto et al., 2013; Ramos et al., 2013).

Some have suggested that nurse cells may contribute nutrients, possibly even iron or heme itself, to developing erythroblasts [Reviewed in Manwani and Bieker (2008)]. Under steady state conditions, the contribution of nurse macrophages is clearly dispensable for normal erythropoiesis; however, under stress erythropoiesis, the nutritional contribution of nurse macrophages may play a more important role. Developing erythroblasts co-cultured with macrophages in transwells showed a marked decrease in the beneficial effects that physical contact between the two cell types normally provides (Ramos et al., 2013). This result indicates that direct contact between the two cell types is required to maximize the positive effect of macrophages, but does not rule out a role for secreted factors.

It was reported that ferritin molecules were localized between the membranes of the central nurse macrophage and developing erythroblasts, and that this ferritin is taken up the erythroblasts by micropinocytosis (Policard and Bessis, 1962). However, the original source of this ferritin remains a mystery. It is possible that a portion of heme and/or iron within a maturing red cell is derived from inter-cellular transport from macrophages to support the initial stages of hemoglobinization, especially under stress conditions. In support of this hypothesis, heme importers Hrg1 and exporters Flvcr1 are both expressed in red cells and macrophages.

#### **HEME MOBILIZATION DURING ERYTHROPHAGOCYTOSIS**

Much like the birth of RBCs, the final moments of a red cell are spent in the care of a macrophage. Human RBCs have a lifespan of about 120 days, after which they become senescent and are recycled by macrophages of the reticuloendothelial system (RES) found in the spleen and liver [Reviewed in Knutson and Wessling-Resnick (2003)]. This process – phagocytosis of RBCs and the breakdown of billions of molecules of hemoglobin – is termed erythrophagocytosis (EP; **Figure 2B**). EP is essential, as the bulk of iron required for the synthesis of new hemoglobin (∼25 mg/day) derives from RBC recycling (Bratosin et al., 1998). By contrast, only about 1 mg per day of dietary iron contributes to erythropoiesis. Heme oxygenase 1 (Hmox1) plays a vital role in this process, freeing iron from its protoporphyrin ring, and enabling the release of this iron for reuse. *Hmox1*−/<sup>−</sup> mice suffer from anemia, reduced serum iron, and accumulation of iron in the spleen and liver, as the ability to recycle heme-iron in these mutants is almost completely crippled (Poss and Tonegawa, 1997). *In vitro* experiments showed that *Hmox1*−/<sup>−</sup> macrophages die when fed RBCs, and that splenic and liver macrophages are absent in *Hmox1*−/<sup>−</sup> mice (Kovtunovych et al., 2010).

Heme extracted from RBCs was postulated by some to be degraded by Hmox1 in the phagolysosome (Beaumont and Canonne-Hergaux, 2005). Subsequently, the newly released iron was either stored in ferritin (Ftn) or returned to the circulating pool of iron via the cellular heme exporter, ferroportin (Fpn). A major problem with this model is that Hmox1, while localizing to the ER via a C-terminal transmembrane segment, has been shown to be oriented with its active site facing the cytosol (Gottlieb et al., 2012). Furthermore, the pH optimum for heme catabolism by

Hmox1 is closer to physiological pH (pH 7.6) than lysosomal pH, and maximal Hmox1 activity requires the presence of biliverdin reductase (Liu and Ortiz de Montellano, 2000; Reed et al., 2010). Studies have shown that heme oxygenases are not present on the phagolysosomal membrane, even though this membrane is partially derived from the ER (Delaby et al., 2012). Thus Hmox1 likely does not have access to heme within the phagolysosomal compartment, suggesting that heme must exit this membrane compartment for degradation.

Hrg1, which has been shown to localize to endocytic compartments in mammalian cells, has recently been reported to transport heme across the phagolysosomal membrane (Delaby et al., 2012; White et al., 2013). Hrg1 is expressed in RES macrophages and is upregulated at both the mRNA and protein level in the presence of heme and during EP. This was true in both an *ex vivo* model using bone-marrow derived macrophages (BMDMs) treated with damaged RBCs, as well as in the livers of mice treated with either heme, damaged RBCs, or the hemolysis agent phenylhydrazine (White et al., 2013). During EP, Hrg1 accumulates on the membrane of the phagolysosome (White et al., 2013), and when Hrg1 is depleted by siRNA, BMDMs are incapable of upregulating the machinery normally required to deal with the influx of heme and iron (including Hmox1, Ftn, Fpn, and Hrg1 itself) indicating a lack of heme export from the phagolysosome.

Interestingly, a polymorphism of Hrg1 (P36L mutation) associated with anemia in four patients was shown to be defective in heme export in both a yeast model of heme transport, and during EP in BMDMs (White et al., 2013). These experiments used a heterogously expressed hemoprotein reporter, horseradish peroxidase (HRP), targeted to the Golgi to interrogate cellular heme availability and showed that heme derived from recycled RBCs could be incorporated into an apohemoprotein by assaying HRP activity in BMDMs during EP. While further experiments are required to definitively show that the increased HRP activity was due to incorporation of heme recycled *in toto,* rather than endogenously synthesized heme, these experiments suggest that not all heme liberated during EP is degraded via Hmox1, and that a portion can be reused by the cell.

## **DIETARY HEME TRANSPORT AND ENTEROCYTE METABOLISM**

Heme is the most bioavailable form of iron in the diet, as ferric iron can form insoluble ferric hydroxide or hydroxyl-iron dimers in the intestine and the uptake of soluble iron is inhibited by iron chelators found in certain foods [Reviewed in West and Oates (2008)]. Early studies using radiolabeled heme showed that heme enters enterocytes, where most of it is degraded and exported as iron to the circulation (**Figure 2C**; Walsh et al., 1955; Weintraub et al., 1965; Brown et al., 1968). Interestingly, these early studies showed that unlike uptake of inorganic iron, heme uptake is not significantly upregulated during iron deficiency (Turnbull et al., 1962). Later studies showed that this process was mediated by a receptor or transporter, as it was pH dependent, saturable, and that its activity was lost in the presence of trypsin (Grasbeck et al., 1979, 1982; Galbraith et al., 1985; Majuri, 1989; Noyer et al., 1998). Studies using zinc mesoporphyrin showed that uptake was inducible with heme depletion and specific to certain cell types, as it was

observed in intestinal Caco-2 and I-407 cells and hepatic HepG2 cells but not in mouse fibroblasts (Worthington et al., 2001).

A recent study analyzed the uptake of 58Fe-heme versus a nonheme iron source (57FeSO4) in pregnant women (Young et al., 2012). Cord blood from neonates taken at time of delivery contained significantly more 58Fe than 57Fe. This study suggests that iron absorbed in the form of heme contributes substantially to body iron stores presumably because dietary heme-iron is more bioavailable for intestinal absorption and therefore is preferentially delivered to the developing fetus. However, it is unclear whether the 58Fe detected is due to degradation of dietary heme within the intestine or, more provocatively, that a portion of ingested 58Fe-heme is delivered intact to the fetus. To differentiate between these possibilities, similar studies must be undertaken using heme labeled in the porphyrin ring.

The relative contributions of receptor-mediated endocytosis versus membrane transport to intestinal heme uptake remain unclear. Early studies of heme uptake in dogs and rats used diaminobenzidine (DAB) staining to visualize intact heme taken up into intestinal epithelial cells (Parmley et al., 1981; Wyllie and Kaufman, 1982). Heme was shown to accumulate on the surfaces of microvilli and in endosomal compartments in the apical cytoplasm, evidence for uptake via an endocytic process. However, these studies do not exclude the possibility that heme is also directly transported across the apical plasma membrane, as heme accumulating in the cytosol may not be sufficiently concentrated to visualize with DAB staining.

It was reported that heme carrier protein 1 (HCP1) was an intestinal heme importer (Shayeghi et al., 2005). Later, biochemical and genetic studies determined that HCP1 was a high affinity folate transporter, as it was capable of mediating folate transport and loss-of-function mutations in human patients were associated with hereditary folate malabsorption (Qiu et al., 2006). Qiu et al. (2006) note that folate supplements completely corrected any hematological defects associated with folate deficiency, indicating that this syndrome was not caused by defective heme uptake. An analysis of transport by HCP1, now termed HCP1/proton coupled folate transporter (PCFT), supported the conclusion that the physiological substrate of HCP1/PCFT is folate, though the protein may have weak heme transport activity. Thus, the mechanism (or mechanisms) of intestinal heme import remains unknown.

It may be useful to apply our knowledge of better-characterized paradigms of intestinal import to study intestinal heme uptake. For example, does heme traverse enterocytes via transcytosis, in a manner similar to vitamin B12, which is endocytosed in clathrincoated vesicles, trafficked to lysosomes, and then secreted from the basolateral intestinal membrane? Does heme, like iron, utilize a transcellular trafficking route, making use of apical importers and basolateral exporters for directed movement into the body? Additionally, early studies showed that the bulk of heme exported from the intestine into the portal circulation is in the form of free iron released from degraded heme. What if heme is not exported from the enterocytes via blood? It is worth noting that cholesterol, another bulky, hydrophobic molecule, is exported into the lymph after being absorbed in the intestine (Abumrad and Davidson, 2012).

#### **HEME TRAFFICKING DURING EMBRYOGENESIS**

It has long been known that metal homeostasis plays a vital role in embryonic and post-embryonic development (Kambe et al.,2008). For example, mouse pups from iron- or copper-depleted dams experience developmental and neurological abnormalities whose severity is dependent on the timing and extent of the maternal deficiency. A number of metal transporters are embryonic lethal when targeted for gene knockout. There are two separate but interconnected processes that should be taken into account when discussing this topic: (a) metal transport within embryonic cells as pluripotent cells grow and differentiate into tissues; and (b) maternal transfer of metals to developing embryos that lack an independent dietary source for these nutrients. We will consider the roles of both modes of transport with regards to heme homeostasis.

The respective phenotypes of the Flvcr1 and Flvcr1a knockout mice provide some insight into the role of heme transport during embryogenesis. As previously mentioned, Keel et al. (2008) report that homozygous Flvcr1 knockout pups (which lack both the mitochondrial and plasma membrane isoforms of Flvcr1) are found dead at either E7.5 or later at E14.5–E16.5. The authors infer from this observation that Flvcr1 is required for definitive erythropoiesis (which begins at ∼E12) and not embryonic erythropoiesis. In support of this, they mention that yolk sac-derived erythroblasts do not express Flvcr1 and appear normal in *Flvcr1*−/<sup>−</sup> mice. If the cause of mortality in *Flvcr1*−/<sup>−</sup> mice is defective heme export from the mitochondria, then we must wonder about the source of heme used for embryonic erythropoiesis. It may be that this heme is maternally derived via one or more heme importers in the early embryo. It is known that homozygous ferrochelatase knockout embryos can be detected at the E3.5 preimplantation stage, but are reabsorbed and undetectable at E9–E10 – a phenotype also observed for uroporphyrinogen decarboxylase knockout animals (Phillips, personal communication; Magness et al., 2002; Phillips et al., 2007). Presumably, the ferrochelatase null phenotype should overlap with the *Flvcr1*−/<sup>−</sup> phenotype, as both would cause heme deficiency in the cell. The greater severity of the ferrochelatase mutant hints that, in the *Flvcr*−/<sup>−</sup> mice, a small portion of heme can exit the mitochondria in an Flvcr1-independent manner to sustain life for a few extra days.

Interestingly, deletion of the Flvcr1a isoform also causes embryonic lethality (Chiabrando et al., 2012). This lethality, due to hemorrhages, edema, and skeletal malformations, is observed later during development, usually between E14.5 and birth. Chiabrando et al. (2012) speculate that many of these defects, also observed in mice with defects in endothelial integrity, are due to heme buildup within endothelial cells leading to widespread oxidative stress and hypoxic conditions in developing embryos. Interestingly, Flvcr1b is upregulated in these mice, presumably leading to increased cytoplasmic heme as evidenced by increased levels of Hmox1, though its activity is insufficient to rescue the embryos. Thus, plasma membrane heme export is an essential regulator of heme homeostasis during development.

While intercellular and intracellular heme transport are critical during embryogenesis, maternal transfer of heme to offspring may also play a critical role in embryonic development. This process has been demonstrated in *C. elegans*, where the small peptide HRG-3 serves as a chaperone for heme delivery to developing embryos and extraintestinal tissues (Chen et al., 2011). When the worm is heme deprived, expression of *hrg-3* in the intestine is upregulated >300-fold. HRG-3 is processed in the secretory pathway into a 45-amino acid chaperone which binds heme in a stoichiometry of 1:2 (heme:protein). Mature HRG-3 is secreted into the worm's circulation and taken up by extracellular tissues and developing oocytes. When *hrg-3* null worms are grown under heme limiting conditions, they show embryonic lethality and delayed growth, indicating a role for *hrg-3* in the distribution of heme from the intestine during early embryonic and larval development (Chen et al., 2011).

It was not surprising to discover a mechanism for heme transport along a maternal-to-embryonic axis in *C. elegans*, as this animal is a heme auxotroph and thus completely reliant on dietary heme. Surprisingly, however, a similar pathway was found in the blood sucking insect *R. prolixus*, a heme prototroph (Walter-Nuno et al., 2013). In a single meal, *R. prolixus* can ingest several times its own weight in blood. To deal with the oxidative stress associated with such a massive amount of heme, *R. prolixus* expresses a 15 kDa protein termed RHBP (Oliveira et al., 1995). RHBP is synthesized in the insect's fat body at all stages of development. It is secreted into the hemolymph, where it and any heme bound to it are taken up into developing oocytes via receptor-mediated endocytosis. Interestingly, silencing of RHBP results in a unique form of embryonic lethality. Eggs laid early during oviposition retain their characteristic red color and develop normally; however, eggs laid at the later time are not viable and are pale due to lack of heme (Walter-Nuno et al., 2013). These eggs showed evidence of fertilization, but no embryo formation. Additionally, analysis of both heme-dependent and heme-independent mitochondrial enzymes showed widespread mitochondrial dysfunction in these later eggs (Walter-Nuno et al., 2013).

*Rhodnius* eggs require maternal heme because they are incapable of producing sufficient endogenous heme to sustain development. The use of RHBP couples the process of ameliorating heme toxicity in the adult to embryonic development – turning a liability into an essential asset. Interestingly, the heme delivered to these eggs is not degraded to release iron (as no heme degradation products are detected in the eggs) and the heme is incorporated *in toto* into embryonic hemoproteins.

## **EVIDENCE FOR HEME TRANSPORT BETWEEN CELLS AND TISSUES**

While the requirement for intracellular heme trafficking, based on the understanding of heme chemistry and cell membrane dynamics, is fairly intuitive and supported by the well-established paradigm for copper, zinc and iron trafficking pathways, the concept of heme trafficking from one tissue to another is controversial. Why move heme from one tissue to another if each cell in an organism is capable of synthesizing its own heme? However, there are several lines to evidence to support the potential existence of these pathways:

(1) While the bulk of heme imported into the intestine is purported to be degraded by heme oxygenases and exported as iron, small amounts of heme have been show to be transported

intact along the basolateral membrane of enterocytes. Studies in polarized monolayers of human enterocyte-like Caco-2 cells have shown that a portion of heme transported intact across the basolateral membrane (Uc et al., 2004). Basolateral transport of heme from intestinal cells was also observed in guinea pigs (Conrad et al., 1966).


The severity of phenotypes associated with loss of heme synthesis indicates that intracellular heme trafficking pathways are not the primary modes to support the heme requirements of animals. The generation of (a) tissue-specific heme synthesis knockout animals; (b) genetic chimeras using *in vivo* tissue-specific reporters; and (c) new modalities of live imaging of heme using label-free microscopy at the tissue and subcellular level would be greatly beneficial in determining the capacity for inter- and intra-tissue heme transport. Ultimately, these pathways may be relevant under specific conditions of aberrant iron or RBC homeostasis, or during pathogenesis. We include in this review a call for further testing and analysis of these ideas.

## **CONCLUSION**

In this review, we have discussed the following topics:


We had three aims with this review. First, we intended to provide a comprehensive outline of the heme requirements in different locations within a cell and define what is known or remains unknown regarding how heme is transported to those places. Though much progress has been made in the past decade, much remains unknown. The field is lacking genetic tools, microscopic techniques capable of imaging heme in live cells, and subcellular sensors of heme levels. We hope this review will spur new ideas and creative thinking regarding how to tackle these questions.

Second, we intended to place what is known about heme transport in the context of physiological processes which are known to or likely require the mobilization of heme. Again, very little is known about heme movement during these events, but we attempted to outline possible mechanisms and players. We hope calling attention to these unknowns will impel members of the field to consider them.

Last, we aimed to question the current dogma that heme is not mobilized from one cell or tissue to other in heme prototrophs. Much as the homeostasis of biometals and lipids relies on intertissue trafficking pathways, it appears likely that heme utilizes similar transport routes *like iron in the blood of the people*, and it is incumbent upon members of the field to determine if this is indeed so.

### **ACKNOWLEDGMENTS**

This work was supported by funding from the NIH DK85035 and DK74797 (to Iqbal Hamza) and the Roche Foundation for Anemia Research (to Iqbal Hamza).

## **REFERENCES**


with minimal contribution from circulating monocytes. *Immunity* 38, 792–804. doi: 10.1016/j.immuni.2013.04.004


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

*Received: 19 March 2014; paper pending published: 18 April 2014; accepted: 12 May 2014; published online: 04 June 2014.*

*Citation: Korolnek T and Hamza I (2014) Like iron in the blood of the people: the requirement for heme trafficking in iron metabolism. Front. Pharmacol. 5:126. doi: 10.3389/fphar.2014.00126*

*This article was submitted to Drug Metabolism and Transport, a section of the journal Frontiers in Pharmacology.*

*Copyright © 2014 Korolnek and Hamza. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*

## Expression of ABCG2 (BCRP) in mouse models with enhanced erythropoiesis

#### *Gladys O. Latunde-Dada1 \*, Abas H. Laftah2, Patarabutr Masaratana3, Andrew T. McKie1 and Robert J. Simpson1*

*<sup>1</sup> Diabetes and Nutritional Sciences Division, School of Medicine, King's College London, London, UK*

*<sup>2</sup> Vascular Sciences Unit, Imperial Centre for Translational and Experimental Medicine, Imperial College, NHLI, London, UK*

*<sup>3</sup> Department of Biochemistry, Faculty of Medicine Siriraj Hospital, Mahidol University, Thailand*

#### *Edited by:*

*Raffaella Gozzelino, Instituto Gulbenkian de Ciência, Portugal*

#### *Reviewed by:*

*Andrei Adrian Tica, University of Medicine Craiova Romania, Romania Emanuela Tolosano, University of Torino, Italy*

#### *\*Correspondence:*

*Gladys O. Latunde-Dada, Diabetes and Nutritional Sciences Division, School of Medicine, King's College London, Franklin Wilkins Building, 150 Stamford Street, London SE1 9NH, UK e-mail: yemisi.latunde-dada@ kcl.ac.uk*

Haem is a structural component of numerous cellular proteins which contributes significantly to iron metabolic processes in mammals but its toxicity demands that cellular levels must be tightly regulated. Breast Cancer Resistance Protein (BCRP/ABCG2), an ATP Binding Cassette G-member protein has been shown to possess porphyrin/haem efflux function. The current study evaluated the expression and regulation of *Abcg2* mRNA and protein levels in mouse tissues involved in erythropoiesis. *Abcg2* mRNA expression was enhanced in bone marrow hemopoietic progenitor cells from mice that were treated with phenylhydrazine (PHZ). *Abcg2* mRNA expression was increased particularly in the extramedullary haematopoietic tissues from all the mice models with enhanced erythropoiesis. Haem oxygenase (*ho1*) levels tended to increase in the liver of mice with enhanced erythropoiesis and gene expression patterns differed from those observed in the spleen. Efflux of haem biosynthetic metabolites might be dependent on the relative abundance of *Abcg2* or *ho1* during erythropoiesis. *Abcg2* appears to act principally as a safety valve regulating porphyrin levels during the early stages of erythropoiesis and its role in systemic haem metabolism and erythrophagocytosis, in particular, awaits further clarification.

**Keywords:** *Abcg2***,** *bcrp***, hypoxia, phenylhydrazine, iron**

## **INTRODUCTION**

Iron turnover in mammals is largely explained by the production (biosynthesis) and breakdown (biodegradation) of erythrocytes. Consequently iron fluxes during erythropoiesis and erythrophagocytosis must be balanced and this is mainly achieved by influences of hepcidin in the maintenance of iron homeostasis. Three essential substrates, iron, porphyrin, and haem as well as their combination, are essential for erythropoiesis but are potentially toxic if allowed to accumulate. Iron, haem, and photosensitized porphyrin are pro-oxidants and they can cause damage to DNA, proteins and cell membrane systems. Availability of iron, a limiting nutrient, and the *de novo* synthesis of haem and porphyrin are therefore strictly coordinated and channeled to avert the accumulation of excess toxic metabolites. Recently the identification of two haem efflux proteins, Breast Cancer Resistant Protein (BCRP or ABCG2) and Feline Leukaemia Virus C Receptor (FLVCR) has defined important mechanisms that protect cells and tissues against noxious free haem (Quigley et al., 2004; Krishnamurthy and Schuetz, 2005a). Abcg2 is a half ATP-Binding Cassette (ABC) G-member transporter with a nucleotide binding domain (NBD) at its amino terminus and a transmembrane domain at its carboxyl end (Krishnamurthy and Schuetz, 2006). It functions conventionally as a plasma membrane protein extruding endogenous and exogenous toxic xenobiotics from the intestine, liver, placenta, and the blood/brain barrier. This transporter, in some instances, confers resistance to anticancer drugs. The broad substrate spectrum of abcg2 are porphyrin metabolites (Suzinges-Mandon et al., 2010). Specifically, *Abcg*2 was shown as a porphyrin efflux protein which functions to rid erythroid cells of excess porphyrins particularly under hypoxic conditions (Krishnamurthy et al., 2004) and thereby protects haematopoietic cells from the potential toxicity of excess free porphyrin compounds. As hypoxia could vary in magnitude depending on the duration and the type of anaemia, the detoxification function of *abcg2* might be pertinent in the various other tissues that handle large quantities of haem and its degradation products e.g., liver.

At the cellular level, haem oxygenase 1 (ho1) catalyses the rate limiting step in the degradation of haem into iron, carbon monoxide, and biliverdin, which is reduced by biliverdin reductase to bilirubin. These by-products also constitute potential substrates for abcg2, and have recently been shown to have antioxidant, anti-inflammatory, and cytoprotective functions (Bilban et al., 2008; Soares and Bach, 2009). Excess haem is presumed to be channeled into the lysosome for catabolism or might be effluxed into circulation where it binds to hemopexin (Tolosano et al., 2010). Haem and its metabolites may be substantially and strategically trafficked or otherwise degraded in a coordinated and regulated manner under different modulators of iron metabolism. Moreover, abcg2 might play a vital role in the transport of haem and porphyrin compounds during erythropoiesis. This study investigates the expression and regulation of *abcg2* and *ho1* in tissues of mice with enhanced erythropoiesis.

## **MATERIALS AND METHODS**

## **ANIMALS AND TISSUE COLLECTION**

CD1 mice were used for all experiments except the hypotransferrinaemic (Hpx) mice that are of Balb/c background. Mice were placed in a hypobaric chamber at 0.5 atm for 72 h to effect hypoxia. Controls were kept at room air pressure. Dietary iron deficiency was induced by feeding 4-week old CD1 male mice with a low-iron diet (Formula TD. 80396, elemental iron concentration: 3–6 mg/kg or control diet containing 48 mg/kg iron, TD. 80394; Harlan-Teklad; Madison, WI, USA) for 3 weeks. Enhanced erythrocyte turnover was induced by intraperitoneal injections of neutralized phenylhydrazine (PHZ) at 60 mg/kg body weight; (Sigma-Aldrich, UK) on 2 consecutive days. All mice were maintained on standard commercial diet (Rodent Maintenance diet; RM1, Special Diet Services, UK) feed and water *ad libitum*. Mice were sacrificed after anaesthesia and neck dislocation, after which tissues were collected. Femurs were excised and flushed with Dulbecco's minimum essential medium (DMEM; Sigma-Aldrich, UK), to collect the hemopoeitic progenitor cells from the bone marrow of control and PHZ-treated CD1 mice. All procedures were approved and conducted in accordance with the UK. Animals (Scientific Procedures Act, 1986).

## **CELL CULTURE** *RT-PCR*

Total RNA was extracted from tissue samples using Trizol reagent (Invitrogen, UK) according to manufacturer's instructions. Quantitative RT-PCR was carried out using an ABI Prism 7000 detection system in a two-step protocol with SYBR Green (ABI, Life Technologies, UK). The efficacy of the amplification was confirmed by a melting curve analysis and gel electrophoresis to confirm the presence of a single product.

Quantitative measurement of each gene was derived from a standard curve constructed from known amounts of PCR product. The results were calculated by the -C*<sup>t</sup>* method that expresses the difference in threshold for the target gene relative to that of 18S RNA. Sequences of primers used, forward, and reverse, respectively, are as follow:

Mouse abcg2 5- -TCGCAGAAGGAGATGTGTTGAG-3- 5- -CCAGAATAGCATTAAGGCCAGG-3- Mouse tfr1, 5- -CATGGTGACCATAGTGCACTCA-3- 5- -AGCATGGACCAGTTTACCAGAA-3- Mouse ho1 5- -CAAGGAGGTACACATCCAAGCC-3- 5- -TACAAGGAAGCCATCACCAGCT-3- Mouse 18s 5- -GAATTCCCAGTAAGTGCGGG-3- 5- -GGGCAGGGACTTAATCAACG-3- Human 18s 5- -AACTTTCGATGGTAGTCGCCG-3- 5- -CCTTGGATGTGGTAGCCGTTT-3-

## *Western blot analysis*

Spleen tissue was homogenized (in a buffer containing 50 mM mannitol, 2 mM Hepes, 0·5 mM PMSF and pH 7·2) with an Ultra Turrax (IKA, Staufen, Germany) homogenizer in (3 × 30 s pulses at full speed). The homogenate was centrifuged at 1500 g for 5 min and the supernatant was centrifuged for 1 h at 15,000 g to obtain the crude membrane fraction. Protein concentration was determined using Bio-Rad reagents (Bio-Rad, Laboratories, Hercules, CA, USA). Fifty (50) μg of membrane extracts were loaded onto a 12% gel in a SDS-PAGE. The proteins separated were then transferred to Hybond ECL-nitrocellulose membrane (Amersham Biosciences, Bucks, UK) using a Bio-Rad semidry transfer apparatus (Trans-Blot<sup>R</sup> SD Semi-Dry Transfer Cell; Bio-Rad, UK). Membranes were blocked with 5% milk for 1 h and probed with ABCG2 BXP-53 monoclonal (Santa Cruz Biotechnology, USA), and, β-actin (Sigma, UK) antibodies diluted in 0.01% milk in TBS. Cross-reactivity was observed with peroxidase-linked anti-IgG by using SuperSignal West Pico (Thermo Scientific, USA).

## *Statistical analysis*

All values are expressed as mean ± s.e.m. Statistical differences between means were calculated with Microsoft Excel 6.0 (Microsoft, Seattle, WA, USA) by using the Student *t*-test correcting for differences in sample variance. When multiple comparisons were necessary, One-Way analysis of variance (ANOVA) was performed using SPSS 14 (SPSS Inc, Chicago, US) with Tukey's *post-hoc* test.

## **RESULTS**

*Abcg2* and *tfr1* mRNA expression were enhanced in bone marrow cells from mice treated with PHZ (**Figure 1**). Increased erythropoietic activity, which typifies Hpx, PHZ-treated, hypoxic, and iron-deficient mice, was associated with significant increases in *abcg2* mRNA expression in the spleen tissues from the mice (**Figure 2**). While changes in *ho1* mRNA expression in spleen tissues varied widely with the mice models (**Figure 3A**), hepatic *ho1* mRNA was induced significantly in Hpx and PHZ-treated mice (**Figure 3B**) instead. Enhancement of abcg2 protein expression in the spleen from mice with enhanced erythropoiesis correlated with the pattern of increased mRNA expression (**Figures 4A,B**).

**FIGURE 1 |** *Abcg2* **and** *TfR1* **mRNA levels in bone marrow precursor cells of CD1 mice.** Erythropoiesis was induced with phenylhydrazine (PHZ) administration (60 mg/kg) over 2 days. Relative mRNA expression was obtained by normalizing mRNA levels to 18S mRNA. Analysis by Students' *T* -test showed significant effects of *abcg2* (*P* < 0.01), and *TfR1* (*P* < 0.005) over the control samples.

There was a slight, but observable, increase in ho1 protein levels in spleen tissues from PHZ-treated, hypoxic, and iron-deficient mice (**Figures 4A,C**).

*Abcg2* mRNA expression showed decreases in mouse model with enhanced erythropoiesis in kidney tissues (**Figure 5A**) but the response were not significantly different. *Ho*1 mRNA expression was significantly enhanced in kidney tissues from Hpx and iron-deficient mouse models (**Figure 5B**). Only modest increases in *ho1* mRNA expression levels were observed in kidney tissues from PHZ-treated and hypoxic mouse models.

## **DISCUSSION**

Abcg2, an efflux protein with broad substrate specificity, is characteristically localized and expressed in epithelial membranes of barrier tissues such as the intestine, placenta, liver and the brain. The regulation of cellular haem and porphyrin levels by *abcg*2 plays a pivotal physiological function in erythropoiesis and iron metabolism in mammals. *Abcg*2 mRNA expression coincides with enhanced stimulation of haemoglobin biosynthesis (Zhou et al., 2005), during differentiation between BFU-E and CFU-E. Globin transcription and iron uptake are enhanced and haem concentration is critical for the initiation of haemoglobin synthesis (Krishnamurthy and Schuetz, 2005a). This fact might be the reason for the increase in *abcg2* mRNA expression in the hemopoeitic precursor cells in mice that were administered with PHZ (**Figure 1**). As haemoglobinization increases with the progression of erythropoiesis, proerythrocytes presumably express low levels of *ho1* and are highly sensitive to low levels of cytosolic haem. *Abcg2* prevents porphyrin toxicity especially under hypoxic conditions, during the mid to the distal stages of erythropoiesis. Flvcr1 also effluxes excess haem during erythropoiesis and erythrophagocytosis (Keel et al., 2008; Chiabrando et al., 2012; Byon et al., 2013). Typical of promoters devoid of the TATA box, *abcg2* promoter has multiple transcription sites (Lin et al., 2001). Zong et al. (2006) showed multiple transcription start sites from several

**and liver of mouse models with enhanced erythropoiesis. (A)** *Ho1* mRNA levels in the spleen of Hpx, PHZ-treated, hypoxic, and iron deficient mice. Analysis by Students' *T* -test showed significant effects of *Ho1* only in iron-deficient spleen tissues, (*P* < 0.01) over the control samples. **(B)** *Ho1* mRNA expression in the liver of Hpx, PHZ-treated, hypoxic, and iron deficient mice. Analysis by Students' *T* -test showed significant effects of *Ho-1* mRNA in Hpx (*P* < 0.05), and PHZ liver tissues, (*P* < 0.01) over the control samples. Data are mean ± s.e.m. of four observations in each group and are expressed as a ratio of 18S mRNA in arbitrary units.

leader exons in *abcg2* gene. Moreover, these promoter regions are enriched with several transcription factors binding sites (Ishikawa et al., 2013). Furthermore, *abcg2* promoter is regulated by tissue specific transcription factors such as GATA 2 (Minegishi et al., 1998). Consequently, *abcg2* gene regulation is varied and multifarious in tissues and during diverse physiological processes.

Similarly, *abcg2* mRNA expression (**Figure 2**) and protein levels (**Figures 4A,B**) were consistently up-regulated in the spleen of mice with enhanced erythropoiesis. *Abcg*2 and, more recently *flvcr1* (Vinchi et al., 2014), have been reported to regulate haem levels in cells during erythropoiesis. Evidence from literature suggests a reciprocal pattern in *flvcr* vs. *abcg2* expression as differentiation progresses (Keller et al., 2006). *Flvcr* null mice are non-viable and this confirms a lack of functional overlap with *abcg2*. *Flvcr* is presumed to efflux haem into circulation from macrophages (Keel et al., 2008) and possibly functions specifically during erythrophagocytosis. The latter report implied that a proportion of haem from phagocytosed, senescent red blood cells in macrophages is not degraded but effluxed into circulation by *flvcr*. This haem "cargo" could subsequently be bound

by hemopexin for delivery to the liver. It has recently been shown that hemopexin is obligatory to the functional haem efflux capability of *flvcr* (Yang et al., 2010). *Flvcr*, in contrast to some other members of the major facilitator superfamily, is a unidirectional, vectorial transporter.

Similarly, splenic expression of *abcg2* might also correlate with its role in detoxification processes. Haem, apart from being endogenously synthesized ubiquitously by erythroid and non-erythroid tissues, is also derived from haemolysis and haemorrhage (Reeder et al., 2002). While PHZ-treated mice are characterized by increased haemolysis and destruction of red blood cells, *ho1* mRNA expression was significantly enhanced in the liver rather than in the spleen of these mice (**Figures 3A,B**). The reasons for variation in the basal levels of *ho1* in the different mice models used in this study remain, however, both puzzling and unclear. The regulation of haem thus favors the depletion of unbound forms (Vinchi et al., 2008) to ameliorate toxicity in cellular and subcellular loci of visceral and peripheral tissues. It is well documented that *ho1* expression is highly inducible in the presence of haem (Schwarzer et al., 1999; Desbuards et al., 2009). Consequently, intracellular and extracellular trafficking of haem is regulated to molecular channeling in organelles and ensures detoxification as necessary. *Ho1* mRNA was generally enhanced in the kidney tissues from mice in the current study (**Figure 5**) although kidney *abcg2* mRNA levels were not significantly different. The kidney is important in iron metabolism for the biogenesis and biodegradation of haem. The former is by a regulatory feedback loop of erythropoietin production and the latter by the glomerullar filtration of haem by megalin and cubilin receptors (Gburek et al., 2002, 2003). Consequently, Abcg2 might not be involved in the efflux of haem/porphyrin in the kidney as it does in erythroid progenitor cells. Perhaps megalin and cubilin function in concert and are complimentarily with *ho1* to exude the resultant porphyrin breakdown products of haem: this remains an unanswered question.

In conclusion, abcg2, effluxes excess porphyrin metabolites and possibly excess haem during erythropoiesis particularly under hypoxic conditions. Egress and detoxification of excess porphyrin/haem by abcg2 and flvcr to maintain cellular homeostasis is presumably complemented by ho1 expression during the

early stages of erythropoiesis. Consequently, abgc2 mRNA levels are stimulated during enhanced erythropoiesis in extramedullary haematopoietic splenic and liver tissues of Hpx, PHZ-treated, hypoxic, and iron-deficient mice. However, abcg2 expression is low in the kidney, and in tissues of this organ ho1 degradation of haem might predominate. Tissue-specific expression of haem influx/efflux proteins and haem catabolism by ho1 must be critically balanced to maintain optimal levels of haem for metabolic processes.

## **ACKNOWLEDGMENT**

The study was supported by a Value in People (VIP) grant award from the Wellcome Trust.

## **REFERENCES**

Bilban, M., Haschemi, A., Wegiel, B., Chin, B. Y., Wagner, O., and Otterbein, L. E. (2008). Heme oxygenase and carbon monoxide initiate homeostatic signaling. *J. Mol. Med*. 86, 267–279. doi: 10.1007/s00109-007-0276-0


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

*Received: 14 January 2014; accepted: 19 May 2014; published online: 04 June 2014. Citation: Latunde-Dada GO, Laftah AH, Masaratana P, McKie AT and Simpson RJ (2014) Expression of ABCG2 (BCRP) in mouse models with enhanced erythropoiesis. Front. Pharmacol. 5:135. doi: 10.3389/fphar.2014.00135*

*This article was submitted to Drug Metabolism and Transport, a section of the journal Frontiers in Pharmacology.*

*Copyright © 2014 Latunde-Dada, Laftah, Masaratana, McKie and Simpson. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*

**MINI REVIEW ARTICLE** published: 11 March 2014 doi: 10.3389/fphar.2014.00042

## Molecular basis of HFE-hemochromatosis

## *Maja Vuji´c\**

Institute of General Zoology and Endocrinology, University of Ulm, Ulm, Germany

#### *Edited by:*

Raffaella Gozzelino, Instituto Gulbenkian de Ciência, Portugal

#### *Reviewed by:*

Stanislav Yanev, Bulgarian Academy of Sciences, Bulgaria Elena Corradini, University of Modena and Reggio Emilia, Italy

#### *\*Correspondence:*

Maja Vuji´c, Institute of General Zoology and Endocrinology, University of Ulm, Helmholtzstrasse 8/1, 89081 Ulm, Germany e-mail: maja.vujic@uni-ulm.de

Iron-overload disorders owing to genetic misregulation of iron acquisition are referred to as hereditary hemochromatosis (HH). The most prevalent genetic iron overload disorder in Caucasians is caused by mutations in the HFE gene, an atypical MHC class I molecule. Recent studies classified HFE/Hfe-HH as a liver disease with the primarily failure in the production of the liver iron hormone hepcidin in hepatocytes. Inadequate hepcidin expression signals for excessive iron absorption from the diet and iron deposition in tissues causing multiple organ damage and failure. This review focuses on the molecular actions of the HFE/Hfe and hepcidin in maintaining systemic iron homeostasis and approaches undertaken so far to combat iron overload in HFE/Hfe-HH. In the light of the recent investigations, novel roles of extra-hepatocytic Hfe are discussed raising a question to the relevance of the multipurpose functions of Hfe for the understanding of HH-associated pathologies.

**Keywords: Hfe, HH, hepcidin, extra-hepatic Hfe, Bmp/Smad signaling**

## **AN ADEQUATE SUPPLY OF IRON IS A PREREQUISITE FOR GOOD HEALTH**

Iron overload disorders comprise a wide range of inherited and acquired disorders of iron metabolism. Hereditary hemochromatosis (HH) encompasses a heterogenous group of inherited iron overload disorders with distinct underlying molecular defects and varying clinical symptoms (Camaschella and Poggiali, 2011). HH begins as mere iron overload which over time can cause serious organ dysfunctions leading to liver failure and cirrhosis, hepatocellular carcinoma, atherosclerosis, arthritis, fatigue, various endocrinopathies including diabetes, heart problems (both arrhythmia and cardiomyopathy, or loss of cardiac muscle function), hypermelanotic pigmentation of the skin, or compromised immune defense (Davies and Enns, 2004; Camaschella, 2005; Guggenbuhl et al., 2005; Pietrangelo, 2010). If remained untreated, HH is a life-threatening disorder. The current mainstay therapy for HH is phlebotomy (venesections), a relatively simple and inexpensive treatment, whereby blood is removed on a weekly basis for several months or more depending on the iron levels. For patients undergoing phlebotomy, liver, and cardiac functions improve after iron-depletion, whereas other HH-associated pathologies (e.g., diabetes mellitus or arthropathy) are often unchanged by the treatment (Brissot and de Bels, 2006; Sahinbegovic et al., 2010). In certain cases, iron chelation therapy may be taken into consideration (Brissot and de Bels, 2006). All current therapies for HH focus exclusively at manipulating the excess of iron.

## **LOW LEVELS OF HORMONE HEPCIDIN HALLMARK HH**

Genetic data in mice and patients demonstrated that a relative deficiency of hormone hepcidin underlies HH disorders (Ahmad et al., 2002; Bridle et al., 2003; Muckenthaler et al., 2003; Niederkofler et al., 2005; Wallace et al., 2005; Babitt et al., 2006; Lesbordes-Brion et al., 2006;Wallace et al., 2009). Hepcidin is a liver-derived peptide that oversees systemic iron changes by maintaining plasma iron

levels within a narrow physiological range. It does so by determining iron absorption and iron release from the cells through binding to its receptor, the iron-export protein ferroportin (Nemeth et al., 2004; Nemeth and Ganz, 2006). Low levels of hepcidin signal for increased ferroportin activity thereby enhancing iron uptake from the diet via enterocytes, iron release from macrophages into the circulation and deposition of the excess of iron in parenchymal cells of tissues leading to a condition of systemic iron overload. The hepcidin-mediated ferroportin regulation can therefore be considered as the critical step for balancing systemic iron levels. Any disturbances in hepcidin regulators such as mutations in HH-genes like *HFE*, *Transferrin receptor 2 (TfR2*; Roetto et al., 2002; Kawabata et al., 2005; Nemeth et al., 2005), *Hemojuvelin (HJV*; Lanzara et al., 2004; Papanikolaou et al., 2004; Huang et al., 2005; Niederkofler et al., 2005; Babitt et al., 2006) or *Hepcidin* itself (Roetto et al., 2003; Lesbordes-Brion et al., 2006), and/or ferroportin (Pietrangelo, 2004; Sham et al., 2009; Le Lan et al., 2011) contribute to iron-related pathophysiology (Ganz and Nemeth, 2006).

## **HFE-HH IS THE MOST COMMON TYPE OF HH IN WESTERN POPULATION**

The most common type of hereditary HH is associated with the mutations in the *HFE* (High Fe) gene (Feder et al., 1996; Levy et al., 1999; Bridle et al., 2003; Muckenthaler et al., 2003). This is the most prevalent mutation in Western populations affecting approx. 1:250 individuals (Pietrangelo, 2010). The *HFE* gene was first identified in 1996 as a major histocompatibility (MHC) class I-like gene in which homozygosity for a missense mutation that results in cystein-to-tyrosine substitution at amino acid 282 of human HFE protein (C282Y), was found in vast majority of HH patients (Feder et al., 1996). Approximately 80% of HH patients are homozygous for C282Y and the frequency of this mutation decreases from the northwest to southwest Europe paralleling the settlements of ancient Celts (Distante et al., 2004).

It is likely that mutation provided survival advantage against from what was then a very poor iron diet. Significantly fewer patients with a clinical diagnosis of HH are heterozygotes for a compound C282Y and H63D mutation (the latter is a substitution of aspartic acid for histidine at the position 63 of the HFE protein), whereas the homozygosity for H63D mutation usually causes little increase in iron absorption and rarely leads to the development of HH. Other mutations (e.g., S65C, V53M, V59M, H63H, Q127H, Q283P, P168X, E168Q, E168X, W168X) are rare and/or have low penetrance (Qaseem et al., 2005; Brissot et al., 2011).

The discovery of the *HFE* gene and the identification of the mutations associated with the HFE-HH were spectacular. The wealth of scientific and clinical data led towards more accurate diagnosis of HFE-HH, improved family screening and provided the first insights into the regulation of iron homeostasis by the *HFE*.

## **HFE-HH IS THE LIVER DISEASE**

The protein encoded by the *HFE* gene is a non-classical MHC class I-like protein which contains a signal sequence, peptidebinding extracellular domains, a transmembrane region, and a small cytoplasmic portion. Within the alpha-2 and alpha-3 extracellular domains are the four cysteine residues that form disulfide bridges representing one of the most conserved structural features of MHC class I molecules. HFE interacts with beta2-microglobulin and this association enables efficient transport of HFE to the cell surface where it interacts with TfR1. C282Y mutation disrupts the disulfide bridges in the extracellular domains of the HFE protein thereby preventing the association of HFE with beta2-microglobulin and TfR1. The lack of HFE interaction with TfR1 increases the affinity of TfR1 for the transferrin-bound iron thereby modulating iron absorption. In contrast to the C282Y mutation, mutant H63D HFE formed stable complexes with TfR1 being in line with the clinical data that H63D HFE mutations marginally affect iron absorption and rarely lead to HH.

The molecular link between the HFE protein and the TfR1 raised the possibility that this regulatory mechanism of iron transport may play a role in the pathogenesis of HH (Fleming et al., 1999; Anderson et al., 2002). The answer to whether Hfe alters cellular iron uptake by serving as a sensor mechanism in duodenal enterocytes was provided through the generation of mice bearing a selective deficiency of *Hfe* in the duodenal enterocytes (Vujic Spasic et al., 2007). Surprisingly, mice lacking *Hfe* in intestinal cells showed no iron accumulation in the liver nor hepatic hepcidin deficiency overruling the traditional hypothesis that duodenal Hfe played a gatekeeper role in controlling systemic iron homeostasis (Vujic Spasic et al., 2007). The question where *Hfe* acts to control iron homeostasis was revealed through the generation of mice with selective deficiency of *Hfe* in hepatocytes which recapitulated most of the anomalies within iron metabolism observed in constitutive *Hfe* mutant mice and HFE-HH patients (Adams, 2003; Wigg et al., 2003; Vujic Spasic et al., 2008; Adams and Barton, 2010). Mice deficient for the hepatocytic-*Hfe* show decreased hepcidin expression which led to uncontrolled iron uptake and iron accumulation in the

liver (Vujic Spasic et al., 2008). Vice versa, hepatocellular transgenic over-expression of *Hfe* in mice lacking endogenous *Hfe* resulted in significant upregulation of hepcidin and normalization of transferrin saturation and liver iron levels (Schmidt et al., 2008). Collectively, these data establish hepatocytic-*Hfe* as the regulator of hepatic hepcidin expression and imply that regulatory cues involved in maintaining iron homeostasis are centered in the liver.

## **HFE IS INVOLVED IN IRON-MEDIATED CONTROL OF HEPCIDIN EXPRESSION VIA THE BMP/SMAD SIGNALING PATHWAY**

The above investigations argue for the role of Hfe in hepatocytes to regulate hepcidin expression and thus iron homeostasis. In the last couple of years first insights into to molecular mechanism coupling Hfe to hepcidin expression have begun to emerge.

It is proposed that Hfe is sequestered by TfR1 protein in an iron-sensing complex located in the hepatocyte cell membrane (**Figure 1**) (Goswami and Andrews, 2006). The close interaction between Hfe and TfR1 is disrupted upon binding of circulating transferrin bound iron which binds with higher affinity to TfR1 (**Figure 1**) (Goswami and Andrews, 2006). *In vitro* studies proposed a model in which Hfe upon dissociation from Hfe/TfR1 complex, interacts with another HH protein, TfR2, and that this partnership is enlarged by the contribution of a membrane-bound bone morphogenetic protein (Bmp) co-receptor Hjv (**Figure 1**) (Goswami and Andrews, 2006; D'Alessio et al., 2012). However a direct interaction between Hfe and TfR2 could not be confirmed *in vivo* suggesting that TfR2 may regulate hepcidin expression in an Hfe-independent manner (Schmidt and Fleming, 2012). Importantly, Hfe-mediated hepcidin expression is abolished by the loss of endogenous Hjv protein implicating for co-dependency between Hfe and the Hjv (Schmidt et al., 2010). The Bmp co-receptor, Hjv, is involved in transmitting the signals initiated upon binding of Bmp ligands, the members of the multifunctional Tgf-β family proteins, to two cognate serine/threonine kinase receptors, type I and II located at the cell surface (**Figure 1**) (Miyazono et al., 2005; Babitt et al., 2006; Babitt et al., 2007). This interaction induces phosphorylation of the intracellular receptor-activated Smad proteins 1/5/8 and subsequent binding of common Smad4 protein to form an active transcriptional complex which directly regulates the expression of numerous target genes including hepcidin (**Figure 1**) (Miyazono et al., 2005; Babitt et al., 2006; Babitt et al., 2007; Truksa et al., 2009). The lack of *Hjv*, *Hfe*, *TfR2*, or combined *Hfe* and *TfR2* deficiency significantly impairs the activity of the Bmp/Smad-signaling pathway resulting in a low hepcidin expression and profound systemic iron overload (Babitt et al., 2006; Corradini et al., 2009, 2011; Kautz et al., 2009; Ryan et al., 2010; Vujic Spasic et al., 2013). Furthermore, mice deficient for hepatic Bmp type I receptors, *Alk2* or *Alk3* (Steinbicker et al., 2011), or *Smad4* expression (Wang et al., 2005), or for ubiquitous *Bmp6* expression (Andriopoulos et al., 2009; Meynard et al., 2009) present with relatively low hepcidin levels in regard to overall body iron overload. The Bmp/Smad signaling pathway therefore emerges as the central signaling event for regulating hepcidin transcription in hepatocytes. Recently, a contribution of the Erk/Mapk signaling cascade to hepcidin regulation has been proposed (Wallace et al., 2009) as its decreased signaling activity

in the liver was associated with the combined deficiency of *Hfe* and *TfR2* suggesting that additional, or parallel signaling pathway to Bmp/Smad may control hepatic hepcidin transcription (**Figure 1**).

## **THERAPEUTIC IMPLICATION OF ATTENUATED BMP/SMAD-MEDIATED HEPCIDIN REGULATION IN HFE-HH**

Given the fact that hepcidin levels are low in Hfe-HH and that Hfe is involved in optimizing iron-response via Bmp/Smad signaling activity in the liver, it would be expected that by modulating hepcidin levels and/or the activity of the Bmp/Smad pathway, the iron homeostatic parameters in HFE-HH patients may be normalized. In contrast to phlebotomy therapies that manipulate the excess of iron, few pilot approaches were conducted with the aim to target the defective molecular mechanisms underlying the HFE-HH.

Short- and long-term hepcidin injections, or hepatic overexpression of hepcidin transgene in *Hfe*-deficient mice resulted in successful reconstitution of hepcidin expression to the levels present in wild type mice. Furthermore, high plasma iron levels present in *Hfe*−/<sup>−</sup> mice were significantly reduced by hepcidin treatments, without affecting hepatic iron load which remained inappropriately high in regard to hepcidin levels (Nicolas et al., 2003; Laftah et al., 2004; Rivera et al., 2005; Viatte et al., 2006; Moran-Jimenez et al., 2010). Neither has exogenous Bmp6 administration to *Hfe*−/<sup>−</sup> mice succeeded to reduce hepatic iron burden, although hepcidin expression was restored to the levels present in wild type mice, followed by a significant drop in serum iron levels and redistribution of iron in the spleen and duodenum in *Hfe*−/<sup>−</sup> mice (Corradini et al., 2010). Due to severe side-effect (peritoneal calcification) that accompanied prolonged exogenous Bmp6 treatment in mice (Corradini et al., 2010), the Bmp6 substitution cannot be currently considered as an optional therapy for HH.

## **HFE-HEPCIDIN-DEPENDENT AND -INDEPENDENT CONTROL OF IRON HOMEOSTASIS**

The above data raise a critical question to whether Hfe-HH is an exclusive consequence of defective Bmp/Smad signaling and low hepcidin expression, or whether HH-associated pathologies may be intensified or refractory to iron-depletion strategies or hepcidin substitution if Hfe exerts distinct,*extra-hepatocytic functions* (**Figure 2**). Noteworthy, both Hfe and hepcidin are expressed in several extra-hepatic tissues (e.g., heart, skeletal muscle, kidney, lungs, thymus, duodenum) but so far, little is known whether Hfe-hepcidin regulation is operating in these cells and how it may impact on cellular and/or systemic iron homeostasis (**Figure 2**). For example, general *Hfe* deficiency was associated with better tolerance of *Hfe*−/<sup>−</sup> mice to severe blood loss in regard to wild type mice or animals kept on an iron-rich diet (Ramos et al., 2011). During erythropoietic stress conditions hepcidin expression is severely hampered which in turn signals for enhanced iron absorption, iron mobilization from the stores and iron utilization by the erythron. It was proposed that in addition to general *Hfe* and hepcidin deficiency, selective *Hfe* actions in erythroid cell may contribute to overall better tolerance of *Hfe*−/<sup>−</sup> mice to severe blood loss by enhancing transferrin-bound iron uptake and thus modulating erythroid iron homeostasis (Ramos et al., 2011) (**Figure 2**). However a direct role of *Hfe* in erythroid cells and its contribution to cellular and overall iron homeostasis remains still to be confirmed.

Furthermore, Hfe may be involved in controlling iron homeostasis in a *non-hepcidin dependent manner*. It was proposed that Hfe may act to control iron release or iron uptake in cells since expression of HFE protein in HT29 human colon carcinoma cells, THP-1 cells, or in monocytes derived from HFE-HH patients inhibited iron release from the cells and resulted in increased iron accumulation without affecting iron uptake (Montosi et al., 2000; Drakesmith et al., 2002; Davies and Enns, 2004). On the other hand, over-expression of HFE in HeLa and hepatic HepG2 cells decreased iron loading in these cells (Gao et al., 2008), suggesting that HFE may exert distinct iron-regulatory functions depending on the site of its expression. In line with this, recent *in vivo* studies demonstrated that *Hfe* actions in macrophages are not required for the control of hepatic hepcidin expression (Makui

hepatic hepcidin transcription.

**FIGURE 2 | Beyond the hepatocytes.** The regulatory cues controlling iron metabolism are centered in the liver where hepatocytic-Hfe directs the production of the liver iron hormone hepcidin. The lack of hepatocytic-Hfe leads to inadequate production of hepcidin which results in increased iron uptake by the duodenum, iron release from macrophages into the circulation and deposition of the excess of iron in numerous tissues causing systemic iron overload (indicated by black arrows). The actions of the Hfe in extra-hepatocytic cells, such as erythroid and macrophages (indicated by grey circle), have recently been proposed suggesting for previously neglected functions of the Hfe in these cells. These selective extra-hepatocytic functions of Hfe are involved in the control of local, tissue-specific iron homeostasis however their impact on systemic iron regulation and the relevance for the Hfe-HH associated pathologies remains still to be discovered.

et al., 2005; Vujic Spasic et al., 2008) and that *Hfe* may display hepcidin-independent iron-modifying effect in these cells (Garuti et al., 2011) (**Figure 2**). Transplantation of *Hfe*+/<sup>+</sup> donor liver into *Hfe*−/<sup>−</sup> recipient mice improved hepatic hepcidin expression and reduced the liver iron burden in *Hfe*−/<sup>−</sup> mice, being in line with the initial clinical transplantation results (Garuti et al., 2011). Surprisingly, liver macrophages (Kupffer cells) were not rescued by the action of the donor, wild type hepatocytic-*Hfe* (Garuti et al., 2011).

These observations bring up a tantalizing thought whether Hfe is only required for the up regulation of hepcidin in response to elevated iron levels and thus for the maintenance of cellular and systemic iron pool, or the functions of this MHC-class I like molecule may be uncoupled from the control of iron homeostasis. Understanding distinct pathways that Hfe employs beyond the hepatocytes and beyond hepcidin regulation may leap translational medicine research in that some symptoms assigned to HH as a consequence of iron overload may be recognized appropriately and may better be treated by organ-specific therapies rather than systemic iron-depletion.

## **ACKNOWLEDGMENT**

This work was supported by the German Research Foundation grant (DFG, VU 75/2-1 to Maja Vuji´c).

## **REFERENCES**


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

*Received: 15 January 2014; paper pending published: 03 February 2014; accepted: 23 February 2014; published online: 11 March 2014.*

*Citation: Vuji´c M (2014) Molecular basis of HFE-hemochromatosis. Front. Pharmacol. 5:42. doi: 10.3389/fphar.2014.00042*

*This article was submitted to Drug Metabolism and Transport, a section of the journal Frontiers in Pharmacology.*

*Copyright © 2014 Vuji´c. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*

**MINI REVIEW ARTICLE** published: 07 May 2014 doi: 10.3389/fphar.2014.00093

## *Laura Silvestri\*, Antonella Nai, Alessia Pagani and Clara Camaschella\**

Division of Genetics and Cell Biology, San Raffaele Scientific Institute, Università Vita-Salute San Raffaele, Milan, Italy

#### *Edited by:*

Paolo Arosio, University of Brescia, Italy

#### *Reviewed by:*

Surjit Kaila Singh Srai, University College London, UK Francesca Vinchi, Heidelberg University, Germany

#### *\*Correspondence:*

Clara Camaschella and Laura Silvestri, Division of Genetics and Cell Biology, San Raffaele Scientific Institute, Università Vita-Salute San Raffaele, Via Olgettina 60, 20132 Milan, Italy e-mail: camaschella.clara@hsr.it; silvestri.laura@hsr.it

Transferrin receptor 2 (TFR2), a protein homologous to the cell iron importer TFR1, is expressed in the liver and erythroid cells and is reported to bind diferric transferrin, although at lower affinity than TFR1. TFR2 gene is mutated in type 3 hemochromatosis, a disorder characterized by iron overload and inability to upregulate hepcidin in response to iron. Liver TFR2 is considered a sensor of diferric transferrin, possibly in a complex with hemochromatosis protein. In erythroid cells TFR2 is a partner of erythropoietin receptor (EPOR) and stabilizes the receptor on the cell surface. However, Tfr2 null mice as well as TFR2 hemochromatosis patients do not show defective erythropoiesis and tolerate repeated phlebotomy. The iron deficient Tfr2-Tmprss6 double knock out mice have higher red cells count and more severe microcytosis than the liver-specific Tfr2 and Tmprss6 double knock out mice. TFR2 in the bone marrow might be a sensor of iron deficiency that protects against excessive microcytosis in a way that involves EPOR, although the mechanisms remain to be worked out.

**Keywords: iron metabolism, transferrin, transferrin receptors, hepcidin, iron deficiency, hemochromatosis**

### **THE SECOND TRANSFERRIN RECEPTOR**

Transferrin receptor 2 (TFR2) is a type II transmembrane glycoprotein, member of the TFR family and homologous to TFR1, which provides iron to the cell by internalization of the transferriniron complex through receptor-mediated endocytosis. TFR2 was cloned during a project aimed at isolating genes encoding new transcription factors (Kawabata et al., 1999). The *TFR2* gene comprises 18 exons and maps to chromosome 7q22 in close proximity to the EPO gene. Two *TFR2* isoforms have been described: the alpha isoform, corresponding to all 18 exons, encodes a protein of about 89 kDa in its unglycosylated form. The beta form, which results from an alternative splicing, lacks exons 1–3, and has 142 additional nucleotides in its first exon (exon 4 of the alpha form). The resulting protein lacks the cytoplasmic and the transmembrane domain (Kawabata et al., 1999) and its function remains unknown. The alpha protein encompasses 801 amino acids and, as TFR1, has a short cytoplasmic tail that contains a consensus sequence for endocytosis, a transmembrane domain and a large extracellular region that comprises a protease-associated domain and two RGD motifs (only one is present in TFR1), that bind diferric transferrin. TFR2 differs from TFR1 for several aspects. First, TFR1 is ubiquitously expressed, whereas the expression of TFR2 is restricted to the hepatocytes and erythroid precursors. Second, TFR1 is post-transcriptionally regulated by iron through the iron regulatory proteins–iron responsive elements (IRP–IRE) interaction, while TFR2 is not regulated by IRPs. TFR2 5 and 3 untranslated regions do not contain IRE elements (Fleming et al., 2000); rather TFR2 promoter shows GATA-1 as well as c-EBP-alpha consensus sequences. Furthermore, although in transfected cells TFR2 may uptake iron loaded transferrin (holo-TF) in transfected cells, *in vivo* it does not rescue the embryo-lethality of *Tfr1*−/<sup>−</sup> mice (Levy et al., 1999), suggesting a function distinct from TFR1 and unrelated to iron transport. Moreover, the affinity

of TFR2 for holo-TF is significantly lower than that of TFR1 (Kd 30 nM vs. 1 nM, respectively; Kawabata et al., 2000; West et al., 2000).

It has been reported that both TFRs bind hemochromatosis protein (HFE) *in vitro*. However, crystallographic studies have shown that HFE binds TFR1 (Bennett et al., 2000) at the same consensus sequences of diferric transferrin, implying a competition between the two ligands. On the contrary, based on *in vitro* data, binding of HFE to TFR2 and holo-Tf would occur simultaneously at two different TFR2 sequence motifs (Gao et al., 2009).

In the liver *Tfr2* expression increases during mouse development, at variance with *Tfr1*, and in adult liver *Tfr2* is much more expressed than *Tfr1* (Kawabata et al., 2001). Our knowledge of the TFR2 hepatic function is still incomplete. In hepatoma cell lines TFR2 is stabilized on cell surface by the addition of holotransferrin to the culture media, an effect due to the increased protein half-life (Enns, 2001). The divergent iron-mediated regulation of the two TFRs is confirmed also *in vivo*: while in iron loaded mice *Tfr1* is downregulated by the loss of IRP-mediated mRNA stabilization, Tfr2 protein level is increased. In agreement with a ligand-mediated stabilization, levels of Tfr2 protein are decreased in the liver of hypotransferrinemic (*hpx*) mice (Robb and Wessling-Resnick, 2004). Thus the major regulation of TFR2 occurs at the protein rather than at RNA level.

In cell culture models, TFR2 localizes in caveolar microdomains (Calzolari et al., 2006), membrane structures involved in the recruitment of receptors that can be activated by ligand binding (Simons and Toomre, 2000). In the absence of holo-transferrin, both TFR1 and TFR2 are internalized by clathrin-mediated endocytosis (Chen et al., 2009), whereas in the presence of the ligand only TFR2, and not TFR1, activates ERK1/2 and p38 MAPK. This has been observed in hepatoma derived and in erythroid cells, supporting the hypothesis that TFR2 may function as a signaling receptor (Calzolari et al., 2006; Poli et al., 2010).

The stabilization of TFR2 by holo-transferrin and its ability to bind HFE led to the current model in which liver TFR2, in conjunction with HFE, represents a sensor of circulating iron and activates hepcidin in response to elevated transferrin saturation (Goswami and Andrews, 2006). In addition it has been shown that the TFR2-HFE interaction on the hepatocyte surface occurs within a multiprotein complex, that *in vitro* includes also the BMP-coreceptor hemojuvelin (D'Alessio et al., 2012). If this complex activates the intracellular signaling to upregulate hepcidin expression *in vivo* remains to be demonstrated. In addition, the binding of TFR2 to HFE has recently been questioned (Rishi et al., 2013) and some evidences are in favor of a distinct function for the two HFEs. Mice with inactivation of both *Tfr2* and *Hfe* have a more severe phenotype compared to single mutant animals (Wallace et al., 2009). This occurs also in humans: patients with mutation in *TFR2* and *HFE* were reported to develop a severe form of juvenile-like hemochromatosis (Pietrangelo et al., 2005). Moreover, patients with *TFR2* mutations do not upregulate hepcidin upon oral iron administration, whereas the iron response is partially preserved in *HFE* patients (Girelli et al., 2011), thus strengthening the distinct and non-overlapping role of HFE and TFR2.

## **TFR2: THE GENE OF HEMOCHROMATOSIS TYPE 3**

In humans inactivating mutations of TFR2 lead to hemochromatosis type 3 (Camaschella et al., 2000), a rare recessive disorder characterized by iron overload, low hepcidin levels (Nemeth et al., 2005) and inability to properly regulate hepcidin after an oral iron challenge (Girelli et al., 2011). The disorder is quite rare among Caucasians and occasionally reported in Japanese, with single families identified in France, Portugal Spain and Taiwan. Currently, less than 30 pathogenic mutations have been described (Camaschella and Roetto, 2011). They are all rare, often private. Some insertions cause frameshift and premature stop codon, others are nonsense and small deletions. All the mutations are loss of function; missense mutations prevalently affect the protein C-terminus, especially the peptidase-like and the dimerization domains, suggesting that these regions have important functional roles (Camaschella and Roetto, 2011). The affected residues, usually highly conserved, might be essential for the proper folding and protein localization, or relevant for interaction with other proteins or for the regulatory function of TFR2.

The type of iron overload caused by mutations in TFR2 differs from the classic type 1 HFE-hemochromatosis, because of an earlier onset and more severe presentation (Camaschella, 2005). A study in patients affected by type 1 (HFE-related) or type 3 (TFR2 related) hemochromatosis showed a different role for TFR2 and HFE in hepcidin activation in response to a single oral iron challenge able to increase transferrin saturation: HFE patients showed a blunted hepcidin response, whereas TFR2 patients showed no response (Girelli et al., 2011). A similar difference in the hepcidin response after an acute iron loading has been observed in *Tfr2Y*245*X*/*Y*245*<sup>X</sup>* and *Hfe*−/<sup>−</sup> mice (Corradini et al., 2011; Ramos et al., 2011). These results led to the conclusion that TFR2 is important to up-regulate hepcidin in response to transferrin saturation.

## **TFR2 IN ERYTHROID CELLS**

In favor of the relevance of TFR2 for the erythroid differentiation are two genetic observations: first, the close proximity of TFR2 and EPO genes on chromosome 7q22 that may suggest a common regulation. Second, the results of different genome-wide association studies: TFR2 single nucleotide polymorphisms (SNPs) have been identified associated with erythroid quantitative traits, such as red cell numbers, indexes and hematocrit (Ganesh et al., 2009; Soranzo et al., 2009; Ding et al., 2012). Although this observation could be due to an indirect effect mediated by serum iron levels, that were not measured in the original studies (Ganesh et al., 2009; Soranzo et al., 2009; Ding et al., 2012), a direct effect of TFR2 on erythropoiesis cannot be excluded.

More recently it was shown that in erythroid cells TFR2 is a partner of erythropoietin receptor (EPOR) that stabilizes the receptor on cell surface. *TFR2* is co-expressed with *EPOR* during erythroid differentiation and its maximal expression precedes that of *TFR1* (Kawabata et al., 2001). TFR2 protein, that associates with EPOR in the endoplasmic reticulum, is needed for the efficient transport of the receptor to the cell membrane (Forejtnikova et al., 2010). The interaction facilitates the stabilization of EPOR, likely contributing to EPO sensitivity and erythroid cell differentiation, both *in vitro* and *in vivo*. However the interaction does not seem to influence EPO binding to EPOR. In addition direct binding of TFR2 to EPO was excluded (Forejtnikova et al.,2010). Interestingly *TFR2* silencing in human erythroid precursors delays their terminal differentiation *in vitro* (Forejtnikova et al., 2010). However, which signaling pathway is activated by TFR2 is still unclear. Moreover, although TFR2 is required for efficient erythropoiesis, *Tfr2* null mice as well as *TFR2* hemochromatosis patients do not show defective erythropoiesis and patients tolerate repeated courses of phlebotomy without developing anemia.

## **ANIMAL MODELS OF** *Tfr2* **INACTIVATION DEVELOP IRON OVERLOAD**

The first animal model of hemochromatosis type 3 was generated by targeted mutagenesis, introducing a premature stop codon (Y245X; Fleming et al., 2002) in the murine *Tfr2* coding sequence. This mutation is orthologous to the mutation (Y250X) originally detected in humans (Camaschella et al., 2000). Young (4 week-old) homozygous Y245X mutant mice had high liver iron concentration, even if maintained on a standard diet, in agreement with the observation of early iron overload in patients. The histological distribution of iron recapitulates features of hemochromatosis, with the typical liver periportal accumulation and low spleen iron stores. As in humans, heterozygous animals were normal. Later on, several murine models of *Tfr2* inactivation were developed (Fleming et al., 2002; Wallace et al., 2007), including, among others, *Tfr2* total (*Tfr2*−/−) and liver-specific (*Tfr2* LCKO) knock-out (Wallace et al., 2007; Roetto et al., 2010) and *Tfr2-Hfe* double knock-out. All these models are characterized by low hepcidin expression and liver iron overload of variable severity (**Table 1**). However, when generated in the same genetic background, *Tfr2* total knockout was shown to have iron overload more severe than *Hfe*−/<sup>−</sup> although less severe than *Hfe/Tfr2* double knock out. These observations are in agreement with the suggested distinct function of the two proteins.


**Table 1 | Hepcidin levels, iron and hematological phenotype in the available murine models of***Tfr2* **inactivation.**

RBC, red blood cells; Hb, hemoglobin.

\*Reduced compared to the level of iron-loaded mice; n.a = not available.

A last model was generated with the M167K substitution in the Tfr2 protein (Roetto et al., 2010): this mutation destroys the methionine, putative start codon of the beta-isoform of the protein. β-Tfr2 is mostly expressed in the spleen (Kawabata et al., 1999; Roetto et al., 2010). Interestingly, the knock-in model *Tfr2KI*, specifically lacking the beta-isoform, is characterized by normal transferrin saturation, liver iron concentration, hepcidin and *Bmp6* levels but show a transient anemia at young age. In addition adult animals accumulate iron in the spleen due to strong reduction of ferroportin mRNA, thus suggesting a possible regulatory effect of β-Tfr2 on splenic ferroportin expression.

## **ANIMAL MODELS OF** *Tfr2* **INACTIVATION IN IRON DEFICIENCY**

*Tfr2*−/<sup>−</sup> mice have slightly less severe iron overload than liverspecific (*Tfr2*LCKO) knock-out (Wallace et al., 2007; Roetto et al., 2010), slightly higher Hb levels (Roetto et al., 2010; Nai et al., 2014) and moderate macrocytosis. The *Tmprss6*−/<sup>−</sup> mice, which have a deletion of the hepcidin inhibitor, the serine protease Tmprss6, is a well established model of iron deficiency anemia with high hepcidin (Du et al., 2008; Folgueras et al., 2008). *Tmprss6*−/−*Tfr2*−/<sup>−</sup> double knock out animals develop iron deficiency with high hepcidin, a phenotype similar to *Tmprss6*−/<sup>−</sup> mice (Lee et al., 2012; Nai et al., 2014) and to *Tmprss6*−/−*Hfe*−/<sup>−</sup> animals (Finberg et al., 2011; Lee et al., 2012). In a single study some degree of erythrocytosis were observed both in *Tfr2*−/<sup>−</sup> and in *Hfe*−/<sup>−</sup> knock-out with deletion of *Tmprss6* (Lee et al., 2012), although results for *Hfe*−/−*Tmprss6*−/<sup>−</sup> are not unequivocal (Finberg et al., 2011).

Deleting *Tmprss6* in the two hemochromatosis type 3 models, *Tfr2*−/<sup>−</sup> and *Tfr2*LCKO mice, revealed similarities but also differences in the hematological phenotype of the resulting double knock-out animals (Nai et al., 2014). Both models have the same degree of anemia, low transferrin saturation and low liver iron content (LIC), a phenotype quite similar to that of the iron deficient *Tmprss6*−/<sup>−</sup> (**Table 1**).

The modification of the phenotype of *Tfr2*−/<sup>−</sup> mice with deletion of *Tmprss6* has important implications. First it indicates that hepatic *TFR2 i*s genetically upstream *TMPRSS6* in the BMP-SMAD signaling pathway, as previously shown for *HFE* (Finberg et al., 2011; Lee et al., 2012). Second, it excludes that TFR2 is a substrate of TMPRSS6, as previously observed in our *in vitro* studies (Pagani, unpublished observation, 2014). Further, it suggests that Tmprss6 is likely hyperactive in *Tfr2*−/<sup>−</sup> mice with iron overload and thus its inhibition might be effective in up-regulating hepcidin production and reducing iron overload, as shown by the use of small interference RNA (siRNA) or allele specific oligonucleotide (ASO) against *Tmprss6* (Guo et al., 2013; Schmidt et al., 2013) in *Hfe*−/<sup>−</sup> animals.

We observed that the phenotype of *Tfr2*-*Tmprss6* double knock-out mice is not exactly the same of the double knockout for *Tfr2*LCKO and *Tmprss6* or of *Tmprss6*−/<sup>−</sup> knock out (**Table 1**). An increased number of red cells are observed only in *Tfr2*-*Tmprss6* double knock-out mice. Also, while hepcidin levels are increased in all models as compared with wild-type animals, they are less elevated in*Tfr2*−/−*Tmprss6*−/<sup>−</sup> mice compared with the other models. It seems that some inhibitory signal lowers hepcidin in *Tfr2*−/−*Tmprss6*−/<sup>−</sup> mice. In principle this may derive from the increased red cell production, exclusively present in *Tfr2*−/−*Tmprss6*−/<sup>−</sup> mice. It is of interest that the observed erythrocytosis is not due to enhanced erythropoietin (Epo) stimulation of erythropoiesis, since Epo levels are similar and consistent with similar degrees of anemia in all models. They are even decreased in *Tmprss6*−/−*Tfr2*−/<sup>−</sup> (Nai et al., 2014). We speculate that this positive modulation of erythropoiesis might result from the lack of *Tfr2* expression in the erythroid compartment (**Figure 1**).

Recently an erythroid function for Tfr2 was independently reported byWallace et al. (2013)who noticed that the triple knockout (*Tfr2*−/−*, Hfe*−/−, *Tmprss6*−/−) mice have more severe iron deficiency than *Tmprss6*−/<sup>−</sup> mice with deletion of either *Hfe* or *Tfr2*. However, the mechanism underlining this difference remains to be worked out.

From all the data available we concluded that Tfr2 in the erythroid compartment might serve to block excessive erythropoietic expansion. If this occurs in normal conditions (**Figure 1** panel A) is difficult to verify because *Tfr2* deletion leads to iron overload. In iron overload indeed the Tfr2 erythroid function is likely masked by the excessive iron availability that increases Hb and also Hb content per single cell (**Figure 1** panel B vs. A; Roetto et al., 2010; (Nai et al., 2014). Tfr2 function becomes more evident in iron

#### **FIGURE 1 | Schematic representation of Tfr2-mediated erythropoiesis modulation. (A)** in normal conditions Tfr2 and erythropoietin receptor (EpoR) interact and localize on the cell surface and control red blood cells production **(B)** Tfr2 inactivation in mice causes iron overload: mean corpuscular hemoglobin (MCH) is enhanced (indicated by the darker red

color of the erythrocytes) due to the increased iron availability. **(C)** Inactivation of Tmprss6 in Tfr2 KO mice increases the number of microcytic erythrocytes, as indicated by the light red color of the erythrocytes. Epo-EpoR interaction occurs in all conditions but is increased in the absence of Tfr2 (panel B and C vs. A).

deficiency, as exemplified by *Tmprss6*−/−*Tfr2*−/<sup>−</sup> double knock out mice (**Figure 1** panel C).

## **CONCLUSION**

From its identification and cloning more than 10 years ago (Kawabata et al., 1999), the correct function of TFR2 in iron metabolism has remained mysterious. Several different roles have been proposed for this receptor: originally published as a second iron importer, after the identification of *TFR2* mutations in hemochromatosis patients its proposed function became that of potential sensor of circulating iron-loaded transferrin and then of hepcidin (co)-activator. Although counteracting iron excess in the circulation remains its major function in the liver, an as well important erythropoietic function is emerging from our and other studies From the few available data it seems that TFR2 might serve as a brake to avoid iron consumption in excessive erythrocyte production in conditions of iron deficiency, likely within the perspective of global body iron economy. A prevalent role of erythroid TFR2 in iron deficiency might explain why this role is not evident in mice nor in patients with type 3 hemochromatosis, who are iron loaded and never experience iron deficiency. Further studies are needed to clarify the molecular mechanisms that mediate the TFR2 function in iron deficiency. However from now on the erythropoiesis status should be considered when interpreting the effect of TFR2 in iron metabolism and homeostasis.

## **ACKNOWLEDGMENTS**

This work was partially supported by MIUR PRIN 2010–2011 (Rome, Italy) and the Italian Ministry of Health (Grant RF-2010- 2312048) to Clara Camaschella.

## **REFERENCES**


are associated with red blood cell traits. *Mayo Clin. Proc.* 87, 461–474. doi: 10.1016/j.mayocp.2012.01.016


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

*Received: 13 March 2014; accepted: 14 April 2014; published online: 07 May 2014. Citation: Silvestri L, Nai A, Pagani A and Camaschella C (2014) The extrahepatic role of TFR2 in iron homeostasis. Front. Pharmacol. 5:93. doi: 10.3389/fphar.2014.00093 This article was submitted to Drug Metabolism and Transport, a section of the journal Frontiers in Pharmacology.*

*Copyright © 2014 Silvestri, Nai, Pagani and Camaschella. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited andthatthe original publication inthis journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*

## The role ofTMPRSS6/matriptase-2 in iron regulation and anemia

## *Chia-YuWang1, Delphine Meynard2 and HerbertY. Lin1\**

<sup>1</sup> Program in Anemia Signaling Research, Division of Nephrology, Program in Membrane Biology, Center for Systems Biology, Massachusetts General Hospital, Harvard Medical School, Boston, MA, USA

<sup>2</sup> INSERM, U1043, CNRS, U5282, Université Paul Sabatier, Centre de Physiopathologie de Toulouse Purpan, Toulouse, France

#### *Edited by:*

Raffaella Gozzelino, Instituto Gulbenkian de Ciência, Portugal

#### *Reviewed by:*

Clara Camaschella, Vita Salute San Raffaele University and San Raffaele Scientific Institute, Italy Carole Beaumont, Institut National de la Santé et de la Recherche Médicale,

France *\*Correspondence:*

Herbert Y. Lin, Program in Anemia Signaling Research, Division of Nephrology, Program in Membrane Biology, Center for Systems Biology, Massachusetts General Hospital, Harvard Medical School, 185 Cambridge Street, CPZN-8216, Boston, MA 02114, USA e-mail: lin.herbert@mgh.harvard.edu

Matriptase-2, encoded by the TMPRSS6 gene, is a member of the type II transmembrane serine protease family. Matriptase-2 has structural and enzymatic similarities to matriptase-1, which has been implicated in cancer progression. Matriptase-2 was later established to be essential in iron homeostasis based on the phenotypes of iron-refractory iron deficiency anemia identified in mouse models as well as in human patients withTMPRSS6 mutations. TMPRSS6 is expressed mainly in the liver and negatively regulates the production of hepcidin, the systemic iron regulatory hormone. This review focuses on the current understanding of matriptase-2 biochemistry, and its role in iron metabolism and cancer progression. In light of recent investigations, the function of matriptase-2 in hepcidin regulation, how it is being regulated, as well as the therapeutic potential of matriptase-2 are also discussed.

**Keywords: iron,***TMPRSS6***, matriptase-2, iron overload, IRIDA**

## **BIOCHEMISTRY OF MATRIPTASE-2**

Type II transmembrane serine protease matriptase-2, encoded by the *TMPRSS6* gene, belongs to the family of type II transmembrane serine proteases (TTSP). Matriptase-2 is comprised of a transmembrane domain, followed by a sea urchin sperm protein, enteropeptidase and agrin (SEA) domain, a stem region containing two complement factor C1r/C1s, urchin embryonic growth factor and bone morphogenetic protein (CUB) domains and three low-density lipoprotein receptor (LDLR) class A repeats, and a C-terminal trypsin-like serine protease domain (Velasco et al., 2002; Ramsay et al., 2008). Matriptase-2 is synthesized as a single chain inactive proenzyme, which auto-activates itself by a cleavage at an arginine residue at the RIVGG consensus site between the prodomain and the catalytic domain (Ramsay et al., 2009b; Altamura et al., 2010). After the auto-activation, it remains membrane-bound through a single disulphide bond linking the pro- and catalytic domains (Ramsay et al., 2009a). Once the catalytic domain is released, it migrates as a single or dimeric species (Silvestri et al., 2008). Matriptase-2 shares high structural and enzymatic similarities with matriptase-1, which contains four LDLR repeats instead of three (Sanders et al., 2010), is expressed in epithelial cells, and has been implicated in the progression of cancers, such as breast, prostate, and colorectal cancer (Oberst et al., 2001; Velasco et al., 2002; Kang et al., 2003; Riddick et al., 2005).

The structural features of matriptase-2 are highly conserved across mammalian species, including human, macaque monkey, dog, cow, mouse and rat, with human protein sharing >80%

identity to matriptase-2 from other species (Ramsay et al., 2008). The expression pattern of *TMPRSS6* determined from mRNA expression studies and analysis of GenBank Unigene database indicates that matriptase-2 is predominantly expressed in the liver (Velasco et al., 2002; Finberg et al., 2008) but also to a lower extent in the kidney, spleen, brain, lung, mammary gland, testis, and uterus (Ramsay et al., 2008). In addition, aberrant expression of *TMPRSS6* is observed in different human cancers such as breast and prostate cancer (Parr et al., 2007; Sanders et al., 2008).

Matriptase-1, a close relative of matriptase-2, is known to be associated with two endogenous inhibitors: hepatocyte growth factor activator inhibitor (HAI)-1 and HAI-2, which inhibit matriptase-1 dependent activation of its physiological substrates, likely through an interaction with the second CUB domain (Szabo et al., 2008; Inouye et al., 2010). With 35% identity and structural similarities with matriptase-1 (Velasco et al., 2002), it is possible that matriptase-2 is also associated with an endogenous inhibitor. Indeed, Maurer et al. (2013) recently demonstrated that HAI-2 is a cognate inhibitor of matriptase-2 that inhibits its proteolytic activity, and thus increases hepcidin expression *in vitro*. However, the physiological role of HAI-2 in the regulation of hepcidin and iron metabolism remains to be investigated.

Following the identification and characterization of matriptase-2, Velasco et al. (2002) also examined the enzymatic activity of the catalytic serine protease domain against extracellular matrix components. It was found that matriptase-2 has the capacity to degrade fibronectin, fibrinogen, and type I collagen. Recently, membrane bound hemojuvelin has also been identified as a substrate for matriptase-2 *in vitro* (Silvestri et al., 2008), providing a straightforward mechanism for the effects of *TMPRSS6* mutations on hepcidin and iron regulations. However, as will be discussed below, evidence exists *in vivo* that is not consistent with this hypothesis.

## **ROLE OF MATRIPTASE-2 IN IRON METABOLISM**

Matriptase-2 is produced mainly by the liver and negatively regulates the production of hepcidin, the systemic iron regulatory hormone encoded by the *HAMP* gene (Du et al., 2008; Finberg et al., 2008). Hepcidin is a peptide secreted by the liver that plays a central role in adjusting iron absorption to meet iron needs of the body (Nicolas et al., 2001). Hepcidin negatively regulates cellular iron export by promoting the degradation of ferroportin (Nemeth et al., 2004), the only known iron exporter present on the surface of duodenal enterocytes, macrophages, and hepatocytes and thus limits iron absorption and iron release. It is now well established that *Hamp* expression is regulated by the bone morphogenetic protein (BMP)/sons of mothers against decapentaplegic (SMAD) signaling pathway (Babitt et al., 2006, 2007).

At the molecular level, BMP6, the endogenous ligand of BMP/SMAD signaling, activates BMP-receptor complex by binding to type I and type II BMP receptors that induces phosphorylation (Andriopoulos et al., 2009; Meynard et al., 2009). The activated complex, in turn, phosphorylates Smad1,5,8/Smad4 complex, which then translocates to nucleus to modulate gene transcription (Wang et al., 2005; Babitt et al., 2006; Kautz et al., 2008). Hemojuvelin (HJV) acts as a coreceptor and is required to fully activate the BMP signaling ability (Babitt et al., 2006). The expression of BMP6 is proportional to hepatic iron concentrations and consistent with *Hamp* mRNA expression (Kautz et al., 2008).

## *TMPRSS6* **MUTATIONS IN MICE AND HUMAN**

Matriptase-2 regulates *Hamp* expression through the BMP/SMAD pathway (Finberg et al., 2010; Lenoir et al., 2011) in an as yet unfully characterized manner. Mice without functional matriptase-2 (both *mask* mice with truncated *Tmprss6* lacking the protease domain and *Tmprss6* knockout mice) showed a hypochromic microcytic anemia and an alopecia (Du et al., 2008; Folgueras et al., 2008). These phenotypes resulted from inappropriately high levels of *Hamp* mRNA expression (Du et al., 2008; Folgueras et al., 2008; Finberg et al., 2010).

Mutations in *TMPRSS6* in humans led to iron-refractory iron deficiency anemia (IRIDA) that is unresponsive to oral iron treatment and only partially responsive to parental iron therapy (Finberg et al., 2008). IRIDA is also characterized by congenital hypochromic, microcytic anemia, low mean corpuscular erythrocyte volume, low transferrin saturation, and defects in iron absorption and utilization (Finberg et al., 2008; Guillem et al., 2008; Melis et al., 2008). Currently, there are 42 different *TMPRSS6* mutations reported in humans, scattered throughout all the different extracellular domains (**Figure 1**).

Interestingly, in contrast to current understanding of autosomal recessive disorder, haploinsufficiency is observed in some *TMPRSS6* mutations (**Figure 1**; Finberg et al., 2008; Pellegrino et al.,2012;Jaspers et al.,2013). Haploinsufficiency is also observed in animal models. Nai et al. (2010) reported that *Tmprss6* heterozygous knockout mice are more susceptible to iron deficiency compared to their wild-type littermates. Finberg et al. (2011) also demonstrated that, compared to mice deficient for *Hfe* alone, heterozygous loss of *Tmprss6* in *Hfe* knockout mice had higher hepcidin levels at 4 weeks of age, which presumably resulted in decreased hepatic iron concentrations at 8 weeks of age.

Human genome wide association studies (GWAS) highlighted the significance of matriptase-2 in control of iron homeostasis by identifying common *TMPRSS6* variants associated with abnormal hematological parameters, including hemoglobin, transferrin saturation, erythrocyte mean cell volume (MCV) and serum iron concentrations (Benyamin et al., 2009; Chambers et al., 2009; Tanaka et al., 2010). Following GWAS, population-based cohort studies were investigated in China and Italy to study the association between serum iron parameters, iron-related diseases and specific *TMPRSS6* single nucleotide polymorphisms (SNPs): rs855791 (V736A) and rs4820268 (D521D). It was found that *TMPRSS6* SNPs was associated with lowered serum iron, hemoglobin, and plasma ferritin levels, consistent with lowered risk of iron overload and increased risk of iron deficiency anemia in Chinese population (An et al., 2012; Gan et al., 2012). A retrospective cohort study in northern Italy also suggested that *TMPRSS6* V736A polymorphism is likely to be a gene modifier in hemochromatosis patients, influencing the susceptibility of cirrhosis (Valenti et al., 2012). Nai et al. (2011) demonstrated that *TMPRSS6* V736A directly modulates *HAMP* expression *in vitro* and that healthy individuals with the homozygous substitution had lower levels of serum hepcidin, higher serum iron and higher transferrin saturation. Taken together, these studies clearly establish *TMPRSS6*/matriptase-2 as an important regulator of iron homeostasis in humans. A recent review focused more on the anemia induced by matriptase-2 mutations is complementary to the current review (De Falco et al., 2013).

## **FUNCTION OF MATRIPTASE-2 IN HEPCIDIN REGULATION**

Matriptase-2 inhibition of hepcidin activation by cleaving membrane hemojuvelin has been established *in vitro* (Silvestri et al., 2008). When overexpressed in HeLa cells, matriptase-2 interacts and induces the cleavage of membrane hemojuvelin at the cell surface, resulting in the generation of soluble hemojuvelin that is released into the cell medium (Silvestri et al., 2008). However, in both *mask* and *Tmprss6* knockout mice, hepatic hemojuvelin levels at the membrane were found unexpectedly to be decreased, compared to wild-type animals (Krijt et al., 2011; Frydlova et al., 2013). In addition, the levels of serum soluble hemojuvelin, which one would expect to be decreased in *Tmprss6* knockout, did not differ from wild-type mice (Chen et al., 2013). Although the possibility that soluble hemojuvelin and fragments are rapidly degraded *in vivo* cannot be excluded, these data suggested that hemojuvelin may not be the endogenous substrate of matriptase-2 and that matriptase-2 functions in a more complicated way *in vivo* than by merely cleaving hemojuvelin to regulate hepcidin and iron.

Several studies have been conducted to study the role of matriptase-2, by crossing*Tmprss6* knockout mice with several iron overload mouse models, including the generations of *Hjv/Tmprss6*, *Bmp6/Tmprss6*, *Hfe/Tmprss6*, and *Tfr2/Tmprss6* double mutant mice (Truksa et al., 2009; Finberg et al., 2011; Lenoir et al., 2011; Lee et al., 2012). In mice lacking both *Hjv* and *Tmprss6*, *Id1*, a target gene of BMP6 signaling, and *Hamp* mRNA levels were low, whereas serum iron, transferrin saturation, and liver iron concentration were high, similar to phenotypes of mice deficient for *Hjv* alone (Truksa et al., 2009; Finberg et al., 2010). These results indicate that if the substrate of matriptase-2 is downstream of hemojuvelin, it is likely to be along the SMAD signaling pathway. It is known that inflammatory cytokines, such as LPS and IL6, can induce *Hamp* expression in the absence of *Hjv* (Niederkofler et al., 2005), presumably via the Stat3 and Stat5 pathways (Verga Falzacappa et al., 2007; Meynard et al., 2013). However, it was surprising to find that the lack of both *Hjv* and *Tmprss6* in mice did not impair the responsiveness of hepcidin to BMP2 and IL6, but did fail to respond to iron challenge (Truksa et al., 2009). In mice deficient for both *Bmp6* and *Tmprss6*, the levels of *Hamp* and *Id1* mRNAs did not differ from mice deficient for *Bmp6* alone; however, their plasma iron levels and hepatic iron stores were significantly lower, suggesting the loss of matriptase-2 ameliorates iron overload conditions in *Bmp6* knockout mice (Lenoir et al., 2011). It is unclear why *Bmp6/Tmprss6* mice had less iron loading compared to mice deficient for *Bmp6* alone, but *Hamp* mRNA levels did not differ between *Bmp6/Tmprss6* and *Bmp6* knockout mice. Whether matriptase-2 has a significant role besides effects on BMP/SMAD signaling in iron metabolism, remain to be investigated.

Mice deficient for *Hfe* or *Tfr2* alone also develop iron overload phenotypes with inappropriately low *Hamp* mRNA expression and high serum iron parameters, compared to wild-type animals (Ahmad et al., 2002; Wallace et al., 2005). It is suggested that *Hfe* competes with transferrin for binding to transferrin receptor-1 and thus inhibits *Hamp* expression (Giannetti and Bjorkman, 2004; Schmidt et al., 2008). Others also showed that *Hfe*

knockout mice had high *Bmp6* mRNA expression but inappropriately low Smad1/5/8 phosphorylation, suggesting *Hfe* facilitates signal transduction initiated by BMP6 (Corradini et al., 2009; Kautz et al., 2009). However, the underlying mechanisms of how *Hfe* and Tfr2 contribute in BMP/SMAD signaling pathway is unclear. Mice deficient for both *Hfe* or *Tfr2* and *Tmprss6*, had high *Hamp* mRNA expression and exhibited iron deficiency microcytic anemia mimicking the phenotypes of mice lacking functional matriptase-2 alone (Finberg et al., 2011; Lee et al., 2012). This suggests that *Hfe* and Tfr2, if involved in BMP/SMAD pathway, are likely to be upstream of matriptase-2 signaling.

## **REGULATION OF MATRIPTASE-2**

Studies have shown that matriptase-2 expression can be modulated by iron status (Meynard et al., 2011; Zhang et al., 2011). In rats under acute iron deprivation, hepatic matriptase-2 protein levels are upregulated to repress hepcidin production (Zhang et al., 2011). Interestingly, matriptase-2 levels are also increased in response to chronic iron treatment and BMP6 administration in mice, possibly to prevent excessive hepcidin production, suggesting a dual role of matriptase-2 in the maintenance of tight systemic iron balance in response to iron (Meynard et al., 2011). In addition, studies also suggest that *TMPRSS6* mRNA expression is suppressed by conditions of inflammation (Meynard et al., 2013) and is upregulated in hypoxia (Lakhal et al., 2011; Maurer et al., 2012) and by erythropoietin (Peng et al., 2010). Human hepatoma Hep3B cells treated with interleukin-6 and mice injected with lipopolysaccharide demonstrated a downregulation of *TMPRSS6* via a decrease in Stat5 phosphorylation, independent of BMP/SMAD pathway (Meynard et al., 2013). Studies using Hep3B cells revealed that *TMPRSS6* is upregulated by HIF-1α and HIF-2α. This upregulation resulted in a decrease in membrane hemojuvelin and thus reducing hepcidin production (Lakhal et al., 2011). In mice, *Tmprss6* mRNA expression is induced by erythropoietin (Peng et al., 2010), which is also shown to be a negative regulator of hepcidin expression (Sasaki et al., 2012). Whether the

downregulation of hepcidin by erythropoietin is dependent on *Tmprss6* or through other unidentified mechanisms remains to be investigated.

## **MATRIPTASE-2 AS A THERAPEUTIC TARGET**

Genetic studies of mice deficient for both *Tmprss6* and *Hfe* or *Tfr2* or *Hbbth*3/+, the mouse model of β-thalassemia intermedia, have shown that iron overload can be prevented by targeting *Tmprss6* (Finberg et al., 2011; Lee et al., 2012; Nai et al., 2012). It is believed that the therapeutic effect did not come from silencing *Tmprss6* directly but from increased hepcidin production, resulting in lowered circulating iron burden (Camaschella, 2013). Studies targeting *Tmprss6* in *Hbbth*3/<sup>+</sup> and *Hfe* knockout mice by injecting silencing RNA (Schmidt et al., 2013) and anti-sense oligonucleotides (Guo et al., 2013) have successfully suppressed *Tmprss6* mRNA expression, leading to elevated hepcidin levels, improved iron overload in *Hfe* knockout and anemia and β-thalassemic mice. It is unclear how the ineffective erythropoiesis is improved by dampening *Tmprss6* expression in *Hbbth*3/<sup>+</sup> mice. However, higher hepcidin level inhibiting iron delivery to the erythroid precursors seems to play a role as evident by the similar effects achieved by overexpression of *Hamp*, iron restriction, and the injection of transferrin to *Hbbth*3/<sup>+</sup> mice (Gardenghi et al., 2010; Li et al., 2010; Finberg, 2013).

One limitation of using this method is that, unlike traditional phlebotomy and chelation therapies, iron is not removed or excreted from the body, and therefore, may not be an ideal treatment for patients with severe iron overload and transfusiondependent thalassemia (Camaschella, 2013). It could, however, improve therapeutic efficacy when used in combination with other traditional therapies by preventing intestinal iron absorption. A key issue for the use of RNA interference for clinical applications is the delivery method. There are safety concerns with viral vectors and non-viral delivery methods, which are still in their early development stage. Concerns have also been raised regarding the potential for off-target effects of siRNAs and their possible induction of interferon-stimulated genes. Other novel inhibitors of *TMPRSS6*, such as small molecule inhibitors, once identified, may eventually become useful therapeutic agents as well.

## **ROLE OF MATRIPTASE-2 IN CANCER**

Numerous members of the type II transmembrane serine protease family have been associated with a variety of different human cancers due to the differential expression patterns observed in these proteases between normal and cancerous tissues and cells (Webb et al., 2011). However, there are only a limited number of studies examining the involvement of matriptase-2 in human cancer, including breast cancer (Hartikainen et al., 2006; Parr et al., 2007; Tuhkanen et al., 2013) and prostate cancer (Sanders et al., 2008; Webb et al., 2012).

The association between matriptase-2 and breast cancer was established by a case control study in eastern Finnish population where they found a SNP (rs733655) in *TMPRSS6* gene associated with increased breast cancer risk (Hartikainen et al., 2006). It was later shown that *TMPRSS6* mRNA expression inhibits breast tumor development and thus correlates with favorable prognostic

outcome in patients (Parr et al., 2007). Recently, Tuhkanen et al. (2013) also demonstrated the association of several *TMPRSS6* variants with breast cancer risk and survival. It was highlighted that matriptase-2 protein levels decrease with tumor progression, and lower gene expression is seen in poor-prognosis-related triplenegative breast cancers (Tuhkanen et al., 2013). Mastriptase-2 is also implicated in tumor invasion and metastasis in prostate cancer *in vitro* (Sanders et al., 2008; Webb et al., 2012). These results indicate the involvement of matriptase-2 in tumor development. However, it is not clear whether the role of *TMPRSS6* in cancer progression is due to its ability to cleave extracellular matrix component such as fibronectin or due to a modification of iron parameters in cancer cells.

*TMPRSS6* expression is predominantly found in low invasive breast cancer cell lines such as MCF-7 and is absent in more invasive breast cancer cell lines such as MDA-MB-231 (Parr et al., 2007). Overexpression of matriptase-2 in MDA-MB-231 leads to a reduction of invasiveness and motility of the transfected cells and suppresses their tumorigenesis when xenografted in athymic nude mice suggesting that matriptase-2 could be involved in cancer progression through its capacity to cleave extracellular matrix components (Parr et al., 2007). However, variations of the iron status and iron regulatory genes expression were not addressed in the transfected cells in this study.

Many cancers exhibit an increased requirement for iron, presumably because of the need for iron as a cofactor in proteins essential to sustain growth and proliferation. The iron exporter ferroportin is expressed in breast cancer cells. Pinnix et al. (2010) showed that cells with high hepcidin and low ferroportin levels tended to be more aggressive. They concluded that having a breast cancer with low hepcidin and high ferroportin levels is an independent predictor of prognosis for a >90% 10-year survival rate (Pinnix et al., 2010), however, the mechanism is still to be investigated. Further studies are required to clarify the role of matriptase-2 in cancer progression.

## **ACKNOWLEDGMENTS**

This study was supported by National Institutes of Health, National Institute of Diabetes and Digestive, and Kidney Diseases grants R01 DK069533 and R01 DK071837 (Herbert Y. Lin).

## **REFERENCES**


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

*Received: 17 March 2014; paper pending published: 07 April 2014; accepted: 29 April 2014; published online: 19 May 2014.*

*Citation: Wang C-Y, Meynard D and Lin HY (2014) The role of TMPRSS6/matriptase-2 in iron regulation and anemia. Front. Pharmacol. 5:114. doi: 10.3389/fphar.2014.00114*

*This article was submitted to Drug Metabolism and Transport, a section of the journal Frontiers in Pharmacology.*

*Copyright © 2014 Wang, Meynard and Lin. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*

## Iron deficiency in the elderly population, revisited in the hepcidin era

## *Fabiana Busti, Natascia Campostrini, Nicola Martinelli and Domenico Girelli\**

Department of Medicine, University of Verona, Verona, Italy

#### *Edited by:*

Paolo Arosio, University of Brescia, Italy

#### *Reviewed by:*

Poli Maura, University of Brescia, Italy Mayka Sanchez, Josep Carreras Leukaemia Research Institute, Spain

#### *\*Correspondence:*

Domenico Girelli, Department of Medicine, University of Verona, Policlinico G.B. Rossi, P.le L.A. Scuro, 37134 Verona, Italy e-mail: domenico.girelli@univr.it

Iron deficiency (ID) is relatively common among the elderly population, contributing substantially to the high prevalence of anemia observed in the last decades of life, which in turn has important implications both on quality of life and on survival. In elderly subjects, ID is often multifactorial, i.e., due to multiple concurring causes, including inadequate dietary intake or absorption, occult bleeding, medications. Moreover, because of the typical multimorbidity of aged people, other conditions leading to anemia frequently coexist and make diagnosis of ID particularly challenging. Treatment of ID is also problematic in elderly, since response to oral iron is often slow, with a substantial fraction of patients showing refractoriness and requiring cumbersome intravenous administration. In the last decade, the discovery of the iron regulatory hormone hepcidin has revolutionized our understanding of iron pathophysiology. In this review, we revisit ID among elderly people in the light of the impressive recent advances on knowledge of iron regulation, and discuss how hepcidin may help in diagnosis and treatment of this common clinical condition.

**Keywords: iron deficiency, anemia, elderly, hepcidin, ferritin**

## **ANEMIA IN ELDERLY, PREVALENCE, AND DEFINITION**

Anemia is a common, multi-factorial condition in elderly. Indeed, the prevalence of anemia increases with age, representing an important health problem among older individuals. Large studies on community-dwelling older adults from the United States and Europe have reported prevalence rates for anemia ranging from 8 to 25% (Patel, 2008). One of the largest population survey, i.e., the third US National Health and Nutrition Examination Survey (NHANES III), indicated that 10.2% of women and 11% of men >65 years of age were anemic (Guralnik et al., 2004). These fractions rose to 26.1 and to 20.1% in subjects older than 85 years old, in males and females, respectively (Guralnik et al., 2004).

There is some debate on which hemoglobin (Hb) threshold should be used to define anemia in the general population and particularly in elderly individuals (Beutler and Waalen, 2006). In many studies, anemia has been defined according to the World Health Organization (WHO) criteria (Blanc et al., 1968) as a Hb level <13 g/dL in men and <12 g/dL in women, respectively. However, these criteria have been criticized since they were based on statistical distributions (i.e., equivalent to two standard deviations below the mean) in reference samples that did not include individuals >65 years of age, making unfeasible their application to older individuals (Mindell et al., 2013). Since Hb values in apparently healthy elderly individuals are generally lower than those in younger adults and the differences between males and females tend to disappear with aging (Patel, 2008), a Hb value < 12 g/dL is now commonly considered indicative of anemia in elderly of both sexes (Izaks et al., 1999; Andrès et al., 2013).

Anemia in elderly is typically hyporegenerative and relatively mild, with Hb levels near 10–11 g/dL in most subjects (Guralnik et al., 2004). Nevertheless, it is associated with a variety of adverse outcomes, including longer hospitalization, disability, and

increased mortality risk (Chaves et al., 2004; Zakai et al., 2005; Culleton et al., 2006; Denny et al., 2006; Penninx et al., 2006; den Elzen et al., 2009; Price et al., 2011). Moreover, it also significantly impacts on the quality of life, being associated with fatigue, cognitive dysfunction, depression, decreased muscle strength, falls, and "frailty," even when Hb levels are merely low-normal (Woodman et al., 2005; Eisenstaedt et al., 2006).

Approximately, one-third of the cases of anemia in elderly can be ascribed to a chronic disease (inflammation and chronic kidney diseases), and one-third is due to nutrient deficiencies (folate, B12, and iron). Iron deficiency (ID), alone or in combination with deficiency of other nutrients, accounts for more than onehalf of this group. The last third remains "unexplained" (Guralnik et al., 2004). Noteworthy, a significant proportion of elderly anemic patients (30–50%) is presumed to have multiple causes of anemia (Petrosyan et al., 2012). Since elderly patients are typically affected by several different pathologic conditions (multimorbidity), and are commonly taking a long list of medications, the precise etiology of anemia is often difficult to determine in a given individual (Andrès et al., 2013), and sometimes remains "unexplained" despite extensive investigation (Guralnik et al., 2004). Thinking in terms of multimorbidity is a key to understanding, diagnosis, and treatment of anemia in the elderly.

## **IRON DEFICIENCY IN ELDERLY**

According to the WHO, ID is by far the most common and widespread nutritional disorder worldwide (http://www.who.int/ nutrition/topics/ida/en/), with estimated one billion people affected, thus constituting a public health condition of epidemic proportions. Besides the large number of children and young women affected in developing countries, ID is the only nutrient deficiency that is also significantly prevalent in industrialized countries [World Health Organization (WHO), 2001; Hershko and Camaschella, 2013], where an additional category at risk is represented by elderly people (Guyatt et al., 1990).

Iron deficiency syndromes include a range of different conditions (Goodnough, 2012). "*Absolute*" ID is defined by the lack of storage iron (Cook, 2005; Fairweather-Tait et al., 2013). In physiological conditions, the total body iron amount (near 3–4 g) is maintained by a fine balance between three distinct factors: body requirements, iron supply (depending on dietary iron intake and duodenal absorption), and blood losses. While an increased iron demand is the main cause of ID in children and fertile females, insufficient dietary iron intake, gastrointestinal (GI) malabsorption and/or increased blood losses are the most common causes of ID in older individuals (see below).

At variance with "absolute" ID, many disorders are characterized by the so-called "*functional*" or "*relative*" ID, defined as the occurrence of iron-restricted erythropoiesis in presence of normal or even increased amounts of body iron stores. This phenomenon is often related to an impaired iron trafficking (i.e., block of iron release from macrophages and hepatocytes, typically during inflammatory diseases) or to increased/ineffective/stimulated erythropoiesis, with iron demand exceeding the supply (i.e., during hemoglobinopathies, chronic hemolytic anemias or treatment with erythropoiesis stimulating agents). Since the focus of this article is on etiology, diagnosis, and management of the absolute

ID in elderly, the readers are referred to others excellent reviews for details on the functional ID syndromes (Goodnough et al., 2010, Goodnough, 2012; Auerbach et al., 2013a).

Whatever the mechanism, both absolute and functional ID reduce iron availability to erythroid precursors, with the development of an iron-restricted erythropoiesis, and finally of anemia. In particular, two ID stages can be distinguished: (a) initial, characterized by reduced transferrin saturation but without anemia; and (b) advanced, when microcytic, hypochromic iron-deficiency anemia (IDA) becomes evident.

In elderly, ID and IDA are nearly always due to chronic GI diseases, which in turn lead to iron loss and malabsorption not infrequently occurring in combination at individual level (**Figure 1**). Indeed, the most frequent cause is represented by *chronic upper and lower GI blood losses*, because of esophagitis, gastritis, peptic ulcer, colon cancer or pre-malignant polyps, inflammatory bowel disease, or angiodysplasia (Eisenstaedt et al., 2006). The prevalence of most of these conditions increases with age, which is particularly true for neoplastic lesions (Eddy, 1990) and angiodysplasia (Sami et al., 2014). Remarkably, GI bleeding is typically increased by concomitant assumption of medications for conditions highly prevalent in elderly individuals, such as non-steroidal anti-inflammatory drugs for osteoarthritis, and antithrombotic therapies for cardiovascular disease, especially for atrial fibrillation.

**Frontiers in Pharmacology** | Drug Metabolism and Transport April 2014 | Volume 5 | Article 83 |

favored by antithrombotic drugs for treatment of cardiovascular diseases that

disease; PPI, proton pump inhibitors; VCE, video capsule endoscopy.

*Iron malabsorption* is also relatively frequent in the elderly. Indeed, further conditions whose prevalence typically increases with age are represented by *Helicobacter pylori* (HP) infection (Pounder and Ng, 1995) and atrophic gastritis. Of note, although, for a long time, celiac disease (CD) has been primarily considered an enteropathy of childhood and young adults, a number of epidemiological studies have reported an increased detection rate in older subjects, with up to one third of newly diagnosed patients being older than 65 years (Patel et al., 2005; Rashtak and Murray, 2009; Vilppula et al., 2009). In this age group, multifactorial anemia is the most frequent clinical presentation (Harper et al., 2007), with micronutrients deficiency (particularly ID) being the leading cause. For poorly understood reasons, the classical triad of malabsorptive symptoms including diarrhea, weight loss and abdominal pain is less common in elderly (Freeman, 2008), making the diagnosis frequently overlooked in this age category. Another factor that could theoretically contribute to iron malabsorption in elderly patients is represented by the frequent long-term use of proton pump inhibitors (PPI), being gastric acid essential for optimal intestinal absorption of the element (Ganz, 2013). However, only few reports have specifically addressed this issue, which remains controversial (Reimer, 2013).

Typically, all the above-mentioned conditions impairing iron absorption share a clinical phenotype of refractoriness to oral iron therapy, recently named "acquired IRIDA" (iron refractory ID anemia; for a review see Hershko and Camaschella, 2013). These conditions should be always considered in elderly subjects with IDA and no evidence of GI blood loss.

Finally, *malnutrition* is an obvious contributing factor to ID in elderly. However, since iron requirement (1–2 mg/day) only corresponds to near 10% of the average daily iron intake, malnutrition is rarely sufficient *per se* to cause IDA, at least in industrialized countries. Nevertheless, evaluation of the patient's nutritional status plays an important role in the diagnostic approach to anemia in the older adult.

## **HEPCIDIN, THE KEY REGULATOR OF IRON HOMEOSTASIS**

Hepcidin, a defensin-like hormone synthesized mainly by the liver, has been discovered in 2001 and recognized as the master regulator of iron metabolism (Ganz and Nemeth, 2011). The active form of hepcidin is a 25-amino acid peptide derived from an 84 amino acid precursor, but at least two others isoforms truncated at the N-terminus, i.e., hepcidin-20 and hepcidin-22, have been also identified in biological fluids (Castagna et al., 2010). The biological meaning of these isoforms is still unclear (Campostrini et al., 2012). Hepcidin acts by binding to its receptor, the transmembrane protein ferroportin, which currently represents the only known cellular iron exporter in mammals (De Domenico et al., 2011). In humans, ferroportin is mainly expressed in cells playing a key role in iron homeostasis, like duodenal enterocytes (absorption of dietary iron), in splenic and hepatic macrophages (recycling iron from erythrophagocytosis), and in hepatocytes (iron stores). The hepcidin-ferroportin binding induces the endocytosis and the lysosomal degradation of both molecules, resulting in decreased intestinal absorption and release of iron from recycling macrophages, both ultimately leading to reduction of plasma iron concentration (Ganz and Nemeth, 2011). Regulation of

hepcidin synthesis is quite complex and includes a number of different pathways [for recent detailed reviews see Ganz (2013) and Meynard et al. (2014)]. ID and increased erythropoietic activity down-regulate hepcidin production, and suppressed or very low hormone concentrations are observed in IDA or anemias with high erythropoietic activity (Ganz et al., 2008). Although the nature of the suppressive signal is still unknown, there is some evidence that, at least in conditions of stimulated erythropoiesis, it could be represented by a circulating factor produced by the erythroid progenitors in the bone marrow (Kautz et al., 2013). On the other hand, hepcidin is strongly induced by inflammation (Nemeth et al., 2003, 2004), in particular by the pro-inflammatory cytokine interleukin-6 (IL-6), and it is responsible for iron-limited erythropoiesis in patients with acute and chronic inflammatory states (Ganz, 2003; McCranor et al., 2013). Nevertheless, recent studies in mouse models (Gardenghi et al., 2014; Kim et al., 2014) have suggested that the iron-restricted anemia induced by inflammation likely recognizes a more complex pathogenesis, only partially dependent on hepcidin.

Currently, two main methods are available for measuring hepcidin in blood and urine, immunoassays based on anti-hepcidin antibodies, and mass spectrometry (MS)-based assays (Castagna et al., 2010; Kroot et al., 2011). The latters are generally preferable, being able to distinguish the iron bioactive 25-mer isoform from other isoforms of uncertain significance, at variance with the incomplete specificity of available antibodies (Castagna et al., 2010; Kroot et al., 2011). Serum hepcidin shows well-defined age- and sex-specific variations at population level as illustrated below (Galesloot et al., 2011; Traglia et al., 2011). The measurement of hepcidin in biological fluids represents a promising tool in the diagnosis and management of conditions characterized by an altered iron homeostasis, including ID/IDA. However, a "gold standard"method available for daily clinical practice at reasonable costs is stills lacking.

### **HEPCIDIN LEVELS IN ELDERLY**

The "unexplained" anemia of elderly has been linked to two putative mechanisms, namely a progressive resistance of bone marrow erythroid progenitors to erythropoietin (EPO), and a chronic subclinical pro-inflammatory state (Vanasse and Berliner, 2010). In this context, hepcidin could theoretically play a substantial role, considering its involvement both in inflammation and in the regulation of iron availability for erythropoiesis. Indeed, previous studies have showed a mild increase of inflammatory markers like tumor necrosis factor alpha (TNF-α) and IL-6, a major hepcidin inducer, in elderly subjects (Andrews, 2004; Ferrucci et al., 2005; Maggio et al., 2006). To date, only two studies, the Nijmegen biomedical studies (NBS; Galesloot et al., 2011), and the Val Borbera studies (VBS; Traglia et al., 2011), have investigated serum hepcidin levels at population level in apparently healthy subjects, including elderly groups. Both studies showed that before the menopause hepcidin levels in women are nearly 50% lower than in males of corresponding ages. After the menopause, hepcidin levels tend to be similar in both sexes, with a slight decrease in the eldest groups. This was evident in both sexes in theVBS (**Figure 2**), but only in males in the NBS. Although not specifically designed to study the anemia of elderly, these studies tended to exclude a

sustained increase of hepcidin in elderly. Accordingly, two studies in elderly anemic patients have failed to detect increased hepcidin levels in urine (Ferrucci et al., 2010) and in serum (Waalen et al., 2011), and even a correlation between hepcidin and IL-6 or TNF-α (Ferrucci et al., 2005).

On the contrary, the Leiden 85-plus Study, a population-based prospective study in 85-year old subjects from Leiden (the Netherlands), showed that in this age group C-Reactive Protein (a proxy for IL-6) was a significant predictor of circulating hepcidin levels, which in turn were relatively higher in a small subgroup (*n* = 29) with unexplained anemia (den Elzen et al., 2013). Several reasons may account for these discrepancies on hepcidin levels in elderly anemic subjects, including the different laboratory methods and clinical settings (Goodnough and Schrier, 2014). Most of the data are retrospective, and properly designed, large-scale, studies are required before drawing definite conclusions. For the moment, the available evidence has generally downsized the initial hypothesis on hepcidin as a major determinant of the unexplained anemia of the elderly. This remains likely a complex condition due to the combination of several age-related changes, such as stem cell aging, low-grade chronic inflammation, subclinical impairment of kidney function, androgen insufficiency, and others still unknown (Guralnik et al., 2004; Makipour et al., 2008).

### **DIAGNOSIS OF IDA IN ELDERLY**

Several guidelines and recommendations have been proposed for the diagnosis of IDA in the general population (Cook, 2005; Goddard et al., 2011), but there is no consensus regarding the optimal approach for the diagnosis and management of IDA in elderly. Nevertheless, it is clear that, besides iron supplementation, the general principle of searching, and if possible treating, the underlying cause(s) should be pursued also in elderly patients (Andrès et al., 2013). Being GI diseases the most common causes of IDA in elderly (**Figure 1**), the diagnostic work-up should often, at least theoretically, include relatively invasive investigations, like endoscopic procedures. This is particularly true since, for example, IDA in elderly often herald the presence of an occult GI malignancy. Of course, old age *per se* is not a contraindication to such procedures, but a particular clinical skill is required in each individual and frail elderly patient to thoroughly evaluate the risk-benefit ratio as well as the prognostic implications.

Anyway, while the diagnostic work-up should be, whenever possible, comprehensive, some conditions deserve peculiar attention in the elderly patient presenting with IDA.

In our experience, a condition particularly challenging is represented by GI angiodysplasia, which in turn is a potentially treatable disease (Richter et al., 1984). Remarkably, bleeding in GI angiodysplasia is often discontinuous, with possible false negativity of occult fecal blood test. Moreover, although most of GI angiodysplasia lesions are localized in the colon (54–82% are in the cecum and in ascending colon), they can escape to a single endoscopy, or be localized in the small bowel, which is not routinely investigated. In this case, additional testing by video capsule endoscopy (VCE) is needed (Sami et al., 2014). Finally, a distinct feature of GI angiodysplasia is its frequent association with another relatively common condition in elderly, i.e., aortic stenosis, which occurs in near one third of patients (Batur et al., 2003). This association, known as Heyde's syndrome, is notably characterized by a coagulopathy, i.e., acquired von Willebrand disease (Vincentelli et al., 2003), due to shear-stress mediated consumption of high molecular weight multimers of the von Willebrand factor (Loscalzo, 2012). Being the latters the most hemostatically competent form of von Willebrand factor, this favors a vicious circle that aggravates the bleeding from GI angiodysplasia and the consequent IDA.

From a laboratory point of view, an accurate diagnosis of IDA in elderly is also challenging because of the high prevalence of concomitant chronic diseases that complicate the interpretation of traditional biomarkers. The red blood cells mean corpuscular volume (MCV) is often a starting index in the evaluation of a patient with anemia, being typically reduced in IDA. However, MCV reduction in elderly is often lacking in early stages and/or blunted by other concomitant nutritional deficiencies, such as those of folic acid or vitamin B12. Similarly, the other common laboratory markers of "absolute" ID, i.e., low serum ferritin and transferrin saturation, and raised transferrin, have a low sensitivity in elderly (Fairweather-Tait et al., 2013). For example, the classical cut-off value of serum ferritin ≤12–15 μg/L, which commonly defines ID in younger adults (Lipschitz et al., 1974; Ali et al., 1978; Carmel, 2008), has been claimed as too stringent in elderly patients. Indeed, in these subjects true ID often occurs at higher ferritin values, since ferritin *per se* raises with aging (Casale et al., 1981), and is an acute-phase reactant that increase during inflammation, infection, malignancy, and other illnesses common in

older people. In a study carried out in hospitalized older patients, a serum ferritin level <50 μg/L resulted more reliable in predicting ID than other traditional cut-off values (Joosten et al., 1991). The low sensitivity of traditional iron biomarkers is demonstrated also by the fact that elderly anemic patients sometimes respond to iron supplementation even if their iron indices at baseline are not abnormal (Price et al., 2011).

Soluble transferrin receptor (sTfR), derived from proteolysis of the membrane transferrin receptor (TR), reflects erythropoietic activity and inversely correlates with the amount of iron available for erythropoiesis. In the past, some evidence supported sTfR measurement as a novel marker of ID in older people, considering that its levels do not increase with age and are not affected by the presence of inflammation (Mast et al., 1998). In particular, the serum sTfR divided by the log of the ferritin (sTfR-ferritin index) was found useful in classifying patients with ACD and concomitant IDA (Punnonen et al., 1997; Rimon et al., 2002). However, currently the lack of standardized reagents for the sTfR assay complicates interpretation of the sTfR-ferritin index in different studies, and limits its use in clinical practice (Pfeiffer et al., 2007).

In the last decade, hepcidin has been suggested as a promising diagnostic marker for iron-related disorders (Goodnough et al., 2010; Kroot et al., 2011). In IDA serum and urinary hepcidin levels are typically reduced and frequently undetectable by currently available assays (Bozzini et al., 2008; Ganz et al., 2008; Castagna et al., 2010). Hepcidin suppression appears also a sensitive indicator of ID without anemia, since decreased levels have been observed prior to a detectable decrease in Hb or hematocrit (Ganz et al., 2008; Pasricha et al., 2011). As mentioned above, hepcidin is induced by inflammatory cytokines and contributes to the pathogenesis of the so-called anemia of chronic disease (ACD), which is characterized by impaired iron utilization, along with inadequate EPO production, and cytokine-induced inhibition of erythroid precursors (Weiss and Goodnough, 2005). The opposite trend of hepcidin in IDA versus ACD has theoretically the potential to differentiate these conditions, both highly prevalent in elderly, and not infrequently coexisting. Of note, preclinical studies have shown that concomitant ID tends to blunt the hepcidin response to pro-inflammatory cytokines (Theurl et al., 2009; Darshan et al., 2010), suggesting the possibility to distinguish in the individual anemic patient the presence of IDA or mixed IDA/ACD (both with low to undetectable hepcidin levels) from ACD alone (with high hepcidin levels). Preliminary data in patients with rheumatoid arthritis (van Santen et al., 2011) or inflammatory bowel disease (Bergamaschi et al., 2013) are consistent with this possibility, but larger data are required for confirmation, particularly in elderly subjects where the distinction between IDA and ACD is expected to be particularly challenging.

## **TREATMENT OF IDA IN ELDERLY**

Currently, no specific guidelines exist for the management of anemia in the elderly. A recent review recommends that iron status should be checked at first in every elderly patient (Goodnough and Schrier, 2014). Once IDA is clearly ascertained or deemed likely (because of ambiguous results of iron markers as discussed above) a therapeutic trial with oral iron should be prescribed, with the aim of correcting both anemia and iron stores. This first-line approach, preferably using divalent compounds like ferrous sulfate or gluconate because of their superior bioavailability (Clark, 2009), is usually considered to be safer for the patient and with a better cost-effective ratio as compared to parenteral iron administration.

In general, Hb levels are expected to rise by approximately 1–2 g/dL every couple of weeks after starting oral iron therapy (Clark, 2008), which should be continued for 3 months after correction of anemia to replenish iron stores. The time needed may be even longer in elderly patients, because of slower bone marrow response. This translates in poor adherence, particularly when concomitant multimorbidity requires the assumption of a huge number of pills per day. Moreover, oral iron supplementation is often poorly tolerated in elderly patients, particularly because abdominal discomfort, as well as poorly absorbed because of the relative high prevalence of malabsorptive conditions (see above).

Thus, intravenous (IV) iron replacement is often required in elderly patients with IDA (Silverstein and Rodgers, 2004; Clark, 2009; Pasricha et al., 2010). Most IV iron formulations are generally effective, well tolerated, and with a lower incidence of serious adverse reaction (e.g., anaphylaxis) than commonly thought by many clinicians (Fishbane, 2003; Auerbach and Ballard, 2010; Auerbach et al., 2013a). Until a few years ago, the most widely used formulations in Europe were iron gluconate or iron sucrose, which are relatively unstable compounds with limited maximal doses per single infusion, i.e., 125 and 200 mg, respectively. Since the mean total iron dose generally required to correct anemia and restore iron is 1000–1500 mg, multiple hospital accesses for repeated infusions are needed. In elderly patients with limited autonomy, this not only increases direct and indirect social costs, but also can substantially hamper the feasibility of the treatment.

Recently, the pharmaceutical industry has made significant progresses through the production of more stable iron compounds that can be safely administered at high doses per single infusion, i.e., 1000–1500 mg, thus allowing a single treatment episode (for a review see Auerbach et al., 2013a). These includes low molecular weight iron dextran (Auerbach et al., 2011), ferumoxytol (Auerbach et al., 2013b), iron isomaltoside (Wikstrom et al., 2011), and ferric carboxymaltose (Evstatiev et al., 2011; Onken et al., 2014). These preparations significantly simplify the IV iron therapy, an effect that is expected to be particularly useful in elderly patients with IDA. However, specific trials in this setting are needed to confirm these exciting promises. Moreover, IV iron has theoretically the potential to generate oxidative stress and inflammatory cytokine production, and to increase susceptibility to certain infections, so that its long-term effects also need further future studies (Auerbach and Ballard, 2010; Goodnough et al., 2010).

Since hepcidin inhibits duodenal iron absorption, it has been suggested that measuring hormone levels may help to determine *a priori* the best way of iron administration, oral, or IV, in a given individual. Indeed, in a recent retrospective study, hepcidin levels were proven useful in identifying IDA patients who did not respond to oral iron supplementation (Bregman et al., 2013). Of note, the cut-off value of serum hepcidin that discriminated between responders (R) and not responders (NR) to oral iron was not particularly high (20 ng/ml), and within the "normal" range for the method employed (Bregman et al., 2013). The positive predictive value of NR for hepcidin levels >20 ng/ml was 81.4%. On the contrary, the majority of R patients had very low to completely suppressed (undetectable) hepcidin levels, as is the general rule for IDA patients (see above). If these data will be confirmed by further studies, hepcidin assay could actually help to individualize iron therapy, by avoiding waste of time with a poorly tolerated

oral therapy when hormone levels are high or "pseudonormal," particularly in elderly patients. In a near future scenario, a possible algorithm for the diagnosis and treatment of IDA in elderly, based on most recent pathophysiological and therapeutic advances in the field, is depicted in **Figure 3**.

## **CONCLUSION**

Iron deficiency is a major cause of anemia in elderly patients, and should be always searched as first diagnostic step. IDA in

elderly is not infrequently multifactorial and difficult to diagnose since traditional biochemical markers of iron status are relatively ambiguous in this age group. Reasoning in terms of multimorbidity is central for a correct approach to IDA in elderly. Being GI blood loss the major cause of IDA in elderly, a correct identification of the bleeding source can be lifesaving, particularly when an occult malignancy is underlying. The discovery of hepcidin has revolutionized our understanding of iron metabolism, and measuring hormone levels may help both in diagnosis and in selecting the best treatment option, particularly as the IV therapy is becoming simpler than ever with the availability of new iron compounds.

## **AUTHOR CONTRIBUTIONS**

Fabiana Busti and Natascia Campostrini co-wrote the paper. Nicola Martinelli analyzed data. Domenico Girelli designed the work and co-wrote the paper. All authors have approved the final manuscript.

#### **ACKNOWLEDGMENTS**

This work was partially supported by the Italian Ministry of University and Research (grant no. 200989KXFN) and by Fondazione Cariverona, project Verona Nanomedicine Initiative to Domenico Girelli.

## **REFERENCES**


in chronic kidney disease. *J. Nephrol.* 24, 589–596. doi: 10.5301/JN.2011. 6248


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

*Received: 28 February 2014; paper pending published: 20 March 2014; accepted: 04 April 2014; published online: 23 April 2014.*

*Citation: Busti F, Campostrini N, Martinelli N and Girelli D (2014) Iron deficiency in the elderly population, revisited in the hepcidin era. Front. Pharmacol. 5:83. doi: 10.3389/fphar.2014.00083*

*This article was submitted to Drug Metabolism and Transport, a section of the journal Frontiers in Pharmacology.*

*Copyright © 2014 Busti, Campostrini, Martinelli and Girelli. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*

## Hemojuvelin and bone morphogenetic protein (BMP) signaling in iron homeostasis

## *Amanda B. Core , Susanna Canali and Jodie L. Babitt\**

*Division of Nephrology, Program in Membrane Biology, Center for Systems Biology, Massachusetts General Hospital, Harvard Medical School, Program in Anemia Signaling Research, Boston, MA, USA*

#### *Edited by:*

*Paolo Arosio, University of Brescia, Italy*

#### *Reviewed by:*

*Silvia Gazzin, Italian Liver Foundation, Italy Kostas Pantopoulos, Lady Davis Institute for Medical Research, Canada Olivier Loréal, Institut National de la Santé et de la Recherche Médicale - UMR 991, France*

#### *\*Correspondence:*

*Jodie L. Babitt, Massachusetts General Hospital, 185 Cambridge St., CPZN-8208, Boston, MA 02114, USA e-mail: babitt.jodie@ mgh.harvard.edu*

## Mutations in hemojuvelin (HJV) are the most common cause of the juvenile-onset form of the iron overload disorder hereditary hemochromatosis. The discovery that HJV functions as a co-receptor for the bone morphogenetic protein (BMP) family of signaling molecules helped to identify this signaling pathway as a central regulator of the key iron hormone hepcidin in the control of systemic iron homeostasis. This review highlights recent work uncovering the mechanism of action of HJV and the BMP-SMAD signaling pathway in regulating hepcidin expression in the liver, as well as additional studies investigating possible extra-hepatic functions of HJV. This review also explores the interaction between HJV, the BMP-SMAD signaling pathway and other regulators of hepcidin expression in systemic iron balance.

**Keywords: hemojuvelin, bone morphogenetic protein, hepcidin, iron, hemochromatosis, repulsive guidance molecule**

## **JUVENILE HEMOCHROMATOSIS IS CAUSED BY MUTATIONS IN THE GENES ENCODING HEPCIDIN OR HEMOJUVELIN**

Juvenile Hemochromatosis (JH) is an autosomal recessive disorder caused by a failure to prevent excess iron entry into the bloodstream, and characterized by progressive tissue iron overload (Pietrangelo, 2010). Although iron's redox properties are critical for its role in many fundamental biological processes from cellular respiration to oxygen transport, iron excess can lead to toxic free radical generation. If left untreated, JH patients develop multiorgan dysfunction as a consequence of iron overload, including cirrhosis, cardiomyopathy, diabetes mellitus, and hypogonadotrophic hypogonadism, before the age of 30 (Pietrangelo, 2010).

The identification of hepcidin as a master regulator of systemic iron balance was a major advance in understanding the pathophysiology of JH (Ganz, 2013). A defensin-like peptide produced predominantly by hepatocytes, hepcidin controls iron entry into the bloodstream from dietary sources, recycled red blood cells, and body storage sites by inducing degradation of the iron exporter ferroportin (Ganz, 2013). Hepcidin expression is stimulated by iron and inflammation to limit iron availability, while hepcidin is inhibited by iron deficiency, anemia, and hypoxia to increase iron availability for erythropoiesis (Babitt and Lin, 2010; Ganz, 2013). Hepcidin deficiency is the common pathogenic mechanism underlying both adult and juvenile-onset hemochromatosis and contributes to the pathogenesis of iron loading anemias such as thalassemia, while its overproduction causes anemia of inflammation and iron refractory iron deficiency anemia (IRIDA) (Ganz, 2013). JH is caused by mutations in the gene encoding hepcidin itself (*HAMP*) or, more commonly, hemojuvelin (*HJV*, also known as *HFE2* or *RGMC*) (Roetto et al., 2003; Papanikolaou et al., 2004).

*HJV* encodes a glycophosphatidylinositol (GPI)-linked membrane protein that is a member of the repulsive guidance molecule (RGM) family (Monnier et al., 2002; Samad et al., 2004). Currently, there are 43 identified *HJV* mutations that cause JH, with G320V being the most frequent (**Table 1**). HJV is expressed in the liver, and JH patients with *HJV* mutations and *Hjv* knockout mice exhibit significantly reduced hepatic hepcidin expression, thereby implicating HJV in the regulation of hepcidin synthesis (Papanikolaou et al., 2004; Huang et al., 2005; Niederkofler et al., 2005).

## **BMP-SMAD SIGNALING VIA HJV IS A CENTRAL REGULATOR OF HEPCIDIN**

A breakthrough in understanding the mechanism of action of HJV in hepcidin regulation came when HJV was discovered to function as a co-receptor for the bone morphogenetic protein (BMP) signaling pathway (Babitt et al., 2006), analogous to its RGM family homologs (Babitt et al., 2005; Samad et al., 2005). Importantly, this BMP signaling function of HJV was demonstrated to be crucial for its role in regulating hepcidin expression (Babitt et al., 2006) (**Figure 1**).

BMPs belong to the Transforming Growth Factor-beta (TGFβ) superfamily of ligands (Shi and Massagué, 2003). In the canonical signaling pathway, BMP ligands bind to type I and type II serine threonine kinase receptors to induce phosphorylation of cytoplasmic SMAD1, SMAD5, and SMAD8 proteins. These SMAD proteins form a complex with SMAD4 and translocate to

### **Table 1 | Mutations of the** *HJV* **gene linked to JH.**


the nucleus to regulate gene transcription. This signaling pathway is further regulated at multiple levels in order to generate a precise signal in a specific cellular context (Shi and Massagué, 2003).

HJV and other RGM family members function as BMP coreceptors that bind selectively to BMP ligands and receptors to enhance SMAD phosphorylation in response to BMP signals (Babitt et al., 2005, 2006; Samad et al., 2005). All RGMs

share the ability to bind to the BMP2/BMP4 subfamily and enhance BMP2/BMP4 signaling (Babitt et al., 2005, 2006; Samad et al., 2005; Wu et al., 2012). Moreover, all RGMs utilize BMP type I receptors ALK2, ALK3, and ALK6, and allow preferential signaling through the BMP type II receptor ACTRIIA (Xia et al., 2007, 2008, 2010). However, HJV is unique from other RGMs in that it exhibits preferential ability to bind to the BMP5/BMP6/BMP7 subfamily compared with RGMA and RGMB (Wu et al., 2012).

The BMP-HJV-SMAD signaling pathway activates hepcidin transcription directly through specific BMP-responsive elements (BMP-REs) on the hepcidin promoter (Casanovas et al., 2009; Truksa et al., 2009a). A mutation in the proximal BMP-RE was associated with a more severe iron overload phenotype in a patient with classical *HFE* hemochromatosis, demonstrating its importance in hepcidin regulation in humans (Island et al., 2009). In mice, liver-specific disruption of *Smad4*, the BMP receptors type I *Alk2* or *Alk3*, or the ligand *Bmp6* result in hepcidin deficiency and iron overload, supporting the important role of these specific BMP-SMAD pathway components, in conjunction with HJV, in hepcidin regulation *in vivo* (Wang et al., 2005; Andriopoulos et al., 2009; Meynard et al., 2009; Steinbicker et al., 2011a).

## **SOLUBLE HJV**

In addition to the GPI-anchored membrane form of HJV, endogenous soluble HJV (sHJV) protein is detectable in human and rodent serum. (Lin et al., 2005; Zhang et al., 2007; Chen et al., 2013). Multiple mechanisms have been proposed for endogenous sHJV generation, including cleavage by the pro-protein convertase furin and the type II transmembrane serine protease TMPRSS6 (Kuninger et al., 2008; Lin et al., 2008; Silvestri et al., 2008a,b). Whereas membrane HJV is a co-receptor for the BMP signaling complex (Babitt et al., 2006), sHJV can antagonize BMP signaling, presumably by binding and sequestering BMP ligands from interacting with cell-surface BMP type I and type II receptors (Babitt et al., 2007) (**Figure 1**). Indeed, the relative binding affinity of HJV for various BMP ligands roughly correlated with the ability of sHJV to inhibit their biological activity (Babitt et al., 2007; Wu et al., 2012).

Although exogenous sHJV inhibits BMP-SMAD signaling, the source, amount, and physiologic role(s) of endogenously produced sHJV *in vivo* are not well-understood. There is some evidence suggesting that endogenous sHJV is increased by iron deficiency and reduced by iron loading (Lin et al., 2005; Zhang et al., 2007; Silvestri et al., 2008a; Brasse-Lagnel et al., 2010; Chen et al., 2013). Interestingly, the furin cleaved form of sHJV appears to be more potent to inhibit BMP signaling and hepcidin compared with the TMPRSS6-cleaved form (Maxson et al., 2010). Whether HJV cleavage mainly represents a mechanism to remove the activating effects of liver membrane HJV, or whether endogenous sHJV has a direct BMP-SMAD inhibiting effect remains uncertain.

## **EXTRA-HEPATIC FUNCTIONS OF HJV**

In addition to the liver, *HJV* mRNA is also highly expressed in skeletal muscle and heart (Niederkofler et al., 2004; Papanikolaou et al., 2004), and has been detected in other tissues (Rodriguez Martinez et al., 2004, Rodriguez et al., 2007; Gnana-Prakasam et al., 2009; Luciani et al., 2011). Tissue specific differences in HJV mRNA regulation and HJV protein glycosylation patterns have also been described (Niederkofler et al., 2005; Fujikura et al., 2011). It was previously hypothesized that skeletal muscle and/or heart could serve as a source of sHJV to suppress hepcidin synthesis in response to iron deficiency or hypoxia (Lin et al., 2005; Zhang et al., 2005). However, mice with a specific knockout of *Hjv* in skeletal ± cardiac muscle do not have altered hepcidin expression or systemic iron balance, at least under basal conditions or with dietary iron changes (Chen et al., 2011; Gkouvatsos et al., 2011). Whether strenuous exercise or hypoxia may uncover a role for muscle hemojuvelin remains uncertain. In contrast, hepatocyte specific *Hjv* knockout mice exhibit an iron overload phenotype similar to global *Hjv* knockout mice (Chen et al., 2011; Gkouvatsos et al., 2011). Thus, hepatic expression of HJV appears to have the most important physiologic role in systemic iron homeostasis regulation *in vivo*.

## **IRON STIMULATES BMP-SMAD SIGNALING TO REGULATE HEPCIDIN**

Iron regulates the activity of the BMP6-SMAD pathway to modulate hepcidin expression. Both circulating and liver iron appear to stimulate this pathway through different mechanisms (Ramos et al., 2011; Corradini et al., 2011a). In mice, liver iron content is positively correlated with liver *Bmp6* mRNA levels and overall activity of the Smad signaling pathway (Kautz et al., 2008; Corradini et al., 2011a). Moreover, hepcidin induction by iron is inhibited by a neutralizing BMP6 antibody (Corradini et al., 2011a). These data suggest that liver iron modulates BMP6-SMAD signaling and hepcidin expression at least in part by regulating expression of *BMP6* mRNA (**Figure 1**). It appears that liver iron regulates BMP6 expression mainly in nonparenchymal cells (Enns et al., 2013), and that iron loading in specific liver cell types may important for this regulation (Daba et al., 2013). However, the mechanism by which hepatic iron levels regulate BMP6 remains unknown. Notably, hepcidin is still increased to a lesser extent by chronic iron loading in *Bmp6* and *Hjv* knockout mice, suggesting that these pathways do not completely account for hepcidin regulation by chronic iron loading (Ramos et al., 2011; Gkouvatsos et al., 2014).

Increases in circulating iron stimulate SMAD1/5/8 phosphorylation and hepcidin expression without affecting *Bmp6* mRNA levels (Corradini et al., 2011a). How circulating iron activates SMAD1/5/8 phosphorylation is unknown, but may involve an interaction with other proteins that are mutated in adult-onset hereditary hemochromatosis (see section HFE and TFR2). HJV liver membrane protein expression itself does not appear to be regulated by iron (Krijt et al., 2012).

Iron administration and BMP6-SMAD signaling also upregulate inhibitory SMAD7 and SMAD6, and TMPRSS6 (see section TMPRSS6), that can act as feedback inhibitors of BMP-SMAD signaling and hepcidin expression (Kautz et al., 2008; Mleczko-Sanecka et al., 2010; Meynard et al., 2011; Corradini et al., 2011a; Vujic Spasi ´ c et al., 2013 ´ ). It has been hypothesized that these pathways may help prevent excessive hepcidin increases by iron to provide tight homeostatic control (Meynard et al., 2011; Corradini et al., 2011a).

## **INTERACTION OF HJV AND THE BMP-SMAD SIGNALING PATHWAY WITH OTHER HEPCIDIN REGULATORS HFE AND TFR2**

Adult-onset hereditary hemochromatosis is a less severe ironoverload disorder that manifests later in life compared with JH, and is associated with mutations in *HFE* or *TFR2* (encoding transferrin receptor 2) (Pietrangelo, 2010). Liver expression of HFE and TFR2 are clearly important for iron homeostasis regulation because mice with a hepatocyte-specific knockout of either gene have a similar iron-overload phenotype compared with global *Hfe* or *Tfr2* knockout mice (Wallace et al., 2007; Vujic Spasi ´ c et al., 2008 ´ ). Moreover, liver transplantation corrects much of the *HFE* hemochromatosis phenotype (Garuti et al., 2010; Bardou-Jacquet et al., 2014). Liver hepcidin expression is inappropriately low in mice and humans with *HFE* or *TFR2* mutations, suggesting that both HFE and TFR2 positively regulate liver hepcidin expression (Ahmad et al., 2002; Fleming et al., 2002; Bridle et al., 2003; Muckenthaler et al., 2003; Kawabata et al., 2005; Nemeth et al., 2005; Piperno et al., 2007). HFE and TFR2 are also postulated to function in iron sensing by the liver. The current working model is that when iron-bound transferrin increases in circulation, it binds to transferrin receptor 1 (TFR1) and displaces HFE, which then signals by some mechanism to stimulate hepcidin expression, possibly through an interaction with TFR2 (Schmidt et al., 2008; Gao et al., 2009).

It has been proposed that HFE and TFR2 may form a "supercomplex" with HJV to stimulate hepcidin expression via the BMP-SMAD pathway. Studies supporting this model have demonstrated that liver BMP-SMAD signaling is impaired in mice and humans with *HFE* and/or *TFR2* mutations, suggesting an interaction at some level between HFE, TFR2 and the BMP-SMAD pathway (Corradini et al., 2009, 2011b; Kautz et al., 2009; Wallace et al., 2009; Bolondi et al., 2010; Ryan et al., 2010). Recently, it was published in an overexpression tissue culture system using tagged proteins that HFE and TFR2 can form a complex with HJV (D'Alessio et al., 2012). However, it is not been shown whether these proteins endogenously interact *in vivo*. Moreover, the more severe iron overload phenotype of *HJV* mutations and combined *HFE/TFR2* mutations compared with either *HFE* or *TFR2* mutations alone suggest that the function of these proteins is not entirely overlapping (Pietrangelo et al., 2005; Wallace et al., 2009). Thus, while it appears that HFE and TFR2 interact at some level with the BMP-HJV-SMAD pathway to regulate liver hepcidin expression (**Figure 1**), the precise molecular mechanisms of how HFE and TFR2 contribute to hepcidin regulation remain an active area of investigation.

## **THE INFLAMMATORY PATHWAY**

In addition to iron, inflammatory stimuli also induce hepcidin expression (Ganz, 2013). The most well-characterized pathway is through IL6 activating the Janus kinase JAK2 to phosphorylate STAT3, which then activates the hepcidin promoter directly via a STAT3-binding motif (Wrighting and Andrews, 2006; Pietrangelo et al., 2007; Verga Falzacappa et al., 2007).

Although inflammation downregulates liver *Hjv* mRNA expression (Krijt et al., 2004; Niederkofler et al., 2005; Constante et al., 2007), liver SMAD1/5/8 signaling is often activated in the context of inflammation (Theurl et al., 2011) and is essential for hepcidin regulation by inflammation. Indeed, blocking BMP signaling with a small molecule BMP type I receptor inhibitor or a sHJV recombinant protein inhibits IL6-induced hepcidin expression in cell culture (Babitt et al., 2007; Yu et al., 2008). Moreover, mice with a hepatocyte-specific knockout of *Smad4* exhibit blunted hepcidin response to IL6 treatment (Wang et al., 2005). Importantly, BMP pathway inhibitors lower hepcidin, increase iron availability for erythropoiesis, and ameliorate anemia in animal models of anemia of inflammation (Theurl et al., 2011; Steinbicker et al., 2011b; Sun et al., 2013).

At least two mechanisms are proposed to account for the crosstalk between the BMP-SMAD and IL6-STAT3 pathways in hepcidin regulation. First, there may be an interaction at the level of the hepcidin promoter, where the proximal BMP-RE and the STAT3 binding site are in close proximity (**Figure 1**). In support of this hypothesis, mutation of the proximal BMP-RE impairs hepcidin promoter activation not only by BMPs, but also by IL6 (Casanovas et al., 2009). Second, inflammation induces hepatic expression of another TGF-β superfamily member, Activin B, which can stimulate hepcidin expression by activating SMAD1/5/8 signaling in hepatoma-derived cell cultures (Besson-Fournier et al., 2012) (**Figure 1**). Whether Activin B contributes to hepcidin regulation by inflammation *in vivo* remains to be determined.

## **TMPRSS6**

The serine protease TMPRSS6 has been implicated in hepcidin inhibition by iron deficiency. Mutations in *TMPRSS6* are linked to IRIDA associated with inappropriately high hepcidin levels (Du et al., 2008; Finberg et al., 2008; Folgueras et al., 2008). Moreover, genome-wide association studies have linked common single nucleotide polymorphisms in *TMRPSS6* to iron status and hemoglobin level, supporting an important role for TMPRSS6 in regulating systemic iron homeostasis and normal erythropoiesis (Benyamin et al., 2009; Chambers et al., 2009; Tanaka et al., 2010). TMPRSS6 is proposed to regulate hepcidin expression through an interaction with HJV and the BMP-SMAD pathway in the liver. Specifically, when both proteins are overexpressed in cell culture, TMPRSS6 binds and cleaves HJV to generate sHJV, thereby inhibiting BMP-SMAD signaling (Silvestri et al., 2008b) (**Figure 1**). In mouse models, the combined deficiency of *Hjv* or *Bmp6* and *Tmprss6* causes iron overload, suggesting that there is a genetic interaction between TMPRSS6 and the BMP6-HJV-SMAD pathway (Truksa et al., 2009b; Finberg et al., 2010; Lenoir et al., 2011). Interestingly, liver membrane expression of Hjv is decreased (Krijt et al., 2011), and serum sHjv levels are unchanged (Chen et al., 2013), in *Tmprss6* knockout mice compared with wildtype mice, which seem contrary to the proposed hypothesis that TMPRSS6 acts to cleave HJV from the liver membrane surface. Future work is needed to fully understand the mechanism of action of TMPRSS6 in hepcidin regulation and iron homeostasis *in vivo.*

## **NEOGENIN**

In addition to TMPRSS6, the deleted in colorectal cancer (DCC) family member neogenin is also proposed to function as an HJV interacting protein that modifies BMP-SMAD signaling and iron homeostasis (**Figure 1**). In particular, neogenin binds to HJV, like other RGM family members (Matsunaga et al., 2004; Zhang et al., 2005; Conrad et al., 2010). Moreover, neogenin mutant mice exhibit reduced hepcidin levels and iron overload consistent with a role for neogenin in regulating hepcidin and systemic iron balance *in vivo* (Lee et al., 2010). However, the mechanism of action of neogenin in hepcidin and iron homeostasis regulation is still not fully understood. In some studies, neogenin increased HJV cleavage (Enns et al., 2012), while in other studies, neogenin reduced HJV secretion (Lee et al., 2010). Moreover, neogenin was variably shown to inhibit (Hagihara et al., 2011), have no effect (Xia et al., 2008), or stimulate BMP signaling (Lee et al., 2010). Whether neogenin and HJV interact in a cell autonomous or cell non-autonomous manner *in vivo* remains unclear, and how this interaction occurs may be important for downstream functional effects.

## **OTHER PATHWAYS**

Hepcidin suppression by erythropoietic drive appears to be mediated by secreted factor(s) released by proliferating red blood cell precursors in the bone marrow (Pak et al., 2006; Vokurka et al., 2006). Two proposed erythroid hepcidin regulators are the TGFβ/BMP superfamily modulators growth and differentiation factor 15 (GDF15) and twisted gastrulation 1 (TWSG1), at least in the context of ineffective erythropoiesis in iron loading anemias (Tanno et al., 2007, 2009) (**Figure 1**). The role of GDF15 and TWSG1 in hepcidin suppression by erythropoietic drive in other contexts has been questioned (Ashby et al., 2010; Casanovas et al., 2013). Recently, erythroferrone has been proposed as a novel erythroid regulator (Kautz et al., 2013), but its mechanism of action is not yet reported.

A number of other hormones, growth factors and signaling pathways have recently been implicated in hepcidin regulation including testosterone, estrogen, hepatocyte growth factor (HGF), epidermal growth factor (EGF), endoplasmic reticulum stress, gluconeogenic signals and the Ras/RAF and mTOR signaling pathways (Oliveira et al., 2009; Vecchi et al., 2009, 2014; Goodnough et al., 2012; Hou et al., 2012; Yang et al., 2012; Guo et al., 2013; Latour et al., 2014; Mleczko-Sanecka et al., 2014). Notably, the majority of these pathways appear to regulate hepcidin through an intersection with the BMP-SMAD pathway at some level (Goodnough et al., 2012; Guo et al., 2013; Latour et al., 2014; Mleczko-Sanecka et al., 2014) (**Figure 1**).

## **CONCLUSION**

Understanding the genetic basis for JH has yielded important insights into the molecular mechanisms of systemic iron homeostasis. Hepcidin and its receptor ferroportin are key regulators of body iron balance, and the BMP-SMAD pathway via the co-receptor HJV is a central regulator of hepcidin production (**Figure 1**). Knowledge of these pathways has already lead to the development of novel therapeutic strategies that target the molecular mechanisms underlying iron homeostasis disorders, with several new treatments currently being evaluated in human clinical trials (Fung and Nemeth, 2013). Future work will be needed to fully understand the mechanisms by which iron levels are sensed by the liver and integrated with other pathways to regulate BMP-SMAD signaling, hepcidin expression, and systemic iron homeostasis.

## **ACKNOWLEDGMENTS**

Amanda B. Core was supported by NIH grant 5T32DK007540- 28. Jodie L. Babitt was supported in part by NIH grant RO1- DK087727 and a Howard Goodman Fellowship Awards from the Massachusetts General Hospital.

## **REFERENCES**


systemic iron homeostasis. *Blood* 115, 3817–3826. doi: 10.1182/blood-2009-05- 224808


hemojuvelin pr176c missense mutation. *Haematologica* 9, 1262–1263. doi: 10.3324/haematol.11247


mice lacking both hemojuvelin and matriptase-2/TMPRSS6. *Br. J. Haematol.* 147, 571–581 doi: 10.1111/j.1365-2141.2009.07873.x


**Conflict of Interest Statement:** Jodie L. Babitt has ownership interest in a startup company FerruMax Pharmaceuticals, which has licensed technology from the Massachusetts General Hospital based on the work cited here and in prior publications. All other authors declare the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

*Received: 28 February 2014; accepted: 21 April 2014; published online: 13 May 2014. Citation: Core AB, Canali S and Babitt JL (2014) Hemojuvelin and bone morphogenetic protein (BMP) signaling in iron homeostasis. Front. Pharmacol. 5:104. doi: 10.3389/fphar.2014.00104*

*This article was submitted to Drug Metabolism and Transport, a section of the journal Frontiers in Pharmacology.*

*Copyright © 2014 Core, Canali and Babitt. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*

## Hepcidin antagonists for potential treatments of disorders with hepcidin excess

## *Maura Poli, Michela Asperti, Paola Ruzzenenti, Maria Regoni and Paolo Arosio\**

Molecular Biology Laboratory, Department of Molecular and Translational Medicine, University of Brescia, Brescia, Italy

#### *Edited by:*

Raffaella Gozzelino, Instituto Gulbenkian de Ciência, Portugal

#### *Reviewed by:*

Domenico Girelli, University of Verona, Italy Zvi Ioav Cabantchik, Hebrew University of Jerusalem, Israel

#### *\*Correspondence:*

Paolo Arosio, Molecular Biology Laboratory, Department of Molecular and Translational Medicine, University of Brescia, Viale Europa 11, 25123 Brescia, Italy e-mail: arosio@med.unibs.it

The discovery of hepcidin clarified the basic mechanism of the control of systemic iron homeostasis. Hepcidin is mainly produced by the liver as a propeptide and processed by furin into the mature active peptide. Hepcidin binds ferroportin, the only cellular iron exporter, causing the internalization and degradation of both. Thus hepcidin blocks iron export from the key cells for dietary iron absorption (enterocytes), recycling of hemoglobin iron (the macrophages) and the release of storage iron from hepatocytes, resulting in the reduction of systemic iron availability. The BMP/HJV/SMAD pathway is the major regulator of hepcidin expression that responds to iron status. Also inflammation stimulates hepcidin via the IL6/STAT3 pathway with a support of an active BMP/HJV/SMAD pathway. In some pathological conditions hepcidin level is inadequately elevated and reduces iron availability in the body, resulting in anemia. These conditions occur in the genetic iron refractory iron deficiency anemia and the common anemia of chronic disease (ACD) or anemia of inflammation. Currently, there is no definite treatment for ACD. Erythropoiesis-stimulating agents and intravenous iron have been proposed in some cases but they are scarcely effective and may have adverse effects. Alternative approaches aimed to a pharmacological control of hepcidin expression have been attempted, targeting different regulatory steps. They include hepcidin sequestering agents (antibodies, anticalins, and aptamers), inhibitors of BMP/SMAD or of IL6/STAT3 pathway or of hepcidin transduction (siRNA/shRNA) or ferroportin stabilizers. In this review we summarized the biochemical interactions of the proteins involved in the BMP/HJV/SMAD pathway and its natural inhibitors, the murine and rat models with high hepcidin levels currently available and finally the progresses in the development of hepcidin antagonists, with particular attention to the role of heparins and heparin sulfate proteoglycans in hepcidin expression and modulation of the BMP6/SMAD pathway.

**Keywords: hepcidin, heparin, anemia of chronic diseases, inflammation, iron metabolism**

## **HEPCIDIN DISCOVERY AND PROPERTIES**

Hepcidin was independently discovered in the years 2000–2001 by various groups. Krause et al. (2000) isolated from human blood ultrafiltrate a 25-residue peptide with antimicrobial activity that they named LEAP-1 (liver-expressed antimicrobial peptide 1). In the same period Pigeon et al. (2001) identified an iron-regulated gene that encoded for the LEAP-1 with high expression in the liver, and much lower expression in the kidney, adipose tissue, heart, and brain. Park et al. (2001) characterized a cysteine-rich peptide and named it hepcidin (hepatic bactericidal protein) for its hepatic origin with a structure typical for an antimicrobial activity. They found homologous cDNAs in the liver of various species from fish to human. A central role of hepcidin in systemic iron homeostasis was soon unambiguously recognized by the finding that inactivation of its gene was associated with severe iron overload in the liver and pancreas (Nicolas et al., 2001). The finding was serendipitous, since the knockout construct was aimed at deleting USF2 gene (upstream stimulatory factor 2) but it also removed the hepcidin adjacent genes. A following specific USF2 knockout mouse had normal hepcidin and iron, while the specific inactivation of hepcidin gene caused iron overload (Nicolas et al., 2002). The importance of hepcidin was conclusively demonstrated by the finding that transgenic mice overexpressing the peptide showed a severe and often lethal anemia (Nicolas et al., 2003). The initial excitement about hepcidin as a central player in the communication of body iron stores to the intestinal absorptive cells (Fleming and Sly, 2001) was further sustained by the finding that patients with homozygous mutations in the hepcidin gene were affected by severe juvenile hemochromatosis (Roetto et al., 2003). The following years were dedicated to the characterization of the gene, its product and the regulation of its expression.

In human the hepcidin gene is in chromosome 19 and encodes a precursor prepropeptide of 84 amino acids that is processed by two sequential cleavages. The first of the signal sequence and the second of the pro-region to produce the mature peptide of 25 amino acids (aa). Furin, a major member of the family of prohormone convertases, is the enzyme involved

**Abbreviations:** BMP, bone morphogenetic protein; SMAD, sons of mothers against decapentaplegic; STAT, signal transducer and activator of transcription; IL-6, interleukin-6; LPS, lipopolysaccharides; GAG, glycosaminoglycan.

in the processing, and it recognizes the consensus sequence (QRRRRR↓DTHF) conserved in mammal and fish hepcidins (Shike et al., 2004; Valore and Ganz, 2008; **Figure 1A**). Chemical or siRNA-mediated inhibition of furin prevents hepcidin maturation but not its secretion from the cell (Valore and Ganz, 2008). The processing may be more complex, since two additional hepcidin N-terminal truncated forms of 22 and 20 aa were originally described (**Figure 1A**). Hepc-25 and hepc-20 are formed intracellularly and possibly processed in the Golgi apparatus by furin-like proteases and both secreted in the blood. Whereas hepc-22 is present only in urine (Park et al., 2001). The role of hepc-20 and hepc-22 is not clear. The NMR structure of the mature hepcidin 25 was obtained from the refolded synthetic peptide (Jordan et al., 2009). It consists of two short β-strands stabilized by interstrand disulfide bonds (**Figure 1B**). It has amphipathic properties with cationic charges and hydrophobic surface like most of antimicrobial peptides (**Figure 1C**), however, the antimicrobial activity of human hepcidin is low and probably not critical. Mouse has two different hepcidin genes (hepc-1 and hepc-2) but only hepc-1 is related with iron metabolism.

Although mammalian hepcidin attracted most interest, it should be mentioned that hepcidin has been described also in fishes, where the size and cysteines are conserved, but the overall amino acid sequence identity with the human one is only 50% (Nemeth and Ganz, 2006). The role of fish hepcidin in systemic iron regulation is unclear.

## **REGULATION OF HEPCIDIN EXPRESSION IN THE LIVER**

Most of the regulation of hepcidin expression in the liver occurs at transcriptional level and it is modulated by iron status, inflammation, and hypoxia (Nemeth and Ganz, 2006). Iron supplementation to the animal readily stimulates liver hepcidin mRNA, but this does not occur in cultured cells, of hepatic or non-hepatic lineage, suggesting an indirect mechanism of iron sensing. A breakthrough for the understanding of the regulatory mechanism was the finding that mice with liver conditional inactivation of the SMAD4 gene developed early and severe liver iron overload, since they did not express detectable hepcidin mRNA even after induction with iron and inflammatory stimuli (Wang et al., 2005). SMAD4 is the key player of the TGF-beta/BMP signaling transduction that involves serine kinases, and phosphorylation of

Poli et al. Hepcidin antagonists

SMAD members to form complexes with SMAD4. These complexes enter the nucleus and bind the responsive elements of the target genes for their activation. The centrality of this regulatory pathway was confirmed by the finding that hemojuvelin (HJV), which is mutated in severe juvenile hemochromatosis (Papanikolaou et al., 2004), acts as a co-receptor of the BMP/SMAD pathway in the liver (Babitt et al., 2006). The 20 different members of the BMP family are produced by various cell types and have different functions, including osteogenesis, embryogenesis, and differentiation. Initial experiments used the osteogenic BMP2 and BMP4 to stimulate hepcidin expression in cultured hepatic cells (Truksa et al., 2006). *In vitro* studies showed that also BMP5, 7 and 9 can induce SMAD pathway and hepcidin expression in primary hepatocytes (Truksa et al., 2006) but after the finding that BMP6 is modulated by systemic iron and, more important, that BMP6−/<sup>−</sup> mice suffer of severe iron overload and the lack of liver hepcidin it was accepted that BMP6 is the major regulator of hepcidin expression (Andriopoulos et al., 2009; Meynard et al., 2009). The dimers of type-II and type-I BMP-receptor participate in BMP/SMAD signaling together with various co-receptors and inhibitors. In the hepatic signaling, ALK2/ALK3 are the predominant BMPR type-I, and ActRIIA is the predominant type-II (Xia et al., 2008) and, of note, the GPI-anchor protein HJV acts as an essential co-receptor for hepcidin expression (Babitt et al., 2006). HJV is a member of the repulsive guidance molecule (RGM) family, which includes RGMa and DRAGON (RGMb), GPI-anchored proteins apparently involved in BMP signaling in different tissues (Corradini et al., 2009). HJV is expressed in skeletal and heart muscle and particularly in the liver where acts as an essential regulator of the signaling. It is also processed by the convertase furin into a soluble form that may act as a decoy and reduce hepcidin expression (Kuninger et al., 2008; Silvestri et al., 2008). It is degraded by the liver-specific serine protease Matriptase-2 (MT2, alias *TMPRSS6*; Silvestri et al., 2009). Inactivation of MT2 in mice and in human results in hepcidin increase and causes iron refractory iron deficiency anemia (IRIDA; Du et al., 2008; Finberg et al., 2008; Folgueras et al., 2008). Neogenin is another transmembrane protein involved in the signaling. It is ubiquitously expressed and interacts with RGM proteins. In the liver it interacts with both HJV and MT2 and may be required for the proper assembly of HJV-BMP ligand-BMPR-I/BMPR-II complex to initiate the BMP signaling and induce hepcidin expression (Zhang et al., 2005; Enns et al., 2012). Neogenin may also form a ternary complex with MT2 and HJV that facilitates HJV cleavage by MT2. Thus Neogenin can stimulate or suppress the BMP signaling by favoring the assembly of BMPs/HJV/BMPR complex or by facilitating MT2 activity, respectively (Zhao et al., 2013). A point not yet fully elucidated regards the involvement of HFE and TFR2. These genes are mutated in type I and III hemochromatosis, both characterized by relatively low hepcidin levels. HFE can bind both TFR1 and TFR2 and a model has been proposed in which HFE and TFR2 act as a complex and contribute to the complete activation of BMP/SMAD pathway (Gao et al., 2010). Other studies indicated that HFE and TFR2 are involved in the activation of MAPK, a pathway that cross talks with the BMP/SMAD and contribute to hepcidin expression (Poli et al., 2010; **Figure 2A**).

Regulation of the BMP/SMAD signaling occurs also after the phosphorylation of SMAD1/5/8 and ensures fine tuning at cytosolic level. SMAD7 antagonizes the recruitment of SMAD4 and the translocation of SMAD1/5/8-SMAD4 complex into the nucleus and recently it has been demonstrated that in hepcidin promoter there is a SMAD7-binding motif for a direct inhibition of the promoter (Mleczko-Sanecka et al., 2010).

Another well characterized pathway of hepcidin regulation involves the inflammatory cytokine IL6 that binds its specific receptor and activates JAK1/2 to phosphorylate STAT3. The phosphoSTAT3 translocates into the nucleus and binds the STAT3 response element (STAT3-RE) in hepcidin promoter, stimulating its transcription (Verga Falzacappa et al., 2007). This pathway can be activated by other cytokines including IL22 and Oncostatin-M which also increase hepcidin transcription, (Chung et al., 2010; Armitage et al., 2011). This inflammatory signaling relies on the BMPs/SMAD pathway to trigger hepcidin expression; in fact SMAD4-KO mice do not respond to the stimulation of IL6 (Wang et al., 2005). Activin-B, which belongs to the TGF-beta family, was shown to be stimulated by inflammation and to induce hepcidin via the BMP/SMAD pathway, thus it may cooperate and enhance the IL6-dependent stimulation and it could represent a connecting component between iron status and inflammation (Besson-Fournier et al., 2012; **Figure 2B**).

## **THE HEPCIDIN–FERROPORTIN AXIS**

The only known receptor for hepcidin is ferroportin, which is the sole cellular iron exporter. Ferroportin (FPN) is expressed by enterocytes of the duodenum, by macrophages that process effete RBC and by liver, and is responsible for the release of iron to transferrin, in a mechanism that needs the assistance of a copper ferroxidase such ceruloplasmin or haephestin (Nemeth et al., 2004; Hentze et al., 2010). The iron export activity of ferroportin is essential for the regulation of systemic iron homeostasis, but its biochemical properties and the mechanism of iron transport are not well characterized. It is a 62.5 kDa protein with 12 transmembrane domains, the N-terminus and possibly also the C-terminus are cytosolic (Liu et al., 2005; **Figure 3A**). Mutation in FPN gene lead to type IV hemochromatosis (ferroportin disease), which is dominantly transmitted. This has suggested that ferroportin forms functional dimers (De Domenico et al., 2005), but direct studies indicated that it is a monomer (Montosi et al., 2001; Liu et al., 2005; Wallace et al., 2010; **Figure 3B**). The ferroportin mRNA has an IRE on the 5 UTR that may be responsible for its iron-dependent regulation (Muckenthaler, 2008). However, the major regulatory step occurs at a post-translational level caused by hepcidin (Nemeth et al., 2004). In fact the binding of hepcidin causes ferroportin internalization and degradation, with the effect to reduce systemic iron availability (Ganz and Nemeth, 2012). Further studies showed that Hepcidin-25 binds an extracellular loop of FPN adjacent to the cytosolic loop containing the two tyrosines required to signal internalization (Nemeth et al., 2006). The binding involves disulfide bridging with FPN Cys326 (Fernandes et al., 2009; **Figure 3A**), ferroportin ubiquitination prevalently at the level of lysines present in a third cytoplasmic loop of ferroportin (possibly Lys 229 and 269), and proteasomal degradation (Qiao et al., 2012; **Figure 3B**).

## **LOCAL EXPRESSION OF HEPCIDIN IN CNS**

Liver is the major producer of hepcidin, and the fine tuning of its expression sets the level of circulating hepcidin to govern systemic iron homeostasis. Various other tissues express hepcidin mRNA and protein suggesting the presence of autocrine or paracrine circuits that may contribute to the regulation of local iron distribution. This may be particularly important in the central nervous system (CNS) that is separated from the rest of the body. The blood–brain barrier (BBB) and the blood/CSF barrier are the major sites of iron exchange with the periphery (Rouault et al., 2009; **Figure 4A**). These barriers act as semipermeable cellular gates characterized by tight junction and specialized transcellular carriers mediating influx or efflux. Transferrin and its receptor take part in most of iron transfer, but the details on the regulatory mechanism are missing. Of interest is that cells of the blood/CSF barrier express proteins of the hepcidin/ferroportin axis (**Figure 4B**). The imaging techniques and histology of the different areas of the brain showed that regions with neurodegeneration exhibit also iron

accumulation in pathological conditions such as Alzheimer's disease (AD; Bishop et al., 2002), Parkinson's disease (PD; Berg et al., 2001; Zecca et al., 2004), and neurodegeneration with brain iron accumulation (NBIA; Hayflick, 2006), while restless leg syndrome (RLS) is associated with reduced CNS iron content (Clardy et al., 2006). There is evidence that the hepcidin/ferroportin axis might play a role in the iron decompartmentalization occurring in these disorders. In fact, not only hepcidin and ferroportin, but also the accessory proteins in iron transfer ceruloplasmin, hephaestin and DMT1, and those involved in hepcidin regulation (BMPs, IL6, Tfr2) are expressed and regulated in the brain (**Figure 4B**). Hepcidin and ferroportin mRNA and protein were detected in different murine and human cerebral areas (Clardy et al., 2006; Zechel et al., 2006) and hepcidin intraventricular injection resulted in downregulation of FPN protein (Wang et al., 2010b). Systemic iron inflammation or acute iron overload induced a significant increase in cerebral hepcidin expression in mice and rats (Malik et al., 2011; Wang et al., 2011; Sun et al., 2012) thus suggesting that the machinery controlling hepcidin

expression in the brain is similar to the hepatic one. Novel evidence suggests that perturbation of the cerebral hepcidin-FPN axis may contribute to local deregulation of iron homeostasis. They include an increase of hepcidin in the brain in a murine model of cerebral ischemia (Ding et al., 2011), and a reduction of both ferroportin and hepcidin detected in lysates obtained from AD patients (Raha et al., 2013). Furthermore, calorie restriction prevented both the increase in cerebral hepcidin mRNA and the impairment of learning and memory observed in an experimental model of aging (Wei et al., 2014). Further characterization of the machinery controlling iron balance in the brain is needed, but attention should be given at the possible modulation of hepcidin expression in CNS caused by pharmacological intervention aimed at regulating systemic iron homeostasis, in particular those involving molecules that can pass the BBB.

## **IRON DISORDERS IN WHICH HEPCIDIN IS PATHOLOGICALLY UPREGULATED**

Hepcidin is the hormone of iron and inflammation and its deregulation occurs in all iron related disorders, including the ones characterized by iron restriction and anemia in which hepcidin is abnormal (Ganz and Nemeth, 2011). Elevated hepcidin levels are associated with secondary iron overload, genetic IRIDA, chronic infectious and inflammatory diseases resulting in anemia of inflammation (AI). The finding that hepcidin is upregulated by the inflammatory cytokine IL6 (Verga Falzacappa et al., 2007) contributed to explain the anemia of chronic diseases (ACD) alias AI. This occurs in a variety of disorders, like infections, chronic kidney diseases (CKDs), rheumatoid arthritis, and cancer (Wang et al., 2012) including multiple myeloma (Maes et al., 2010), a severe malignant disorder of plasma cells. Currently ACD therapy includes erythropoiesis-stimulating agents and intravenous iron (Goodnough et al., 2010), which may have adverse effects and are scarcely effective (Glaspy, 2012; Fung and Nemeth, 2013). For example inflammation often induces erythropoietin resistance (Macdougall and Cooper, 2002). Alternative treatments have been proposed and hepcidin antagonists seem to be the best candidates to treat these disorders (Wang et al., 2012).

## **THERAPEUTIC APPROACHES TO NEUTRALIZE HEPCIDIN EXCESS**

The mechanisms involved in the regulation of hepcidin expression are complex and partially known, and the approaches can use different targets to downregulate hepcidin or its function, as described in recent reviews (Sun et al., 2011; Fung and Nemeth, 2013).

## **BMP/BMPR COMPLEX**

One obvious target is the BMPs/BMPR complex. This was initially tested by developing anti-BMP6 antibodies to abolish the interaction between BMP6 and its receptors. The iron-restricted anemia of HFE transgenic mice due to high hepcidin was effectively cured with 10-day treatment with anti-BMP6 (Corradini et al., 2010). However, similar treatment was not effective in ACD possibly due to the expression of other BMPs, as it occurs in multiple myeloma with high BMP2 (Maes et al., 2010). Another target is HJV, the major co-receptor of the BMP/SMAD signaling. A treatment of healthy rodents with soluble HJV.Fc blocked SMAD phosphorylation, decreased hepcidin expression, mobilized splenic iron content and increased serum iron levels (Babitt et al., 2007). In an ACD rat model, a long 4-week treatment showed a recovery of anemia with the inhibition of SMAD1/5/8 phosphorylation; increase of splenic ferroportin levels and of serum iron (Theurl et al., 2011). Another promising approach is to block the phosphorylation of type I BMP receptor. The initially tested molecule was dorsomorphin, a non-specific kinase inhibitor (Yu et al., 2008), and then a more selective BMP inhibitor, coded LDN-193189 (Cuny et al., 2008), in animal models. A 4-week treatment of anemic rats with chemically induced arthritis reduced hepatic hepcidin mRNA levels, increased serum iron concentration, increased ferroportin expression in splenic macrophages, and improved hemoglobin

levels and hematocrit (Theurl et al., 2011). It was also effective in treating mice with acute inflammatory anemia induced by turpentine injections (Steinbicker et al., 2011). A limit of this chemical inhibition of hepcidin is the lack specificity for BMP inhibition, since it can also potently inhibit VEGF and components of the MAPK/ERK pathway and show toxicity (Vogt et al., 2011). Liver is the organ most easily targeted by siRNAs, and the studies on the silencing of HJV and TfR2 are ongoing (Akinc et al., 2011).

## **HEPARIN**

The observation that BMPs are heparin binding molecules and that heparin modifies the osteogenic activity of BMP2/4 stimulated (Poli et al., 2011) to verify the effect of heparin on hepcidin expression. It was shown that commercial heparins are potent hepcidin inhibitor *in vitro* in HepG2 cells and *in vivo* in healthy mice and that act by inhibiting the BMP6/SMAD signaling. Heparins are well characterized molecules with some 70 years of clinical experience, and appealing drugs for the treatment of anemia. The major drawback of their strong anticoagulant activity can be overcome. In fact the anticoagulant activity is mostly linked to high binding affinity to antithrombin, which is limited to a specific pentasaccharide, named AT-bs, absent in some heparins, that can

be chemically modified (**Figure 5**). The main modifications to reduce or abolish the anticoagulant property are summarized in **Figure 5B** and they are: *N*-desulfation or *N*-acetylation, 2-/6- *O*-desulfation, supersulfation or, more simply, the treatment of heparins by reduction and oxidation, to obtain the so called ROheparins (Casu et al., 2002). This splits the glycol bonds, increasing molecular flexibility and improving the interaction with targets other than antithrombin. These glycol-split heparins retain various biological functions, including anti-heparanase activity that reduces tumor growth and metastasis in animal models (Ritchie et al., 2011). Some of these compounds are in clinical trials and they have shown little or no toxicity. These glycol-split heparins showed to be potent hepcidin inhibitors *in vitro*, in HepG2 cells and primary hepatocytes, and *in vivo* in mice (Poli et al., 2014). *In vivo* these heparins reduced hepcidin in 6 h with concomitant increase of serum iron and decrease of spleen iron. They inhibited hepcidin also after an acute lipopolysaccharide (LPS) stimulation, and in a mouse model of anemia induced by a single injection of heat-killed *Brucella abortus* (HKBA) these heparins improved the recovery of anemia. The available data indicate that heparins act by sequestering of BMP6 and inhibiting the SMAD1/5/8 signaling. These findings also indirectly suggest a role of liver heparan sulfate proteoglycans (HSPGs) in hepcidin regulation. The main

structure of heparin is composed by 70% of *N*-sulfated region (NS, IdoA2SO3- -GlcNSO36SO3), *N*-acetylated region (NA, GlcA-GlcNAc) and mixed NA/NS (GlcA-GlcNSO3; **Figure 5A**). Heparin is structurally analogous to the heparan sulfates (HSs) exposed on the surface of all cells that are known to modulate critical biological events, such as embryonic development, growth regulation and maintenance of normal tissue structure and function (Turnbull et al., 2001). In fact they can act as "receptors" for circulating proteins, including several cytokines and angiogenic growth factors (Casu et al., 2010). Heparin is utilized as a model to study the interaction of molecules with cellular heparan sulfates and to modulate their biological activity (Rusnati et al., 2005). In fact it was recently demonstrated that HSPGs act as coreceptors of BMP2 and BMP4 in facilitating receptor oligomerization (Kuo et al., 2010). The consequences of the BMPs binding to HSPGs vary, depending on the BMP member, cell type targeted and if HSPGs are cell-associated (co-receptor action) or in a free form (antagonist effect). Accordingly, alteration of cell-associated HSPGs by heparinases or by chlorate treatments reduced (Irie et al., 2003) or increased BMP signaling (Jiao et al., 2007). Interestingly, the HS in the liver are highly sulfated, and

their inactivation by conditional-KO of key enzymes has effects on lipid homeostasis, but hepcidin has not been analyzed yet (Stanford et al., 2010).

## **IL6/STAT3 AXIS**

Inflammation induces hepcidin expression mainly through IL6/STAT3 pathway, which can be blocked by anti-IL6 antibody. Siltuximab, an anti-IL6 monoclonal antibody drug used in clinic, was shown to be effective in reducing hepcidin expression in patients with Castleman's disease (CD) and in improving their anemia (van Rhee et al., 2010). The antibody was used also in patients with renal cell carcinoma and multiple myeloma resulting in a decrease of hepcidin and an increase of hemoglobin (Schipperus et al., 2009; Kurzrock et al., 2013). Tocilizumab is another antibody in clinical use that acts on IL6 activity by binding IL6 receptor. CD patients, treated with it, showed a reduction of serum hepcidin and correction of anemia after 6–12 month treatment (Song et al., 2010). Tocilizumab was used also in monkeys with collagen-induced arthritis with an improvement of anemia (Hashizume et al., 2010). Chemical agents like AG490 that inhibits STAT3 phosphorylation (Caceres-Cortes, 2008) or

PpYLKTK that disrupts pSTAT3 dimerization and DNA binding were investigated in cancers with elevated JAK/STAT activity (Zhang et al., 2010). Both compounds decrease IL6-dependent hepcidin expression in differentiated mouse hepatocytes (Fatih et al., 2010) and AG490 also *in vivo* in healthy mice (Zhang et al., 2011).

## **ANTI-HEPCIDIN AGENTS**

A direct approach is to downregulate hepcidin using RNA interference, taking advantage of the observation that liver is an easy target for siRNAs. This implies the design of RNAi without offtarget effects, sufficiently stable *in vivo*, biocompatible and with specific delivery to liver but not to other organs (Wang et al., 2010a). High affinity anti-hepcidin antibodies have been produced and have been engineered to be used *in vivo* and to analyze their effects. They improved the inflammatory anemia in mice induced by HKBA only when co-administrated with erythropoietic stimulating agents (Sasu et al., 2010). Fully humanized mAb against hepcidin (LY2787106) is currently in Phase I for the treatment of cancer-related anemia. Hepcidin blocking proteins were obtained by modifying the lipocalins, natural proteins that bind small hydrophobic ligands and cell surface receptors (Flower, 1996; Schlehuber and Skerra, 2005). They were engineered to produce anticalin PRS-080 that exhibits sub-nanomolar affinity for human hepcidin. Monkeys treated with PRS-080 showed an effective iron mobilization, and studies are in progress on anticalin safety and tolerability *in vivo*. Spiegelmers are synthetic compounds designed to inhibit other molecules. Spiegel means mirror in German and they are mirror-images L-enantiomeric oligonucleotides that bind the targets in a manner similar to antibodies or aptamers. Being nuclease resistant and immunologically passive is suited for *in vivo* application. NOX-H94 is a structured L-oligoribonucleotide, that binds human hepcidin with high affinity, blocking its biological function (Schwoebel et al., 2013). In monkey NOX-H94 prevented the onset of anemia induced by IL6, in human volunteers, it increased indices of iron availability and was safe and well tolerated (Riecke et al., 2012), it also delayed the onset of hypoferremia in volunteers treated with LPS (Van Eijk et al., 2013). The Phase II clinical trials with NOX-H94 are ongoing for patients with anemia of cancer.

## **ALTERATION OF HEPCIDIN–FERROPORTIN INTERACTION**

Antibodies that block ferroportin binding to hepcidin without affecting its functionality have been described (Leung et al., 2012). They have been engineered and are now in a Phase I trial. A high throughput screening approach discovered a thiol modifier compound (fursultiamine) that prevented ferroportin–hepcidin interaction sequestering the Cys326-HS residue (essential for hepcidin binding, **Figure 3A**) and blocking internalization of ferroportin (Fung and Nemeth, 2013). It could be an interesting agent to be evaluated *in vivo*.

## **ERYTHROID FACTORS**

Growth differentiation factor 15 (GDF15), is a member of the transforming growth factor-β superfamily. It is produced in erythroid precursor cells and is strongly upregulated in disorders with increased ineffective erythropoiesis, such as β-thalassemia, congenital dyserythropoietic anemias. It was shown to downregulate hepcidin mRNA expression in primary human hepatocytes (Tanno et al., 2007). A synthetic low molecular weight compound (K7174) that enhances GDF15 expression in HepG2 cells was described, and it also reduced hepcidin (Fujiwara et al., 2013). It was claimed a potential therapeutic option to treat ACD. However, this contrasts with the finding that GDF15 deficient mice have normal hepcidin expression and that GDF15 is not required to balance iron homeostasis in response to blood loss (Casanovas et al., 2013). More recently another erythroid factor with strong effect on hepcidin expression was identified, it was named erythroferrone it is stimulated by active erythropoiesis and it suppresses hepcidin expression in hepatic cells (Kautz et al., 2013). It is an important potential targetfor the control of hepcidin expression.

## **ANIMAL MODELS OF INFLAMMATORY ANEMIA**

The hepcidin antagonists are expected to find clinical use mainly for the treatment of inflammatory anemia which, although widely diffused in clinical practice, has few animal models with different properties, as recently reviewed (Rivera and Ganz, 2009). Here is a short description of the ones so far described in past and recent papers focusing only on the well-known model.

### **LIPOPOLYSACCHARIDE**

Lipopolysaccharide injections in the mice induce an inflammatory response, with upregulation of IL6, an increase in Socs3 mRNA, Crp mRNA and hepcidin mRNA and protein and a decrease in serum iron, but generally do not induce anemia (Poli et al., 2014). The activation of hepcidin is fast (4–6 h) and decreases just as quickly. It was shown that anemia could be induced after a single dose of LPS followed a week later by an injection of Zymosan A (a preparation from yeast wall). This, so named ZIGI mouse model, is characterized by high IL6 and hepcidin, increase in spleen iron content with a decrease of liver ferroportin and anemia 5 days after Zymosan injection (Lasocki et al., 2008). The LPS pre-treatment reduces the strong septic shock-like response triggered by Zymosan A which leads to multiple organ dysfunctions (Volman et al., 2005). This interesting model has not been used yet to test the efficacy of hepcidin antagonists.

## **TURPENTINE**

Turpentine is used to trigger sterile inflammatory response in different animal models. Mice treated with a single subcutaneous injection of turpentine (5 ml/kg) showed induction of hepcidin and hyperferremia (Sakamori et al., 2010). Anemia was described after 3-week of daily treatments, that was accompanied by a reduction of mean corpuscular volume (MCV) and serum iron and a 2–7 fold increase of hepcidin.

#### **HEAT-KILLED** *Brucella abortus*

Heat-Killed *Brucella abortus* agent is the vaccine to prevent Brucellosis in large animals. When injected in mice induce an inflammatory response with anemia. It seems the easiest mouse model of inflammatory anemia, and it was used to verify the activity of hepcidin antagonists like anti-hepcidin antibodies (Sasu et al., 2010) and glycol-split heparins (Poli et al., 2014) to improve anemia *in vivo*. This model of inflammatory anemia has been recently analyzed in depth by the groups of Ganz (Kim et al., 2014) and Rivella (Gardenghi et al., 2014). Both showed that HKBA-treated mice developed a severe anemia, with a nadir after 14 days, followed by a partial recovery after 28 days. They showed hypoferremia and iron-restricted erythropoiesis with normal iron stores, shortened erythrocyte lifespan, and reduced erythropoiesis. The IL6-KO and hepcidin-KO mice showed a milder anemia and a faster recovery confirming the role of inflammation and of hepcidin in the development of anemia in this model. The anemia has multifactorial pathogenesis and hepcidin (that is induced transiently at 6 h after the HKBA injection) appears to play an important role in it. It remains to be evaluated if a model closer to ACD in human can be produced by repeated injections of lower doses of HKBA.

**RAT MODEL OF ANEMIA OF INFLAMMATION**

Anemia of inflammation can be obtained in rats with different treatments that have been used for many years. Nowadays a good rat model of ACD is obtained with a single intraperitoneal injection of group A streptococcal peptidoglycan-polysaccharide (PG-APS) with rhamnose. This treatment caused arthritis with involvement of multiple joints (Cromartie et al., 1977). Recently this rat model was analyzed in depth for all iron parameter. After 3 weeks of PG-APS treatment, the rats showed an increase of serum IL6, hepcidin, and ferritin. Spleen ferroportin decreased, resulting in a reduction of iron release from macrophages. These rats develop anemia after 2-week treatment that persists for at least 3 months (Theurl et al., 2009). This anemia has typical features of human ACD (mild to moderate normocytic and normochromic anemia) with inflammation. This rat model was used to test the effect of different hepcidin inhibitors with positive results (Theurl et al., 2011).


#### **Table 1 | Hepcidin inhibitors and corresponding targets.**

#### *Modified adenine-induced kidney disease rat model*

Another interesting rat model that develops CKD is obtained in rodent with a 0.75% adenine diet (modified adenine) for 3 weeks followed by a control diet for 5 weeks. This protocol improved survival (90%) maintaining persistent kidney disease and more severe anemia (Sun et al., 2013). This model was used to evaluated the effect of the BMP inhibitor LDN-193189 (Sun et al., 2013). Adenine-treated rats showed increased liver hepcidin mRNA, decreased serum iron, increased spleen iron content, low hemoglobin, and low erythropoietin levels. LDN-193189 treatment reduced hepatic hepcidin mRNA, mobilized stored iron and increased hemoglobin content of reticulocytes.

#### **GENETIC MODEL OF ANEMIA**

Iron refractory iron deficiency anemia is an autosomal recessive human disorder characterized by congenital hypochromic, microcytic anemia, very low mean corpuscular erythrocyte volume, low transferrin saturation, poor response to oral iron supplementation and partial response to parenteral iron therapy. The mouse models that mimic this disorder were obtained by two groups (Du et al., 2008; Folgueras et al., 2008). One was produced by chemically induced mutation that causes splicing defect in the transmembrane serine protease in gene *Tmprss6*. The other one is the KO model produced by a duplication of an entire region of *Tmprss6* gene. In the two models matriptase-2 is inactivated and is characterized by progressive loss of body but not facial hair ("*mask* phenotype") and microcytic anemia. The *mask* phenotype results from reduced absorption of dietary iron and iron retention in duodenal enterocytes, low ferroportin, and iron deficiency anemia caused by high levels of hepcidin (Folgueras et al., 2008). *Mask* homozygotes are slightly smaller than their heterozygous littermates, and adult female homozygotes are infertile whereas male homozygotes retain fertility. This mouse with a direct implication of hepcidin upregulation is probably the best model to study the long term effects of hepcidin inhibitors with the aim at solving anemia.

## **CONCLUSION**

This review shows that many laboratories are studying different pharmacological means to neutralize hepcidin expression or activity in order to cure inflammatory anemia. They produced a number of promising approaches, and some of them have been tested in animal models. Most of them seemed to be effective in reducing hepcidin expression or activity under acute conditions, but it is still unclear if and how they are efficient in the treatment of anemia. One of the problems is the lack of adequate animal models for inflammatory anemia, as indicated above. Mice models are rather complex, and rat models seems to mimic more closely the human disease, but the absence of transgenic rats for hepcidin and inflammatory cytokines does not allow a detailed characterization. Monkeys have been used to induce inflammatory response, but not anemia (Cooke et al., 2013). The described antagonists (**Table 1**) originate from different and novel biotechnological techniques, including humanized anti-hepcidin antibodies, aptamers, anticalin, siRNAs, and the old traditional heparin. Some of them are in clinical trials, and perhaps in a few

years we will know if the downregulation of hepcidin really meets the expectation to improve the anemia in most, or some chronic diseases. All the antagonists have some advantages and problems. For example the humanized antibodies have a long half-life *in vivo*, but their production is highly expensive. Aptamer and anticalin inactivate hepcidin, but their fate is unclear. The specific and efficient delivery of siRNA is complex. Non-anticoagulant heparins are probably the best known, most convenient and safer agents. After 70 or more years of use in clinic, most of problems and side effects of heparins are known. They include thrombocytopenia, elevation in serum aminotransferase, hyperkalemia, alopecia and osteoporosis, but they occur rarely and are transient. The removal of anticoagulant activity has been resolved and the clinical trials of these agents as hepcidin antagonist should not be far away. Also the negative effects of treatments for hepcidin inhibition should be taken into account. They are expected to increase systemic iron availability and absorption, which may favor irondependent oxidative damage in some parenchymal tissues. If the pharmacological agents are capable to cross BBB and enter the brain, they may alter CNS iron homeostasis with unpredictable effects. In addition, hepcidin is known to have an antimicrobial activity, which is considered low, but its biological role needs to be established.

## **ACKNOWLEDGMENTS**

We are grateful di Dr. Annamaria Naggi for the support in the study of heparin chemistry, and to Dr. Dario Finazzi for reading the manuscript and helpful suggestions. This work was partially supported by Fondazione Cariplo grant no. 2012-0570 and by MIUR-PRIN-11 to Paolo Arosio.

## **REFERENCES**


systemic iron homeostasis in phlebotomized mice. *Haematologica* 98, 444–447. doi: 10.3324/haematol.2012.069807


(Tmprss6) is an essential regulator of iron homeostasis. *Blood* 112, 2539–2545. doi: 10.1182/blood-2008-04-149773


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

*Received: 28 February 2014; paper pending published: 23 March 2014; accepted: 07 April 2014; published online: 28 April 2014.*

*Citation: Poli M, Asperti M, Ruzzenenti P, Regoni M and Arosio P (2014) Hepcidin antagonists for potentialtreatments of disorders with hepcidin excess. Front. Pharmacol. 5:86. doi: 10.3389/fphar.2014.00086*

*This article was submitted to Drug Metabolism and Transport, a section of the journal Frontiers in Pharmacology.*

*Copyright © 2014 Poli, Asperti, Ruzzenenti, Regoni and Arosio. This is an openaccess article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*

## Physiological implications of NTBI uptake by T lymphocytes

## *Jorge P. Pinto1, João Arezes 1, Vera Dias 1, Susana Oliveira1, Inês Vieira1, Mónica Costa1,2, Matthijn Vos 3, Anna Carlsson3, Yuri Rikers 3, Maria Rangel <sup>4</sup> and Graça Porto1,5,6\**

*<sup>1</sup> Molecular and Cellular Biology Division, Basic and Clinical Research on Iron Biology, Instituto de Biologia Molecular e Celular, Universidade do Porto, Porto, Portugal*

*<sup>2</sup> Faculdade de Medicina, Universidade do Porto, Porto, Portugal*

*<sup>3</sup> Europe NanoPort, FEI, Eindhoven, Netherlands*

*<sup>4</sup> Chemistry Department, REQUIMTE, Instituto de Ciências Biomédicas Abel Salazar, Universidade do Porto, Porto, Portugal*

*<sup>5</sup> Clinical Hematology, CHP-HSA - Santo António General Hospital, Porto, Portugal*

*<sup>6</sup> Molecular Immunology and Pathology, Instituto de Ciências Biomédicas Abel Salazar, Universidade do Porto, Porto, Portugal*

#### *Edited by:*

*Paolo Arosio, University of Brescia, Italy*

#### *Reviewed by:*

*Emanuela Tolosano, University of Torino, Italy Stefania Recalcati, University of Milan, Italy*

#### *\*Correspondence:*

*Graça Porto, Molecular and Cellular Biology Division, Basic and Clinical Research in Iron Biology, Instituto de Biologia Molecular e Celular, Rua do Campo Alegre, 823, 4150-180 Porto, Portugal*

*e-mail: gporto@ibmc.up.pt*

In iron overload disorders a significant fraction of the total iron circulates in the plasma as low molecular weight complexes not bound to transferrin, known as non-transferrin-bound iron (NTBI). By catalyzing the formation of free radicals, NTBI accumulation results in oxidative stress and cellular damage, being a major cause of organ toxicity. NTBI is rapidly and preferentially cleared from circulation by the liver and the myocardium, the main disease targets in iron overload conditions. We have recently demonstrated that human peripheral blood T lymphocytes take up NTBI *in vitro*, with a pattern that resembles that of hepatocytes. Since T lymphocytes constitute a numerically important component of the circulating cell pool, these findings support a putative role for this cell type in the systemic protection against iron toxicity. Here we tested the hypothesis that the circulating peripheral blood T lymphocyte pool constitutes an important storage compartment for NTBI and is thus a modifier of NTBI deposition in target organs. First we show that NTBI uptake by human T lymphocytes increases the expression of the iron-storage protein ferritin and of the iron exporter ferroportin *via* an IRE-dependent mechanism. NTBI retention by T lymphocytes is shown to be critically controlled by the hepcidin-mediated modulation of ferroportin both *in vitro* and *in vivo*. Finally, the protective effect of T lymphocytes was tested by analyzing the patterns of iron accumulation in the T lymphocyte-deficient mouse model *Foxn1nu* before and after reconstitution with T lymphocytes by adoptive transfer. The results confirmed a significant increase of liver and pancreas iron accumulation in T lymphocyte-deficient mice. NTBI accumulation in the liver and spleen was prevented by reconstitution with syngeneic T lymphocytes. Altogether, our results demonstrate that T lymphocytes are important components of a circulating "NTBI storage compartment" and show its physiological relevance as a modifier of tissue iron overload.

#### **Keywords: iron, homeostasis, NTBI, lymphocytes, hemochromatosis**

## **INTRODUCTION**

Iron, the most abundant transition metal in mammalian systems, is essential for various vital metabolic processes. In physiologic conditions iron circulates in the plasma bound to transferrin, the main iron transporter protein, and this constitutes the major iron source for iron-avid processes, such as erythropoiesis (Hentze et al., 2010).Circulating iron which is not bound to transferrin, heme or ferritin (here designated as nontransferrin-bound iron—NTBI) becomes relevant in iron overload disorders, appearing in plasma even before transferrin becomes fully saturated (Breuer et al., 2000; Esposito et al., 2003).

NTBI is the main source of iron for storage in the liver (Zimelman et al., 1977; Brissot et al., 1985). However, in contrast to transferrin-bound-iron, NTBI is potentially toxic, causing cellular damage, not only of the plasma membrane but also of various intracellular organelles, due to its involvement in the formation of reactive oxygen species (reviewed in Brissot et al., 2012). The uptake of NTBI by hepatocytes is thus viewed as a clearance mechanism of potentially toxic circulating iron that could otherwise cause damage to other cell types. Not surprisingly, hepatocytes are the first target of iron toxicity in situations of severe iron overload such as in HFE-Hereditary Hemochromatosis (HFE-HH) or transfusion-dependent betathalassemia. When the clearance capacity of liver is exceeded, other organs, namely pancreas, heart or hypophysis, are also affected by NTBI uptake and accumulation, leading to the fullblown clinical picture of severe iron overload with liver cirrhosis, diabetes, cardiomyopathy, and hypogonadotrophic hypogonadism (Pietrangelo, 2004). Therefore an effective removal of circulating NTBI is a major goal in the management of iron overload disorders.

We have recently demonstrated that T lymphocytes are capable of taking up and accumulating the same NTBI species as hepatocytes (Arezes et al., 2013). Together with previous evidence showing that T lymphocytes synthesize ferritin in greater amounts than non-T lymphoid cells (Dörner et al., 1980) these results support the hypothesis that T lymphocytes may have an important role in iron handling in the blood circulation, where they could act as natural NTBI buffers and constitute a barrier to protect other tissues from iron-mediated toxicity. In order to test this hypothesis we analyze here the capacity of T lymphocytes to store/export NTBI and assess the physiological implications of NTBI storage by T lymphocytes as a modifier of tissue iron load.

## **METHODS**

## **ISOLATION OF HUMAN PERIPHERAL BLOOD CELLS**

Peripheral Blood Mononuclear cells (PBMCs) were obtained from apparently healthy volunteer blood donors, randomly recruited at Santo António Hospital Blood Bank (Porto, Portugal) who gave their consent to participate in this study, approved by the Santo António Hospital Ethical Committee. Cells were isolated by gradient centrifugation over Lymphoprep (Nycomed). After lysis of erythrocytes, cells were resuspended in RPMI (GibcoBRL) supplemented with 10% fetal calf serum (FCS; GibcoBRL) and plated. CD3+, CD4+, and CD8+ cells were purified from PBMCs using magnetic-activation cell sorting (MACS), after incubation with specific microbead-conjugated antibodies (Miltenyi Biotec), according to manufacturer's instructions.

## **NTBI UPTAKE**

Uptake of non-transferrin-bound iron (NTBI) was assessed using 55Fe-citrate (Grootveld et al., 1989). 55Fe-citrate stock solutions were prepared by mixing 55FeCl3 (5–10 mCi, in 0.1 M HCl; Amersham) with unlabeled trisodium citrate, for a final citrate concentration of 100μM. The pH was maintained at 7.4 and solutions were allowed to rest for 20 min before being diluted 33 fold in uptake medium and added to cells. Specific activity in the uptake medium was approximately 30 counts· min−1· pmol−<sup>1</sup> Fe. All Fe:citrate solutions were freshly prepared before use and discarded after each experiment. To prevent competition of T lymphocyte-secreted transferrin for citrate-bound iron, unless otherwise indicated cells were depleted of intracellular transferrin by incubation for 1 h in serum-free RPMI. Since the estimated time of recycling of transferrin molecules in lymphocytes is approximately 30 min (Holtzman, 1939), 1 h should suffice for all endogenous transferrin to be secreted and, due to the absence of extracellular iron, be prevented from being endocytosed again. Cells were then washed and incubated with RPMI + 20% FCS <sup>+</sup> <sup>5</sup>μ<sup>M</sup> 55Fe-citrate (as 5μ<sup>M</sup> 55FeCl3<sup>+</sup> <sup>100</sup>μM citric acid), at 37◦C. Given the 5μM is the typical NTBI concentration reported in sera from *thalassemia major* patients (Evans et al., 2008) and 100μM citric acid falls within the interval of citrate concentrations normally present in human blood plasma (Lentner, 1984). In some experiments the iron chelator desferrioxamine (DFO; Sigma) was added to the medium providing a final concentration of 5μM. The pH of the incubation medium was maintained at 7.4. After incubation, cells were washed 3× with ice-cold washing buffer (20μM DFO, in PBS, pH 7.4), lysed with a 0.1% NaOH, 0.1% Triton X-100 solution and intracellular Fe was measured in a MicroBeta Trilux β-counter (Perkin Elmer), for 1 min. No significant impact of iron treatments in cell viability was observed, using trypan blue exclusion and maintenance of proliferative potential following activation with anti-human anti-CD3 and anti-human anti-CD28 for CD3+, CD4+, and CD8+T lymphocytes (Arezes et al., 2013).

## **IRON EXPORT**

T lymphocytes were depleted of transferrin, as described above, and incubated, unless otherwise stated, in RPMI + 5μM Fecitrate (or 5μ<sup>M</sup> 55Fe-citrate) <sup>+</sup> 20% FCS for 2 h. Cells were washed 2× with washing buffer and incubated for different timeperiods in RPMI + 20% FCS + 5μM DFO (for short-term experiments), to prevent re-uptake of exported iron, or in RPMI + 20% FCS (for long-term experiments). At each time-point, the supernatants were collected and Fe was quantified as described in *NTBI uptake*.

To analyze the effect of hepcidin in iron export, CD4+ and CD8+ T lymphocytes were iron-loaded with 5μM Fe-citrate (or 5μM 55Fe-citrate), as described above, washed and incubated in export medium (RPMI + 20% FCS) supplemented with 100 or 600 ng/ml of human synthetic hepcidin (Peptides International), or with an equal volume of PBS (Mock), for up to 72 h. FPN levels were analyzed by western blot analysis (see below) after 24 h of incubation and intracellular <sup>55</sup> Fe levels were quantified after up to 72 h of incubation.

## **ASSESSMENT OF LABILE IRON POOL**

CD4+ and CD8+ T lymphocytes were loaded with 0.25μM of calcein acetoxymethyl ester dye (Molecular Probes) at 37◦C for 10 min, and rinsed twice with PBS to remove unincorporated dye. These conditions were empirically determined by us to be the most suitable for detection of iron-induced changes in calcein fluorescence in PBMCs (Pinto et al., unpublished data). Cells were then incubated with 5μM of Fe-citrate (5:100) + 20% FCS or with 5μM of Na-citrate + 20% FCS, for up to 180 min, and the fluorescence intensity was measured at defined time points in a fluorescent plate reader, using the conditions λexc = 488 nm and λem = 517 nm. In some experiments, the membrane permeable iron chelator salicylaldehyde isonicotinoyl hydrazone (SIH, kindly provided by Dr. Prem Ponka, McGill University, Canada) was added to the cells (10μM final concentration) for 30 min following incubation with iron.

## **WESTERN BLOT ANALYSIS**

Cellular extracts were isolated by the carbonate fractioning procedure, using a protocol adapted from Fujiki et al. (1982). Briefly, harvested CD3+ lymphocytes were resuspended in 120 mM sodium carbonate supplemented with 1× Complete EDTA-free protease inhibitor cocktail (Roche Diagnostics) and incubated on ice for 30 min. An aliquot of this suspension was saved as total fraction. The suspensions were transferred to polycarbonate tubes and centrifuged at 38,000 rpm for 45 min, at 4◦C, using a Beckman 70.1 Ti rotor. The supernatants (corresponding to the cytoplasmic fractions) were decanted and the membrane pellets were directly dissolved, at 65◦C, in 2× Laemmli buffer, with occasional vortexing and sonication. Both total and soluble extracts were subjected to trichloroacetic acid precipitation and processed for SDS-PAGE.

Cell extracts were submitted to SDS-PAGE and immunoblotted for ferritin (1:1000 rabbit anti-ferritin antibody; Abcam) and/or ferroportin (1:100 rabbit anti-ferroportin antibody; Novus Biologicals). To control for the purity of cell fractions, some of the extracts were hybridized with an anti-alpha 1 Na/K ATPase antibody (1:2000; Abcam). Immuno-reactive proteins were detected with Super Signal West Dura (Pierce). For control of protein loading, membranes were stripped and re-hybridized with 2μg/ml of a mouse anti-human β-actin or goat polyclonal anti-β-actin antibodies (both from Abcam).

## **INTRACELLULAR Fe DETECTION BY ENERGY DISPERSIVE X-RAY ANALYSIS**

Lymphocytes were separated from monocytes in PBMCs by allowing monocytes to adhere to the substrate for 1.5 h and collecting the supernatant (lymphocyte-fraction). Cells were depleted of transferrin and incubated with 5μM Fe-citrate (5:100) for 24 h. Cells were pelleted, fixed in 2.1% glutaraldehyde/2% formaldehyde, washed with 0.1 M phosphate buffer and fixed for 1 h with 1% OsO4, at 4◦C. After dehydration, the samples were embedded in Epon 812 and ultrathin sections obtained. For high resolution fast elemental mapping of iron-loaded lymphocytes, a FEI Tecnai Osiris transmission electron microscope with ChemiSTEM technology was used. Images were recorded using a Fischione High Angle Annular Dark Field (HAADF) detector. A FEI super-X EDX detector was used, which provides a solid collection angle of 0.9 sterads fully integrated with the A-TWIN objective lens. The elemental maps were recorded with a 20μs dwell time per pixel (spectra) with a 1K × 1K frame size (approximately 20 s scan time per total frame). As plastic sections are susceptible to shrinkage and beam damage, care was taken not to expose the sample outside of the acquisition period by blanking the beam when no image was acquired. Before scanning, samples were pre-eradiated in TEM mode at low magnification for 30 min to reach a stable state with respect to shrinkage. During and after the period of acquisition, the scanning area was assessed with respect to the contrast of the image, which was taken as a measurement of stability. No additional shrinkage or carbon deposit occurred during the scanning period as the contrast of the image did not change after each consecutive scan. We assumed therefore that, after pre-eradiation, the sample was in a stable state and no additional mass loss occurred, nor contamination deposited on the sample during acquisition, showing that both the beam current was acceptable for the imaging and the column contamination was negligible. The spectra were processed using Bruker Esprit software and the elemental map of Fe was calculated. The Energy Dispersive X-ray (EDX) maps were processed with an opening filter followed by a Gaussian smoothing.

## **PLASMID CONSTRUCTION AND LUCIFERASE REPORTER ASSAY**

The promoter region of *FPN1* spanning from 845 bp 5 to the ATG codon to 50 bp after *FPN1* transcription initiation site was amplified and ligated into the pGL4 luciferase reporter vector (Promega). pGL4 has an SV40 promoter, driving the transcription of a luciferase chimeric mRNA which includes the *FPN1* 5 UTR. The *FPN1* 5 CAGUG IRE sequence, which is involved in the formation of the terminal –AGU- loop (Hentze and Kuhn, 1996) was mutagenized into CCCCG with the Quickchange® II site direct mutagenesis kit (Stratagene), according to the manufacturer's protocol. The luciferase chimeric constructs containing the wt and mutated *FPN1* IREs were transfected into CD4+ and CD8+ cells using AMAXA nucleofection, as previously described (Pinto et al., 2010). The day after nucleofection, cells were incubated for 2 h with 5μM of Fe-citrate (5:100), 5μM of Fe-citrate (5:100) + 5μM of DFO or with RPMI, and lysed. The luciferase activity was measured with the Dual-Luciferase Reporter Assay System (Promega), according to the manufacturer's instruction. The luciferase activity was normalized with renilla luciferase activity, used as transfection control. To control for changes in promoter-driven transcription, luciferase mRNA levels were assessed in parallel, with no significant changes observed between each experimental condition (data not shown).

## **GENE EXPRESSION**

Total RNA was extracted using the RNeasy Plus Mini kit (Qiagen), with on-column DNA digestion (Qiagen). cDNA was synthesized using the Superscript First-Strand Kit (Invitrogen) and qRT-PCR was performed in an iCycler iQ5 PCR detection system (Bio-Rad), using specific primers. Glyceraldehyde-3-phosphate dehydrogenase (*GAPDH*) and 18S rRNA mRNA expression were used as internal controls. Since no significant differences were observed between the two controls, all qRT-PCR results displayed were normalized for *GAPDH* mRNA expression. Melting curve experiments previously established that the signal for each amplicon was specific and not derived from primer dimers. For every gene a series of four serial dilutions was used during optimization of the procedure. Relative expression levels were calculated as 2−--*CT* [--CT = (Gene of interest CT − *GAPDH* CT)sample − (Gene of interest CT − *GAPDH* CT)reference]. All experiments involving qRT-PCR were performed at least in triplicate, with two to three replicates each.

## **EXPERIMENTAL PROCEDURES WITH MOUSE MODELS**

All mice models used were 7–8 week old females, maintained at the IBMC's Animal Care Facility. C57BL/6 mice were supplied by Charles River or Jackson Laboratory. *Foxn1nu* and *Foxn1*+/<sup>−</sup> were purchased from the Jackson Laboratory. All experimental procedures were approved by and performed according to the guidelines of the IBMC's Animal Care Committee.

For the analysis of the effect of exogenous hepcidin on ferroportin expression mice (*n* = 6) were maintained on an iron-sufficient diet (35 mg/kg chow; Harlan Laboratories) from weaning and intravenously injected daily with mouse hepcidin (Peptides International; 50μg/mouse in 0.3 ml PBS) for 3 consecutive days. The control group (*n* = 6) was injected with PBS only. Animals were killed 3 days after the first dose injection and peripheral blood CD3+ cells isolated with MACS, as described above.

To study the effect of endogenous hepcidin on ferroportin expression mice (*n* = 12/experimental condition) were maintained on the iron-sufficient diet from weaning or on iron-rich diet (iron-sufficient diet enriched with 2.5% iron carbonyl—Sigma) for 2 weeks. Mice were then intravenously injected each 24 h with 100μl of 50μM 55Fe-citrate (50:100), to achieve a systemic, traceable, 55Fe-citrate concentration of 5μM. Two days after the first 55Fe injection mice were sacrificed and livers and peripheral blood CD3+ cells were collected. *Hamp1* mRNA expression was quantified in the livers by qRT-PCR and Fpn1 expression and intracellular 55Fe accumulation quantified in CD3+ T lymphocytes, as described above.

The effect of each diet on blood iron-related parameters was assessed in 4 animals/experimental condition by analyzing levels of serum iron, total iron binding capacity, and resulting transferrin saturation by standard routine procedures at the Clinical Chemistry Laboratory of Santo António Hospital (Porto).

For the study of the modifier role of T lymphocytes in NTBI accumulation *Foxn1nu* and *Foxn1*+/<sup>−</sup> mice were fed the iron-rich diet for 2 weeks and intravenously injected, every 2 days, with 100μl of a 50μM 55Fe-citrate (50:100) solution, during 7 days. The animals were killed 8 days after the first dose injection and the livers and spleen collected. Incorporation of 55Fe was quantified following the same procedure described in *NTBI uptake*.

For T lymphocyte reconstitution of *Foxn1nu* mice, CD3<sup>+</sup> lymphocytes were isolated from the spleens of *Foxn1*+/<sup>−</sup> animals (on iron-sufficient diet) using MACS, resuspended in 100μl of endotoxin-free PBS and intravenously injected via the lateral tail vein (5 <sup>×</sup> <sup>10</sup><sup>6</sup> or 15 <sup>×</sup> <sup>10</sup><sup>6</sup> T lymphocytes/injection) on *Foxn1nu* mice that had been fed the iron-sufficient diet from weaning or the iron-rich diet for 2 weeks. Control (Mock) animals were injected with an equal volume of PBS. The purity of the T lymphocyte fraction was verified by FACS, using an FITC-labeled anti-mouse CD3 antibody (BD Pharmingen) and confirmed to be >93%. Immediately following T lymphocyte transfer mice on iron-rich diets were injected with 100μl of a 50μM 55Fe-citrate (50:100) solution, which was repeated 24 h later. Forty-eight hours after T lymphocyte transfer all mice were sacrificed and the levels of 55Fe in the liver and spleen quantified. For quantification of hemosiderin deposition, livers were fixed in 4% formaldehyde, embedded in paraffin, mounted onto slides, and stained with Prussian blue and hematoxylin-eosin counterstain, using standard procedures. Hemosiderin was detected using the Perls' Prussian Blue method.

## **STATISTICAL ANALYSIS**

The results are expressed as mean values ± 1 standard deviation (SD). Statistical differences between means were calculated using the Student's unpaired *t*-test. For experiments involving multiple comparisons, One-Way or Two-Way analysis of variance (ANOVA) was used. When significant differences were detected, the data were re-analyzed using the Fisher's least significance difference test. Statistical significance was set at *P* < 0.05. All statistical analyses were performed using STATGRAPHICS Centurion XV (Statpoint Technologies).

#### **RESULTS**

## **NTBI ACCUMULATION BY HUMAN PERIPHERAL BLOOD T LYMPHOCYTES**

To test the ability of T lymphocytes to store iron acquired as NTBI, CD4+, and CD8+ T lymphocytes were incubated with 5μM of Fe-citrate for 0–12 h and the expression of the iron storage protein Ferritin H (FTH) quantified by Western blot analysis. We observed that both T lymphocyte populations increase FTH levels in response to Fe-citrate, an effect abrogated by the supplementation of the culture medium with 5μM desferrioxamine (DFO) (**Figure 1A**). The changes in FTH expression are in agreement with our previous results showing that a plateau in intracellular iron is reached earlier than 6 h in the presence of a constant NTBI concentration (Arezes et al., 2013) and suggest the capacity of the cells to store a fraction of the intracellular iron in the Ferritin core.

The fraction of iron that is not included in the storage compartment should be associated with proteins other than ferritin (*functional iron*) or enter the *transit iron* pool, also known as Labile Iron Pool (LIP). Using the calcein assay, which is based on the quenching of the fluorescence of the permeant calcein acetoxymethyl ester by iron, we assessed the contribution of NTBI to the LIP in T lymphocytes. We observed a significant time-dependent decrease in the fluorescence of intracellular calcein exclusively for T lymphocyte populations exposed to 5μM of Fe-citrate, indicating an increase in intracellular labile iron (**Figure 1B**). The difference between iron-loaded and control cells reaches its maximum values after 30 min of incubation, suggesting that an equilibrium is reached at this point between the formation of new labile iron and incorporation of iron in the storage pool + iron export. Validation of the specific effect of iron in the fluorescence quenching of calcein was obtained by recovering fluorescence exclusively on iron-treated cells upon incubation with the membrane-permeable iron chelator SIH.

To further characterize intracellular iron accumulation in T lymphocytes we used EDX analysis to generate maps of intracellular iron distribution in ultrathin sections of cells incubated with 5μM Fe-citrate (5:100) for 24 h. Using this approach we confirmed that T lymphocytes accumulate iron from Fe-citrate and showed that intracellular iron shows distinct distribution patterns, ranging from homogeneously-sized particles with an area of approximately 20 nm<sup>2</sup> to large aggregates with a more amorphous appearance, some of which >1μm in diameter (**Figure 2A**). A more detailed analysis of the iron-rich structures showed that their iron content is highly variable (**Figures 2B,C**), revealing also a consistent enrichment in nitrogen, oxygen, phosphorous, and chlorine (**Figure 2C**), which suggests the association of intracellular iron with other compounds, that could include phosphate and proteins.

## **NTBI EXPORT BY HUMAN PERIPHERAL BLOOD T LYMPHOCYTES** *Iron export*

In order to have a relevant role as a component of the *iron storage* compartment, T lymphocytes would need to selectively retain or, alternatively, export intracellular iron acquired as NTBI according to systemic signals. To test this hypothesis we analyzed the export of iron acquired as Fe-citrate by T lymphocytes. CD4+ and CD8+ T cells were exposed to 5μM Fe-citrate for 2 h, a time point at which cells have reached the maximum iron content (Arezes et al., 2013), and allowed to export iron into an ironfree medium for up to 6 h. We observed that iron export by T lymphocytes follows a linear pattern, with a delay in CD8+ cells,

expression by NTBI in T lymphocyte populations. Western blot quantification of Ferritin H (FTH) expression in CD4+ and CD8+ cells incubated with 5μM of Fe-citrate (upper panel) or Fe-citrate + 5μM DFO (middle panel) for 0–12 h. β-actin expression was used as loading control. Images are representative of three experiments without significant inter-experimental variation. Bottom panel: Scanning densitometry of FTH Western blot. Data are normalized for β-actin levels and are expressed as arbitrary units (AU). ∗*P* < 0.05 and ∗∗*P* < 0.005 (One-Way ANOVA) between relative FTH levels at 0 h and posterior time-points, for each population and experimental condition.

which only match CD4+ lymphocyte's export rate after 60 min of export (**Figure 3A**). After 60 min in an iron-free medium, iron export by T lymphocytes corresponds to approximately 3% of intracellular levels, demonstrating a slow release of iron acquired as NTBI by these cells. This is confirmed by the quantification of intracellular 55Fe remaining in both T lymphocyte populations throughout time, which further shows that, after 72 h of export, T lymphocytes maintain approximately 20% of the initial iron load acquired as NTBI (**Figure 3B**).

## *Role of FPN1*

The trans-membrane domain protein SLC40A1 (ferroportin, FPN1) is the only known cellular iron exporter (Donovan et al., 2005). Little is known about the expression of FPN1 by T lymphocytes and evenless onits regulation by cellular and systemic stimuli. We observed clear expression of the protein in membrane extracts of untreated CD3+ cells, while it was undetectable in the cytosolic fraction (**Figure 3C**). The high purity of the cell fractionation was further confirmed by the clear detection of the plasma membrane marker protein alpha 1 sodium potassium ATPase in one of the membrane extracts depicted in **Figure 3C**, with only residual detection in the cytoplasmic extracts (**Supplementary Figure 1**). Expression of FPN1 increased significantly in T lymphocytes after

lymphocytes incubated with 5μM of Fe-citrate or 5μM of Na-citrate (Mock). Each point represents the average (*n* = 3) ± 1 SD. ∗∗*P* < 0.005 (One-Way ANOVA) between fluorescence levels at 0 h and posterior time-points, for each population and experimental condition. Bottom panel: following incubation with 5μM of Fe-citrate for 180 min, 10μM of SIH were added to cells (arrow) and changes in calcein fluorescence monitored for 30 min. Data from a representative experiment (*n* = 3), with minimum inter-experimental variation.

exposure to 5μM Fe-citrate for 3 h and are maintained at least up to 6 h of incubation (**Figure 3D**). No changes in FPN1 were observed in cells supplemented with PBS, confirming the role of iron in the observed FPN1 increase. Altogether, these results suggest the involvement of ferroportin in the trans-membranar transport of iron in T lymphocytes.

Although 5μM Fe-citrate is in the range of the NTBI concentrations commonly detected in iron overload disorders, in severe cases of iron overload values as high as 30μM have been reported (Batey et al., 1980). To test whether iron export by T lymphocytes is dependent on the extracellular NTBI concentration, cells were incubated with increasing doses of Fe-citrate (XFe:100μM citrate) for 2 h and allowed to export iron into an iron-free medium. We observed a dose-dependent increase in iron export with increasing NTBI concentrations (**Figure 3E**). Interestingly, the increase in iron export cannot be accounted by augmented FPN1 levels, as no significant changes in FPN1 expression were detected between incubations with 20 and 100μM Fe-citrate (**Figure 3F**). The increased iron export might thus be caused by an increase in the flux of Fe(*II*) ions transported by each FPN1 molecule rather than to a augmented number of FPN1 molecules at the cell surface.

In hepatocytes and macrophages intracellular iron regulates FPN1 translation, by a mechanism involving the IRP/IRE

X-ray analysis of a representative image showing intracellular cytoplasmic iron in a CD3+ cell. White represents iron. Numbered areas analyzed for elemental composition; "1" represents the total area depicted. **(B)** Elemental composition of the four areas depicted in **(A)** with the peak

machinery (Lymboussaki et al., 2003). To test if this mechanism is involved in the response of FPN1 to NTBI in T lymphocytes, we cloned the FPN1-5 UTR in the pGL4 vector, in which a SV40 promoter drives transcription to produce a luciferase chimeric mRNA with the FPN1-5 UTR as its 5 UTR. In CD4+ and CD8+ cells nucleofected with the FPN1/luciferase construct, the expression of luciferase increased approximately 2-fold following incubation with 5μM Fe-citrate for 2 h (**Figure 3G**), in comparison with cells incubated in iron-free medium or in medium containing Fe-citrate + DFO. This result matches the increase in FPN1 observed by immuno-blot (**Figure 3D**). Sequential mutagenesis of the FPN1-5 -UTR previously shown to impair IRP binding to FPN1-IRE (Hentze and Kuhn, 1996) resulted in a significant increase (approximately 4-fold) of the basal FPN1 expression in the two cell populations and, importantly, abrogated the response to Fe-citrate. Together, these results confirm the involvement of the IRE/IRP system in the NTBI-mediated increase in FPN1 expression. This post-transcriptional mechanism is probably the major contributor to the FPN1 increase in response to NTBI, as no changes in *FPN1* mRNA expression were observed in response to 5μM Fe-citrate (**Supplementary Figure 2**), arguing against a transcriptional regulation of FPN1 by NTBI in these cells.

## *The hepcidin-FPN1 axis*

S = sulfur, Cl = chlorine, Fe = iron.

Besides the IRE/IRP translational regulation, FPN1 expression is also regulated post-translationally by the hormone hepcidin, which is secreted mainly by the liver in response to several stimuli, including iron overload (Nemeth et al., 2004). Since the characterization of iron export above was performed in a hepcidin-free medium, which does not correspond to most *in vivo* situations in which NTBI is present, we re-analyzed iron export by T lymphocytes, using the same experimental setup, in the presence of human synthetic hepcidin at doses present in human serum in normal and in iron-overload conditions (Ganz et al., 2008). Exogenous hepcidin caused a dose-dependent decrease in FPN1 levels in both T lymphocyte populations (**Figure 4A**) together with a significant dose-dependent increase in iron retention over 72 h (**Figure 4B**). To control for possible hepcidin-mediated changes in NTBI uptake that could explain the iron-retention results we measured iron uptake in T lymphocytes incubated with 5μM of 55Fe-citrate for 30 min, in the presence of human hepcidin. No significant differences were found between hepcidin-supplemented and hepcidin-free conditions (**Supplementary Figure 3**), which leads us to ascribe the effect of hepcidin exclusively on iron export.

second per ev (cps/ev). **(C)** Relative (%) elemental composition of regions signaled in **(A)**; C = carbon, N = nitrogen, O = oxygen, P = phosphorous,

**FIGURE 3 | Iron export by T lymphocytes. (A)** Time-dependent iron export by T lymphocytes. Each point is a mean value (*n* = 3) ± 1 SD. **(B)** Impact of iron export in intracellular Fe levels. Each point is a mean value ± 1 SD of two experiments each with three replicates. **(C)** Ferroportin (FPN1) expression in CD3+ T lymphocytes. Membrane (M), cytoplasmic (C), and total (T) extracts of CD3+ cells were blotted and membranes incubated with anti-FPN1 and anti-β-actin antibodies. Results obtained with CD3+ cells from three different blood donors are presented. **(D)** Western blot quantification of FPN1 expression in CD4+ and CD8+ T lymphocytes incubated with 5μM of Fe-citrate (upper panel) or PBS (middle panel) for 0–6 h. β-actin expression was used as loading control. Images are representative of four experiments without significant inter-experimental variation. Bottom panel: Scanning densitometry of FPN1 Western blots. Data are normalized for β-actin levels and are expressed as arbitrary units (AU); ∗*P* < 0.05 (One-Way ANOVA) between relative FPN1 levels at 0 h and posterior time-points, for each

The previous results show that iron export by T lymphocytes is virtually totally abolished following exposure to hepcidin levels typically present in iron overload and inflammation contexts (600 ng/ml), suggesting that systemic hepcidin levels may significantly modify the amount of iron which is released by T lymphocytes back into the circulation. To test this hypothesis, C57Bl/6 mice were intravenously injected with mouse synthetic hepcidin, or PBS, for 3 consecutive days and Fpn1 expression quantified in peripheral blood CD3+ cells. Confirming the *ex vivo* results, hepcidin injections caused a marked decrease of Fpn1 in mouse peripheral blood T lymphocytes (**Figure 4C**). To further explore the involvement of systemic hepcidin in the control of iron export by T lymphocytes, C57Bl/6 mice were population and experimental condition. **(E)** Dose-dependent iron export by T lymphocytes. Each point is the mean of two experiments with two replicates each. **(F)** Western blot quantification of FPN1 expression in T lymphocytes incubated with different Fe-citrate concentrations. β-actin expression was used as loading control. Images are representative of three experiments without significant inter-experimental variation. Bottom panel: Scanning densitometry of FPN1 Western blot. Data are normalized for β-actin levels and are expressed as arbitrary units (AU). ∗*P* < 0.05 (One-Way ANOVA) between relative FPN1 levels at 0 h and posterior time-points. **(G)** Involvement of the IRE/IRP system in the NTBI-induced modulation of FPN1 in T lymphocytes. Cells were transfected with a chimeric construct including wild-type (WT) or mutated (Mut) FPN1-IRE and incubated with Fe-citrate (Fe-Cit), Fe-Cit + DFO or RPMI (No-Fe) and the luciferase activity quantified. Each point is a mean value (*n* = 3) ± 1 SD. <sup>∗</sup>*P* < 0.05 (One-Way ANOVA) for each experimental condition relative to WT IRE-No Fe control.

fed an iron-rich diet for 2 weeks, in order to saturate transferrin and induce an increase in systemic hepcidin levels. Controls were fed a diet previously shown to be iron-sufficient (Ramos et al., 2011). As expected, dietary iron induced a significant increase in serum iron parameters, including transferrin saturation, which increased from 53 ± 8% in control animals to 87 ± 4% in mice on iron-rich diet (**Table 1**). Two weeks later, all mice were intravenously injected with 100μl of 50μM 55Fecitrate (50:100), to achieve a systemic 55Fe-citrate concentration of 5μM. Analysis of the animals 48 h after injection showed, as expected, a significant increase (about 20-fold) of liver *Hamp1* mRNA levels in iron-loaded mice (**Figure 4D**). Most importantly, peripheral blood CD3+ cells in iron-overloaded mice had

**FIGURE 4 | Modulation of FPN1 expression and iron export by hepcidin in human and mouse T lymphocytes. (A)** Modulation of Ferroportin (FPN1) by hepcidin in human T lymphocytes. Western blot quantification of FPN1 in CD4+ and CD8+ T lymphocytes incubated with the indicated concentrations (ng/ml) of hepcidin for 24 h. β-actin expression was used as loading control. Images are representative of three experiments without significant inter-experimental variation. Bottom panel: Scanning densitometry of FPN1 Western blot. Data are normalized for β-actin levels and are expressed as arbitrary units (AU). ∗∗*P* < 0.005 (One-Way ANOVA) between relative FPN1 levels in Mock- and hepcidin-supplemented conditions. **(B)** Modulation of iron retention by hepcidin. Each point represents the mean value ± 1 SD (*n* = 3) of intracellular iron in human T lymphocytes incubated with 5μM of 55Fe-citrate for 2 h and allowed to export iron in an NTBI-free medium, in the presence (100 or 600 ng/ml) or absence (Mock) of hepcidin, for up to 72 h. ∗*P* < 0.05

and ∗∗*P* < 0.005 (One-Way ANOVA) between the time-dependent distribution of iron levels in each experimental condition and the no-iron control, for each cell population. **(C,D)** *In vivo* modulation of mFpn1 by hepcidin and iron in mice T lymphocytes. **(C)** Western blot quantification of mFpn1 expression in T lymphocytes from C57Bl/6 mice injected with synthetic mouse hepcidin (+ Hepcidin) or with PBS (Mock). β-actin expression was used as loading control. Each lane represents a pool of CD3+ cells from 3 animals. **(D)** Upper panel: qRT-PCR quantification of *Hamp1* expression in livers (*n* = 3 per bar) of C57Bl/6 mice fed iron-rich (+Fe) or iron-sufficient (Mock) diets followed by injection of 50μM 55Fe-citrate. Western blot quantification of Fpn1 expression (*middle panel*) and intracellular 55Fe quantification (*bottom panel*) in peripheral blood T lymphocytes collected from the same animals. In middle and bottom panels data from each lane represents a pool of T lymphocytes from 6 animals. ∗∗*P* < 0.005 (Student's *t*-test).

a marked reduction in Fpn1 expression and significantly higher intracellular 55Fe levels. Altogether, these results demonstrate, *in vivo*, the role of hepcidin in the control of NTBI retention vs. export by peripheral blood T lymphocytes.

## *Systemic impact of NTBI uptake by T lymphocytes*

The results above led us to hypothesize that NTBI storage by T lymphocytes may represent a modifier of accumulation of this iron form in target organs. To test this hypothesis, we first



*UIBC, unsaturated iron-binding capacity; TIBC, total iron-binding capacity; Tf, transferrin.*

compared iron accumulation in target organs of T lymphocytedeficient *Foxn1nu* and T lymphocyte-normal *Foxn1*+/<sup>−</sup> mice fed an iron-rich diet in which traceable NTBI was generated by intravenous injection of 55Fe-citrate. We observed significantly higher 55Fe levels in the livers than in the other organs analyzed, confirming that the traceable iron in our experimental setup mostly incorporates the NTBI pool (**Figure 5A**). In accordance with previous results obtained with other lymphocyte-deficient models (de Sousa et al., 1994; Santos et al., 2000; Cardoso et al., 2002), there is a significant increase in iron accumulation in livers and pancreas of *Foxn1nu* mice, in comparison with *Foxn1*+/<sup>−</sup> controls. No significant differences were observed for the spleen and heart. Next the specific involvement of T lymphocytes in this result was tested by reconstituting a group of *Foxn1nu* animals in iron-rich diets with 5 <sup>×</sup> <sup>10</sup><sup>6</sup> or 15 <sup>×</sup> <sup>10</sup><sup>6</sup> T lymphocytes isolated from the spleens of histocompatible *Foxn1*+/<sup>−</sup> animals on iron-normal diets. In *Foxn1nu* animals reconstituted with 5 <sup>×</sup> <sup>10</sup><sup>6</sup> T lymphocytes liver iron levels are approximately 20% lower than in non-reconstituted controls, a difference which increases to approximately 35% when 15 <sup>×</sup> <sup>10</sup><sup>6</sup> T lymphocytes were used (**Figure 5B**). The same result, although with lower statistical significance, was observed for the spleen. Histological analysis of *Foxn1nu* liver sections showed no significant hemosiderin deposition in the livers of *Foxn1nu* or T lymphocyte-reconstituted *Foxn1nu* mice on iron-normal diets (**Figure 5C**), but high levels of stainable iron were detected in the liver parenchyma of *Foxn1nu* iron-loaded animals. Confirming the results obtained with 55Fe, reconstitution of the T lymphocyte pool with 15 <sup>×</sup> <sup>10</sup><sup>6</sup> T lymphocytes reduced the levels of liver stainable iron. Changes in liver iron deposition in response to T lymphocyte reconstitution cannot be ascribed to alterations in intestinal iron absorption or iron retention induced by hepcidin, since no changes in hepcidin mRNA levels were observed in response to T lymphocyte transfer (**Supplementary Figure 4**). Altogether, these results demonstrate the physiological role of T lymphocytes as modifiers of NTBI deposition in target organs, which schematic representation is proposed in **Figure 6**.

## **DISCUSSION**

The major players and mechanisms in systemic iron homeostasis classically consider four major compartments, involving particular cell types. The *functional* compartment (erythroid precursors or other proliferating cell pools), the *uptake* compartment (enterocytes), the *recycling* compartment (spleen macrophages), and the *storage* compartment (hepatocytes and macrophages). The described model, however, neglects the contribution of all remaining cell types which may contribute to the fine-tuning of the homeostasis of this essential yet dangerous element. In this work we demonstrate the capacity of peripheral blood circulating T lymphocytes to store iron acquired in the form of NTBI, which is the iron presentation associated with toxicity in ironoverload disorders. In addition, we describe how this capacity, modulated by the hepcidin-ferroportin axis, has important physiological consequences. By doing so, we introduce the concept of the "*circulating storage* compartment" and establish a role for T lymphocytes as important players in this new component of iron homeostasis.

The capacity to take up NTBI has been demonstrated for a variety of circulating blood cells, including reticulocytes and erythrocytes (Zhang et al., 2008; Prus and Fibach, 2011), lymphocytes (Arezes et al., 2013), monocytes, eosinophils, basophils, neutrophils, and platelets (Pinto et al., unpublished results and Hausmann et al., 1988). It is thus possible that all or at least some of these cell types play a role in the buffering of NTBI from target organs. That will depend not only on the quantitative NTBI retention capacity but also on the selectivity of each cell type for the NTBI species present in circulation in each particular context. Among these cells, T lymphocytes constitute a particularly interesting population and an obvious target for analysis. They are known for some time to be negatively associated with the severity of iron overload both in mice (de Sousa et al., 1994; Santos et al., 1996; Cardoso et al., 2002) and in HFE-HH human patients (Porto et al., 1997; Barton et al., 2005; Cruz et al., 2006), and we show now that they possess the necessary features to be acknowledged as important players in NTBI homeostasis:

#### **FAST AND SELECTIVE NTBI IMPORT**

NTBI clearance from the circulation has been previously described to be a rapid and efficient procedure. Once in contact with the liver, NTBI is cleared with a half-life of <30 s, in comparison with approximately 50 min for TF-bound iron (Brissot et al., 1985). The same rapid pattern of NTBI uptake was previously described by us for T lymphocytes, although these cells show a lower capacity to accumulate NTBI than hepatocytes (Parkes et al., 1995; Arezes et al., 2013).

#### **IRON UTILIZATION/STORAGE AFTER EXPOSURE TO NTBI**

We show here an increase in the LIP of T lymphocytes exposed to NTBI. The nature and distribution of LIP in any cellular system remains essentially unknown. Our results with EDX analysis in T lymphocytes show a heterogeneous distribution of intracellular iron and the enrichment of iron-rich structures with nitrogen, oxygen, phosphorous and chlorine, suggesting the association of iron with compounds such as phosphate and proteins. This association may result in protection of the cell from oxidative damage mediated by labile iron, an hypothesis supported by previous studies showing the lack of NTBI-induced ROS in Jurkat cells that accumulate NTBI as phosphate nanoparticles (Jhurry et al., 2012, 2013).

Intracellular iron incorporated in the LIP fraction can either be utilized in iron-requiring processes or stored, mostly associated

with ferritin. Evidence that T lymphocytes are equipped to integrate iron acquired as NTBI in the cell's metabolism is provided in our previous study showing an increase in activationmediated proliferation in response to this iron form (Arezes et al., 2013). Evidence for iron storage is provided here by the observation of an NTBI-induced up-regulation of the ferritin levels in T lymphocytes, suggesting an increase in the iron-storage capacity of these cells in response to NTBI uptake. The few studies previously addressing the response of ferritin to NTBI in lymphocytes show disparate results and their interpretation is hindered by the use of distinct iron donors, most of them non-physiological, and of frequently non-physiological NTBI concentrations (Pelosi et al., 1986; Seligman et al., 1991; Djeha and Brock, 1992; Chitambar and Wereley, 2001). We believe that the use of ferric citrate as iron donor, the maintenance of citrate concentrations between the physiological interval of 60–140μM (Lentner, 1984) and the use of iron concentrations in the range of those

lymphocytes or injected with PBS (Mock). Data are depicted as lower

found commonly in iron overload situations (Evans et al., 2008) constitute the correct experimental setup and should become the standard procedure for future studies involving biological systems and NTBI.

## **NTBI EXPORT IN A REGULATED MANNER**

Magnification = 100×.

We demonstrate that iron acquired as NTBI by T lymphocytes is exported via ferroportin in an IRP/IRE-regulated manner. *In vitro*, and in the absence of exogenous hepcidin, both T lymphocyte populations (CD4+ and CD8+) export approximately 3% per hour of intracellular Fe, which compares with what has been reported for monocytic/macrophagic cell lines—approximately 6% (Ludwiczek et al., 2003). However, these values might not represent the *in vivo* iron export of these cells, as the demonstration of the modulation of T lymphocyte ferroportin, and concomitant Fe export, by circulating hepcidin, places the effective iron export of T lymphocytes,

and thus their NTBI storage capacity, under the systemic control. This finding, together with a previous study reporting the expression of ferroportin by erythroblasts and its modulation by hepcidin (Zhang et al., 2011), places this protein in an even more central place in iron homeostasis and highlights the need to consider the contribution of other "non-classical" players in pathologies associated with inappropriate expression of hepcidin. Nevertheless, these results do not preclude the existence of alternative iron export pathways, such as iron exported in association with ferritin, a mechanism first proposed for T lymphocytes (Dörner et al., 1980) and later described for macrophages (Cohen et al., 2010) and hepatocytes (Nemeth, pers. comm.).

Our results show that CD4+ and CD8+ T lymphocytes do not differ significantly in the export/retention of iron acquired as NTBI. Previous results showing iron overload in mice deficient in CD8+ T lymphocytes (de Sousa et al., 1994; Santos et al., 1996; Cardoso et al., 2002) and the inverse correlation recurrently observed between CD8+ T lymphocyte numbers and the severity of iron overload in HFE-HH human patients (Porto et al., 1997; Barton et al., 2005; Cruz et al., 2006) might thus need to be reinterpreted not as an indication that CD8+ T cells play a unique role in iron homeostasis but instead that the lack or a reduction in number of these cells affects the fine-tuning of iron homeostasis, namely NTBI distribution. This conclusion is supported by previous results showing higher severity of iron overload in β2m−Rag1<sup>−</sup> mice—deficient in CD4<sup>+</sup> and CD8<sup>+</sup> T lymphocytes and B lymphocytes—than in β2m<sup>−</sup> animals deficient in CD8+ T lymphocytes (Santos et al., 2000). The lack of differences in NTBI retention between CD4+ and CD8+ cells is also in agreement with our previous findings of similar NTBI uptake by the two cell types (Arezes et al., 2013). In that study we did not find evidence for the involvement of neither DMT1 nor ZIP14, the most likely NTBI transporter candidates (Trinder et al., 2000; Liuzzi et al., 2006), leaving the door open for the identification of a still elusive transport mechanism.

Here we demonstrate the ability of transferred T lymphocytes to retain circulating NTBI and thus reduce its accumulation in the liver and spleen. This result cannot be ascribed to hepcidin-mediated changes in iron metabolism induced by T lymphocyte transfer, since (1) no significant differences in liver hepcidin expression were found in response to transference alone and (2) no visible changes in liver iron deposition were observed between non-transferred and transferred animals on an iron-sufficient diet. This is in agreement with a previous report showing the absence of alterations in iron homeostasis in NOD/SCID mice transferred with syngeneic T lymphocytes (Bair et al., 2009). The lack of increased iron deposition in the livers of T lymphocyte-depleted animals on an iron sufficient diet also illustrates one important conclusion from our results which is that the absence of T lymphocytes *is not the cause* of iron overload and NTBI accumulation but instead that, by acting as a first line of retention, T lymphocytes *are modifiers* of NTBI accumulation when this iron form is present in the blood circulation.

At the present time we cannot exclude the possibility that other mechanisms besides NTBI storage could, at least in part, underlie the observed reduction in organ iron deposition. An alternative explanation is the synthesis of cytokines or other molecules by T lymphocytes upon contact with iron, which could in turn impact, directly or indirectly, on the handling of iron by other tissue/cell types, as previously reported (Ten Elshof et al., 1999; Meyer et al., 2002; Sharma et al., 2009).

The capacity of T lymphocytes to store and release NTBI may have implications beyond their capacity to modify systemic iron overload. It is reasonable to consider that at the local tissue level they could interfere with the availability of iron in the extracellular milieu (as proposed in **Figure 6**) and thus influence the growth rate of adjacent cells, playing a role in normal cell/tumor growth and tissue remodeling, such as already demonstrated for polarized M2 macrophages (Recalcati et al., 2012). This report does not address the impact of NTBI uptake in the *in vivo* T lymphocyte metabolism besides iron storage and export. Although in our experimental conditions T lymphocytes retained their viability and proliferation potential, at this point we cannot exclude the possibility that accumulation of iron from NTBI may interfere with other normal T lymphocyte functions and immune surveillance. Previous studies have shown that lymphocytes from iron-overloaded animals have a reduced capacity to generate allo-specific cytotoxic responses (Good et al., 1987) and that ferric salts can alter a variety of lymphocyte functions *in vitro* (Nishiya et al., 1980; Brock, 1981; Bryan et al., 1981, 1986; van Asbeck et al., 1982; Bryan and Leech, 1983). In addition, specific abnormalities in CD8+ T lymphocyte functions have been described in HFE-HH patients, including defective lymphocyte-specific protein tyrosine kinase (p56lck) activity, decreased cytotoxic activity, a decreased number of CD8+ T cells expressing the costimulatory molecule CD28, but also an increased number of CD8+ T cells lacking CD28, and an abnormally high percentage of HLA-DR-positive activated T cells (Arosa et al., 1994, 1997; Arosa, 2002).

In conclusion, this work demonstrates that T lymphocytes are important components of a circulating NTBI storage compartment and show its physiological relevance as modifiers of tissue iron overload. On a broader scope, it provides a mechanistic support for the possibility of circulating T lymphocytes acting *in vivo* as a key component in systemic iron homeostasis, by playing a role in the surveillance of iron toxicity and of its possible use by pathogens and tumor cells, as first postulated 32 years ago (de Sousa, 1981).

## **AUTHOR CONTRIBUTIONS**

Jorge P. Pinto conceptualized the idea, designed and performed the research, analyzed data, and wrote the paper; João Arezes performed the research, analyzed data and wrote the paper; Vera Dias, Susana Oliveira, Inês Vieira, Mónica Costa, Matthijn Vos, Anna Carlsson, and Yuri Rikers performed the research; Maria Rangel analyzed data and wrote the paper; Graça Porto contributed with vital analytical tools, contributed to design of the research, analyzed data and wrote the paper.

## **ACKNOWLEDGMENTS**

We are especially grateful to Maria de Sousa (University of Porto, Portugal), who was the first to suggest a role for lymphocytes in the control of iron toxicity and was a constant inspiration for the present work. We are also grateful to Caroline Enns and An-Sheng Zhang (Oregon Health & Science University, Portland, USA) and to Elizabeta Nemeth (UCLA, USA) for helpful discussions and suggestions during the preparation of this manuscript. This work was funded by FEDER funds through the Operational Competitiveness Programme—COMPETE, by national funds through FCT—Fundação para a Ciência e a Tecnologia under the projects FCOMP-01-0124-FEDER-015823 (PTDC/SAU-MET/113011/2009), FCOMP-01-0124-FEDER-007046 (PTDC/ BIA-BCM/66818/2006), FCOMP-01-0124-FEDER-037277 (PEst-C/SAU/LA0002/2013), by the INOVA Foundation and by the American Portuguese Biomedical Research Fund (APBRF). João Arezes, Vera Dias, Susana Oliveira, Inês Vieira, and Mónica Costa are recipients of FCT fellowships (www.fct.pt). Jorge P. Pinto is supported by Programa Ciência (sponsored by POPH-QREN (4.2), with match-funding from the European Social Fund and Portuguese funds from the MCTES).

## **SUPPLEMENTARY MATERIAL**

The Supplementary Material for this article can be found online at: http://www.frontiersin.org/journal/10.3389/fphar.2014. 00024/abstract

**Supplementary Figure 1 | Assessment of the purity of the plasma membrane fraction from human CD3**<sup>+</sup> **cells.** Membrane (M; *n* = 1), cytoplasmic (C; *n* = 3), and total (T; *n* = 1) protein fractions from human CD3+ cells were blotted and the membrane incubated with the plasma membrane marker anti-alpha 1 Na/K ATPase antibody.

**Supplementary Figure 2 | Effect of NTBI on** *FPN1* **mRNA expression in human peripheral blood T lymphocytes.** qRT-PCR analysis of *FPN1* mRNA expression in human peripheral blood CD4+ and CD8+ T lymphocytes incubated with 5μM Fe-citrate or PBS for up to 6 h. Data are depicted as fold change relative to *FPN1* mRNA levels at 0 h for each cell population. Experiments were performed three times with 2 replicates per experiment. Each bar represents the mean ± 1 SD.

**Supplementary Figure 3 | Effect of hepcidin on Fe-citrate uptake by T lymphocytes.** Intracellular 55Fe levels in CD4<sup>+</sup> and CD8<sup>+</sup> T-lymphocytes incubated with 5μM of 55Fe-citrate, for 30 min, in the presence of 100 or 600 ng/ml of human synthetic hepcidin. Mock controls were incubated without hepcidin. Experiments were performed three times with 3 replicates per experiment. Each point represents the mean ± 1 SD.

**Supplementary Figure 4 | Effect of T lymphocyte reconstitution on** *Hamp1* **mRNA expression on** *Foxn1nu* **mice.** qRT-PCR quantification of *Hamp1* mRNA expression on livers of *Foxn1nu* mice on iron-rich (+NTBI) or iron-sufficient (no NTBI) diets I.V. injected with 15 <sup>×</sup> 106 CD3<sup>+</sup> <sup>T</sup> lymphocytes or with PBS (Mock). Data are depicted as fold change relative to *FPN1* mRNA levels of Mock animals on iron-sufficient diet. Each bar represents the mean ± 1 SD of the values obtained for three animals. Statistical analysis was performed using One-Way ANOVA.

## **REFERENCES**


Holtzman, E. (1939). *Lysosomes*. New York, NY: Plenum Press.


Lentner, C. (1984). *Geigy Scientific Tables, Vol. 3*. Basle: Ciba-Geigy.


homeostasis of erythroblasts. *Blood* 118, 2868–2877. doi: 10.1182/blood-2011- 01-330241

Zimelman, A. P., Zimmerman, H. J., McLean, R., and Weintrauh, L. R. (1977). Effect of iron saturation of transferrin on hepatic iron uptake: an *in vitro* study. *Gastroenterology* 72, 129–131.

**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

*Received: 20 December 2013; paper pending published: 19 January 2014; accepted: 11 February 2014; published online: 26 February 2014.*

*Citation: Pinto JP, Arezes J, Dias V, Oliveira S, Vieira I, Costa M, Vos M, Carlsson A, Rikers Y, Rangel M and Porto G (2014) Physiological implications of NTBI uptake by T lymphocytes. Front. Pharmacol. 5:24. doi: 10.3389/fphar.2014.00024*

*This article was submitted to Drug Metabolism and Transport, a section of the journal Frontiers in Pharmacology.*

*Copyright © 2014 Pinto, Arezes, Dias, Oliveira, Vieira, Costa, Vos, Carlsson, Rikers, Rangel and Porto. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*

## Iron at the interface of immunity and infection

## *Manfred Nairz, David Haschka, Egon Demetz and Günter Weiss\**

Department of Internal Medicine VI-Infectious Diseases, Immunology, Rheumatology, Pneumology, Medical University of Innsbruck, Innsbruck, Austria

#### *Edited by:*

Raffaella Gozzelino, Instituto Gulbenkian de Ciência, Portugal

#### *Reviewed by:*

Marcelo Torres Bozza, Universidade Federal do Rio de Janeiro, Brazil Kostas Pantopoulos, Lady Davis Institute for Medical Research, Canada

#### *\*Correspondence:*

Günter Weiss, Department of Internal Medicine VI-Infectious Diseases, Immunology, Rheumatology, Pneumology, Medical University of Innsbruck, Anichstraße 35, A-6020 Innsbruck, Austria e-mail: guenter.weiss@i-med.ac.at

Both, mammalian cells and microbes have an essential need for iron, which is required for many metabolic processes and for microbial pathogenicity. In addition, cross-regulatory interactions between iron homeostasis and immune function are evident. Cytokines and the acute phase protein hepcidin affect iron homeostasis leading to the retention of the metal within macrophages and hypoferremia. This is considered to result from a defense mechanism of the body to limit the availability of iron for extracellular pathogens while on the other hand the reduction of circulating iron results in the development of anemia of inflammation. Opposite, iron and the erythropoiesis inducing hormone erythropoietin affect innate immune responses by influencing interferon-gamma (IFN-γ) mediated (iron) or NF-kB inducible (erythropoietin) immune effector pathways in macrophages. Thus, macrophages loaded with iron lose their ability to kill intracellular pathogens via IFN-γ mediated effector pathways such as nitric oxide (NO) formation. Accordingly, macrophages invaded by the intracellular bacterium Salmonella enterica serovar Typhimurium increase the expression of the iron export protein ferroportin thereby reducing the availability of iron for intramacrophage bacteria while on the other side strengthening anti-microbial macrophage effector pathways via increased formation of NO or TNF-α. In addition, certain innate resistance genes such as natural resistance associated macrophage protein function (Nramp1) or lipocalin-2 exert part of their antimicrobial activity by controlling host and/or microbial iron homeostasis. Consequently, pharmacological or dietary modification of cellular iron trafficking enhances host resistance to intracellular pathogens but may increase susceptibility to microbes in the extracellular compartment and vice versa.Thus, the control over iron homeostasis is a central battlefield in host–pathogen interplay influencing the course of an infectious disease in favor of either the mammalian host or the pathogenic invader.

**Keywords: iron, anemia of chronic disease, bacteria, nitric oxide, interferon, hepcidin, macrophage**

## **INTRODUCTION**

The control over iron homeostasis is decisive in host–pathogen interaction (Schaible and Kaufmann, 2004; Nairz et al., 2010; Cassat and Skaar, 2013). This is due to the fact that iron is central for several metabolic processes for both, prokaryotic and eukaryotic cells that the metal affects microbial proliferation and pathogenicity and in addition significantly impacts on immune cell plasticity and innate immune responses. These multiple functional aspects of iron are based on its ability to transfer electrons needed during metabolic processes and to catalyze the formation of highly reactive radicals (Papanikolaou and Pantopoulos, 2005; Koskenkorva-Frank et al., 2013). The latter can act as signaling molecules but also intoxicate microbes or damage surrounding cells and tissues. Many microbes are highly dependent on a sufficient supply of iron and take up this metal by multiple and divergent pathways or steel it from iron deposition sites of the host (Winkelmann, 2002; Cassat and Skaar, 2013; Frawley et al., 2013). The activation and expression of such microbial iron acquisition systems is linked to their pathogenicity and proliferation (Rabsch et al., 1999; Schrettl et al., 2004; Crouch et al., 2008; Andrews-Polymenis et al., 2010; Cassat and Skaar, 2013). On the other hand, iron plays important roles in anti-microbial

host responses, first by synergistic effects towards anti-microbial radical formation (Mastroeni et al., 2000; Esposito et al., 2003; Papanikolaou and Pantopoulos, 2005; Koskenkorva-Frank et al., 2013) but second, by directly altering immune cell proliferation and anti-microbial immune effector pathways (Nairz et al., 2010). Thus, the host immune system affects the availability of iron for microbes via the activity of cytokines, cellular proteins/peptides and hormones to gain control over pathogen proliferation and to strengthen specific immune effector pathways, a strategy for which the term "nutritional immunity has been coined.

## **ALTERATION OF IRON HOMEOSTASIS DURING INFECTION AND INFLAMMATION**

The most frequent and best known example visualizing the interaction between iron, immunity and infection is anemia of chronic disease (ACD) also termed as anemia of (chronic) inflammation (Cartwright, 1966; Spivak, 2002; Weiss and Goodnough, 2005). ACD is considered to be the second most frequent anemia worldwide and it develops specifically in patients suffering from chronic inflammatory diseases, such as auto-immune disorders, cancer, chronic infections or in patients undergoing dialysis (Weiss

and Goodnough, 2005). The underlying pathophysiology involves mainly (i) iron retention within the monocyte/macrophage system, (ii) a blunted formation and activity of the red blood cell hormone erythropoietin (Epo), and (iii) an impaired proliferation and differentiation of erythroid progenitor cells (Weiss and Schett, 2013). All these pathophysiological pathways are driven by inflammatory signals, and the mononuclear-phagocyte system (MPS) is in the center of such alterations. Specifically, macrophages are of central importance for maintaining a sufficient supply of iron for erythropoiesis due to their role in iron re-utilization from senescent erythrocytes which are taken up by these immune cells by a process called erythrophagocytosis before being degraded leading to recovery of heme which is further processed by the enzyme heme oxygenase-1 (HO-1) yielding equal amounts of iron, biliverdin and carbon-monoxid (Delaby et al., 2005; Knutson et al., 2005; Soe-Lin et al., 2009). Under physiological conditions iron recycling by macrophages accounts for approximately 95% of the daily needs of the metal for erythropoiesis and other physiological processes (Hentze et al., 2010; Pantopoulos et al., 2012). However, during inflammation this process is blunted resulting in an impaired delivery of iron for erythropoiesis. Thereby, cytokines and acute-phase proteins affect body iron homeostasis and macrophage iron metabolism leading to an inflammation driven diversion of iron traffic which is characterized by low circulating iron concentrations and high levels of the iron storage protein ferritin , the latter reflecting iron retention in the MPS (Thomas and Thomas, 2005; Weiss and Goodnough, 2005).

In case of an infection, auto-immune disease or cancer immune cells are activated and produce a myriad of cytokines, some of which exerting specific effects on iron homeostasis. Cytokines such as interleukin (IL)-1, IL-6, or IL-22 as well as bacterial LPS or endoplasmic reticulum stress induce the formation of the master regulator of iron homeostasis, hepcidin, in the liver (Nemeth et al., 2003; Vecchi et al., 2009; Armitage et al., 2011). Hepcidin affects cellular iron homeostasis upon binding to the only known iron export protein ferroportin, thereby leading to ferroportin internalization and degradation which subsequently reduces cellular iron export (Nemeth et al., 2004b). As a consequence of this interaction, the absorption of iron from the diet is reduced due to hepcidin mediated reduction of ferroportin expression in enterocytes, thereby resulting in a reduction of circulating iron levels which is further aggravated by the inhibition of iron export from macrophages by the same mechanism (Nemeth et al., 2004a; Kemna et al., 2005; Roy et al., 2007; Theurl et al., 2009). Moreover, macrophages produce minute amounts of hepcidin in response to inflammatory stimuli such as IL-6 or LPS thereby blocking iron export in an autocrine fashion (Peyssonnaux et al., 2006; Theurl et al., 2008), which is meant to result from a nutritional immune strategy of the body to reduce the availability of iron for extracellular pathogens (**Figure 1**).

In parallel, cytokines exert subtle and hepcidin independent effects on the regulation of iron homeostasis on multiple levels (Weiss and Schett, 2013). First, tumor necrosis factoralpha (TNF-α) impairs duodenal iron absorption by an as yet not fully elucidated mechanism (Johnson et al., 2004). Second, TNF-α, IL-1, IL-6, and interferon-gamma (IFN-γ) affect macrophage iron homeostasis by different avenues. They increase the uptake of transferrin and non-transferrin bound iron by modulating the expression of transferrin receptor-1 and divalent metal transporter-1, respectively (Byrd and Horwitz, 1993; Fahmy and Young, 1993; Mulero and Brock, 1999; Ludwiczek et al., 2003). In parallel, based on damaging of erythrocytes by inflammation born radicals, the half-life of erythrocytes decreases and erythrophagocytosis is stimulated. In addition, many of these cytokines as well as the anti-inflammatory cytokines IL-4, IL-10, and IL-13 promote efficient iron storage within macrophages/monocytes by increasing the expression of ferritin, both at the transcriptional and posttranscriptional level (Byrd and Horwitz, 1993; Weiss et al., 1997a; Mulero and Brock, 1999; Arosio and Levi, 2002; Tilg et al., 2002; Ludwiczek et al., 2003). Macrophage iron content is further expanded via an inhibitory effect of IFN-γ and LPS on ferroportin transcription thereby reducing cellular iron egress (Yang et al., 2002; Ludwiczek et al., 2003).

In summary the combination of these regulatory effects lead to iron retention in circulating monocytes and macrophages which present with low ferroportin expression and increased intracellular ferritin levels (Theurl et al., 2006) along with a reduction of circulating iron concentration, the diagnostic hallmark of ACD.

## **INTERRELATIONSHIP BETWEEN IRON AVAILABILITY, INNATE IMMUNE FUNCTION AND CONTROL OF INFECTIONS**

Considering the prevalence of ACD the question arises whether there is a specific benefit for the host upon development of ACD. Obviously, the retention of iron in the monocyte–macrophage system reduces circulating iron levels and thus the availability of this essential nutrient for extracellular microbes (**Figure 1A**). This iron withholding strategy appears to be of benefit to combat infections with circulating microbes or pathogens residing in tissues with high iron contents such as the liver (Ganz, 2009; Weinberg, 2009). In addition, the pathophysiological mechanisms underlying the development of ACD also significantly impact on the efficacy of host responses against infections. First, the reduced biological activity of Epo may ameliorate anti-bacterial immune responses. This is based on the fact that the major erythropoiesis stimulating hormone Epo also exerts effects on cells and tissues outside the bone marrow, and such effects are transduced by a heterodimeric receptor which differs from the receptor expressed on erythroid progenitor cells (Brines and Cerami, 2012). Thereby, Epo inhibits pro-inflammatory immune effector pathways in macrophages via inhibition of NF-kB activation which results in reduced expression of inducible nitric oxide synthase (iNOS), TNF-α, IL-6, and IL-12 by inflammatory macrophages and an impaired immune control of infections with bacteria such as *Salmonella enterica serovar typhimurium* (*S. typhimurium*; Nairz et al., 2011). Accordingly, the reduced Epo activity observed in ACD results in a sustained pro-inflammatory immune response and an improved control of *S. typhimurium* septicemia. Second, iron exerts multiple effects on immune effector functions. This is on the one hand based on the role of iron for the differentiation and proliferation of immune cells, including antigen presenting cells and lymphocytes (reviewed by Oppenheimer, 2001; Weiss, 2002; Cairo et al., 2011). Moreover, iron affects anti-microbial immune function of

with iron lose their ability to clear infections with intracellular pathogens, such as *Salmonella*, *Mycobacteria*,*Chlamydia*,*Candida*, or *Legionella* (Mencacci et al., 1997; Chlosta et al., 2006; Nairz et al., 2007; Paradkar et al., 2008; Botella et al., 2012; Bellmann-Weiler et al., 2013) whereas reduction of iron levels or addition of iron chelators can improve infection control by withholding iron from microbes and by increasing anti-microbial immune effector functions. Thereby iron chelators antagonize a direct inhibitory effect

et al., 1997; Dlaska and Weiss, 1999). iNOS catalyzed high output formation of the labile radical NO by macrophages (Nathan and Shiloh, 2000; Bogdan, 2001) which exerts direct anti-microbial effector functions. This pathway appears to be partly responsible for the beneficial clinical effects of iron chelators observed in several experimental infection systems (Mencacci et al., 1997; Fritsche et al., 2001; Ibrahim et al., 2007; Bellmann-Weiler et al., 2013; Nairz et al., 2013). Along this line children suffering from cerebral malaria benefited from the addition of the iron chelator

desferrioxamine to a standard anti-malaria treatment, as reflected by faster recurrence from coma and clearance of plasmodia, although this did not translate into a survival benefit (Gordeuk et al., 1992; Mabeza et al., 1999), although anti-plasmodial innate immune responses were positively affected by iron chelation *in vivo* (Thuma et al., 1996; Weiss et al., 1997b). Of interest, via its inhibitory potential on IFN-γ iron affects T-helper (TH) cell differentiation favoring the expansion of TH-2 cells which produce a number of macrophage-deactivating cytokines such as IL-4 or IL-13 (Weiss, 2002). Accordingly, iron overload and iron chelation have been shown to modulate the TH-1/TH-2 in mice exposed to different microbes (Mencacci et al., 1997; Ibrahim et al., 2007; Nairz et al., 2013). However, a recent clinical trial in patients after allogenic bone marrow transplantation who received the iron chelator desferrasirox in an attempt to improve the control of mucormycosis, a devastating invasive fungal infection in severely immuno-compromised patients, failed to show a clinical benefit (Spellberg et al., 2012). In contrast, the mortality rate was even higher in those subjects receiving desferrasirox and the reason for that has been elusive thus far. Hypothetically this may be partly related to an effect of desferrasirox on innate immune function, which could have been resulted in aggravation of graft versus host disease, a notion which should be verified at least retrospectively.

However, from all the evidence listed above it appears obvious that ACD develops from the endeavor of the body to withhold iron from invading, extracellular pathogens and to strengthen at the same time anti-microbial immune responses (**Figure 1A**).

Accordingly, clinical trials which were performed to supplement iron to children in developing countries based on the notion that iron deficiency is associated with growth and mental retardation produced unpredicted results (Schumann et al., 2007). These studies demonstrated that iron supplementation resulted in higher incidence of or higher mortality from infections such as malaria, diarrhea or bacterial meningitis (Sazawal et al., 2006; Soofi et al., 2013). The pathways underlying these devastating outcomes remain elusive thus far. However, they may be linked to iron mediated modulation of anti-microbial immune defense mechanisms or traced back to increased availability of the metal for pathogens in the setting of subclinical parasitemia or bacteremia. The cause-effective association between iron availability and the clinical course of malaria has been well established by several recent studies.

Those studies provided evidence for a reduced risk of malaria in general and severe malaria in iron deficient pregnant women and children (Kabyemela et al., 2008; Gwamaka et al., 2012; Jonker et al., 2013) whereas other studies found interesting associations between iron delivery for erythropoiesis, circulating hepcidin levels and the prevalence of malaria in tropical countries (de Mast et al., 2010; Prentice et al., 2012; Atkinson et al., 2014).

Apart from plasmodia, the availability of iron is of importance for other parasitic infection such as Leishmaniosis or Trypanosomiasis. The pathogenicity of *Leishmania* is linked to the expression of different microbial iron acquisition molecules and a sufficient supply of iron (Mittra et al., 2013). Accordingly, inhibition of cellular iron export can promote *Leishmania* proliferation (Ben-Othman et al., 2014) whereas over-expression of ferroportin with subsequent limitation of intracellular iron availability limits intracellular leishmanial growth (Rafiee et al., 2014). In addition, drugs such as quercetin exert their anti-microbial activity against *Leishmania donovani* by interfering with microbial iron acquisition (Sen et al., 2008). Of note, *L. donovani* manipulates macrophage iron homeostasis to increase its own iron supply (Das et al., 2009). However, via its radical promoting capacity iron can also exert protective effects in *Leishmania* infection by strengthening radical dependent host response as shown in a mouse model of *Leishmania infantum* infection (Vale-Costa et al., 2013). Similarly, *Trypanosma* species are also highly dependent on a sufficient supply of iron which depending on the subtype is acquired by classical sources (transferrin) or taken up via ZIP family transporters (Taylor and Kelly, 2010). Given the central role of iron for the pathogenesis of trypansoma infections, the host aims at limiting iron availability to these parasites which results in alterations of macrophage iron homeostasis along with the development of anemia (Stijlemans et al., 2008, 2010). Of interest, the induction of the Nrf2 pathway exerted protective effects in a model of *Trypanosma cruci* infection in mice which was counterbalanced by the addition of iron sulfate (Paiva et al., 2012), suggesting that Nrf2 exerted its anti-trypanosmal activity by increasing the expression of ferroportin (Nairz et al., 2013).

## **METABOLIC IRON RESPONSES AND IMMUNE CONTROL OF INTRACELLULAR PATHOGENS**

Although still insufficiently understood, it is hypothesized that the presence of extracellular pathogens in the circulation induces iron restriction in the monocyte/macrophage system via the action of hepcidin and several cytokines, whereas the metabolic responses to intracellular microbes appear to be different (**Figure 1**). Macrophages infected with *S. typhimurium* increase the expression of ferroportin and stimulate iron export (Nairz et al., 2007). This leads to a reduced availability of iron for intracellular bacteria along with an activation of anti-microbial immune effector mechanisms due to counter-balancing the negative regulatory effects of iron on IFN-γ inducible immune pathways (Oexle et al., 2003; Nairz et al., 2007; Weiss and Schett, 2013). Similar observations have been made with other intracellular pathogens such as Chlamydia spp. or Legionella (Chlosta et al., 2006; Paradkar et al., 2008; Bellmann-Weiler et al., 2010) Of interest, ferroportin is also expressed in mycobacteria containing phagosomes where it pumps iron into the cytoplasma (Van Zandt et al., 2008), a pathway which limits the availability of this essential nutrient for this bacterium (Schnappinger et al., 2003; Olakanmi et al., 2013). This indicates that the stimulation of iron export via ferroportin is an efficient defense strategy against infection with intracellular microbes by limiting their access to iron and by strengthening anti-microbial immune effector pathways (Nairz et al., 2010).

Recent evidence suggest, that the intracellular bacterium *Salmonella typhimurium* can counteract these metabolic immune defense strategies by inducing hepcidin expression in hepatocytes via activation of estrogen related receptor (ERR)-gamma, thereby resulting in hypoferremia and macrophage iron retention with a subsequently increased availability of the metal for intracellular *Salmonella* (Kim et al., 2014). A reverse agonist of this pathway

counter-acting ERR-gamma and hepcidin mediated macrophage iron retention led to an improved control of *Salmonella* infection (Kim et al., 2014). This is in a line with previous observations demonstrating that modulation of the hepcidin/ferroportin axis impacts on intracellular proliferation of *Salmonella* and the course of infection in mice (Nairz et al., 2009b; Wang et al., 2009).

Importantly, several innate resistance mechanisms exert at least part of their anti-microbial activity via restriction of iron to microbes.

## **NO**

High output formation of NO by immune cells is of central importance for immune control of infections and cancer (Nathan and Shiloh, 2000; Bogdan, 2001). An interaction of NO with iron homeostasis has been well documented based on the fact that (i) NO has a high affinity for iron and that many of the cytotoxic effects of NO are based on targeting of central iron sulfur clusters in enzymes by the labile radical (Nathan and Shiloh, 2000; Bogdan, 2001), that (ii) NO controls intracellular iron trafficking by regulating the binding affinity of iron regulatory proteins (IRP) to specific RNA stem loop structures within the non-coding regions of critical iron metabolism genes (Drapier et al., 1993; Weiss et al., 1993; Pantopoulos et al., 1996), and finally, that (iii) cellular iron content controls NO expression by regulating iNOS transcription (Weiss et al., 1994; Melillo et al., 1997; Dlaska and Weiss, 1999). Recent evidence now suggests that part of the antimicrobial effect of NO can be attributed to a regulatory activity of the radical on iron homeostasis. Thereby, NO activates the binding affinity of the transcription factor Nrf-2 to the ferroportin promoter, resulting in increased ferroportin expression and iron export (Nairz et al., 2013). This results in a reduced availability of iron for intra-macrophage bacteria and a strengthening of antimicrobial, IFN-γ driven immune responses. Of note, iNOS−/<sup>−</sup> mice present with macrophage iron overload and their resistance against infection with *Salmonella* can be increased upon application of a membrane permeable iron chelator such as desferrasirox (Nairz et al., 2013).

## **Nramp1**

The natural resistance macrophage protein 1 (Nramp1 or Slc11a1) has been characterized as a late phagosomal protein conferring resistance to infection with *Salmonella*, *Mycobacteria*, and *Leishmania* (Blackwell et al., 2000; Forbes and Gros, 2001). Nramp1 exerts its protective effect against such infections via acidification of the microbe containing phagosome, but also by altering the cellular distribution of divalent metals such as zinc, manganese or iron which all play decisive role in host–pathogen interaction (Hood and Skaar, 2012; Diaz-Ochoa et al., 2014). In addition, the functional expression of Nramp1 strengthens anti-microbial immune effector pathways such as the formation of TNF-α or NO (Barton et al., 1995; Fritsche et al., 2003). These effects can be traced back to prolonged activity of pro-inflammatory signaling pathways and inhibition of the expression of anti-inflammatory cytokines, such as IL-10 (Fritsche et al., 2008). Evidence accumulates, that Nramp1pumps iron out of macrophages thereby reducing iron levels in the cytoplasma and within the phagolysosome rendering the metal

less available for intracellular bacteria (Barton et al., 1999; Zwilling et al., 1999; Kuhn et al., 2001; Nairz et al., 2009a; Sohn et al., 2011). As a further consequence of intracellular iron depletion, pro-inflammatory immune effector pathways are augmented. One of these Nramp1-inducible responses is the increased formation of lipocalin-2 (Lcn2 or NGAL), which blocks another source of iron for bacteria (Fritsche et al., 2012).

## **LIPOCALIN-2**

Lcn2 is produced by several cells in the body including neutrophils and macrophages (Chakraborty et al., 2012). Among many other functions it binds the bacterial siderophore enterobactin, which is produced by Gram-negative bacteria such as *Escherichia coli*, *Klebsiella*, or *Salmonella* spp. to scavenge iron and to redeliver the metal to the microbe where it is taken up via specific receptors. Mice expressing Lcn2 are more resistant to infections with such Gram-negative bacteria as compared to Lcn2−/<sup>−</sup> littermates which is cause-effectively due to the bacterial iron withholding capacity of Lcn2 (Flo et al., 2004; Berger et al., 2006). Of note, the varying dependence of bacteria from siderophore mediated iron uptake can cause a growth advantage of certain bacteria among others. It has recently been demonstrated that the probiotic bacterium *E. coli* Nissle controls the growth of pathogenic *Salmonella* in the intestine by over-coming iron restriction by Lcn2 (Deriu et al., 2013). Accordingly, resistance of *Salmonella* to Lcn2 mediated iron restriction can cause a growth advantage of this pathogen in the gut (Raffatellu et al., 2009). It is of interest, that Lcn2 not only affects microbial iron delivery but also host iron homeostasis. This is most likely due to binding of a recently identified mammalian siderophore by Lcn2 which then can shuttle iron across cellular membranes (Bao et al., 2010; Devireddy et al., 2010). Of note, recent evidence also suggests that the mammalian siderophore 2,5-DHBA can be utilized by Gram-negative bacteria as a source of iron and that macrophages reduce the expression of 2,5-DHBA when exposed to LPS or Gram negative bacteria which is also considered to be part of the "iron withholding" defense network of innate immune cells (Liu et al., 2014).

Lcn2 appears to be of importance to mount alterations of iron host homeostasis on the cellular and systemic level thereby contributing to hypoferremia and intracellular iron depletion in systemic *Salmonella* infection (Nairz et al., 2009b; Srinivasan et al., 2012). An increased expression of Lcn2 by macrophages along with reduced intramacrophage iron content and impaired bone morphogenetic signaling may be largely responsible for the reduced susceptibility of Hfe−/<sup>−</sup> mice, a model for hereditary hemochromatosis, against infection with the intracellular bacteria *Salmonella* and *Mycobacteria* (Olakanmi et al., 2007; Corradini et al., 2009; Nairz et al., 2009b), which may also be a reason of the high penetrance of this genes in people of Northern and Western European origin (Pietrangelo, 2004). Accordingly, Lcn2 expression positively affects the course of infection with other intracellular pathogens, even if they do not express siderophores such as Chlamydia (Bellmann-Weiler et al., 2013), whereas cellular iron export and increased delivery of the metal to the extracellular space can be detrimental as observed in a model of pneumococcal infection, where Lcn2 expression by neutrophils resulted in increased mortality of mice (Warszawska et al., 2013).

This again provides evidence that contrasting pathways for the regulation of iron homeostasis according to infection with either intracellular or extracellular pathogens exist (Chan et al., 2009; Warszawska et al., 2013; Fang and Weiss, 2014) which are still insufficiently understood (**Figure 1**).

Owing to the importance of these pathways for immune defense against infection with intracellular pathogens the central TH-1 cytokine IFN-γ exerts part of its anti-microbial effects by stimulating immune responses which restrict the availability of iron for microbes. IFN-γ induces the expression of Nramp1, Lcn2, and NO thereby reducing intracellular iron content in macrophages and limiting the growth of bacteria such as *Salmonella* (Gruenheid et al., 1997; Fritsche et al., 2008; Nairz et al., 2008, 2009a; Andrews-Polymenis et al., 2010; Nairz et al., 2013). Thus, IFN-γ is central for host response in *Salmonella* infection based on these nutritional iron effects but also due to its ability to induce a myriad of anti-bacterial effector mechanisms in macrophages including oxygen and nitrogen radical formation or maturation of the bacterial containing phagosome (Mastroeni et al., 2000; McCollister et al., 2005; Henard and Vazquez-Torres, 2011). Of note, neutrophils have recently been shown to produce IFN-γ to further stimulate anti-bacterial immune pathways (Spees et al., 2014) whereas *S. typhimurium* has been shown to impair the proliferation of CD4 cells, a major source of IFN-γ (Atif et al., 2014).

## **TARGETING IRON HOMEOSTASIS IN INFECTIOUS DISEASE-TO WALK A TIGHTROPE**

While iron supplementation in eras with a higher burden of infectious diseases caused detrimental effects toward the risk of malaria or invasive bacterial infections (Sazawal et al., 2006; Soofi et al., 2013), iron supplementation in HIV infected patients resulted in an impaired control of malaria but also in a beneficial or at least neutral course of HIV infection at as reflected by CD4+-cell counts or progression of the disease which was partly dependent on the degree of anemia and base-line iron status (Esan et al., 2013; Prentice et al., 2013; Zlotkin et al., 2013). However, HIV infection is often associated with reactivation of *Mycobacterium tuberculosis* infection, and iron has been shown to be an essential nutrient for such bacteria which goes along with the observation that iron loading is associated with an increased risk for tuberculosis and an adverse clinical course of this infection (Moyo et al., 1997; Gangaidzo et al., 2001; Schaible and Kaufmann, 2004). Accordingly, iron supplementation in subjects with latent tuberculosis and an impaired immune control, e.g., on the basis of HIV infection, is hazardous (McDermid et al., 2013). Of interest, a recent study suggested that both, iron deficiency and iron loading, are associated with an adverse clinical course of tuberculosis, both in HIV positive and negative subjects (Isanaka et al., 2012a,b). This may relate to the divergent effects of iron on the immune system, on the one hand iron is a prerequisite for immune cell proliferation and differentiation (Thorson et al., 1991; Weiss, 2002; Porto and De Sousa, 2007), and a catalyzer for the formation of anti-microbial radicals (Mastroeni et al., 2000; Papanikolaou and Pantopoulos, 2005), whereas on the other hand it impacts on innate immune effector functions (Weiss and Schett, 2013) and positively affects microbial proliferation (Schaible and Kaufmann, 2004; Weinberg, 2009; Nairz et al., 2010; Cassat and Skaar, 2013; **Figure 1B**). Thus, a certain balance of iron, not too less and not too much, is needed to strengthen immune responses to successfully combat infections (Drakesmith and Prentice, 2012). Of note, alterations of iron homeostasis have been shown to affect the composition of the human microbiome and may thereby alter the proliferation of pathogenic bacteria (Deriu et al., 2013).

Accordingly, pharmacological concepts to modify iron homeostasis and iron trafficking in an attempt to combat infection have always to keep in mind that a positive effect on one infection may have devastating effects on the course of another infection as seen in models of malaria where tolerance induction via overexpression of HO improves the course of malaria but increases the risk for bacterial infections such as Salmonellosis (Portugal et al., 2011; Gozzelino et al., 2012). Opposite, reduction of intracellular/macrophage iron levels upon therapeutic application of the calcium-antagonists nifedipine resulted in significantly improved survival of mice with *S. typhimurium* (Mair et al., 2011). Strictly speaking, any therapeutic strategy, e.g., iron chelation, iron mobilization, hepcidin or anti-hepcidin pharmacological approaches, which help to control the course of one, e.g., intracellular infection, may increase the availability of iron for a pathogen residing in a different compartment, e.g., in the extracellular space, along with the unpredictable effects on host immune response (Nairz et al., 2010; Drakesmith and Prentice, 2012). Thus, major research efforts must be undertaken to better understand the diverse roles of iron in infection and in immune control of infections. Specifically, we also need to address the metabolic alterations of iron homeostasis in response to different pathogens, not only in terms of their primary tissue localization but also in relation to their needs for iron and the immune mechanisms being involved in their control.

## **REFERENCES**


transmission setting: community-based, randomised, placebo-controlled trial. *Lancet* 367, 133–143. doi: 10.1016/S0140-6736(06)67962-2


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

*Received: 21 April 2014; paper pending published: 27 May 2014; accepted: 10 June 2014; published online: 16 July 2014.*

*Citation: Nairz M, Haschka D, Demetz E and Weiss G (2014) Iron at the interface of immunity and infection. Front. Pharmacol. 5:152. doi: 10.3389/fphar.2014.00152*

*This article was submitted to Drug Metabolism and Transport, a section of the journal Frontiers in Pharmacology.*

*Copyright © 2014 Nairz, Haschka, Demetz and Weiss. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited andthatthe original publication inthis journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*

## Heme on innate immunity and inflammation

## *Fabianno F. Dutra\* and Marcelo T. Bozza\**

Laboratório de Inflamação e Imunidade, Departamento de Imunologia, Instituto de Microbiologia, Universidade Federal do Rio de Janeiro, Rio de Janeiro, Brazil

#### *Edited by:*

Raffaella Gozzelino, Instituto Gulbenkian de Ciência, Portugal

#### *Reviewed by:*

Leo Otterbein, Beth Israel Deaconess Medical Center/Harvard Medical School, USA Viktória Jeney, University of Debrecen, Hungary

#### *\*Correspondence:*

Fabianno F. Dutra and Marcelo T. Bozza, Laboratório de Inflamação e Imunidade, Departamento de Imunologia, Instituto de Microbiologia, Universidade Federal do Rio de Janeiro, Avenida Carlos Chagas Filho, 373, Rio de Janeiro 21941-902, Brazil e-mail: fabianno44@yahoo.com.br; mbozza@micro.ufrj.br

Heme is an essential molecule expressed ubiquitously all through our tissues. Heme plays major functions in cellular physiology and metabolism as the prosthetic group of diverse proteins. Once released from cells and from hemeproteins free heme causes oxidative damage and inflammation, thus acting as a prototypic damage-associated molecular pattern. In this context, free heme is a critical component of the pathological process of sterile and infectious hemolytic conditions including malaria, hemolytic anemias, ischemiareperfusion, and hemorrhage. The plasma scavenger proteins hemopexin and albumin reduce heme toxicity and are responsible for transporting free heme to intracellular compartments where it is catabolized by heme-oxygenase enzymes. Upon hemolysis or severe cellular damage the serum capacity to scavenge heme may saturate and increase free heme to sufficient amounts to cause tissue damage in various organs. The mechanism by which heme causes reactive oxygen generation, activation of cells of the innate immune system and cell death are not fully understood. Although heme can directly promote lipid peroxidation by its iron atom, heme can also induce reactive oxygen species generation and production of inflammatory mediators through the activation of selective signaling pathways. Heme activates innate immune cells such as macrophages and neutrophils through activation of innate immune receptors. The importance of these events has been demonstrated in infectious and non-infectious diseases models. In this review, we will discuss the mechanisms behind heme-induced cytotoxicity and inflammation and the consequences of these events on different tissues and diseases.

**Keywords: heme, iron, hemolysis, ROS, inflammation, innate immunity, programed cell death, cytotoxicity**

## **HEME PHYSIOLOGY**

Heme consists of a tetrapyrrole ring with an iron (Fe) atom bound in the center, coordinated to the pyrrole rings (Kumar and Bandyopadhyay, 2005). Heme is present in organisms of all kingdoms being the prosthetic group of several proteins fundamental for the life of aerobic organisms (Wagener et al., 2003b). In vertebrates, heme is ubiquitously expressed and its functions are determined by the polypeptide interacting with it (Dawson, 1988). The capacity of the chelated Fe to undergo a reversible change in the oxidation state renders heme a versatile biological catalyst acting in oxidative reactions, electron transfer processes, and delivering molecular oxygen (O2) to cells (Ryter and Tyrrel, 2000). Heme is presented in different forms, mostly in forms *a*, *b*, and *c*. Heme *a* and *c* are important in the physiology of electron transport and are found in cytochrome *c* oxidase (Tsukihara et al., 1995) and cytochrome *c* reductase (Xia et al., 1997; Zhang et al., 1998), respectively. Heme *b* is the most common type in mammals and is part of several proteins (Larsen et al., 2012) including the gas carriers hemoglobin (Hb; Park et al., 2006) and myoglobin (Evans and Brayer, 1988; Vojtechovsky et al., 1999), hemeproteins related to heme release during hemolysis and tissue damage. Importantly, heme *b* interaction with heme oxygenase (HO; Lad et al., 2003), the enzyme responsible for heme intracellular catabolism, and hemopexin (Hx; Paoli et al., 1999), a plasmatic heme scavenger, is essential for the regulation of free heme availability and Fe recycling (Kovtunovych et al., 2010; Tolosano et al., 2010).

Besides its physiological importance, heme has a potent oxidative capacity oxidizing lipids (Tappel, 1953, 1955; Vincent et al., 1988) and proteins (Aft and Mueller, 1984; Vincent, 1989), and damaging DNA (Aft and Mueller, 1983). Thus, heme can be a dangerous molecule once released from hemeproteins. Some diseases are characterized by high amounts of hemeproteins out from their physiological environments. The consequences of heme toxicity can be appreciated in hemolytic diseases such as β-thalassemia, sickle-cell disease (SCD), ischemia-reperfusion (IR), and malaria (Katori et al., 2002; Pamplona et al., 2007;Vinchi et al., 2013). Extracellular and intracellular proteins have essential functions controlling free heme availability, guarding tissues from its deleterious effects. Through evolution mammals acquired several protective mechanisms against heme toxicity. Under physiologic conditions or mild to moderate hemolysis, haptoglobin (Hp) binds Hb (Melamed-Frank et al., 2001). Because of its large molecular size, this complex is maintained in the intravascular space, preventing the association of otherwise free Hb with nitric oxide (NO; Reiter et al., 2002) and inhibiting the release of free heme (Melamed-Frank et al., 2001). The complex Hp:Hb binds to CD163 (Kristiansen et al., 2001), present in macrophages and hepatocytes (Philippidis et al., 2004; Quaye, 2008), which mediates the endocytosis of Hb–Hp complexes for degradation. However, during hemolytic diseases high amounts of Hb are released in the intravascular environment, saturating the Hp molecules and thus accumulating free Hb (Muller-Eberhard, 1970). In the presence of reactive oxygen species (ROS), Hb is oxidized to methemoglobin (MetHb; Balla et al., 1993), characterized by the change in the oxidative state of the Fe present in the heme molecule from ferrous (Fe+2) to ferric (Fe+3). MetHb is unstable and rapidly releases free heme (Balla et al., 1993). In this context, Hx scavenges free heme, binding it with very high affinity (Paoli et al., 1999). Hx inhibits the oxidative property of heme (Eskew et al., 1999) and mediates heme transportation to intracellular compartments through the macrophage receptor CD91 (Hvidberg et al., 2005), a critical step on heme catabolism. Once inside the cells, heme is catabolized by HO enzymes, generating equimolar amounts of biliverdin, carbon monoxide (CO), and Fe (Tenhunen et al., 1968). Differently from biliverdin and CO, which have anti-inflammatory effects (Otterbein et al., 2000; Baranano et al., 2002), free Fe is highly oxidative and can promote free radicals generation through the Fenton reaction, which catalyzes hydroxyl radicals from the reaction of Fe with H2O2 (Fenton, 1894). These radicals initiate lipid-chain peroxidation by obtaining electrons from other molecules, including unsaturated fatty acid (RH), generating an alkyl radical (R•; Ryter and Tyrrel, 2000). Thus, labile Fe can induce cytotoxic effects (Gozzelino et al., 2012). In this context, the HO function is complemented by ferritin, a ubiquitous intracellular protein complex that stores Fe in a soluble and non-toxic form (Eisenstein et al., 1991). Ferritin is composed of a heavy and a light chain which cooperates in the overall uptake and store of Fe in a non-toxic form protecting cells from hydroxyl radicals generated through the Fenton reaction (Arosio et al., 2009). While the ferritin heavy chain (FtH) presents a ferroxidase activity that converts the Fe from Fe+<sup>2</sup> into Fe+<sup>3</sup> (Broxmeyer et al., 1991), the ferritin light chain (FtL) is essential to promote Fe nucleation (Levi et al., 1992). Cells deficient on FtH are more susceptible to oxidative damage, while increased amounts of FtH protects cells from death induced by challenges such as Fe, tumor necrosis factor (TNF), heme, heme plus TNF, or oxidized low-density lipoprotein (LDL; Juckett et al., 1995; Pham et al., 2004; Gozzelino et al., 2012). All these effects require the ferroxidase activity of FtH.

Not only vertebrates need to manage high amounts of free heme. Hematophagous arthropods, upon digestion of Hb in the mid-gut, are exposed to enormous quantities of free heme and ROS. Different blood-sucking species have developed independent adaptations to deal with heme toxicity, including heme aggregation (Oliveira et al., 1999), heme degradation (Paiva-Silva et al., 2006; Pereira et al., 2007), expression of heme-binding proteins (Oliveira et al., 1995), and induction of antioxidant systems (Paes et al., 2001; Oliveira et al., 2011).

## **DIRECT CYTOTOXIC EFFECTS OF HEME**

The cytotoxicity of heme was initially associated with a direct effect due to its structural properties, specifically the capacity of the hydrophobic porphyrin ring to intercalate lipid layers and the presence of the atom of Fe, considered essential for heme-induced ROS generation. Heme can destabilize biological membranes increasing its permeability and the chance of lysis (Schmitt et al., 1993), as it happens with erythrocytes (Chiu and Lubin, 1989). This effect relies in the amphipathic properties of the porphyrin ring rendering heme the tendency to allocate in non-polar niches including lipoproteins and lipid

membranes where it might establish dangerous contacts with lipids, important targets of peroxidation by ROS (Rose et al., 1985; Light and Olson, 1990a,b). Once intercalated into cellular plasma membranes heme amplifies cellular susceptibility to oxidative-mediated injury by oxidants such as H2O2 or those derived from activated inflammatory cells (Balla et al., 1991a,b, 1993). Because heme has an atom of Fe in its structure and can sensitize cellular membranes to H2O2, it was proposed that heme-induced cytotoxicity could be derived from free radicals generated through the Fenton reaction (Vincent, 1989; Chiu et al., 1996). However, conflicting data makes uncertain that heme can directly catalyze this reaction. Studies using electron paramagnetic resonance (EPR) spin-trapping techniques, in cell free conditions, suggested that neither MetHb (Mao et al., 1994), nor free heme (Van der Zee et al., 1996), can catalyze the Fenton reaction generating hydroxyl radicals. Furthermore, heme induces ROS generation dependent on enzymatic reactions. In fact, it was demonstrated that heme triggers ROS production by the NADPH oxidase enzyme (Graça-Souza et al., 2002; Arruda et al., 2004; Moraes et al., 2012; Barcellos-de-Souza et al., 2013) and by mitochondria (unpublished), through activation of specific signaling pathways (Graça-Souza et al., 2002; Porto et al., 2007; Fernandez et al., 2010; Fortes et al., 2012). Alternatively, hemeinduced formation of radical species relies on the conversion of low-reactive organic hydroperoxides (ROOH) into highly reactive alkoxyl (RO•) and peroxyl (ROO•) radicals (Tappel, 1953, 1955; Van der Zee et al., 1996). These radicals may initiate further lipid peroxidation forming alkyl radicals that in the presence of O2 form more peroxyl radicals leading to a facile propagation of free radical reactions. Ultimately, alkyl radicals are rearranged to form toxic conjugated diens and stimulate the autocatalytic lipid peroxidation cascades (Ryter and Tyrrel, 2000) which could mediate heme-induced toxicity. Together, these data suggest that the cytotoxic heme effects can be mediated by ROS from different sources generated both through non-enzymatic and enzymatic reactions. This leads to at least four possible mechanisms of heme-induced ROS: (1) heme might directly promote ROS generation by the Fe present in its structure through the Fenton reaction, (2) Fe might be released from heme to exert its oxidative effects, (3) heme can activate signaling pathways to induce ROS generation enzymatically, or (4) heme can convert organic hydroperoxides into free radicals.

## **INFLAMMATORY EFFECTS OF HEME**

Heme activates innate immune cells and non-hematopoietic cells promoting inflammation. Heme injection in mice leads to vascular permeability, leukocyte migration from the intravascular environment to tissues and increase of acute-phase proteins (Lyoumi et al., 1999; Wagener et al., 2001b), hallmarks of acute inflammation. Heme activates endothelial cells inducing the expression of the adhesion molecules ICAM-1 (intercellular adhesion molecule 1), VCAM-1 (vascular cell adhesion molecule 1), E-selectin, Pselectin, and von Willebrand factor (VWF; Wagener et al., 1997; Belcher et al., 2014) and causes neutrophil migration (Graça-Souza et al., 2002; Porto et al., 2007). The adhesion molecules enable neutrophils to attach firmly to the endothelium and migrate to tissues parenchyma. Thus, leukocytes must be activated and

express adhesion molecules complementary to those expressed by the endothelium. A seminal study demonstrated that the ability of heme to activate neutrophils depend on protein kinase C (PKC) activation and ROS generation, inducing the expression of adhesion molecules and modifying actin cytoskeleton dynamics, necessary features for neutrophils migration (Graça-Souza et al., 2002). In fact, heme can induce neutrophil migration by acting as a chemotactic molecule (Porto et al., 2007) or by inducing the production of leukotriene B4 (LTB4) by macrophages (Monteiro et al., 2011). Heme-induced neutrophils activation leads to extracellular traps (NETs) release through a mechanism dependent on ROS generation (Chen et al., 2014). Moreover, heme induces the production of the chemokine interleukin (IL)-8 by endothelial cells (Natarajan et al., 2007) and by neutrophils (Graça-Souza et al., 2002). Although neutrophils have important functions controlling infection, these cells can promote vascular and tissue injury by generating ROS, secreting proteases, and releasing extracellular chromatin (NETs; reviewed in Mócsai, 2013). Consequently, the prolonged activity of these cells is detrimental for tissue homeostasis. In this context, heme inhibits neutrophils apoptosis, increasing their longevity, and possibly enhancing harmful stimuli from these cells (Arruda et al., 2004, 2006).

Macrophages stimulate inflammatory responses by secreting cytokines, chemokines, and lipid mediators. TNF, IL-6, and IL-1β are pleiotropic cytokines that regulate cell death, increase vascular endothelial permeability, recruit immune cells to inflamed tissues and induce the production of acute-phase proteins. Part of these effects is mediated by the activation of fibroblasts and endothelial cells, especially by TNF and IL-1β. Moreover, IL-1β and TNF modifies the hypothalamus threshold of the body temperature causing fever. KC (keratinocyte-derived chemokine) is a chemokine that attracts neutrophils to sites of inflammation and LTB4 is a lipid mediator that functions as a chemoattractant molecule and also activates leukocytes. Heme activates macrophages inducing the production of TNF, KC (Figueiredo et al., 2007), IL-1β (unpublished), and LTB4 (Monteiro et al., 2011). Although KC and IL-1β functions were not investigated during heme-induced inflammatory effects, TNF and LTB4 were described as essential inflammatory mediators during inflammatory events induced by heme. As described before, LTB4 has an important function regulating heme-induced neutrophils migration (Monteiro et al., 2011). On the other hand, TNF secretion induced by heme is essential for the activation of the programed necrotic cell death pathway, which is denominated necroptosis (Fortes et al., 2012).

Hemoglobin is used as a nutrient by *Plasmodium*, the protozoan parasite that causes malaria, during its replication stage inside erythrocytes in the course of the infection in mammals (Francis et al., 1997). Hb digestion generates high amounts of free heme which is detoxified by its polymerization into a crystal structure named hemozoin (Hz; Slater et al., 1991; Slater, 1992; Pandey and Tekwani, 1996). Once released from ruptured erythrocytes, Hz can be phagocytized and accumulated inside macrophages altering their functions. Similarly to heme, Hz also induces the production of several inflammatory mediators by macrophages such as cytokines and chemokines, and induces leukocytes migration (reviewed

by Olivier et al., 2014). In nature, Hz produced by *Plasmodium* spp. carries DNA from the parasite. Because DNA is immunogenic and can activate innate immune receptors (Yanai et al., 2009; Coban et al., 2010), the exact role of pure Hz in the activation of innate immunity is not completely understood (discussed later).

Together, these inflammatory responses triggered by heme indicate that heme is a damage-associated molecular pattern (DAMP), a group of endogenous molecules derived from damaged cells and extracellular matrix degradation capable of promoting and exacerbating immune responses. These concepts challenged the idea that the cytotoxic and inflammatory effects of heme were exclusively mediated by the oxidative capability of the Fe associated with the amphipathic property of the porphyrin ring. Moreover, albeit the controversy about Hz immunogenicity, Hz composed of heme and *Plasmodium* DNA can be defined as a pathogen-associated molecular pattern (PAMP) in the course of malaria infection, promoting inflammation and contributing to pathogenesis.

## **HO-1 PROTECTION AGAINST HEME**

Plasma scavengers of heme and/or Hb once saturated allow heme to exert its inflammatory properties. In fact, hemolysis or heme injection in *Hx*−/<sup>−</sup> mice cause increased inflammation and severe renal damage compared to wild type (WT) mice (Tolosano et al., 1999; Vinchi et al., 2008). Mice lacking HO-1 (*Hmox1*−/−) are highly susceptible to pathologic conditions associated with increased serum heme concentration. For instance, *Hmox*−/<sup>−</sup> mice develops acute renal failure and marked mortality when submitted to rhabdomyolysis, a pathological condition that increases serum myoglobin which can be oxidized and release heme (Nath et al., 2000). Furthermore, *Hmox1*−/<sup>−</sup> mice are susceptible to liver IR which is characterized by tissue damage in sites that are reperfused after ischemia injury and hemolysis (Devey et al., 2009). Thus, HO-1 regulation by heme is an essential feedback mechanism that maintains tissue homeostasis through anti-inflammatory and anti-oxidant activities in pathologic situations associated to heme (reviewed by Gozzelino et al., 2010). SCD and β-thalassemia are genetic diseases associated to erythrocytes that are prone to lysis due to defective Hb production (Heinle and Read, 1948; Pauling et al., 1949; Ingram, 1957; discussed later). Asymptomatic humans with sickle-cell trait are protected against malaria disease. Interestingly, this fact was suggested to be dependent on heme-induced HO-1 expression. Transgenic sickle-cell hemizygous mice are protected against malaria disease (Ferreira et al., 2011). Besides the fact that these mice do not develop anemia, hemolysis is evidenced by a modest serum heme increase which is sufficient to increase HO-1 expression and CO production (Ferreira et al., 2011). CO inhibits Hb oxidation and subsequently heme release, thus blocking heme accumulation in serum and preventing heme from exerting its inflammatory effects in the course of malaria disease (Ferreira et al., 2011). This case demonstrates the importance of the feedback production of HO-1 induced by heme to maintain tissue homeostasis. Moreover, heme accumulates in wounded areas and promotes inflammation through the expression of adhesion molecules and the recruitment of leukocytes (Wagener et al., 2003a). During the resolution phase of

inflammation HO-1 expression in leukocytes reduces adhesion molecules expression and leukocytes migration, thus contributing to wound healing (Wagener et al., 2003a). Although heme is a proinflammatory molecule, the feedback activity of HO-1 inhibits exacerbated inflammation and maintains homeostasis. However, the efficiency of HO-1 to counteract heme depends on the health state of the individual. Heme administration in mice without a genetic or pathologic condition that cause susceptibility to heme might have a therapeutic effect dependent on the activity of HO-1. In fact, heme-induced HO-1 expression prevents injury induced by indomethacin in the intestine (Yoriki et al., 2013), unilateral ureteral obstruction in the kidney (Correa-Costa et al., 2010), nutricional steatohepatitis (Nan et al., 2010), and IR in the liver (Xue et al., 2007).

## **INNATE IMMUNE RECEPTORS ACTIVATED BY HEME**

The immune system is characterized by the exquisite recognition of exogenous and endogenous molecules, and by its central role in tissue homeostasis as well as contributing to the pathogenesis of several diseases. The physiological consequence of an infection is the activation of an immune response essential to pathogen control or elimination (Medzhitov, 2008). Similarly, sterile tissue damage triggers inflammatory responses fundamental to tissue repair (Nathan and Ding, 2010). For years the mechanism of innate cell activation and function was considered non-specific, in opposition to the ability of B and T lymphocytes to recognize antigens by clonal receptors. The specific recognition of molecules from infectious agents and from endogenous origin by innate immune cells is performed by evolutionary conserved membrane and cytosolic receptors also known as pattern recognition receptors (PRRs; reviewed in Takeuchi and Akira, 2010). The presence and function of these receptors on mammals were originally proposed by Janeway (1989) and first confirmed by his group (Medzhitov et al., 1997). PRRs are germline-encoded and represent an essential arm of innate immune cells to sense invading microbes through conserved PAMPs and to detect endogenous stress signals through DAMPs. Several families of innate immune receptors have been described in recent years.

#### **HEME TRANSPORTERS**

For decades it has been appreciated that the transport of nutritional heme across cellular membranes of intestinal epithelial cells is an active and regulated process, more efficient than free Fe transport, requiring energy, and a selective transporter. The molecular nature of mammalian heme transporters remained elusive for many years, in part due to the lack of molecular and genetic tools and also as a consequence of a number of membrane proteins able to bind and transport heme at low affinity. In fact, several transporters of heme and other porphyrins have been described in recent years, although the *in vivo* importance of some of these molecules on heme trafficking is still object of debate. The cytoplasmic heme importer, named heme carrier protein 1 (HCP1), is expressed in the cytoplasmic membrane of duodenal epithelial cells and is believed to participate on dietary heme absorption (Shayeghi et al., 2005; Latunde-Dada et al., 2006). It was also demonstrated the critical *in vivo* importance of this putative heme transporter as

a mediator of extracellular folate import by intestinal epithelial cells (Qiu et al., 2006). Similarly, the ATP-binding cassette, Abcg2, involved on heme export is also associated with multidrug resistance (Krishnamurthy et al., 2004; Krishnamurthy and Schuetz, 2006). A second exporter of heme is the feline leukemia virus subgroup C receptor 1 (FLVCR1), in this case the lack of the transporter in mice affects erythroid differentiation and cell survival (Quigley et al., 2004; Keel et al., 2008). A mitochondrial isoform of FLVCR1 named FLVCR1b was recently demonstrated to export heme from the mitochondria to the cytosol (Chiabrando et al., 2012). The heme transporter ATPbinding cassette, Abcb6, controls the translocation of protoporphyrin between the cytosol and the mitochondria (Krishnamurthy et al., 2006). The heme-responsive gene-1 (HRG1) participates in the translocation of heme between the cytosol and endosomal compartments (Rajagopal et al., 2008).

An interesting study demonstrated that heme binds with high affinity to a heme-binding motif present at the intracellular domain of Slo1 channels, and inhibits transmembrane K currents by decreasing the frequency of channel opening (Tang et al., 2003). This was the first demonstration of an acute signaling effect of heme dependent on a specific transmembrane protein. Heme interactions with these membrane proteins led us to hypothesize that heme activates cells of the innate immune system through binding to an innate immune receptor.

#### **HEME IS AN AGONIST OF TOLL-LIKE RECEPTOR 4**

Using mouse strains deficient in different adaptors and innate immune receptors, we observed that heme-induced TNF production was dependent on MyD88 and Toll-like receptor 4 (TLR4; Figueiredo et al., 2007). TLRs are receptors present in mice and humans, which are distributed among the cellular membrane and intracellular compartments, as endosomes and lysosomes (Takeuchi and Akira, 2010). Each TLR recognize one or more PAMPs derived from different microbes such as bacteria, fungi, parasites, and virus (Takeuchi and Akira, 2010). Besides microbial agonists, TLRs triggers inflammatory responses to endogenous inflammatory molecules. DAMPs that activate TLRs include the high-mobility group box 1 (HMGB1) nuclear protein, myeloidrelated proteins (Mrp8/14), amyloid β, oxidized LDL (oxLDL), DNA, and RNA in the form of immune complexes, mitochondrial DNA (Oka et al., 2012) and extracellular matrix components such as tenascin-C, fibronectin, and hyaluronan (reviewed by Yu et al., 2010). Other DAMPs such as ATP, formyl peptides, uric acid crystals, and cholesterol crystals activate other innate immune receptors (Mariathasan et al., 2006; Martinon et al., 2006; Duewell et al., 2010; discussed later).

A critical step in characterizing an endogenous molecule as an agonist of an innate immune receptor is the necessity to exclude the presence of any microbial contaminant. In fact, in the case of heat shock proteins (HSPs), the effect on TNF secretion was later attributed to endotoxin contamination in the protein preparations (Gao and Tsan, 2003). Several strategies were used to exclude that the TLR4-induced activation by heme was due to contaminant (Figueiredo et al., 2007). Highly purified heme free of endotoxin contamination was used, as well as polymyxin B, anti-TLR4/MD2, and lipid A antagonist, all of which inhibited

the effects of LPS but did not interfere with the induction of TNF by heme. Similarly, protoporphyrin IX (PPIX), an antagonist of heme, did not inhibit the activity of LPS on TNF secretion (Figueiredo et al., 2007). The TLR4 engagement by LPS triggers two distinct pathways by the recruitment of the adaptor molecules MyD88 (Muzio et al., 1997; Wesche et al., 1997) and TRIF [TIR (Toll–IL-1 receptor)-domain-containing adapterinducing interferon (IFN)-β; Hoebe et al., 2003; Yamamoto et al., 2003]. The MyD88-dependent pathway leads to the activation of mitogen-activated protein kinases (MAPKs) and NF-κB (nuclear factor kappa-light-chain-enhancer of activated B cells) transcription factors inducing the expression of inflammatory cytokines such as TNF, IL-6, IL-1β, and KC (Takeuchi and Akira, 2010). The TRIF-dependent pathway activates the IRF3 (IFN regulatory transcription factor-3), which induce the expression of type I IFN, and late-phase NF-κB (Takeuchi and Akira, 2010). Heme seems to activate only the MyD88-dependent pathway and is ineffective to induce several cytokines or the expression of costimulatory molecules typical of TLR4 activation by LPS. It can induce MAPK activation and the expression of TNF and KC, but not of IFNβ or the IFN-dependent gene IP-10 (Figueiredo et al., 2007; **Figure 1**). Even the transcription of certain MyD88 dependent genes is not induced by heme when compared to LPS. In fact, different TLR4 ligands trigger signaling pathways differently. While, monophosphoryl lipid A, an analog of the TLR4

agonist tetra-acylated lipid A, induces a biased response to the TRIF pathway (Mata-Haro et al., 2007), the vesicular stomatitis virus glycoprotein G induces a biased response to the TRAM pathway to specifically induce IRF7 activation (Georgel et al., 2007). Moreover, human embryonic kidney (HEK) cells transfected with human TLR4 secretes IL-8 upon stimulation with heme (Piazza et al., 2011). In this case, the experiments were performed in the absence of CD-14 and MD-2 transfection.

The differences in signaling pathways activated by different agonists of the TLR4 might be due to (1) a different binding site on the TLR4, compared to LPS, or perhaps (2) the involvement of different co-receptors. (1) Molecular determinants of heme structure for TLR4 activation were determined (Figueiredo et al., 2007). Heme (Fe-PPIX) analogs, porphyrin rings without Fe (PPIX) or with metal substitutions (Pd-PPIX and Sn-PPIX), do not induce TNF production. Fe-mesoporphyrin IX, a heme analog with two ethyl groups substituting its vinyl groups, also is unable to activate TLR4. These data suggests that coordinated Fe and the vinyl groups are necessary for heme effect on TNF production. PPIX functions as an antagonist for TNF production induced by heme but not by LPS (Figueiredo et al., 2007). Conversely, LPS antagonists do not inhibit TNF production induced by heme. Interestingly, the sequence comparison of TLR4 and globins from different species suggests the presence of conserved heme-binding sequences in TLR4 (unpublished). These conserved sequences

**FIGURE 1 | Heme activates TLR4 in macrophages.** The TLR4 activates two distinct pathways: MyD88 and TRIF. In macrophages, heme induces a biased MyD88 activation and the secretion of the pro-inflammatory cytokines TNF and KC. TLR4 activation leads to MAPKs and NFκB activation, which are necessary to TNF secretion. Heme induces ROS generation independently of TLR4. However, ROS is necessary to induce TNF secretion and MAPK activation. Thus, TLR4

activation and ROS generation seem to be complementary and both are required for MAPKs and IκB degradation. It will be important to determinate the source of ROS and the mechanisms triggering heme-induced ROS generation. The mitochondria and the NADPH oxidase complexes seem to be involved. mROS scavenger (Mito-TEMPO) and NADPH oxidases inhibitors (apocynin and DPI) block TNF production induced by heme.

present a histidine which is essential for heme to bind globins. It is tempting to speculate that this histidine is essential for heme induced TLR4 activation. In fact, this same conserved histidine, together with two others non-conserved histidines, was shown to be important for the TLR4 activation induced by nickel in humans cells (Schmidt et al., 2010). The lack of these non-conserved histidines in the murine TLR4 disables the nickel capacity to activate this receptor. Moreover, this conserved histidine is important to Nickel, but not to LPS, to activate the TLR4. (2) Depending on the TLR4 activator, different co-receptors will be involved. LPS depends on MD2 and CD14 to activate TLR4 (Park et al., 2009). Amyloid β and oxLDL induce the assembly of a CD36– TLR4–TLR6 complex (Stewart et al., 2010), while hyaluronan is recognized by a complex formed by TLR4, MD2, and CD44 (Taylor et al., 2007). Heme depends on CD14 but not MD2 to activate TLR4 and induce TNF (Figueiredo et al., 2007; Piazza et al., 2011).

The crystal structure formed by heme polymerization, Hz, has pro-inflammatory properties dependent on TLR9 activation. However, it is uncertain whether pure Hz can directly activate TLR9. Because Hz isolated from *P. falciparum* cellular extract may contain DNA, this might be the TLR9 activator. In fact, it was shown that DNase treatment of *Plasmodium* Hz abolishes cytokine production by dendritic cells (DCs; Parroche et al., 2007). However, there is evidence that synthetic Hz, which is free of DNA, induces TLR9-dependent responses in DCs (Coban et al., 2005). Moreover, synthetic Hz induces conformational changes in recombinant TLR9, similar to those induced by CpG DNA, a well characterized TLR9 agonist (Coban et al., 2010). Interestingly, heme was able to induce similar conformational changes in the recombinant TLR9 (Coban et al., 2010). However, we demonstrated that TLR9 was not involved in heme-induced TNF production because WT and *Tlr9*−/<sup>−</sup> macrophages produce similar amounts of TNF when stimulated with heme (Figueiredo et al., 2007). Moreover, this demonstration suggests that mitochondrial DNA, a TLR9 activator (Oka et al., 2012), is not involved in the production of TNF by heme.

## **G PROTEIN-COUPLED RECEPTORS**

As mentioned before, heme induces LTB4 production by macrophages to stimulate neutrophil migration *in vivo* (Monteiro et al., 2011). Surprisingly, heme can also induce neutrophil migration *in vitro* thus indicating that heme acts as a chemotactic molecule (Graça-Souza et al., 2002; Porto et al., 2007). Chemokines and chemotactic molecules can induce neutrophil migration through G protein-coupled receptors (GPCR) which are sensitive to pertussis toxin (reviewed in Rot and von Andrian, 2004). Pertussis toxin inhibits heme-induced ROS generation and neutrophil migration, thus suggesting the involvement of a GPCR (Porto et al., 2007). Pharmacological inhibition of several pathways activated by GPCR [PI3K, PKC, phospholipase C (PLC)β, Rho, p38] demonstrated that ROS generation and neutrophil migration depends on the activation of signaling pathways characteristic of chemotactic receptors (Porto et al., 2007). Interestingly, PI3K and MAPK activation delays neutrophil apoptosis through a NADPH oxidase-dependent mechanism (Graça-Souza et al., 2002). This indicates that similar pathways might be involved in ROS generation, survival, and

migratory mechanisms triggered by heme in neutrophils. Furthermore, although several heme analogs were able to induce neutrophil migration, mesoporphyrins, molecules lacking the vinyl groups in their rings, were not chemotactic for neutrophils and selectively inhibited the migration induced by heme (Porto et al., 2007). We believe that mesoporphyrins functions as antagonist molecules blocking the activation of an extracellular receptor. Cytochrome *c* has a conserved sequence responsible for its interaction with heme. This same sequence is present in the Slo1 channel and was proved to be essential for its binding to heme (Tang et al., 2003). Interestingly, we found this conserved sequence in two chemotactic receptors expressed in neutrophils. Together, these results raises the possibility that heme activates neutrophils through the activation of a GPCR.

## **MECHANISMS INVOLVED IN HEME-INDUCED CELL DEATH AND INFLAMMATION**

### **CYTOTOXIC EFFECTS CONTROLLED BY SPECIFIC SIGNALING PATHWAYS**

Because of its lipid intercalating ability, heme was assumed to induce its cytotoxic and inflammatory effects mediated by its pro-oxidant and amphipathic properties. However, new possibilities emerged because of the new characterized mechanisms of inflammatory responses induced by heme. Heme activates TLR4 to induce cytokine production and necrotic cell death (Figueiredo et al., 2007; Fortes et al., 2012; **Figure 2**). Hemeinduced necrosis depends on TNF produced by TLR4 activation, because *Tlr4*−/<sup>−</sup> and *Tnfr*−/<sup>−</sup> macrophages are resistant to the cytotoxic effects of heme (Fortes et al., 2012). However, heme sensitizes *Tlr4*−/<sup>−</sup> macrophages to TNF-induced cell death, indicating that TLR4 is essential to heme-induced TNF production, but not directly required for the cell death induced by heme. The necrotic cell death induced by heme was dependent on ROS generation, since antioxidants prevented it (Fortes et al., 2012). Moreover, *Rip1*−/<sup>−</sup> and *Rip3*−/<sup>−</sup> cells were resistant to heme-induced necrotic cell death (Fortes et al., 2012), characterizing the type of programed cell death as necroptosis (Cho et al., 2009; He et al., 2009; Zhang et al., 2009). Interestingly, ROS induction is essential but not sufficient to cause cell death. In fact, TLR4 and TNFR deletions do not change the capacity of heme to induce ROS generation (Figueiredo et al., 2007). Therefore, it is clear that heme triggers ROS-induced sensitization of macrophages to TNFR-mediated cell death through RIP1 and RIP3 activation to promote necroptosis (**Figure 2**). Moreover, inhibition of Jun N-terminal kinase (JNK) also decreases heme-induced ROS and necroptosis (Fortes et al., 2012; **Figure 2**). Heme-induced necroptosis is reversed by deferoxamine, a Fe chelator. Fe can mediate non-enzymatic ROS generation by the Fenton reaction. However, since deferoxamine binds to heme (Baysal et al., 1990; Lu et al., 2012), we cannot exclude the possibility that its inhibitory effect is directly related to heme and not to free Fe. In fact, the protective effect of pharmacologic inhibition of NADPH oxidase and mitochondrial ROS (mROS) on hemeinduced TNF and macrophage necroptosis suggests that enzymatic ROS production is essential for the cytotoxic effects of heme (unpublished).

Heme also induces programed cell death in non-hematopoietic cells. Stimulation of astrocytes with heme, as a model of hemorrhagic injury, causes cell death with characteristics of programed necrosis including the loss of plasma membrane integrity, reversion by necrostatin-1, a selective inhibitor of RIP1, and by antioxidants (Laird et al., 2008). Extracellular-signal-regulated kinase (ERK) activation is also involved in astrocyte cell death induced by heme (Regan et al., 2001). Interestingly, heme-induced ERK activation in these cells is ROS-independent (Regan et al., 2001) indicating the involvement of a distinct pathway that might involve TLR4 signaling. In fact, TLR4 is involved in intracerebral hemorrhage (ICH) induced by heme (Lin et al., 2012). Microglia cells, another cell type present in the central nervous system, undergoes cell death through heme-induced activation of JNK and p38 (Cai et al., 2011). Moreover, heme induces apoptosis in human brain vascular endothelial cells (HBVEC) by STAT3 (signal transducer and activator of transcription 3)-dependent activation of matrix metallopeptidase 3 (MMP3; Liu et al., 2013). Heme also sensitizes other non-hematopoietic cells to TNF-induced cell death (Seixas et al., 2009; Larsen et al., 2010; Sukumari-Ramesh et al., 2010; Gozzelino et al., 2012). This cell death has morphological and biochemical characteristics of apoptosis including caspase activation, membrane blebbing, nuclear shrinking, and fragmentation, as well as chromatin condensation and formation of apoptotic bodies (Seixas et al., 2009). Treatment with

leads to MAPKs activation and TNF production. While JNK increases ROS generation, TNF induces RIP1–RIP3 necrosome which triggers necroptosis.

> antioxidants, increased expression of HO-1 or Hx also revert the heme/TNF-induced cell death in hepatocytes (Seixas et al., 2009; Gozzelino et al., 2010, 2012; Larsen et al., 2010; Gozzelino and Soares, 2011).

inhibition to induce necroptosis. Moreover, HO-1 has a protective effect

during heme-induced necroptosis.

Heme induces HO-1 that in turn protects cells from death (Vile et al., 1994; Poss and Tonegawa, 1997a; Ferris et al., 1999; Brouard et al., 2002; Arruda et al., 2004; Seixas et al., 2009; Gozzelino et al., 2010; Kovtunovych et al., 2010; Larsen et al., 2010). The ability of HO-1 to promote cytoprotection has been attributed to its role controlling heme amounts as well as Fe homeostasis and oxidative stress (Gozzelino et al., 2010). Mice lacking HO-1 (*Hmox1*−/−) has Fe overload, increase cell death and widespread inflammation (Poss and Tonegawa, 1997a,b; Kovtunovych et al., 2010). *In vitro* embryonic fibroblasts from *Hmox1*−/<sup>−</sup> mice are more sensitive to heme or H2O2 and present increase oxidative stress (Poss and Tonegawa, 1997a). Moreover, *Hmox1*−/<sup>−</sup> mice have reduced macrophage numbers in the spleen and liver due to cell death (Kovtunovych et al., 2010). This increased macrophage cell death was confirmed *in vitro* upon erythrophagocytosis, further suggesting that the *in vivo* exposure to heme is the cause of the increased macrophage death observed in *Hmox1*−/<sup>−</sup> mice. This result is corroborated by the increased necroptosis observed in *Hmox1*−/<sup>−</sup> macrophages exposed to heme compared to WT (Fortes et al., 2012). Treatment with antioxidants fully blocked heme-induced cell death of *Hmox1*−/<sup>−</sup> macrophages.

Importantly, increased expression of FtH also protects different cell types from the cytotoxic effects of heme, TNF or heme and TNF (Balla et al., 1992; Berberat et al., 2003; Cozzi et al., 2003; Gozzelino et al., 2012). These results suggest that Fe derived from heme catabolism participates in the heme-induced sensitization to cell death. The mechanism depends on the sustained activation of JNK induced by Fe-derived ROS. The increased JNK activation inhibits FtH expression, increasing the amount of labile Fe and consequently the amounts of ROS. The antioxidant property of FtH blocks TNF-induced JNK activation, reducing cell death (Pham et al., 2004; Kamata et al., 2005; Gozzelino et al., 2012). Thus, it seems that JNK activation is a common trigger of hemeinduced cell death in a variety of cells (macrophages, hepatocytes, and microglia; **Figure 2**). Thus heme induces programed cell death in different types of cells by the regulation of specific signaling pathways.

## **THE SPLEEN TYROSINE KINASE PATHWAY OF INFLAMMATION**

The cytotoxic effect of heme has important impact in different pathological conditions. This is in part due to the loss of essential tissue functions and in part due to the release of cellular contents with inflammatory activities. High amounts of free heme are found in infectious diseases, such as malaria and sepsis (Pamplona et al., 2007; Larsen et al., 2010), suggesting that heme can amplify inflammatory responses during infection. In fact, it was demonstrated that dying hepatocytes release HMGB1 upon heme and TNF challenge, which in turn increases the systemic inflammatory response (Larsen et al., 2010). Another mechanism involved in heme-induced inflammation during infectious conditions is the amplification of cytokine production induced by heme with microbial molecules. Hb synergizes with LPS enhancing the production of pro-inflammatory cytokines by macrophages (Yang et al., 2002). The opposite effect of globin indicates that heme moiety is the responsible for the potentiating effect of Hb. A low dose of LPS on its own that causes an insignificant cytokine secretion, together with heme results in substantial production of cytokines, thus suggesting that this synergism may be particularly important under conditions of low agonist concentrations, helping the control of infectious agents (Fernandez et al., 2010). This amplification response can also be deleterious under higher concentrations of heme and microbial molecules. *In vivo*, injection of heme and LPS induces a significant increase in the concentrations of TNF and IL-6 when compared to the challenge with LPS alone (Fernandez et al., 2010). Moreover, the co-injection of non-lethal doses of heme and LPS induces 100% lethality.

Interestingly, although purified Hx inhibits the synergy between heme and PAMPs, the synergy only occurs in the presence of serum, a condition that protects the cells against heme-induced TLR4-dependent TNF production and necroptosis. Heme synergizes with ligands of TLR2, TLR3, TLR4, TLR9, NOD1 (nucleotide-binding oligomerization domain 1), and NOD2 to increase TNF and IL-6 production by macrophages (**Figure 3**). The synergy between heme and LPS, a TLR4 activator, is induced by the enhanced activation of MAPKs (ERK, p38, JNK) and NF-κB (Fernandez et al., 2010; **Figure 3**). The mechanism depends on spleen tyrosine kinase (Syk) phosphorylation which regulates the ROS production induced by heme (**Figure 3**). Heme-induced Syk phosphorylation modulates ROS generation by activating PKC and calcium signaling through PLCγ, two downstream molecules regulated by Syk, which are involved in the synergy between heme and LPS (Fernandez et al., 2010). In fact, Syk modulates signaling pathways activated by TLRs cytokine production by macrophages and DC (Brown et al., 2003; Gantner et al., 2003; Turnbull et al., 2005). Syk is a signaling molecule activated by receptors that signal through immunoreceptor tyrosine-based activation motifs (ITAMs) suggesting the involvement of a receptor in the synergy between heme and PAMPs (Underhill and Goodridge, 2007; **Figure 3**). On the other hand, Syk is also activated independently of cell surface receptors by disturbances in lipid rafts domains (Ng et al., 2008; **Figure 3**). Because of heme's capacity to intercalate cellular membranes we cannot exclude the possibility that Syk might be activated by the direct contact of heme with cellular membranes. It will be important to investigate whether heme-induced Syk phosphorylation requires a receptor.

## **THE NLRP3 INFLAMMASOME**

While TNF and IL-6 are mainly regulated at the transcriptional and translational levels, cytokines such as IL-1β and IL-18 requires two steps to be produced. These cytokines are expressed as zymogens, pro-IL-1β and pro-IL-18, which are regulated by the synthesis of their mRNA by TLRs. However, IL-1β and IL-18 maturation requires cleavage of their zymogens by caspase-1 (Cerretti et al., 1992; Thornberry et al., 1992), which is activated independently of TLR signaling. The cleavage of these pro-forms is necessary for their biological effects. Caspase-1 is activated by a multienzymatic platform called the inflammasome, which is composed of NLRs (NOD-like receptors), ASC (apoptosis-associated speck-like protein), and caspase-1 (Martinon et al., 2002; Wen et al., 2013). NLRP3 (nucleotide-binding domain, leucine rich family, pyrin containing 3 gene) is the best characterized inflammasome receptor and is activated by several DAMPs such as ATP, Hz, amyloid β, uric acid crystals, and cholesterol crystals (Mariathasan et al., 2006; Martinon et al., 2006; Halle et al., 2008; Shio et al., 2009; Duewell et al.,2010; mechanisms reviewed in Latz et al.,2013). Three mechanisms are involved in the NLRP3 inflammasome activation: K+ efflux (Petrilli et al., 2007; Muñoz-Planillo et al., 2013), lysosome damage (Hornung et al., 2008), and mitochondrial dysfunctions that lead to mROS generation (Zhou et al., 2011) and the release of cardiolipin (Iyer et al., 2013) and mitochondrial DNA (Shimada et al., 2012).

In this context, we found that heme induces IL-1β production by activated LPS-primed macrophages promoting NLRP3 dependent processing of IL-1β (unpublished). Although heme is capable of activating TLR4, it does not induce IL-1β expression (Figueiredo et al., 2007). Heme (Fe-PPIX), but not its analogs, porphyrin rings without Fe (PPIX) or with metal substitutions (Pd-PPIX and Sn-PPIX), induces IL-1β processing and release (unpublished). The activation of NLRP3 by heme required K+ efflux, Syk phosphorylation, mROS, and NOX2. On the other hand, the mechanism was independent of ATP release and lysosomal damage (unpublished). Our results are in sharp contrast with a recent published report that describes the ability of heme to

**FIGURE 3 | Heme activates Syk and amplifies cytokine production induced by PAMPs.** Heme induces Syk phosphorylation in macrophages. The Syk pathway is essential for heme-induced ROS production. Heme-induced ROS generation increases MAPKs and NFκB activation and consequently, cytokine production. Heme amplifies cytokines induced by cell surface receptors (TLR2, TLR4, TLR5), endosome receptors (TLR3, TLR9), and cytosolic receptors (NOD1 and NOD2). Moreover, heme amplifies MyD88- (TNF and IL-6) and TRIF-dependent (IP-10) cytokines. The mechanism by which

activate the NLRP3 inflammasome (Li et al., 2014). (1) Heme and PPIX caused the maturation of IL-1β in the absence of serum (Li et al., 2014). In our experimental conditions, heme did not activate the inflammasome in the absence of serum (unpublished). This same result was previously published (Dostert et al., 2009). Interestingly, the synergism of heme with PAMPs regarding the production of TNF and IL-6 is also dependent on serum (Fernandez et al., 2010; Lin et al., 2010). The reason why heme requires serum to induce part of its inflammatory effects needs further investigation. (2) Heme and PPIX primed the macrophages and induced IL-1β mRNA expression (Li et al., 2014). It is possible that the procedure to elicit peritoneal macrophages used in their study primed the cells for subsequent inflammasome activation. We observed that PPIX did not induce pro-IL-1β or inflammasome activation, suggesting that the iron in the porphyrin ring is essential for the effect of heme (unpublished). (3) Heme-induced IL-1β secretion depended on P2X receptors (Li et al., 2014). This observation suggests that ATP release is involved. In fact, ATP release from necrotic cells activates NLRP3 (Iyer et al., 2009). Importantly, their experimental procedure was heme triggers Syk activation is not known. However, there are two possibilities. Heme could trigger Syk activation through an unknown surface receptor or through interaction with lipid rafts. The source of ROS controlled by Syk is not known but because PKC, NOX, and mitochondria inhibitors blocks the synergism between heme and PAMPs it is possible to consider that Syk controls ROS generation by NOX and mitochondria. Although NOX2 is the principal NADPH oxidase complex in phagocytes we cannot exclude the possibility that other NOX might be involved.

made without serum. This could have increased the cell death of macrophages stimulated with heme and the subsequent released of ATP. Differently from their results, we demonstrated that heme induced similar amounts of IL-1β in WT and P2X7-deficient macrophages (unpublished). Moreover, the use of oxidized ATP (ATP antagonist) and apyrase (degrades ATP) corroborate that heme-induced inflammasome activation is independent of ATP (unpublished).

One of the mechanisms responsible for the NLRP3 activation involves the phagocytosis of crystal structures, subsequent lysosome damage and cathepsin B release to the cytosol (Hornung et al., 2008). In fact, phagocytosis and cathepsin B inhibition blocks the inflammasome activation induced by crystals, including Hz (Shio et al., 2009). Like heme, Hz also depends on ROS generation and Syk phosphorylation to activate NLRP3 (Shio et al., 2009). Heme, on the other hand, does not require phagocytosis or lysosome damage to induce IL-1β secretion (unpublished). Thus, heme activates the NLRP3 inflammasome with a distinct mechanism compared to ATP and Hz.

*In vivo*, heme injection induced IL-1β production and caspase-1-dependent neutrophil migration in mice peritoneum (unpublished). Moreover, *Caspase-1*−/<sup>−</sup> mice were resistant to mortality induced by hemolysis. On the other hand, heme can be used to prevent inflammation through the induction of HO-1. In fact, preventive induction of HO-1 induced by heme inhibits the NLRP3 inflammasome activation in the liver (Kim and Lee, 2013) and the lung (Luo et al., 2014) in different pathological conditions.

## **POSSIBLE ROLES OF HEME-INDUCED INNATE IMMUNE RECEPTOR ACTIVATION ON PATHOGENESIS SICKLE CELL DISEASE AND β-THALASSEMIA**

Pathologies associated with intravascular hemolysis are the most commonly related to heme inflammatory effects. SCD and β-thalassemia are molecular blood disorders caused by mutations in genes encoding Hb (Heinle and Read, 1948; Pauling et al., 1949; Ingram, 1957). In SCD, a single point mutation in the Hb gene encodes a protein that polymerizes under low-oxygen conditions causing red blood cells (RBC) deformation (sickle shape; Edelstein et al., 1973; Browne et al., 1998). Repeated sickling episodes decrease RBC elasticity rendering them susceptible to hemolysis (Hebbel, 1991). β-Thalassemia is the result of a mutation in the β-globin chain gene that impairs the β-globin chain production and leads to accumulation of the α-globin which can aggregate causing hemolysis and erythroid precursor premature death (Khandros et al., 2012). Chronic episodes of hemolyses increase the concentrations of Hb and heme which are considered to be the major responsible for vascular inflammation in these diseases (Muller-Eberhard et al.,1968;Chies and Nardi,2001; **Figure 4**). In fact, hematin injection in healthy volunteers induces thrombophlebitis and disturbed homeostasis (Simionatto et al., 1988). Moreover, polymorphisms in the promoter region of the *Hmox1* gene is associated with decreased rates of hospitalization of patients with acute chest syndrome (ACS; Bean et al., 2012), a major life-threatening condition for patients with SCD (Gladwin and Vichinsky, 2008). Repeated vaso-occlusion episodes are characteristic of SCD and can lead to tissue damage due to IR (Kaul and Hebbel, 2000; **Figure 4**). Heme induces the expression of adhesion molecules in the vasculature (Wagener et al., 1997, 2001a; Belcher et al., 2014). The adherence of leukocytes and reticulocytes to the endothelium causes stasis and painful crisis, a hallmark of SCD pathogenesis (Belcher et al., 2000). β-thalassemic patients also experience signs of vascular inflammation and vasoocclusion due to endothelial adhesion of aggregated thalassemic erythrocytes and reduced NO bioavailability (Bunyaratvej et al., 1995; Hovav et al., 1999; Hahalis et al., 2008). In this disease, cardiac complications seem to be the main cause of mortality (Engle et al., 1964; Spirito and Maron, 1990).

Transgenic SCD mice present typical signs of multiorgan pathology and vascular inflammation such as oxidative stress, endothelial activation, and reduced NO bioavailability (Pászty et al., 1997; Ryan et al., 1997; Belcher et al., 2003). Also, β-thalassemic mice present the same signs of vasculopathy (Vinchi et al., 2013). This phenotype is similar to that of *Hx*−/<sup>−</sup> mice during hemolysis or heme overload (Tolosano et al., 1999; Vinchi et al., 2008). In both models, vascular inflammation was associated

with an increase is serum heme concentration and Hx depletion. This fact is also observed in patients with hemolytic anemia (Muller-Eberhard et al., 1968; Adisa et al., 2013). Hx administration protects transgenic SCD mice and β-thalassemic mice from heme-induced oxidative stress and vascular inflammation (Vinchi et al., 2013). Hx works by chelating and delivering heme to the liver where it is catabolized by HO-1. In this context, Hx-heme complex-induced HO-1 protects the liver against heme cytotoxic effects (Vinchi et al., 2013). The increased HO-1 expression reduces vascular inflammation, vaso-occlusion and liver damage (Belcher et al., 2000; Vinchi et al., 2008, 2013). While Hx treatment protects SCD mice with severe respiratory symptoms from acute lung inflammation (ALI; Ghosh et al., 2013) and SCD mice from stasis (Belcher et al., 2014), Hx also improves cardiovascular functions in β-thalassemic mice (Vinchi et al., 2013).

Recent studies demonstrated that heme-induced endothelial cell activation is at least in part mediated by TLR4 signaling. Heme-induced TLR4 activation in endothelial cells leads to NF-κB activation, adhesion molecules expression and Weibel– Palade body degranulation (Belcher et al., 2014), which contains the VWF and P-selectins, molecules involved in SCD vasculopathy (Matsui et al., 2001; Chen et al., 2011a; **Figure 4**). Blockade of adhesion molecules, including VWF and P-selectin, inhibits heme-induced stasis in SCD mice (Belcher et al., 2014). TLR4 is also involved in heme-induced ACS. The TLR4 specific antagonist, TAK-242, which inhibits TLR4 signaling by the intracellular domain, protects transgenic SCD mice from heme-induced stasis and ACS (Ghosh et al., 2013; Belcher et al., 2014). In fact, *Tlr4*−/<sup>−</sup> mice or TAK-242 administration prevented ALI and respiratory distress induced by heme (Ghosh et al., 2013). These results suggest that TLR4 could be a new therapeutic target for treating SCD.

Furthermore, antioxidants proved to be efficient modulators of vascular functions in hemolytic diseases. In fact, PKC inhibition and antioxidants also protects transgenic SCD mice from hemeinduced stasis (Belcher et al., 2014). PKC was shown to mediate heme-induced ROS generation in macrophages by the Syk pathway and in neutrophils by a signaling pathway characteristic of a GPCR (Porto et al., 2007; Fernandez et al.,2010). Because ROS generation induced by heme is TLR4-independent (Figueiredo et al., 2007), it is possible that another receptor might collaborate with TLR4 to promote vascular inflammation.

Another possible mechanism involved in the pathogenesis of SCD is the release of NETs by neutrophils (Chen et al., 2014). NETs are fibers composed of chromatin (DNA and histones) decorated with antimicrobial peptides. In contrast to its functions in host defense against pathogens, which NET exert by trapping and killing them, exaggerated NET formation is implicated in a number of pathologies (reviewed by Zawrotniak and Rapala-Kozik, 2013). In this context, NET formation has been shown to contribute to disseminated intravascular coagulation in the course of inflammatory diseases and therefore to morbidity and mortality in sepsis (Fuchs et al., 2010; Massberg et al., 2010; von Bruhl et al., 2012). Thus, this effect could contribute to vaso-occlusion and ACS in the course of SCD. In fact, patients with SCD presented NET formation in plasma during steady state conditions,

when compared to healthy individuals (Schimmel et al., 2013). The amount of NET formation was increased during painful crisis and ACS, and correlated with the length of hospitalization. Moreover, TNF administration in SCD mice induced NET formation within the pulmonary microvasculature (Chen et al., 2014). DNAse I treatment prevented lung inflammation and vascular permeability induced by TNF in SCD mice. In this study heme was reported to mediate NET formation in SCD mice. In vitro, TNFprimed neutrophils released NETs after the stimulation with heme. Furthermore, Hx treatment prevented TNF-induced NET formation and hypothermia in SCD mice demonstrating an important role for heme in the induction of NET formation in this model. Together, these data suggest that heme-induced NET formation might contribute to the pathogenesis of SCD (**Figure 4**).

## **ATYPICAL UREMIC HEMOLYTIC SYNDROME**

Atypical hemolytic uremic syndrome (aHUS) is characterized by an over activation of the complement alternative pathway (CAP) due to genetic and acquired abnormalities (Noris and Remuzzi, 2009). The pathogenesis of this disease involves glomerular endothelial damage, leukocyte activation, thrombotic microangiopathy (blood clots in small blood vessels), and mechanical hemolysis. However, the observed mutations are insufficient to induce complement abnormal activation requiring a triggering stimulus for disease manifestation. A broad range of precipitating events is related with aHUS including infections, drugs, autoimmune conditions, transplants, pregnancy, and metabolic conditions (reviewed in Loirat and Fremeaux-Bacchi, 2011). It was suggested that heme derived from hemolysis could act as a secondary hit capable of amplifying aHUS pathogenesis (deCiutiis et al., 1978; Mold et al., 1995; Ruiz-Torres et al., 2005; Pawluczkowycz et al., 2007). In fact, heme activates the CAP in the serum and on endothelial cells surfaces (Frimat et al., 2013). This effect is enhanced by mutations associated to CAP overactivation during aHUS. Heme-induced exocytosis of Weibel– Palade bodies from endothelial cells induces the expression of P-selectins in the plasma membrane, which are known to bind the complement C3b protein and trigger CAP activation, and to release the prothrombotic VWF (Frimat et al., 2013). Moreover, heme induces the release of C5a and C5b9, fragments of the C5 complement protein known to induce endothelial cell activation and permeabilization. Thus the combined effect of heme-induced CAP and TLR4 activation in endothelial cells (Frimat et al., 2013; Belcher et al., 2014) may act together to induce

Weibel–Palade body degranulation, contributing to renal damage in aHUS.

## **INTRACEREBRAL HEMORRHAGE**

Intracerebral hemorrhage is a bleeding into the brain parenchyma. Although the mass effect of an edema derived from an expanding hematoma mediates part of the brain injury, there are evidences pointing inflammation and cytotoxic degradation products of blood as mediators of ICH injury (Xi et al., 2006). While Hx injection protects brain tissues from rat cerebral IR (Dong et al., 2013), *Hx*−/<sup>−</sup> mice are susceptible to ICH (Chen et al., 2011b) suggesting a role for free heme in this process. In fact, bleeding will result in hemolysis and free heme release. Depending on experimental conditions, HO-1 can be deleterious or beneficial in neurodegenerative models. Interestingly, in a mouse model of ICH, *Hmox*−/<sup>−</sup> mice present reduced brain damage (Wang and Doré, 2007) suggesting that cerebral cells are prone to the oxidative effects of free Fe derived from heme catabolism. In fact, the treatment with the Fe chelator deferoxamine ameliorates brain injury after ICH or subarachnoid hemorrhage (Gu et al., 2009; Lee et al., 2010; Chen et al., 2011c).

TLR4 has been involved in brain injury during ICH (Teng et al., 2009; Sansing et al., 2011; Fang et al., 2013; Wang et al., 2013). Exogenous heme induces brain injury through TLR4 induced inflammation (Lin et al., 2012). In fact, *Tlr4*−/<sup>−</sup> or anti-TLR4 treatment suppresses heme-induced neuroinflammation, edema, and neurologic deficit. Moreover, heme activates microglia cells to produce TNF, IL-6, and IL-1β through TLR4 signaling indicating that heme can directly activate brain cells (Lin et al., 2012). Heme-induced IL-1β production during neuroinflammation suggests the involvement of the NLRP3 inflammasome. In macrophages, heme induces IL-1β secretion via mROS generation and NLRP3 activation (Li et al., 2014; unpublished data). Interestingly, NLRP3 knockdown and mROS inhibitors reduce brain edema and improve neurological functions (Ma et al., 2014).

## **ATHEROGENESIS**

Atherosclerosis is a chronic inflammatory disease characterized by the formation of plaques in the vessel walls (Libby et al., 2002). Plaque formation is initiated by lipid accumulation in subendothelial spaces and metabolic dysfunction in infiltrating monocyte-derived macrophages which contributes to foam cells development (Moore and Tabas, 2011). Part of these cells die and release their fat and cholesterol-laden membranes into the intercellular space recruiting more monocytes in a chronic process that increases the formation of the necrotic core (Guyton and Klemp, 1996; Virmani et al., 2005). The necrotic core rupture can lead to serious complications. The release of plaque components into the blood stream form clots that damages the heart muscle and slowly develop ischemic heart disease (Falk et al., 1995; Arroyo and Lee, 1999). Also, if these clots are big enough they can block blood flow and induce tissue necrosis. On the other hand, the necrotic core rupture exposes inflammatory components which can bind platelet cells at the rupture site and induce their accumulation forming thrombus which can be large enough to produce vessel lumen narrowing (Virmani et al., 2005).

Low-density lipoprotein is the major lipid involved in plaque formation. In the inflammatory environment, subendothelial LDL is oxidized and becomes a potent inflammatory molecule. oxLDL activates intraplaque macrophages by the TLR4–TLR6– CD36 complex engagement leading to the activation of the NFκB pro-inflammatory pathway and facilitating endocytosis (Stewart et al., 2010). oxLDL endocytosis leads to the formation of crystalline substances inside lysosomes causing the destabilization of this organelle as well the release of proteases, events that lead to the NLRP3 inflammasome activation (Duewell et al., 2010). Indeed, TLRs and NLRP3 have been associated with atherosclerosis development.

Low-density lipoprotein oxidation and its inflammatory effects can be enhanced by the development of new vessels inside the necrotic core (Virmani et al., 2005). The major stimulus for this angiogenic process seems to be hypoxia (Bjornheden et al., 1999; Sluimer et al., 2008). In fact, the plaque atheroma composition decreases oxygen availability rendering the necrotic core structure a hypoxic condition. Furthermore, inflammatory cells can actively contribute to angiogenesis by secretion of angiogenic factors, such as vascular endothelial growth factor (VEGF), fibroblast growth factor 2 (FGF2), and plateletderived growth factor BB (PDGFBB) ligands (Cheng et al., 2013). The new vessels generally arise from vasa vasorum (Kwon et al., 1998), a network of small blood vessels that supply the walls of large blood vessels. Plaque new vessels are immature and present a fragile structure which is prone to leakage and highly vulnerable to rupture, thus causing intraplaque hemorrhage (Kolodgie et al., 2003; Dunmore et al., 2007; Sluimer et al., 2009). This event is responsible for increased entry of erythrocytes and inflammatory cells inside the necrotic core. Erythrocytes that enter developing atherosclerosis lesions are prone to lysis by lipid oxidation products. Erythrocytes lysis increases the amount of extracellular Hb which in a highly oxidative environment leads to MetHb formation and the release of free heme. LDL oxidation induced by heme increases the oxidation of lipids and inflammatory stimuli, contributing to further plaque development (Balla et al., 1991a). In fact, hemeinduced LDL oxidation is highly cytotoxic for endothelial cells and LDL oxidation seems to be mediated by Fe (Jeney et al., 2002; Nagy et al., 2010). This group of events, triggered by plaque hemorrhage, increases plaque vulnerability and consequently raises the risk of thrombus formation and ischemic heart disease (Michel et al., 2011). Expression of HO-1 provides protection against several cardiovascular diseases, including atherosclerosis (reviewed by Wu et al., 2011). Interestingly, a 6-year-old patient with HO-1 deficiency experienced a severe atherosclerotic pathology (Yachie et al., 1999). This patient suffered from persistent intravascular hemolysis, anemia, endothelial damage, iron deposition, and cellular sensibility to heme cytotoxic effects. These observations resemble those of *Hmox*−/<sup>−</sup> mice, thus demonstrating the importance of HO-1 functions in the maintenance of iron recycling and tissue homeostasis.

After hemolysis, sustained interaction between Hb and ROS can lead to ferrylhemoglobin (ferrylHb) formation, which is characterized by an increase in the Fe oxidative state to Fe+<sup>4</sup> (Harel and Kanner, 1988; Patel et al., 1996). ferrylHb undergoes intermolecular cross-linking of its globin chains forming aggregates, which induces the expression of adhesion molecules in vascular endothelial cells that support the recruitment of macrophages into the vessel wall (Silva et al., 2009; Potor et al., 2013). Similarly to heme, ferrylHb activates endothelial cells through NFκB activation (Silva et al., 2009). However, the mechanism responsible for NFκB activation is different. While TLR4 and ROS mediate heme-induced activation of macrophages and endothelial cells (Figueiredo et al., 2007; Belcher et al., 2014), ferrylHb activates endothelial cells independently of TLR4 and ROS (Silva et al., 2009). Moreover, ferrylHb is unable to induce cytokine secretion by endothelial cells (Silva et al., 2009), another difference to heme which induces IL-8 production (Natarajan et al., 2007). Like MetHb, ferrylHb is unstable and releases free heme to further increase oxLDL formation (Potor et al., 2013). Thus, heme and ferrylHb can induce direct effects, such as endothelial cells activation, and indirect effects, like LDL oxidation to increase atheroma plaque development and atherosclerosis pathogenesis.

Besides heme pro-inflammatory capacity in atherosclerotic plaque development, it was proposed that heme could mediate an atheroprotective event. Recently, heme was shown to activate human monocytes by the induction of activating transcription factor 1 (ATF-1), coupling the expression of HO-1 and liver X receptor (LXR)-β to induce a specific macrophage phenotype (Boyle et al., 2012). HO-1 is well known for its homeostatic functions creating an antioxidant and cytoprotective environment against heme and Fe deleterious effects (Gozzelino et al., 2010). LXR-β protects against lipid overload by activating a lipid exportation program regulated by proteins such as LXR-α and ATP binding cassette transporter A1 (ABCA1), preventing foam cells formation. This phenotype induced by heme is distinct from M1, M2, and Mox (reviewed in Moore et al., 2013) and therefore was named Mhem. Mhem was described in macrophages from atheroma plaques of patients with fatal coronary artery disease (Boyle et al., 2009). In this context, ATF-1 is able to convert macrophages to an atheroprotective state coordinating Fe and lipid metabolism.

The Mhem phenotype contrasts with the Fe-loaded M1 macrophages phenotype which presents an enhanced production of TNF and hydroxyl radicals, and has the capacity to induce precocious fibroblast senescence, impairing wound healing (Sindrilaru et al., 2011). Macrophages with this phenotype were described in the skin of patients with chronic venous leg ulcers (CVUs). This disease is caused by chronic venous valve insufficiency that leads to hypertension in the lower-limb veins, with persistent erythrocyte extravasation. Erythrophagocytosis by tissue macrophages and release of heme-bound Fe seems to be the trigger of this pro-inflammatory phenotype. It will be interesting to investigate the presence of Mhem in other pathologies such CVUs to understand the environments and factors involved in the development of Mhem and its capacity to modulate proinflammatory microenvironments also induced by heme and Fe. Moreover, it will be interesting to characterize the putative involvement of innate immune receptors activated by heme in these processes.

### **HEMOLYSIS DUE TO INFECTION**

Vascular hemolysis is a relevant factor in some infectious diseases. The severe pathology observed in malaria and hemorrhagic fevers is a result of complex events that triggers a vigorous inflammatory response. The systemic inflammatory response and the disseminated intravascular coagulation observed in severe cases of these diseases are similar to septic shock and are also dependent on the production of high amounts of inflammatory mediators such as TNF, IL-6, and IL-1β. Heme homeostasis is critically involved in the development of sepsis and malaria (Pamplona et al., 2007; Larsen et al., 2010; Gozzelino et al., 2012). In both cases it is possible to observe high amounts of heme in the circulation (**Figure 4**). In malaria, it was demonstrated that heme acts together with high amounts of ROS to induce the disruption of the blood–brain barrier permitting the migration of inflammatory cells, plasmatic proteins, and *Plasmodium* antigens, which cause damage to brain tissues (Pamplona et al., 2007). In the CLP (cecal ligation and puncture) model of sepsis, heme sensitizes hepatocytes to necrotic cell death rendering mice more susceptible to sepsis (Larsen et al., 2010). Moreover, there is a negative correlation between Hx serum concentration and tissue damage in patients with septic shock. Therefore, patients with septic shock presenting higher serum concentrations of Hx develop decreased tissue damage and have a better survival outcome, suggesting an important role for heme during sepsis (Larsen et al., 2010). As discussed, Hx inhibits several pro-inflammatory effects induced by heme such as ROS generation, cell death (Larsen et al., 2010; Gozzelino et al., 2012), and the synergistic effect (Lin et al., 2010) with PAMPs. HO-1 has a central role in maintaining tissues homeostasis against heme noxious effects, since *Hmox*−/<sup>−</sup> mice are extremely susceptible to CLP and cerebral malaria (Pamplona et al., 2007; Larsen et al., 2010). In both cases, HO-1 maintains tissue homeostasis independently of parasite burden. Thus, HO-1 confers tolerance against these infectious diseases. On the other hand, HO-1 induction confers host resistance, rather than tolerance, to *Mycobacterium* infection (Silva-Gomes et al., 2013). *Hmox1-/-* mice infected with *M. avium* presented a susceptible phenotype with an increased pathogen burden due to impaired protective granuloma formation. HO-1 deficiency induced heme accumulation and macrophage cell death which contributed to *M. avium* proliferation. In fact, heme administration to macrophages in vitro increased *M. avium* proliferation and heme injection in infected mice prevented granuloma formation. Moreover, *Hmox1-/-* mice infected with M. tuberculosis died while WT and heterozygous mice (*Hmox1+/-*) survived. Thus, HO-1 plays a critical role during *Mycobacterium* infection by preventing heme-induced granuloma macrophage death and bacterial proliferation. Besides its role in tissue tolerance and bacterial control, HO-1 antioxidant activity might contribute to intracellular bacterial multiplication. In fact, HO-1 impairs resistance to *Plasmodium* infection in the liver (Epiphanio et al., 2008) and to non-typhoid *Salmonella* (NTS), a common complication of *P. falciparum* infection (Cunnington et al., 2012a,b). In fact, hemolysis triggered by *Plasmodium* infection causes premature mobilization of bone-marrow granulocytes with impaired antioxidant defenses due to HO-1 expression (Cunnington et al., 2012a,b). This leads to uncontrolled bacterial load and lethal bacteremia. Together these observations reinforce the concept that

heme might be used as a target for adjuvant therapies during infectious diseases.

### **FUTURE PERSPECTIVES**

The last 10 years of research on heme inflammatory properties brought new insights to understand its functions in hemolytic and hemorrhagic pathologies. Besides the fact that heme presents direct cytotoxic properties, it is now clear that heme can activate specific receptors and signaling pathways to promote ROS generation, inflammation, and programed cell death. Important studies demonstrated that heme-induced TLR4 activation is involved in the pathogenesis of SCD and ICH. Interestingly, a specific inhibitor of TLR4 signaling is very efficient preventing heme-induced vasculopathy and lung injury in mouse models of these diseases. Moreover, heme activates the NLRP3 inflammasome inducing IL-1β processing and secretion. The critical role of the inflammasome in the pathological response following lethal hemolysis was unveiled. Moreover, new studies are helping to understand the beneficial properties of heme scavenging proteins such as Hx. In fact, Hx was shown to protect mice against infectious (malaria and sepsis) and non-infectious (SCD, β-thalassemia, and cerebral IR) models of diseases. Characterizing the role of cell death pathways induced by heme and the participation of heme transporters on hemolytic diseases promises interesting perspectives. Thus, understanding the molecular signaling pathways affected by heme might prove useful to the identification of new options for treating pathological conditions that course with increased extracellular heme and inflammation.

#### **ACKNOWLEDGMENTS**

We thank Miriam Werneck for critical reading of the manuscript. Marcelo T. Bozza received financial support from Conselho Nacional de Pesquisa (CNPq), Fundação de Amparo à Pesquisa do Estado do Rio de Janeiro (FAPERJ), and INCT Dengue, Brazil. Fabianno F. Dutra has a postdoctoral fellowship from Programa Nacional de Pós-Doutorado (PNPD) CAPES/FAPERJ.

#### **REFERENCES**


comparison between alpha- and beta-thalassemia. *Southeast Asian J. Trop. Med. Public Health*. 26(Suppl. 1), 257–260.


Chen, L., Zhang, X., Chen-Roetling, J., and Regan, R. F. (2011b). Increased striatal injury and behavioral deficits after intracerebral hemorrhage in hemopexin knockout mice. *J. Neurosurg.* 114, 1159–1167. doi: 10.3171/2010.10.JNS10861


structures of human heme oxygenase-1. *J. Biol. Chem.* 278, 7834–7843. doi: 10.1074/jbc.M211450200


Nathan, C., and Ding, A. (2010). Nonresolving inflammation. *Cell* 140, 871–882. doi: 10.1016/j.cell.2010.02.029


for the pathogenesis of anemia in malaria. *J. Immunol.* 179, 5543–5552. doi: 10.4049/jimmunol.179.8.5543


antioxidant properties. *Free Radic. Biol. Med.* 28, 289–309. doi: 10.1016/S0891- 5849(99)00223-3


receptor signaling pathway. *Science* 301, 640–643. doi: 10.1126/science. 1087262


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

*Received: 09 March 2014; paper pending published: 03 April 2014; accepted: 29 April 2014; published online: 27 May 2014.*

*Citation: Dutra FF and Bozza MT (2014) Heme on innate immunity and inflammation. Front. Pharmacol. 5:115. doi: 10.3389/fphar.2014.00115*

*This article was submitted to Drug Metabolism and Transport, a section of the journal Frontiers in Pharmacology.*

*Copyright © 2014 Dutra and Bozza. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*

## Iron, anemia and hepcidin in malaria

## *Natasha Spottiswoode1,2 , Patrick E. Duffy1 and Hal Drakesmith2 \**

<sup>1</sup> Laboratory of Malaria Immunology and Vaccinology, National Institute of Allergy and Infectious Diseases, National Institutes of Health, Bethesda, MD, USA <sup>2</sup> MRC Human Immunology Unit, Weatherall Institute of Molecular Medicine, University of Oxford, Oxford, UK

#### *Edited by:*

Raffaella Gozzelino, Instituto Gulbenkian de Ciência, Portugal

#### *Reviewed by:*

Carla Cerami, University of North Carolina at Chapel Hill, USA Aubrey Cunnington, Imperial College London, UK

*\*Correspondence:* Hal Drakesmith, MRC Human Immunology Unit, Weatherall Institute of Molecular Medicine, University of Oxford, Oxford, OX3 9DS, UK e-mail: hdrakes@hammer. imm.ox.ac.uk

Malaria and iron have a complex but important relationship. Plasmodium proliferation requires iron, both during the clinically silent liver stage of growth and in the diseaseassociated phase of erythrocyte infection. Precisely how the protozoan acquires its iron from its mammalian host remains unclear, but iron chelators can inhibit pathogen growth in vitro and in animal models. In humans, iron deficiency appears to protect against severe malaria, while iron supplementation increases risks of infection and disease. Malaria itself causes profound disturbances in physiological iron distribution and utilization, through mechanisms that include hemolysis, release of heme, dyserythropoiesis, anemia, deposition of iron in macrophages, and inhibition of dietary iron absorption. These effects have significant consequences. Malarial anemia is a major global health problem, especially in children, that remains incompletely understood and is not straightforward to treat. Furthermore, the changes in iron metabolism during a malaria infection may modulate susceptibility to co-infections. The release of heme and accumulation of iron in granulocytes may explain increased vulnerability to non-typhoidal Salmonella during malaria. The redistribution of iron away from hepatocytes and into macrophages may confer host resistance to superinfection, whereby blood-stage parasitemia prevents the development of a second liver-stage Plasmodium infection in the same organism. Key to understanding the pathophysiology of iron metabolism in malaria is the activity of the iron regulatory hormone hepcidin. Hepcidin is upregulated during blood-stage parasitemia and likely mediates much of the iron redistribution that accompanies disease. Understanding the regulation and role of hepcidin may offer new opportunities to combat malaria and formulate better approaches to treat anemia in the developing world.

**Keywords: hepcidin, malaria, iron, anemia, global health**

## **INTRODUCTION**

Among the many nutrients required for human survival, iron plays a unique role in determining disease susceptibility. This relationship is especially well studied in the case of malaria. Parasites of the genus *Plasmodium* cause malaria, although "malaria" is generally used to refer to symptomatic *Plasmodium falciparum* infection, unless otherwise specified. Malaria is currently one of the most geographically widespread and deadly diseases, and responsiblefor the deaths of an estimated 600,000 people per year (Malaria World Report, 20131). The controversial relationships between malaria and iron have been the subject of widespread discussion and debate by the global health community since 2006, when a randomized large-scale trial on the island of Pemba found that iron supplementation in children was linked with an increase in malaria-related mortality (Sazawal et al., 2006). Two Cochrane reviews (Ojukwu et al.,2009; Okebe et al.,2011) published subsequent to this finding have not resolved this conundrum for physicians and policymakers. Furthermore, while iron supplementation appears to be linked with increased malarial mortality, malaria infections are a major global cause of anemia (Kassebaum et al., 2014); and measures taken to decrease malaria at a population level also decrease anemia (Meremikwu et al., 2012).

The interactions between malaria and iron have only lately begun to be understood at the molecular level. Primarily, the discovery of the iron regulatory hormone hepcidin has given us new understanding of human iron physiology and pathophysiology. Hepcidin serves to block iron absorption from the diet and also to route iron in the body into macrophages and away from the serum. Hepcidin plays a complex but vital role in both the iron restriction that occurs during malaria infection, and in determining iron status and thereby influencing disease susceptibility. In this review, we examine the known interactions between physiological iron deficiency or repletion, iron supplementation, malaria, and hepcidin, and offer recommendations and suggestions for future work.

## **IRON DEFICIENCY PROTECTS FROM MALARIA INFECTION**

Variations in the iron levels of susceptible hosts may modulate the frequency and clinical severity of malaria infections. Gwamaka et al. (2012) collected detailed data from a large cohort of Tanzanian children (birth – 3 years). In this vulnerable population, iron deficiency at healthy aparasitemic visits was strongly associated with decreased future risk of parasitemia and severe malaria (Gwamaka et al., 2012). Iron deficiency in this study was defined by low ferritin (<30 ng/mL) in individuals with low C-reactive protein (CRP). A higher ferritin cutoff (<70 ng/mL) was used to

<sup>1</sup>http://www.who.int/malaria/publications/world\_malaria\_report\_2013/en/

define individuals with higher CRP as iron deficient; plasma ferritin is considered to be representative of iron stores in healthy individuals but increases acutely in infections.

A further study in a slightly older cohort of Kenyan children (8 months–8 years) found that iron repletion (defined as ferritin ≥ 12 ng/mL, with transferrin saturation ≥10%, children with high CRP excluded), was predictive of clinical malaria episodes in the year following measurement (Nyakeriga et al., 2004). Separately, Jonker et al. (2012b) noted a similar effect in Malawian children (6 months–5 years): iron-deficient children (defined as plasma ferritin <30 ng/mL), had a lower incidence of clinical malaria the subsequent year.

Studies performed in pregnant women also may indicate an association between iron deficiency and protection from clinical malaria, although somewhat less data are available. Two crosssectional studies have found that at the time of delivery, placental malaria was associated with iron replete status (Kabyemela et al., 2008; Senga et al., 2011). A limitation of this approach is that all currently used measures of iron status, such as ferritin, can be distorted by inflammation and infection, and thus cross-sectional studies may be of limited utility in understanding this relationship. Only one study has so far attempted to examine the predictive values of iron status on future placental or peripheral parasitemia in pregnant women (Senga et al., 2012). This study measured iron status by examining zinc protoporyphyrin (ZPP) levels. At both the first antenatal visit and at delivery, levels of ZPP indicative of iron repletion were associated with parasitemia; however, analyzing these data for ZPP as a predictive measure were complicated by the elevation of ZPP by concurrent parasitemia. The authors concluded that in this population, with a high incidence of parasitemia and associated inflammation, ZPP alone is not a valid measure of iron status, a concern that also could be applied to pediatric populations. Although the available data suggest a relationship, further studies that recruit women early in pregnancy, stratify carefully by gravidity and malaria transmission intensity, and follow their outcomes closely are required to definitively answer the question of whether iron status in women can predict malaria risk, as in pediatric populations.

## **IRON SUPPLEMENTATION AND MALARIA INFECTION**

While iron deficiency appears to offer some protection against malaria, iron supplementation may increase the vulnerability of susceptible populations to infection. Consequently, the use of population-scale iron supplementation in malaria-endemic areas is currently highly controversial. Early studies postulated a link between oral intake of iron and malaria susceptibility (Murray et al., 1978), as exemplified by flare-ups of latent infections following refeeding (Murray et al., 1975; Murray and Murray, 1977) or differing malaria susceptibility associated with different diets (Murray et al., 1980). In 2006, the "Pemba trial," a large-scale, randomized trial of iron supplementation in an area with very high malaria transmission, was stopped prematurely when trial monitoring boards found a link between the supplementation of children with iron and folic acid and subsequent malaria infection and mortality (Sazawal et al., 2006). Several studies that followed the Pemba trial produced results that appeared to contradict this finding, showing either no association between supplementation

and malaria risk (Desai et al., 2003; Ouedraogo et al., 2008), or a protective effect of iron supplementation (Zlotkin et al., 2013). However, many of the studies that followed the Pemba trial were smaller in scale and/or introduced measures to combat malaria, such as bednets or intermittent preventative chemoprophylaxis, which may have masked any increase in malaria susceptibility.

Two Cochrane reviews have since been published in attempts to reconcile the apparently disparate findings of the Pemba study and subsequent trials (Ojukwu et al., 2009; Okebe et al., 2011). Both reviews concluded that iron supplementation did not increase the risk of malaria infection in children when "regular malaria surveillance and treatment services"were provided. However, many areas of the world have inadequate malaria surveillance and treatment, coupled with high levels of iron deficiency and anemia. We would posit that two major questions remain unanswered by the current literature.

First, to what degree must malaria be controlled, or treatment and prevention practices put in place, before the benefits of giving iron outweigh the risks? This question remains practically difficult to answer due to the ethical difficulties inherent in providing iron supplementation without introducing malariareduction measures, but further guidance is required for clinicians, nutritionists, and policymakers.

A second and related question is: should iron supplementation be restricted to those children who are anemic and/or iron deprived? The updated Cochrane review found that the benefits of iron supplementation were greatest among children who had the lowest hemoglobin levels at baseline (Okebe et al., 2011). Not only might these children reap the greatest benefits from supplementation, they may also be less likely to exhibit increased malaria incidence or severity as a consequence of increased iron stores. In a substudy of the Pemba trial, in which children were more closely monitored and given malaria-reduction measures such as insecticide-treated bednets (ITNs), the authors did not find that iron supplementation associated with increased malaria incidence or mortality (Sazawal et al., 2006). Moreover, children in this substudy who were iron deficient and anemic at baseline showed a significant *reduction* in adverse events, including malaria episodes, when supplemented with iron and folic acid (Sazawal et al., 2006). This finding was apparently echoed by a subsequent study that examined the effect of food fortification (micronutrient powders with or without iron) on malaria risk in Ghanaian children (Zlotkin et al., 2013), which found that children who were both iron deficient and anemic at baseline showed reduced incidence of malaria following iron supplementation. However, as was mentioned in an editorial on this study (Prentice et al., 2013) in the Ghanian trial, the ironcontaining micronutrient powders were not actually effective at reducing anemia, making these results challenging to interpret. Moreover, iron supplementation in this trial was found to be associated with increased hospital admissions during the intervention period.

Targeting iron supplementation toward iron-deficient or anemic children is complicated by both practical and fiscal obstacles. As previously stated, most indicators of iron status can be confounded by infection or inflammation, limiting their interpretation in the field, and the widespread use of such tests may be difficult to fund. A solution to the problem may be to find more efficient ways to avoid giving iron to individuals who are iron replete, infected, or at high risk of infection (Atkinson et al., 2014). For example, in many areas, malaria transmission and prevalence of iron deficiency is seasonal; iron supplementation given outside of the malaria season may be less likely to increase susceptibility to infection, although studies are required to assess whether sub-patent parasitemias during the dry season may be increased by iron supplementation and cause illness.

When considering supplementation guidelines in pregnant women, it should be considered that primigravidae are both more susceptible to malaria infection and less prone to iron deficiency. One study demonstrated that intravenous iron supplementation increased malaria frequency in primigravidae but not multigravidae (Oppenheimer et al., 1986). Studies focusing on the effects of randomized oral iron supplementation have produced differing results (Menendez et al., 1994; Nacher et al., 2003), but relatively few have been performed thus far. In a large cohort study, it was found that iron-replete primigravidae are significantly more likely to experience placental malaria, but this effect loses significance in multigravidae (Kabyemela et al., 2008; Friedman et al., 2009). Summarizing the available evidence, adjusted supplementation guidelines that recommend lower iron supplementation to primigravidae may be advantageous.

## **HOW THE MALARIA PARASITE BENEFITS FROM IRON**

Iron is a limiting factor for the growth of many bacterial or protozoan pathogens. Iron-chelating agents such as desferrioxamine have been shown to restrict malaria growth *in vitro* (Raventos-Suarez et al., 1982), in murine models of malaria infection (Fritsch et al., 1985; Ferrer et al., 2012), and in malaria-infected monkeys (Pollack et al., 1987) (reviewed in Mabeza et al., 1999). Consequently, iron-chelating agents have been considered as an adjunct or primary antimalarial therapy.

In humans, preliminary work on the use of the iron chelators desferrioxamine, or the orally administered deferiprone, as an adjunct to standard antimalarial therapy seemed promising (Traore et al., 1991). However, after further studies showed no clear benefit from administration of chelators, and one seemed to hint at a slight increase in mortality in the trial group treated with desferrioxamine (Thuma et al., 1998), a Cochrane review recommended that trials testing iron chelators for malaria treatment be discontinued (Smith and Meremikwu, 2003).

The development of antimalarial drugs that target the parasite's access to iron might meet with more success if we had a better understanding of how *Plasmodium* parasites acquire iron during their various life stages. Blood-stage parasites have been theorized to acquire iron from serum transferrin (Rodriguez and Jungery, 1986), from iron produced during the breakdown of hemoglobin, or from a free pool of intracellular iron (Hershko and Peto, 1988), but this important issue remains unresolved. Very little work has been done that investigates the acquisition of iron by the obligate liver or mosquito life stages of malaria infection; these continue to be fruitful areas for future research. For example, identification of *Plasmodium*-encoded

iron transporters would both increase our understanding of the mechanism by which different life stages obtain their iron, and provide new drug targets aimed at inhibiting parasite growth.

Iron repletion may also have effects on parasite growth through different mechanisms than direct utilization by the parasite. A full synopsis of the literature is beyond the scope of this review, but iron has many effects on the immune system (McDermid and Prentice, 2006). In malaria specifically, it has been suggested that parasitized red blood cells from iron-deficient hosts may be more efficiently phagocytized, based on evidence from a murine model (Matsuzaki-Moriya et al., 2011). Finally, some *Plasmodium* species, notably *P. vivax*, live preferentially in young reticulocytes. Reticulocyte production can be restricted in highly anemic hosts, and this has been proposed as a mechanistic explanation for a study that showed an inverse association between severe anemia and *P. vivax* infection (Manning et al., 2012).

## **HUMAN IRON CONTROL: THE HORMONE HEPCIDIN**

Systemic mammalian iron metabolism is controlled at the level of iron absorbance from the diet and iron recycling through macrophages. Approximately 1 mg of iron is absorbed from the diet every day, roughly equivalent to the iron that is lost daily in poorly regulated activities such as sweating, any bleeding, and the sloughing off of enterocytes. At the same time, iron already in the body is constantly being recycled as macrophages phagocytose senescent or damaged red blood cells, digest the heme and extract the iron they contain, and export that iron back into the circulation.

The export of iron across the basolateral membrane of enterocytes and the recycling of iron through macrophages are dependent on the same transport protein: ferroportin, the sole currently identified mammalian iron export protein (Abboud and Haile, 2000; Donovan et al., 2000; McKie et al., 2000). The release of iron from enterocytes into the bloodstream is the final step of absorption of iron from the diet. Hepcidin, discovered a decade ago by three groups working independently (Krause et al., 2000; Pigeon et al., 2001; Park et al., 2001), is a 25-amino acid protein that binds to ferroportin and causes it to be internalized and degraded (Nemeth et al., 2004). The effect of ferroportin inhibition by hepcidin is therefore to block uptake of dietary iron from the intestine, and to increase the accumulation of iron in macrophages. The result is a decrease in serum iron levels, which routes iron away from pathogens that could potentially exploit circulating iron, but may also render the host anemic by restricting iron availability to the erythron.

Evidence from both animal models and from human genetic lesions suggests that the hepcidin–ferroportin interaction is functionally non-redundant. Mice underexpressing hepcidin are severely iron-overloaded (Nicolas et al., 2001), mimicking the human genetic disorder hereditary hemochromatosis, which (in rare cases) is caused by genetic lesions in the hepcidin gene itself (Roetto et al., 2003), in genes that encode hepcidin-regulatory factors (Bridle et al., 2003), or by mutations in ferroportin that render it resistant to hepcidin control (Drakesmith et al., 2005). Conversely, mice that overexpress hepcidin are fatally anemic (Nicolas et al., 2002a), while patients with mutations that lead to chronic hepcidin overexpression suffer from iron-refractory iron-deficiency anemia (Finberg et al., 2008).

Hepcidin increases in response to high iron conditions (Pigeon et al., 2001), but it is also upregulated in response to infectious and inflammatory stimuli (Nicolas et al., 2002b). Hepcidin is controlled homeostatically primarily via the bone morphogenetic protein (BMP) pathway. BMPs signal through the phosphorylation of SMAD transcription factors, which bind to a well-studied site on the hepcidin promoter and increase hepcidin transcription (Wang et al., 2005). Inflammation and infection also induce interleukin (IL)-6 or IL-22, which can upregulate hepcidin through the phosphorylation of STAT3 (Nemeth et al., 2003; Armitage et al., 2011). BMP/SMAD signaling has also recently been linked to hepcidin upregulation in infectious or inflammatory conditions by the molecule activin B, another member of the transforming growth factor β (TGFβ) superfamily that is upregulated by inflammatory stimuli but acts to upregulate hepcidin through SMAD signaling (Besson-Fournier et al., 2012). Hepcidin is suppressed by anemia, hypoxia, or erythropoietic drive, but, at the time of writing, the pathways underlying hepcidin suppression are less well articulated than are those involved in its upregulation.

## **UPREGULATION OF HEPCIDIN IN MALARIA: CAUSES AND CONSEQUENCES**

Hepcidin is upregulated in infections by bacterial, fungal, and viral pathogens (Armitage et al., 2011). Multiple studies have found that hepcidin is upregulated in malaria infection in symptomatic and asymptomatic natural human infections (Howard et al., 2007; de Mast et al., 2009a, 2010), in experimentally controlled human infections (de Mast et al., 2009b), and in murine models of malaria infection (Portugal et al., 2011; Wang et al., 2011). Resolution of infection leads to normalization of hepcidin levels (de Mast et al., 2009a).

Hepcidin's upregulation in malaria has several important consequences. First, the upregulation of hepcidin leads to iron accumulation in macrophages and a decrease in serum iron, possibly contributing to the dyserythropoiesis and anemia that can accompany malaria infections. Additionally, hepcidin upregulation directly blocks dietary iron absorption: children with post-malarial anemia have high hepcidin levels and poorly incorporate orally administered iron into their red blood cells (Prentice et al., 2012).

The mechanisms whereby hepcidin is upregulated in malaria infection have yet to be fully characterized. IL-6 has been shown to be correlated with hepcidin in some studies (Casals-Pascual et al., 2012; Burte et al., 2013; Jonker et al., 2013), but in another study, urinary IL-6 and hepcidin were not significantly associated after a multiple, stepwise linear regression in infected humans (de Mast et al., 2009a). In one *ex vivo* study, human peripheral blood mononuclear cells co-incubated with *Plasmodium-*infected red blood cells showed a significant upregulation of hepcidin mRNA without concomitant IL-6 message increase (Armitage et al., 2009), but the relative contribution of peripheral blood mononuclear cells to systemic hepcidin levels is unknown. Finally, a study looking at the mechanisms by

which blood-stage malaria infection can prevent the establishment of a liver-stage infection, which is thought to be modulated by hepcidin, found that liver-stage inhibition was preserved in mice treated with anti-IL-6 antibodies (Portugal et al., 2011). In brief, the role of IL-6 in hepcidin regulation in malaria remains controversial, and it is, as yet, unclear which other pathways may contribute to hepcidin upregulation in malaria infection.

Despite the paucity of knowledge on precisely how hepcidin is upregulated in malaria, the fact that both infection and iron repletion lead to increased hepcidin, and that increased hepcidin prevents oral iron absorption, suggests its potential utility as a point of care test to guide iron supplementation in areas of high infectious burden. Low hepcidin would indicate both the absence of infection and the probability of efficient iron absorption, while high hepcidin would indicate either adequate iron stores or ongoing infection. In the latter situation, iron supplementation would likely be unnecessary due to adequate iron stores, poorly absorbed as a direct result of high hepcidin, and/or potentially harmful if hepcidin is elevated due to an ongoing infection.

## **UPREGULATION OF HEPCIDIN IN MALARIA: FOCUSING ON CO-INFECTION**

Hepcidin has been recently shown to play a crucial role in determining the multiplicity of malaria infections within a single host. The obligate liver stage of the malaria parasite requires iron: hepcidin peptide injection or hepcidin overexpression by transgene or viral vector can reduce parasite survival at the crucial hepatic bottleneck (Portugal et al., 2011). The hepcidin upregulation initiated by one blood-stage infection thereby blocks the establishment of a second infection (Portugal et al., 2011).

The physiological redistribution of iron as a consequence of hepcidin upregulation may also have a significant effect on host susceptibility to other bacterial, viral, or protozoa parasites (see **Figure 1**). In a blood-stage malaria infection, raised hepcidin is expected to contribute to increased macrophage iron levels, as does increased erythrophagocytosis. This increase in bioavailable macrophage iron may benefit pathogens that exploit the macrophage niche (van Santen et al., 2013). In particular, hepcidin upregulation may help to explain the association between malaria infections and susceptibility to non-typhoid salmonella (NTS). The epidemiological link between malaria and NTS is well established (Mabey et al., 1987). Iron has been implicated in the contribution of malaria to NTS susceptibility through increases in both free heme and heme-oxygenase expression (Cunnington et al., 2012). By routing iron to accumulate in macrophages, the hepcidin response to malaria may also render the host more vulnerable to NTS directly (van Santen et al., 2013). Similarly, tuberculosis could conceivably benefit from the increased iron availability in its macrophage niche, but the specific role of iron and hepcidin in this important co-infection has not been examined at the time of writing.

An important question is whether malaria-induced alteration of iron metabolism may modulate susceptibility to viral infections and vice versa. Malaria infections are associated with

**metabolism. (A)** Possible implications of blood-stage malaria infection on host susceptibility to other infections. Blood-stage parasitemia causes hepcidin upregulation, which in turn routes iron away from hepatocytes and lowers serum iron levels, blocking the erythron's access to iron and causing anemia. Lowered hepatocyte iron levels prevent the establishment of a second malaria infection (superinfection) by blocking liver-stage growth. Hepcidin also causes iron levels to increase in macrophages and potentially in enterocytes as well, thus possibly giving an advantage to pathogens that

Toxoplasma) and those that require iron in enterocytes (Toxoplasma). **(B)** Hypothesized effect of HCV infection on malaria susceptibility. Long-term HCV infection causes hepcidin suppression, thus increasing liver iron stores and plausibly increasing malaria susceptibility. **(C)** Potential impact of intestinal helminth infection. Helminths cause periodic blood loss, which restricts iron to the liver, potentially blocking malaria liver-stage infection as in **(A)**. Decreases in serum restrict iron availability for erythropoiesis, causing anemia, which in turn causes the downregulation of hepcidin.

higher HIV viral load (Kublin et al., 2005). In turn, HIV-infected individuals are at greater risk of malarial infection (Patnaik et al., 2005). HIV affects iron metabolism in a complex manner: HIV is commonly associated with anemia, and in one study, hepcidin levels were shown to be elevated in HIV infection and inversely correlated with CD4 counts (Wisaksana et al., 2013). Conversely, increased iron stores in HIV-infected individuals are associated with mortality (McDermid et al., 2007, 2009) and risk of tuberculosis infection (McDermid et al., 2013). In an HIV-infected host, how do perturbations in iron and hepcidin modulate host susceptibility to malaria? Interestingly, HIV antiviral drugs have been found to demonstrate anti-*Plasmodium* activity at asexual (Skinner-Adams et al., 2004), liver (Hobbs et al., 2009), and gametocyte (Hobbs et al., 2013) life stages. However the effect of combination therapy on *Plasmodium* in HIV-1 positive patients, and the interactions of anti-retrovirals with antimalarials and their overall effect on iron metabolism are difficult to predict.

Chronic hepatitis C virus (HCV) infection leads to hepcidin suppression and concomitant hepatic iron overload (Girelli et al., 2009). Might this increased iron lead to increased malaria susceptibility? Exposure of mice to HCV RNA induces a type 1 interferon response that has the effect of reducing parasite liver infection 48 h later (Liehl et al., 2014), but a long-term productive HCV infection and consequent iron overload may affect malaria risk very differently. More studies are needed to establish how chronic viral infections alter iron metabolism and malaria susceptibility.

Another unresolved issue is the effects of malaria on toxoplasmosis susceptibility. *Toxoplasma gondii* is carried asymptomatically by a large proportion of the human population, and it can cause life-threatening infections in pregnant women and the immunosuppressed. *Toxoplasma* infects hosts via enterocytes and can live in host macrophages (Hunter and Sibley, 2012). One study has shown that interferon gamma causes the death of *Toxoplasma* in enterocytes in a manner dependent on the depletion of intracellular iron stores (Dimier and Bout, 1998). Upregulation of hepcidin by a malaria infection would likely route iron toward enterocytes and macrophages, thus plausibly increasing host susceptibility and/or tolerance of *Toxoplasma*. If *Toxoplasma* does require intracellular iron stores to withstand the host response, then hepcidin upregulation and subsequent routing of iron to macrophages and enterocytes might increase host susceptibility to *Toxoplasma* infection.

A major cause of iron deficiency and anemia in the developing world is intestinal helminth infection. Infection with helminths is strongly associated with anemia in children (Calis et al., 2008; Jonker et al., 2012a); and de-worming treatment at the population level decreases anemia levels (reviewed in Smith and Brooker, 2010). Complex data exist on the contributions of intestinal helminth infections to malaria susceptibility and severity (Salgame et al., 2013). Early studies claimed that high-population helminth infection was associated with a striking lack of malaria, which the authors suggested was due to the nutritional perturbations associated with the worms (Murray et al., 1977). This finding was corroborated by more recent evidence suggesting that helminth infection in young children may be associated with fewer and later

malaria episodes (Lyke et al., 2005), or protection from some of the potentially severe manifestations of malaria infections (Nacher et al., 2001, 2002). These effects remain controversial; further work has shown associations between helminths and higher parasitemia (Degarege et al., 2012) or severe malaria (Le Hesran et al., 2004). A single study found that infection with filarial worms ameliorated the drop in hemoglobin and reduced inflammatory cytokine production during malaria infection (Dolo et al., 2012). The impact of helminth infections on hepcidin is not well established, but it is plausible that helminth infection may, through periodic intestinal bleeding, cause constitutionally low hepcidin, as the body attempts to recoup the lost iron. At the same time, by causing iron deficiency, helminth infection may decrease host susceptibility to malaria or accentuate the effects of a hepcidin increase in infection.

Most laboratory studies on iron/infection interactions to date have focused on the effect of a single pathogen. Throughout human history, however, as well as in developing nations where infectious disease is still highly prevalent, susceptible individuals are frequently prey to multiple pathogens simultaneously. Understanding how pathogens affect iron metabolism and thereby modulate host susceptibility to other infections is likely to prove important both for its implications on treatment and on public health recommendations.

## **SUPPRESSION OF HEPCIDIN IN SEVERE MALARIA SYNDROMES**

Although the majority of studies on hepcidin and malaria have demonstrated an upregulation of hepcidin in malaria infection, three recent studies have shown that in certain circumstances, hepcidin suppression may also occur. One study found that among all children presenting with malaria, those with severe anemia had the lowest hepcidin levels (Casals-Pascual et al., 2012). A further study (Burte et al., 2013) demonstrated that children with uncomplicated malaria had higher hepcidin levels than those who could be classified as either presenting with severe anemia (in all studies cited, severe anemia was defined as Hb ≤5 g/L) or cerebral malaria. Finally, a group of children with severe malarial anemia exhibited very low hepcidin serum levels (over 50% were undetectable by the study's method; Jonker et al., 2013).

Taken together, these studies clearly indicate that in severe malarial anemia, the signaling pathway that suppresses hepcidin can override the activation pathways associated with parasitemia. The mechanisms of hepcidin suppression by erythropoietic drive, hypoxia, or iron deficiency have not yet been well defined, but some groups have posited the existence of a bone marrow-secreted factor that suppresses hepcidin during erythropoiesis (Ginzburg and Rivella, 2011). In the two human studies that compared erythropoietin levels with serum hepcidin in malaria infection, erythropoietin and hepcidin were negatively associated (Casals-Pascual et al., 2012; Jonker et al., 2013). One animal study has also demonstrated a significant negative correlation between serum erythropoietin and hepcidin liver message in mice infected with *P. berghei* (Wang et al., 2011).

Further studies utilizing animal models of severe malarial anemia will likely be required to explore this aspect of iron control in malaria infection. In addition, the role of iron and hepcidin in cerebral malaria requires investigation.

## **MALARIA PREVENTION EFFORTS REDUCE ANEMIA PREVALENCE**

A recently published meta-analysis of global anemia prevalence estimated that in 2010, the prevalence of anemia was 32.9%, accounting for 8.8% of total years of life lived with disability worldwide (Kassebaum et al., 2014). In malaria-endemic countries, malaria is a major contributor to anemia at the population level: the authors estimated that in sub-Saharan African, 24.7% of anemia is attributable to malaria (Kassebaum et al., 2014). Therefore, interventions that reduce the prevalence of malaria could be expected to result in a reduction in the severity and prevalence of anemia.

The global community is currently using multiple complementary tools to reduce malaria at the population level. Intermittent preventative treatment (IPT) refers to the practice of administering antimalarial drugs presumptively during the especially vulnerable periods of pregnancy or early childhood. Insecticidetreated bed nets (ITNs) reduce nocturnal bites from *Anopheles* mosquitoes (Hill et al., 2006). Both interventions serve to protect the recipient and to reduce transmission at a population level (Greenwood, 2004).

A recent Cochrane review examining the effects of IPT for children under 5 concluded that of five West African trials examined, there was likely a significant reduction overall in moderate anemia incidence in areas with seasonal *P. falciparum* malaria transmission (Meremikwu et al., 2012). Of the two trials, which considered severe anemia as a separate outcome, both also found a significant reduction as a result of IPT (Dicko et al., 2011; Konate et al., 2011). This is consistent with the findings of a previous Cochrane review, which had also concluded that IPT in pregnant women was similarly associated with decreased rates of both severe anemia and all anemia (Garner and Gulmezoglu, 2006).

Fewer studies have examined the effects of the introduction of ITNs only on anemia prevalence, but earlier introduction of ITNs is associated with a statistically significant reduction in anemia prevalence in children at 6–12 months of age (Muller et al., 2006). In another study, hemoglobin levels were found to be significantly associated with the use of ITNs in young children (Holtz et al., 2002).

Why does reducing malaria infection have such a profound effect on population anemia levels? The relationship between malaria control and anemia risk may be partially dependent on hepcidin and its effects on iron absorption and utilization (as well as by reducing the well-known inhibitory effects of malaria on erythropoiesis). The hepcidin increases associated with malaria infection prevent efficient iron uptake. Iron supplementation has been shown to be less effective in areas with high malaria transmission (Gera et al., 2007).

Two studies have delineated the effects of malaria infection and hepcidin increase on the incorporation of orally administered iron into erythrocytes (Doherty et al., 2008; Cercamondi et al., 2010). In a study of young Beninese women, asymptomatic parasitemia was associated with poor incorporation of orally administered iron into RBCs, an effect that did not extend to

parentally administered iron. Treatment and resolution of parasitemia resulted in a decrease in serum hepcidin and improved absorption of oral iron supplements (Cercamondi et al., 2010).

Similarly, young children with post-malarial anemia were shown to poorly absorb orally administered iron supplements (Doherty et al., 2008). Hepcidin levels in sera of these children were measured subsequently; these hepcidin levels were shown to be the best predictors of iron absorption (Prentice et al., 2012). Furthermore, despite poor oral absorption of iron, these children showed a more rapid hematological recovery than anemic and aparasitemic children, suggesting that hepcidin upregulation in malaria may contribute to relocalization of iron to macrophages, rather than true iron deficiency.

Populations with a high prevalence of malaria infections may therefore suffer both from anemia as a direct consequence of malaria infections, but also from poor utilization of oral iron as a result of chronically upregulated hepcidin.

## **DISCUSSION AND RECOMMENDATIONS**

Anemia continues to be one of the most common causes of disability worldwide; while *Plasmodium* is one of the most prevalent human pathogens. The relationship between the two is complex: low iron status may protect against malaria infection, but malaria infection in turn is linked with anemia at both the individual and population levels. New clues to this relationship have been obtained through our improved understanding of iron metabolism.

Specifically, the relatively recent identification of the human iron hormone hepcidin has allowed us to begin to explore these relationships in more depth. Hepcidin is upregulated in malaria infection, likely contributing to anemia through the relocalization of iron to macrophages and the prevention of iron uptake from the diet. However, in certain severe malaria syndromes, hepcidin may be suppressed. The mechanisms behind both hepcidin's increase and its suppression in malaria are currently unclear.

At the moment, action on three fronts is required. First, we must implement our improved understanding of the iron– malaria relationship toward optimizing malaria prevention and anemia treatment. Research findings support more nuanced iron supplementation regimens, rather than the blanket approach of supplementing, or not supplementing, whole populations. Supplementation could be scheduled around the malaria season (Atkinson et al., 2014), routed to the most iron-deprived children or pregnant women, or targeted based on hepcidin levels as a proposed biomarker (Pasricha et al., 2014). At the very least, the findings of the last decade of research since the Pemba trial argues very strongly that iron supplementation should be carried out in concert with malaria control efforts; and that improved malaria control may itself be effective as a means of decreasing anemia at the population level.

Second, clinically focused research is required to develop better and rationally informed therapeutics for those who become infected. For example, understanding the mechanisms underlying hepcidin upregulation in malaria could be the first step toward the development of drugs that, given concurrently with antimalarial treatment, repress hepcidin and speed recovery from anemia. Carefully designed *in vivo* and *in vitro* studies could explore the role that hepcidin may play in mediating the outcomes of co-infections between malaria and other protozoan, bacterial, or viral pathogens, providing vital information for treating individuals and populations exposed to multiple pathogens simultaneously.

Finally, in the longer term, basic research should focus on an improved understanding of the host–pathogen tug-of-war over iron metabolism. How does malaria acquire iron at a molecular level in its many life stages? Do malaria pathogens show different phenotypes in iron-deprived hosts? Armed with new tools to explore pathogen biology and iron metabolism, we have a chance to answer questions of immediate clinical importance, and to advance the basic science behind our understanding of the host–parasite relationship.

## **ACKNOWLEDGMENTS**

Natasha Spottiswoode and Patrick E. Duffy are supported by the Intramural Research Program of NIAID, NIH. Natasha Spottiswoode was also supported through the NIH Oxford – Cambridge Scholars Program during the period of writing. Hal Drakesmith is supported by the Medical Research Council, UK and by the Bill and Melinda Gates Foundation.

## **REFERENCES**


different types of anemia in african children. *Sci. Transl. Med.* 6, 235re233. doi: 10.1126/scitranslmed.3008249


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

*Received: 12 March 2014; paper pending published: 29 April 2014; accepted: 11 May 2014; published online: 30 May 2014.*

*Citation: Spottiswoode N, Duffy PE and Drakesmith H (2014) Iron, anemia and hepcidin in malaria. Front. Pharmacol. 5:125. doi: 10.3389/fphar.2014.00125*

*This article was submitted to Drug Metabolism and Transport, a section of the journal Frontiers in Pharmacology.*

*Copyright © 2014 Spottiswoode, Duffy and Drakesmith. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited andthatthe original publication inthis journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*

## Influence of host iron status on *Plasmodium falciparum* infection

#### *Martha A. Clark1, Morgan M. Goheen1 and Carla Cerami <sup>2</sup> \**

*<sup>1</sup> Microbiology and Immunology, University of North Carolina, Chapel Hill, NC, USA*

*<sup>2</sup> Epidemiology, Gillings School of Global Public Health, University of North Carolina at Chapel Hill, Chapel Hill, NC, USA*

#### *Edited by:*

*Raffaella Gozzelino, Instituto Gulbenkian de Ciência, Portugal*

#### *Reviewed by:*

*Leann Tilley, Melbourne University, Australia Akira Kaneko, Karolinska Institutet, Sweden*

#### *\*Correspondence:*

*Carla Cerami, Epidemiology, Gillings School of Global Public Health, University of North Carolina at Chapel Hill, CB# 7435, Chapel Hill, NC 27599, USA e-mail: ccerami@unc.edu*

Iron deficiency affects one quarter of the world's population and causes significant morbidity, including detrimental effects on immune function and cognitive development. Accordingly, the World Health Organization (WHO) recommends routine iron supplementation in children and adults in areas with a high prevalence of iron deficiency. However, a large body of clinical and epidemiological evidence has accumulated which clearly demonstrates that host iron deficiency is protective against falciparum malaria and that host iron supplementation may increase the risk of malaria. Although many effective antimalarial treatments and preventive measures are available, malaria remains a significant public health problem, in part because the mechanisms of malaria pathogenesis remain obscured by the complexity of the relationships that exist between parasite virulence factors, host susceptibility traits, and the immune responses that modulate disease. Here we review (i) the clinical and epidemiological data that describes the relationship between host iron status and malaria infection and (ii) the current understanding of the biological basis for these clinical and epidemiological observations.

**Keywords: malaria, iron, iron deficiency anemia,** *Plasmodium falciparum***, iron supplementation**

Iron deficiency and malaria are significant co-morbidities in large portions of the developing world, and both maladies disproportionately affect pregnant women and children. Malaria causes an estimated 250 million infections and 500,000 deaths annually. Iron deficiency is estimated to affect one quarter of the world's populations causing substantial morbidity. Fortunately, iron deficiency is easily treated with iron supplementation (Okebe et al., 2011). Accordingly the World Health Organization (WHO) recommends routine iron supplementation for children and adults in areas with high prevalence of iron deficiency (Haider et al., 2013; Low et al., 2013). However, the wisdom of universal iron supplementation campaigns in malaria endemic regions has recently been questioned due to clinical evidence that suggests iron deficiency protects against malaria, and that iron supplementation of women and children may increase the incidence of malaria when given without malaria prophylaxis or access to adequate health care (Nyakeriga et al., 2004; Sazawal et al., 2006; Tielsch et al., 2006; Kabyemela et al., 2008; Senga et al., 2011; Veenemans et al., 2011; Gwamaka et al., 2012; Jonker et al., 2012; Esan et al., 2013; Zlotkin et al., 2013). This situation has created a dilemma for health policy makers and health care workers in malaria endemic regions of the world (Prentice et al., 2013).

Despite these clinical and epidemiological studies, the extent to which the human host's iron status affects risk to and severity of malaria infection is unknown. Differences in study design and confounding factors (such as acquired immunity to malaria and hemoglobinopathies) have made the clinical and epidemiological studies difficult to interpret (Prentice et al., 2007). Furthermore, though iron and malaria have been and continue to be studied the exact biological relationship between host iron and malaria virulence remains largely unclear.

## **IRON DEFICIENCY AND IRON DEFICIENCY ANEMIA**

Iron deficiency is a condition in which there is insufficient iron in the body to maintain normal physiologic functions. Iron deficiency can be categorized into three stages: iron deficiency without anemia, iron deficiency with mild anemia, and iron deficiency with severe anemia. Iron deficiency anemia occurs when iron stores are exhausted and the supply of iron to tissue is compromised; this condition is defined as anemia with biochemical evidence of iron deficiency. Iron deficiency is most prevalent and severe in young children and women of reproductive age, but can also occur in older children, adolescents, adult men, and the elderly. It is estimated that 50% of pregnant women and 40% of preschool children in the developing world are iron deficient (WHO | Assessing the iron status of populations, 2007; Kassebaum et al., 2014). Often, iron deficiency develops slowly and is not clinically diagnosed until severe anemia is apparent (Stoltzfus, 2003).

Studies suggest that iron deficiency impairs the growth, cognition, and neurological development of children from infancy through adolescence, impairs immune function, and is associated with increased morbidity rates (De-Regil et al., 2011, 2013; Wang et al., 2013). Iron deficiency during pregnancy is associated with multiple adverse outcomes for both mother and infant, including increased risk of hemorrhage, sepsis, maternal mortality, perinatal mortality, and low birth weight (Peña-Rosas et al., 2012a,b). Iron deficiency anemia can be a direct cause of death or contribute indirectly. For example, during child birth an anemic mother cannot afford to lose more than 150 mL of blood, compared with a healthy mother who can lose up to 1 liter of blood and still survive. Thus, the WHO recommends iron supplementation for all men, women, and children in areas where malnutrition is prevalent (WHO | Guidelines on food fortification with micronutrients, 2006).

Host iron metabolism is intimately linked to the host response to infection and inflammation. In the face of infection and inflammation, the human host protein hepcidin becomes elevated and initiates signaling which results in reduced iron absorption into the body along with the redistribution of body iron stores. As a consequence many of the biomarkers utilized to assess host iron status are sensitive to both iron as well as infection. For example, low serum ferritin (serum ferritin reflects total body iron reservoirs) is indicative of iron deficiency. However, ferritin is also an acute phase protein which is elevated in the context of infection, and as a result is not a reliable marker of human iron status in the presence of infection or inflammation. Like serum ferritin, transferrin saturation and transferrin receptor levels are biochemical markers of human iron status that are also sensitive to infection and inflammation. As a result evaluating an individual's iron status during an infection has proven difficult (Aguilar et al., 2012), and the scientific community has struggled to establish formal guidelines.

## **MALARIA**

In 2012 malaria caused an estimated 207 million infections and over 600,000 deaths; 90% of these deaths occurred in sub-Saharan Africa, and 77% occurred in children under five (WHO | World Malaria Report, 2013). At least five species of the eukaryotic Apicomplexan parasite from the genus *Plasmodium* cause malaria in humans with *Plasmodium falciparum* being the most common and deadly. Following the bite of a malaria parasite infected mosquito, the sporozoite stage of the parasite enters the bloodstream and travels to the liver, where it subsequently infects liver hepatocytes. Malaria replication in the liver is asymptomatic. Next, the merozoite form of the parasite leaves the liver and enters into circulation to infect host red blood cells (RBCs). During the erythrocytic stage of infection, the parasite repeatedly invades, replicates within, and egresses from host RBCs. This erythrocytic stage of infection is responsible for all symptoms of disease (Miller et al., 2013), and the severity of disease is directly associated with parasite burden (Chotivanich et al., 2000; Dondorp et al., 2005).

A wide range of symptoms can be observed in malaria patients. Clinically however, malaria is categorized as either uncomplicated or complicated. Complicated malaria is further divided into three overlapping syndromes: cerebral malaria, severe anemia, and metabolic acidosis. The clinical syndrome observed in each individual patient is influenced by multiple variables: parasite species, host immune status, and genetic background, as well as the use and timing of antimalarial drugs (Taylor et al., 2010).

## **CLINICAL STUDIES LINKING IRON AND MALARIA INFECTION**

Host iron has received significant attention at the clinical level as a major factor that may regulate malaria virulence. The results of clinical studies conducted prior to 2002 which examined the relationship between host iron status and malaria risk are reviewed in three meta-analyses (Shankar, 2000; Oppenheimer, 2001; Gera and Sachdev, 2002). In the interim, two large iron supplementation trials as well as several smaller clinical studies have shed further light on the relationship between host iron status and malaria infection (**Table 1**). Clinical trials that have examined the relationship between host iron and malaria fall into two basic categories: those that observe the rate of malaria in individuals with iron deficiency anemia, and those that look at the rate of malaria infection in individuals given iron supplementation. Differences in study design exist within both study types, and include: the definition of study participant iron status, the administration of iron alone or with folate, and access to health care. Despite these differences, assessment of the outcome of the clinical studies has led to the general consensus that iron deficiency is protective against malaria, and iron supplementation increases malaria risk in the absence of access to adequate health care (Prentice and Cox, 2012; Spottiswoode et al., 2012; Stoltzfus, 2012).

While these clinical studies and meta-analyses have been indispensable for determining the relationship between host iron status and malaria risk, it is not clear how iron deficiency protects and why iron supplementation increases risk. Immunity to malaria and high prevalence of genetic traits such as G6PD deficiency and hemoglobinopathies in the study populations limit the capacity of clinical studies to parse out causation. Furthermore, relatively little is known with regards to the role host iron plays in malaria pathogenesis. Iron impacts a broad range of biological processes that have the potential to shape malaria pathogenesis. As a result, even with the most ideal of clinical study designs; the prerequisite knowledge of which aspects of malaria pathogenesis should be studied is largely absent. A better grasp on the underlying biological principals that govern (i) the protection of iron deficiency against malaria and (ii) the increased risk of malaria associated with iron supplementation is critical for managing iron supplementation campaigns in malaria endemic regions.

## **BIOLOGICAL IMPORTANCE OF IRON**

Iron is an essential nutrient for nearly every living organism including humans and the malaria parasite. Iron impacts a broad range of biological processes; including host and parasite cellular function, erythropoiesis and immune function. The capacity of iron to fluctuate between two oxidation states, ferrous (Fe2+) and ferric (Fe3+), makes it indispensable for many critical biological processes, including DNA replication, cellular respiration, and oxygen transport. However, the same useful biphasic properties of iron which make it indispensable also contribute to its high cytotoxicity. As a result the human host tightly regulates iron availability and usage.

Access to iron is particularly important in the context of host-pathogen interactions. When confronted with infection and inflammation the human host reallocates its iron reservoirs in an effort to deprive invading pathogens of iron. The human protein hepcidin—a rheostat of systemic iron homeostasis—signals the body to decrease absorption of iron in the proximal duodenum and orchestrates the movement of iron from serum into



*(Continued)*

#### **Table 1 | Continued**


storage within the liver and macrophages (Roy, 2013). As a result of reduced serum iron, erythropoiesis—a process exquisitely sensitive to iron levels—slows in the face of infection as well as inflammation. The human host's active reduction in bioavailable iron protects against a wide range of pathogens (Armitage et al., 2011). Not surprisingly, as many pathogens require access to host iron sources to survive and grow, pathogens have evolved sophisticated iron acquisition systems, and the iron acquisition systems of many bacterial and fungal species have been well described (Skaar, 2010). By comparison how the malaria parasite acquires, regulates, and utilizes iron remains a mystery.

## **IRON METABOLISM IN THE MALARIA PARASITE**

Iron is essential for the survival of the malaria parasite. The parasite multiplies 8–32 times in the course of a single intraerythrocytic lifecycle. Iron is an essential cofactor for the DNA replication enzyme ribonucleotide reductase, and as a result iron is required to fuel this rapid intra-erythrocytic proliferation (Rubin et al., 1993). Iron is also utilized by the parasite for pyrimidine (Krungkrai et al., 1990; Van Dooren et al., 2006) and heme biosynthesis (Sato and Wilson, 2002; Dhanasekaran et al., 2004; Sato et al., 2004; Nagaraj et al., 2008, 2009, 2010, 2013). As with the human host, the malaria parasite must also balance its need for iron against the cytotoxicity of iron.

The malaria parasite metabolizes host hemoglobin in its acidic digestive vacuole in order to acquire necessary amino acids; however, as discussed below, the parasite does not utilize the iron in host heme. Plasmodium aspartic and cysteine proteases degrade host hemoglobin and release large quantities of toxic iron-laden heme (Goldberg et al., 1990; Subramanian et al., 2009). Apicoblast parasites neutralize the cytotoxic heme produced during hemoglobin metabolism by sequestering the heme in an inert crystal, hemozoin (Rudzinska et al., 1965; Chugh et al., 2013). Despite neutralizing a substantial portion of host heme into hemozoin, some residual heme remains free and becomes oxidized, generating free oxygen radicals (Francis et al., 1997). The parasite possesses powerful thioredoxin and glutathione systems to maintain intracellular redox equilibrium (Jortzik and Becker, 2012). However, even when these redox systems are functioning at full capacity, oxidative stress significantly increases as the parasite matures and replicates within host erythrocytes (Fu et al., 2010). In fact, many antimalarials, including artemisinin, appear to target the parasite's ability to detoxify reactive oxygen species (ROS) (Rosenthal and Meshnick, 1996; Klonis et al., 2013; Ariey et al., 2014). For example, it was recently found that mutations in *PF3D7\_1343700* (Kelch) can confer resistance to artemisinin. The authors speculate that these mutations cause a disruption of the parasite's ability to detoxify ROS because the efficacy of artemisinin depends on its ability to generate oxygen radicals and some kelch-containing proteins in other organisms have been shown to be involved in the regulation of cytoprotection (Ariey et al., 2014).

Given the relationship between iron, heme, and ROS, it is possible that perturbations in host iron regulation might also affect the malaria parasite's redox equilibrium. Iron responsive proteins (IRPs) and their accompanying iron responsive elements are critical for maintaining cellular iron homeostasis in the human host. IRPs and iron responsive elements are responsible for mobilizing iron when demands are high and moving iron into storage when excess iron may promote ROS formation (Hentze et al., 2010). Loyevsky et al. identified and characterized a *P. falciparum* IRP, the expression of which was affected by iron starvation as well as iron supplementation (Loyevsky et al., 2001, 2003; Hodges et al., 2005). However, a search of gene databases failed to identify *Plasmodium* homologs of ferritin, ferroportin, metallothione, a ferrioxamine-based transport system or ferredoxin or siderophore biosynthesis pathways—all proteins and processes utilized by other organisms to acquire, regulate, and store iron (Scholl et al., 2005). Clearly, much remains unknown regarding parasite iron biology.

## **IRON CHELATORS AND THEIR CONTRIBUTION TO THE ELUCIDATION OF MALARIA IRON BIOLOGY**

Realizing the importance of iron for the malaria parasite, researchers have invested extensive time and effort into the investigation of the antimalarial activity of iron chelating agents. These studies have also provided insight into malaria parasite iron biology. In contrast to mammalian cells, which are sensitive to millimolar concentrations of iron chelators, erythrocytic stage malaria parasites are sensitive to micromolar concentrations of iron chelators *in vitro* and in animal models (Cabantchik et al., 1996). The cytotoxicity of iron chelators is dependent upon the stage of intra-erythrocytic maturation of the malaria parasite and the hydrophobicity of the iron chelator (Lytton et al., 1994). For example, the hydrophilic chelator hydroxamate-based deferoxamine (DFO) has cytostatic activity against the ring stage and cytotoxic activity against the late trophozoite and schizont erythrocytic stages of the parasite (Whitehead and Peto, 1990; Lytton et al., 1994; Cabantchik et al., 1999).

The cytotoxicity of iron chelators against the malaria parasite suggests that the mechanism of action of iron chelators is more complex than simple iron deprivation. Alternative mechanisms have been suggested for some chelators, including the direct inhibition of parasite ribonucleotide reductase activity (Lederman et al., 1984; Lytton et al., 1994). Furthermore, as iron chelators can modulate host immune function, iron chelator antimalarial activity may be a result of modification of the host immune response (Golenser et al., 2006; Li et al., 2012).

Caution must be taken when considering the use of iron chelators to inform our understanding of the biological relationship between iron deficiency and malaria infection. The evidence that iron chelators do more than merely deprive the parasite of iron introduces potential confounding factors into studies that utilize iron chelators as a model for iron deficiency. Furthermore, most iron chelators cannot chelate iron associated with heme, ferritin, or transferrin. Because the iron saturation of each of these host iron reservoirs are reduced in iron deficiency, iron chelators are not suitable for studying the effect of host iron reduction on the malaria parasite.

That said, evidence that chelation of chelatable extracellular and intra-erythrocyte iron does not impact erythrocytic stage *P. falciparum* growth, suggests that chelatable host iron is not necessary for the erythrocyte stage of infection (Scott et al., 1990). Furthermore, work by Moormann et al. shows that parasite nuclear and mitochondrial transcripts decrease in the presence of the iron chelator DFO (Moormann et al., 1999). These results are consistent with a normal cellular response to iron deprivation. In conclusion, iron chelators are obviously indispensable in the study of iron biology. However, in the case of malaria caution must be taken.

## **HOST IRON RESERVOIRS AVAILABLE TO ERYTHROCYTIC STAGE MALARIA**

It is inarguable that iron is essential to erythrocytic stage malaria and therefore possible that alterations in host iron levels may tip the balance between inhibiting or promoting parasite growth and virulence. Consequently, the question of how the parasite acquires host iron becomes central. A healthy iron-replete human has 3–4 total grams of iron, which is distributed in hemoglobin contained within circulating RBCs (2.5 g), in iron containing proteins (400 mg), in serum bound to transferrin (3–7 mg), and in storage proteins such as ferritin (1 g). Host iron reservoirs available to erythrocytic stage malaria parasite include: (1) transferrin and non-transferrin bound iron (NTBI) in the serum and (2) intra-erythrocytic iron contained within hemoglobin, ferritin, as well as trace amounts freely bioavailable iron in the RBC cytosol (**Figure 1**).

Iron deficiency affects these host iron reservoirs by significantly reducing the availability of both serum iron and intra-erythrocytic iron. Iron supplementation results in brief spikes in serum iron levels (Schümann et al., 2012, 2013), but has little immediate effect on intra-erythrocyte iron. However, approximately 2 weeks following iron supplementation, average intra-erythrocyte iron levels slowly begin improving as new ironreplete RBCs enter into circulation. It is well-documented that virulence of many bacteria is directly associated with the availability of host iron, and as a result iron supplementation can exacerbate infections (Doherty, 2007). Whether described changes in serum and intra-erythrocyte iron stores affect erythrocytic stage malaria infection remains unknown.

**FIGURE 1 | Host Iron available to erythrocytic stage** *P. falciparum***.** Host iron immediately available to the erythrocytic stage of *P. falciparum* include serum and intra-erythrocytic iron. Serum iron ranges from 10 to 27μM. Transferrin bound iron is the predominant form of iron in the serum, though trace amounts of non-transferrin bound iron (NTBI) are present. In some pathologic conditions such as hemochromatosis, NTBI may be

## **THE RELATIONSHIP BETWEEN SERUM IRON AND ERYTHROCYTIC STAGE MALARIA**

The relationship between host serum iron and parasitized RBCs (pRBCs) is especially intriguing (**Table 2**). Because transferrin has an extremely high affinity for iron (1023M−<sup>1</sup> at pH 7.4), NTBI is scarce in healthy individuals. There is strong evidence that transferrin associates with pRBCs but not uninfected RBCs. Work by Pollack et al. shows that pRBCs take up Fe<sup>59</sup> bound to human transferrin, and a recent publication by our own group demonstrates that incubation of pRBCs with transferrin and ferric citrate increases the bioavailable iron in pRBCs (Pollack and Fleming, 1984; Clark et al., 2013). The idea that the parasite is able to acquire transferrin bound iron is further supported Surolia et al. who demonstrated that gelonin toxicity toward *P. falciparum* is 25 times greater when the gelonin is bound to transferrin (Surolia and Misquith, 1996). Moreover, Fry et al. report transferrin reductase activity associated with pRBCs but not uninfected RBCs (Fry, 1989). Additionally, two groups have reported the identification of a *P. falciparum* transferrin receptor in the RBC membrane of pRBCs (Haldar et al., 1986; Rodriguez and Jungery, 1986). However, a later study by Pollack et al. concluded that transferrin binding of pRBCs is nonspecific (Pollack and Schnelle, 1988), and additional studies were unable to detect any acquisition of transferrin bound iron by pRBCs (Peto and Thompson, 1986; Sanchez-Lopez and Haldar, 1992).

Despite strong evidence that transferrin associates with pRBCs, neither iron depletion nor iron supplementation of malaria culture media has any observable effect on parasite growth (Peto and Thompson, 1986; Scott et al., 1990; Sanchez-Lopez and Haldar, 1992; unpublished data Clark et al.). These results challenge the idea that serum iron, specifically transferrin significantly greater. While iron deficiency anemia is characterized by a significant decline in serum iron. RBC iron is found within hemoglobin (20 mM), ferritin (0.7 nM), and as bioavailable iron (1–10μM). Iron deficiency anemia significantly reduces RBC iron, specifically hemoglobin iron. Shown in the figure are: *Pf*, *P. falciparum*; DV, digestive vacuole; N, parasite nucleus; and EC, endothelial cell.

bound iron, contributes to the protection of iron deficiency from malaria and the increased risk of malaria associated with iron supplementation. Yet, it should be noted that malaria culture media contains tenfold less iron than human sera and all existing studies have utilized culture adapted *P. falciparum* laboratory lines. It is possible laboratory lines have adapted to an iron-starved extracellular environment. Furthermore, because hemoglobin is an essential nutrient for erythrocytic stage malaria, it is impossible to "starve" the parasite of iron *in vitro* and this may in turn limit the ability to study the effect of serum iron on *P. falciparum*.

## **THE RELATIONSHIP BETWEEN INTRA-ERYTHROCYTIC IRON AND ERYTHROCYTIC STAGE MALARIA**

Much less is known about the ability of the malaria parasite to access intra-erythrocytic iron (**Table 3**). An individual RBC contains 100 fg (20 mM) of iron, the majority of which is contained within hemoglobin. It is estimated that if the parasite were able to access only 1% of this hemoglobin iron all of its iron demands would be fulfilled (Hershko and Peto, 1988; Gabay and Ginsburg, 1993). However, as discussed above, the parasite incorporates the majority of heme released as a result of hemoglobin digestion into hemozoin (Chugh et al., 2013). Despite identification of a *Plasmodium* heme oxygenase-like protein, which would facilitate release of iron from host heme (Okada, 2009), the parasite does not exhibit enzymatic heme oxygenase activity nor possess a canonical heme oxygenase pathway (Sigala et al., 2012). Even without inherent heme oxygenase activity, it remains possible that non-enzymatic mechanisms release enough iron from trace heme to meet the iron requirements of the parasite. Possible mechanisms include heme breakdown by glutathione or hydrogen peroxide, the conditions for which are predicted to exist within erythrocytic stage parasites (Ginsburg et al., 1998; Loria

#### **Table 2 | Relationship between host serum iron and** *P. falciparum.*


#### **Table 3 | Relationship between RBC iron and** *P. falciparum.*


et al., 1999). However, as the parasite synthesizes heme *de novo*, it does not seem likely that the parasite draws iron from host heme (Nagaraj et al., 2013).

In addition to hemoglobin, RBCs contain residual amounts of biovaialble iron (1–10μM) as well as iron stored within ferritin (0.7 nM), and it is possible that the parasite is capable of utilizing one or both of these erythrocyte iron reservoirs. Currently, however, there is no reported evidence to either support or refute these possibilities (Scholl et al., 2005). However, despite a lack of evidence that the parasite accesses host intra-erythrocytic iron, recent work by our group has shown that pRBC bioavailable iron content increases as the parasite matures from ring stage to schizont. This observation suggests that iron is released from some form of storage as the parasite develops within host RBCs (Clark et al., 2013). Whether the iron is released from parasite or host storage remains an open question.

Although the precise host iron source(s) the malaria parasite acquires remains unclear, all the potential host iron reservoirs (serum and intra-erythrocyte) available to erythrocytic stage malaria are affected by iron deficiency as well as iron supplementation. Therefore, it is reasonable to hypothesize that iron deprivation and excess iron contribute to the relationship between host iron and malaria risk observed in the clinical studies discussed earlier. That said, even during iron deficiency, the erythrocytic stage of the parasite inhabits the most iron rich environment in the human body. As such it is alternatively possible that neither iron deficiency nor iron supplementation perturb iron reservoirs enough to significantly impact the parasite.

## **MICROCYTIC IRON DEFICIENT RBCs AND MALARIA**

In addition to affecting host iron reservoirs, iron deficiency also induces changes in RBC physiology. One such difference between iron-replete and iron-deficient RBCs is the substitution of zinc for iron in hemogloblin when iron is limiting. This results in zinc protoporphoryin IX levels ten times higher in iron deficient as compared to iron-replete RBCs (Wong et al., 1996). As zinc protoporphoryin IX inhibits hemozoin extension *in vitro*; it is reasonable to hypothesize that that elevated zinc protoporphoryin IX in iron deficient erythrocytes provides protection against malaria infection by impeding parasite growth (Iyer et al., 2003).

Additional changes to RBC physiology caused by iron deficiency include: microcytosis, greater susceptibility to oxidative stress, reduced ATP content, and decreased deformability (Yip et al., 1983; Acharya et al., 1991; Nagababu et al., 2008; Brandão et al., 2009). Furthermore, iron deficient RBCs experience enhanced eryptotic cell death (Kempe et al., 2006). The altered physiology of microcytic iron deficient RBCs may therefore protect against erythrocytic stage malaria infection. Research by Koka et al. indicates that propagation of the erythrocytic stage of *P. falciparum* strain BinH is reduced in iron deficient RBCs (Koka et al., 2007). However, earlier work by Luzzie et al. observed abnormal parasite morphology but no difference in the growth of *P. falciparum* strain UPO in iron deficient as compared to ironreplete RBCs (Luzzi et al., 1990). The differences between these studies may be explained by the use of different *P. falciparum* isolates which feasibly could have different sensitivities to iron deficient RBCs.

Accelerated host clearance of iron deficient pRBCs is an additional explanation for the protection afforded by iron deficiency against malaria. Results from two studies that examined malaria infection in iron deficient mice both observed a higher clearance rate of pRBCs in iron deficient as compared to iron-replete mice (Koka et al., 2007; Matsuzaki-Moriya et al., 2011). Specifically, Matsuzaki et al. observed elevated phagocytosis of pRBCs in iron deficient as compared to iron-replete mice, and proposed that the increased phagocytosis rate may be attributable to greater phosphatidylserine levels on iron deficient pRBCs as compared to iron-replete pRBCs. Koka et al. similarly observed greater phosphatidylserine levels on *P. falciparum* human iron deficient pRBCs. Ultimately, these limited data suggests that iron deficiency may provide protection against malaria infection by both impeding erythrocytic stage malaria growth and increasing phagocytosis of iron deficient pRBCs. However, only further investigation will reveal the true relationship between iron deficient RBCs and *P. falciparum*.

## **PERTURBATIONS IN ERYTHROPOIESIS AND MALARIA**

In the absence of sufficient iron for heme synthesis, the human host's erythropoietic rate decreases. Conversely, iron supplementation of individuals with iron deficiency anemia results in a strong erythropoietic response; because the body attempts to recover RBC numbers and replace less viable iron deficient RBCs (**Figure 2**). It is well-known that *P. vivax* exclusively infects the very youngest RBCs (reticulocytes). However, *P. vivax* is not the only *Plasmodium* species that prefers young RBCs. In fact many species of *Plasmodium,* including *P. falciparum,* preferentially infect young RBCs, and furthermore young RBC support greater parasite replication than more mature RBCs (Wilson et al., 1977; Pasvol et al., 1980; Lim et al., 2013). Thus, significant elevation in the erythropoietic rate could put an individual at increased risk of erythrocytic stage *P. falciparum* infection. Tian et al. have investigated this hypothesis in the context of pregnant women, who are at greater risk of malaria infection than their non-pregnant counterparts and experience increased erythropoietic rates to meet the

anemia and iron supplementation each profoundly influence human erythropoiesis, and this may influence erythrocytic stage malaria infection. Iron deficiency induced reduction in the erythropoietic rate and synthesis of iron supplementation and subsequent replacement of microcytic iron deficient RBCs with young iron-replete RBCs may increase an individual's risk of erythrocytic stage *P. falciparum* infection.

oxygen demands of the growing fetus. The authors report that *P. falciparum* growth is significantly greater in the on average younger RBCs taken from pregnant women as compared to the on average older RBCs taken from non-pregnant women (Tian et al., 1998).

Murine models have additionally been used to shed light on the relationship between erythropoiesis and malaria infection. Interestingly, when Chang et al. manipulated the timing of erythropoiesis during the course of a malaria infection it was observed that reticulocytosis early in infection significantly increased infection and morbidity, while reticulocytosis late in infection decreased mortality (Chang et al., 2004). These observations are consistent with recent work by Zhao et al. showing that lipocalin 2, which is elevated during malaria infection, provides protection from malaria infection in mice by limiting reticulocytosis (Zhao et al., 2012).

Furthermore, mathematical modeling by Cromer et al. makes several key predictions that support a role for erythropoiesis in driving the protection from malaria associated with iron deficiency anemia and increased risk associated with iron supplementation. First, their model predicts that low reticulocyte production rate—as would be observed in iron deficiency—in combination with a parasite that prefers reticulocytes, could result in a less severe infection. Second, high reticulocyte production as would be observed in iron deficient individuals responding to iron supplementation—could increase severity of malaria infection (Cromer et al., 2009). These results indicate that limiting reticulocytosis early in infection is important for limiting erythrocytic stage malaria infection and further support the hypothesis that iron supplementation-induced reticulocytosis significantly increases the risk of erythrocytic stage *P. falciparum* infection.

Together, these observations provide insight into potential cellular mechanisms contributing to the protection of iron deficiency against malaria, and the increased risk of malaria associated with iron supplementation. With regard to iron deficiency, altered RBC physiology may limit *P. falciparum* propagation within iron deficient RBCs and increase clearance of iron deficient pRBCs. Furthermore, the reduced erythropoietic rate and subsequent reduction in an iron deficient individual's hematocrit may additionally contribute to protection. Conversely, the increased erythropoietic rate triggered by iron supplementation paired with the preference of *P. falciparum* for young RBCs may be partially responsible for the increased risk of malaria infection that is associated with iron supplementation.

#### **CONCLUSIONS AND FUTURE QUESTIONS**

Overall, the available evidence supports a link between (i) iron deficiency and protection from malaria infection and (ii) iron supplementation and increased risk of malaria. However, there is still much to be learned. Furthermore, study of the competition between the malaria parasite and the human host for iron can serve as a translational model to identify critical molecular mechanisms of *P. falciparum* pathogenesis (see questions in **Table 4**). Most importantly, however, such research will help the global health community reach their goal of devising a strategy for safely administering iron supplementation in malaria endemic regions.

#### **Table 4 | Questions for future translational research.**

#### **PARASITE**

How does the malaria parasite regulate iron?

What host iron sources are utilized by the malaria parasite?

Does the malaria parasite store iron?

Are parasite virulence factors regulated by iron?

Can merozoites sense host intra-erythrocytic iron?

**RELATIONSHIP BETWEEN PARASITE AND HOST**

Is iron limited enough during iron deficiency or in such excess following iron supplementation to respectively inhibit and exacerbate erythrocytic stage *P. falciparum* infection?

Do iron deficiency and iron supplementation affect erythrocytic stage *P. falciparum* microvascular adhesion, or host endothelial cell activation? Do iron deficiency or iron supplementation impact parasite *var* gene expression or PfEMP1 protein levels on the RBC membrane?

Are there specific strains of *P. falciparum* that are better equipped to infect people with iron deficiency?

What is the effect of host iron deficiency and iron supplementation on *P. falciparum* gametocytogenesis?

What are the effects of changing RBC population dynamics on malaria infection?

How are the host innate and adaptive immune responses to malaria affected by iron deficiency and iron supplementation?

Is anemia of inflammation protective against malaria?

How does the presence of iron deficiency anemia modify the effects of HbS, HbC, or HbE on parasite growth, maturation, microvascular adhesion, or endothelial cell activation?

How do other malaria-protective polymorphisms, such as type O blood group antigen and glucose-6-phosphate dehydrogenase (G6PD) deficiency, interact with iron deficiency in mitigating malaria pathogenesis?

#### **AUTHOR CONTRIBUTIONS**

Martha A. Clark, Morgan M. Goheen, and Carla Cerami wrote and edited the manuscript. All authors have read and approved the final manuscript.

## **ACKNOWLEDGMENTS**

The work was supported by the National Institute of Child Health and Human Development under award number U01HD061235 (to CCH). We thank Dan Raiten, Steven R. Meshnick, and Con Beckers for many useful discussions.

#### **REFERENCES**


meta-analysis of randomized controlled trials. *CMAJ* 185, E791–E802. doi: 10.1503/cmaj.130628


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

*Received: 15 January 2014; paper pending published: 12 March 2014; accepted: 04 April 2014; published online: 06 May 2014.*

*Citation: Clark MA, Goheen MM and Cerami C (2014) Influence of host iron status on Plasmodium falciparum infection. Front. Pharmacol. 5:84. doi: 10.3389/fphar. 2014.00084*

*This article was submitted to Drug Metabolism and Transport, a section of the journal Frontiers in Pharmacology.*

*Copyright © 2014 Clark, Goheen and Cerami. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*

## Iron overload in Plasmodium berghei-infected placenta as a pathogenesis mechanism of fetal death

## *Carlos Penha-Gonçalves, Raffaella Gozzelino and Luciana V. de Moraes\**

Instituto Gulbenkian de Ciência, Oeiras, Portugal

#### *Edited by:*

Paolo Arosio, University of Brescia, Italy

#### *Reviewed by:*

Francesca Vinchi, Heidelberg University, Germany Demba Sarr, University of Georgia, USA

#### *\*Correspondence:*

Luciana V. de Moraes, Instituto Gulbenkian de Ciência, Rua da Quinta Grande 6, 2780-156 Oeiras, Portugal e-mail: lmoraes@igc.gulbenkian.pt

Plasmodium infection during gestation may lead to severe clinical manifestations including abortion, stillbirth, intrauterine growth retardation, and low birth weight. Mechanisms underlying such poor pregnancy outcomes are still unclear. In the animal model of severe placental malaria (PM), in utero fetal death frequently occurs and mothers often succumb to infection before or immediately after delivery. Plasmodium berghei-infected erythrocytes (IEs) continuously accumulate in the placenta, where they are then phagocytosed by fetalderived placental cells, namely trophoblasts. Inside the phagosomes, disruption of IEs leads to the release of non-hemoglobin bound heme, which is subsequently catabolized by heme oxygenase-1 into carbon monoxide, biliverdin, and labile iron. Fine-tuned regulatory mechanisms operate to maintain iron homeostasis, preventing the deleterious effect of iron-induced oxidative stress. Our preliminary results demonstrate that iron overload in trophoblasts of P. berghei-infected placenta is associated with fetal death. Placentas which supported normally developing embryos showed no iron accumulation within the trophoblasts. Placentas from dead fetuses showed massive iron accumulation, which was associated with parasitic burden. Here we present preliminary data suggesting that disruption of iron homeostasis in trophoblasts during the course of PM is a consequence of heme accumulation after intense IE engulfment. We propose that iron overload in placenta is a pathogenic component of PM, contributing to fetal death. The mechanism through which it operates still needs to be elucidated.

**Keywords: pregnancy malaria, heme, iron, fetal death, trophoblast**

## **INTRODUCTION**

Malaria is an infectious disease that affects millions of individuals every year and remains one of the major causes of morbidity and mortality worldwide (WHO | World Malaria Report, 2013). The disease is transmitted by *Anopheles* mosquito carrying *Plasmodium* parasites. Once injected in the blood stream, *Plasmodium* replicate and mature in the liver (liver stage) before infecting erythrocytes (blood stage; Schofield and Grau, 2005). The blood stage of infection is known to be associated to clinical manifestations of the disease (Mackintosh et al., 2004; Haldar et al., 2007). *Plasmodium* parasites proliferate inside red blood cells; within these erythrocytes circulate freely in the bloodstream and are under risk of removal by the spleen – the organ responsible for eliminating old and modified erythrocytes (Janicik et al., 1978). To avoid this host defense mechanism, intracellular parasites export to the erythrocyte membrane molecules that interact with receptors on the vascular endothelium (Krücken et al., 2005), enabling adherence/sequestration of the infected erythrocyte (IE) in the microvasculature of specific organs such as lung, heart, and brain (Pongponratn et al., 1991). IE sequestration is considered one of the major factors in triggering organ inflammation, leading to severe clinical forms of disease such as cerebral malaria (Carter et al., 2005; Ponsford et al., 2012). It has recently been shown that proliferation and maturation of mutant parasites lacking ability to adhere to host endothelium is reduced in mice when compared to *wild-type* parasites

(Fonager et al., 2012) strengthening the hypothesis that adherence impacts *Plasmodium* proliferation during the blood stage of infection.

In 2007 over 125 million women living in areas with *Plasmodium falciparum* and or *P. vivax* transmission became pregnant and at risk of developing malaria (Dellicour et al., 2010). Pregnant women bitten by infected mosquitoes can develop placental malaria (PM; known also as pregnancy malaria) – a disease characterized by adverse pregnancy outcomes such as abortion, stillbirth, premature delivery and low birth weight babies, which in turn increases infant morbidity. These clinical features can be recapitulated in a BALB/c mouse model of infection at mid-stage pregnancy (Neres et al., 2008) and are associated to accumulation of *Plasmodium* IEs in the placenta (Fried and Duffy, 1996).

The placenta is a favored niche for IE sequestration and there are two different explanations for this phenomenon. Placental cells, specifically trophoblasts which are epithelial cells of fetal origin, express chondroitin sulfate A (CSA) – a sulfated glycosaminoglycan – on their surface (Duffy and Fried, 1999). This offers a binding site for VAR2CSA, a protein encoded for the human malaria parasite *P. falciparum,* exposed on the membrane of IEs (Salanti et al., 2003, 2004; Duffy et al., 2006; Srivastava et al., 2010). Physical CSA–VAR2CSA interaction allows the adhesion of IEs to placental tissue. Microcirculatory dynamics of the placenta is another important factor in IE sequestration. We recently described that the speed of maternal blood circulation in the mouse placenta is heterogeneous: there are areas of high, moderate, and low blood flow (de Moraes et al., 2013). In placentas infected with *P. berghei* – a rodent parasite that lacks the VAR2CSA molecule (Jemmely et al., 2010) – accumulation of IEs is increased in low maternal blood flow regions (de Moraes et al., 2013). This highlights the relevance of placental tissue configuration in promoting sequestration of *Plasmodium* IEs. We hypothesized that IE sequestration may occur through specific interaction of IEs with the trophoblast membrane and is favored by IE arrest in maternal regions characterized of low blood flow.

Placental IE sequestration leads to a local inflammatory response characterized by monocyte infiltration (Abrams et al., 2003). Trophoblasts are also capable of phagocytosing IEs (de Moraes et al., 2013) yeast and bacteria (Amarante-Paffaro et al., 2004) in an attempt to eliminate invading microbes. Trophoblast responses to infection may be deleterious to the developing fetus. Despite many evidences that link PM to placental inflammation, mechanisms underlying poor pregnancy outcomes are still unclear.

Our preliminary data in a mouse model of PM show an association between iron overload in trophoblast of *P. berghei*-infected placentas and fetal death. Iron accumulation was observed in 43% of placentas from dead fetuses and was associated to the dose of injected IEs. Iron deposits were never detected in placentas from live embryos. Our preliminary results also show downregulation of mRNA expression of the heme exporter feline leukemia virus C receptor 1a (FLVCR1a) in infected placentas suggesting a role for this molecule in dysregulation of iron homeostasis in PM. Here, we discuss how iron is regulated during pregnancy and develop a hypothesis to explain infection-induced iron overload in trophoblasts. This may help explain the poor pregnancy outcomes brought about by *Plasmodium* infections during gestation.

## **IRON AND PREGNANCY MATERNAL IRON HOMEOSTASIS**

Iron is essential for all living organisms. It is important in a variety of biological functions involving reduction and oxidation reactions which are crucial for cell survival and proliferation (Crichton and Charloteaux-Wauters,1987). Iron can exchange electrons with a number of different substrates, necessitating tight control of the reactivity of this metal (Hentze et al., 2004). Control is provided by cellular and systemic mechanisms that have been evolved to maintain iron homeostasis and prevent its participation in the Fenton chemistry (Fenton, 1894). Disruption of iron homeostasis leads to production of highly reactive hydroxyl radicals, the cytotoxic effect of which is associated with tissue iron overload, organ dysfunction and disease severity, recently demonstrated in the case of severe forms of malaria (Gozzelino et al., 2012).

Dietary iron is absorbed by the duodenum, transported into the cytosol by the divalent metal transporter-1 (DMT1; Fleming et al., 1997; Gunshin et al., 1997; Veuthy and Wessling-Resnick, 2014) and then either stored within multimeric subunits of ferritin (Harrison and Arosio, 1996) or released into the circulation by the iron exporter ferroportin (Donovan et al., 2000; McKie et al., 2000). Extracellular iron is bound to transferrin (Tf) and delivered

to cells by interaction with Tf receptors (TfR;Aisen, 1998) where it is mainly used for heme biosynthesis and erythropoiesis (Andrews, 2000).

During pregnancy, maternal iron absorbance increases substantially throughout the gestational period [from 0.8 mg/day in the first trimester to 7.5 mg/day in the third trimester (Bothwell, 2000)] to fulfill the needs of the developing fetus. Increased iron absorption is also required for maternal erythropoiesis, to maintain hemoglobin (Hb) levels (Bothwell, 2000) and to increase the iron stores to compensate for blood loss during delivery (Milman, 2006). Studies have shown that 40–70% of iron that is transferred to the fetus derives from maternal iron stores (Murray and Stein, 1970; Bothwell, 2000), suggesting that fetus can mobilize iron from this resource; this will be further discussed.

Maternal iron status during pregnancy can be monitored by biomarkers such as Hb, hematocrit, serum ferritin (SF), serum soluble TfR (sTfR), and total body iron (TBI). SF concentration is a well established marker for iron reserves; SF levels ≤30 μg/L are indicative of low iron reserves and levels ≤12 μg/L are associated with iron deficiency (Milman, 2006). Soluble TfR are detached receptors from young erythrocytes which in high concentrations indicate iron deficiency at cellular level and may be useful marker to monitor erythropoiesis (Feelders et al., 1999). Both SF and sTfR measurements during pregnancy yield reliable information on maternal iron status; the sTfR-index (sTfR/log SF) is a practicable parameter for patients with depleted iron stores (Punnonen et al., 1997) and during pregnancy may correct for plasma volume expansion differences (Cao and O'Brien, 2013).

Another parameter for assessment of iron homeostasis is hepcidin, a hormone mainly produced by the liver that controls iron efflux. Hepcidin binds to ferroportin, induces internalization and degradation of this molecule, preventing iron export from the cell (Nemeth et al., 2004). Levels of hepcidin vary according to the iron concentrations in circulation, and kinetics of this hormone during pregnancy is indicative of the status of maternal iron stores (Simavli et al., 2014). In a cohort of healthy Northern European pregnant women, reduced levels of hepcidin correlated with decreased ferritin and iron concentrations, which occur simultaneously to an increase in sTfR-index in the third trimester (van Santen et al., 2013). As proposed in this study, low hepcidin production might reflect increased iron demands of the developing fetus and increase in maternal erythropoietic activity during pregnancy, which reduces maternal iron stores.

The effect of *Plasmodium* placental infection on maternal and cord blood iron parameters [hepcidin, iron, ferritin, sTfR, Tf saturation (TS)] has been considered. In a population from Gabon, anemic women with PM had a trend for lower levels of sTfR compared to anemic women without PM suggesting that parasitemia could affect proliferation of erythroid progenitors (Van Santen et al., 2011). A slight but not significant increase in ferritin concentrations was also observed in these anemic and infected pregnant women. Studies conducted in Malawi showed that infected pregnant women exhibited increased levels of ferritin compared to non-infected and that cord blood ferritin was correlated with increased maternal parasitemia and lower birth weight (Abrams et al., 2005). Moreover Hb levels in the fetus

were not affected by infection suggesting that fetal iron stores are preserved. Other studies corroborate these observations showing that neither maternal or placental infection had an effect on Hb levels (Mokuolu et al., 2009; Van Santen et al., 2011) as well on iron parameters in the cord blood (Van Santen et al., 2011).

### **IRON TRANSPORT FROM MOTHER TO FETUS**

Both in mouse and human placentas, maternal blood is directly in contact with trophoblasts. These cells, more specifically named cytotrophoblast and syncythiotrophoblasts, are arranged in layers to form a barrier separating maternal from fetal blood and allow flux of gas and nutrient to the fetus through different mechanisms (Sibley et al., 2010; **Figure 1A**). Iron is transferred across the placenta via proteins involved in iron transfer; the detailed mechanisms regulating trophoblasts uptake of iron from maternal circulation and transfer to the fetus are still unclear. It has been suggested that iron uptake from maternal serum may occur primarily via DMT1 (Gruper et al., 2005). DMT1 is a transmembrane glycoprotein that transports divalent iron into the cytoplasm; it is located both in the cell membrane and on late endosomes (Andrews, 1999). Expression of DMT1 on the apical side of human trophoblasts (facing maternal blood) suggests

DMT1 could play a major role in iron uptake by trophoblasts also during pregnancy (Gruper et al., 2005). In human term placentas, DMT1 is located predominantly on the maternal side of syncytiotrophoblasts, more rarely on the fetal side (Li et al., 2012b). Fetal iron uptake is also mediated by the expression of TfR on the apical membrane of human placenta (Kroos et al., 1996; Georgieff et al., 2000; Bastin et al., 2006). This suggests that the mechanism also operates to ensure iron transport from maternal circulation to the fetus. TfR is induced during pregnancy, in response to iron deficiency; it is rapidly transcytosed to the apical membrane of syncytiotrophoblasts (Gruper et al., 2005), as demonstrated in rodent (Gambling et al., 2001; Cornock et al., 2013) and human studies (Li et al., 2012a) of maternal iron deficiency.

Mechanisms of iron transfer from mother to fetus must be well *orchestrated* to guarantee adequate iron supply to the developing embryo. The cross-talk between mother, placenta (trophoblast) and fetus is not yet well understood but the *conductor* of the *symphony* seems to be the fetus. Studies have shown that fetal liver iron status directly correlates with TfR expression on placenta and maternal liver, suggesting that the embryo regulates iron uptake by the mother and at the trophoblast level (Gambling et al., 2009). In a rat model of dietary iron deficiency using

bar 25 μm.

maternal blood space (MBS), endothelial cell (EC), and fetal

two different strains, placental TfR and DMT1 expression were upregulated, possibly to increase iron transport and supply to the fetus (Cornock et al., 2013). These studies also showed interstrain differences in maternal liver iron contents in controls not associated to expression of duodenal iron transporters but rather placental efficiency in iron uptake and transfer to the embryo. In the placenta TfR seems to be modulated by hepcidin; studies on hepcidin transgenic embryos have shown that expression of TfR mRNA was significantly downregulated in the placentas, suggesting that hepcidin is constitutively produced by the fetal liver and possibly controlled by fetal iron contents (Gambling et al., 2009) modulating the iron uptake by trophoblast (Martin et al., 2004). Underlying mechanisms are still unclear. Taken together these data strongly suggest that fetal iron status regulates trophoblast iron uptake.

Much evidence supports ferroportin 1 (FPN1) as the major iron export molecule due to its high expression on the basal membrane of the syncytiotrophoblast (Donovan et al., 2000; McKie et al., 2000; Bastin et al., 2006). However, the observation that FPN mRNA expression (Gambling et al., 2001; Li et al., 2008) and protein levels (Gambling et al., 2009; Cornock et al., 2013) on trophoblasts of rat placentas were not altered regardless of maternal iron status suggests that regulation of iron efflux from trophoblast may not involve the FPN-hepcidin pathway (Gambling et al., 2009; Cornock et al., 2013). The observation that DMT1 is also localized on the syncytiotrophoblast basal membrane raises the possibility that this molecule could be involved in iron efflux to the fetal serum (Georgieff et al., 2000).

### **IRON DEFICIENCY AND PREGNANCY MALARIA**

In malaria endemic regions, pregnant women with nutritional iron deficiency who are infected with *Plasmodium* may be protected from PM (Kabyemela et al., 2008) but are at risk of developing severe anemia (Guidotti, 2000) increasing the risk of perinatal mortality and morbidity (Levy et al., 2005). The apparent contradiction stems from two distinct effects of iron: low levels may impair fetal development; high concentration could promote parasite replication, as this metal is strictly required for the viability of these pathogens (Cabantchik et al., 1996; Ganz, 2009). Iron supplementation trials have shown contradictory results on pregnancy outcomes during malaria infection (Nacher et al., 2003; Friedman et al., 2009). Iron and acid folic supplementation has been recommended for all women of child-bearing age in malaria endemic regions with the aim of overcoming iron deficiency. Safety of this strategy in terms of malaria risk has not been assessed and remains a concern. Increased malaria risk has been observed in pregnant women who receive intravenous iron (Oppenheimer et al., 1986) but oral iron supplementation increased neither the prevalence nor severity of malaria in Gambian women (Menendez et al., 1994). In contrast, iron deficiency has been associated with decreased susceptibility to *P. falciparum* infection in pregnant women in Tanzania (Kabyemela et al., 2008) but PM appeared to protect Malawian pregnant women from maternal iron deficiency (Senga et al., 2011). Likewise Sudanese women with *P. falciparum*PM had low frequency of anemia (Adam et al., 2012). These epidemiological findings suggest impairment of iron metabolism over the course of malaria infection during

pregnancy. On the one hand, decreased iron availability appears to protect against PM, probably by depriving the parasite of iron required for expansion during blood stage infection. On the other hand, PM appears to protect pregnant women from developing iron deficiency, possibly by impairing the mechanism of iron transport in the placenta. Dysregulation of iron metabolism and distribution in malaria infection may be exacerbated during pregnancy, due to increased maternal erythropoiesis and fetal needs. There is little information on whether alterations in the intricate interaction between maternal and fetal iron metabolism in PM impose restrictions on fetal growth and survival. There are so far no hypotheses to explain the interplay between iron homoeostasis, pregnancy status, and malaria infection.

## **IRON DYSREGULATION IN THE PM MODEL TROPHOBLAST RESPONSE TO INFECTION**

A critical factor in PM pathogenesis is sequestration of IEs in the placenta. This process may occur through interaction of VAR2CSA exposed on IE membrane with CSA expressed on trophoblasts (Fried and Duffy, 1996) or by arrest of IEs in maternal low blood flow areas (de Moraes et al., 2013). In *P. falciparum* infection, VAR2CSA is so far the best characterized molecule involved in cytoadherence (Salanti et al., 2003, 2004) and a favored target of PM vaccines (Kane and Taylor-Robinson, 2011). Trophoblasts can respond to infection after IE binding. *In vitro* studies with *P. falciparum* have shown that following IEtrophoblast interaction placental cells produce chemokines such as macrophage inflammatory protein-1 alpha (MIP-1α/CCL3) and macrophage migration inhibitory factor (MIF) via activation of c-Jun N-terminal kinase I (JNK I) which in turn recruit peripheral blood mononuclear cells to the site (Lucchi et al., 2008); this inflammatory milieu contributes to PM pathology (**Figure 2**). Trophoblasts have also been shown to respond to *Plasmodium* products such as hemozoin (crystallized heme from Hb digestion during the blood stage of *Plasmodium* infection; Lucchi et al., 2011), *P. chabaudi* IEs (Poovassery and Moore, 2009), and pathogenic bacteria (Griesinger et al., 2001) releasing pro-inflammatory factors.

Despite VAR2CSA-CSA interactions being exclusive to *P. falciparum*-induced PM, the observation that other *Plasmodium* parasites are capable of causing malaria during pregnancy both in humans (Poespoprodjo et al., 2008) and mice (Neres et al., 2008; Rodrigues-Duarte et al., 2012) suggests involvement of other parasite molecules that could target CSA. This is supported by *in vitro* data showing that *P. berghei* IE binding is CSA-dependent (Neres et al., 2008). The contribution of VAR2CSA and non-VAR2CSA molecules to PM is currently being addressed by the characterization of a PM mouse model induced by transgenic *P. berghei* parasite that expresses VAR2CSA molecule (de Moraes et al., unpublished data). Our recent study of blood circulation patterns in mouse placenta using intra-vital microscopy showed that IEs and non-IEs are arrested in regions where maternal circulation is stationary (de Moraes et al., 2013) suggesting that here *P. berghei* IE sequestration might be independent of molecular interactions with trophoblasts.

Stationary IEs in low maternal blood spaces can be engulfed by trophoblasts. Phagocytic activity of trophoblasts occurs in early stages of pregnancy, during implantation of blastocyst (Tachi et al., 1970; Bevilacqua and Abrahamsohn, 1989); engulfment of maternal components such as endometrial cells, red blood cells, and epithelial cells creates space for embryo attachment to the endometrium and its nutrition (Bevilacqua and Abrahamsohn, 1989). After maturation, trophoblasts seem to retain their phagocytic activity. Studies have shown that trophoblasts from mid-stage gestation are capable of phagocytosing erythrocytes (Albieri and Bevilacqua, 1996), *P. chabaudi* IEs (Poovassery and Moore, 2009), *P. berghei* parasitic material *in vitro* (Pavia and Niederbuhl, 1991) as well as yeast and bacteria *in vivo* (Amarante-Paffaro et al., 2004). Our recent observation using intra-vital microscopy show that in placentas from gestational day 18, *P. berghei* IEs are engulfed by trophoblasts supporting the idea that phagocytic function is extended until termination of pregnancy (de Moraes et al., 2013; **Figures 1B,C**).

### **DISRUPTION OF INTRACELLULAR HEME HOMEOSTASIS AS A MECHANISM INVOLVED IN EXCESSIVE IRON ACCUMULATION IN TROPHOBLAST**

In our PM model, abortions, stillbirths, underweight babies, and maternal mortality are consequences of *Plasmodium* infection during pregnancy. These features are associated with irreversible placental damage, due to IE accumulation. Here we present a novel hypothesis as to the mechanisms underlying fetal death in PM. Fetal death/loss during mid-gestation (abortions) has already been shown to be associated to pro-inflammatory cytokine response to *Plasmodium* infection in the mouse (Poovassery and Moore, 2009; Poovassery et al., 2009) and in humans (Okoko et al., 2003), and to microcirculatory impairments (Avery et al., 2012). In our model fetal death was observed at end-stage gestation (G19/20): 8/30 fetuses were expelled at G19 and were very small for gestational age. The remaining fetuses were assessed *in utero*, were in general fully developed but underweight. Our preliminary data show iron accumulation in trophoblasts in 43% of the infected placentas from dead fetuses (**Table 1**), which was associated to the dose/number of *Plasmodium*-IEs used in infection of pregnant mice (**Figure 3** and **Table 2**).

Iron deposits were assessed in placental sections by Perl's Prussian blue staining (**Figure 4**). The pattern of excessive iron staining in placentas from dead fetuses (**Figures 4E,F**) was observed neither in non-infected (**Figures 4A,B**) nor in infected placentas from live but underweight fetuses (**Figures 4C,D**); this strongly



Association between phenotypes (iron accumulation and death): p < 0.01 (chisquare test; df = 1).

**infection.** Pregnant mice were infected with either 10<sup>4</sup> or 10<sup>6</sup> Plasmodium berghei-infected erythrocytes (IEs). Placentas from Plasmodium-infected pregnant mice were collected at G19/20 and processed for iron measurements as previously described (Gozzelino et al., 2012). Briefly placentas were dehydrated overnight, dissolved in 3 M HCl/trichloroacetic acid (TCA) 10% for 24 h. Labile iron was detected by colorimetric reagent (bathophenanthroline-disulfonic acid; BPTS) and absorbance was measured by spectrophotometer (SmartSpec 3000, Bio-Rad); \*\*p < 0.01.

**Table 2 | Iron accumulation in placentas from dead fetuses according to IE dose.**


Association between phenotypes (iron accumulation and IE dose): p < 0.01 (chisquare test; df = 1).

suggests that iron overload in trophoblasts is associated with the most severe pathological events. Iron was accumulated inside trophoblasts (**Figures 4E,F**) and, in particular, at the interface between trophoblasts and fetal blood capillary and, possibly in fetal circulation (**Figure 4F**). Although further experiments would be required to validate the latter observation, staining was not observed in maternal blood spaces (**Figures 4E–G**) supporting the specificity of the results obtained. We also observed loss of placental tissue integrity and cellular damage in regions of iron positive staining (**Figure 4E**), which suggests that iron overload in trophoblasts might induce programmed cell death, presumably mediated by oxidative stress. Increased apoptosis of trophoblasts has been linked to adverse pregnancy outcomes such as low birth weight due to intrauterine growth retardation (IUGR; Endo et al., 2005), spontaneous abortion (Olivares et al., 2002; Nakashima et al., 2008), and in a mouse model of intrauterine fetal death (Mu et al., 2003). In an experimental model of iron overload during pregnancy, iron accumulation in trophoblasts showed a fivefold increase but fetal iron content was not modified, suggesting that transfer of iron across the placenta is, to quote, *a rate-liming step* (Martin et al., 2004). Despite these evidences, we

would not rule out the possibility that, in our model, excessive iron in the placenta is toxic for the fetus; the observation that iron accumulates at the interface with fetal capillaries and possibly in the fetal circulation allows us to speculate toward this direction.

The mechanism proposed to explain iron accumulation in our model relies on dysregulation of intracellular heme homeostasis leading to iron overload presumably occurring as a consequence of increased IE phagocytosis by trophoblasts. These cells are capable of engulfing and digesting red blood cells (Albieri and Bevilacqua, 1996) possibly for fetal nutrition at the beginning of gestation (Bevilacqua and Abrahamsohn, 1989); hence, it is possible that the mechanisms involved in iron recycling from Hb in the trophoblast occurs as in macrophages (Bratosin et al., 1998). We would expect that following phagocytosis, disrupted erythrocytes would lead to intracellular accumulation of cell-free Hb, a tetramer responsible for oxygen transport (Bratosin et al., 1998) which upon oxidation releases its heme prosthetic groups (Schechter, 2008). Non-Hb bound heme (free heme) sensitizes non-hematopoietic cells to undergo programmed cell death in response to pro-inflammatory agonists such as tumor necrosis factor (TNF) released in the course of the infection (Seixas et al., 2009). This deleterious effect is caused by the iron atom contained within the protoporphyrin ring of the heme molecule and its ability to produce highly reactive hydroxyl radicals via its participation in the Fenton chemistry (Gozzelino et al., 2012).

Degradation of heme by the heme catabolizing enzyme heme oxygenase 1 (HO-1) prevents heme-induced iron cytotoxicity and generates equimolar amounts of labile iron (Fe2+), carbon monoxide (CO), and biliverdin (Tenhunen et al., 1968; Choi and Alam, 1996; Soares and Bach, 2009; Gozzelino et al., 2010). Expression of HO-1 is induced in a variety of stress responses, including to heme itself and is encoded by *HMOX1* gene (Tenhunen et al., 1968). This is achieved by suppressing the activity of Bach-1, transcriptional repressor of *Hmox-1* gene (Sun et al., 2002; Gozzelino et al., 2010). The crucial role of HO-1 has already been demonstrated in cerebral and severe forms of malaria (Pamplona et al., 2007; Seixas et al., 2009; Ferreira et al., 2011); increased expression of this detoxifying enzyme strongly correlates with the ability to survive the infection. The protection afforded by HO-1 relies on the inhibition of heme sensitization to programmed cell death.

To limit pro-oxidant effects of iron, a regulatory mechanism evolved to couple heme catabolism and maintenance of iron homeostasis. This metabolic iron adaptation is conferred by the expression of ferritin heavy/heart chain (FtH; Gozzelino and Soares, 2014). Conversion of Fe2<sup>+</sup> into inert Fe3<sup>+</sup> is attributed to the ferroxidase activity of FtH, a property necessary to neutralize the redox activity of labile iron and prevent generation of oxidative stress (Eisenstein et al., 1991; Harrison and Arosio, 1996). This allows storage of iron inside the multimeric complex of ferritin, formed by 24 subunits of FtH and ferritin light/liver chain (FtL; Arosio and Levi, 2002) at proportions that varies between different tissues (Harrison and Arosio, 1996). The iron storage capacity of the multimeric complex of ferritin can incorporate up to 4500 iron atoms (Harrison and Arosio, 1996), as inorganic ferrihydrite aggregates from which

Plasmodium-infected pregnant females were evaluated at G19 for iron accumulation by Prussian blue staining in placental sections. **(A**,**B)** Non-infected placenta showing maternal blood space (MBS) and fetal capillary (FC) separated by trophoblast. **(C**,**D)** Infected placenta from a live fetus that shows no iron accumulation in trophoblasts and fairly preserved tissue structure; **(E**,**F)** Infected placenta from a dead fetus showing

tissue structure. Iron seems to be accumulated in phagocytic trophoblasts trophoblast and fetal capillary (arrows in **F**) but not in MBS **(E–G)**. Areas delimitated in **A**,**C**,**E** are magnified (crop and zoom) in **B**,**D**,**F,** respectively. Scale bars: 25 μm **(A**,**C**,**E)** and 10 μm **(G)**. Arrowheads point to infected erythrocytes.

are then released according to cellular requirements (Koorts and Viljoen, 2007). The mechanism of release is currently not understood.

The cytoprotective effect of FtH relies on its anti-oxidant properties, which reduce free radicals production (Balla et al., 1992). Reactive oxygen species (ROS) are responsible for sustained activation of JNK (Kamata et al., 2005) which induces caspase-3 cleavage and programmed cell death (Gozzelino et al.,2012). Through inhibition of ROS, FtH diminishes JNK phosphorylation (Pham et al., 2004) preventing cell death. In malaria infection, FtH was shown to prevent tissue damage and reduces disease severity by providing a metabolic adaptation to tissue iron overload; this enables mice and humans to survive to the infection (Gozzelino et al., 2012). Further experiments would be required in our PM model to assess the mechanisms by which FtH can protect against trophoblast damage.

Intracellular accumulation of free heme is also controlled by the expression of proteins that regulate heme trafficking inside the cells, known as heme transporters (Yuan et al., 2013). Extracellular release of excessive heme is mediated by the expression of membrane proteins that act as heme exporters, such as BCRP/ABCG2 (Krishnamurthy and Schuetz, 2005) and two isoforms of the FLVCR1 (Keel et al., 2008; Chiabrando et al., 2012). FLVCR1a is a cell surface molecule and regulates heme extracellular exit (Keel et al., 2008; Vinchi et al., 2014); FLVCR1b is a mitochondrial protein that controls heme efflux into the cytosol for hemoglobinization and erythroid differentiation (Chiabrando et al., 2012).

Our preliminary data also show that FLVCR1a mRNA expression in infected placentas is downregulated (data not shown) suggesting that reduced FLVCR1a expression together with intense trophoblast phagocytic activity may contribute to intracellular heme accumulation (**Figure 5**). Placenta has been shown to express high amounts of FLVCR1 (Jaacks et al., 2011) but the functional roles of this molecule in this tissue are poorly investigated. FLVCR1 is required for erythroid differentiation maintaining the balance of intracellular free heme during erythropoiesis (Quigley et al., 2004) and has an important role in iron recycling by macrophages (Keel et al., 2008). It has been suggested that expression of FLVCR1 in the placenta may reverse the flow of heme from this tissue to maternal circulation, so to prevent the fetoplacental unit from iron toxicity (Cao and O'Brien, 2013). Analysis of FLVCR1 expression in a population of pregnant adolescents at high risk of iron deficiency showed that FLVCR1 was downregulated in placentas of anemic mothers at term (Jaacks et al., 2011). In our PM model it is possible that placental FLVCR1a downregulation may be linked to maternal anemia induced by *Plasmodium* infection. Engulfing high numbers of sequestered IEs in low maternal blood flow regions by trophoblasts expressing low levels of FLVCR1a could result in heme accumulation. We have not yet determined if Hb levels and FLVCR1a expression correlates to IE expansion in the placenta, which could possibly explain iron overload in placentas from mice infected with higher parasite numbers.

We could argue that mechanisms involved in the putative hemeinduced cellular damage observed in infected placentas could be associated with a decrease in expression of cytoprotective proteins, such as HO-1 and ferritin. These proteins cooperate to degrade heme and neutralize the iron released from the protoporphyrin ring of the heme molecule (Gozzelino and Soares, 2014). Whilst HO-1 is responsible for heme catabolism, expression of ferritin is crucial to neutralize iron generated through heme degradation and store it inside its multimeric structure (Gozzelino and Soares, 2014). This would lead to insufficient function to prevent the damage induced by intracellular heme accumulation. In fact, administration of anti-oxidant compounds that mimic the cytoprotective effect of HO-1 prevented hememediated cytotoxicity and pathological outcomes of *Plasmodium* infection in mice (Seixas et al., 2009). However, a recent study in mice with conditional disruption of *Flvcr1a* gene in the liver showed that in the absence of FLVCR1a heme and iron accumulated in this organ and that genes such as *Ho-1, Fpn, H-Ft* and *L-Ft* were upregulated suggesting that FLVCR1a is associated to heme degradation (Vinchi et al., 2014). Therefore it is possible that in our model heme degradation pathway in trophoblasts is upregulated to compensate for the impaired function of heme export by FLVCR1a, as previously suggested (Vinchi et al., 2014).

Heme is synthesized in the mitochondria by a series of enzymatic reactions and exported to the cytosol to be incorporated in hemoproteins (reviewed in Chiabrando et al., 2014). Mice lacking FLVCR1a in the liver also showed downregulation of genes involved in the heme biosynthesis after injection of heme precursor 5-aminolevulinic acid (ALA) whereas in controls treatment induced FLVCR1a upregulation (Vinchi et al., 2014). Extrapolation of this evidence to our model allow the speculation that free heme accumulation after intense phagocytosis of IEs by trophoblast could repress heme *de novo* biosynthesis which in turn downregulates FLVCR1a expression.

Whether there is a role for HO-1 in protection against PM remains to be established. The notion that this protein is essential for processes such as oocyte maturation, fertilization, fetus implantation, and acceptance, Bainbridge and Smith (2005) and Zenclussen et al. (2007) suggests that the positive potential of HO-1 in PM is worth to be tested.

## **FINAL REMARKS**

Our PM model represents a severe form of the disease; mechanisms underlying fetal death are still unclear. Our initial observation that only 43% of placentas from dead fetuses showed iron accumulation in trophoblasts (**Table 1**) suggests that iron overload in PM is a heterogeneous and progressive process and potentially represents one of the mechanisms to explain fetal demise. The amount of iron in the placenta was significantly associated with infection dose (**Figure 2** and **Table 2**) implying that the higher the degree of infection the higher the probability of iron accumulation. Pregnancy outcomes are improved when mice are infected with IE doses 100× times lower than usual; incidence of fetal death is greatly decreased and birth weight is comparable to normal (unpublished) suggesting that here infection does not have an important impact on fetal development. It could also indicate that lower infection levels may not deprive the organism of the capacity to activate cytoprotective

**according to IE load.** Left panel: **(A)** increased phagocytosis of IEs and possibly non-IEs by cytotrophoblast (CTB) and syncythiotrophoblasts (STB) in low blood flow regions in maternal blood space (MBS) where IE and non-IE are sequestered; **(B)** disruption of erythrocytes in the phagocytic vacuole leads to free heme exposure; downregulation of FLVCR1a on trophoblasts might cause free heme accumulation due to decreased heme which may exceed ferritin binding capacity; **(D)** free iron (Fe2+) induces cell death by ROS. Right panel: **(A)** lower placental infection might not affect FLVCR1a expression and phagocytosis is decreased; **(B)** excessive free heme is exported from the cell; **(C)** iron from heme catabolism is neutralized by ferritin; and **(D)** trophoblasts are protected from programmed cell death.

molecules controlling regulation of heme/iron homeostasis. This is further supported by the observation that iron accumulation in infected placentas is associated with the number of injected IEs.

The main hypothesis we formulate in this article is that increased phagocytosis of IEs by trophoblasts leads to intracellular iron overload as a consequence of accumulation of free heme after IE disruption. Based on preliminary evidences we suggest that infection induces downregulation of FLVCR1a which affects heme homeostasis triggering cytotoxic events in trophoblasts and impairing fetal development.

Whether *Plasmodium* infection induces dysregulation of molecules involved in heme-iron uptake, storage and/or efflux leading to intracellular iron overload is highly speculative. Nevertheless, our observations that iron was accumulated at the interface between trophoblast basal membrane andfetal capillary encourage the investigation of FPN expression levels in the infected placenta. We could further elaborate that insufficient function of FLVCR1a in exporting intracellular heme together with inefficient iron export through FPN (if shown to be downregulated) might both contribute to iron accumulation in trophoblasts. Our current observations warrantfurther efforts to investigate iron metabolism in PM.

#### **ACKNOWLEDGMENTS**

The authors would like to thank the Histopathology Unit at Instituto Gulbenkian de Ciência for performing histological sections and Richard Hampson for proof reading the manuscript. This work was supported by fundings from Fundação de Ciência e Tecnologia (FCT), Portugal (EXPL-IMI-IMU-0428/2013). Raffaella Gozzelino and Luciana V. de Moraes are recipients of Post-Doctoral fellowships SFRH/BPD/44256/2008 and 44486/2008, respectively.

### **REFERENCES**


growth restriction and without maternal symptoms. *Int. J. Mol. Med.* 16, 79–84. doi: 10.3892/ijmm.16.1.79


cytokine responses. *Infect. Immun.* 77, 4998–5006. doi: 10.1128/IAI. 00617-09


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

*Received: 02 May 2014; accepted: 12 June 2014; published online: 01 July 2014. Citation: Penha-Gonçalves C, Gozzelino R and de Moraes LV (2014) Iron overload in Plasmodium berghei-infected placenta as a pathogenesis mechanism of fetal death. Front. Pharmacol. 5:155. doi: 10.3389/fphar.2014.00155*

*This article was submitted to Drug Metabolism and Transport, a section of the journal Frontiers in Pharmacology.*

*Copyright © 2014 Penha-Gonçalves, Gozzelino and de Moraes. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*

## Behavioral decline and premature lethality upon pan-neuronal ferritin overexpression in Drosophila infected with a virulent form of Wolbachia

## *Stylianos Kosmidis1,2 , Fanis Missirlis 3,4 \*, Jose A. Botella5 , Stephan Schneuwly5 ,Tracey A. Rouault <sup>3</sup> and Efthimios M. C. Skoulakis1*

<sup>1</sup> Neuroscience Division, Biomedical Sciences Research Centre "Alexander Fleming", Vari, Greece

<sup>2</sup> Department of Neuroscience, Columbia University, New York, NY, USA

<sup>3</sup> Molecular Medicine Program, Eunice Kennedy Shriver National Institute of Child Health and Human Development, National Institutes of Health,

Bethesda, MD, USA

<sup>4</sup> Departamento de Fisiología Biofísica y Neurociencias, Centro de Investigación y de Estudios Avanzados del Instituto Politécnico Nacional, México City, México

<sup>5</sup> Institute of Zoology, University of Regensburg, Regensburg, Germany

#### *Edited by:*

Paolo Arosio, University of Brescia, Italy

#### *Reviewed by:*

Bing Zhou, Tsinghua University, China María-José Martínez-Sebastián, University of Valencia, Spain

#### *\*Correspondence:*

Fanis Missirlis, Departamento de Fisiología Biofísica y Neurociencias, Centro de Investigación y de Estudios Avanzados del Instituto Politécnico Nacional, Avenida Instituto Politécnico Nacional 2508, Zacatenco, 07360 México City, District Federal, México e-mail: fanis@fisio.cinvestav.mx

Iron is required for organismal growth. Therefore, limiting iron availability may be a key part of the host's innate immune response to various pathogens, for example, in Drosophila infected with Zygomycetes. One way the host can transiently reduce iron bioavailability is by ferritin overexpression. To study the effects of neuronal-specific ferritin overexpression on survival and neurodegeneration we generated flies simultaneously over-expressing transgenes for both ferritin subunits in all neurons. We used two independent recombinant chromosomes bearing UAS-Fer1HCH, UAS-Fer2LCH transgenes and obtained qualitatively different levels of late-onset behavioral and lifespan declines. We subsequently discovered that one parental strain had been infected with a virulent form of the bacterial endosymbiont Wolbachia, causing widespread neuronal apoptosis and premature death. This phenotype was exacerbated by ferritin overexpression and was curable by antibiotic treatment. Neuronal ferritin overexpression in uninfected flies did not cause evident neurodegeneration but resulted in a late-onset behavioral decline, as previously reported for ferritin overexpression in glia. The results suggest that ferritin overexpression in the central nervous system of flies is tolerated well in young individuals with adverse manifestations appearing only late in life or under unrelated pathophysiological conditions.

**Keywords: metal, symbiosis, rickettsia, popcorn,** *w***Mel, insect, immunity, aging**

## **INTRODUCTION**

Iron is essential for the growth of microorganisms and animals because it serves as a cofactor in many enzymes (Sheftel et al., 2012). Symbiotic relationships will therefore require the tuning of iron homeostasis between the organisms, whereas in parasitism, iron sequestration becomes an important factor in the antagonism between parasite and host (Cassat and Skaar, 2013). *Drosophila melanogaster* is used very widely to study basic biological questions and as a model of human disease (Rajan and Perrimon, 2013), but our knowledge of how iron homeostasis is maintained in *Drosophila* is still rather limited [reviewed in (Mandilaras et al., 2013; Tang and Zhou, 2013b)]. It is known that iron is normally stored in specialized intestinal cells expressing ferritin within their secretory system (Locke and Leung, 1984; Mehta et al., 2009). Ironloaded ferritin can be excreted to the intestinal lumen and is also found in the hemolymph (Locke and Leung, 1984). Disrupted ferritin function by mutation or RNA interference (RNAi) results in embryonic or first instar larval lethality (Missirlis et al., 2007; Li, 2010; Tang and Zhou, 2013a). Overexpression of ferritin also leads to excess iron sequestration and functional iron deficiency, which does not impede development to adulthood (Missirlis et al., 2007; Gutierrez et al., 2013; Tang and Zhou, 2013a). Neuronal

ferritin overexpression was found to be beneficial when flies were fed with aluminum (Wu et al., 2012), whereas glial ferritin overexpression resulted in ferritin-iron inclusions in a subset of glia of the optic lobes and mild behavioral defects with a late-onset of appearance (Kosmidis et al., 2011). However, the effects of neuronal ferritin overexpression in otherwise wild type individuals have not been studied to date and were the first objective of this study.

*Drosophila melanogaster* laboratory cultures commonly host a symbiotic relationship with the α-proteobacteria *Wolbachia* species, in which infected females show a reproductive advantage over non-infected, with no other overt fitness costs associated with the presence of the bacterium (Werren et al., 2008; Saridaki and Bourtzis, 2010). As an exception, a virulent *Wolbachia* strain causing degeneration and early death has been identified (Min and Benzer, 1997). Its name, *popcorn* (*w*MelPop) reflects the appearance of the bacteria visualized *in situ* with electron microscopy (EM) and is thought to result from its increased proliferation. The severity of *w*MelPop-mediated phenotypes has been shown to vary with ambient temperature and the genetic background of the host (Reynolds et al., 2003). A whole-genome sequence comparison between variants of *w*MelPop and *w*Mel revealed only minor genetic changes between the pathogenic and symbiotic strains, providing no direct clues to help explain differences in virulence (Woolfit et al., 2013). Intriguingly, a small genomic region encoding 24 bacterial genes (out of a total of 1,111 annotated in the species) was either triplicated or absent in the pathogenic strains (Woolfit et al., 2013). Finally, two recent imaging studies documented in anatomical detail the presence of *w*MelPop in the fly brain (Albertson et al., 2013; Strunov et al., 2013).

Previous results show that*Wolbachia* influences iron homeostasis in *D. melanogaster* and in the closely related species *D. simulans* (Brownlie et al., 2009; Kremer et al., 2009). When grown under iron-limiting conditions, *D. melanogaster* suffers reduced fecundity, but when infected with *w*Mel, the flies laid significantly more eggs in four out of six experimental trials (Brownlie et al., 2009). The beneficial effect of *w*Mel was only seen under iron stress conditions, but the likely relevance of these experiments were supported by the finding that two out of the three natural populations tested were iron-deficient as collected in the wild. In a different paradigm, where *Wolbachia* is an obligate parasite of the wasp *Asobara tabida*, a study identified that both ferritin chains were induced in non-infected wasps (Kremer et al., 2009). To investigate whether *Wolbachia* affected ferritin iron in other insect species, iron accumulation in infected and uninfected *D. simulans* was determined. No difference was seen under normal dietary conditions, but when iron was supplemented in the diet infected individuals accumulated significantly less iron. Therefore, it appears that in a variety of insect species the presence of *Wolbachia* enhances tolerance to iron stress. We describe here the effect of neuronal ferritin overexpression in infected and uninfected *D. melanogaster* adults as they age and provide the first example where imposing an iron stress exacerbates the virulence of a pathogenic form of *Wolbachia*.

## **MATERIALS AND METHODS**

## **FLY STOCKS**

The X-chromosome insertion of the pan-neuronal driver *elav-Gal4* was obtained from the Bloomington *Drosophila* Stock Center (#458). The construction of transgenic stocks *UAS-Fer1HCH* and *UAS-Fer2LCH* have been described in detail elsewhere (Missirlis et al., 2007). The *w*MelPop infection likely occurred from a balancer stock (not recorded) used to recover the *UAS-Fer1HCH*, *UAS-Fer2LCH* recombinant on the X-chromosome. *Wolbachia*infected *D. simulans* flies were a kind gift from Kostas Bourtzis.

## **MICROSCOPY**

Adult brains were prepared, cut, and stained as described (Kretzschmar et al., 1997). Ultrathin Epon plastic sections were post stained with 2% uranyl acetate, followed by Reynolds' lead citrate and stabilized for transmission electron microscopy (TEM) by carbon coating. Examination was done under an optical microscope or with a Zeiss EM10C/VR (Oberkochen, Germany) electron microscope at 40−80 kV.

## **BEHAVIORAL ASSAYS**

Female adult flies collected 0−2 days after eclosion were kept throughout their lifetime at 25◦C in groups of 20 individuals per vial. For elicited-escape response tests flies were adapted for at least 20 min in an environment of 25◦C and 70−80% humidity illuminated by red light. They were then placed in 14 mL polystyrene Falcon tubes individually and each fly was vortexed at the highest speed for 2−3 s and was tested twice in negative geotaxis and horizontal escape assays (Kosmidis et al., 2011). All flies were recorded for at least 1 min or until they reached their target of 10 cm. A minimum 20 flies per genotype were tested for each experimental point and the data analyzed parametrically for statistical significance using the JMP and Prism software. Lifespan determinations used a minimum of 120 flies per genotype.

## **PCR-DETECTION OF** *Wolbachia*

The presence of *Wolbachia* was determined by PCR using 16s rDNA *Wolbachia*-specific primers (Bourtzis et al., 1996). The primer sequences used were as follows: 5 -TTGTAGCCTGCTATG-GTATAACT-3 and 5 -GAATAGGTATGATMTCATGT-3 .

## **RESULTS**

## **PHENOTYPES OBSERVED BY FERRITIN OVEREXPRESSION IN ALL FLY NEURONS**

We used the pan-neuronal driver *elav-Gal4* to overexpress both ferritin subunits simultaneously from recombinant chromosomes bearing transgene insertions of *UAS-Fer1HCH* and *UAS-Fer2LCH.* As the X-chromosome was modified in one stock to carry the two *UAS* transgenes and in the other stock to carry the *Gal4* insertion in the *elav* locus, we used only female flies in our experiments. Following eclosure, adult females were kept with their male siblings for 2 days to ensure successful mating and thereafter, female cohorts were kept under non-crowded conditions with biweekly transfers to fresh food. Flies bearing the transgenes alone were used as controls, but we note that *UAS-Fer1HCH, UAS-Fer2LCH*/+; +/+; +/+ females behaved differently depending on whether their mothers were infected with *Wolbachia* or not and this observation (reported further below) was unknown to us when experiments were originally planned and performed. In **Figure 1** our controls were derived from uninfected *white* (*w*) mutant mothers. We also tested the effect of neuronal overexpression of single subunits, as they have been previously shown to rescue degeneration associated with soluble β-amyloid overexpression (Rival et al., 2009). Once weekly during the lifespan of the flies, we monitored the time required for individual flies to climb 10 cm following a brief stimulus that elicited a negative geotactic response (**Figure 1**). The control flies used in this experiment showed a modest two-fold increase (*p* < 0.05) in the time required to perform this test during the sixth week of testing. Flies overexpressing either ferritin subunit alone behaved similar to controls through the fifth week, but presented significant performance impairment (a fourfold increase in time, *p* < 0.001) when the aging-related decline initiated at week six. The flies overexpressing both ferritin subunits from the transgenes on the third chromosome showed a modest compromise at weeks 5 and 6 (threefold and sixfold increases; *p* < 0.0001). The flies overexpressing both ferritin subunits from the transgenes on the X chromosome showed a robust mobility impediment one week earlier than controls (fourfold increase in time by week five,

*p* < 0.001), with a dramatic tenfold increase in the time required to climb 10 cm by week six (*p* < 0.0001; **Figure 1A**). Similar findings were found when testing horizontal escape responses, which presumably require less physical effort from individual flies (**Figure 1B**). During their sixth week post-eclosion, *elav-Gal4/UAS-Fer1HCH, UAS-Fer2LCH;* +/+; +/+ flies could hardly walk.

overexpressing Fer1HCH and Fer2LCH, respectively. Bars 4 and 5 correspond to flies simultaneously overexpressing both Fer1HCH and Fer2LCH from recombinant transgenes on the third and X chromosomes, respectively.

We then prepared plastic-embedded sections of adult fly heads to monitor the neuroanatomical integrity of the brains of 40-day old flies, at the transition between weeks 5 and 6. No apparent degeneration was evident in *elav-Gal4*/+; +/+; +/+ controls (**Figure 2A**), flies overexpressing the single ferritin subunits (**Figures 2C,D**) and those overexpressing ferritin from the third chromosomal recombinant (**Figure 2E**). In stark contrast, animals overexpressing ferritin from the X-chromosome showed severe neuronal degeneration, evident throughout the brain, including the optic lobes (**Figures 2B,F**). Furthermore, lifespan determinations, clearly distinguished between ferritin overexpressing flies derived from the third and X-chromosomal recombinants, with the latter presenting precipitous mortality during the sixth week of aging, 2 weeks earlier than controls (**Figure 3A**). Therefore, in addition to the significant difference in the onset of the mobility impediment between the two genotypes of flies overexpressing both ferritin subunits, brain pathology, and reduced lifespan were also associated only with the ferritin overexpression induced from the X-chromosomal transgenes.

## **MATERNAL EFFECT ASSOCIATED WITH REDUCED LIFESPAN OF FERRITIN OVEREXPRESSING FLIES**

in weeks 5 and 6 (p < 0.001).

**(B)** Only flies overexpressing both ferritin subunits from the X-chromosome were significantly compromised when responding in the horizontal direction

To investigate the potential cause of the differential shortened lifespan, neurodegeneration and early onset behavioral decline of the X-chromosome driven ferritin overexpression, we tested whether there was a maternal effect associated with these phenotypes. We crossed homozygous female *UAS-Fer1HCH, UAS-Fer2LCH* to male *elav-Gal4* or to *w* control flies and *vice-versa*. *UAS-Fer1HCH, UAS-Fer2LCH/elav-Gal4* flies were therefore derived from *UAS-Fer1HCH, UAS-Fer2LCH* or *elav-Gal4* mothers, whereas *UAS-Fer1HCH, UAS-Fer2LCH/*+ control flies were derived from *UAS-Fer1HCH, UAS-Fer2LCH,* or *w* mothers. The results clearly show that lifespan shortening was observed only in progeny of *UAS-Fer1HCH, UAS-Fer2LCH* mothers (**Figure 3B**). In agreement with this notion the decrease was also seen in control *UAS-Fer1HCH, UAS-Fer2LCH/*+ animals that do not overexpress ferritin since they do not carry the *elav-Gal4* driver, but their transgene carrying chromosome is maternal in origin. Therefore, the behavioral, neurodegenerative, and reduced life span phenotypes are not associated with over-expression of the ferritin subunits, but rather with the maternal origin of the transgene-bearing chromosome.

## *Wolbachia* **CLUSTERS DETECTED BY TEM AND PCR IN THE SEVERELY AFFECTED FERRITIN OVEREXPRESSING FLIES**

Transmission electron microscopy images from affected *UAS-Fer1HCH, UAS-Fer2LCH/elav-Gal4* flies revealed the presence of

ration. **(E)** Flies overexpressing ferritin from the recombinant insertions on the third Chromosome also show no apparent neurodegeneration. **(F)** Close-up of image shown in **(B)**. Filled arrowheads denote the absence of single neuronal cell bodies lined up in the optic medulla. Horizontal arrows indicate areas of extensive neuronal damage.

bacterial infections, with the characteristic morphology attributed to *w*MelPop (**Figure 4A**; Min and Benzer, 1997; Strunov et al., 2013). To verify the presence of *Wolbachia* independently, we performed genomic PCR using bacteria-specific primers on the *UAS-Fer1HCH, UAS-Fer2LCH* and *elav-Gal4* parental lines and on *UAS-Fer1HCH, UAS-Fer2LCH/elav-Gal4* progeny derived from presumably infected and uninfected mothers (**Figure 4B**). We verified the presence of *Wolbachia* in mothers and progeny derived from the stock bearing the *UAS-Fer1HCH, UAS-Fer2LCH* recombinant on the X-chromosome. We also assessed negative geotaxis on *UAS-Fer1HCH, UAS-Fer2LCH/elav-Gal4* flies and in agreement with the life span experiment, we only observed a large time delay in infected individuals at 5 weeks of age (**Figure 4C**). Although the severe neurodegeneration that follows *w*MelPop infection has been previously documented, it has been assumed that the bacterial infection leads to neuronal necrosis (Min and Benzer, 1997). Our TEM analysis demonstrated numerous pyknotic nuclei within CNS neurons (**Figure 5**), consistent with the characteristic nuclear changes that accompany programmed

cell death. Treatment of the *UAS-Fer1HCH, UAS-Fer2LCH* stock with the antibiotic tetracycline for four consecutive generations eliminated the infection (**Figure 6A**). *UAS-Fer1HCH, UAS-Fer2LCH/elav-Gal4* flies derived from the tetracycline-treated *UAS-Fer1HCH, UAS-Fer2LCH* mothers retained their brain integrity at 40 days of age (**Figures 6B,C**) and, by qualitative observations, showed no signs of premature mobility decline and mortality.

## **DISCUSSION**

#### **IRON AND** *Wolbachia* **SYMBIOSIS**

The maternally inherited bacterium *Wolbachia* infects arthropods and nematodes (Serbus et al., 2008). In filarial nematodes, *Wolbachia* is an obligate symbiont required for the successful reproduction of the worm (Bandi et al., 1999). The hypothesis that *Wolbachia* contributes metabolic provisioning to its hosts originated from whole genome sequencing projects showing that the entire heme biosynthesis pathway was present in *Wolbachia* but absent in *Brugia malayi* (Foster et al., 2005). This hypothesis was tested in *D. melanogaster* supporting the notion that the presence of *Wolbachia* assisted the flies under conditions of iron stress (Brownlie et al., 2009). Further support to the idea that iron is important in symbiotic relationships of *Wolbachia* and its hosts came from the finding that ferritin expression was induced in the absence of *Wolbachia* from *A. tabida* (Kremer et al., 2009). However, we still have no molecular understanding of the irondependent interactions between*Wolbachia* and its hosts. The same is true for interactions of *Wolbachia* with other metals (Wang et al., 2012) or with the host's antioxidant defense systems (Brennan et al., 2008).

Our study provided evidence that *w*MelPop, or a *w*MelPoplike strain, causes neuronal death, likely by inducing the apoptotic pathway in infected brains, with apoptosis seen also in the optic lobes, which generally were found to harbor significantly less bacteria (Albertson et al., 2013; Strunov et al., 2013). Ferritin overexpression, which likely results in depletion of readily bioavailable iron (Gutierrez et al., 2013), conferred no apparent protection to infected flies, but instead further decreased their shortened lifespan due to *Wolbachia* infection (**Figure 3B**). Our experiments did not discriminate between the possibility that ferritin overexpression benefited the pathogen, encouraging its propagation, or, alternatively, that ferritin overexpression on top of pathogen infection further deprived bioavailable brain iron. Our results point to a different outcome between *w*MelPop and *w*Mel infections when responding to iron deficiency, with the former killing their host more effectively when the latter appear to provide an advantage to their host (Brownlie et al., 2009), although our observations call for further, differently controlled, experiments to settle this question. It will be of interest to investigate whether iron-related genes are included in the genomic region that appears to distinguish the pathogenic *w*MelPop strains from the more common endosymbiotic *w*Mel variants (Woolfit et al., 2013).

## **IRON LIMITATION AND INSECT IMMUNITY**

The clearest evidence that iron availability can determine the outcome of an infection in *Drosophila* was provided by a study of

#### **FIGURE 3 | Lifespan determinations of cohorts of female adult** *D. melanogaster.* Day 0 marks eclosion, flies were kept with the opposite sex for the first five days, then they were separated on groups of 20 and transferred to fresh food regularly. **(A)** The same genotypes assayed in **Figure 1** for behavior were used here. The experiments were performed three times independently (each time with a minimum cohort of 120 flies per genotype) and standard deviations are shown between the three

experiments. Note the early onset of mortality in the flies overexpressing ferritin from the X-chromosomal recombinant. **(B)** A similar single experiment demonstrating that a maternal component determines the early onset mortality. Note that the same genotypes have markedly different lifespans, whose length is predicted by the mothers used for the crosses (see text for details). Overexpression of ferritin was never seen to have a beneficial effect on lifespan.

infection with the human fungal pathogen *Zygomycetes* (Chamilos et al., 2008). Provision of iron to the fungus resulted in more aggressive infections whereas treatment of flies with an iron chelator suppressed the infection. This outcome is consistent with the broadly recognized role of iron in infection: its availability is good for the pathogen and bad for the host

(Drakesmith and Prentice, 2012). Although very limited studies have been conducted investigating the role of iron during infections in *Drosophila* to date, the emerging picture appears to be complex and specific to each microbe studied. Hence, three qualitatively different interactions have been noted: (i) that iron limitation stops an infection (Chamilos et al., 2008), (ii) that

**FIGURE 5 | (A–D)** Four independent electron micrographs from Wolbachia infected flies showing characteristic apoptotic nuclei, suggesting neuronal death results from apoptosis.

iron limitation exacerbates the pathogenesis of an infection (this study), or (iii) that symbiotic infection becomes beneficial to the host under iron limitation (Brownlie et al., 2009). Clearly the role of iron in *Drosophila* immunity is an exciting open field for study.

## **FERRITIN OVEREXPRESSION IN THE CENTRAL NERVOUS SYSTEM OF** *Drosophila*

We previously described the effect of ferritin overexpression in all *Drosophila* glia (Kosmidis et al., 2011) and in terms of behavior and lifespan, the major conclusions are very similar to these reported here for neurons: there seems to be no decrease in lifespan of flies overexpressing ferritin and there is only a small but readily quantifiable late-onset behavioral defect in negative geotactic responses. Although there is a detectable negative effect from ferritin overexpression at old age in wild type conditions, it has been clearly shown that ferritin overexpression can be protective under conditions of metal stress (Missirlis et al., 2007; Wu et al., 2012). Ferritin overexpression is thought to generate an iron limitation, but ferritin may be also able to sequester other metals with a significantly lower capacity (Gutierrez et al., 2013). One particularity of overexpressing ferritin in glia is that ironloaded ferritin accumulated in a subset of glia in the optic lobes forming inclusions (Kosmidis et al., 2011). No such accumulation was detected with ferritin overexpression in neurons. In addition, RNAi experiments against each ferritin subunit resulted in behavioral alterations when neurons were targeted and lethality when ferritin was suppressed in glia (Mandilaras and Missirlis, 2012; Tang and Zhou, 2013a). Our results suggest that ferritin overexpression, and iron chelators more generally (Soriano et al., 2013),

**from the***Wolbachia* **infection. (A)** PCR detection of the Wolbachia ribosomal 16S gene in (i) infected D. simulans (ii) tetracycline-treated D. simulans (iii) infected UAS-Fer1HCH, UAS-Fer2LCH on X (D. melanogaster) (iv) tetracycline-treated UAS-Fer1HCH, UAS-Fer2LCH on X **(B)** Plastic embedded sections of heads from infected (mothers from (iii) and **(C)** tetracycline-treated (mothers from iv) Elav/UAS-Fer1HCH, UAS-Fer2LCH adults at 40 days of age.

should be considered as a therapeutic means of intervention, but a more general recommendation for iron chelation in otherwise healthy adults is not supported from the model organism *D. melanogaster*.

## **AUTHOR CONTRIBUTIONS**

All authors designed the original set of experiments aiming to address the consequences of ferritin overexpression in neurons of wild type flies and participated in the data analysis. Fanis Missirlis generated the transgenic flies and measured the life spans. Stylianos Kosmidis performed the behavioral tests, the molecular biology to detect *Wolbachia* and the antibiotic treatments. Jose A. Botella and Stephan Schneuwly performed the electron microscopy analysis. Tracey A. Rouault and Efthimios M. C. Skoulakis had continuous input during each stage of the project, including during the writing of the manuscript, with Fanis Missirlis in a coordinating role.

## **ACKNOWLEDGMENTS**

This research was supported by the Intramural Program of the *Eunice Kennedy Shriver* National Institute of Child Health and Human Development in the USA and the CONACYT project 179835 in Mexico. We thank Kostas Bourtzis for providing *Wolbachia* infected flies and advice for detecting the bacteria.

## **REFERENCES**


oxygen species (ROS) production. *Neurobiol. Aging* 33, 199.e1–199.e12. doi: 10.1016/j.neurobiolaging.2010.06.018

**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

*Received: 21 February 2014; paper pending published: 10 March 2014; accepted: 20 March 2014; published online: 04 April 2014.*

*Citation: Kosmidis S, Missirlis F, Botella JA, Schneuwly S, Rouault TA and Skoulakis EMC (2014) Behavioral decline and premature lethality upon pan-neuronal ferritin* *overexpression in Drosophila infected with a virulent form of Wolbachia. Front. Pharmacol. 5:66. doi: 10.3389/fphar.2014.00066*

*This article was submitted to Drug Metabolism and Transport, a section of the journal Frontiers in Pharmacology.*

*Copyright © 2014 Kosmidis, Missirlis, Botella, Schneuwly, Rouault and Skoulakis. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*

## The role of iron in anthracycline cardiotoxicity

## *Elena Gammella1, Federica Maccarinelli <sup>2</sup> , Paolo Buratti 1, Stefania Recalcati <sup>1</sup> and Gaetano Cairo1\**

<sup>1</sup> Department of Biomedical Sciences for Health, University of Milano, Milano, Italy

<sup>2</sup> Department of Molecular and Translational Medicine, University of Brescia, Brescia, Italy

#### *Edited by:*

Paolo Arosio, University of Brescia, Italy

#### *Reviewed by:*

Antonio Lax, Universitary Hospital Virgen de la Arrixaca, Spain Hossein Ardehali, Northwestern University, USA

*\*Correspondence:* Gaetano Cairo, Department of Biomedical Sciences for Health, University of Milano, Via Mangiagalli 31, 20133 Milano, Italy e-mail: gaetano.cairo@unimi.it

The clinical use of the antitumor anthracycline Doxorubicin is limited by the risk of severe cardiotoxicity. The mechanisms underlying anthracycline-dependent cardiotoxicity are multiple and remain uncompletely understood, but many observations indicate that interactions with cellular iron metabolism are important. Convincing evidence showing that iron plays a role in Doxorubicin cardiotoxicity is provided by the protecting efficacy of iron chelation in patients and experimental models, and studies showing that iron overload exacerbates the cardiotoxic effects of the drug, but the underlying molecular mechanisms remain to be completely characterized. Since anthracyclines generate reactive oxygen species, increased iron-catalyzed formation of free radicals appears an obvious explanation for the aggravating role of iron in Doxorubicin cardiotoxicity, but antioxidants did not offer protection in clinical settings. Moreover, how the interaction between reactive oxygen species and iron damages heart cells exposed to Doxorubicin is still unclear. This review discusses the pathogenic role of the disruption of iron homeostasis in Doxorubicinmediated cardiotoxicity in the context of current and future pharmacologic approaches to cardioprotection.

**Keywords: iron, anthracyclines, heart, doxorubicin, reactive oxygen species**

## **INTRODUCTION**

Doxorubicin (DOX) is a widely used anthracycline anticancer drug that intercalates in DNA and inhibits topoisomerase II. More than four decades after discovery, DOX still ranks among the most effective agents in breast cancer and numerous other neoplastic diseases. Unfortunately, the clinical use of DOX is limited by the possible development of cardiomyopathy and congestive heart failure (CHF). The various forms of anthracycline-mediated cardiotoxicity, as well as the reasons for the particular susceptibility of cardiac cells, have been reviewed elsewhere (Minotti et al., 2004a; Berthiaume and Wallace, 2007; Octavia et al., 2012; Carvalho et al., 2014); here we will summarize the major clinical consequences of this toxicity, and then review the biochemical mechanisms of anthracycline cardiotoxicity, focusing on the role of iron.

Since a tight relationship between the quantity of DOX accumulated in the heart and the incidence of cardiac events has been identified, the lifetime cumulative dose of DOX usually does not exceed 500 mg/m2; this reduces considerably the incidence of treatment-requiring cardiac events, but sometimes can preclude successful completion of chemotherapy. Moreover, cardiomyopathy may develop at lower doses in the presence of risk factors like age, hypertension, arrhythmias, coronary disease, etc. (Gianni et al., 2008). Therefore, the dose-adjusted chance of cardiotoxicity may be more elevated than expected, and the necessity to adjust doses limits the therapeutic capability of DOX and other anthracyclines. It should also be considered that anthracycline cardiotoxicity is often exacerbated by combination therapies involving classic drugs like taxanes or new drugs targeting growth factor receptors or signal transducing molecules activated in tumors (Gianni et al., 2008).

Since at present there is no specific and effective pharmacological treatment for DOX-related cardiomyopathy, cardiac transplantation often remains the only option for patients with DOX-dependent advanced cardiomyopathy and CHF. It is therefore evident that the risk of developing cardiac events because of anthracycline treatment is a life-threatening problem for a high number of cancer survivors. In particular, given the peculiar delay in the appearance of toxic consequences, which may become manifest several years after completing chemotherapy (Minotti et al., 2004a), DOX cardiotoxicity represents a serious problem in pediatric oncology (Lipshultz et al., 2013a).

Despite convincing evidence linking the levels of DOX in the heart and the development of cardiomyopathy, the precise mechanisms through which DOX eventually induces cardiac damage are much less clear; this uncertainty is also dependent on obstacles to investigation posed by species-related differences in anthracycline metabolism and susceptibility to DOX-derived reactive oxygen species (ROS; see below), as well as to inherent difficulties in developing animal models reliably recapitulating chronic and delayed cardiotoxicity.

Anthracycline-induced cardiotoxicity is a multifactorial process based on mechanisms only in part distinct from those targeting cancer cells. In fact, it has been demonstrated that the interaction of DOX with topoisomerase, which is key for the antitumor effect of anthracyclines, is also relevant for cardiotoxicity (Zhang et al., 2012). Other mechanisms include suppression of the cardiacspecific gene expression program, persistent alterations of calcium compartmentalization, changes in adrenergic function, inhibition of the expression of several sarcomeric proteins, promotion of free radical reactions beyond the limited antioxidant defence of cardiomyocytes that eventually lead to lipid peroxidation (Minotti

et al., 2004a). Great attention is paid to the latter mechanism because a number of studies have shown that reductive activation of the quinone moiety of DOX (**Figure 1**) by a number of reductases, including the reductase domain of endothelial nitric oxide synthase (Fogli et al., 2004), eventually results in the formation of superoxide anion (O2 •−) and hydrogen peroxide (H2O2) that may cause oxidative stress and cell injury in cardiomyocytes (Minotti et al., 2004b; Berthiaume andWallace, 2007; Octavia et al., 2012; Stˇerba et al., 2013; Carvalho et al., 2014). In this context, the role of ROS-generating mitochondrial quinone reductases appears particularly relevant, as also indicated by the abundance of mitochondria in the heart. In addition to biochemical evidence, the role of ROS produced by the reductases was underlined by studies in genetically modified mice showing a relation between cardiotoxicity and either deletion of enzymes involved in anthracycline metabolism or overexpression of antioxidant proteins (reviewed in Minotti et al., 2004a). Interestingly, a high-risk variant of a gene involved in free radical generation significantly increased the susceptibility to DOX-dependent CHF in survivors of hematopoietic cell transplantation (Armenian et al., 2013).

Other pathways of DOX metabolism are also highly relevant for cardiotoxicity. Reduction of the carbonyl group in ringA to alcohol (**Figure 1**) changes DOX into DOXol that, being more polar than DOX, accumulates at higher levels andfor longer times in the heart, thus helping to understand chronic cardiotoxicity. Moreover, it has been shown that DOXol is remarkably more potent than its parent molecule at dysregulating iron homeostasis (see below). The

role of DOXol in cardiotoxicity is reinforced by studies showing correlation between polymorphisms in the genes coding for carbonyl reductases that form DOXol and cardiovascular morbidity (Blanco et al., 2008).

However, in addition to the reductive metabolism described above, DOX can undergo oxidative degradation catalyzed by peroxidases. Recent elegant studies identified H2O2-activated myoglobin as the catalyst of this reaction, and 3-methoxyphthalic acid, the final product of DOX oxidative degradation (**Figure 1**) was detected in hearts of mice exposed to DOX or human heart stripes incubated with DOX (Menna et al., 2010). Importantly, 3 methoxyphthalic acid proved to be much less toxic to cultured cardiomyocytes and mice than equimolar amounts of DOX, possibly because this metabolic pathway prevents quinone-semiquinone redox cycling. These findings suggest that anthracycline oxidative degradation may represent a salvage pathway that diminishes DOX cardiotoxicity (Menna et al., 2010).

## **IRON AND ANTHRACYCLINE CARDIOTOXICITY**

The relative weight of each individual mechanism in the development of anthracycline-induced cardiotoxicity is uncertain, but the role of iron in DOX cardiotoxicity is widely recognized (Minotti et al., 1999; Xu et al., 2005; Simu◦nek et al., 2009).

Perhaps the most compelling proof is provided by the effectiveness of the iron chelator dexrazoxane at preventing cardiotoxicity, both in animal and clinical studies. The use of dexrazoxane in patients has been debated after reports showing myelotoxicity and interference with anticancer effects of DOX, and others

#### **FIGURE 1 | Structure and simplified scheme of molecular**

**transformations of Doxorubicin.** Redox cycling between the quinone and semiquinone forms (ring C) of Doxorubicin (DOX) leads to oxygen radicals formation. The residue involved in DOXol formation following two-electron

reduction of the carbonyl group in ring A is marked in blue. Oxidative pathways involving a hydroquinone-derived semiquinone lead to formation of a diquinone (rings B and C), and eventual degradation of DOX with formation of 3-methoxytphthalic acid as remnant of ring D.

with opposite conclusions (discussed by Stˇerba et al., 2013). Nevertheless, the observation that dexrazoxane protects cancer patients against cardiotoxicity demonstrates that alterations in iron homeostasis are involved in anthracycline-induced CHF.

Additional and complementary experimental evidence for the role of iron in DOX cardiotoxicity was provided by studies showing that iron overload increases the cardiotoxic effects of the drug. Early studies showed that iron loading exacerbated DOX toxicity in cell culture (Hershko et al., 1993; Link et al., 1996). Moreover, increasing iron stores in rats by dietary iron loading enhanced DOX cardiotoxicity (Panjrath et al., 2007), and mice lacking HFE, a model that mimics the iron overload found in human hereditary hemochromatosis, showed higher susceptibility to DOX-dependent cardiac damage (Miranda et al., 2003). Importantly, a recent study showed that DOX-induced myocardial injury was associated with the frequency of HFE gene mutations in survivors of childhood acute lymphoblastic leukemia (Lipshultz et al., 2013b).

## **IRON-DEPENDENT MECHANISMS OF ANTHRACYCLINE CARDIOTOXICITY**

It is considered that by forming anthracycline-iron complexes iron potentiates the toxicity of DOX-derived ROS transforming relatively safe species like O2 •− and H2O2 into the much more reactive and toxic hydroxyl radical OH• or iron-peroxo complexes that damage DNA, proteins, and lipids (Minotti et al., 1999; Xu et al., 2005; Simu◦nek et al., 2009). Moreover, it has been proposed that the redox cycling of the quinone moiety (**Figure 1**) would enable anthracyclines to increase the cellular levels of iron by mobilizing it from ferritin (see below), and thus amplify iron-mediated oxidative stress. Indeed, several lines of evidence in preclinical models indicated that iron and oxygen radicals conspire at inducing cardiotoxicity. However, the so-called "ROS and iron hypothesis" has been called into question; in fact, while iron chelators like dexrazoxane mitigate anthracycline cardiotoxicity, none of several antioxidants proved to be protective against chronic cardiotoxicity in clinical settings (Minotti et al., 2004a; Octavia et al., 2012; Stˇerba et al., 2013).

Accordingly, we showed that DOX doses lower than those usually adopted in experimental studies, but in the range of the plasma levels found in patients undergoing slow infusion chemotherapy induced apoptosis of cardiac-derived H9c2 myocytes in the absence of ROS production, thus supporting the idea that oxidative stress does not play a role in DOX-dependent apoptotic death of cardiomyocytes (Bernuzzi et al., 2009).

We recently provided evidence suggesting an alternative mechanism at the basis of the protective capacity of iron chelators against cardiotoxicity (Spagnuolo et al., 2011). Treatment with dexrazoxane induced hypoxia inducible factor (HIF) in H9c2 cardiomyocytes. Moreover, it prevented the activation of apoptosis caused by treatment of H9c2 cells with low but pharmacologically relevant doses of DOX. The role of HIF was demonstrated by experiments involving pharmacological or genetic manipulation of HIF activity: the inhibition of HIF activity blunted the protective effect of iron chelation. Conversely, overexpressing HIF conferred protection from DOX-dependent cell death also in the absence of dexrazoxane. HIF exerted its defensive effect

by activating the expression of protective proteins, as treatment with the chelator induced the synthesis of anti-apoptotic proteins known to be HIF targets.

These results suggest that the HIF pathway may contribute to the cardioprotective effect of iron chelation. Moreover, these results reinforce the idea that the protection against DOX toxicity provided by the iron chelator may be ROS-independent.

## **EFFECT OF DOXORUBICIN ON PROTEINS OF IRON METABOLISM**

The data reported above, together with evidence discussed below, indicate that the role of iron in DOX cardiotoxicity may be mediated not by DOX–iron interactions, but by profound interference with proteins involved in intracellular iron traffic (**Figure 2**). This concept is in agreement with the essential role of iron in regulating fundamental cellular processes like DNA synthesis, energy production, calcium channel function, redox balance, etc. (Hentze et al., 2004).

## **IRON REGULATORY PROTEINS**

The relationship between iron and anthracycline cardiotoxicity may be related to disruption of cardiac iron homeostasis obtained by means of targeted interaction of DOX with iron regulatory proteins (IRP1 and IRP2), the key regulators of intracellular iron metabolism (Cairo and Pietrangelo, 2000; Rouault, 2006; Cairo and Recalcati, 2007; Muckenthaler et al., 2008). IRPs form the

basis of a homeostatic mechanism that optimizes the metabolic utilization of iron while avoiding its involvement in potentially toxic reactions. In mammalian cells, iron levels are finely controlled by means of opposite but balanced regulation of ferritin, the iron sequestering protein, and transferrin receptor (TfR1), the major iron uptake protein (Cairo and Pietrangelo, 2000). The post-transcriptional control of the cellular levels of ferritin and TfR1 is exerted by IRP-1 and IRP-2, cytosolic proteins that bind to iron regulatory elements (IRE) in ferritin and TfR1 mRNAs (Recalcati et al., 2010). Under conditions of iron scarcity, binding of IRPs to IRE represses the translation of ferritin mRNAs and at the same time enhances the half-life of TfR1 mRNA. This results in increased iron availability for the requirements of the cell. On the contrary, when iron is abundant, IRP activity is decreased, and thereby ferritin mRNAs are translated whereas TfR1 mRNA is degraded; hence, this mechanism facilitates iron storage. It is noteworthy that IRPs can regulate mRNAs for other proteins closely related to iron utilization (mitochondrial aconitase and δ-aminolevulinic acid synthase), uptake (DMT1), and release (ferroportin). Hence, their influence extends over a number of regulatory pathways controlling the iron status of the cell (Cairo and Recalcati, 2007).

IRP1 is a bifunctional protein that is also able to act as an aconitase, thus converting citrate to isocitrate, when it contains a [4Fe–4S] cluster (i.e., in iron-rich cells). On the other hand, when intracellular iron levels are low and the cluster is not formed, IRP-1 becomes an RNA-binding protein. Switching between these two forms therefore represents a mechanism allowing aconitase/IRP1 to adapt iron trafficking to the cell requirements. By providing isocitrate to isocitrate dehydrogenase, IRP-1 may also control the cellular levels of NADH+ H+. IRP2, which is similar to IRP1 but cannot assemble a [4Fe–4S] cluster, is ubiquitinated and degraded by the proteasome in response to iron, whereas it builds up in iron poor cells. IRP-2 degradation also occurs following oxidative modifications induced by reactive oxygen and nitrogen intermediates (Cairo and Pietrangelo, 2000; Rouault, 2006; Cairo and Recalcati, 2007; Muckenthaler et al., 2008).

How DOX affects IRPs ? Pathways linking interactions of IRPs with DOX, DOXol, and quinone-derived ROS, have been characterized, and shown to play a role in cardiotoxicity (Cairo et al., 2002). Observations conducted on cell-free systems or cardiac cell lines have demonstrated that after interaction with DOXol IRP-1 first looses the [4Fe–4S] cluster and then is converted by ROS into a"null"protein lacking both RNA binding and enzymatic activities (Minotti et al., 1998, 2001). Other studies provided data indicating a slightly different mechanism, by which such a "null" protein would be formed by DOX–iron complexes (Kwok and Richardson, 2002). Moreover, ROS-mediated damage targets IRP-2 to proteolysis (Minotti et al., 1998, 2001). The attack of DOX, and in particular of its secondary alcohol metabolite, to IRPs may dysregulate iron homeostasis in the cell not only by disrupting IRP's regulatory function, but also through the release of iron atoms from the Fe–S cluster of IRP1; this, rather than ferritin (see below), may represent a source of iron that facilitates iron-mediated free radical formation and apoptosis. The demonstration that anthracyclines can interact with IRE regions of mRNAs (Canzoneri and

Oyelere, 2008), thus affecting IRP-mediated regulation of several proteins of iron metabolism, has indicated another possible pathway through which DOX can disrupt iron balance.

What could be the pathologic consequences of a DOXdependent inactivation of both IRPs? It should be kept in mind that only mice lacking both IRPs are not viable; this indicates that one IRP is sufficient to maintain iron balance, even though IRP-1-deficient mice did not show any phenotype whereas mice lacking IRP-2 have neurological and hematological abnormalities (Recalcati et al., 2010). In addition to the problems related to the relative importance of each IRP, another complicating factor comes from conflicting results obtained in animal models. Studies in a rat model of chronic cardiotoxicity found no evidence of impaired IRP-1 binding activity following DOX administration (Cusack et al., 2006), but evidence of IRP-1/aconitase sensitivity to DOX was obtained by other authors (Sacco et al., 2003). On the other hand, studies exploiting IRP-1-deficient mice showed that an acute DOX treatment caused cardiac damage irrespective of the presence or absence of IRP-1 (Corna et al., 2006).

Taking into consideration these inconsistencies, one should appreciate the aforesaid pharmacokinetic and metabolic limitations that restrict the validity of the preclinical models of DOX cardiotoxicity. However, it could be concluded that, by allowing ferritin mRNAs translation, an acute loss of IRP-2 might permit free iron sequestration in newly formed ferritin shells before it triggers oxidative damage. This effect would be consistent with the protective effect of ferritin (see below). On the other hand, a chronic loss of IRP-2 is expected to be more injurious, particularly in case IRP-1 is also irreversibly damaged. In fact, the embryonic mortality of IRP-1 and IRP-2 double deficient mice anticipates that complete disruption of the IRP/IRE system is not easily compensated.

## **FERRITIN**

Another possible mechanism linking iron and anthracycline metabolism involves ferritin. While *in vitro* experiments using purified components suggested that the semiquinone free radical and O2 •− produced by redox cycling of DOX quinone moiety caused iron release from ferritin (Thomas and Aust, 1986), the role of ferritin *in vivo* seems quite opposite. In fact, recent data suggest that the increase in ferritin synthesis induced by DOX could be a defensive mechanism to limit the amount of iron available for ROS production in the heart, and thus prevent oxidative injury, in line with the known antioxidant function of this protein (Cairo et al., 1995; Arosio and Levi, 2010). We demonstrated that ferritin is induced in H9c2 cardiomyocytes (Corna et al., 2004) and mouse hearts (Corna et al., 2006) exposed to DOX and protected the cells against iron toxicity. Newly formed ferritin can therefore sequester iron and may paradoxically represent a defense for cardiomyocytes. These findings were recently supported by two studies demonstrating that NF-*k*B-mediated ferritin H chain induction plays an important role in the mitochondrial protection offered by metformin against DOX cardiotoxicity in isolated cardiomyocytes (Asensio-López et al., 2013a,b). Moreover, it has been shown that exposure of cardiomyocytes to DOX leads to higher accumulation of iron into ferritin by mechanisms that

impair iron release from ferritin and/or lysosomal/proteasomal degradation of this iron-storage protein (Kwok and Richardson, 2003, 2004).

As reported above, a number of observations indicate that cardiac mitochondria are preferential targets of anthracyclines. Additional evidence was recently provided by data showing that the anthracycline epirubicin has an intracellular distribution different from that of DOX: it accumulates in acidic organelles (lysosomes, recycling endosomes) and spares the mitochondria (Salvatorelli et al., 2006). This appears to be one of the reasons for the reduced cardiotoxicity of this analog.

In this context, recent results point to a critical role for mechanisms controlling mitochondrial iron availability in anthracycline cardiotoxicity. In particular, it has been recently shown that exposure to DOX induces iron accumulation specifically in mitochondria, and mice with heart-specific deletion of ABCB8, which is involved in iron export out of the mitochondria, were more sensitive to DOX cardiotoxicity, whereas the opposite effect was obtained withABCB8 overexpression (Ichikawa et al.,2014). Mitochondrial ferritin (FtMt), which is a recently identified ferritin type that accumulates specifically in the mitochondria where readily accumulates iron (Arosio and Levi, 2010), may be another key player. FtMt expression, which is not controlled by intracellular iron levels, is high in mitochondria-rich tissues, such as the testis, brain, heart. Despite the lack of evident phenotype in FtMt-deficient mice (Bartnikas et al., 2010), studies in cultured cells overexpressing FtMt showed that, by sequestering iron inside the mitochondria, FtMt is able to reduce iron-mediated oxidative damage (Campanella et al., 2009). Moreover, it has been recently demonstrated that ectopic expression of FtMt in HeLa cells was able to reduce DOX-mediated cytotoxicity (Cocco et al., 2013). Although this effect was not obtained in cardiac cells, these results strongly suggested that FtMt may play a relevant role in DOX cardiotoxicity. Indeed, FtMt induction was recently found in the heart of DOX-treated mice (Ichikawa et al., 2014).

Given the role of mitochondria in cellular iron homeostasis, their importance as producers and targets of ROS, and their interaction with anthracyclines, further investigation of the effects of DOX on mitochondrial iron metabolism is required to increase our knowledge of DOX cardiotoxicity.

#### **CONCLUSIONS AND PERSPECTIVES**

Understanding the molecular basis of cardiotoxicity is important to improve the therapeutic use of anthracyclines, which still remain among the most effective antitumor drugs, and to identify less cardiotoxic analogs. In this context, the role of iron in the pathogenesis of cardiac damage is undisputed, but the precise molecular mechanisms underlying its effects remain incompletely understood (**Figure 2**). While early studies pointed to a role as amplifier of the free radical generation initiated by DOX redox cycling, more recent evidence suggests that the cardiotoxic role of iron should not be restricted to the oxidative stress scenario. The fact that anthracycline cardiotoxicity is mitigated by iron chelators but not by antioxidants, together with biochemical evidence showing that proteins important for cellular iron homeostasis are specifically targeted by DOX,indicates that iron and anthracyclines

can also induce oxidant-independent cell injury. Exploring more deeply the role of iron in the pathogenic mechanisms of DOX cardiotoxicity will reveal useful insights for the development of improved therapeutic strategies against anthracycline-dependent heart damage and CHF.

## **ACKNOWLEDGMENTS**

This work was supported by grants from Ministero dell'Istruzione, dell'Università e della Ricerca, and Ministero della Salute.

## **REFERENCES**


induced by doxorubicin in HeLa cells. *Mol. Biol. Rep.* 40, 6757–6764. doi: 10.1007/s11033-013-2792-z


evidence for distinct metabolic pathways and implications for iron-mediated cardiotoxicity of antitumor therapy. *Cancer Res.* 61, 8422–8428.


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

*Received: 23 January 2014; paper pending published: 29 January 2014; accepted: 12 February 2014; published online: 26 February 2014.*

*Citation: Gammella E, Maccarinelli F, Buratti P, Recalcati S and Cairo G (2014) The role of iron in anthracycline cardiotoxicity. Front. Pharmacol. 5:25. doi: 10.3389/fphar.2014.00025*

*This article was submitted to Drug Metabolism and Transport, a section of the journal Frontiers in Pharmacology.*

*Copyright © 2014 Gammella, Maccarinelli, Buratti, Recalcati and Cairo. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*

## Epidemiological associations between iron and cardiovascular disease and diabetes

## *Debargha Basuli 1, Richard G. Stevens <sup>2</sup> , Frank M. Torti <sup>3</sup> and Suzy V. Torti 1\**

<sup>1</sup> Molecular Biology and Biophysicis, University of Connecticut Health Center, Farmington, CT, USA

<sup>2</sup> Division of Epidemiology and Biostatistics, Department of Community Medicine and Health Care, University of Connecticut Health Center, Farmington,

CT, USA <sup>3</sup> Internal Medicine, University of Connecticut Health Center, Farmington, CT, USA

#### *Edited by:*

Paolo Arosio, University of Brescia, Italy

#### *Reviewed by:*

Luca Valenti, Università degli Studi di Milano, Italy Igor Theurl, Medical University of Innsbruck, Austria

#### *\*Correspondence:*

Suzy V. Torti, Molecular Biology and Biophysicis, University of Connecticut Health Center, 263 Farmington Avenue, Farmington, CT 06030, USA e-mail: storti@uchc.edu

Disruptions in iron homeostasis are linked to a broad spectrum of chronic conditions including cardiovascular, malignant, metabolic, and neurodegenerative disease. Evidence supporting this contention derives from a variety of analytical approaches, ranging from molecular to population-based studies. This review focuses on key epidemiological studies that assess the relationship between body iron status and chronic diseases, with particular emphasis on atherosclerosis ,metabolic syndrome and diabetes. Multiple surrogates have been used to measure body iron status, including serum ferritin, transferrin saturation, serum iron, and dietary iron intake. The lack of a uniform and standardized means of assessing body iron status has limited the precision of epidemiological associations. Intervention studies using depletion of iron to alter risk have been conducted. Genetic and molecular techniques have helped to explicate the biochemistry of iron metabolism at the molecular level. Plausible explanations for how iron contributes to the pathogenesis of these chronic diseases are beginning to be elucidated. Most evidence supports the hypothesis that excess iron contributes to chronic disease by fostering excess production of free radicals. Overall, epidemiological studies, reinforced by basic science experiments, provide a strong line of evidence supporting the association between iron and elevated risk of cardiovascular disease and diabetes. In this narrative review we attempt to condense the information from existing literature on this topic.

**Keywords: iron, cardiovascular disease, diabetes mellitus, metabolic syndrome, epidemiologic studies**

## **INTRODUCTION**

Cardiovascular disease (CVD) and diabetes are major health problems worldwide. In the United States, approximately one in four deaths are due to heart disease, making it the leading cause of death for both men and women (Heron et al., 2009; Heidenreich et al., 2011)1. Coronary heart disease is the most common type of heart disease, and costs the US over 100 billion each year (Go et al., 2013). Risk factors include high blood pressure, high LDL cholesterol, smoking, diabetes, and obesity (MMWR Morb Mortal Wkly Rep 60, 2001). Diabetes is itself a significant health problem that is reaching epidemic proportions, with a global prevalence of 382 million people in 2013. It is estimated that by 2035 this will rise to 592 million2.

The search for risk factors and methods of prevention for both CVD and diabetes are major efforts of the medical and research community. The role of iron as a risk factor for CVD and diabetes has drawn attention in part due to the concept that it may be a risk factor susceptible to simple dietary modification. Although this is an oversimplification, many (not all) reports suggest that there is indeed an association between iron and both CVD and diabetes, as detailed in this review.

Epidemiological studies have been a powerful tool to probe the association between iron and CVD and diabetes. Several types of study design have been employed. Each of these has its benefits and limitations. Briefly, epidemiological studies can be divided into observational and experimental studies. The difference between an observation and experimental study is that in the latter, an outcome is studied in a population in the absence or presence of an intervention by the investigator. In an observational study, there is no intervention and the investigator simply "observes" and analyses the relationship between exposure and disease outcome. Observational studies include cohort studies, case-control studies and cross sectional studies. A cohort study is an analysis of risk factors where a disease-free study population is identified and followed prospectively over time and a subsequent evaluation is done to find the association between the exposure and disease outcome. While this kind of study can provide strong scientific evidence of an association between risk factors and disease and a temporal framework to assess causality, it is limited by the requirement for a large sample size and long follow-up duration. Often several biases can adulterate the evidence. A case control study on the other hand starts with groups with and without an outcome and evaluates how much a suspected exposure might have contributed to the present outcome status. Thus in comparison to cohort studies, case control studies are

<sup>1</sup>http://www.cdc.gov/heartdisease/facts.htm (accessed February 2, 2014). 2http://www.idf.org/diabetesatlas

relatively quicker to conduct, inexpensive and require comparatively fewer subjects. Cross sectional studies collect and analyze the data on exposure and disease at one specific time point. Such studies cannot evaluate cause and effect relationships since there is no temporal assessment. **Table 1** shows the level of evidence of different types of epidemiological studies. In the hierarchy of evidence-based medicine, experimental studies (or more specifically randomized controlled trials) are recognized as level I of scientific evidence. However, the consensus over this has recently changed as observational studies have been reported to be as effective as randomized controlled trials in estimating the impact of medical interventions on disease outcomes (Benson and Hartz, 2000; Concato et al., 2000). Of course, it must also be emphasized that for a potential hazard, such as elevated body iron, a randomized controlled trials cannot be performed ethically, although it could be done for a study of iron reduction by, for example, phlebotomy.

Virtually all of these epidemiological analyses have been used to probe the relationship between iron and CVD or diabetes. PubMed searches using the terms "iron heart disease epidemiology" or "iron diabetes epidemiology" identify over 500 papers for each search term. In this narrative review, we have not attempted to be comprehensive, but to focus on key epidemiological studies that have investigated these issues. We provide some historical context, but emphasize recent, well-controlled studies with large sample size.

## **IRON AND CARDIOVASCULAR DISEASE**

Cardiovascular disease is a broad term that includes ischemic and non-ischemic irregularities. Association with iron has been mainly studied and found in ischemic cardiovascular diseases caused by atherosclerosis. To measure ischemic disease outcome, several different endpoints have been used, including coronary heart disease (CHD), carotid artery plaque formation, coronary artery calcium deposition, carotid intima thickness, and atherosclerosis. CHD has been measured by myocardial infarction and cardiogenic angina occurrences and deaths from such incidents. The role of iron in CVD has generally been explored in a group of individuals using one of these defined endpoints. For the purposes of this review, we have included all of these clinical entities under the umbrella of CVD and have not attempted to differentiate among them.

The most common measurement used in the assessment of body iron has been serum ferritin. Serum ferritin was shown to correlate with body iron stores in the 1970s, and is still used clinically for this purpose (Jacobs et al., 1972; Cook et al., 1974; Jacobs and Worwood, 1975; Wang et al., 2010). However, serum ferritin can also be elevated by acute and chronic inflammation (Wang et al., 2010). Accounting for the contribution of these variables thus becomes an important component of studies that use serum ferritin as a measure of body iron, as discussed below. Less frequently, the ratio of soluble transferrin receptor to ferritin has been used, as it has been suggested that this is a more precise measure for body iron store than ferritin alone (Skikne et al., 1990). Catalytically available iron has also been measured in some studies, with the goal of measuring reactive rather than total iron. This approach derives from the consideration that the preponderance of iron in the body is bound to proteins and is not available for participation in the potentially deleterious reactions that are thought to underlie much of the toxicity of iron, such as the formation of reactive oxygen species. A limitation of this approach is that since catalytically available iron represents a relatively small fraction of total iron, its measurement is technically challenging.

The hypothesis that iron status could influence the risk of coronary heart disease was first proposed by Sullivan in the 1980s. Sullivan hypothesized that the higher occurrence of CHD in men and post-menopausal women than in pre-menopausal women was due to higher iron stores in them compared to menstruating women (Sullivan, 1981, 1989). Some earlier studies supported the hypothesis. In a cohort of 2873 Framingham women, an increase in incidence of CHD and disease severity was observed in women who had either natural or surgical menopause (Gordon et al., 1978). In some early prospective studies, a weak association between high blood hemoglobin and hematocrit and risk of CHD was noted (Cullen et al., 1981; Bottiger and Carlson, 1982; Knottnerus et al., 1988). Hemoglobin and hematocrit are not good surrogates for body iron status and during this period serum ferritin was emerging as the best measurement of body iron status (Cook et al., 1974; Kaltwasser and Werner, 1989). The first report in humans on the association between serum ferritin and CHD risk was published in Salonen et al. (1992a). In this cohort of randomly selected 1931 Eastern Finnish men, serum ferritin concentration had a significant association with ischemic heart disease risk. Subjects with serum ferritin ≥200 μg/l had a 2.2-fold (95% CI, 1.2–4.0; *p* < 0.01) higher risk of acute myocardial infarction compared to men with lower serum ferritin. Total blood leucocyte count was adjusted in the statistical analysis to rule out the potential confounding effect of inflammation or chronic


**Table 1 | Evidence level provided by epidemiological studies of different design.**

Song and Chung (2010).

vascular disease that would elevate serum ferritin independent of body iron status. The association was stronger in men with higher concentrations of low density lipoproteins (RR = 1.8, 95% CI, 0.9–3.5, NS in men with low LDL and a RR = 4.7, 95% CI, 1.4–16.3, *p* < 0.05 in men with high LDL). After this report, the group conducted another nested case-control study within the Kuopio Ischemic Heart Disease Risk Factor Study (KIHD) cohort and found that men with high body iron stores were at increased risk of acute myocardial infarction (AMI), confirming their original observation (Tuomainen et al., 1998). In this study body iron status was measured by ratio of soluble TfR and ferritin which some authors suggest as a better measure of body iron than serum ferritin alone (Cook et al., 1974; Skikne et al., 1990).

The first prospective study in women was conducted in 11,471 Dutch post-menopausal female subjects aged 49–70 years (van der et al., 2005). In the study, the multivariate hazard ratio of ischemic strokes in the highest tertile of serum ferritin concentration was 2.23(95% CI, 1.05–4.73) compared to the lowest. An interesting finding common to some of these studies was the interaction of LDL with serum ferritin in increasing the risk of ischemic events. A plausible biological mechanism underlying this interaction may be the ability of iron to produce reactive oxygen species. Iron catalyzes the Fenton reaction which produces potent oxidants that increase the risk of atherosclerosis by promoting the peroxidation of lipids (McCord, 1991; Salonen et al., 1992b; Berliner and Heinecke, 1996). Local release of iron from ferritin by superoxide radical generated by ischemia/reperfusion injury to blood vessels may further exacerbate this damage (Thomas et al., 1985).

Many of the studies discussed above focused on cardiovascular events such as acute myocardial infarction. However, myocardial infarction is a complex endpoint resulting from multiple potential pathogenesis pathways. To circumvent this limitation, other studies used preclinical atherosclerosis as the dependent variable and explored its relationship to serum ferritin. For example, Kiechl et al. (1997) reported that serum ferritin level was closely related to incidence of carotid atherosclerosis and progression of previous atherosclerotic lesions in a cohort of Italian men and women. In a cross sectional study that included German men and women, there was an association of serum ferritin with carotid plaque prevalence in both men (OR: 1.33; 95% CI, 1.08–1.44) and women (OR, 1.29; 95% CI, 0.98–1.75) (Wolff et al., 2004). When the study population was divided into ferritin octiles, both men and women showed a dose-dependent relationship between serum ferritin and atherosclerotic plaques. Subjects with malignancy and liver diseases were excluded to eliminate the confounding effect of inflammation and mild liver disease, but no adjustments were made for any inflammatory markers. Thus, acute or chronic inflammatory conditions could have confounded the findings in this study by affecting serum ferritin levels at the time of measurement.

Several recent studies have shown that serum ferritin is independently associated with preclinical measures of vascular diseases. Sung et al. (2012)showed that ferritin levels in a large cohort of 12,033 young Korean men were independently associated with coronary artery calcium content, a marker of early coronary artery sclerosis. In a similar study, Valenti et al. (2011) showed that in a small study population of non-alcoholicfatty liver patients, carotid intima medial thickness and carotid plaque were independently associated with increased ferritin levels.

The potentially damaging effect of iron on the heart, pancreas, liver and other organs was made evident in part through the study of hemochromatosis, a disorder in which excess iron is absorbed and deposited in tissues. Patients with untreated hemochromatosis can exhibit diabetes, liver damage, and cardiac injury among other symptoms (Wolff, 1993; Witte et al., 1996; Powell et al., 1998). Mutations in HFE gene (the hemochromatosis gene) are one cause of hemochromatosis. Using a mouse model for hereditary hemochromatosis, Turoczi et al. (2003)showed an interaction of dietary iron intake and HFE gene status (KO vs. wildtype) in degree of ischemia/reperfusion injury to heart; HFE KO mice showed greater ventricular dysfunction, myocardial infarct size, and cardiomyocyte apoptosis compared to wild type mice on a standard diet, and an even greater degree of damage in the HFE KO mice fed a high iron diet (Turoczi et al., 2003). In human subjects, a similar increase in cardiovascular death was observed in women heterozygous for the HFE gene (Roest et al., 1999). However, no association between the HFE genotype and atherosclerosis has been found in hemochromatosis patients in spite of iron overload status in these patients (van der et al., 2006; Engberink et al., 2010). Valenti et al. (2011) found that the prevalence of carotid plaques was highest in patients with hyperferritinemia independent of HFE genotype. Although the risk of atherosclerotic heart disease appears unrelated to HFE genotype, hemochromatosis patients do have a higher risk of iron-related non-ischemic cardiovascular irregularities (Gaenzer et al., 2002; Dunn et al., 2008).

Multiple mechanisms likely underlie the association of iron with CVD. In addition to the ability of iron to promote lipid peroxidation, recent studies have implicated the peptide hormone hepcidin in atherosclerosis. Hepcidin is a central regulator of iron absorption and recycling (Ganz, 2013). Hepcidin acts by binding ferroportin, an iron efflux pump present in both enterocytes and macrophages. Binding of hepcidin to ferroportin triggers ferroportin degradation (Nemeth et al., 2004), thus inhibiting the delivery of iron to the circulation through the enterocyte as well as inhibiting iron recycling in macrophages. Valenti et al. (2011) observed that serum hepcidin was independently associated with carotid plaques, suggesting that hepcidin-induced iron accumulation may be involved in the process of atherogenesis in subjects negative for HFE mutations. Specifically, hepcidin may induce excessive iron trapping within macrophages, resulting in an increase in oxidative stress, transformation into foam cells, and ultimately to atherosclerotic vascular disease (Sullivan, 2007, 2009; Theurl et al., 2008). This hypothesis is known as the "iron hypothesis" and was proposed by Sullivan (2009). Since hepcidin is decreased in patients with hereditary hemochromatosis, this model provides a potential explanation for the previous observation that atherosclerosis risk is not increased in subjects with hereditary hemochromatosis (van der et al., 2006). However, a recent experimental study provided evidence that hepatic hepcidin expression is not correlated with atherosclerosis progression in a mouse model Kautz et al. (2013). Further, the authors reported that increasing macrophage iron accumulation in mice

with atherosclerosis either through a genetic mutation in the ferroportin gene or through parenteral iron administration failed to increase the size of atherosclerotic lesions or lesion calcification. The study challenged the "iron hypothesis."

An alternative approach to the use of serum ferritin to assess the relationship between iron and CVD has been to assess the relationship between catalytic iron and heart disease. Catalytic iron is the iron that is not bound to transferrin or ferritin and is available to take part in chemical reactions to produce oxidant products. This can be measured using a bleomycin detectable iron assay (BDI; von Bonsdorff et al., 2002). Results from such studies are equivocal. While some studies found a relation between catalytic iron with CVDs (Rajapurkar et al., 2012), a recent study with a larger study population failed to report any association of catalytic iron with risk of MI and recurrent ischemic events (Steen et al., 2013). However the study showed that in a cohort of 1701 patients with AMI or unstable angina, catalytic iron was associated with stepwise increase in all-cause mortality (multivariate adjusted HR = 3.97, 95% CI 1.09–14.1, *p* = 0.036, highest quartile vs. baseline) when followed for a median of 10 months. Although no significant association of ischemic events with iron was reported, most of the deaths were related to ischemic complications, and thus the contribution of catalytic iron could not be ruled out completely. It is to be noted that serum catalytic iron does not reflect the intraplaque iron which might be a more proximal factor for vascular ischemic injuries (Nelson et al., 1992; Castellanos et al., 2002). Unfortunately, these studies did not report the relationship between serum ferritin and outcome. It thus remains unaddressed whether or not catalytic iron is more strongly associated than serum ferritin with outcome.

Another approach to testing a potential link between iron and CVDs has been the study of dietary iron intake and risk. Zhang et al. (2012) reported that dietary intake of total iron was positively associated with deaths from strokes and CVD in a cohort of 23,083 Asian men with a multivariate hazard ratio of 1.43 (95% CI, 1.02–2, *p* = 0.009) for stroke and 1.27 (95% CI, 1.01–1.58, *p* = 0.023) for CVD after adjustment for other CVD risk factors; iron intake was not associated with these outcomes in women, however. In another large prospective study in almost 50,000 European men, there was a positive association of dietary iron, more specifically heme iron (a form of iron that is more readily absorbed by the gut than inorganic iron), with strokes in a follow-up period of 11.7 years (Kaluza et al., 2013). Adjustments were made for red meat consumption to rule out confounding by other known risk factors for stroke such as *N*-nitroso compounds and heterocyclic compounds (Forstermann, 2008). The association was observed in normal weight men and not in overweight or obese men, most likely because of decreased iron absorption due to increased hepcidin in the chronic inflammatory state associated with obesity (Greenberg and Obin, 2006). Adipocyte hepcidin expression is known to be positively correlated with obesity (Bekri et al., 2006). This association is still to be properly evaluated in women.

Although the foregoing studies appear well-conducted, there are also a number of other well-conducted studies that have found no association of markers of body iron and risk of CVD (Baer et al., 1994; Danesh and Appleby, 1999; Gupta et al., 2000; Knuiman et al., 2003; Sun et al., 2008b; Friedrich et al., 2009). Discordant results among studies may in part be due to imprecision in the surrogate markers used to measure iron status (ferritin, total iron binding capacity, transferrin saturation, serum iron, and dietary iron intake), which are all indirect measures of body iron stores. Because these variables are subject to temporal and measurement variations, there is undoubtedly exposure misclassification in the subjects. This non-differential misclassification would reduce the ability of a study to identify a true association should one exist. An additional potential problem, particularly in cross-sectional studies, is that observed elevations in serum ferritin may represent an effect rather than a cause of underlying pathology.

If iron is associated with CVD, can interventions that reduce iron reduce risk? Unfortunately, only a limited number of intervention studies have been conducted, but results of these studies are at least suggestive that modulating iron can reduce risk. In animal models, treatment with deferoxamine, an iron chelator, during ischemia improved recovery and reduced reperfusion-induced oxygen radical formation in rabbit hearts (Williams et al., 1991). Paraskevaidis et al. (2005) reported that deferoxamine infusion ameliorated lipid peroxidation and improved long term outcome in patients having coronary artery bypass surgery. In a recent randomized controlled single blinded study, Zacharski et al. (2011) showed that a lower ferritin level predicted improved outcome and iron reduction by phlebotomy improved outcomes by preventing or delaying non-fatal myocardial infarction and stroke in young age patients with peripheral artery disease.

## **IRON AND METABOLIC SYNDROME AND DIABETES**

Metabolic syndrome refers to a collection of risk factors that increase the likelihood of heart disease, diabetes and stroke. They include a large waistline, high triglyceride, low HDL cholesterol, high blood pressure and high fasting blood sugar. The presence of three of these five risk factors, many of which are associated with obesity, is defined as metabolic syndrome3.

Multiple studies have shown that excess body iron is associated with one or more components of metabolic syndrome (Jehn et al., 2004; Bozzini et al., 2005; Choi et al., 2005; Gonzalez et al., 2006; Sun et al., 2008a). To study the association of iron with metabolic syndrome in normal individuals, a cross-sectional study in 6044 US adults was conducted. The results showed a significant association of ferritin level with metabolic syndrome and insulin resistance (IR) after excluding hemochromatosis cases and adjusting for age, race/ethnicity, C-reactive protein, smoking, alcohol intake, and BMI (Jehn et al., 2004). Other studies in western populations showed similar associations (Jehn et al., 2004; Bozzini et al., 2005; Gonzalez et al., 2006; Wrede et al., 2006). A positive association of serum ferritin with the prevalence of metabolic syndrome in a study population of 8441 people including both sexes and from different provinces of China was recently reported (Li et al., 2013). Mendler et al. (1999) showed that in a cohort of patients with unexplained hepatic iron overload, IR was also often observed. Such patients with non-alcoholic fatty liver disease also tend to have elevated ferritin levels (Valenti et al., 2003,

<sup>3</sup>http://www.nhlbi.nih.gov/health/health-topics/topics/ms/

2006, 2007, 2010) which is now considered a feature of metabolic syndrome (Marchesini et al., 2001, 2003; Angulo, 2002). The constellation of hepatic steatosis, mild to moderate iron overload in both hepatocytes and macrophages, increased serum ferritin levels, and insulin resistance is commonly referred to as dysmetabolic iron overload syndrome or DIOS (Barisani et al., 2008; Riva et al., 2008; Datz et al., 2013). DIOS is detected in about one third of the patients with NAFLD and MetS and may be predisposing factor to the development of type 2 DM and CVDs [Dongiovanni et al., 2011; Valenti et al., 2012; See Dongiovanni et al. (2011) and Datz et al. (2013) for detailed discussion and review of this topic]. Iron depletion in such patients has been shown to improve histological liver damage and abnormal liver function when compared to lifestyle modification (Valenti et al., 2014).

Similarly, many epidemiological studies have reported statistically significant associations of body iron with diabetes, although results from all studies are not entirely consistent. In one longitudinal study of overweight/obese individuals with an impaired glucose tolerance test, there was no association between ferritin levels with risk of diabetes (Rajpathak et al., 2009). In another similar study, adjustment for BMI and components of the metabolic syndrome produced a null result (Jehn et al., 2007). However, in a prospective study in China, a country with the largest diabetic population in the world, an almost twofold increased risk of type 2 diabetes was observed among middle aged and elderly persons in the highest quintile of ferritin level compared with those in lowest after adjusting for known risk factors including high sensitivity C reactive protein (hsCRP), BMI, γ-glutamyl transferase (GGT), and adiponectin (Sun et al., 2013). The result was consistent with findings from similar prospective cohort studies in Caucasian populations (Salonen et al., 1998; Jiang et al., 2004; Forouhi et al., 2007; Montonen et al., 2012). In the EPIC (European Prospective Investigation of Cancer)-Norfolk cohort, serum ferritin was an important and independent predicting factor for development of diabetes after adjustment for conventional risk factors as well as vitamin C levels, CRP, IL6, liver function test (ALT, GGT), fibrinogen and adiponectin (Forouhi et al., 2007). In a case cohort study among 27,548 participants of the EPIC Postdam study in Germany, the sTFR-ferritin ratio was significantly inversely related to the risk of type 2 diabetes, and ferritin concentration was associated with higher risk (Montonen et al., 2012). The result was independent of biomarkers of inflammation, hepatic fat, IR, and dyslipidemia.

In evaluating these studies, a few considerations must be borne in mind. In epidemiological studies investigating the relation between iron and diabetes, serum ferritin is the most commonly used indicator of body iron stores. As mentioned in the introduction, the use of ferritin in assessing body iron stores has been somewhat challenging because ferritin can be elevated in inflammation, cancer, and liver disease (Wang et al., 2010). Serum ferritin concentration can also be increased in some conditions like obesity and metabolic syndrome which are associated risk factors for type 2 diabetes (Lee et al., 2009). It thus becomes difficult to discern whether the association of ferritin with diabetes is due to other concomitant conditions or serum ferritin levels increase as a result of diabetes (a case of reverse causation). In

addition, serum ferritin has been correlated with dyslipidemia biomarkers (Halle et al., 1997), hepatic enzymes (Choi et al., 2005), and negatively associated with adiponectin, an insulin sensitizing adipokine that is decreased in diabetic patients (Forouhi et al., 2007). Thus adjustments for these components become very important.

The exact molecular mechanism of iron-related pathology in metabolic syndrome and diabetes is not clearly understood. Iron is a powerful pro-oxidant and can cause cellular damage by producing reactive oxygen species in different tissues of the body (Andrews, 1999; Rajpathak et al., 2009). Insulin producing pancreatic β cells have been shown to be particularly susceptible to oxidative injury, in part due to decreased expression of antioxidant enzymes such as dismutase, catalase, and glutathione peroxidase (Tiedge et al., 1997). Thus iron deposition in β cells can lead to apoptosis and consequently to decreased insulin synthesis and secretion (Tiedge et al., 1997; Ferrannini, 2000; Cooksey et al., 2004). A recent study in a mouse model of iron overload showed that iron deposition enhances fatty acid oxidation and decreases glucose oxidation in skeletal muscle by inhibiting pyruvate dehydrogenase (PDH) enzyme activity thus increasing IR (Huang et al., 2011). Glucose oxidation is decreased in adipose tissue (Merkel et al., 1988; Green et al., 2006). Iron accumulation also results in an abnormal increase in hepatic glucose production (Mendler et al., 1999; Ferrannini, 2000; Green et al., 2006), inappropriate hepatic insulin extraction, and affects insulin secretion in the pancreas (Niederau et al., 1984). A recent study of 492 subjects demonstrated an association between markers of iron metabolism, adipocyte insulin resistance, and adiponectin (an insulin-sensitizing adipokine), consistent with a model in which iron contributes to T2DM by inducing insulin resistance in adipocytes (Wlazlo et al., 2013). Consistent with this model, mice fed a high iron diet exhibited an accumulation of iron within adipocytes and altered transcription of adipokines involved in glycemic control (Gabrielsen et al., 2012; Dongiovanni et al., 2013). In particular, iron downregulated adiponectin (an adipokine with insulin sensitizing action; Gabrielsen et al., 2012) and increased resistin (an adipokine with hyperglycemic action; Dongiovanni et al., 2013).

The association of dietary iron and diabetes has also been examined. Iron in the diet exists as heme (organic) and nonheme (inorganic) forms. Some studies have shown that the risk of diabetes can be increased by heme iron in the diet (Jiang et al., 2004; Lee et al., 2004; Rajpathak et al., 2006). Most of these studies were conducted in a healthy US population. In Asian populations, a similar association was reported in a cross sectional study of 2997 Chinese people (Luan de et al., 2008). Consistent results were obtained in an observational cohort of Mediterranean people (Fernandez-Cao et al., 2013). Two recent meta-analyses concluded that higher heme iron poses higher risk of type 2 diabetes (Bao et al., 2012; Zhao et al., 2012). One of the studies reported that there was no significant association with total iron, non-heme iron or iron supplements in the diet (Bao et al., 2012). However, these studies did not separate heme iron *per se* from other components of red meat. Red meat bears a high correlation with heme iron and has been shown to be associated with the risk of

type 2 diabetes (Pan et al., 2011). Therefore confounding by other components from red meat cannot be ruled out with certainty. Screening for HFE mutation was not conducted in these studies, and hence a contribution of genetic interaction cannot be ruled out either.

Despite the limitations of epidemiological studies, intervention studies support the association between excess iron, metabolic syndrome and diabetes. Iron reduction by phlebotomy and chelation therapy produced an improvement in glucose tolerance not only in patients with hemochromatosis (Dymock et al., 1972; Inoue et al., 1997), but also in healthy donors. Houschyar et al. (2012)reported that reduction of body iron stores through phlebotomy had therapeutic effects in metabolic syndrome patients, including lowered blood pressure and improvement in glycemic control and cardiovascular risks. Fernandez-Real et al. (2002)found that bloodletting in high ferritin type 2 diabetes patients resulted in decrease in glycated hemoglobin and also improved insulin secretion and sensitivity.

#### **CONCLUSION**

Epidemiological studies provide evidence that elevated iron stores are a risk factor for developing cardiovascular and metabolic abnormalities. Such results have been verified in diverse ethnic and geographic populations. Although mechanistic insights have been limited, iron-dependent pathophysiological pathways involved in these two conditions may exhibit some differences. In diabetes and metabolic syndrome, iron may contribute to risk following deposition in the liver, pancreas, and skeletal muscle, where it can enhance oxidative damage and contribute to insulin deficiency and resistance. In CVD, iron within macrophages and foam cells predisposes to theformation of atherosclerotic plaques. Hepcidin may promote plaque destabilization by preventing iron export from the intralesional macrophages leading to ischemic events. Although additional mechanisms are likely involved, **Figures 1** and **2** illustrate some pathways through which excess iron can increase risk of CVD, metabolic syndrome, and diabetes.

Regarding iron in the diet, there is still insufficient data to formulate guidelines on dietary iron restrictions in the at-risk or general population. This is primarily because dietary iron exists in two very different forms – heme and non-heme iron. Study findings are more inclined toward the association of heme iron (mainly from meat) and disease risk rather than non-heme iron. Although some studies have attempted to assess whether iron supplementation is linked to disease risk, particularly diabetes in women, the results have been inconsistent (Rajpathak et al., 2006; Bo et al., 2009; Chan et al., 2009).

Further research is required to identify more predictors of body iron stores that may help in reducing the risk of cardiovascular or metabolic disease. Experiments are needed to unravel the underlying biological mechanism of this association. Additionally, more randomized controlled studies are warranted to evaluate the clinical outcome of patients placed on iron restricted diets or subjected to iron depletion therapy so that therapeutic recommendations can be made.

#### **ACKNOWLEDGMENT**

Supported in part by NIH R01 CA171101 (Frank M. Torti).

#### **REFERENCES**


without evidence for coronary artery disease. *Atherosclerosis* 128, 235–240. doi: 10.1016/S0021-9150(96)05994-1


adipocyte insulin resistance and plasma adiponectin: the Cohort on Diabetes and Atherosclerosis Maastricht (CODAM) study. *Diabetes Care* 36, 309–315. doi: 10.2337/dc12-0505


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

*Received: 10 March 2014; accepted: 30 April 2014; published online: 20 May 2014. Citation: Basuli D, Stevens RG, Torti FM and Torti SV (2014) Epidemiological associations between iron and cardiovascular disease and diabetes. Front. Pharmacol. 5:117. doi: 10.3389/fphar.2014.00117*

*This article was submitted to Drug Metabolism and Transport, a section of the journal Frontiers in Pharmacology.*

*Copyright © 2014 Basuli, Stevens, Torti and Torti. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited andthatthe original publication inthis journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*

## Atherogenesis and iron: from epidemiology to cellular level

## *Francesca Vinchi 1,2, Martina U. Muckenthaler 1,2, Milene C. Da Silva1,2, György Balla3,4, József Balla5 and Viktória Jeney3,5\**

*<sup>1</sup> Department of Pediatric Oncology, Hematology and Immunology, University of Heidelberg, Heidelberg, Germany*

*<sup>2</sup> Molecular Medicine and Partnership Unit, University of Heidelberg, Heidelberg, Germany*

*<sup>3</sup> MTA-DE Vascular Biology, Thrombosis and Hemostasis Research Group, Hungarian Academy of Sciences, Debrecen, Hungary*

*<sup>4</sup> Department of Pediatrics, University of Debrecen, Debrecen, Hungary*

*<sup>5</sup> Department of Medicine, University of Debrecen, Debrecen, Hungary*

#### *Edited by:*

*Raffaella Gozzelino, Instituto Gulbenkian de Ciência, Portugal*

#### *Reviewed by:*

*Andrei Adrian Tica, University of Medicine Craiova Romania, Romania Gregory M. Vercellotti, University of Minnesota, USA*

#### *\*Correspondence:*

*Viktória Jeney, Department of Medicine, University of Debrecen, Nagyerdei krt. 98, Debrecen 4012, Hungary e-mail: jeneyv@ internal.med.unideb.hu*

Iron accumulates in human atherosclerotic lesions but whether it is a cause or simply a downstream consequence of the atheroma formation has been an open question for decades. According to the so called "iron hypothesis," iron is believed to be detrimental for the cardiovascular system, thus promoting atherosclerosis development and progression. Iron, in its catalytically active form, can participate in the generation of reactive oxygen species and induce lipid-peroxidation, triggering endothelial activation, smooth muscle cell proliferation and macrophage activation; all of these processes are considered to be proatherogenic. On the other hand, the observation that hemochromatotic patients, affected by life-long iron overload, do not show any increased incidence of atherosclerosis is perceived as the most convincing evidence against the "iron hypothesis." Epidemiological studies and data from animal models provided conflicting evidences about the role of iron in atherogenesis. Therefore, more careful studies are needed in which issues like the source and the compartmentalization of iron will be addressed. This review article summarizes what we have learnt about iron and atherosclerosis from epidemiological studies, animal models and cellular systems and highlights the rather contributory than innocent role of iron in atherogenesis.

**Keywords: atherosclerosis, iron, hemoglobin, heme, intraplaque hemorrhage, LDL, oxidative stress, macrophages**

## **INTRODUCTION**

## **A ROLE FOR IRON IN ATHEROSCLEROSIS: THE "IRON HYPOTHESIS"**

The correlation between iron and heart disease was initially proposed by Sullivan in 1981. According to his "iron hypothesis," iron overload promotes cardiovascular disease, while on the contrary, sustained iron depletion/deficiency exerts a primary protective effect against ischemic heart disease. This theory, continually debated for more than 30 years, was based on the observation that male gender is associated with higher risk of cardiovascular disease, but that the protective effect in women is diminished after menopause (Sullivan, 1981, 1989). Based on results from the Framingham study (Kannel et al., 1976), Sullivan first hypothesized that the regular menstrual loss of iron, rather than the effect of estrogen, protects women against coronary heart disease (CHD). The failure of postmenopausal estrogen replacement to prevent coronary events further supported the iron hypothesis and its link to gender differences in atherosclerosis (Hulley et al., 1998).

The presence of redox-active iron, as well as high expression levels of H- and L-ferritin in human atherosclerotic lesions provided indirect support for the iron hypothesis (Smith et al., 1992; Pang et al., 1996). L-ferritin levels are increased in coronary arteries from patients with coronary artery disease (CAD), indicating that iron accumulates in atherosclerotic plaques (You et al., 2003). Additionally, cholesterol levels in lesions correlate with iron deposits (Stadler et al., 2004). Within the plaque, iron deposition and ferritin induction may be observed in endothelial cells and macrophages in early human lesions, and additionally in vascular smooth muscle cells (VSMCs) in late lesions.

Iron accumulates in human atherosclerotic lesions (Sullivan, 2009) via different mechanisms. Under normal conditions iron circulates in the bloodstream bound to its carrier protein transferrin (Tf). However, non-transferrin bound iron (NTBI) may be generated during chronic iron overload disorders such as sickle cell disease, thalassemia and transfusional iron overload (Brissot et al., 2012). NTBI is thought to be easily accessible to many cell types within the plaque, likely accumulating in endothelial cells, macrophages, and VSMCs (**Figure 1**).

Iron can enter into the atherosclerotic lesion in the form of free hemoglobin (Hb), that is released upon intravascular hemolysis or intraplaque hemorrhage (Kolodgie et al., 2003). Intraplaque hemorrhage originates from leaky neovessels invading the intima from the vasa vasorum as the intima thicken, and contributes significantly to the enlargement of the necrotic core (Sakakura et al., 2013). Increasing evidence indicates that plaque neovascularization and vasa vasorum density accompanied by intraplaque hemorrhage is a strong marker for plaque vulnerability (Moreno et al., 2004; Carlier et al., 2005; Virmani et al., 2005; Michel et al., 2011; Sakakura et al., 2013). Following intraplaque hemorrhage, red cells can be taken up by macrophages or they burst extracellularly releasing free Hb. Hb is prone to oxidation, especially in the highly oxidative milieu of the atheroma leading to the formation

monocytes/macrophages, and vascular smooth muscle cells (VSMCs).

many cell types in the atherosclerotic plaque: endothelial cells,

progression to foam cells. Iron also affects VSMC proliferation and migration into the lesion, favoring plaque progression.

of metHb and higher oxidation states such as ferrylHb, which can release heme (**Figure 1**). Altogether, Hb oxidation products, heme and iron exert pro-oxidant and pro-inflammatory effects targeting different cellular (i.e., endothelial cells, smooth muscle cells, macrophages) and acellular components (i.e., low-density lipoprotein, LDL) of the atherosclerotic vessel wall (**Figure 1**).

The finding that human atherosclerotic plaques contain redoxactive iron, that could promote free radical formation and lipid peroxidation, further suggested a role for iron in atherosclerosis that may be eventually responsible for progressive oxidative damage in atherosclerotic lesions. This review article summarizes our current knowledge about the role of Hb, heme, and iron in atherosclerosis by discussing the results of epidemiological studies, and observations in animal models and cellular experiments.

## **ALTERED IRON HOMEOSTASIS AND ATHEROSCLEROSIS: EPIDEMIOLOGICAL STUDIES, HUMAN CASES CORRELATION BETWEEN MARKERS OF IRON STORES AND DEVELOPMENT OF CAD**

To evaluate whether iron accumulation in the atherosclerotic plaque is a cause, rather than a consequence, of cardiovascular disease, several epidemiological and perspective studies were conducted since the nineties and many are still ongoing.

The results of several human studies strongly suggested a relationship between body iron levels and atherosclerosis. According to these epidemiological studies, high systemic iron levels, monitored by serum ferritin levels or transferrin saturation, positively correlated with increased risk of myocardial infarction (Salonen et al., 1992; Morrison et al., 1994; Tuomainen et al., 1998; Holay et al., 2012) cardiovascular disease (Rajapurkar et al., 2011), peripheral arterial disease (PAD) (Menke et al., 2009), and mortality rates (Lauffer, 1990). This association was stronger in men with high serum LDL levels, suggesting a synergistic role of high iron and high LDL levels (Salonen et al., 1992; Morrison et al., 1994). Finally, a clear proatherogenic role for iron was suggested by the observation that a 10 mg/L increase in serum ferritin level raised the probability of having at least two atherosclerotic plaques by 3% (Ahluwalia et al., 2010).

Body iron stores correlated with asymptomatic carotid atherosclerosis in healthy men (Syrovatka et al., 2011), an association becomes even more evident in symptomatic atherosclerosis. Plaques from symptomatic patients showed higher iron concentrations, signs of cap rupture and increased cap macrophage activity compared with asymptomatic plaques (Gustafsson et al., 2013). This suggests that the presence of iron in carotid plaques positively correlates with plaque vulnerability for rupture.

Additionally, the description of serum ferritin levels as a risk indicator of carotid lesion progression highlights a clear association between atherosclerosis progression and iron stores (Kiechl et al., 1997).

In agreement with this, serum iron levels directly correlate with cardiovascular disease severity. Serum iron levels were significantly higher in patients with severe atherosclerosis compared to those showing normal, mild, and moderate sings of CADs, thus further strengthening the hypothesis that high iron levels could affect atherosclerosis severity (Bagheri et al., 2013). Collectively, these epidemiological studies clearly identified high body iron levels as a risk factor for atherosclerosis and cardiovascular diseases.

Serum ferritin levels are frequently used to assess body iron status but increasing evidence suggests that this parameter additionally serves as a more general marker of inflammation (Kalantar-Zadeh et al., 2004; Manousou et al., 2011). Thus, some studies evaluated the relationships between serum ferritin, inflammatory cytokines and cardiovascular disease (Haidari et al., 2001; Depalma et al., 2010). Ferritin levels positively correlated with IL-6 and C-reactive protein (hsCRP) levels and were higher in patients that died of acute myocardial infarction vs. survivors, further supporting a rationale for measurement of ferritin levels in patients with atherosclerosis.

#### **BLOOD DONATION AND THE RISK OF CAD**

The incidence of atherosclerosis in premenopausal women was less than half of that observed in men of the same age (Kiechl and Willeit, 1999). The sex difference disappeared within 5 years after menopause, likely due to increased body iron stores. According to these observations, in 1991 Sullivan proposed that blood donation could prevent cardiovascular disease (Sullivan, 1991). Several studies confirmed the cardiovascular protective effect of blood donation. Blood donation was positively associated with a reduced risk of cardiovascular disease, in particular in non-smoking men with high serum LDL levels (Meyers et al., 1997, 2002; Tuomainen et al., 1997; Salonen et al., 1998). This result was in agreement with the iron hypothesis, according to which the protective effect of blood donation would be more pronounced in men that have a higher body iron load than women. A first randomized clinical trial (FeAST) showed that phlebotomy resulted in clinical benefits and reduction of death in young patients affected by PAD (Sullivan and Katz, 2007; Zacharski et al., 2007). High-frequency blood donation was associated with reduced body iron stores and improved vascular function and blood pressure, reduced oxidative stress, improved markers of cardiovascular risk in blood donors (Zheng et al., 2005; Houschyar et al., 2012). These findings are complemented by the observation that endothelial dysfunction is attenuated by iron chelation in patient with CAD (Duffy et al., 2001). Altogether these findings suggest that iron depletion, by blood donation or iron chelation, significantly lowers the risk of cardiovascular disease, thus supporting the iron hypothesis.

#### **ASSOCIATION OF IRON OVERLOAD AND CAD IN HEMOCHROMATOSIS**

If the iron hypothesis is correct, individuals with iron overload would be expected to show an increased risk and incidence of cardiovascular diseases, thus being optimal study model to test the validity of the hypothesis.

An interesting observation comes from the study of American blacks that compared to American whites and Hispanics are well known for higher ferritin levels throughout their entire life, likely explained by nutritional and genetic factors rather than increased iron intake (Zacharski et al., 2000). Interestingly, the incidence of CHD is higher in African–American than in white men and women (Gillum et al., 1997; Sacco et al., 1998), suggesting an association between body iron and cardiovascular disease.

Hereditary hemochromatosis (HH) is a genetic disorder associated with progressive iron overload, resulting in oxidative stress and organ failure. HH is more common among individuals of Northern European descent and is caused by inherited mutations in proteins implicated in iron transport and regulation, such as the upstream regulators of hepcidin, the human hemochromatosis protein (HFE), hemojuvelin, transferrin receptor (TfR)- 2, as well as hepcidin and ferroportin (FPN) (Hentze et al., 2010).

Hemochromatotic patients show vascular dysfunction and increased expression of adhesion molecules that positively correlates to iron overload and NTBI levels (Gaenzer et al., 2002; Kartikasari et al., 2006; Van Tits et al., 2007). These patients further show functional and structural alterations in midsize muscle arteries. In particular, arterial wall thickness is increased before the onset of cardiovascular complications, suggesting that this is an early abnormality in HH. This alteration is reverted by phlebotomy-induced iron depletion, which can also improve the endothelium-dependent vasodilation and the initial radial artery wall stiffening associated with HH (Failla et al., 2000; Gaenzer et al., 2002).

Different cohort studies reported a significantly greater risk of myocardial infarction, cerebrovascular mortality and cardiovascular mortality in carriers of the HFE mutation (Cys282Tyr) (Roest et al., 1999; Tuomainen et al., 1999; Rasmussen et al., 2001). Additionally, patients with genetic hemochromatosis have significant eccentric hypertrophy of the radial artery, although not showing arterial hypertension or evidence of cardiovascular disease.

In contrast to the above reported studies, others failed to find an association between hemochromatosis and the presence or frequency of atherosclerosis and did not succeed in establishing a link between body iron stores and cardiovascular diseases in human populations (Miller and Hutchins, 1994; Sullivan and Zacharski, 2001; Munoz-Bravo et al., 2013). The disagreement among epidemiological studies may result from variations in the validity and reliability of the indicators of iron status. Additionally, the magnitude of the relative risk associated with iron overload might be small, thus the association being obscured by stronger risk factors. Further prospective and experimental studies are needed to confirm the association between the iron status and atherosclerosis.

## **THE "REFINED IRON HYPOTHESIS": A PROTECTIVE ROLE FOR IRON-DEPLETED MACROPHAGES IN ATHEROSCLEROSIS**

Controversial results from epidemiological studies investigating different types of atherosclerotic events and using various markers for body iron levels present a confusing picture regarding the iron hypothesis. In addition, several studies failed to observe an increased risk or incidence of cardiovascular events in hemochromatotic patients, thus further increasing the confusion concerning an eventual association between iron overload and atherosclerosis (reviewed in Munoz-Bravo et al., 2013). Finally, the description of a potentially protective effect of hemochromatosis against atherosclerosis and cardiovascular diseases was perceived as a "paradox" and considered as clear evidence against the iron hypothesis (Miller and Hutchins, 1994; Sullivan and Zacharski, 2001; Munoz-Bravo et al., 2013). On the basis of these observations Sullivan presented a refinement of his "iron hypothesis" (Sullivan and Zacharski, 2001).

Since then the peptide hormone hepcidin has been identified as the master regulator of iron homeostasis. Hepcidin inhibits iron export by binding to FPN and promoting its degradation. By inhibiting FPN, hepcidin prevents iron release from enterocytes into the bloodstream and decreasing iron release from macrophages, thereby reducing the amount of iron systemically available. HH is hallmarked by low levels of hepcidin and/or increased expression of the iron exporter FPN. Therefore, in hemochromatotic patients the FPN-hepcidin circuitry is impaired, leading to increased duodenal iron absorption and reduced iron retention in macrophages (Hentze et al., 2010; Ganz and Nemeth, 2011). Considering the key role of the macrophages in atherogenesis, the selective iron depletion in this cell type was proposed as a mechanism of protection against foam cell formation and atherosclerotic lesion progression (**Figure 2**). According to this view, the hypothesis postulated by Sullivan that iron depletion protects against atherosclerosis may apply even to hemochromatotic individuals.

According to this view, hepcidin levels may act as a potential iron-dependent risk factor for atherosclerosis by regulating macrophage iron accumulation and atherosclerotic plaque progression (**Figure 2**). Recently, hepcidin was suggested as a predictor of carotid atherosclerosis. Serum ferritin was found to associate with vascular damage, common carotid thickness and presence of carotid plaques in all patients but not those showing a reduction in hepcidin levels due to heterozygous HFE mutations (Valenti et al., 2011b). Additionally, hepcidin levels and macrophage iron positively correlate with the release of IL-6 and macrophage chemoattractant protein 1 (MCP-1), and vascular damage in high-risk individuals (Valenti et al., 2011a). Collectively, these findings suggest an involvement of iron-loaded macrophages in inflammation and vascular alterations. On the other hand, monocytes from hemochromatotic patients showed reduced ability to accumulate iron and reduced upregulation of MCP-1 and IL-6 (Valenti et al., 2011a). The anti-inflammatory properties of iron-depleted macrophages may help to explain the lack of increased incidence of atherosclerosis in hemochromatotic patients.

Anyway, a direct and definitive demonstration of the refined iron hypothesis in human is still lacking and further studies are needed to fully elucidate the impact of macrophage-stored iron, as well as circulating iron and tissue-stored iron on human lesion formation and progression.

#### **THALASSEMIAS AND SICKLE CELL ANEMIA**

β-thalassemia and sickle cell anemia are hereditary blood disorders characterized by anomalies in the synthesis of the β-globin chains of Hb. β-thalassemic and sickle patients show increased plasma iron turnover, iron absorption and tissue iron deposition. Additionally, they have frequent hemolytic events that lead to the release of Hb and heme into the circulation, further increasing the amount of redox-active iron available for the production of reactive oxygen species and lipid peroxidation (Livrea et al., 1998; Brizzi et al., 2002). The release of Hb upon hemolytic events and the enhanced absorption of iron, to support inappropriate erythropoiesis, contribute to the pathogenesis of vasculopathy, a well-known predisposing factor for cardiovascular diseases. Moreover, these patients, usually presenting with severe anemia, require regular red blood cell transfusions (Vichinsky, 2005), further exacerbating iron overload and iron-driven oxidative stress (McLeod et al., 2009).

Iron-dependent peroxidative tissue injury results in arterial stiffness and dysfunction, frequently occurring in thalassemic patients (Kremastinos et al., 1999; Hahalis et al., 2008). Iron overload in patients with beta-thalassemia major lead to alterations in the arterial structures and in the thickness of the carotid arteries (Cheung et al., 2002; Tantawy et al., 2009). Moreover, carotid thickness positively correlated with age, Hb, ferritin and cholesterol levels in these patients (Cheung et al., 2006; Tantawy et al., 2009). As a result, CAD is a quite common cardiovascular complication in thalassemia (Ramakrishna et al., 2003; Ferrara and Taylor, 2005; Aessopos et al., 2007). Patients on a regular transfusion regimen progressively develop clinical manifestations of iron overload associated with heart dysfunction and left ventricular failure (Borgna-Pignatti et al., 2004). Interestingly, iron chelation therapy in thalassemia patients improves arterial function and stiffness (Cheung et al., 2008).

Ischemic complications are the major causes of morbidity and mortality in patients with sickle cell disease (Platt et al., 1994; Switzer et al., 2006). Ischemic events in these patients have been attributed to the effects of Hb polymerization, resulting in sickled cells trapped in the microcirculation (Francis and Johnson, 1991).

**a role for macrophage-retained iron in atherosclerosis.** Iron can accumulate in macrophages as inorganic iron and Hb-iron, upon erytrophagocytosis or hemolysis. Once stored in the cell, iron can be made available to the bloodstream via FPN-mediated export. According to the refined iron hypothesis, high hepcidin levels are

Hepcidin is known to bind to FPN, thus promoting its degradation and blocking iron export. This increases intracellular ROS levels and decreases cholesterol efflux. As a result, the oxidative status alters and LDL accumulation occurs, promoting foam cell formation, inflammation and eventually plaque instability.

Nevertheless, different factors other than red blood cell sickling, could contribute to these events, atherosclerosis being one of this.

SCD is an uncommon risk factor for atherosclerosis. However, in the last decades, together with the increased life expectancy of SCD patients, the risk to develop atherosclerosis is significantly increasing. Endothelial dysfunction, hyperhomocysteinemia and activation of platelets are the most likely mechanisms for the development of atherosclerosis in SCD patients (Elsharawy et al., 2009). The presence of excessive circulating Hb, heme, and iron in SCD could have in principle a crucial role in atherosclerosis development, even though a clear experimental proof of this is still missing. Conversely, a paradoxical protective effect of SCD on atherosclerosis and thrombosis was observed in ApoE-null mice transplanted with bone marrow from mice carrying the sickle cell mutation. This effect was abolished by inhibition of HO-1, suggesting that this protection relies on the activity of this enzyme, whose induction is sustained in SCD, due to the high circulating Hb and heme levels (Wang et al., 2013). These observations have the limit that mice were analyzed after 23–28 weeks from bone marrow transplantation, giving an idea of the onset of atherosclerosis but not of the late phases of the disease, in which HO-1 activity could be eventually overwhelmed, thus promoting atherosclerosis progression.

The most common sites of atherosclerosis in these patients are represented by large cerebral arteries (Rothman et al., 1986). Approximately 75% of strokes in sickle cell disease are the result of occlusion of cerebral vessels (Moran et al., 1998). Also pulmonary and splenic arteries are common sites of atherosclerosis in sickle cell disease. One-third of the sickle patients had histological evidence of medial hypertrophy and intimal proliferation in these arteries (de Chadarevian et al., 2001; Graham et al., 2007).

Therefore, thalassemia and sickle cell anemia patients are considered at high atherogenic risk in view of the perturbation of the Hb/heme/iron metabolism that predisposes these patients to oxidative status alterations (Belcher et al., 1999; Switzer et al., 2006).

#### **HAPTOGLOBIN POLYMORPHISM AND CAD**

Extracellular Hb may exert proatherogenic effects via different mechanisms. Free Hb scavenges nitric oxide, an important vasodilator and signaling molecule (reviewed in Rother et al., 2005). Moreover, oxidized Hb species trigger pro-oxidant (reviewed in Balla et al., 2005) and pro-inflammatory effects on vascular endothelium (Silva et al., 2009), and cause lipidperoxidation (Jeney et al., 2002; Potor et al., 2013).

Efficient mechanisms have evolved to remove extracellular Hb from the circulation to limit its deleterious effects. Haptoglobin (Hp) is an acute-phase plasma protein with the primary function to capture cell-free Hb and chaperon it to macrophages for degradation (reviewed in Alayash, 2011). Hp binding facilitates the removal of Hb from circulation via endocytosis through the CD163 macrophage scavenger receptor (Kristiansen et al., 2001).

The Hp gene is polymorphic in humans, whereby the two functional alleles (hp1 and hp2) can form three genotypes: Hp1- 1, Hp2-1, and Hp2-2 with heterogeneous protein structure and functional differences (reviewed in Goldenstein et al., 2012). Differences between antioxidant properties of Hp1-1 and Hp2-2 were examined. An early study showed that Hp1-1 protein is more potent in inhibiting the oxidative actions of extracorpuscular Hb (Melamed-Frank et al., 2001). Contradictory, a recent study described no differences between the two phenotypes in protecting against Hb-driven toxicity (Lipiski et al., 2013). When applied *in vivo* following Hb injection both Hp1-1 and Hp2-2 attenuate Hb-induced blood pressure response with equal efficacy, restrict trans-endothelial diffusion of extracellular Hb equally, and prevent Hb redistribution and renal iron deposition in the same way (Lipiski et al., 2013). Both phenotypes show similar abilities to stabilize the ferryl Hb state, to restrict heme release from the complex, and to prevent Hb-driven LDL oxidation *in vitro* (Lipiski et al., 2013). Immunomodulatory effects of the two phenotypes were compared as well. The Hp1-1-Hb complex induces more robust anti-inflammatory macrophage signaling, leading to the secretion of anti-inflammatory cytokines than that of Hp2-2-Hb complex (Philippidis et al., 2004; Landis et al., 2013).

The Hp polymorphism was investigated as a possible genetic determinant in cardiovascular disease. These epidemiologic studies revealed that the Hp2-2 genotype is a risk factor for cardiovascular complications in both type I and type II diabetic patients (reviewed in Costacou and Levy, 2012). In particular, the Hp2-2 genotype is associated with elevated amounts of iron in atherosclerotic carotid plaques, accompanied by increased levels of oxidation-specific epitopes, increased macrophage infiltration and decreased VSMCs, all events promoting plaque instability (Lioupis et al., 2011, 2012; Purushothaman et al., 2012). In addition, the Hp2-2 genotype is associated with increased circulating oxLDL levels when compared to Hp1-1 or Hp2-1 genotypes (Brouwers et al., 2004). A correlation between the Hp2- 2 genotype, carotid plaque instability and increased risk of major cardiovascular diseases was recently described (Ijas et al., 2013).

Collectively, these findings suggest that detoxification of extracellular Hb by Hp acts in an atheroprotective manner. In addition, the Hp2-2 genotype represents a non-modifiable risk factor for cardiovascular diseases. Because Hp1-1 and Hp2-2 inhibit the oxidative actions of extracorpuscular Hb equally, therefore disease association is most probably explained by other functions or properties of the Hp molecule.

#### **HEME OXYGENASE-1 (HO-1) AND CARDIOVASCULAR DISEASE**

Heme oxygenases catabolize heme to equimolar amounts of biliverdin, carbon monoxide, and free iron, followed by the conversion of biliverdin into bilirubin by biliverdin reductase (Singleton and Laster, 1965; Tenhunen et al., 1968). HO-1 is a stress-inducible isoform of heme oxygenases, encoded by the hmox-1 gene which possesses antioxidant, anti-apoptotic and anti-inflammatory properties (reviewed in Gozzelino et al., 2010; Durante, 2011). These protective mechanisms partially rely on the ability of HO-1 to extract iron from heme. The released iron induces the expression of ferritin, the 24-subunit complex of heavy (H) and light (L) chains, with enormous iron-storage capacity (Eisenstein et al., 1991; Harrison and Arosio, 1996). In addition, both bilirubin and CO, the other two end products of heme degradation exhibit direct anti-oxidant and antiinflammatory activities (Gozzelino et al., 2010).

An important, but somewhat neglected function of HO-1 is its role in iron recycling (Poss and Tonegawa, 1997). Erythrophagocytosis, subsequent HO-1-mediated heme degradation and iron release from macrophages is a major mechanism in iron recycling, accounting for about 90% of total body iron turnover (reviewed in Hentze et al., 2010).

Accumulating evidences suggest the protective role of HO-1 in atherosclerotic vascular disease (reviewed in Chan et al., 2011). Both the antioxidant bilirubin and the vasodilator CO may contribute to this atheroprotective effect (Siow et al., 1999; Mayer, 2000; Parfenova et al., 2012; Erkan et al., 2013). Low bilirubin levels are associated with endothelial dysfunction and increased intima-media thickness (Erdogan et al., 2006), whereas high plasma bilirubin concentrations are linked to low incidence of cardiovascular disease (Schwertner et al., 1994) and stroke (Kimm et al., 2009). Differences in plasma bilirubin levels may arise from the variation of HO-1 activity in humans.

In the human hmox-1 promoter a GT repeat microsatellite polymorphism exists, leading to higher hmox-1 transcriptional activity and subsequently higher HO-1 expression in individuals having shorter GT repeats compared to subjects with longer GT repeats. A number of studies investigated the relationship between this gene polymorphism and the risk of cardiovascular disease, with conflicting results. Some studies revealed that shorter GT repeats in the hmox-1 promoter region are associated with lower incidence and/or progression of CAD (Kaneda et al., 2002; Liang et al., 2013), whereas others argue against a relevant role of this polymorphism in cardiovascular diseases (Lublinghoff et al., 2009).

Progressive atherosclerotic lesion destabilization with subsequent plaque rupture is a key event predisposing to acute thrombus formation and coronary artery occlusion (Schwartz et al., 2007). Autopsy studies reveal that the risk of plaque rupture mainly depends on the composition of the plaque rather than its size (Kolodgie et al., 2004). Severe macrophage infiltration, a necrotic core and a thin fibrous cap are the main characteristics of vulnerable plaques (Kolodgie et al., 2004). In humans, HO-1 expression is increased in atherosclerotic lesions and closely correlates with plaque instability and pro-inflammatory markers. The observation that HO-1 induction reverses plaque progression from a vulnerable plaque to a more stable phenotype suggests that HO-1 expression may act as a compensatory atheroprotective mechanism (Cheng et al., 2009).

By contrast, HO-1 deficiency in humans leads to severe vascular pathologies (Yachie et al., 1999). A 6-year old boy with inactivating mutations of the HO-1 gene presented with severe intravascular hemolysis associated with persistent endothelial damage. Autopsy examination revealed the presence of aortic fatty streaks and fibrous plaques at this young age, highlighting the atheroprotective function of HO-1 (Yachie et al., 1999). More recently, another case of HO-1 deficiency in a young girl was reported, with evidence of severe endothelial damage, as suggested by raised inflammatory markers, von Willebrand factor and coagulopathy (Radhakrishnan et al., 2011). Since free circulating heme promotes endothelial damage, the lack of functional HO-1 likely results in a form of vasculitis or endothelial injury syndrome. This may therefore increase their susceptibility to develop atherosclerosis.

Taken together, these findings prove a crucial role for HO-1 in the maintenance of vascular homeostasis and counteraction of atherosclerosis.

## **ALTERED IRON HOMEOSTASIS AND ATHEROSCLEROSIS: ANIMAL MODELS**

## **IRON OVERLOAD AND IRON DEFICIENCY IN ATHEROSCLEROSIS**

The effect of iron in atherogenesis was tested using different hypercholesterolemic animal models. In an initial study that intramuscular administration of iron dextrane augmented the formation of atherosclerotic lesions in hypercholesterolemic rabbits (Araujo et al., 1995). In contrast, another group using the same rabbit model described that iron dextrane injection significantly decreased lesion formation by about 50% by reducing plasma cholesterol levels (Dabbagh et al., 1997). Kirk et al. observed a reduction (about 50%) in plaque area in apoE deficient mice fed with a 2% carbonyl iron containing standard diet in spite of that dietary iron overload caused a modest (30%) rise in plasma triglyceride and cholesterol levels (Kirk et al., 2001).

Other studies took the opposite approach and examined the effect of iron restriction on atherogenesis. In this regard, atherosclerotic lesions in mice fed a low-iron diet were significantly smaller than those found in control littermates (Lee et al., 1999). Reduced plaque size in the low-iron group was associated with lower levels of circulating autoantibodies to oxLDL, and the diminished occurrence of thiobarbituric acid reactive epitopes in the lesions (Lee et al., 1999). This was explained by the observation that dietary iron restriction increases plaque stability via elevated collagen and reduces matrix metalloproteinase-9 expressions in the lesion (Lee et al., 2003). Consistently, iron chelation by DFO lowers the iron content of the lesions and inhibits atherosclerotic lesion development in cholesterol-fed rabbit (Minqin et al., 2005) as well as in apoE deficient mice (Zhang et al., 2010). Other than an effect on atherosclerosis, several studies showed that iron depletion by chelation significantly reduces endothelial activation and vascular dysfunction in animal models (Ishizaka et al., 2005). Recently, a combined therapy of iron chelator and antioxidant was observed to restore ironinduced brain vascular dysfunction in rats (Sripetchwandee et al., 2014), supporting the idea that iron promotes earlier steps in atherogenesis.

### **HEMOCHROMATOSIS MODELS**

Although there is support for the idea that iron is detrimental for atherosclerosis, the validity of the original iron hypothesis has not been tested in models of genetic iron overload, such as hemochromatotic mice. To date, several mouse models of hemochromatosis are available, such as HFE-null, Hamp-null, HJV-null, and BMP6-null mice (Fleming et al., 2011) but atherosclerosis progression has not been assessed in any of them. Future studies will have to dissect the contribution of systemic iron overload and macrophage iron deficiency in hemochromatotic mouse models for atherosclerosis in order to better understand the outcome of the epidemiological studies.

## **ANIMAL MODELS TO ASSESS THE IMPACT OF MACROPHAGE IRON ON ATHEROSCLEROSIS**

The key role of macrophages in atherosclerosis was extensively studied in animal models. Lipid-laden foam cells are macrophages derived from circulating monocytes that migrate into the vessel wall. Inhibition of monocyte migration, by disrupting a variety of chemokine/chemokine receptor interactions, was shown to inhibit atherosclerosis development. The osteopetrotic (op) mouse, spontaneously deficient in macrophage-colony stimulating factor (M-CSF), displayed a reduction of 86% in plaque volume, demonstrating the essential role of macrophages in atherogenesis (Smith et al., 1995). Quite recently, a CD11b– diphtheria toxin receptor transgenic mouse line was generated, whereby diphtheria toxin administration conditionally ablates monocytes/macrophages (Stoneman et al., 2007). In atherogenesis experiments, diphtheria toxin markedly decreased monocyte numbers by 50% and altered plaque development and composition, reducing collagen content and necrotic core formation, thus demonstrating that monocytes/macrophages are critical for atherogenesis.

The crucial role of macrophages in atherosclerosis raised the possibility of selective intraplaque macrophage depletion achievable as a specific therapeutic intervention to counteract plaque progression. This approach now gains increasing attention in cardiovascular medicine. Several successful strategies have recently been reported to induce macrophage cell death in atherosclerotic plaques (Martinet and De Meyer, 2007). Its feasibility is currently debated and object of several studies, aimed at locally deleting macrophages, without affecting this cell type in other tissue compartments. However, local therapies can be administered only for a relatively short time, with the limitation that macrophages may reinfiltrate the plaque after treatment.

Assessment of the impact of macrophage-associated iron on atherosclerosis could eventually provide additional mechanisms/pathways that could be targeted in macrophages to prevent/reduce atherosclerosis. Animal studies were initiated to evaluate the role of iron in macrophages, thus revisiting the iron hypothesis. Although not tested in hemochromatotic mice, atherosclerosis was studied in mice with macrophage iron depletion triggered by drug administration. The pharmacological suppression of hepcidin in mice decreased macrophage iron content, and increased cholesterol efflux, thus resulting in reduced foam cell formation (Saeed et al., 2012). In particular, the reduction of macrophage-associated iron levels lowered the formation of ROS and increased the expression of cholesterol transporters, namely ABCA1 and ABCG1. This leads to improved lipid efflux by macrophages, correlating with reduced foam cell formation and atherosclerosis (**Figure 2**). This approach is limited by the use of a BMP signaling pathway inhibitor to achieve hepcidin suppression. BMP signaling inhibitors are in fact expected to effect on many other biological processes involved in the formation of the atherosclerotic plaque, other than those directly dependent on hepcidin reduction. Future studies that apply drugs that directly and specifically reduce hepcidin expression or that counteract its activity are needed to examine whether hepcidin suppression by itself affects progression of atherosclerosis.

In agreement with these findings, hepcidin recently emerged as a positive regulator of atherosclerotic plaque destabilization, via regulating macrophage iron homeostasis (Li et al., 2012). Hepcidin production in the carotid artery was achieved by adenoviral infection in a mouse model of accelerated atherosclerosis. Although a change in plaque size was not observed, hepcidin overexpression significantly affected plaque composition, increasing intraplaque macrophages and decreasing VSMCs and collagen amounts. Additionally, hepcidin overexpression increased trapped iron as well as oxidized-LDL levels in intraplaque macrophages. This correlated with increased oxidative stress and expression of pro-inflammatory cytokines by macrophages and enhanced plaque vulnerability, suggesting that hepcidin plays a critical role in plaque destabilization.

Collectively, these findings indicate that the interactions of hepcidin, trapped iron, and accumulated lipids are critical for proatherosclerotic activation of macrophages leading to plaque destabilization (**Figure 2**). The suppression of hepcidin by specific shRNA exerts effects opposite to those reported above. These studies described a unique role for hepcidin in promoting atherosclerosis progression and plaque instability and provided evidence of a protective function of the iron-spared macrophage, at least partially clarifying the paradoxical issues observed in hemochromatosis.

A complementary approach to test the effect of iron-loaded macrophages on atherosclerosis was recently pursued (Kautz et al., 2013). Atherosclerosis was studied in the flatiron (ffe) mouse (Zohn et al., 2007), a model that specifically accumulates iron in macrophages. Contrary to the refined iron hypothesis, atherosclerosis was not increased in mice with elevated macrophage iron. In addition, increased macrophage iron levels triggered by parenteral iron administration also failed to promote atherosclerosis. These findings dispute that macrophage iron loading could be an aggravating factor in the pathogenesis of atherosclerosis.

#### **EFFECTS OF HO-1 IN ANIMAL MODELS OF ATHEROSCLEROSIS**

The role of HO-1 in atherosclerotic lesion formation was first investigated in apoE deficient mice, overexpressing HO-1. Overexpression of HO-1 in the vasculature was achieved by direct injection of an adenovirus expressing HO-1 (Adv-HO-1) into the left ventricles of anesthetized animals. HO-1 overexpression inhibits lesion formation and reduces iron overload in apoE deficient mice (Juan et al., 2001). Reduced iron deposition in aortic tissues of Adv-HO-1-treated mice might be explained by the observation that HO-1 overexpression augments iron recycling from cells (Ferris et al., 1999). To further examine the role of HO-1 in atherogenesis, mice deficient in both HO-1 and apoE were generated. When compared to apoE deficient mice these double knock-out mice exhibited accelerated and more advanced lesion formation in response to a cholesterol rich diet (Yet et al., 2003). Interestingly, aged HO-1 knock-out mice exhibit severe aortitis and coronary arteritis with mononuclear cell infiltration accompanied by fatty streak formation, even on a standard chow diet (Ishikawa et al., 2012).

Expression of HO-1 is strongly regulated by its substrate heme, in a Bach1-mediated manner. Bach1 is a transcriptional repressor of the hmox-1 gene that becomes inactive and undergoes ubiqitination and degradation upon heme binding (Zenke-Kawasaki et al., 2007). Consequently, deletion of the bach1 gene leads to sustained HO-1 expression in various tissues. The effect of bach1 deletion in atherosclerosis was studied in Bach1 apoE double deficient mice (Watari et al., 2008). In these mice HO-1 was upregulated in the vasculature, mainly in the vascular endothelium (Watari et al., 2008). Elevated HO-1 expression was accompanied by reduced plaque area compared with that in apoE deficient mice, supporting the anti-atherogenic nature of HO-1 (Watari et al., 2008). Overexpression of HO-1 inhibited lesion progression into vulnerable plaques, whereas inhibition of HO-1 activity augmented plaque vulnerability (Cheng et al., 2009).

Biological effects of a wide variety of molecules depend on the upregulation of HO-1 by these compounds (Bach, 2005). Accordingly, there are several anti-atherosclerotic compounds that exert their protective effects via the induction of HO-1. For example the anti-oxidant probucol, has been shown to protect from atherosclerosis by a HO-1 pathway that is independent of radical scavenging in various models of vascular diseases (Wu et al., 2006). Recently, HO-1 was found to be the molecular target of Tanshinone IIA, a lipophilic bioactive compound extracted from Salvia miltiorrhiza Bunge that exert anti-atherogenic effect via suppressing cholesterol accumulation in macrophages (Liu et al., 2014). In addition, the polyphenolic compound quercetin as well attenuates endothelial dysfunction and atherosclerosis in apoE deficient mice in a HO-1 dependent manner (Shen et al., 2013).

Taken together, these results support a protective function for HO-1 in atherosclerotic lesion formation and progression.

## **EFFECT OF IRON ON MAIN PLAYERS IN ATHEROGENESIS LIPID METABOLISM AND LDL OXIDATION**

Elevated iron stores reflected by increased plasma ferritin levels are positively correlated with the prevalence of certain diseases such as metabolic syndrome, diabetes and obesity (Jehn et al., 2004, 2007; Lecube et al., 2008; Sun et al., 2008). All of these diseases are associated with abnormal lipid metabolism, but until recently there were few studies addressing whether elevated iron levels and lipid metabolism are directly correlated. A first study showed that HH associated with primary hypertriglyceridemia (Solanas-Barca et al., 2009), which can be improved by periodic therapeutic phlebotomy (Casanova-Esteban et al., 2011). In rats with dietary iron overload a significant increase in triglycerides, free cholesterol, cholesteryl ester, and high-density lipoproteincholesterol levels was observed (Brunet et al., 1999). By contrast, intraperitoneal injection of iron-dextrane enhanced serum triglyceride levels but not serum cholesterol levels in an independent study (Silva et al., 2008). Excess iron directly modulates activities of several key enzymes for cholesterol and triglyceride homeostasis—e.g., 3-hydroxy-3-methylglutaryl coenzyme A reductase, cholesterol 7alpha-hydroxylase, acyl-CoA: cholesterol acyltransferase and lipoprotein lipase - which might explain perturbations of lipid metabolism in conditions of iron overload (Brunet et al., 1999).

Other than affecting lipid metabolism, iron mediates the oxidative modification of LDL, a clear contributing factor to the pathogenesis of atherosclerosis (Heinecke et al., 1984). The molecular mechanism of iron-catalyzed LDL oxidation was extensively studied. Redox active iron that undergoes oxidation and reduction is an absolute necessity to catalyze lipid peroxidation (Lynch and Frei, 1993; Miller et al., 1993). Iron-mediated oxidation of LDL is dependent on superoxide anion (O−• <sup>2</sup> ) that acts as a Fe3<sup>+</sup> reducing agent, but requires neither H2O2 nor production of hydroxyl radical (OH•) by the Fenton reaction (Lynch and Frei, 1993).

The unlikely existence of iron in free catalytically active form in normal body fluids initiated the search for physiologically more relevant iron compounds with the ability to oxidize LDL. In fact most of the iron in the human body is found in heme that serves as a prosthetic group in Hb and other heme proteins. This ubiquitous iron compound is a very efficient catalyst of LDL oxidation (Balla et al., 1991a). Studies revealed that initiation and propagation of heme-induced lipid-peroxidation is independent of Fenton chemistry similarly to that of iron-mediated LDL oxidation. The initial interaction between heme and H2O2 might lead to the formation of ferryl and perferryl radicals, those can be responsible for initiating lipid peroxidation (Klouche et al., 2004). During heme-mediated LDL oxidation, oxidative scission of the heme ring occurs and iron is released (Balla et al., 1991a). Both heme degradation and LDL oxidation are effectively inhibited by lipid soluble antioxidants and iron chelators (Balla et al., 1991a; Pocsi et al., 2008).

Several lines of evidence suggest that heme-mediated oxidation of LDL occurs *in vivo*. High amount of heme in the plasma of the HO-1 deficient boy was correlated with extensive LDL oxidation (Jeney et al., 2002). During heme-mediated LDL oxidation heme reacts with proline and arginine residues in apolipoprotein B-100 and a unique oxidation product, gammaglutamyl semialdehyde is formed, that is subsequently reduced to 5-hydroxy-2-aminovaleric acid (HAVA). HAVA is a hallmark of heme-mediated LDL oxidation, as other agents known to trigger LDL oxidation, such as HOCl, H2O2 alone or in combination with Cu2<sup>+</sup> or Fe2<sup>+</sup> induce only minor HAVA formation (Julius and Pietzsch, 2005). Elevated concentrations of HAVA were found in LDL of patients with impaired glucose tolerance and with diabetes mellitus suggesting that heme-mediated LDL oxidation occurs in these patients (Julius and Pietzsch, 2005).

Cell-free Hb when oxidized releases heme and induces oxidative modification of LDL (Jeney et al., 2002). This effect was abolished by the heme-scavenging protein Hx and by Hp or cyanide, agents that either bind free heme or strengthen the heme-globin bond, highlighting the role of heme release in this process (Miller et al., 1996; Jeney et al., 2002). Recently a feedforward process in atheromatous lesions with the interactions of atheroma lipids and cell free Hb was described. This vicious cycle includes lipid-hydroperoxide mediated oxidation of Hb, spontaneous heme release, oxidative heme scission, iron release, and further lipid peroxidation (Nagy et al., 2010; Jeney et al., 2013; Potor et al., 2013).

Collectively, these results confirm that excess iron, heme, and cell-free Hb act in an atherogenic manner.

## **ENDOTHELIAL CELL ACTIVATION AND DYSFUNCTION**

Upon steady-state condition, endothelial cells provide an antithrombotic and antiadhesive surface in the vasculature. Lowgrade inflammation is a characteristic of the atherosclerotic lesions in which endothelial cell activation occurs, triggering vasoconstriction, thrombosis as well as leukocyte adhesion, and transmigration (Libby, 2002). This pro-inflammatory response relies on the upregulation of a variety of genes encoding vasoconstrictive, pro-thrombic, pro-inflammatory, chemotactic, and adhesive molecules (reviewed in Pober and Sessa, 2007). Redoxsensitive mechanisms involving the activation of redox-regulated transcription factor nuclear factor-kB (NF-kB) have been implicated in the expression of these vascular inflammatory molecules (Marui et al., 1993; Kunsch and Medford, 1999).

Accumulating evidences suggest the critical role of redox active iron in mediating the pro-inflammatory response in endothelial cells. Chelation of iron by DFO leads to decreased induction of E-selectin, vascular cell adhesion molecule-1 (VCAM-1), and intercellular adhesion molecule-1 (ICAM-1) in endothelial cells stimulated by tumor necrosis factor alpha (TNF alpha) (Zhang and Frei, 2003). Switching to *in vivo* models, iron chelation inhibits the lipopolysaccharide-mediated induction of soluble cellular adhesion molecules, monocyte chemoattractant protein-1 (MCP-1) and activation of NF-kB in mice (Zhang et al., 2010). In humans, iron chelation by DFO improves nitric oxidemediated endothelium-dependent vasodilation in patients with CAD, highlighting a role for iron in impaired nitric oxide action in atherosclerosis (Duffy et al., 2001).

The direct association between excess iron and endothelial dysfunction has been established upon physiological and pathological conditions. Administration of iron into healthy individuals provoked endothelial dysfunction accompanied by increased generation of superoxide radical in whole blood (Rooyakkers et al., 2002). Hemodialysis (HD) patients who receive intravascular iron along with erythropoiesis-stimulating agents to treat functional iron deficiency and subsequent anemia, as well as iron-overload patients, provide a unique opportunity to study the effect of iron on vascular function. There are conflicting data regarding the effect of iron on vascular function, cardiovascular risk and overall mortality in HD patients. Serum ferritin was reported as a marker of mortality in HD patients, but whether ferritin levels were regulated by iron in these patients is not clear (Kalantar-Zadeh et al., 2001). High serum ferritin level in HD patients (*>*600μg/L) is associated with increased overall 4-year mortality even in the absence of infection (Kletzmayr and Horl, 2002). A cohort study concluded that iron supplementation at a dose of 1000 mg or less over 6 month does not have any adverse effect, whereas iron supplementation at higher doses is associated with elevated morbidity (Feldman et al., 2004).

Recently, more mechanistic studies were performed to show the involvement of endothelial dysfunction in iron-triggered cardiovascular complications. Intravenous administration of iron increased the levels of circulating soluble adhesion molecules in HD patients which was associated with higher risks for cardiovascular events (Kuo et al., 2012). Consistently, endothelial cells treated with iron sucrose, a widely used iron drug, changed their morphology and showed an increased ability to recruit monocyte (Kamanna et al., 2012). Iron sucrose treatment causes marked reduction in acetylcholine-mediated relaxation in rat aorta rings, thus further confirming the detrimental effect of iron on endothelial function (Kamanna et al., 2012). Iron overload diseases are associated with the presence of NTBI in the serum. In serum from hemochromatosis patients, NTBI levels were found to be positively correlated with the expressions of adhesion molecules, ICAM-1, VCAM-1, and E-selectin but not to the inflammatory marker CRP (Kartikasari et al., 2006).

Hemolytic diseases are also associated with endothelial dysfunction, therefore several studies addressed whether cell free Hb or heme can harm endothelial cells directly in these pathologies. Heme strongly sensitizes endothelial cells to oxidant-mediated killing and its plasma scavenger, Hx, completely inhibits this effect (Balla et al., 1991b). Hb when oxidized to metHb can transfer heme to the endothelium and exert the same sensitizing effect as free heme (Balla et al., 1993). More recently globinglobin cross-linked Hb multimers were identified in complicated atherosclerotic lesions (Nagy et al., 2010). The formation of these species can be triggered by inorganic and organic peroxides and involves the generation of ferrylHb and globin radicals (reviewed in Jeney et al., 2013). Interestingly, these globin-globin cross-linked Hb multimers are the exclusive species inducing pro-inflammatory response in endothelial cells *in vitro*. As a pro-inflammatory agonist, globin-globin cross-linked Hb multimers trigger the formation of intercellular gaps disrupting the integrity of the endothelial cell monolayer, induce the expression of adhesion molecules, E-selectin, ICAM-1, and VCAM-1 leading to increased monocyte adhesion (Silva et al., 2009; Potor et al., 2013).

Recently, the study of mouse models of hemolytic diseases (β-thalassemia and sickle cell disease mice) proved that heme largely contributes to endothelial activation and dysfunction and cardiovascular alterations (Tolosano et al., 2010; Vinchi and Tolosano, 2013). These effects can be strongly counteracted by the administration of an Hx-based therapy (Vinchi et al., 2008, 2013). Most of these effects have been described to rely on heme ability to activate TLR4 in endothelial cells. Heme-mediated TLR4 activation leads to Weibel-Palade body (WPB) mobilization and degranulation, thus promoting the expression of P-selectin and VWF, and NF-κB activation in endothelial cells *in vitro* and vessel wall surfaces *in vivo* (Belcher et al., 2014). By activating TLR4 pathway, heme triggers vascular stasis and occlusion, common complications associated with hemolytic disorders such as sickle cell disease. TLR4-null mice transplanted with sickle bone marrow do not exhibit heme-induced vaso-occlusion and activation of WPB/NF-κB. The ability of Hb and heme to induce stasis is abolished by the administration of the Hb and heme scavengers, Hp and Hx in a mouse model of SCD (Belcher et al., 2014). In addition heme has been recently described as a trigger of the acute chest syndrome, one of the major complications associated with SCD. In a sickle mouse model, respiratory failure due to ACS was avoided by treatment with recombinant Hx. The activation of TLR4 by heme in vascular tissues was likely responsible for this lethal type of acute lung injury. Pharmacologic inhibition of TLR4 protected sickle mice from heme-induced ACS (Ghosh et al., 2013).

These recent findings highlight a crucial role for the TLR4 activated signaling pathway in Hb/heme-mediated activation of endothelial cells and macrophages. From the point of view of atherosclerosis, a role for TLR4 in the initiation and progression of the disease is widely recognized. TLR4 is expressed on the cell surface of the main cell types involved in atherosclerosis, endothelial cells, platelets and macrophages. Its activation is required to enhance the expression of adhesion molecules and cytokines (e.g., MCP1), thus promoting the recruitment of monocytes and initiating the inflammatory response. The enhanced cytokine and chemokine release by TLR4 activation could stimulate EC and VSMC migration and proliferation, thus accelerating plaque progression (Pasterkamp et al., 2004). Additionally, oxLDL upregulate TLR4 expression and induce cytokine expression at least partially via TLR4 activation (Pasterkamp et al., 2004; den Dekker et al., 2010). Also platelets participate in atherogenesis and show clear signs of increased activity in individuals with established cardiovascular and thrombotic disease. Increased activation of platelets via TLR4 binding could increase the risk of atherosclerosis and thrombosis (Jayachandran et al., 2010) and heme could potentially promote this event. Some mouse models and human studies also support a role of TLR4 in the progression of atherosclerotic disease. Individuals with TLR4 deficiency may be at increased risk for infection but at lower risk for cardiovascular disease (Jayachandran et al., 2010). Besides heme scavenging by Hp and Hx, targeting TLR4 as a signaling receptor downstream of heme could be an alternative therapeutic approach to reduce heme-driven pro-atherogenic effects.

Heme and oxidized Hb species can also threaten vascular endothelial cell integrity indirectly by their ability to mediate the oxidative modification of LDL (reviewed in Balla et al., 2005). Lipid hydroperoxides are transiently formed during LDL oxidation and responsible mostly for oxLDL-mediated endothelial damage and for initiation of redox signaling (Nagy et al., 2005, reviewed in Chapple et al., 2013).

Altogether these finding indicate that excess iron, extracellular Hb and heme have detrimental effects on the vascular endothelium leading to endothelial dysfunction.

## **THE EFFECT OF IRON ON MACROPHAGE POLARIZATION AND FUNCTION**

During atherogenesis, blood monocytes are recruited to the vascular endothelium and attracted to the subendothelial space where the deposition of LDL occurs. These monocytes are later differentiated into macrophages and foam cells. Atherosclerosis macrophages are one of the most important cell populations, as they contribute to the progression of the lesions.

Macrophages are innate immune system cells therefore they exhibit great plasticity. Different stimuli and environments can lead to diverse phenotypes. Their functions comprise inflammatory responses, antimicrobial activity, tissue remodeling and iron recycling (Khallou-Laschet et al., 2010; Leitinger and Schulman, 2013).

Macrophages are key players in the regulation of iron homeostasis as they recycle 20–25 mg of iron per day from senescent erythrocytes. Macrophages engulf aged or damaged erythrocytes and catabolize heme via HO-1 activity. Heme-derived iron is then exported from phagocytic vesicles by the natural resistanceassociated macrophage protein 1 (NRAMP1) and divalent metal transporter 1 (DMT1) expressed within phagolysosomal membranes. Iron is either stored coupled to ferritin or exported as ferrous iron via FPN, the only known iron exporter (Hentze et al., 2010). Interestingly, several studies demonstrated that much of the iron within plaques is associated with macrophages and foam cells. The exact source of iron still needs to be elucidated. However, it is well known that an important contribution is made by Hb-contained iron that is released from microhemorrhage within the plaque (Boyle et al., 2009; Saeed et al., 2012).

The identification of different macrophage subtypes that polarize in response to a specific microenvironment (Leitinger and Schulman, 2013), in both human and murine atherosclerotic lesions, raised the possibility that iron itself could affect macrophage plasticity. A putative involvement of iron in the polarization of some macrophage subtypes has been recently demonstrated in atherosclerosis (**Figure 3**).

Two major subtypes of macrophages have been extensively studied and described: the classical activation (M1) and the alternative activation (M2) macrophages (**Figure 3**). M1 macrophages are polarized after exposure to IFNγ and/or microbial products such as LPS. These macrophages are characterized by a strong pro-inflammatory activity with the production of several inflammatory cytokines: IL-1β, IL-6, IL-8, IL-12, and TNFα (Butcher and Galkina, 2012). In terms of iron metabolism, M1 macrophages are prone to a low turn-over of iron with low expression of CD163, HO-1, FPN and high expression of ferritin, suggesting an iron retention phenotype with decreased iron recycling and export capacity (Recalcati et al., 2010). In chronic venous leg ulcers and wound healing models, macrophage iron overload induces an unrestrained pro-inflammatory M1 phenotype, via enhanced production of TNFα and hydroxyl radicals, suggesting that iron accumulation in macrophages contributes to a pro-inflammatory phenotype (Sindrilaru et al., 2011). Similarly, macrophage exposure to heme could lead to a pro-inflammatory activation of these cells. In fact, heme has been described as an extracellular signaling molecule able to affect the innate immune response thanks to its ability to bind and activate TLR4. By activating TLR4 heme, induces the secretion of (TNF-alpha) by macrophages (Figueiredo et al., 2007), suggesting that heme retains the ability to polarize macrophages toward an M1 rather an M2 phenotype. Whether plaque-associated macrophages are polarized toward the M1 or M2 phenotype in hemolytic sickle animal models or patients still needs to be investigated to address this point.

In atherosclerosis M1 macrophages were detected in both human and mouse lesions, in the lipid core of the plaque. M1 macrophages were the prevalent macrophage subtype in advanced lesions (Khallou-Laschet et al., 2010). It is postulated that M1 macrophages might contribute to the formation of the necrotic core, since inflammatory macrophages are prone to

**FIGURE 3 | High macrophage plasticity in atherosclerosis.** In the atherosclerotic plaque, macrophages differentiate into different phenotypes. The two extreme phenotypes are represented by M1 and M2 macrophages. M1 macrophages show strong pro-inflammatory properties, thus potentially being involved in lesion progression. M1 macrophages show high expression levels of iNOS, MHCII, and inflammatory cytokines, such as IL-6 and TNF-a. M2 macrophages are considered anti-inflammatory and are involved in tissue repair and remodeling. M2 specific markers are Arginase 1, Ym1, and IL-10. The M2 phenotype is reported as anti-atherogenic. In addition, several other macrophage phenotypes are observed in the atherosclerotic plaque. Mhem macrophages originate as a consequence of intraplaque hemorrhage and are endowed with high Hb handling ability. These anti-atherogenic macrophages express high levels of the heme-degrading enzyme HO-1 and the Hp-Hb scavenger receptor CD163. Additionally, Mheme macrophages express the cholesterol exporters ABCA1 and ABCG1, thus efficiently activating reverse cholesterol efflux. Mox macrophages are generated upon oxidized phospholipid stimulation. They show anti-oxidant properties, as they express genes involved in the anti-oxidant responses such as HO-1, Txnrd1, and Srxn1. Their potentially athero-protective effect still needs to be demonstrated. M4 macrophages differentiate in response to the chemo-attractant CXCL4, thus showing pro-inflammatory and pro-atherogenic effects. These macrophages express low levels of CD163 and high levels of MHCII and CD86. M1 and M4 macrophages promote, while M2 and Mhem macrophages counteract foam cell formation, thus having opposite effect on atherosclerosis progression.

evolve in foam cells, eventually leading to apoptosis and cell content release. Moreover, the release of TNFα, IL-1β, IL-6 and other inflammatory cytokines by M1 macrophages in the lesion environment may contribute to the activation of endothelial cells (increasing the expression of LFA-1, VCAM-1, ICAM-1, CCL2, CD62P, and CD62E) and smooth muscle cells (increasing the expression of CCL2, CCL9, CX3CL1, CXCL10, CXCL16, and VCAM-1) and an overall increase in oxidative stress by the production of reactive oxygen and nitrogen species. In addition, M1 macrophages are associated with the response of Th1 lymphocytes, which is in accordance to an increased inflammatory response (Butcher and Galkina, 2012). All these events are expected to promote atherosclerotic plaque progression.

The alternative M2 macrophages are polarized after exposure to IL-4 or IL-13 and display an anti-inflammatory phenotype. M2-like macrophages have been described in wound healing as well as in association with tumors and with human carotid atherosclerotic plaques (Bouhlel et al., 2007). This subtype of macrophages has enhanced capacity for phagocytosis, tissue remodeling and matrix metalloproteases production (Martinez et al., 2006; Mosser and Edwards, 2008). In contrast to M1, M2 macrophages have higher expression of CD163, HO-1 and FPN and low expression of ferritin, suggesting that these macrophages have an iron release phenotype with increased iron uptake, recycling and export but low iron retention (Recalcati et al., 2010; Cairo et al., 2011). In atherosclerosis, M2 macrophages are mainly found in early lesions and are characterized by the expression of CD68 and mannose receptor (Chinetti-Gbaguidi et al., 2011). They preferentially localize in the area of the plaque overlying the lipid core (Khallou-Laschet et al., 2010). M2 macrophages are less susceptible to become foam cells and they also display a lower ability to handle lipids and to export cholesterol, due to the downregulation of the cholesterol exporter ABCA1 and the LDL carrier apoE. Also upregulation of genes involved in phagocytosis suggests that M2 macrophages in atherosclerosis have an enhanced phagocytic activity by clearing up cellular debris and dead cells (Chinetti-Gbaguidi et al., 2011). M2 macrophages are associated with a Th2 type response (Butcher and Galkina, 2012). These macrophages do not contribute to the activation of endothelial cells or smooth muscle cells since they have anti-inflammatory properties (Kleemann et al., 2008). Altogether—less susceptibility to become foam cells, high phagocytic activity and antiinflammatory properties—place these macrophages as protective for the atherosclerotic lesion development.

Recently, new subtypes of macrophages have been described in the context of atherosclerosis, supporting the idea of an increasing diversity of macrophage subsets within the lesions.

The platelet-derived chemokine CXCL4 promotes the differentiation of monocytes to macrophages toward an M4 macrophage subtype (Gleissner et al., 2010b) (**Figure 3**). There is no doubt that CXCL4 is important for atherosclerosis since the deletion of the PF4 gene that encodes CXCL4 reduces atherosclerotic lesions in apoE deficient mice (Sachais et al., 2007). M4 macrophages display a distinct transcriptome when compared to M1 and M2 macrophages. The major characteristic of this subtype relies on the downregulation of CD163, the Hp-Hb scavenger receptor, which indicates that M4 macrophages are not able to clear Hb after plaque hemorrhage (Gleissner et al., 2010a,b). The incapacity of Hb uptake is consistent with the absence of HO-1 upregulation which has a protective and anti-inflammatory effect in atherosclerotic lesion (Gleissner, 2012). This also might have some implication for iron handling but further studies are necessary to characterize this macrophage subtype regarding iron turnover. In addition M4 macrophages showed reduced expression of cholesterol scavenger receptors, leading to a decreased ability to clear modified LDL. Immunohistochemistry of human post-mortem coronary arteries revealed the presence of CD68+ CD163+ as well as CD68+ CD163− macrophages, showing a correlation in the expression levels of CD163 and CXCL4 (Gleissner et al., 2010a). Whether this macrophage subtype promotes or protects against atherosclerotic plaque progression still needs to be addressed. On the basis of their reduced Hb clearance ability, a detrimental role of M4 macrophages in atherosclerosis could be speculated. Future research will be required to establish whether M4 macrophages represent a promising therapeutic target in human atherosclerosis.

Atherosclerotic lesions are characterized by the accumulation of oxidized phospholipids that also play a role in macrophage polarization. A novel macrophage phenotype denominated Mox macrophages was identified in the lesions of mice deficient for the LDL receptor (Kadl et al., 2010) (**Figure 3**). Mox macrophages were identified in atherosclerotic plaques in mice and accounted for 30% of all CD11b+/F4/80+ cells in established lesions. *In vitro* treatment of bone marrow-derived macrophages (BMDMs) with oxidized phospholipids reproduced differentiation toward this macrophage subtype that is distinct from both M1 and M2 subtypes. Considering the pro-oxidant action of iron, increased iron levels would be expected to enhance lipid oxidation, thus promoting Mox polarization. Whether this occurs in conditions of body iron overload has not been demonstrated.

Mox macrophages show a characteristic expression profile, including the upregulation of HO-1, thioredoxin reductase1 and sufiredoxin-1, whose expression is dependent on the redoxsensitive transcription factor Nrf2 (Kadl et al., 2010; Butcher and Galkina, 2012). These Mox-specific genes may have important functions in controlling oxidative status in an oxidizing environment and protecting cells from dying in oxidatively damaged tissue. It was demonstrated that failure of Nrf2 expression leads to various diseases related to oxidative stress, inflammation, and xenobiotic metabolism in mice. Based on these findings, a protective role of Nrf2-driven Mox macrophages in atherogenesis would be expected. Surprisingly, a recent study showed that Nrf2-null mice were protected against diet-induced atherosclerosis. Whether these Nrf2-driven Mox macrophages contribute to the initiation or progression of atherosclerotic lesion formation remains to be investigated.

Intraplaque hemorrhage is one of the key events in advanced atherosclerotic lesions leading to iron accumulation and increased oxidative stress, thus contributing to lesion development. Erythrophagocytosis is an important source of iron in plaqueassociated macrophages and increased ferritin correlates with macrophage infiltration in human atheroma (Yuan et al., 1996). The recent description of hemorrhage-associated macrophages in atherosclerotic lesions further confirmed that hemorrhagederived Hb is a source of iron for intraplaque macrophages and directs their polarization into a specialized phenotype, able to handle high Hb/iron amount (Boyle et al., 2009; Finn et al., 2012). These macrophages show high Hb handling capacity and antiatherogenic properties and were named Hemorrhage-associated macrophages (HA-mac), Hb-stimulated macrophages, M(Hb) or heme-directed macrophages (M-hem) (**Figure 3**).

Macrophages associated with hemorrhage areas were characterized as CD163 high and HLA-DRlow (Boyle et al., 2011b). Moreover, as a consequence of enhanced Hb clearance, HA-mac macrophages have increased HO-1 and FPN expression, leading to facilitated heme catabolism and reduced intracellular free iron. Thus, they show antioxidative characteristics, increased expression of cholesterol exporters and resist foam cell formation both *in vivo* and in response to cholesterol loading. The reduction in intracellular free iron available for ROS formation causes increased expression of cholesterol exporters, via the activation of the LXRs (liver X receptors) pathways. HA-mac macrophages are distinct from the macrophages found in the lipid core and seem to play an atheroprotective role. *In vitro* stimulation of monocytes with Hb-Hp complexes showed a differentiation toward an HAmac phenotype, suggesting that Hb released upon hemorrhage might model monocytes recruited to the lesion toward a specific HA-mac subtype (Boyle et al., 2011b). After treatment of human blood monocytes with heme HO-1and CD163 are upregulated, a process depending onNrf2 and the activating transcription factor 1 (Boyle et al., 2011a). Altogether, these findings suggest that ironspared macrophages may have a protective role, as postulated by Sullivan, and that the pharmacological manipulation of iron homeostasis may be a promising target to increase macrophage reverse cholesterol transport, thus limiting atherosclerosis.

Mhem macrophages exemplify how iron can affect macrophage differentiation and function, in such a way that they can handle large amounts of Hb and iron, thus limiting ironmediated oxidative effects and preventing lesion progression.

In atherosclerosis, macrophage activity and iron metabolism might be intrinsically connected. It is interesting to note that macrophage polarization is driven according to the specific microenvironment of the atherosclerotic lesion. The description of the different macrophage subtypes reported above suggests that also iron, in the form of Hb or via LDL oxidation, can differentially affect macrophage polarization. How broad is the range of macrophage subtypes generated in response to iron and how these subtypes contribute to atherosclerosis progression is not clear yet. Further studies are required to estimate the contribution of different iron sources to macrophage polarization and their impact on the atherosclerosis process.

## **THE EFFECT OF IRON ON VSMC PHENOTYPE SWITCH**

VSMC are the predominant cell type of the medial layer of the vessel wall. Under physiological conditions, VSMC show high contractility and a low proliferation rate. These properties are essential for VSMC to perform its primary function, contraction and dilatation of vessels to regulate blood pressure and flow. However, VSMC are not terminally differentiated cells but show the capacity to switch to synthetic, inflammatory, osteochondrogenic or macrophage-like, phenotypes upon certain stimuli. The synthetic phenotype is characterized by loss of contractility, increased motility and high proliferation rate (Campbell and Campbell, 1985). Synthetic VSMC are involved in fibrous cap formation during atherogenesis. Inflammatory VSMC phenotype is defined by cytokine secretion (e.g., IL-8, IL-6) and cell adhesion molecule expression (e.g., VCAM-1), that can regulate monocyte/macrophage adhesion and recruitment (Orr et al., 2010). Under certain pathological condition, VSMCs can undergo phenotypic transition into osteoblastlike cells, whereby they synthesize excessive extracellular matrix with parallel loss of their original function (Jono et al., 2000; Giachelli et al., 2001; Giachelli, 2003), reviewed in Sallam et al. (2013). Osteoblast specific markers are present in calcified atherosclerotic lesions, highlighting the relevance of these events in atherosclerosis (Dhore et al., 2001; Engelse et al., 2001). Finally, VSMC can differentiate into macrophage-like cells. These cells are enlarged and characterized by lipid inclusions in the cytoplasm with immunoreactivity to α-smooth muscle actin and vimentin, specific markers of VSMC. These cells are present in human atherosclerotic lesions (Vukovic et al., 2006)

Some effort was made to study the effect of iron on the phenotype switching of VSMC. Iron chelation by desferoxamine (DFO) significantly inhibited VSMC proliferation, a hallmark of the synthetic phenotype *in vitro* (Porreca et al., 1994; Wong et al., 2012), although opposing results show that iron decrease VSMC growth (Mueller et al., 2006). Iron chelation inhibits the pathological vascular remodeling response induced by balloon injury and pulmonary hypertension (Porreca et al., 1994; Wong et al., 2012). Accumulating evidence indicates that heme, and in particular, products of heme catabolism by HO-1 regulate VSMC growth (reviewed in Durante, 2003). Carbon monoxide directly inhibits VSMC proliferation by arresting cells in the G0/G1 phase of the cell cycle, whereas biliverdin and bilirubin induce VSMC apoptosis (Morita et al., 1997; Liu et al., 2002; Peyton et al., 2002).

Recently, by studying the effect of iron on osteochondrogenic differentiation of VSMC, iron was reported to inhibit inorganic phosphate (Pi)-mediated osteoblastic transition and subsequent mineralization of VSMCs *in vitro* (Zarjou et al., 2009). Importantly, iron inhibited the Pi-mediated increase in the expression of core binding factor-1 (Cbfa-1), the key osteoblastspecific transcription factor orchestrating the production of osteoblast-specific proteins, such as alkaline phosphatase and osteocalcin (Zarjou et al., 2009). Ferritin was identified as the major protective molecule behind iron-mediated inhibition of mineralization. The inhibitory effect of ferritin is strictly dependent on its ferroxidase activity but not on its iron-storage ability (Zarjou et al., 2009). Although a direct evidence of a role for iron in calcification *in vivo* is lacking, recently it has been described that iron and calcium show a highly significant spatial inverse correlation within the atherosclerotic lesions (Rajendran et al., 2012).

Although increasing evidence suggests the critical role of VSMC phenotype switch in atherogenesis, the role of iron in these mechanisms still remains to be elucidated. Further *in vitro* and *in vivo* studies are essential to clarify the particular role of iron in differentiation of VSMC into synthetic, inflammatory, osteochondrogenic, or macrophage-like phenotypes.

## **CONCLUSIVE REMARKS**

Over the last 30 years, several studies in animals and humans assessed the effect of increased body iron levels on atherosclerosis, yielding conflicting results. In the last decade, our understanding of Hb and iron biology underwent a radical revision. This significantly helped in understanding the atherogenic effects of iron and iron-containing molecules. Numerous experiments support the idea that oxidized Hb, Heme, and iron—by interacting with plaque lipids, promoting endothelial dysfunction, dictating macrophage polarization, modulating VSMC phenotype and proliferation—may affect the atherogenic process. Complex systems have evolved to control and dispose cell-free Hb, heme, and iron but these systems may be eventually overwhelmed upon excessive hemorrhage or hemolysis and upon pathological iron overload. However, to date, the impact of iron on atherosclerosis is still debated. Future studies are required to clearly address whether iron overload is a risk factor for atherosclerosis and what iron source - systemic, tissue or macrophage iron—mainly affects the atherosclerotic process. Comprehensive understanding the role of iron on atherogenesis may lead to the development of improved diagnostics and therapeutics meant to interrupt the pathologic actions of excess iron.

## **AUTHOR CONTRIBUTIONS**

All authors contributed to the conception and design of this review. Viktória Jeney, Francesca Vinchi, and Milene Costa Da Silva wrote the manuscript, designed and made the figures. The manuscript was critically revised by Martina U. Muckenthaler, József Balla, and György Balla.

## **ACKNOWLEDGMENTS**

The research group is supported by the Hungarian Academy of Sciences (11003). This work was supported by Hungarian Government grants OTKA-K75883 (György Balla), OTKA-K83478 (József Balla), OTKA- PD83435 (Viktória Jeney), European Reintegration Grant FP7-PEOPLE-2010-268332 (Viktória Jeney), and by the TÁMOP 4.2.2.A-11/1/KONV-2012- 0045 projects. The project is co-financed by the European Union and the European Social Fund. This work was supported by a Postdoctoral-Fellowship granted to Francesca Vinchi from the Medical Faculty of the University of Heidelberg, Germany (http://www*.*medizinische-fakultaet-hd*.*uni-heidelberg*.*de).

## **REFERENCES**


and between atherosclerotic plaques. *Magn. Reson. Med*. 71, 885–892. doi: 10.1002/mrm.24687


response to atherogenic phospholipids via Nrf2. *Circ. Res.* 107, 737–746. doi: 10.1161/CIRCRESAHA.109.215715


dysfunction and increased cardiovascular risk among hemodialysis patients. *PLoS ONE* 7:e50295. doi: 10.1371/journal.pone.0050295


low-density lipoprotein (LDL) and increased cytotoxic effect by LDL oxidation in heme oxygenase-1 (HO-1) deficiency. *Cell. Mol. Biol.* 51, 377–385.


patients with peripheral arterial disease: a randomized controlled trial. *JAMA* 297, 603–610. doi: 10.1001/jama.297.6.603


Zohn, I. E., De Domenico, I., Pollock, A., Ward, D. M., Goodman, J. F., Liang, X., et al. (2007). The flatiron mutation in mouse ferroportin acts as a dominant negative to cause ferroportin disease. *Blood* 109, 4174–4180. doi: 10.1182/blood-2007-01-066068

**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

*Received: 25 February 2014; paper pending published: 23 March 2014; accepted: 14 April 2014; published online: 05 May 2014.*

*Citation: Vinchi F, Muckenthaler MU, Da Silva MC, Balla G, Balla J and Jeney V (2014) Atherogenesis and iron: from epidemiology to cellular level. Front. Pharmacol. 5:94. doi: 10.3389/fphar.2014.00094*

*This article was submitted to Drug Metabolism and Transport, a section of the journal Frontiers in Pharmacology.*

*Copyright © 2014 Vinchi, Muckenthaler, Da Silva, Balla, Balla and Jeney. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*

## The role of iron metabolism as a mediator of macrophage inflammation and lipid handling in atherosclerosis

## *Anwer Habib and Aloke V. Finn\**

Division of Cardiology, Department of Medicine, Emory University School of Medicine, Atlanta, GA, USA

#### *Edited by:*

Raffaella Gozzelino, Instituto Gulbenkian de Ciência, Portugal

#### *Reviewed by:*

Jozsef Balla, University of Debrecen – Medical and Health Science Center, Hungary Joseph J. Boyle, Imperial College London, UK

#### *\*Correspondence:*

Aloke V. Finn, Division of Cardiology, Department of Medicine, Emory University School of Medicine, 101 Woodruff Circle, WMB 319B, Atlanta, GA 30322, USA e-mail: avfinn@emory.edu

Iron is an essential mineral needed for normal physiologic processes. While its function in oxygen transport and other important physiologic processes is well known, less is understood about its role in inflammatory diseases such as atherosclerosis. Existing paradigms suggest iron as a driver of atherosclerosis through its actions as a pro-oxidant capable of causing lipid oxidation and tissue damage. Recently we and others have identified hemoglobin (Hb) derived iron as an important factor in determining macrophage differentiation and function in areas of intraplaque hemorrhage within human atherosclerosis. Hb associated macrophages, M(Hb), are distinct from traditional macrophage foam cells because they do not contain large amounts of lipid or inflammatory cytokines, are characterized by high levels of expression of mannose receptor (CD206) and CD163 in addition to producing anti-inflammatory cytokines such as IL-10. Despite the well-known role of iron as an catalyst capable of producing lipid peroxidation through generation of reactive oxygen species (ROS) such as hydroxyl radical, we and others have shown that macrophages in areas of intraplaque hemorrhage demonstrate reduced intracellular iron and ROS which triggers production of anti-inflammatory cytokines as well as genes involved in cholesterol efflux. These data suggest that manipulation of macrophage iron itself may be a promising pharmacologic target for atherosclerosis prevention through its effects on macrophage inflammation and lipid metabolism. In this review we will summarize the current understanding of iron as it relates to plaque inflammation and discuss how further exploration of this subject may lead to new therapies for atherosclerosis.

**Keywords: iron, macrophages, atherosclerosis, inflammation, lipid metabolism**

## **IRON IN THE VASCULATURE**

Iron is a powerful catalyst resulting in the production of a hydroxyl radical through the oxidation of its ferrous (Fe2+) to ferric form (Fe3+) through the Fenton reaction (Crichton et al., 2002). In the endothelium, heme-derived iron is thought to catalyze oxidation of low density lipoproteins by itself or in conjunction with myeloperoxidase or lipoxygenase located on the endothelial surface (Balla et al., 1991; Miller et al., 1997; Camejo et al., 1998; Jeney et al., 2002; Yoshida and Kisugi, 2010). Hemolysis is often a result between the interaction of erythrocytes and mature atheromas resulting in the transition of ferrous to ferric forms of hemoglobin (Hb) which additionally leads to lipid oxidation (Nagy et al., 2010). Furthermore this oxidized form of Hb can also act as a pro-inflammatory agonist targeting vascular endothelial cells (Silva et al., 2009). Ferritin, a hepatic protein, may counteract some redox activity via ferroperioxidase in the vasculature (Balla et al., 1992), however, overall oxidation can be largely unchecked in these distinct pathologic environments as intraplaque hemorrhage. Oxidized low density lipoproteins (LDL) have a high affinity for the LDL receptor on macrophages leading their development into foam cells. Foam cells provide the major inflammatory component of atherosclerotic plaques. Foam cells density and necrotic core size within atherosclerotic plaques is thought to be a key determinant of plaque vulnerability for rupture (Sakakura et al., 2013). While the role of iron as a pro-oxidant has

been established *in vitro* (Smith et al., 1992; Juckett et al., 1995; Pang et al., 1996; Silva et al., 2009), there is not a clear association of increased serum iron and increased incidence of coronary artery disease (CAD; Miller and Hutchins, 1994). In disease states of iron overload, such as hemochromatosis, an autopsy study found the extent of CAD to be less than the general population while another prospective study of 9000 individuals found carriers of the hemochromatosis genotypes C282Y to not have an increased risk for ischemic heart disease or myocardial infarction (Miller and Hutchins, 1994; Ellervik et al., 2005). Furthermore, dietary iron overload in Apo E−/<sup>−</sup> mice reduces rather than exacerbates the severity of atherosclerosis (Kirk et al., 2001). These experimental data challenge the prevailing idea of iron as a pro-oxidant capable of accelerating coronary artery disease.

## **SYSTEMIC IRON REGULATION AND LINKS TO INFLAMMATION**

The majority of iron needed to regulate normal bodily functions is recycled from senescent red cells by the reticuloendothelial system. Additional demand for iron due to various environmental challenges such as anemia is fine-tuned by adjusting iron absorption via enterocytes. In some disease states, such as hemochromatosis, the regulation of iron is disturbed leading to excess iron entering the body. There are many systems within body that regulate the balance of iron. For the purposes of this review, we will focus on those within the macrophage. The regulation of movement of iron through various organs in the body is critical to maintaining iron homeostasis. Ferroportin (FPN), a transporter which mediates exit of iron from macrophages into the circulation, is an extremely important mechanism for immediate control of available and circulating serum iron. Although regulated at multiple levels, the peptide hormone, hepcidin, is the key regulator of FPN. Hepcidin binds to FPN inducing its internalization and degradation (Nemeth et al., 2004). Hepcidin induced downregulation of FPN thus inhibits cellular iron export from macrophages. The hepcidin-FPN axis is a major regulatory mechanism that maintains iron homeostasis in response to changing requirements. Also known as an acute phase reactant, hepcidin responds to inflammation resulting in adjustments to FPN levels which alters the regulation of body iron status (Ganz, 2003). The importance of this mechanism is observed in hereditary hemochromatosis where often either the expression or function of hepcidin is disturbed. In these situations, FPN is elevated because of low circulating hepcidin levels leading to increased gut iron absorption and pathologic deposition of iron in tissues.

Interestingly, mice deficient in the hemochromatosis gene, Hfe, have attenuated inflammatory responses to bacterial challenge associated with decreased macrophage TNF-α and IL-6 after exposure to the canonical Toll-like receptor 4 agonist lipopolysaccheride (LPS). These data suggest that these animals have impairment in Toll-like receptor 4 (TLR4) signaling (Wang et al., 2009). These defects could be replicated by exposing wild type murine macrophages to iron chelators, suggesting low intracellular iron within Hfe KO macrophage may lead to impaired TLR4 signaling. Thus, these results suggest iron overload in the setting of hemochromatosis may be associated with dampening of inflammation rather than exacerbating it.

## **LOCAL IRON REGULATION BY MACROPHAGES AND LINKS TO ANTI-INFLAMMATION**

In addition to helping to maintain systemic iron homeostasis, macrophages are intimately involved in preventing toxic effects of iron release during events involving hemolysis including in the setting of intraplaque hemorrhage. We and others have previously shown the importance of intraplaque hemorrhage, an event which leads to the deposition of erythrocyte-derived iron, in human atherosclerotic lesions (Kolodgie et al.,2003). In a relatively large number of human coronary plaques from sudden coronary death victims, we observed a greater frequency of previous intraplaque hemorrhages in plaques prone to rupture compared to early lesion morphologies or stable plaques. Hemorrhage itself contributes to the deposition of free cholesterol and enlargement of the necrotic core in atherosclerotic plaques through the accumulation of erythrocyte membranes that are rich in cholesterol. These findings were paralleled by an increase in macrophage density, which supports previous observations that hemorrhage itself is an inflammatory stimulus.

During hemorrhage, the pro-oxidant environment of atherosclerosis promotes erythrocyte lysis and accumulation of free Hb, which, if not eliminated, may cause tissue damage by releasing free iron which increases oxidative stress through the Fenton reaction. During hemolysis, free Hb binds to the plasma protein haptoglobin and hemoglobin:haptoglobin (HH) complexes are formed. CD163, the receptor for this complex, is expressed exclusively on the surface of macrophages and binds to HH, mediating its endocytosis. Conversely the interaction of haptoglobin itself with CD163 is impaired in highly oxidized environment (Vallelian et al., 2008), suggesting a more favorable interaction in the form of HH complexes. The heme subunit of Hb is then degraded by the heme oxygenase (HO-1) enzymes. The HO-1 pathway, which produces anti-oxidants carbon monoxide and biliverdin also releases free iron (Fe2+). Once iron has been released by HO-1, it is either utilized by the cell, stored as ferritin in a redox inactive form, or exported via FPN and converted to less redox active ferric iron (Fe3+) via ceruloplasm. Although the role of HO-1 in atherosclerosis has been studied in detail, an exact understanding of the molecular events in macrophages which orchestrate responses to iron and how this affects macrophage function remains incompletely understood. In addition, because hemorrhage, iron, and macrophages are not infrequently found in advanced atherosclerosis, the findings of these studies have important implications for our understanding of how iron itself event influences this disease.

The macrophage is the major inflammatory cell involved in atherosclerosis progression (Libby, 1995; Ross, 1999). While the role of lipid-rich foam cell macrophages which up-regulate proteolytic enzymes leading to plaque rupture has been extensively studied, less attention has been paid to alternative macrophage phenotypes which exist in atherosclerosis (Libby, 1995). It has been classically thought that macrophages exist in two subtypes: (1)"classically"activated (M1) macrophages, which are induced by Th1 cytokines such as tumor necrosis factor α (TNF-α) and LPS, and (2) alternative M2 cells, stimulated by Th2 cytokines such as IL-4 or IL-13 which produce anti-inflammatory cytokines such as IL-10 (Gordon, 2003).

Studies done by Boyle et al. (2009), in addition to our lab, suggest a third macrophage phenotype [M(Hb) or Mhem], induced by ingestion of HH complexes leading to an anti-inflammatory effect via production of anti-inflammatory cytokines such as IL-10 and production of anti-inflammatory metabolites produced during heme metabolism (Boyle et al., 2009; Finn et al., 2012).

## **CD 163, INTRAPLAQUE HEMORRHAGE, AND MACROPHAGE POLARIZATION**

Boyle et al. (2009, 2011) were the first to explore the effects of intraplaque hemorrhage on macrophage phenotype. Advanced atherosclerotic plaques were examined for immunostaining for CD163 and HLA-DR, a sign of macrophage activation. Macrophages were found to express either CD163 or HLA-DR. The CD163high macrophages were found in areas of intraplaque hemorrhage and displayed evidence of less oxidative damage. This phenotype could be reproduced by exposure of human monocytes to HH complexes. More recently, our lab has expanded this work to demonstrate that macrophages in areas of human coronary intraplaque hemorrhage represent a subtype distinct from foam cells or the previously reported M2 phenotype.

These cells, characterized by high surface mannose receptor (MR, CD206) and CD163, exhibit reduced expression of pro-inflammatory cytokines such as tumor necrosis factor alpha (TNFα), and are devoid of lipids typical of foamy macrophages (**Figure 1**; Finn et al., 2012). The term M(Hb) or Hb associated macrophages (Mhem) was used to refer to this subset since induced by ferrous Hb not IL-4 or hemorrhage (Bouhlel et al., 2007; Boyle et al., 2009). These cells demonstrate a unique iron handling signature associated with activation of the nuclear receptor liver × receptor alpha (LXRα), upregulation of ferroportin (FPN) and CD163. The activation of LXRα in addition to HO-1 was thought to be via oxidative stress from heme release and phosphorylation of activating transcription factor 1 (ATF-1; Boyle et al., 2012). Cultured human monocytes exposed to HH complexes have reduced free intracellular iron and reactive oxygen species (ROS) levels likely due to increased sequestration of iron by ferritin and by increased export of free iron outside the cell via FPN. This reduction in free iron and ROS could be reversed by pre-treating with cells with hepcidin, suggesting the importance of FPN in this effect. Moreover, M(Hb) macrophage demonstrate resistance to lipid loading, lowered expression of genes involved in lipid uptake (i.e., SR-A1, SR-A2, CD36, SR-B1) that characterize foam cells and increased reverse cholesterol through ATP binding cassette (ABC) transporters (i.e., ABCA1, ABCG1) involved

in Apo-A1 cholesterol efflux to high density lipoproteins (HDL; **Figure 2**).

Our work suggests that iron itself does not result in increased oxidative stress and lipid retention with atherosclerotic plaque macrophages. Instead areas of hemorrhage demonstrate the opposite findings with little evidence of oxidative damage as assessed by 8- hydroxyguanine staining and diminished macrophage foam cell formation. To demonstrate the causal effect of lowering intracellular iron in the phenotype of M(Hb) cells, we treated HH differentiated macrophages with hepcidin and found that ABCA1 expression was significantly reduced. Moreover this was associated the downregulation of LXRα activity, a major transcriptional driver or ABCA1. This suggests the importance of macrophage intracellular iron levels driving cholesterol efflux in M(Hb) cells. Additionally differentiation of human macrophages with anti-oxidants such as superoxide dismutase (SOD) increased ABC transporter expression suggesting lowered ROS as a final common trigger for increasing cholesterol efflux. This suggests that manipulation of macrophage iron levels via the hepcidin-FPN axis represents a promising avenue to retard atherosclerosis development via up-regulation of macrophage cholesterol efflux.

#### **FIGURE 1 | Identification of M(Hb) macrophages in an area of hemorrhage in a human coronary fibroatheroma. (A)** Cryosection shows a fibroatheroma with a necrotic core (NC, arrows). Movat pentachrome staining. **(B–I)** represent the area within the black box in "a." **(B)** Accumulation of inflammatory cells in an area of prior hemorrhage adjacent to the NC, H&E. **(E)** Iron (Fe) accumulation near the periphery of the necrotic core. **(D)** identification of macrophages by CD68 shows strong staining within the cell cluster adjacent to the necrotic core. **(E)** Intense staining for

the mannose receptor (MR, CD206) within the cell cluster; note, however, the adjacent necrotic core shows negative staining. **(F)** The same MR positive macrophages within the cluster are also strongly positive for CD163, while the necrotic core remains negative. **(G)** Shows that the same cluster of cells is negative for lipid (ORO) while the adjacent necrotic core is strongly positive. The area of CD206/CD163 positive macrophages does not stain for CD36 **(H)** or TNFα **(I)**. Reproduced from Finn et al. (2012) permission pending.

## **MACROPHAGE DIVERSITY IN HUMAN ATHEROSCLEROSIS – ROLE OF M(Hb) vs. M2 MACROPHAGES**

Recent studies such as those from Chinetti-Gbaguidi et al. (2011) have looked IL-4 induced M2 macrophages in human atherosclerotic plaques. However, unlike M(Hb) where intraplaque hemorrhage provides a precipitant for its differentiation, the source for driving IL-4 remains unclear. Additionally, IL-4 differentiated M2 macrophages demonstrate mannose upregulation but not CD163 and do not demonstrate the same iron handling signature in that they show no increase in FPN expression and minimal changes in HO-1 and ferritin heavy chain (Bories et al., 2013). However, when M2 macrophages were exposed to iron, both FPN, HO-1, and LXRα-dependent genes such as ABCA1 were induced, mimicking the phenotype of M(Hb) macrophages. These data suggest, regardless of the stimulus (Hb or less physiologic FeCl3), iron is an essential factor driving the phenotype found in areas of intraplaque hemorrhage.

Hemoglobin: haptoglobin differentiated macrophages resist exogenous lipid loading to a much greater extent compared to IL-4 differentiated M2 macrophages and are characterized by an entirely different expression pattern of lipid handling genes (Boyle et al., 2012). However, M(Hb) demonstrated reduced expression of the scavenger receptors CD36 and increased expression of cholesterol efflux genes ABCA1/ABCG1, M2 macrophages demonstrate the opposite pattern with increased CD36 expression and reduced expression of ABCA1 and cholesterol efflux (Chinetti-Gbaguidi et al., 2011). Furthermore, a microarray analysis of 2400 genes showed a distinct gene transcriptome signature of M(Hb) versus M2 macrophages (Boyle et al., 2012).

Our work suggests that liver x receptor alpha (LXRα), an inducible transcription factor known to be important in human macrophage ABC transporter transcription, may play a central role in the response to heme-derived iron ingestion. LXRα can be activated by oxysterols which can also be produced by iron loading. The works of Bories et al. (2013) indicates LXRα appears to direct the upregulation of FPN and the repression of hepcidin, a protein which inhibits iron transport out of macrophages by degrading FPN. LXRα is likely a critical mediator of iron responses in macrophages especially M(Hb) with roles in lipid handling and inflammatory responses through transcriptional control of FPN/hepcidin.

## **HEPCIDIN-FPN AXIS: MODULATION OF MACROPHAGE DIVERSITY TO IMPROVE ATHEROSCLEROTIC PROGRESSION**

Given the link between macrophage the hepcidin→FPN axis, macrophage intracellular iron and the atheroprotective phenotype of M(Hb) we examined the effect of inhibitors of hepcidin on macrophage lipid metabolism (Yu et al., 2008; Saeed et al., 2012). Bone morphogenic protein-1 (BMP-1) signaling is involved in hepcidin gene transcription via SMAD 1/5/8 phosphorylation (Yu et al., 2008). BMP-1 inhibitors, such as dorsomorphin, and LDN, potently inhibit hepcidin production by blocking BMP-1 receptors, ALK 2/3/6 preventing its downstream effects on SMAD (Boergermann et al., 2010; Saeed et al., 2012). Effects of this BMP-1 inhibition on macrophage polarization lead to increased ABCA1/G1 expression, decreased cytokine and ROS production and increased FPN production (Saeed et al., 2012). These effects again were mitigated through hepcidin repletion (Saeed et al., 2012). Interestingly, LDN treatment delayed atherosclerotic progression in transgenic ApoE knockout mice and increased serum iron suggesting a potent effect in reducing intracellular iron content and plaque progression (Saeed et al., 2012). It must be stated, however, that inhibition of BMP signaling could reduce atherosclerosis via additional mechanisms not explored by us Derwall et al. (2012). However, the long-term effects of such manipulations which increase serum and likely tissue iron via up-regulation of FPN remains unclear. Given the pivotal role of hepcidin in regulating iron homeostasis, its chronic inhibition could potentially result in an iron overload-like state, which may limit the actual clinical adoption of such as strategy.

Further support for our data come from others work which has shown shown that overexpression of hepcidin both *in vitro* and *in vivo* murine ApoE carotid plaque model increases plaque instability especially in the setting of macrophage iron loading (Li et al., 2012). AdditionallyWang et al. (2009) demonstrated that similarly targeted inhibitors of BMP signaling significantly attenuated infectious and non-infectious enterocolitis in a mouse model, again reinforcing the anti-inflammatory effect of this strategy which may be mediated in part through TLR4 inhibition (Wang et al., 2009). Given these findings, it suggests that the hepcidin-FPN axis is an important modulator of inflammation and determinant of macrophage polarization.

## **CONCLUSION**

Our knowledge of the effects of iron on inflammation and atherosclerosis continues to evolve. Recent studies on human atherosclerosis demonstrate that areas of intraplaque hemorrhage where iron is abundant demonstrate reduced ROS, tissue damage, lipid retention and inflammation. These data challenge existing paradigms that iron is a catalyst capable of producing ROS which accelerates atherosclerosis. Our data point to an important role for LXRα, FPN, hepcidin in controlling macrophage iron levels and thereby determining these cells lipid handing and inflammatory potential. These studies suggest that approaches to reduce intracellular macrophage iron that involve downregulation of hepcidin either directly (i.e., via shRNA) or indirectly (i.e., BMP-1 inhibitors) and may present a therapeutic benefit for advanced atherosclerotic lesions and perhaps other inflammatory conditions. However, given side effects that would occur by interfering with the FPN/hepcidin axis, more investigation is necessary to define this strategy of local modulation of inflammation to prevent atherosclerosis progression.

#### **REFERENCES**


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

*Received: 26 May 2014; paper pending published: 20 June 2014; accepted: 04 August 2014; published online: 27 August 2014.*

*Citation: Habib A and Finn AV (2014) The role of iron metabolism as a mediator of macrophage inflammation and lipid handling in atherosclerosis. Front. Pharmacol. 5:195. doi: 10.3389/fphar.2014.00195*

*This article was submitted to Drug Metabolism and Transport, a section of the journal Frontiers in Pharmacology.*

*Copyright © 2014 Habib and Finn. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*

## H-ferritin ferroxidase induces cytoprotective pathways and inhibits microvascular stasis in transgenic sickle mice

*Gregory M. Vercellotti 1,2 \*, Fatima B. Khan1,2 , Julia Nguyen1,2 , Chunsheng Chen1,2 , Carol M. Bruzzone1,2 , Heather Bechtel <sup>3</sup> , Graham Brown1,2 , Karl A. Nath4 , Clifford J. Steer <sup>5</sup> , Robert P. Hebbel1,2 and John D. Belcher 1,2*

<sup>1</sup> Division of Hematology, Oncology and Transplantation, Department of Medicine, University of Minnesota Medical School, Minneapolis, MN, USA

<sup>2</sup> Vascular Biology Center, Department of Medicine, University of Minnesota Medical School, Minneapolis, MN, USA

<sup>3</sup> Mercy Clinic Children's Cancer and Hematology, St. Louis, MO, USA

<sup>4</sup> Division of Nephrology and Hypertension, Department of Medicine, Mayo Clinic/Foundation, Rochester, MN, USA

<sup>5</sup> Division of Gastroenterology, Hepatology and Nutrition, Department of Medicine, University of Minnesota Medical School, Minneapolis, MN, USA

#### *Edited by:*

Paolo Arosio, University of Brescia, Italy

#### *Reviewed by:*

Raffaella Gozzelino, Instituto Gulbenkian de Ciência, Portugal Leo Otterbein, Beth Israel Deaconess Medical Center–Harvard Medical School, USA

#### *\*Correspondence:*

Gregory M. Vercellotti, Division of Hematology, Oncology and Transplantation–Vascular Biology Center, Department of Medicine, University of Minnesota Medical School, 420 Delaware Street SE, MMC 480, Minneapolis, MN 55455, USA

e-mail: verce001@umn.edu

Hemolysis, oxidative stress, inflammation, vaso-occlusion, and organ infarction are hallmarks of sickle cell disease (SCD). We have previously shown that increases in heme oxygenase-1 (HO-1) activity detoxify heme and inhibit vaso-occlusion in transgenic mouse models of SCD. HO-1 releases Fe2<sup>+</sup> from heme, and the ferritin heavy chain (FHC) ferroxidase oxidizes Fe2<sup>+</sup> to catalytically inactive Fe3<sup>+</sup> inside ferritin. FHC overexpression has been shown to be cytoprotective. In this study, we hypothesized that overexpression of FHC and its ferroxidase activity will inhibit inflammation and microvascular stasis in transgenic SCD mice in response to plasma hemoglobin. We utilized a Sleeping Beauty (SB) transposase plasmid to deliver a human wild-type-ferritin heavy chain (wt-hFHC) transposable element by hydrodynamic tail vein injections into NY1DD SCD mice. Control SCD mice were infused with the same volume of lactated Ringer's solution (LRS) or a human triple missense FHC (ms-hFHC) plasmid with no ferroxidase activity. 8 weeks later, LRS-injected mice had ∼40% microvascular stasis (% non-flowing venules) 1 h after infusion of stroma-free hemoglobin, while mice overexpressing wt-hFHC had only 5% stasis (p < 0.05), and ms-hFHC mice had 33% stasis suggesting vascular protection by ferroxidase active wt-hFHC.The wt-hFHC SCD mice had marked increases in splenic hFHC mRNA and hepatic hFHC protein, ferritin light chain (FLC), 5-aminolevulinic acid synthase (ALAS), heme content, ferroportin, nuclear factor erythroid 2-related factor 2 (Nrf2), and HO-1 activity and protein. There was also a decrease in hepatic activated nuclear factorkappa B (NF-κB) phospho-p65 and vascular cell adhesion molecule-1 (VCAM-1). Inhibition of HO-1 activity with tin protoporphyrin demonstrated HO-1 was not essential for the protection by wt-hFHC. We conclude that wt-hFHC ferroxidase activity enhances cytoprotective Nrf2-regulated proteins including HO-1, thereby resulting in decreased NF-κB-activation, adhesion molecules, and microvascular stasis in transgenic SCD mice.

**Keywords: H-ferritin, sickle cell disease, inflammation, endothelium, vaso-occlusion**

## **INTRODUCTION**

Heme-driven oxidative stress and inflammation play critical roles in vaso-occlusion, endothelial cell dysfunction and chronic vasculopathy in SCD (Kato et al., 2007, 2009; Belcher et al., 2010a; Hebbel, 2011; Ghosh et al., 2013). Sickle red blood cells hemolyze, releasing hemoglobin into the vasculature, which, when oxidized to methemoglobin, can release toxic heme that promotes oxidative stress and inflammation (Jia et al., 2007; Mollan and Alayash, 2013; Schaer et al., 2013; Belcher et al., 2014). Normally, plasma hemoglobin and heme are removed safely by haptoglobin and hemopexin, but these cytoprotective scavenging proteins are depleted in the plasma of SCD patients and mice (Muller-Eberhard et al., 1968; Wochner et al., 1974; Mollan and Alayash, 2013;Vinchi et al., 2013; Belcher et al., 2014).

Transgenic sickle mice have vascular inflammation which is critical for blood cell adhesion and vaso-occlusion (Belcher et al., 2003; Kaul et al., 2004; Zhang et al., 2013). We have recently shown that free heme can activate endothelium *in vivo* and *in vitro* via the toll-like receptor 4 (TLR-4) causing Weibel-Palade body exocytosis with expression of P-selectin and von Willebrand factor on their surfaces and activation of the pro-inflammatory transcription factor NF-κB (Belcher et al.,2014). Supplemental haptoglobin or hemopexin can prevent endothelial activation and Hb/hemeinduced vaso-occlusion (Schaer et al., 2013; Belcher et al., 2014). Detoxification of heme requires HO, either the inducible HO-1 or the constitutive HO-2 (Otterbein et al., 2003; Maines and Gibbs, 2005). We have shown that although HO-1 is increased in sickle patients and mice, pharmacologic or gene therapy augmentation of HO-1 activity provides protection against inflammation and vaso-occlusion in sickle mice (Nath et al., 2001; Jison et al., 2004; Belcher et al., 2006, 2010b). HO-1 degrades heme releasing Fe2+, carbon monoxide, and biliverdin/bilirubin. We and others have

shown that CO, either inhaled or delivered by hemoglobin as MP4CO, has salutary effects in sickle mice and in human endothelial cells *in vitro* (Beutler, 1975; Belcher et al., 2006, 2013; Beckman et al., 2009). Biliverdin/bilirubin has marked antioxidant and antiinflammatory effects, both of which are observed in SCD (Belcher et al., 2006; Kapitulnik and Maines, 2009).

Ferritin, an iron storage protein, plays an important role in iron and heme-catalyzed oxidative damage. Ferritin is composed of 24 subunits of two types: heavy (H) and light (L; Harrison and Arosio, 1996). Ferritin heavy chain (FHC) with its ferroxidase activity detoxifies heme and protects cells against heme and redox-active iron (Levi et al., 1988; Epsztejn et al., 1999; Cozzi et al., 2000; Arosio and Levi, 2002). Released Fe2<sup>+</sup> from heme is oxidized via FHC ferroxidase activity, and safely stored as the catalytically inactive Fe3+. For several years, we have considered ferritin a cytoprotective antioxidant stratagem of the endothelium. In fact, we originally reported that FHC ferroxidase that takes up iron can protect endothelial cells against oxidative injury whereas ferroxidase-nullferritin does not (Balla et al.,1992,2007). Multiple investigators have shown, *in vitro* and *in vivo*, that FHC can protect cells and organs (Epsztejn et al., 1999; Cozzi et al., 2000; Berberat et al., 2003; Xie et al., 2005). A recent review couples heme and iron metabolism via FHC with the pathogenesis of systemic infections and inflammation through control of labile pro-oxidant iron (Gozzelino and Soares, 2014).

The ratio of ferritin H and L subunits is tissue specific and affects iron storage and availability; this differential expression may influence heme-catalyzed oxidative damage (Harrison and Arosio, 1996). In SCD, iron overload due to enhanced iron absorption, hemolysis, and red blood cell transfusions leads to multi-organ dysfunction (Walter et al., 2009; Porter and Garbowski, 2013; Ware and Kwiatkowski, 2013). Serum ferritin, which primarily reflects apo-light chain, is increased in SCD patients and correlates with total body iron burden. Yet, for unknown reasons, levels of ferritin are insufficient to handle the catalytic heme-iron burden in SCD patients. We posited that selectively increasing FHC ferroxidase activity would provide a cytoprotective mechanism in SCD mice. However, effective means of influencing these mechanisms *in vivo* are lacking. Ferritin-inducing reagents such as heme and iron, as well as application of recombinant FHC, have been shown to protect endothelial cells from heme-peroxide challenge in cell culture (Balla et al., 1992, 2000; Lin and Girotti, 1997; Lanceta et al., 2013). Ferritin levels are controlled by cellular iron levels through a post-translational interaction with iron-response proteins 1 and 2 (IRP-1 and IRP-2), releasing these proteins from iron-binding response elements on ferritin mRNA (Rouault, 2006; Wang and Pantopoulos, 2011). Overexpression of FHC ferroxidase through transfection of a tetracycline responsive promoter or through an adenovirus had cytoprotective effects in cultured endothelial, HeLa and L929 cells; and in rat livers subjected to ischemia/reperfusion injury (Cozzi et al., 2000; Berberat et al., 2003; Xie et al., 2005). Since ischemia/reperfusion physiology underpins the pathogenesis of SCD, we hypothesized that overexpression of FHC with ferroxidase activity will attenuate hemoglobin-mediated vasoocclusion in mouse models of SCD (Hebbel et al., 2009). We utilized a novel non-viral delivery system, *SB* transposase, for

wild type (wt)-FHC and ferroxidase-null missense (ms)-FHC expression in sickle mice as previously described (Belcher et al., 2010b).

## **MATERIALS AND METHODS**

## **CONSTRUCTION OF** *SLEEPING BEAUTY* **(***SB***) TRANSPOSASE/wt-hFHC AND ms-hFHC**

Sticky ended directional cloning of wt human fTH-1 was completed by use of linker PCR primers. Human sequence gene specific PCR primers were designed to include unique restriction endonuclease sites also found in the pORF5 MCS and six additional nucleotides 5 to either the start or stop codons. Total RNA was purified from a human umbilical endothelial cell preparation. Gene specific reverse transcription was followed by five cycles of touch down PCR followed by 37 cycles of PCR using the two gene specific primers (Titan one Tube, Roche). PCR product of ∼584 bp was verified by agarose gel electrophoresis. Ferroxidasenull 3 ms fTH-1 (E62K, H65G, and K86Q) was created using the same primers and a PCR target template containing the 3 ms coding region (kind gift of Dr. Paolo Arosio, Università di Brescia) and a standard PCR reaction. The 3 ms fTH-1 (ms-hFHC) was used as a negative control for ferroxidase activity. pORF5-MCS (InvivoGen) and the PCR products were digested with the matching restriction endonucleases to create direction specific sticky ends. The vector and fTH-1 PCR products were cleaned to remove residual restriction endonucleases, multi cloning site fragment, and primer tail trim by size exclusion with Qiagen PCR clean up. The PCR amplified fTH-1 and pORF-MCS were ligated at 3:1 molar ends ratio overnight using T4 ligase (New England Biolabs) and then transformed into competent cells. Colonies present on ampicillin selection bacterial plates were screened for gain-of-mass and restriction mapping on agarose gel electrophoresis, ligation gap PCR reactions, and DNA was confirmed by sequencing.

pORF-MCS/human fTH-1 plasmid DNA was digested with unique restriction endonucleases AsiSi and SwaI (NEB) to release the, hEF1-eIF4g prom/human fTH-1/SV40pAn cassette from the plasmid basic elements. The hEF1-eIF4g prom/human fTH-1/SV40pAn was purified using agarose gel mass separation and PCR clean up (Qiagen), and then blunted with Klenow reaction. A blunt opening between the IR/DR sequences of the SB100X plasmid was created at the unique restriction endonuclease EcoRV site. The digested SB100X plasmid was phosphatased to reduce self religation. Blunt hEF1-eIF4g prom/human fTH-1/SV40pAn and blunt pKT2/meIF/ SB100X/pAn were ligated at 3:1 molar ends ratio overnight at 16◦C and transformed into competent *E. coli*. Colonies present on kanamycin selection bacterial plates were screened for gain of mass and DNA was confirmed by Sanger sequence.

## wt−hFHC ... 61HE**E**RE**H**AEKLMKLQNQRGGRIFLQDI**K**KPD ms−hFHC ... 61HE**K**RE**G**AEKLMKLQNQRGGRIFLQDI**Q**KPD

## **MICE**

All animal experiments were approved by the University of Minnesota's Institutional Animal Care and Use Committee. NY1DD mice were chosen as our sickle mouse model (Fabry et al., 1992). Equal numbers of males and females were obtained and were infused as described below at ages 8–17 weeks. The mice were housed in specific pathogen-free housing to minimize common infectious sources of inflammation and fed standard chow diet.

## **GENE TRANSFER INTO MICE**

DNA (25 μg) was diluted into 0.1 ml LRS/g body weight of recipient mouse, with a maximum of 2.5 ml. DNA given was either the *SB*-wt-fTH-1 or the *SB*-ms-fTH-1 DNA plasmid as described above. Vehicle control mice were injected with LRS alone. Mice were anesthetized and DNA was injected via tail veins over a course of 5–6 s. Further experiments were conducted 8 weeks after infusion to allow stable expression of resultant proteins (Belcher et al., 2010b).

## **MEASUREMENT OF VASCULAR STASIS**

Dorsal skin fold chambers were placed as described previously (Kalambur et al., 2004; Belcher et al., 2005). Mice recovered for three days to allow healing. At baseline, mice were anesthetized and free-flowing vessels within the visible field were identified via intravital microscopy and mapped. After baseline selection of flowing venules, mice were infused via tail vein with stroma-free hemoglobin (8μmol/kg; a gift from Dr. Mark Young, Sangart Inc.) to simulate a hemolytic event. The same venules were re-examined 1 and 4 h after hemoglobin infusion. Venules with no observable blood flow were counted as static. The percentage of static vessels was calculated by dividing the number of static venules at 1 or 4 h by the total number of venules examined.

#### **HARVEST OF ANIMAL TISSUES**

Blood samples and organs were then harvested 4 h after hemoglobin infusion as described previously (Belcher et al., 2005). Mice were asphyxiated in CO2 and blood was collected via terminal cardiac puncture. Organs were removed and processed for immunohistochemistry, RNA analysis, and homogenate preparation. Portions for immunohistochemistry were immediately dropped in phosphate-buffered formalin, while other portions were wrapped in foil, frozen in liquid nitrogen, and stored at <sup>−</sup>85*o*C.

#### **mRNA ANALYSIS**

Total RNA was isolated from frozen spleen sections, and human FHC and mouse HO-1 mRNAs were quantified using probe based quantitative PCR against murine GAPDH as a reference gene (Roche Applied Sciences Universal Probe Library).

## **WESTERN BLOT ANALYSIS OF LIVER TISSUE FOR FHC, FLC, HO-1, Bach-1, Nrf2, NF-κB, VCAM-1, ALAS, AND FERROPORTIN**

An equal amount of protein (30 μg) from liver microsomes, nuclei, or cytosol was loaded into lanes in an SDS buffer and subjected to electrophoresis on 10 or 15% polyacrylamide gels (Bio-Rad) as previously described (Belcher et al., 2005, 2013, 2014). The samples were transferred to polyvinylidene fluoride membranes (Millipore) via electrophoresis. The membranes were probed with rabbit primary antibodies against human FHC (Origene, #TA301280), mouse HO-1 (Stressgen, #OSA111), mouse NF-κB p65 (Cell Signaling #3034) mouse NF-κB phospho-p65

(Cell Signaling, #3031), mouse 5-aminolevulinic acid synthase (ALAS; GeneTex, #GTX104139), mouse ferroportin (Novus Biologicals, #NBP1-21502), mouse ferritin light chain (FLC; Origene, #TA307874), mouse VCAM-1 (Abcam, #174279), and mouse GAPDH (Sigma, #G9495). Sites of binding were visualized via the appropriate secondary IgG conjugated to horseradish peroxidase (Santa Cruz). Final detection of bands was done with ECFTM substrate (GE Healthcare) and read on a StormTM Reader (GE Healthcare). Membranes were stripped using Restore Stripping Buffer (Thermo Scientific) and re-probed as described above.

#### **HO-1 ACTIVITY IN LIVER MICROSOMES**

HO-1 activity was measured as previously described from fresh liver microsome preparations (Belcher et al., 2010b).

#### **HEME CONTENT OF LIVER MICROSOMES**

Heme content was measured via the pyridine hemochromogen method (Balla et al., 1991). Briefly, liver microsomes were combined with pyridine in microcentrifuge tubes, incubated and then sodium hydroxide was added to each. Aliquots were pipetted in quadruplicate into microtiter plates. Fresh sodium hydrosulfite was added to duplicates, and potassium ferricyanide to other duplicates. The plate was read at 557 and 541 nm, the difference of the two was calculated for the oxidized and reduced samples, and the final concentration was determined by the difference between the two, expressed as μmols heme/mg protein.

### **HEK-293 CELL TRANSFECTION**

Human embryonic kidney (HEK-293) cells (ATCC CRL-1573) were grown in 10% FBS DMEM high glucose (Life Technologies) with 500 uM sodium pyruvate (Life Technologies), 4 mM L-Glutamine (Life Technologies), 1.5 g/L sodium bicarbonate (Sigma) and 1% antibiotic (Life Technologies). For transfection, cells were seeded overnight on glass chamber slides and transfected at 70% confluence, 2 h prior to transfection cells were washed twice with 5% FBS DMEM high glucose (Life Technologies) with 500 μM sodium pyruvate, 4 mM L-Glutamine, 1.5 g sodium bicarbonate without antibiotic. Cells were transfected with 0.1 μg of DNA/cm<sup>2</sup> with human wt- and ms-fTH-1 in pORF plasmids using Lipofectamine LTX (Life Technologies) as per the instructions of the manufacturer. The growth medium was replaced 24 h post-transfection.

## **IMMMUNOFLUORESCENCE**

In order to investigate the cellular distribution of FHC, HEK-293 cells were fixed with 3% paraformaldehyde and permeabilized with 0.5% Triton X-100 48 h after transfection. Slides were incubated with polyclonal goat anti-FHC (Abcam, #ab80587) and then donkey cy3 (red)-conjugated anti-goat antibody (Jackson ImmunoResearch, #705–165). Slides were stained with DAPI (blue) to illuminate nuclei. Images were taken with confocal microscopy using 60<sup>×</sup> objective and merged with Adobe PhotoshopTM. The percentage of nuclear FHC was estimated using Photoshop by counting the number of FHC pixels co-localized with DAPI (nuclear) divided by the total number of FHC pixels in the image.

#### Vercellotti et al. H-ferritin modulates sickle vaso-occlusion

#### **STATISTICAL ANALYSIS**

All analyses were performed using SigmaStat 2.0 (SPSS, Chicago, IL, USA). Comparison of outcomes for treatment groups was performed using one-way ANOVA.

### **RESULTS**

Human wt- and ferroxidase-null ms-FHC *SB* transposase DNA constructs in LRS were infused hydrodynamically into the tail veins of NY1DD sickle mice. A single death occurred due to injection-related bleeding complications. After 8 weeks, studies were performed as described below.

Wt-hFHC and ms-hFHC mRNAs were transcribed in the spleens of the sickle mice after 8 weeks (**Figure 1A**). Wt-hFHC protein was expressed in liver microsomes and cytosol (**Figure 1B**). LRS injected animals did not demonstrate presence of human mRNA or human FHC protein (**Figures 1A,B**).

Overexpression of FHC in NY1DD sickle mice was tested for its ability to modulate hemoglobin-induced stasis. 8 weeks after hydrodynamic infusion of LRS or *SB* FHC plasmids, dorsal skin fold chambers were placed and free-flowing venules were identified and mapped. After selection of flowing venules, stromafree hemoglobin was infused to induce stasis. 1 and 4 h after hemoglobin infusion, the same venules were re-examined and the static venules were counted. Vascular stasis induced by hemoglobin was inhibited in the skin of sickle mice overexpressing wt-hFHC (5.2 and 2.5% at 1 and 4 h, respectively, *p* < 0.05) but not in LRStreated sickle mice (39.8 and 45.3% at 1 and 4 h, respectively), or in sickle mice expressing ms-hFHC (33 and 27% at 1 and 4 h, respectively; **Figure 2**).

We have previously shown that HO-1 is a modulator of inflammation and vaso-occlusion in transgenic sickle mice both by induction of HO-1 with heme as well as by HO-1 gene therapy using a *SB* transposase (Belcher et al., 2010b). In sickle mice expressing wt-hFHC or ms-FHC, there was a significant (*p* < 0.05) increase in HO-1 mRNA in the liver (**Figure 3A**). HO-1 protein was increased in liver microsomes in mice expressing wt-hFHC, but not ms-FHC or LRS injected animals (**Figure 3B**). Furthermore, there was no increase in HO-1 activity in the ms-FHC or the LRS injected animals, while wt-hFHC significantly increased bilirubin production as a measure of HO-1 activity (**Figure 3C**; *p* < 0.001); that such elevation in bilirubin arose from HO-1 was indicated by the marked reduction in bilirubin production when the competitive inhibitor of HO activity, tin protoporphyrin (SnPP), was administered intraperitoneally (40 μmols/kg) for 3 days prior to organ harvest.

Since HO-1 activity increased with wt-hFHC gene therapy, it was important to establish whether HO-1 was responsible for the inhibition of stasis as previously shown (Belcher et al., 2006, 2010b). Thus, mice injected with wt-hFHC in LRS were given SnPP intraperitoneally to block HO-1 activity. Treatment of animals overexpressing wt-hFHC with SnPP to inhibit HO-1 activity did not reverse wt-hFHC's ability to inhibit stasis (**Figure 3D**). Therefore, wt-hFHC protects against vascular stasis in sickle mice even when HO-1 activity is blocked.

To explore why HO-1 was induced by wt-hFHC, we examined whether known modulators of HO-1, such as heme, Bach-1 and Nrf-2, played a role. Heme content of liver microsomes was

**transgenic sickle mice.** NYDD1 sickle mice were injected hydrodynamically with Sleeping Beauty (SB) transposase vector with either a wild-type (wt) human fTH-1 gene or a triple missense (ms) fTH-1 gene devoid of ferroxidase activity. Control NY1DD sickle mice were infused hydrodynamically with lactated Ringer's solution (LRS). 8 weeks after hydrodynamic infusion, the spleens and livers were removed and flash frozen. **(A)** Total mRNA (n = 4 mice/treatment group) was isolated from the spleens and transcription of the human wt- and ms-fTH transgene mRNA was demonstrated by qRT-PCR, values are expressed as mean + SD of the ratio of FHC to GAPDH mRNA. **(B)** Western blots of liver microsomes and cytosol from treated sickle mice were immunostained for FHC expression. Note that because the third mutation at residue 86 is in a loop section suspected to be a conformational epitope (Addison et al., 1984), it is likely that our primary anti-FHC antibody did not bind our triple ms-hFHC protein on western blots.

significantly increased in wt-hFHC mice compared to ms-FHC or LRS mice (*p* < 0.05; **Figure 4A**). Surprisingly, nuclear Bach-1 and Nrf-2 that the regulate HO-1 expression negatively and positively, respectively (Sun et al., 2004; Liu et al., 2013), were both increased in the nuclei of wt-hFHC expressing mice compared to LRS- or ms-hFHC-treated mice (**Figures 4B,C**).

To determine the basis for the protective effects of wt-hFHC on hemoglobin-induced stasis in SCD, we considered the involvement of NF-κB-mediated adhesion molecule expression. We have previously shown that NF-κB activation plays a critical role in

vascular inflammation, endothelial cell adhesion molecule expression and blood cell adhesion in sickle mice. In an NF-κB p50 knockout bred into sickle NY1DD mice, hemin did not induce stasis, compared to wt sickle mice (Kollander et al., 2010). Animals that received gene therapy with wt-hFHC had diminished hepatic NF-κB activation as evidenced by significantly (confirmed by densitometry, data not shown) decreased phosphorylation of NF-κB p65 in the nucleus compared to ms-FHC or LRS injected animals (**Figure 5**). Similarly, VCAM-1 is an NF-κB-driven adhesion molecule that is required for vaso-occlusion in SCD mice and was decreased in liver microsomes of animals overexpressing wt-hFHC compared to animals treated with ms-hFHC or LRS (**Figure 5**).

Ferritin has been implicated as a nuclear transcription factor for a variety of genes (Cai et al., 1998; Pountney et al., 1999; Alkhateeb and Connor, 2010) and could potentially affect NF-κB activation. Thus, we examined nuclear FHC localization by immunofluorescence in HEK-293 cells transfected with wt- or ms-hFHC. There were marked increases in hFHC expression in cells transfected with wt- and ms-hFHC compared to untreated cells (**Figure 6A**). In untreated cells ∼68% of the FHC was colocalized with the nuclear DAPI stain. Nuclear hFHC expression was 90 and 96% in cells transfected with wt-hFHC and ms-hFHC, respectively. In SCD mice, liver nuclear extracts demonstrated significantly increased nuclear hFHC protein on western blot in mice overexpressing wt-hFHC relative to mice treated with ms-hFHC or LRS (**Figure 6B**). The difference in ms-hFHC expression seen by immunofluorescence (**Figure 6A**, non-denatured protein) and western blot (**Figure 6B**, denatured protein) may be due to differences in antibody preparations used and antibody recognition of triple ms-hFHC and wt-hFHC under denaturing and non-denaturing conditions.

We examined whether other proteins involved in iron or heme metabolism were affected by increases in wt-hFHC. Ferroportin, ALAS and FLC were markedly increased in the livers of SCD mice overexpressing wt-hFHC compared to mice treated with mshFHC or LRS (**Figures 7A–C**). All three iron-related proteins were markedly increased in mice treated with wt-hFHC compared to mice treated with ms-hFHC or LRS.

## **DISCUSSION**

Our results indicate that overexpression of wt-hFHC decreases NF-κB activation andVCAM-1 expression, increases HO-1 expression, and attenuates hemoglobin-mediated vaso-occlusion in SCD mice. Furthermore, the ferroxidase activity of FHC was essential to this protection, as ms-hFHC did not protect.

We and others have previously shown that anti-inflammatory therapies or treatment with antibodies targeting adhesion molecules, including VCAM-1, E-selectin, ICAM-1, P-selectin, α4β1, αVβ3, vWF, or PECAM-1, inhibit vaso-occlusion in SCD mice (Kaul et al., 2004; Belcher et al., 2005, 2006, 2010b, 2013, 2014). Therefore, it is not surprising that decreases in inflammatory tone as reflected by lower NF-κB activation and VCAM-1 expression in SCD mice overexpressing wt-hFHC were accompanied by declines in vascular stasis. However, the mechanism of FHC anti-inflammatory action remains speculative.

Hemoglobin S is known to auto-oxidize 1.7-fold faster than hemoglobin A resulting in a higher rate of conversion to methemoglobin and generation of superoxide (·O2 −) radicals and subsequent amplification of oxidative stress (Hebbel et al., 1988). To that end, scavenging free hemoglobin and heme with haptoglobin and hemopexin, two plasma proteins depleted in hemolyzing sickle cell patients, is a critical defense for the vasculature in SCD. We and others have shown the beneficial effects of supplemental haptoglobin or hemopexin in states of free hemoglobin or free heme excess (Schaer et al., 2013). Detoxifying heme through the induction of HO-1 is another stratagem that can provide vasculoprotection in SCD (Belcher et al., 2006, 2010b).

A clear protective benefit has been shown with upregulation, either pharmacologically or by gene therapy, of the HO-1 system in murine models of SCD. In SCD patients, individuals with a short GT nucleotide repeat in the *hmox-1* gene promoter, hence greater levels of inducible HO-1 protein, had fewer hospitalizations for acute chest syndrome (Bean et al., 2013). Furthermore, the products of HO-1 including carbon monoxide and biliverdin decrease vaso-occlusion in sickle mice (Belcher et al., 2006). Iron is released from heme by HO-1 and ultimately placed as Fe3<sup>+</sup> in the core of ferritin which can safely store ∼4500 atoms of iron. Ferritin's role in SCD is usually considered malevolent as elevated serum ferritin reflects iron overload. Ferritin's potential cytoprotective role in SCD has not been explored.

The basis for the present study on ferritin in SCD was initially published in 1992 (Balla et al., 1992). We had shown that heme, markedly aggravated endothelial cytotoxicity engendered by oxidants. In contrast, however, if cultured endothelial cells were briefly pulsed with heme and then allowed to incubate for a prolonged period (16 h), the cells became highly resistant to oxidant-mediated injury and to the accumulation of endothelial

ANOVA. **(B)** Microsomal membranes (n = 4 mice per treatment group) were isolated from the livers, run on a western blot (30 μg of microsomal protein per lane), and immunostained for HO-1 and GAPDH protein expression. **(C)** Heme oxygenase (HO) enzymatic activity was measured by measuring bilirubin production using 2 mg protein of liver microsomes per reaction (n = 4 mice per treatment group). HO activity in

**(D)** Microvascular stasis was measured 1 and 4 h after infusion of stroma-free hemoglobin (8 μmol/kg) via the tail vein in mice treated with LRS, ms- and wt-FHC (n = 4 mice per group) as seen in **Figure 2**. Inhibition of HO activity with SnPP (40 μmol/kg/day × 3 days, intraperitoneally) did not block the inhibitory effect of wt-FHC on stasis. Values are mean % stasis ± SD, p < 0.05 for wt-FHC mice vs. LRS and ms-FHC, as calculated by one-way ANOVA.

lipid peroxidation products. This protection was associated with the induction for both HO-1 and ferritin. Differential induction of these proteins suggested that ferritin was the primary cytoprotectant. A site-directed mutant of ferritin (heavy chain Glu62---Lys; His65----Gly), which lacks ferroxidase activity and is deficient in iron sequestering capacity, was completely ineffective in protecting cells from injury.

In these studies, FHC was designed to integrate ubiquitously using an *SB100X* transposase driven by a constitutive hEF1-eIF4g promoter. The transposase, delivered in *cis* with the wt- and msfTH-1 transgene, binds to IR/DR sites flanking the transgene, and catalyzes the excision of the flanked transgene, mediating its insertion into the target host genome with an apparently equal preference for AT-rich dinucleotide insertion sites in introns, exons, and intergenic sequences (Vigdal et al., 2002). We looked primarily at the liver and spleen in these animals but did not attempt to identify and characterize specific integration sites. Transient (∼4–6 weeks) episomal expression of the transgene can also occur in organs.

As can be seen in **Figure 1**, the spleen and liver showed evidence for integration of the hFHC genes eight weeks after infusion, with increased levels of mRNA in the spleen and expression of wt-hFHC protein observed in the liver. Most notably, sickle animals expressing wt-hFHC were resistant to hemoglobininduced stasis (**Figure 2**), while ms-hFHC- or LRS-treated animals had significant vaso-occlusion.

In this study, HO-1 was up-regulated at both the transcriptional and translational levels in mice treated with wt-hFHC. This co-expression of FHC and HO-1 suggests intertwined physiology in cytoprotection rather than independent pathways. However, in the study published by Balla et al. (1992), inhibition of HO-1 with SnPP in endothelial cells overexpressing wt-FHC did not reverse the protection afforded by ferritin. A residual effect of endogenous CO produced by HO-1 has not been ruled out. CO has been shown to modulate TLR4 responses in acute pancreatitis (Xue and Habtezion, 2014) and possibly responses to heme in sickle mice. Overexpression of HO-1 or administration of exogenous CO decreases inflammatory tone assessed by NF-κB activation. The decrease in NF-κB activation in wt-hFHC mice (**Figure 5**) could in part be related to increased HO-1 or through anti-oxidative effects of the ferroxidase activity of wt-hFHC.

Our initial studies demonstrating the cytoprotective properties of HO-1 emphasized the attendant induction of ferritin as a necessary accompaniment for HO-1 to confer its protection (Nath et al., 1992). Our prior studies demonstrating the protective effects of ferritin in the heme-exposed endothelium revealed that ferritin can elicit cytoprotection without the concomitant need for intact HO activity (Balla et al., 1992). Our present studies thus complete the loop by demonstrating that ferritin itself feeds back to induce HO-1, but such induction of HO-1 is not essential for ferritin to evince its protective effects. This latter finding is germane to recent observations in a mutant murine model in which H-ferritin is conditionally deleted in the renal proximal tubule; such mice are markedly sensitive to acute kidney injury, despite exhibiting higher expression of HO-1 in the kidney (Zarjou et al., 2013). Thus induced HO-1 may not prove protective if ferritin is absent, and ferritin can exert protection even in the absence of functional HO activity.

**FIGURE 5 | Decrease in nuclear phospho- and total NF-κB p65 and VCAM-1 in sickle mice overexpressing wt-FHC. (A)** Nuclear extracts were isolated from livers, and 30 μg of nuclear extract protein from each liver was run on a western blot and immunostained for phospho- and total NF-κB p65 in wt-, ms-, and LRS-treated mice (n = 4). **(B)** Microsomal membranes were isolated from livers, and 30 μg of microsomal protein from each liver was run on a western blot and immunostained for VCAM-1.

LRS-, ms-, and wt-FHC treated mice (n = 4) were isolated; 30 μg of nuclear extract protein from each liver was run on western blot and immunostained for human FHC. Note: different primary antibodies to FHC were used for **(A)** and **(B)**.

Bach-1 was increased in liver nuclear extracts from mice overexpressing wt-hFCH (**Figure 4B**) despite the marked increases in HO-1 (**Figures 3A–C**). Bach-1 and small Maf heterodimers that bind to Maf recognition elements (MARE) repress HO-1 transcription. Heme binds to Bach-1 and displaces Bach-1/Maf heterodimers from MARE allowing Nrf2 and other transcription factors to bind and induce HO-1 transcription. The displaced Bach-1 is normally exported from the nucleus and degraded. The elevated nuclear Bach-1 levels, in the face of marked increases

in Nrf2 and HO-1, could possibly be related to intracellular transport or new production of heme. Along with heme (**Figure 4A**), ALAS, the rate-limiting enzyme in heme biosynthesis was increased in mice overexpressing wt-hFHC (**Figure 7B**). It is possible the wt-hFHC mice increased heme biosynthesis with altered trafficking of heme/iron in the cell. In animals treated with wt-FHC, significantly higher heme loads and Nrf2 could potentially overwhelm Bach-1 leading to induction of other cytoprotective molecules. Alternatively, nuclear FHC could prevent the phosphorylation of Bach-1. Antioxidant-induced phosphorylation of tyrosine 486 on Bach-1 is essential for the nuclear export of Bach-1. Perhaps formation of heme/Bach-1 complexes, even without nuclear export and degradation, may be sufficient to inhibit the formation of Bach-1/small Maf heterodimers and binding to MARE.

ferroportin, **(B)** mitochondrial ALAS, and **(C)** cytosolic FLC.

Direct nuclear effects of FHC have been described (Pountney et al., 1999; Surguladze et al., 2004; Alkhateeb and Connor, 2010). FHC was protective of DNA from UV damage in the corneal lens cells (Cai et al., 1998). A DNA-binding motif in H-ferritin raised the novel possibility of a role for ferritin as a conventional transcription factor associated with the beta-globin locus promoter (Broyles et al., 2001; Alkhateeb and Connor, 2010). We found wt-hFHC in the nuclei of HEK-293 transfected cells and in nuclear extracts of liver cells *in vivo* (**Figures 6A,B**). The effect of FHC in the nucleus appears to be specific to the heavy chain, in part because the light chain is undetectable (Alkhateeb and Connor, 2010). We found increased expression of stress-responsive proteins Nrf-2 and HO-1 in mice overexpressing wt-hFHC raising the question whether nuclear ferritin may play a role in driving an anti-inflammatory cassette of genes. We found increased ferroportin which could drive iron out of cells, light chain ferritin to store iron and ALAS to utilize iron in the mitochondria, findings that are all in support of this possibility (**Figure 7**). As elegantly recently proposed, FHC can have remarkable effects in a variety of inflammatory states due to its ability to metabolically adapt tissues to iron overload (Gozzelino and Soares, 2014).

In conclusion, our results support the notion that overexpression of FHC ferroxidase protects sickle mice from injury associated with heme-mediated vaso-occlusive disease. Although ferritin has been posited as cytoprotective for decades, upregulation of FHC has proven a challenge especially in sickle patients. As with HO-1, clearly ferritin is more friend than foe in hemolytic disease. Newer therapeutics, including gene therapy approaches may prove to be a valuable asset to the management of SCD.

## **AUTHOR CONTRIBUTIONS**

Gregory M. Vercellotti designed the research, analyzed data, and wrote the paper. Fatima Khan performed research and wrote the paper. Heather Bechtel performed research and analyzed data. Karl A. Nath designed experiments and wrote the paper. Julia Nguyen performed the research and analyzed the data. Chunsheng Chen performed the research and analyzed the data. Carol Bruzzone performed the research, analyzed the data, and wrote the paper. Graham Brown performed and designed experiments and analyzed data. Clifford J. Steer designed the research, analyzed the data, and wrote the paper. Robert P. Hebbel designed experiments and wrote the paper. John D. Belcher designed the research, analyzed data, and wrote the paper.

#### **ACKNOWLEDGMENTS**

The 3 ms fTH-1 (E62F, H65G, K86Q) was a generous gift from Dr Paolo Arosio, Università di Brescia, Brescia, Italy. Stromafree hemoglobin was a gift from Dr. Mark Young, Sangart Inc., San Diego, CA, USA. This work was supported by NIH grants from NHLBI (P01 HL55552, R21/33 HL096469-01, and RO1 HL115467-01).

## **REFERENCES**

Addison, J. M., Treffry, A., and Harrison, P. M. (1984). The location of antigenic sites on ferritin molecules. *FEBS Lett.* 175, 333–336. doi: 10.1016/0014- 5793(84)80763-2

Alkhateeb, A. A., and Connor, J. R. (2010). Nuclear ferritin: a new role for ferritin in cell biology. *Biochim. Biophys. Acta* 1800, 793–797. doi: 10.1016/j.bbagen.2010.03.017


transcriptional repressor Bach1. *Oxid. Med. Cell. Longev.* 2013, 984546. doi: 10.1155/2013/984546


heme-Induced endothelial toxicity in mouse models of hemolytic diseases. *Circulation* 127, 1317–13129. doi: 10.1161/CIRCULATIONAHA.112.130179


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

*Received: 13 January 2014; accepted: 31 March 2014; published online: 17 April 2014. Citation: Vercellotti GM, Khan FB, Nguyen J, Chen C, Bruzzone CM, Bechtel H, Brown G, Nath KA, Steer CJ, Hebbel RP and Belcher JD (2014) H-ferritin ferroxidase induces cytoprotective pathways and inhibits microvascular stasis in transgenic sickle mice. Front. Pharmacol. 5:79. doi: 10.3389/fphar.2014.00079*

*This article was submitted to Drug Metabolism and Transport, a section of the journal Frontiers in Pharmacology.*

*Copyright © 2014 Vercellotti, Khan, Nguyen, Chen, Bruzzone, Bechtel, Brown, Nath, Steer, Hebbel and Belcher. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*

## Randomized controlled trials of iron chelators for the treatment of cardiac siderosis in thalassaemia major

## *A. John Baksi and Dudley J. Pennell\**

NIHR Cardiovascular Biomedical Research Unit, Royal Brompton & Harefield NHS Foundation Trust & Imperial College London, London, UK

#### *Edited by:*

Raffaella Gozzelino, Instituto Gulbenkian de Ciência, Portugal

#### *Reviewed by:*

Stanislav Yanev, Bulgarian Academy of Sciences, Bulgaria Andrei Adrian Tica, University of Medicine Craiova Romania, Romania

#### *\*Correspondence:*

Dudley J. Pennell, Cardiovascular Magnetic Resonance Unit, Royal Brompton Hospital, Sydney Street, London SW3 6NP, UK e-mail: dj.pennell@rbht.nhs.uk

In conditions requiring repeated blood transfusion or where iron metabolism is abnormal, heart failure may result from accumulation of iron in the heart (cardiac siderosis). Death due to heart failure from cardiac iron overload has accounted for considerable early mortality in β-thalassemia major. The ability to detect iron loading in the heart by cardiovascular magnetic resonance using T2\* sequences has created an opportunity to intervene in the natural history of such conditions. However, effective and well tolerated therapy is required to remove iron from the heart. There are currently three approved commercially available iron chelators: deferoxamine, deferiprone and deferasirox. We review the high quality randomized controlled trials in this area for iron chelation therapy in the management of cardiac siderosis.

**Keywords: cardiac siderosis, iron, chelation trials,T2\* imaging, thalassaemia**

## **INTRODUCTION**

In conditions of iron overload, either due to recurrent blood transfusion or abnormal iron metabolism, iron accumulation within organs can have fatal consequences. Beta-thalassemia major (TM) is a condition of abnormal hemoglobin resulting in profound anemia that requires lifelong regular blood transfusion, typically from the age of 2 years. As a consequence of the repeated transfusions and the body's inability to clear excess iron, iron can build up within tissues. The accumulation of iron within the heart (cardiac siderosis) can result in heart failure and this has been the commonest cause of death in TM. The development of iron chelators offered a solution to this problem, but long-term experience showed persistently high and premature cardiac mortality. This has been related in large part to conventional measures of iron loading such as serum ferritin and liver iron concentration being poor surrogates of iron loading in the heart. It is possible to have high levels of iron loading in the liver without significant cardiac loading and vice versa. Consequently cardiac siderosis may not be identified until late, when it presents with heart failure that is associated with a very poor prognosis.

Over recent decades, the observation that iron reduces magnetic resonance (MR) signal has lead to the development of targeted MR sequences to assess the tissue iron concentrations in the liver and the heart. The use of cardiovascular magnetic resonance (CMR) to quantify the parameter T2∗ as a measure of myocardial iron concentration has been particularly important as it remains the only technique available for this purpose (Anderson et al., 2001; Baksi and Pennell, 2014). T2∗ assessment has now been robustly validated,(Kirk et al., 2009; Carpenter et al., 2011) is well established and is widely used (Carpenter et al., 2013). In addition to being able to identify iron loading in the heart and liver, the T2∗ technique can be used to assess the efficacy of iron chelators and to guide therapy. Prior to the development of CMR T2∗, establishing the efficacy of iron chelation therapies on myocardial siderosis was not possible with any accuracy in the absence of cardiac biopsy. Little data was available to accurately guide the application of iron chelators or assess their safety and patient compliance with treatment.

There are few randomized control trials (RCT) of iron chelation for cardiac siderosis. In addition to the paucity of data, the robustness of the presented data is variable. We describe and review the published trials, relating them to the recent guidelines for the management of β-thalassemia major by the American Heart Association (AHA; Pennell et al., 2013). The efficacy of the iron chelators for liver iron is not discussed in this review, and we focus on heart failure due to cardiac siderosis as the main driver of mortality. We only discuss RCTs using T2∗ MR as an endpoint, because it remains the only validated technique to measure myocardial iron. Myocardial T2∗ is considered "normal" when >20 ms. Cardiac iron loading is considered "present" with T2∗ <20 ms, and is considered severe with T2∗ <10 ms because the rate of heart failure increases dramatically below this threshold (Anderson et al., 2001).

## **COMMERCIALLY AVAILABLE IRON CHELATORS**

Currently, there are three commercially available products with FDA and EU approval for iron chelation in the clinical setting. As well as varying in the mode of administration, there are significant differences between the three chelators with regard to efficacy, tolerability, and patient compliance.

## **DEFEROXAMINE**

The first iron chelator approved for clinical use was deferoxamine. The greatest drawback with this agent is that due to poor gut uptake, it needs to be given parenterally (either subcutaneously, or in heart failure due to iron overload by intravenous infusion). The short half life results in typical administration for 8–12 h per day for at least 5 days of the week. This has a profound impact on patient compliance and considerable work has been done to provide devices for subcutaneous drug delivery. Once

severe cardiac iron loading is present, months to years of careful therapy is needed to clear the iron. Even when used as a continuous intravenous infusion, deferoxamine only clears myocardial iron at 5% per month (Pennell et al., 2006; Tanner et al., 2007). Side effects of therapy include deafness, visual disturbance and growth retardation as well as a high frequency of skin reactions.

## **DEFERIPRONE**

Deferiprone was licensed around 30 years later than deferoxamine. Deferiprone is rapidly absorbed from the gastrointestinal tract and is therefore taken orally, usually three times a day. The oral formulation has significant compliance advantages over deferoxamine. Significant side effects include agranulocytosis, gastrointestinal disturbance and arthropathy. Given the risk of neutropenia, weekly monitoring of the neutrophil count is advised. Deferiprone is commonly used in combination with deferoxamine for enhanced iron clearance.

#### **DEFERASIROX**

The most recent of the chelators is deferasirox. This is taken orally with rapid absorption and can be taken once daily. Adverse events are typically gastro-intestinal upset, rash and raised creatinine. More serious adverse events include Fanconi syndrome, and eye and ear toxicity. Monitoring of renal and hepatic function are recommended for deferasirox.

## **RANDOMIZED CONTROLLED TRIAL DATA COMPARING IRON CHELATORS IN THE HEART**

There have been three RCTs examining the role of the iron chelators in cardiac siderosis using T2∗ CMR as an endpoint. The first RCT compared oral deferiprone monotherapy against subcutaneous deferoxamine (Pennell et al., 2013). The achieved dose of deferiprone was 92 mg/kg per day and deferoxamine was 43 mg/kg for 5.7 days per week. The improvement in myocardial T2∗ was significantly greaterfor deferiprone than deferoxamine (27% vs. 13%; *P* = 0.023). Left ventricular (LV) ejection fraction (EF) increased significantly more in the deferiprone treated group (3.1% vs. 0.3% absolute units; *P* = 0.003). This RCT established that deferiprone was superior to deferoxamine over 1 year for the removal of cardiac iron, and the improvement of LV EF in patients with asymptomatic myocardial siderosis. A further report from this trial also showed that right ventricular EF also improved more in the deferiprone patients, than with deferoxamine (Alpendurada et al., 2010).

The second RCT compared combination treatment of deferiprone plus deferoxamine against deferoxamine alone in 65 patients with myocardial T2∗ from 8 to 20 ms representing a range from severe to mild cardiac siderosis (Pennell et al., 2006). There were significant improvements in the combined treatment group compared with deferoxamine alone in myocardial T2∗ (ratio of change in geometric means +50% versus +24%; *P* = 0.02). The combined group also showed significantly great improvement in absolute LV EF (2.6% versus 0.6%; *P* = 0.05), and absolute endothelial function (8.8% versus 3.3%; *P* = 0.02). A further report from this trial showed that right ventricular EF also improved more in the combined group, than with deferoxamine alone (Alpendurada et al., 2012).

The third RCT (CORDELIA trial) compared oral deferasirox against subcutaneous deferoxamine (Pennell et al., 2014). This was a prospective, randomized comparison of deferasirox (target dose 40 mg/kg per day) vs. subcutaneous deferoxamine (50– 60 mg/kg per day for 5–7 days/week) for myocardial iron removal in 197 thalassemia major patients with myocardial siderosis (T2∗ 6–20 ms) and no signs of cardiac dysfunction. The geometric mean of the myocardial T2∗ improved with deferasirox from 11.2 ms to 12.6 ms at 1 year (+12%) and with deferoxamine from 11.6 ms to 12.3 ms (+7%). The between-arm ratio of the geometric means was 1.056 with the 95% confidence intervals of 0.998 and 1.133). The lower boundary was greater than the pre-specified margin of 0.9, establishing non-inferiority of deferasirox vs. deferoxamine (*P* = 0.057 for superiority of deferasirox). Mean LVEF remained stable and within the normal range after 1 year of treatment with deferasirox (66.9– 66.3%) and deferoxamine (66.4–66.4%). The change in mean LVEF after 1 year was not different between the two treatments (*P* = 0.54).

There are two meta-analyses of trials reporting cardiac iron, but they both have significant limitations in including trials using non-T2∗ assessment technology and also non-randomized trials (Mamtani and Kulkarni, 2008; Xia et al., 2013). This demonstrates the paucity of data in the literature and emphasises the need to focus on the three gold standard randomized controlled trials described above.

## **OTHER DATA AND GUIDELINES FOR USE**

#### **DEFERIPRONE VERSUS DEFEROXAMINE**

The superiority of deferiprone monotherapy over standard deferoxamine therapy has been suggested by retrospective and crosssectional studies (Anderson et al., 2002; Pepe et al., 2006, 2011). RCT data indicate that deferiprone is the most efficacious of these agents for reducing myocardial iron, clearing cardiac iron at nearly double the rate of deferoxamine, and also improving EF, an effect not observed with deferoxamine (Pennell et al., 2013). This remains the only prospective randomized control trial of deferiprone monotherapy compared to deferoxamine, and further data would be welcome, as recommended by a Cochrane review (Fisher et al., 2013). The AHA guidelines on heart management in thalassaemia major, indicate that deferiprone monotherapy is useful in patients with cardiac siderosis, and it is also suitable for patients with reduced LVEF or asymptomatic LV dysfunction (Carpenter et al., 2013).

## **COMBINED DEFERIPRONE AND DEFEROXAMINE VERSUS DEFEROXAMINE MONOTHERAPY**

The RCT by Tanner showed that combined treatment with deferiprone and deferoxamine was superior to deferoxamine alone in removing cardiac iron and improving EF. These findings accord with other trial data (Daar and Pathare, 2006; Kattamis et al., 2006; Kolnagou and Kontoghiorghes, 2006; Christoforidis et al., 2007) and a Cochrane review which again suggested that further trial data would be helpful (Fisher et al., 2013). The AHA Guidelines indicate that combined treatment is commonly used where cardiac iron loading is moderate to severe or when LV function is impaired (Carpenter et al., 2013).

#### **MONOTHERAPY WITH DEFERASIROX**

The efficacy of deferasirox monotherapy with regard to its ability to reduce myocardial iron has now been established in a prospective randomized control trial of good size (Alpendurada et al., 2012). Non-inferiority to deferoxamine was demonstrated. Also of note was LV function remained stable during treatment with both deferasirox and deferoxamine, and consequently, the AHA guidelines does not recommend the use of deferasirox as first line treatment where there is established depression of LV EF, or in those with severe iron loading (Carpenter et al., 2013). Data investigating the impact of combination therapy with deferasirox and deferiprone would be direct comparison of interest, as would a trial of deferasirox against deferiprone.

#### **ACUTE HEART FAILURE DUE TO IRON OVERLOAD**

Continuous intravenous deferoxamine is the default consensus treatment for acute heart failure due to myocardial iron toxicity. At present, there is very limited RCT data in this setting. A small trial of iron chelation in heart failure found both deferoxamine monotherapy and combination therapy with deferiprone and deferoxamine to be effective in improving LVEF and myocardial T2∗ (Porter et al., 2013). A non-randomized trial found combined treatment to be effective in severe cardiac siderosis with reduced LV EF (Tanner et al., 2008).

## **CONCLUSION**

The assessment of myocardial iron by T2∗ CMR is now established as fundamental to the best practice management of thalassaemia. This has enabled the investigation of which chelation treatment regime is suitable for myocardial iron loading. Unfortunately, many published studies have suboptimal design and are essentially observational with low patient numbers, and do not contribute significantly to the evidence base for determining cardiac treatment. T2∗ CMR is pivotal to inform management and prolong survival. This is reflected no more strongly than in that data showed a 71% reduction in mortality in the UK βthalassemia major population consequent on the introduction of routine screening for cardiac siderosis by CMR T2∗(Modell et al., 2008).

Whilst all the available iron chelation therapies appear to be effective if given in high enough doses with patient compliance, from the available evidence, deferiprone appears to have superior efficacy compared to deferoxamine, and the effect of deferoxamine is superior when combined with deferiprone compared to deferoxamine alone. Additionally, deferasirox appears to have equivalent efficacy to deferoxamine. Each of these therapies has advantages and disadvantages and tailoring therapy to individual patients is important to optimize outcome. Despite the current evidence base, further larger, well-designed, randomized controlled trials are desirable, and greater data on compliance is needed. Well targeted application of these therapies in combination with T2∗ CMR will improve the prognosis of thalassaemia major patients.

## **AUTHOR CONTRIBUTIONS**

A. John Baksi and Dudley J. Pennell contributed to data review and manuscript preparation.

### **REFERENCES**


a consensus statement from the American Heart Association*. Circulation* 128, 281–308. doi: 10.1161/CIR.0b013e31829b2be6


Xia, S., Zhang, W., Huang, L., and Jiang, H. (2013). Comparative efficacy and safety of deferoxamine, deferiprone and deferasirox on severe thalassaemia: a meta-analysis of 16 randomized controlled trials. *PLoS ONE* 8:e82662. doi: 10.1371/journal.pone.0082662

**Conflict of Interest Statement:** A. John Baksi has no commercial or financial relationships that could be construed as a potential conflict of interest. Dudley J. Pennell is a shareholder in Cardiovascular Imaging Solutions (London, UK) and consultant to Siemens. He is also a consultant to ApoPharma, Novartis, Shire and Bayer.

*Received: 06 August 2014; paper pending published: 02 September 2014; accepted: 08 September 2014; published online: 23 September 2014.*

*Citation: Baksi AJ and Pennell DJ (2014) Randomized controlled trials of iron chelators for the treatment of cardiac siderosis in thalassaemia major. Front. Pharmacol. 5:217. doi: 10.3389/fphar.2014.00217*

*This article was submitted to Drug Metabolism and Transport, a section of the journal Frontiers in Pharmacology.*

*Copyright © 2014 Baksi and Pennell. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*

## The iron regulatory capability of the major protein participants in prevalent neurodegenerative disorders

## *Bruce X.Wong1 and James A. Duce1,2 \**

<sup>1</sup> Oxidation Biology Unit, The Florey Institute of Neuroscience and Mental Health, The University of Melbourne, Parkville, VIC, Australia <sup>2</sup> School of Molecular and Cellular Biology, Faculty of Biological Sciences, University of Leeds, Leeds, UK

#### *Edited by:*

Raffaella Gozzelino, Instituto Gulbenkian de Ciência, Portugal

#### *Reviewed by:*

Dominic James Hare, University of Technology at Sydney, Australia Joshua Dunaief, University of Pennsylvania, USA

#### *\*Correspondence:*

James A. Duce, School of Molecular and Cellular Biology, Faculty of Biological Sciences, University of Leeds, Leeds LS2 9JT, UK e-mail: j.a.duce@leeds.ac.uk

As with most bioavailable transition metals, iron is essential for many metabolic processes required by the cell but when left unregulated is implicated as a potent source of reactive oxygen species. It is uncertain whether the brain's evident vulnerability to reactive speciesinduced oxidative stress is caused by a reduced capability in cellular response or an increased metabolic activity. Either way, dys-regulated iron levels appear to be involved in oxidative stress provoked neurodegeneration. As in peripheral iron management, cells within the central nervous system tightly regulate iron homeostasis via responsive expression of select proteins required for iron flux, transport and storage. Recently proteins directly implicated in the most prevalent neurodegenerative diseases, such as amyloid-β precursor protein, tau, α-synuclein, prion protein and huntingtin, have been connected to neuronal iron homeostatic control. This suggests that disrupted expression, processing, or location of these proteins may result in a failure of their cellular iron homeostatic roles and augment the common underlying susceptibility to neuronal oxidative damage that is triggered in neurodegenerative disease.

**Keywords: Alzheimer's disease, Parkinson's disease, Huntington's disease, prion disease, amyloid-β precursor protein, tau, α-synuclein, prion protein**

## **INTRODUCTION**

The comparative ease in transition of valency states, most commonly between 2+ and 3+, over other metals makes iron one of the most useful metals in oxidative biology. In a cellular environment, over half of all enzymes are metalloproteins, a proportion of which establish complexes with Fe3<sup>+</sup> and/or Fe2<sup>+</sup> (Waldron et al., 2009). Within an aerobic environment iron is continuously redox cycling between Fe3<sup>+</sup> and Fe2<sup>+</sup> to produce reactive oxygen species (ROS). When liganded within proteins this cycling is safely guarded, however, when unprotected the ferrous form is prolific at producing very reactive and damaging hydroxyl radicals from O2 and H2O2 via Haber–Weiss and Fenton reactions (Fenton, 1894; Haber and Weiss, 1934). ROS production leads to DNA, lipid, and protein damage (Crichton and Ward, 2014); intrinsic factors to the increased oxidative stress and cellular damage in many neurodegenerative diseases.

## **IRON WITHIN THE BRAIN**

The brain not only requires iron for the usual fundamental metabolic process such as mitochondrial respiration and DNA synthesis, it is also essential for neurotransmitter synthesis and metabolism as well as myelin synthesis (Crichton, 2009). As iron mismanagement in either direction may severely damage the cell, homeostatic mechanisms have been evolutionarily incorporated to maintain optimal cell function (Wang and Pantopoulos, 2011). Unsurprisingly the proteins required to regulate cellular iron homeostasis in the brain are very similar to those used in the body's periphery and rely on the two cytosolic labile iron pool sensors; iron response protein (IRP) 1 and 2, to bind

to their respective iron regulatory elements in the untranslated mRNA (UTR) of iron responsive proteins (Muckenthaler et al., 2008). Binding to IRE's located in the 5- -UTR prevent translation whereas IRP binding to 3- -UTR IRE's protect mRNAs against nuclease degradation. This canonical*cis–trans* iron regulatory system increases IRP binding to the IRE when iron is required and allows: increased iron uptake through proteins such as transferrin receptor 1 (TfR1) and divalent metal ion transporter 1 (DMT1); impaired iron storage through ferritin (Ft); and reduced export via ferroportin (Fpn; Muckenthaler et al., 2008). When iron is in excess, the IRPs are no longer able to bind, allowing Ft and Fpn mRNA translation and increased mRNA degradation of TfR1 and DMT1.

Despite these common iron-regulated proteins being expressed within most cell types of the brain, the population of the cells present within the brain is diverse and dynamic in their function and requirement of iron. Therefore, differences in how each cell type regulates its iron is evident by the expression of proteins required for import, storage, and export of iron in neurons compared to neuroglia such as oligodendrocytes, astrocytes, and microglia. Circulatory iron transport in the brain is significantly carried out through association with small molecules such as citrate, ascorbate or ATP, however, the iron carrier transferrin is still present (Malecki et al., 1999; Ke and Qian, 2007). Despite this, neurons express TfR abundantly and are likely to acquire iron through the classical holotransferrin endocytosis followed by DMT1-mediated entrance of iron into the cytosol (Pelizzoni et al., 2012). In contrast, astrocytes and oligodendrocytes do not express TfR and therefore take up non-transferrin bound

iron (NTBI) either through DMT1 (Lane et al., 2010) or some alternative routes involving the metal inward transporters; "transient receptor potential cation channel, subfamily C, member 6" (TRPC6; Mwanjewe and Grover, 2004; Giampa et al., 2007), "L-type voltage-dependent calcium channels" (L-VDCCs; Gaasch et al., 2007; Lockman et al., 2012) or "Zrt- and Irt-like proteins" (Zip8 and Zip14; Pinilla-Tenas et al., 2011; Jenkitkasemwong et al., 2012). Once inside, iron is mostly stored in Ft; however, abundance varies age-dependently between cell type with neurons typically containing the least and microglia containing the most (Benkovic and Connor, 1993). Iron may alternatively be stored by neuromelanin in select neuronal types, particularly those known to poorly express Ft such as the melanized dopaminergic neurons of the basal ganglia (Zecca et al., 2003; Snyder and Connor, 2009). Excess iron is exported from all cell types using Fpn, a unique iron efflux pore that is abundantly expressed in neurons, microglia, astrocytes, and oligodendrocytes (Song et al., 2010). However, spatiotemporal expression in neurons is variable (Moos and Rosengren Nielsen, 2006). A glycophosphatidylinositol-anchored form of ceruloplasmin (CP) expressed in astrocytes is known to facilitate iron efflux through Fpn (Jeong and David, 2003) and a similar role is proposed for hephaestin expressed in oligodendrocytes (Schulz et al., 2011) and amyloid-β precursor protein (APP) in neurons (Duce et al., 2010).

As well as each cell type's ability to regulate its own iron content, a continual homeostatic interplay between neurons and the neuroglia is apparent as with most protein regulatory pathways. A good example of this is that despite there being limited presence of the soluble form of CP in the interstitial fluid (Singh et al., 2013), non-neuronal CP depletion causes age-dependent neuronal iron accumulation and cognitive impairment (Jeong and David, 2003). The full regulatory mechanisms of brain iron homeostasis has yet to be fully understood and are likely to include a number of redundant pathways for iron import, storage, and export that can be implemented in order to protect neurons from iron-induced oxidative stress.

Iron accumulation is evident in the aging brain from a range of animals including humans and whilst a heterogeneous distribution of iron is present within the brain, most regions have a continual increase in iron with lifespan (Aquino et al., 2009; Bilgic et al., 2012). Despite neuronal and neuroglial accumulation of iron with age, it has generally been considered not to associate with severe pathology, indicating that these cells are still capable of safely liganding the metal and guarding against oxidative stress-induced cellular damage. Evidence of this is shown with a correlative increase in the expression of the iron storage complex proteins Ft and neuromelanin with age (Connor et al., 1990; Zecca et al., 2001). To wholly understand the iron-induced pathology in neurodegenerative diseases, continued research is required to better understand the full homeostatic pathways for iron regulation in the brain.

## **IRON IN NEURODEGENERATIVE DISEASE**

Most of the brain's iron is concentrated in the substantia nigra pars compacta (SN) and basal ganglia, together reaching comparable levels to that observed in the liver; a known peripheral repository of iron (Griffiths and Crossman, 1993; Haacke et al., 2005). Disrupted iron homeostasis focally in this region of the brain is evident in rare human disorders generally classed as "neurodegeneration with brain iron accumulation" (NBIA) disorders. The phenotypic symptoms of choreoathetosis, dystonia, parkinsonism, spasticity, and rigidity that are present in all forms of NBIA are predominantly associated with neuronal iron accumulation within this region of the brain [reviewed by Rouault (2013)]. The four most frequent subtypes have gene mutations in either; *PANK2* [pantothenate kinase-associated neurodegeneration (PKAN)], *PLA2G6* [PLA2G6-associated neurodegeneration (PLAN)], *C19orf12* [mitochondrial-membrane protein-associated neurodegeneration (MPAN)], or *WDR45* [beta-propeller protein-associated neurodegeneration (BPAN)]. A further five more rare NBIA disorders include Aceruloplasminemia (a deficiency in CP) and Neuroferritinopathy (a deficiency in Ft), and it is only in these two conditions that the functional mutation is in a known iron-regulated protein.

It has become increasingly evident that iron dyshomeostasis may not be pathologically restricted to NBIA disorders, but also a common underlying phenotype in more prevalent forms of neurodegenerative disease. The accumulation of iron in excess of that observed with age within these diseases may induce a variety of adverse effects, including increased oxidative stress, protein aggregation, mitochondrial dysfunction, and an imbalance in neurotransmitters; all of which are prevalent with neuropathology (Duce and Bush, 2010; Crichton and Ward, 2014).

The existence of a definitive correlation between brain iron homeostasis and neurotoxicity associated with these more prevalent forms of neurodegenerative disease still remains to be seen. Despite some studies suggesting that iron is non-specifically coprecipitated with the aggregated proteins pathologically observed with Alzheimer's, Parkinson's, Huntington's, and prion diseases (Altamura and Muckenthaler, 2009), evidence described in this review provides support that iron has a role in the pathogenesis of these diseases and that previously known proteins associated with these disease pathologies are involved in neuronal iron homeostasis.

## **PARKINSON'S DISEASE**

Parkinson's disease (PD) is the second most prevalent age-related neurodegenerative disease affecting 1–2% of the population over 65. It has received the most attention in explaining iron's contribution to the pathogenesis of neurodegeneration. Patients present with motor dysfunction broadly arising from a loss of dopaminergic neurons within the pars compacta region of SN, while the SN reticulate is relatively unaffected. Elevated levels of total iron and a shift in the equilibrium of iron to the oxidized state within a region that already has a high level of iron in the brain is considered to contribute to the oxidative stress-induced neurotoxicity (Dexter et al.,1991; Halliwell,1992;Wypijewska et al.,2010). However, more recently it has been observed that the combination of iron with dopamine is a greater risk factor than each element on their own, and that the SN pars compacta has a greater "irondopamine index" than other regions of the brain (Hare et al., 2014). Altered redox-active labile iron in PD is compounded by a

loss of the buffering capacity of iron storage proteins; neuromelanin (Faucheux et al., 2003) and Ft (Connor et al., 1995) as well as iron-catalyzed aggregation of α-synuclein to form Lewy bodies in surviving neurons (Lotharius and Brundin, 2002). Contributing factors that can explain PD-increased iron accumulation are the elevated expression of the iron import transporter DMT1 (Salazar et al., 2008) and reduced expression of the iron export pore protein Fpn (Song et al., 2010) as well as CP ferroxidase activity (Olivieri et al., 2011; Ayton et al., 2012). *In vivo* imaging of iron by transcranial sonography (TCS) and T2 ∗ -weighted magnetic resonance imaging (MRI) has strengthened the iron hypothesis of PD by illustrating a strong correlation for SN iron levels with disease severity and duration (Menke et al., 2009; Ulla et al., 2013). Typically, intraneuronal iron would be controlled by IRP1/2 response to the cytosolic labile iron pool. However, iron accumulation with an iron-regulated protein profile that correlates with decreased iron, infers a breakdown in the neuron's iron regulatory system with PD. Support for this theory come from an inability for IRP to correctly respond in models of PD. Upregulation of IRP1/2 induces a downregulation in Fpn, thus exacerbating iron accumulation in a 6-hydroxydopamine model of PD (Song et al., 2010), but is unable to control Ft mRNA translation despite the elevated labile iron pool in PD (Hirsch, 2006).

## **ALZHEIMER'S DISEASE**

Accounting for 50–80% of all dementia cases, AD is the most common neurodegenerative disease of individuals over 65. The neuropathological hallmarks of AD are an accumulation of extracellular amyloid plaques comprising mainly of amyloid-β (Aβ), and the presence of intraneuronal neurofibrillary tangles that are comprised of hyperphosphorylated tau. Aβ is proteolytically derived from APP, a ubiquitously expressed type 1 transmembrane protein also predominantly expressed on the neuronal surface. A small number of AD cases who tend to have an earlier onset of disease are caused by autosomal dominance in familial mutations. These mutations are present within regions of APP or its γ-secretase cleavage proteins; presenilin 1 and 2, and promote the amyloidogenic processing of APP to increase Aβ generation. The trisomy mutation associated with Down's syndrome also increases Aβ accumulation leading to early AD pathology and is considered to be caused by the increased copy number of APP that lies within chromosome 21.

While strong evidence suggests that Aβ is the principal cause of neurotoxicity and may be a significant contributor to synaptic dysfunction inAD (Roberts et al.,2012), iron accumulation in affected brain regions, as reported in both post mortem and MRI studies (Falangola et al., 2005; Jack et al., 2005;Antharam et al., 2012), may also be a factor in the increased oxidative stress observed in AD (Castellani et al., 2007). Hippocampal iron accumulation localized in neurofibrillary tangle-containing neurons and the neuritic processes surrounding senile plaques in AD (Quintana et al., 2006) correlates well with cognitive decline (Ding et al.,2009).Within the same regions, proteins regulated by iron are also altered whereby Ft is elevated (Grundke-Iqbal et al., 1990; Morris et al., 1994; Bouras et al., 1997; LeVine, 1997) and both CP expression and activity are lowered (Connor et al., 1993; Torsdottir et al., 2010). Despite

these changes, transferrin expression levels are a little less clear with some evidence of a decrease (Connor et al., 1992) while later reports showing a localized increase within the frontal lobe (Loeffler et al., 1995). As previously suggested, the transport of iron by transferrin within the brain is minor and only required by select cells. This may account for the uncertainty in protein observations despite weak association with AD risk of the C2 variant to the *Tf* gene (Robson et al., 2004; Bertram et al., 2007). A known partner of Tf called HFE protein is also expressed in glia and neurons around neurofibrillary tangles and senile plaques of AD (Connor and Lee, 2006) and over the previous decade numerous genetic association studies have illustrated *HFE* gene mutations increase the risk of AD (Nandar and Connor, 2011). In particular the mutations H63D and C282Y (Sampietro et al., 2001; Pulliam et al., 2003; Blazquez et al., 2007) cause peripheral iron accumulation in AD and possibly has a link with the *APOE* gene. The HFE protein carrying the H63D mutation has also been shown to upregulate the phosphorylation of tau (Hall et al., 2010).

## **HUNTINGTON'S DISEASE**

Huntington's disease (HD) is a progressive neurodegenerative disorder characterized by motor, psychiatric and cognitive disturbances that progress to dementia (The Huntington's disease Collaborative Research Group, 1993). Prevalence in Europe, North America, and Australia is ∼5.70 per 100,000 (Pringsheim et al., 2012). HD is caused by a dominant CAG expansion in the exon 1 encoded region of the *huntingtin* gene resulting in the expression of polyglutamine-expanded mutant huntingtin protein (The Huntington's disease Collaborative Research Group, 1993). Similar to most neurodegenerative diseases and iron accumulative disorders, numerous mechanisms have been implicated in the pathogenesis of HD including oxidative stress (Browne and Beal, 1994), energetic dysfunction (Panov et al., 2002; Cui et al., 2006), transcriptional dysregulation (Nucifora et al., 2001; Dunah et al., 2002), and defective axonal transport (Trushina et al., 2004). Iron dysregulation occurs in human HD (Dexter et al., 1991; Rosas et al., 2012) and brain field map MRI values of gene-positive individuals have suggested that alterations of brain iron homeostasis occur before the onset of clinical signs (Rosas et al., 2012). Genetic mouse models of HD have also accurately recapitulated the elevated levels of brain iron (Fox et al., 2007; Chen et al., 2013). These findings were interpreted to indicate a compensatory response to iron stress occurring in HD striatum.

## **PRION DISEASES**

Prion diseases are a group of disorders whereby extreme cellular destruction leads to vacuolization and spongiosis of large areas of the brain. Within humans the most common form of prion disease is Creutzfeldt–Jacob disease (CJD) of which ∼80% is sporadic cases. Despite being comparatively rare considering other forms of neurodegenerative disease, its high risk of infectivity both within and between species has prompted intense research into the disorder. Prion disease is now known to, at least in part, be caused by the conversion of the prion protein from its regular form (PrPC) into more of a β-sheeted isoform termed PrP-scrapie (PrPSc; Prusiner, 1998; Aguzzi and Falsig, 2012). Little doubt remains that PrPSc is able to initiate infection when inoculated into a recipient animals brain (Wang et al., 2010) and that this is done by PrPSc autocatalyzing its conversion from PrPC. However, the mechanism by which PrPSc induces neurotoxicity remains to be fully elucidated as PrPSc levels poorly correlate with disease progression (Caughey and Baron, 2006).

Accumulation of redox-active iron, partly co-aggregated with Ft and in association with PrPSc plaques, has been reported in CJD brains (Petersen et al., 2005; Singh et al., 2009a, 2012). However, it appears that the increase in total iron may be biologically unavailable as the IRP response in the disease indicates iron deficiency. An increase in both Tf and its receptor as well as transcriptional changes with Ft and IRP1/2 are reported with Tf increase correlating with PrPSc levels (Kim et al., 2007; Singh et al., 2009a). In reflection to the prion diseased brain, cerebrospinal fluid (CSF) has decreased levels of Tf and increased total ferroxidase activity (Singh et al., 2011; Haldar et al., 2013). When used in combination these CSF markers of disease have an accuracy of 88.9% in detecting CJD over other forms of neurodegenerative disease (Haldar et al., 2013).

## **THE ROLE OF PATHOLOGICAL PROTEINS IN IRON**

As mentioned previously, a number of key iron homeostatic proteins such as CP and Ft, have been known for some time to cause NBIA disorders as well as be implicated in the more prevalent neurodegenerative diseases. However, recently a number of key proteins traditionally associated with the pathogenesis of the neurodegenerative diseases described above have also been implicated in an iron regulatory role in neurons. This has strengthened the argument that redox-active iron is a major facilitator of neurotoxicity in these diseases. Correlative studies on iron accumulation and altered pathology also support the theory that changes in iron homeostasis may be a feature in the early progress of the disease.

## **AMYLOID-β PRECURSOR PROTEIN**

As the name infers APP is the precursor of Aβ; the prevalent peptide found in senile plaques from a range of amyloidogenic diseases including AD. Proteolytic processing of APP from the neuron is predominantly through cell surface α-secretase cleavage followed by cleavage with the γ-secretase complex. This non-amyloidogenic processing of APP excludes Aβ production due to the α-secretase cleavage site residing within the Aβ peptide sequence. The alternative processing of APP through the amyloidogenic pathway requires β-secretase instead of α-secretase to produce Aβ as one of its products. The amyloidogenic processing of APP requires the protein to be endocytosed to allow optimal pH conditions for β-secretase cleavage. As yet it is not functionally clear as to why two intricate proteolytic pathways are required to cleave APP, however, the altered cellular location and function of cleaved products could be a likely reason.

Despite a reduced affinity compared to other transitional metals, iron binds Aβ and induces Aβ aggregation (Huang et al., 2004; Ha et al., 2007; Bousejra-ElGarah et al., 2011). This iron interaction is via His6, His13, and His14 of Aβ and is thought to be facilitated in a more reduced environment such as the

brain due to the prevalence for the Fe2<sup>+</sup> form of iron to bind Aβ (Bousejra-ElGarah et al., 2011). ROS generated through iron aggregated Aβ is toxic to neurons (Liu et al., 2011) and may partly contribute to the neurotoxicity present within the iron enriched environment around senile plaques (Meadowcroft et al., 2009; Gallagher et al., 2012). Evidence also supports the same histidine residues in Aβ binding to the iron center as well as porphyrin ring of heme (Atamna et al., 2009; Yuan and Gao, 2013). While the interaction with heme reduces Aβ aggregation (Zhao et al., 2013) it is unclear whether oligomeric Aβ, now considered to be the neurotoxic species within AD, are preferentially formed instead (Thiabaud et al., 2013). Heme binding to Aβ also restricts the bioavailability of regulatory heme and the complex formed has been shown to have peroxidase activity (Atamna and Boyle, 2006).

Translational regulation of APP through an IRE within the 5- UTR implies an interaction with iron status whereby increased cytosolic free iron levels translationally upregulate APP expression (Rogers et al., 2002). APP has recently been identified as a facilitator of neuronal iron efflux through an interaction with Fpn (Duce et al., 2010). While some controversy surrounds the exact mechanism of how APP is involved in the release of iron from the cell (Ebrahimi et al., 2012; Honarmand Ebrahimi et al., 2013), APP within a neuronal environment still appears to be essential to efflux iron (Duce et al., 2010; Wan et al., 2012). Depletion on APP in both cultured neurons and mouse models leads to intracellular iron retention that can be rescued upon the addition of APP to the extracellular environment (Duce et al., 2010) or by overexpression of cellular APP (Wan et al., 2012). Of significance, children suffering from Down's syndrome that have an increased expression of APP have a reported high risk of iron deficiency and anemia (Dixon et al., 2010; Tenenbaum et al., 2011), however, further investigation is required to confirm whether this is due to an increase in APP facilitated iron efflux. As with Fpn, it appears that the surface presence of APP is essential for its role in iron efflux. When APP trafficking to the cell surface is impaired (Lei et al., 2012) or altered by processing through the amyloidogenic pathway, as with the AD-associated familial mutation in APP, iron accumulation arises (Wan et al., 2011).

## **TAU**

Hyperphosphorylated tau has mostly been recognized as the principal component of neurofibrillary tangles, a pathological hallmark in a number of neurodegenerative disorders including AD. Various repeat motifs on tau are known to bind iron in a pHand stoichiometric-dependent manner that results in the promotion of phosphorylation and aggregation of the protein (Ma et al., 2006; Malm et al., 2007; Zhou et al., 2007). While the affinity for iron within a physiological environment has yet to be established for tau, binding of Fe2<sup>+</sup> appears to preferentially induce phosphorylation of tau (Lovell et al., 2004; Chan and Shea, 2006) despite Fe3<sup>+</sup> being the favored state in causing aggregation of tau once it has been hyperphosphorylated (Yamamoto et al., 2002; Amit et al., 2008).

Recently, it has been identified that tau may be required in the iron-modulatory role of APP (Lei et al.,2012). Mice deficient in tau have neuronal iron accumulation that can be reduced in primary cultures by the extracellular addition of APP or moderate chelators such as clioquinol (Lei et al., 2012). Tau has been implicated in axonal trafficking of proteins including APP (Islam and Levy, 1997) and it is proposed that impaired trafficking of APP to the cell surface in tau−/<sup>−</sup> neurons restricts APP's ability to facilitate iron efflux through Fpn leading to intracellular iron accumulation (Lei et al., 2012).

#### **α-SYNUCLEIN**

Variance in α-synuclein is sufficient to cause PD in humans and animal models suggesting a central role in PD pathogenesis (Hardy, 2010). This is apparent through several observations; the overexpression of wild-type α-synuclein through gene duplication is sufficient to cause parkinsonian symptoms (Singleton et al., 2003; Chartier-Harlin et al., 2004; Fuchs et al., 2007); the majority of familial cases of PD are associated with mutations in α- synuclein (Polymeropoulos et al., 1997; Kruger et al., 1998; Zarranz et al., 2004; Hardy, 2010); and aggregated α-synuclein is a core protein found in Lewy bodies. It has been proposed that this ubiquitously expressed protein is involved in synaptic vesicle formation however, it is currently not understood why the dopaminergic neurons of the SN that are targeted in PD are more susceptible to α-synuclein aggregation and toxicity. One theory as to the vulnerability of the SN is the high levels of iron within the region that can augment α-synuclein aggregation. Iron has been shown to bind to the C-terminal region of α-synuclein and under oxidizing conditions, such as that provided in dopamine's presence, denatures the protein and promotes further aggregation (Cappai et al., 2005; Binolfi et al., 2006; Bharathi et al., 2007). Redox-active iron is detected in association with αsynuclein aggregates in Lewy bodies (Castellani et al., 2000), a phenotype that is likely to cyclically promote iron-mediated oxidation and exacerbate the aggregation of α-synuclein as well as the other proteins co-aggregated within the Lewy body (Giasson et al., 2000).

As with APP a strong indication of the importance of α-synuclein in the regulation of neuronal iron was through the identification of an iron-response element in its 5- UTR that is required to increase translation when intraneuronal iron is high (Friedlich et al., 2007; Febbraro et al., 2012). In support of an ironassociated role of α-synuclein, it has recently been identified to modulate cellular iron homeostasis through its ability to reduce Fe3<sup>+</sup> into the biologically active Fe2<sup>+</sup> form (Davies et al., 2011). In accordance with its ferrireductase activity, α-synuclein has a greater affinity for Fe3<sup>+</sup> (Rouault and Tong, 2005) and within a normal physiological environment such as the SN may provide a consistent supply of Fe2<sup>+</sup> for neuronal metabolic processes such as enzymatic synthesis of neurotransmitters.

#### **PRION PROTEIN**

As well as the significant role iron has in prion disease pathogenesis through PrPSc – Ft generated ROS that was described above, recent reports have also suggested that PrPC has a normal physiological role in iron uptake. Cell surface presentation of PrP<sup>C</sup> is required for this function and it appears the copper binding octapeptide repeat region of the protein may have ferrireductase activity (Singh et al., 2013). When the non-modified form of PrPC is overexpressed in cultured cells, iron uptake, and storage is increased (Singh et al., 2009c, 2013) and similarly when expression is depleted by gene knockout in mouse models, tissue iron deficiency correlates with changes to proteins responsive to iron (Singh et al., 2009b). Intriguingly, a familial mutation in PrPC (P102L) classically associated with the prion disorder called Gerstmann–Sträussler–Scheinker (GSS) disease is shown to increase ferrireductase activity and increase levels of intracellular labile iron (Singh et al., 2013). Accumulating evidence now suggests that PrP<sup>C</sup> may have a role in the transferrin and NTBI import into the cell similar to DMT1 or Zip14. Of note, PrP<sup>C</sup> bears a phylogenetic relationship to the ZIP family (Schmitt-Ulms et al., 2009) and has recently been implicated in neuronal zinc import when complexed to NMDA receptors (Watt et al., 2012). Similar to the recent identification that Zip14 is able to transport iron as well as zinc, it is worth noting that the NMDA- PrP<sup>C</sup> complex involved in zinc import may also be implicated in neuronal iron import under certain conditions.

## **HUNTINGTIN**

Mutant huntingtin protein aggregates to form inclusion bodies that represent a pathological hallmark of HD. As with most aggregated protein structures formed in neurodegenerative disease, these bodies bind iron and act as centers of oxidative stress with large amounts of oxidized protein present (Firdaus et al., 2006). Genetic mouse models of HD that transgenically overexpress mutant huntingtin accurately recapitulate the elevated levels of brain iron in the disease (Fox et al., 2007; Chen et al., 2013) and huntingtin knockdown in zebra fish models result in an iron deficiency phenotype (Hilditch-Maguire et al., 2000; Lumsden et al., 2007; Henshall et al., 2009). This suggests that huntingtin is not only regulated by iron but also involved in iron homeostasis. However, iron does not interact directly with N-terminal huntingtin fragments (Fox et al., 2007; Chen et al., 2013) indicating that the effect of huntingtin on iron may mediate downstream influences on iron homeostatic pathways.

## **CONCLUSION; BRAIN IRON DYSHOMEOSTASIS AS A THERAPEUTIC TARGET**

With increasing evidence indicating that iron dyshomeostasis may be a mechanism of exacerbating disease pathology in these more prevalent forms of neurodegenerative disease, there is an escalating realization for its use as a viable target for new therapeutic design. Instrumental work carried out on therapeutic design in the body's periphery (Higgs et al., 2012; Zhou et al., 2012) has increasingly been implemented to investigate their value at restoring iron homeostasis within the brain (Zecca et al., 2004; Badrick and Jones, 2011; Zorzi et al., 2012). However, a significant hurdle in the use of these drugs has been the relative impermeability to the blood brain barrier for some of the more effective peripheral tissue therapeutics and the necessity to target iron in brain rather than the periphery. This barrier is required to isolate and protect the brain from the peripheral circulatory system and transport of drugs across must either occur via active transport using receptors or small lipid soluble molecules that can diffuse across the cellular plasma membrane [for review see (Zheng and Monnot, 2012)].

Iron selective chelators such as desferrioxamine have had limited success in the brain when administered peripherally, largely due to their size and impermeability of the blood brain barrier (Richardson, 2004). However, a number of smaller molecular compounds with varying affinity for iron such as deferiprone, deferasirox, and clioquinol as well as their derivatives, have had promising outcomes in preclinical trials on models of neurodegenerative disease (Kaur et al., 2003; Atamna and Frey, 2004; Molina-Holgado et al., 2008; Rival et al., 2009; Prasanthi et al., 2012). Recently there have been a series of reviews comprehensively discussing preclinical studies on metal affinity compounds [for example Duce and Bush (2010), Ward et al. (2012), Weinreb et al. (2013)]. Deferiprone has already been clinically approved for the peripheral iron overload disorder thalassemia, and used in other neurodegenerative diseases such as Friedreich's ataxia (Boddaert et al., 2007). A recent pilot clinical trial to test for safety and efficacy indicated 6-months deferiprone treatment of early stage PD patients decreased motor handicap progression (as measured by Unified PD Rating Scale) as well as SN iron deposition (as measured by R2∗MRI; Devos et al., 2014) and a further trial is currently underway (Clinical trial #'s NCT01539837). Similarly, clinical trials in AD patients using a clioquinol derivative called PBT2 that has weaker affinity for iron than other translational metals such as copper and zinc, reduced Aβ in CSF as well as improved executive cognitive function (Lannfelt et al., 2008) and is currently in a further clinical trial on AD and HD patients (Clinical trial #'s NCT00471211 and NCT01590888).

While metal affinity compounds have been the focus of most iron therapeutic research in the past decade (Duce and Bush, 2010; Badrick and Jones, 2011; Roberts et al., 2012; Zorzi et al., 2012), it is evident that care must be taken so as to prevent excessive binding which could result in removal of too much iron from the neuronal environment as the reduction in levels below that required for normal physiological maintenance could be just as detrimental to survival (as indicated in anemia). Inaccurate administrative dose of a chelator may therefore only compound the disease phenotype in which they are being used to ease or alter the disease phenotype toward an iron deficiency-like pathology that is just as harmful to the patient. Upon better understanding of the brain's capacity in regulating iron homeostasis both within and between cell types perhaps an alternative approach in the future may be to utilize the brain's own iron homeostatic system to restore the balance of iron. In so doing, recompartmentalizing iron to areas within the brain that have a better capability to cope with oxidative stress-induced by redox-active iron may go a long way toward alleviating the common underlying defects that occur in these more prevalent neurodegenerative diseases.

#### **ACKNOWLEDGMENTS**

This work was supported by funding from the Australian Research Council, the Australian National Health and Medical Research Council (NHMRC) as well as Alzheimer's Research UK. The Florey Institute of Neuroscience and Mental Health acknowledges

the strong support from the Victorian Government and in particular the funding from the Operational Infrastructure Support Grant.

### **REFERENCES**


Browne, S. E., and Beal, M. F. (1994). Oxidative damage and mitochondrial dysfunction in neurodegenerative diseases. *Biochem. Soc. Trans.* 22, 1002–1006.


are a direct effect of polyglutamines. *Nat. Neurosci.* 5, 731–736. doi: 10.1038/ nn884


sporadic Alzheimer's disease. *Neurobiol. Aging* 22, 563–568. doi: 10.1016/S0197- 4580(01)00219-6


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

*Received: 28 February 2014; paper pending published: 08 March 2014; accepted: 02 April 2014; published online: 21 April 2014.*

*Citation: Wong BX and Duce JA (2014) The iron regulatory capability of the major protein participants in prevalent neurodegenerative disorders. Front. Pharmacol. 5:81. doi: 10.3389/fphar.2014.00081*

*This article was submitted to Drug Metabolism and Transport, a section of the journal Frontiers in Pharmacology.*

*Copyright © 2014 Wong and Duce. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*

## Neurodegeneration with brain iron accumulation: update on pathogenic mechanisms

## *Sonia Levi 1,2 \* and Dario Finazzi 3,4 \**

<sup>1</sup> Proteomic of Iron Metabolism, Vita-Salute San Raffaele University, Milano, Italy

<sup>2</sup> San Raffaele Scientific Institute, Milano, Italy

<sup>3</sup> Department of Molecular and Translational Medicine, University of Brescia, Brescia, Italy

<sup>4</sup> Spedali Civili di Brescia, Brescia, Italy

#### *Edited by:*

Paolo Arosio, University of Brescia, Italy

#### *Reviewed by:*

Gert Fricker, University of Heidelberg, Germany Fanis Missirlis, Centro de Investigación y Estudios Avanzados del Instituto Politécnico Nacional, Mexico Nardo Nardocci, Fondazione IRCCS Istituto Neurologico Carlo Besta, Italy

#### *\*Correspondence:*

Sonia Levi, Proteomic of Iron Metabolism, Vita-Salute San Raffaele University, Via Olgettina 58, 20132 Milano, Italy; San Raffaele Scientific Institute, Via Olgettina 58, 20132 Milano, Italy e-mail: levi.sonia@hsr.it; Dario Finazzi, Department of Molecular and Translational Medicine, University of Brescia, Viale Europa 11, 25125 Brescia, Italy e-mail: dario.finazzi@unibs.it

Perturbation of iron distribution is observed in many neurodegenerative disorders, including Alzheimer's and Parkinson's disease, but the comprehension of the metal role in the development and progression of such disorders is still very limited.The combination of more powerful brain imaging techniques and faster genomic DNA sequencing procedures has allowed the description of a set of genetic disorders characterized by a constant and often early accumulation of iron in specific brain regions and the identification of the associated genes; these disorders are now collectively included in the category of neurodegeneration with brain iron accumulation (NBIA). So far 10 different genetic forms have been described but this number is likely to increase in short time. Two forms are linked to mutations in genes directly involved in iron metabolism: neuroferritinopathy, associated to mutations in the FTL gene and aceruloplasminemia, where the ceruloplasmin gene product is defective. In the other forms the connection with iron metabolism is not evident at all and the genetic data let infer the involvement of other pathways: Pank2, Pla2G6, C19orf12, COASY, and FA2H genes seem to be related to lipid metabolism and to mitochondria functioning, WDR45 and ATP13A2 genes are implicated in lysosomal and autophagosome activity, while the C2orf37 gene encodes a nucleolar protein of unknown function. There is much hope in the scientific community that the study of the NBIA forms may provide important insight as to the link between brain iron metabolism and neurodegenerative mechanisms and eventually pave the way for new therapeutic avenues also for the more common neurodegenerative disorders. In this work, we will review the most recent findings in the molecular mechanisms underlining the most common forms of NBIA and analyze their possible link with brain iron metabolism.

**Keywords: brain, iron, neurodegeneration, NBIA disorders, oxidative stress, pathogenesis**

## **INTRODUCTION**

The recent advance in magnetic resonance imaging (MRI) techniques and the increased ability to identify causative genes led to the recognition of a new group of disorders named neurodegeneration with brain iron accumulation (NBIA). It is a set of degenerative extrapyramidal monogenic disorders with radiological evidence of focal accumulation of iron in the brain, usually in the basal ganglia (Yoshida et al., 1995; Gregory et al., 2009; Gregory and Hayflick, 2011). They are characterized by early or late onset, with the main symptoms associated to problems in the movement, spasticity, and cognitive impairment. The diagnosis of NBIA is made on the basis of the combination of representative clinical features along with MRI evidence of iron accumulation (Kruer and Boddaert, 2012). Imaging details can be so specific to orient the successive genetic analysis required to include the disease in a genetically confirmed category or in the idiopathic group, that still represent about 30% of all cases (Dusek et al., 2013).

Different disorders were grouped under this category (**Table 1**): neuroferritinopathy, aceruloplasminemia, pantothenate kinaseassociated neurodegeneration (PKAN), phospholipase 2, group

VI-associated neurodegeneration (PLAN), mitochondrial membrane protein-associated neurodegeneration (MPAN), fatty acid hydroxylase-associated neurodegeneration (FAHN), β-propeller protein-associated neurodegeneration (BPAN), and two other forms, the Kufor–Rakeb disease and the Woodhouse–Sakati syndrome, which are not always associated to brain iron accumulation. Very recently a new form was identified. Exome sequencing revealed the presence of recessive missense mutations in the COASY gene, encoding coenzyme A (CoA) synthase, in one NBIAaffected subject and confirmed in a second unrelated patient (Dusi et al., 2014). The authors proposed COASY protein-associated neurodegeneration (CoPAN) as the name to classify this disorder. The main characteristics of the different types of genetic NBIA syndromes are summarized in **Table 1**. They show variable incidence, and, from an epidemiological point of view, it is also important to note that these diseases are extremely rare and some are generally confined to specific populations (Dusek and Schneider, 2012, 2013).

We are particularly interested in iron homeostasis and the mechanisms underlining its accumulation in the brain. By this

#### **Table 1 | List of the neurodegeneration with brain iron accumulation disorders.**


point of view, the genes so far associated with different NBIA forms can be distinguished between those that code for proteins directly involved in iron metabolism and those that encode proteins responsible for other functions such as fatty acid metabolism and lysosomal activity (**Table 2**). The proposal of hypotheses about iron accumulation mechanisms may be relatively easy in the first case; while for the other type of defects it could represent a difficult challenge.

Here we will try to critically resume the main results obtained from *in vitro* and *in vivo* studies of the pathogenic molecular mechanisms with the aim to highlight the iron involvement in the NBIA pathogenesis, which is still far to be clarified, while for a comprehensive review of clinical symptoms, phenotype and how to make a differential diagnosis we refer to several recent papers appeared in the literature (Kruer and Boddaert, 2012; Rouault, 2013; Schneider et al., 2013).

## **NBIA CAUSED BY DEFECTS IN GENES CODING FOR PROTEINS OF IRON METABOLISM**

Until now, only two genes coding for iron proteins have been identified as responsible of NBIA subtypes: the ceruloplasmin gene (*CP*) causing aceruloplasminemia (Miyajima et al., 1987) and the L-ferritin gene (*FTL*) altered in neuroferritinopathy (Curtis et al., 2001). Defects in these genes lead to early deposits of iron in the striatum, thalamus, globus pallidus (GP), dentate nuclei, cortex, retina; the clinical symptoms typically manifest in adulthood, suggesting that the brain possesses excellent compensatory mechanisms to buffer the harmful action mediated by iron. Neurologic manifestations include blepharospasm, oro-lingual-mandibular dystonia, dysarthria, chorea, parkinsonism, ataxia, and cognitive decline (Dusek and Schneider, 2012).

## **ACERULOPLASMINEMIA**

Aceruloplasminemia (MIM 604290) is an autosomal recessive inherited disease. It was originally described by Miyajima in a Japanese female in 1987 (Miyajima et al., 1987). Effectively, it affects particularly the Japanese population with a prevalence of approximately 1 per 2,000,000 in non-consanguineous marriages, while other 35 families have been described around the world (Kono, 2013). The symptoms include neurological signs with first appearance in adulthood (fourth or fifth decade of life), usually preceded by diabetes mellitus, retinal degeneration (Kono, 2012) and microcytic anemia, unresponsive


to treatment with iron, accompanied by undetectable serum ceruloplasmin, high serum ferritin levels (from 3 to 40 times the normal levels), and low levels of sideremia (Ogimoto et al., 2011). MRI analysis of aceruloplasminemia patients reveals abnormal low intensities in the liver as well as in the striatum, thalamus, and dentate nucleus of the brain on T1 and T2 weighted images, which are consistent with iron deposition (Miyajima, 2003).

Up to now analysis of autopsies from six cases of aceruloplasminemia patients have been reported (Morita et al., 1995; Yoshida et al., 1995; Gonzalez-Cuyar et al., 2008; Kaneko et al., 2012). They showed severe destruction of the basal ganglia and dentate nucleus, with considerable iron deposition in neuronal and glial cells. In the majority of them the cerebral cortex showed mild iron deposition in glial cells without neuronal involvement, however in one case, in which the pathology lasted for the longest time, iron overload was detected also in the cerebral cortex, indicating that the neuropathologic process in aceruloplasminemia worsened during the time and extended beyond the basal ganglia to the cerebral cortex (Kaneko et al., 2012). A sustained accumulation of iron occurs also in retina and cerebellum other than in the pancreas and myocardium (Kaneko et al., 2012). Iron deposition was detected in perivascular areas, localizing to terminal astrocytic processes, especially in the basal ganglia, which show neuronal loss and accumulation of large iron-rich globular structures that appear to be the remains of dead astrocytes (Gonzalez-Cuyar et al., 2008). Accordingly, enlarged or deformed astrocytes and spheroid-like globular structures are characteristic neuropathological findings in aceruloplasminemia.

*CP* is a single-copy gene on chromosome 3, which contains 20 exons with a total length of about 65 kb (Patel et al., 2000), and encodes ceruloplasmin (Cp). The genetic analysis of patients affected by aceruloplasminemia revealed more than 40 distinct causative mutations (Kono, 2013).

Cp is a glycoprotein of the α2-globulin fraction of the serum. It is a multicopper ferroxidase, containing 95% of the copper in the plasma. Its functional role is to facilitate iron export, mediated by ferroportin, from cells. It oxidizes the Fe2<sup>+</sup> to Fe3<sup>+</sup> so that the ferric iron can bind to transferrin present in the extracellular environment. In central nervous system (CNS), Cp is expressed as a glycosylphosphatidylinositol (GPI)-linked form in the astrocytes (Patel et al., 2002). Its action is essential in this cerebral cell type for which the Cp is the only existing ferroxidase (Jeong and David, 2003). In the absence of Cp activity, the ferrous iron that enters the CNS cannot be oxidized and is internalized in large amount, through transferrin-independent, non-regulated pathway (Brissot et al., 2012). The excess import of iron, associated to the export inability due to ferroportin malfunctioning in the absence of Cp, leads to the remarkable accumulation of iron within astrocytes observed in the pathology. Thus it is reasonable to think that iron sequestration by astrocytes may induce iron deficiency and death in neurons, which are astrocytes-depended for iron acquisition (Jeong and David, 2006). Other cells in the CNS, including oligodendrocytes, express hephestin as alternate ferroxidase (Wang et al., 2007) and are not dependent on the action of Cp; this explains the specificity of astrocytes and neuronal death. In brain tissues and cerebral fluid there is also evidence of a marked increase in oxidative stress such as lipid peroxidation and protein carbonylation, in support to excess iron-toxicity (Kono and Miyajima, 2006).

The molecular pathogenesis of aceruloplasminemia was investigated by analysis of Cp mutants expressed in mammalian cell culture (Hellman et al., 2002; Kono et al., 2007, 2010; di Patti et al., 2009) and by characterizing murine models (Harris et al., 1999; Patel et al., 2002; Yamamoto et al., 2002).

The *in vitro* biological analysis of Cp mutants revealed three different types of pathological mechanisms, all resulting in loss of the protein ferroxidase activity. The protein structural modifications induced by mutations can lead to: (i) retention of Cp in the endoplasmic reticulum (ER), (ii) miss-incorporation of copper into apoceruloplasmin, and (iii) impaired ferroxidase activity (Hellman et al., 2002; Kono et al., 2007, 2010; di Patti et al., 2009). All these events hinder iron export from the cell, leading to cellular iron overload.

The *in vivo* experiments were performed on different mice models obtaining variable results. The first knockout mice generated on a C57BL/6J genetic background were described by Harris et al. (1999). The mice showed increased iron content with lipid peroxidation in the brain, but they did not suffer of neurological symptoms; however, a more recent work revealed that this *CP*-deficient young adult mice showed an anxiety phenotype, without discernable effects on learning and memory or motor performance (Texel et al., 2012). The authors determined that in contrast to peripheral tissues, iron levels in the hippocampus were significantly reduced in *CP*-KO mice and, paradoxically, that the anxiety phenotype resulted from reductions in the levels of iron, serotonin, and brain-derived neurotrophic factor expression in the hippocampus (Texel et al., 2012).

A second knockout mice model was developed with a different strategy by Patel et al. (2002). Also in this case *CP*-null mice showed increased iron deposition and lipid peroxidation in several regions of the CNS, but, in addition, they showed deficits in motor coordination that were associated with a loss of brainstem dopaminergic neurons. Astrocytes isolated from the CNS of these *CP*-null mice were used to investigate the functional role of Cp in CNS (Jeong and David, 2003). The authors assessed the 59Fe influx and efflux from astrocytes and concluded that GPI-Cp is essential for iron efflux, while is not involved in regulating iron influx. Furthermore, they identified the co-localization of GPI-Cp on the astrocyte cell surface with the divalent metal transporter IREG1 (now renamed ferroportin) and defined that the harmonized actions of GPI-Cp and IREG1 is required for iron efflux from neural cells. If disruption of this equilibrium occurs, it could lead to iron accumulation in the CNS and neurodegeneration, highlighting the importance of Cp in maintaining iron homeostasis in brain.

In 2002, Yamamoto described a third type of knockout obtained on C57BL/10 and BALB/c genetic background (Yamamoto et al., 2002). The mice had hepatic iron overload but no brain iron accumulation was detected. This murine model was then also investigated for the age dependent expression of hephestin (Cui et al., 2009). The authors detected regional difference in age-dependent hephestin expression and concluded that hephestin may compensate for the loss of Cp in a region-specific manner (Cui et al., 2009). These interesting results might explain the evidence that adult *CP*-null mice have increased iron deposition in the cerebellum and brainstem, while other regions (such as the caudate and putamen), appeared to have normal iron levels (Patel et al., 2002). Unfortunately, the latter brain regions show iron accumulation in aceruloplasminemia patients, indicating the limit of this murine model in recapitulating the aceruloplasminemia human phenotype.

The murine model that lacks the action of both Cp and hephestin develops symptoms consistent with those shown by aceruloplasminemia patients (Hahn et al., 2004; Schulz et al., 2011). Together with iron accumulation in both gray and white matter oligodendrocytes (Schulz et al., 2011) these mice also showed macular degeneration, iron overload and increased oxidative stress in the retina, (Hadziahmetovic et al., 2008). The use of a chelating agent, such as oral deferiprone (DFP), was shown to be protective against the increased oxidative stress and retinal degeneration (Hadziahmetovic et al., 2011; Wolkow et al., 2012). Despite the good results obtained in the mouse model, the use of DFP on a patient resulted in a worsening of symptoms (Mariani et al., 2004), while variable results ranging from mild amelioration to no effect on neurological symptoms were obtained in patients treated with deferasirox (Skidmore et al., 2008; Roberti Mdo et al., 2011). However iron-chelating therapy appeared effective in reducing the hepatic and pancreatic iron overload (Finkenstedt et al., 2010).

Thus, the cascade of events leading to neuronal death remains to be fully elucidated but the overall data strongly suggest that oxidative stress, driven by heavy metal accumulation, represents the primary cellular cytotoxic process, determining the neuronal damage in affected brain regions. Nevertheless, it has been described that some mutated forms of Cp can also accumulate in aggregates and lead to death of astrocytes through an iron-independent pathway (Kono et al., 2006).

## **NEUROFERRITINOPATHY**

The Neuroferritinopathy (OMIM, 606159, also labeled as hereditary ferritinopathies or NBIA3) is a rare monogenic autosomaldominant progressive movement disorder caused by mutations in the gene encoding the L chain of ferritin (FtL). Ferritin (Ft) is the main protein iron storage from prokaryotes to mammals and is characterized by a highly conserved structure that consists of a virtually spherical shell with an internal cavity that can accommodate up to 4500 iron atoms. In vertebrates Ft consists of 24 subunits of two types, H and L, which are assembled in different proportions (Arosio and Levi, 2010). The three dimensional structure of the two chains is very similar: a bundle of four parallel helices, with a long loop that connects helix B and helix C, and a fifth smaller helix called E, at the C-terminus, which is directed toward the center of the cavity. The H chain has a ferroxidase center, where the oxidation of iron occur (Lawson et al., 1991), while the L chain facilitates the mineralization of the iron in the cavity supporting the ferroxidase activity of the H chain (Levi et al., 1994), thus the heteropolymer incorporates iron more efficiently than homopolymers, both *in vitro* and *in vivo* (Santambrogio et al., 1993).

The disease was initially described in members of an English family (Curtis et al., 2001; **Table 3**), which carried the insertion of an adenine in *FTL* gene, resulting in alteration of the


**Table 3 | Reported cases of neuroferritinopathy.**

C-terminus of the protein encoded both in terms of sequence and of length (Curtis et al., 2001). To date, six other alterations have been identified, all of them localized in exon 4 of the *FTL* gene. All the mutations alter the helix E of the FtL and determine a change in the C-terminal part of the protein (**Figure 1**). In addition, a missense mutation, causing A96T substitution on helix C, it has been described as causative of the disorder (Maciel et al., 2005). The average age of onset is 39 years and the main clinical manifestations are those that characterize the extrapyramidal disorders. Despite all the mutations determine common manifestations such as oro-buccal dyskinesia, chorea, and dystonia, it seems that there may be subtle phenotypic variations between them in terms of age of onset, progression of the disease, or the presence of cognitive deterioration (Lehn et al., 2012).

In these patients, routine blood tests are usually normal, only serum ferritin may be low (**Table 3**), while electrophysiological analysis of the cerebrospinal fluid did not show appreciable differences (Lehn et al., 2012). Standard histochemical analysis of muscle biopsies are usually normal, however, Chinnery et al. (2007) found a significant percentage of cytochrome *c* oxidase-negative fibers in two of nine patients analyzed, in addition to isolated or combined defects of respiratory chain complexes in five of six patients analyzed. Iron deposits in the cerebellum, basal ganglia and motor cortex are visible by MRI, either with traditional gradient echo sequences (T2∗)

or with susceptibility weighted imaging (Ohta and Takiyama, 2012). The neuropathological data available were obtained from patients with c.442dupC (p.His148ProfsX33; Mancuso et al., 2005), the c.460dupA (p.Arg154LysfsX27; Curtis et al., 2001) and c.498dupTC (p.Phe167SerfsX26; Vidal et al., 2004). The examination of the brains of these patients showed mild cerebral and cerebellar atrophy as well as cavitation of the putamen. The main neuropathological findings were the presence of intracytoplasmic and intranuclear aggregates of Ft in the glial cells and in some neuronal subtypes, deposits of iron, gliosis, and neuronal death. Glial cells that contain aggregates were found mainly in the caudate, putamen, and GP, and these areas also showed the death of nerve cells and extracellular deposits of Ft. In the cerebral cortex, aggregates were found in cells of the perineural and perivascular glia. The presence of aggregates in neurons was clearly visible in the putamen, the GP, thalamus, in cerebellar granule, and Purkinje cells (Curtis et al., 2001; Vidal et al., 2004; Mancuso et al., 2005).

The aggregates appeared as homogeneous eosinophilic bodies, that were stained with antibodies anti-FtL, anti-FtH or against the mutated form of the *FTL*; inclusions contained both Fe2<sup>+</sup> and Fe3+, the presence of which was analyzed respectively using Turnbull blue and Perls' Prussian blue. Using the transmission electron microscopy, nuclear aggregates appeared as granules of about 100 Å, reminiscent of the structure of Ft and occupy a large part of the nucleoplasm. The presence of aggregates was also reported in cells of other tissues

that cause large alterations of the C-terminal region of the subunit.

Position of the A96T mutation is indicated by a blue arrow along the C-helix.

such as skin, liver, kidney, and muscle (Vidal et al., 2004; Mancuso et al., 2005).

Biochemical analysis of the isolated aggregates identified FtH, FtL, and FtL variant as components, suggesting that the variant form assembles in the ferritin heteropolymers (Vidal et al., 2004). Extensive work has been done on variant recombinant proteins with the attempt to define the structural alteration induced by mutations (Baraibar et al., 2008, 2010; Luscieti et al., 2010). The crystallization of the p.Phe167SerfsX26 homopolymer (Baraibar et al., 2008, 2010; Luscieti et al., 2010) showed that the mutated L chain overlaps exactly with the wild type chain up to glycine 157 while the downstream portion of the amino acids sequence is not solvable, suggesting that the final part of the chain is unstructured (Baraibar et al., 2008, 2010; Luscieti et al., 2010). This alteration of the C-terminal and the exposure of this part of the chain determines a disorganization of the channel described above with two main consequences: a reduced physical stability of the protein due to the loss of stabilizing interactions along the axis of symmetry and the formation of a wider and permeable quaternary channel. These alterations of the hydrophobic channel in the heteropolymer occur also when only one of four subunits, that make up the channel, is changed, thus indicating a dominant negative effect exerted by the mutant protein (Luscieti et al., 2010). Muhoberac et al. (2011) confirmed a greater propensity of ferritin to precipitation induced by iron and a lower functionality of the heteropolymer containing variant chains. Recently, it has also been reported a greater propensity to oxidation of the mutated chains, both *in vitro* and *in vivo*, stressing that oxidative stress is a key component of the pathogenesis (Baraibar et al., 2012).

Overexpression of neuroferritinopathy variants in cells led to the proposal of a hypothetical pathogenetic pathway (Cozzi et al., 2006, 2010). Stable transfection of the p.Arg154LysfsX27 in HeLa cells revealed that the mutant chain assembles in the Ft heteropolymers, although in low proportion (on average less than four subunits for polymer); its expression is associated with an increase in the endogenous ferritin chains expression and a decrease in the expression of transferrin receptor 1 (TfR1). In addition, an increase of reactive oxygen species (ROS) production is evident after treatment with H2O2 in comparison to control cells. The polymer containing the mutated chains incorporates iron with lower efficiency and is degraded much faster than the wild type protein. This leads to an increased release of iron in the cytoplasm, which stimulates the expression of the endogenous ferritin chains, that assemble with the mutant one thus establishing a vicious cycle (Cozzi et al., 2006; **Figure 2**). Similar data were obtained also with

expression of p.Phe167SerfsX26 in neuroblastoma cell lines (Cozzi et al., 2010). When compared to control cells, both variant lines showed an increase of endogenous ferritins, an increase in the LIP after treatment with iron and an enhanced production of ROS after treatment with H2O2, accompanied by an intensification in protein ubiquitination. Both cell lines develop aggregates positive for FtL and iron, which grow in number and size upon iron supplementation, and a defect in the proteasomal activity, which is rescued by treatment with iron chelator and antioxidant agents (Cozzi et al., 2010). These data are confirmed by the analysis of fibroblasts from a patient carrying the c.498dupTC mutation (Barbeito et al., 2010). These cells show an altered management of iron, an accumulation of ferritin and markers of oxidative stress (Barbeito et al., 2010). In basal conditions, the patient's fibroblasts show a significant increase compared with control fibroblasts in the amount of total iron, while the LIP is not altered; the expression of H-, L-, and mutated-L Ft is increased, while that of TfR1 is decreased. Furthermore, also the binding capacity of IRP to IRE is decreased. The whole phenotype is consistent with an increase in the intracellular free iron. Even in these cells the level of ROS is significantly increased compared to that of controls (Barbeito et al., 2010).

*In vivo* data were obtained on a mouse model expressing the human cDNA of *FTL* with the mutation c.498dupTC. The expression of the transgene in the animal caused the formation of nuclear and cytoplasmic aggregates of ferritin throughout the CNS and in other organs (Vidal et al., 2008). The size and the number of nuclear aggregates increase with the aging of the animal, as is the case for patients (Vidal et al., 2004). The model shows a progressive neurological phenotype, a decreased mobility and a reduced life expectancy as well as an increase in the

amount of iron in the brain with altered levels of the associated proteins. In fact, in brain the expression of H- and L-Ft is increased while TfR1 expression is decreased. The transgenic mice also show an accumulation of oxidized DNA in the mitochondria of the brain but no significant damage to the nuclear DNA (Deng et al., 2010) even if oxidative stress markers such as protein carbonylation and lipid peroxidation are elevated (Barbeito et al., 2009).

In conclusion, the overall data obtained by *in vitro* analysis of protein suggest that the pathogenesis could be caused by a reduction in ferritin iron storage capacity and by enhanced toxicity associated with iron-induced ferritin aggregates, whereas data on cellular models, confirmed by the study on transgenic mouse model, imply that the pathogenesis could be mostly related to irondependent oxidative damage. Thus, time should be invested now in the development of therapeutic agents aimed at blocking the detrimental cascade of oxidative events. Some promising results have been recently obtained in more common diseases, such as Parkinson's disease (Devos et al., 2014), and might be extended to these cases.

## **NBIA CAUSED BY DEFECTS IN GENES CODING FOR PROTEINS INVOLVED IN LIPID METABOLISM AND MEMBRANE HOMEOSTASIS**

## **PANTOTHENATE KINASE-ASSOCIATED NEURODEGENERATION**

Pantothenate kinase-associated neurodegeneration (PKAN or NBIA type I) is the most frequent syndrome among NBIA disorders and it was the first one to be associated to mutations in a specific gene, namely pantothenate kinase 2 (*PanK2*; Zhou et al., 2001). It accounts for about 50% of all NBIA cases and may present in two distinct and age-dependent forms, the classic one with early

onset and the atypical one, with later onset of the disease. The classic form usually manifests in the first decade of life, more often before the age of 6 years (Hayflick et al., 2003). Gait and postural difficulties are the most common symptoms often associated to extrapyramidal features such as dystonia, dysarthria, and choreoathetosis. Spasticity and hyperreflexia as well as retinopathy are often present. The atypical form most commonly presents in the second decade of life and its clinical features are less homogeneous, with less severe extrapyramidal and pyramidal signs. Difficulties with speech and psychiatric symptoms with cognitive decline are often observed.

The MRI analysis is a fundamental step in the diagnostic process given that the vast majority of the patients show a specific pattern defined as "eye of the tiger" at T2-weighted MR images. It corresponds to bilateral areas of hypointensity in the medial GP with central spots of hyperintensity (Hayflick et al., 2003). The area of reduced signal is due to iron accumulation in the tissue, as confirmed by the histology (Kruer et al., 2011), whereas the central hyperintense signal may be linked to the dramatic rarefaction of the tissue and the microglial activation (Sethi et al., 1988; Kruer et al., 2011). On the basis of MR data the iron content in the GP of PKAN patients would be about twice as much that found in controls (391 vs 177 μg/ml; Dezortova et al., 2012); the larger proportion of it would be inside ferritin as antiferromagnetic crystals, but a minor part would have different, not yet determined characteristics and have ferrimagnetic properties.

According to a recent study performed on six genetically confirmed PKAN cases (Kruer et al., 2011), the pathology of the disease was almost exclusively located in the CNS, and particularly in the GP where a strong reduction of neurons and synapsis was evident. This was accompanied by the presence of reactive astrocytes, of numerous large and granular spheroids, corresponding to degenerating neurons, and of small and more eosinophil spheroids corresponding to swollen dystrophic axons. Iron accumulation was evident in the GP already on hematoxylin–eosin stain, particularly in theform of perivascular hemosiderin deposits, sometimes within macrophages. The Perl's staining showed the presence of iron within the GP of PKAN brain with an intensity significantly higher than that observed in control brain; the accumulated metal was particularly evident in the cytoplasm of astrocytes, but could also be detected in some neurons, and with a more diffuse pattern in the neuropil. Ferritin staining partially overlapped with the Perl's result and was significantly increased in astrocytes, with a granular pattern, and in the neuropil; in some degenerating neurons the iron staining seemed to be more intense than that of the storage protein, ferritin. This could be in line with the suggestion of two types of iron in the GP of patients proposed from the interpretation of MR data (Dezortova et al., 2012). Overall there was no evident correlation between the intensity of iron accumulation and other histological or clinical features.

An interesting observation was the positive staining for ubiquitin found both in many degenerating neurons and in residual intact neurons present in the GP, possibly indicating the accumulation of ubiquitinated proteins as an early event in the degenerative process. Less consistent was the staining for tau or amyloid precursor protein (APP) in these cells. The latter was instead particularly

evident in neuroaxonal spheroids. In contrast with previous studies performed on Hallervorden–Spatz cases without a genetic confirmation of the disease (Galvin et al., 2000; Neumann et al., 2000; Wakabayashi et al., 2000), no Lewy bodies were detected, thus challenging the interpretation of PKAN as a synucleinopathy.

Pantothenate kinase-associated neurodegeneration is an autosomal recessive syndrome (OMIM number 234200) associated to mutation in the *PanK2* gene. Mutations are usually missense, with c.1561G > A (p.G521R) and c.1583C > T (p.T528M) being the most common, but deletions, duplications, and exon splice site variations have been identified (Hartig et al., 2006). The predicted functional severity of the mutations seems to correlate with the age of onset, but not with the loss of ambulation and the course of the disease (Hartig et al., 2006).

Four *PanK* genes are present in the human genome, which code for PanK enzymes. The different isoforms (PanK1a and b, Pank2, Pank3, and Pank4) share a common C-terminal catalytic domain of about 350 amino acids with about 80% identity and differ for the N-terminal portion of the protein, which contributes to determine their specific cellular localization and regulatory properties (Leonardi et al., 2005; Alfonso-Pecchio et al., 2012; Garcia et al., 2012). PanK2 gene consists of six core exons and different initiating exons; the transcript variant 1 (NM\_153638) is the most common (Polster et al., 2010) and code for a protein of 570 amino acids, with a long N-terminal domain. Targeting signals within the N-terminal part of the protein drive it to mitochondria, where two sequential proteolytic cleavages by the mitochondrial processing peptidase lead to the mature, long-lived and active protein, with a molecular mass of 48 kD (Kotzbauer et al., 2005). The alternate use of a downstream CTG codon for translation initiation was described, that would give rise to a protein of 50.6 kD, always with a mitochondrial localization (Zhou et al., 2001; Johnson et al., 2004). Other transcript variants are described, but they are predicted to code short, not functional cytosolic proteins. Different approaches have shown that the mature PanK2 protein resides within the mitochondrial intermembrane space (Alfonso-Pecchio et al., 2012; Brunetti et al., 2012). Interestingly, a recent analysis of PanK proteins cellular compartmentalization (Alfonso-Pecchio et al., 2012) confirmed a preliminary observation by Hörtnagel et al. (2003) and showed a nuclear localization for PanK2, with the identification of nuclear targeting and export signals driving the entry of the protein into the nucleus and the subsequent exit for the final trafficking to the mitochondria. The functional protein exists as a dimer. The mouse *PanK2* gene structure is the same of the human gene, yet there are conflicting reports as to the cellular localization of the murine protein, alternatively documented in the cytosol (Leonardi et al., 2007b; Alfonso-Pecchio et al., 2012) or in the mitochondria (Brunetti et al., 2012).

A full comprehension of the mechanism linking defects in PanK2 functioning with the neurodegenerative process and iron accumulation in the brain is lacking yet, even though very recent data provided new perspectives, also by the therapeutic point of view. To define the biological mechanism underpinning the development of this disorder different aspects await for clarification. Is the perturbation of CoA biosynthesis the main pathogenic trigger or does PanK2 have other biological functions? What are the relevance and the specific role of Pank2 in CoA biosynthesis and what is the nature of its relationship with the other isoforms of the enzyme? What is the connection linking Pank2 activity and iron metabolism? Why is the brain and particularly specific cerebral areas and neurons so vulnerable to the degenerative process? Even though many of the previous questions remain unanswered, much progress has been recently obtained by different approaches ranging from biochemical *in vitro* studies to the use of cellular and animal models of the disease.

The recognized function of Pank2 is to catalyze the first and limiting step of CoA biosynthesis, that is the phosphorylation of pantothenate (vitamin B5) to 4- -phosphopantothenate (**Figure 3**). This was shown *in vitro*, in transfected cells, where the increased catalytic activity was predominantly found in mitochondria (Kotzbauer et al., 2005; Zhang et al., 2006b). The enzymatic activity was inhibited by CoA (IC50 = 50 μM) and more strongly by a CoA ester (IC50 = 1 μM). This block could be overcome by palmitoylcarnitine, that works as a potent activator of PanK2 function (Leonardi et al., 2007a). This explain how Pank2 can function *in vivo* with the physiological concentration of CoA and acyl-CoA esters and suggests that the presence of enzyme within the mitochondria inner space could be important to sense the levels of palmitoylcarnitine and modulate CoA biosynthesis according the requirement for β-oxidation. Interestingly the *in vitro* analysis of

different PanK2 mutants showed that in many cases the cellular localization, the enzymatic activity, and the regulatory properties are all maintained or only modestly affected (Kotzbauer et al., 2005; Zhang et al., 2006a; Leonardi et al., 2007a). This is for instance the case with the c.1583C > T (T528M) mutation that occurs with high frequency in both early and late onset diseases. Experiments performed with wild type and mutant hPank2 in the fumble (*fbl*) fly model (Wu et al., 2009b) partially questioned the reliability of these *in vitro* enzymatic assays and showed a better correlation between loss of Pank2 catalytic activity and rescue potency. Nonetheless the alteration of PanK2 catalytic activity may not be the only cause of the neurodegenerative process and the possibility of other yet unknown functions of the protein has to be considered. As a matter of fact no definitive evidence is available showing decreased level of CoA in human Pank2 deficient cells or tissues. Recently, the CoA levels in mice with the genetic deletion of *Pank2* alone or in combination with *Pank1* (dKO) were analyzed; no decrease was detected in any tissue in adult *Pank2*-KO mice, while a significant reduction was found in the brain of dKO pups (Garcia et al., 2012). The study suggests that the expression of the other PanK isoforms as well as that of CoA-degrading enzyme such as Nudt7 and Nudt19 may compensate for the absence of Pank2 function. Whether this occurs also in the human brain is not known.

Even though the precise role of Pank2 in CoA biosynthesis in not fully understood yet, the strong connection between this metabolic pathway and neurodegenerative processes is confirmed by the recent identification of mutations in *COASY* in two subjects with clinical and MRI features of NBIA (Dusi et al., 2014). COASY is a bifunctional enzyme with 4- -PP adenyltransferase (PPAT) and dephospho-CoA kinase (DPCK) activities, catalyzing the last two steps in the CoA biosynthesis. Interestingly, in contrast to previous reports showing localization of COASY at the outer mitochondrial membrane (Tahiliani and Neely, 1987; Zhyvoloup et al., 2002), these authors found the enzyme in the mitochondrial matrix, probably anchored to the inner membrane. In the light of this finding the possibility of a CoA biosynthetic process inside the mitochondria has to be further considered and may explain the specificity of Pank2 in this metabolic pathway.

Relevant insight into the biological mechanism connecting defects of Pank2 function, CoA metabolism and the neurodegenerative process came from the drosophila model of the disease. *Drosophila* has only one PanK gene, *fbl*. It gives rise to different transcripts; the longest one (Fbl/L) seems to code for a protein of 512 amino acids with a mitochondrial localization, while the other proteins are cytosolic (Wu et al., 2009b). *Fbl1*, a hypomorphic allele, is due to a P-element insertion near the fbl locus. Fbl1 flies show defects in mitosis, male and female sterility, locomotor dysfunction, neurodegeneration, and reduced life-span (Afshar et al., 2001; Bosveld et al., 2008). The levels of CoA in the mutant flies are significantly lower than in control animals and they are also reduced in *Drosophila* S2 cells exposed to specific fbl silencing (Rana et al., 2010); in deficient animals and cells, mitochondria are severely affected, with damaged and ruptured cristae and membranes, and oxidized proteins are significantly increased. The reduction of *de novo* CoA synthesis is also associated with a marked decrease in the acetylation of histones and tubulin (Siudeja et al., 2011), proteins that are involved in neurodegenerative processes (Fischer et al., 2010). The restoration of appropriate acetylation levels by use of HDAC inhibitors partially corrected many aspects of the phenotype of *Pank2*/*fbl*-deficient cells, such as radiation sensitivity, climbing activity, and homozygous exclusion rate. Furthermore, reduced CoA levels led to hyperphosphorylation and inhibition of twinstar, the fly homolog of cofilin, an essential factor for actin remodeling (Siudeja et al., 2012). These biochemical changes are mediated by pathways that involve Cdi kinase and slingshot phosphatase activities and led to morphological changes in cells and compromised the differentiation and neurite formation capability of human neuroblastoma cells exposed to retinoic acid. Of extreme relevance the fact that the addiction of pantethine to the diet of the mutant flies or to the culture medium of *Pank2/fbl*-deficient cells determined the correction of many of the described biochemical and functional features (Rana et al., 2010; Siudeja et al., 2011, 2012). The metabolite is probably reduced to pantetheine, and then converted to 4- -phosphopantetheine, an intermediate of the canonical *de novo* biosynthesis pathway downstream of PanK2 function. This piece of evidence reinforces and strengthen the direct connection between CoA metabolism and the neurodegenerative process of PKAN. Very recently an alternate drosophila model has been generated with exclusive suppression of *fbl* in tissues with circadian

clock cells, that is eyes, fat and specific neural cells (Pandey et al., 2013). The phenotype of these flies recapitulated aspects observed in the *flb1* hypomorphic animals, with increased sensitivity to oxidative stress, developmental lethality, and shorter lifespan; the transcriptome analysis revealed features partially overlapping with those observed in flies exposed to paraquat, that is in pro-oxidant conditions, but also a specific signature with perturbed expression of genes involved in mitochondrial pathways, cytoskeleton assembly, cell surface receptor signaling, and eye pigment biosynthesis, that could be particularly relevant in the development of retinal degeneration. Altogether the analysis confirmed the relevance of mitochondria and oxidative stress in the pathogenesis of the disease, but also indicated a wide range of transcriptional effects induced by Pank2/Fbl defects and CoA deficiency, that could explain phenotypes observed in different models, such as those related to cytoskeleton function and protein acetylation (Siudeja et al., 2011, 2012) and to iron metabolism (Poli et al., 2010).

The analysis of the mouse model of the disease was at first disappointing, but more recently it has provided new important insight into the disease pathogenesis and possibly opened new therapeutic perspectives. *Pank2* null mice were obtained by Kuo et al. (2005) and were followed for more than 1 year. When compared to littermates they showed a 20% reduction in size and were infertile because of a block in spermiogenesis, with absence of elongated and mature spermatids in the testis. Over time they showed a progressive retinal impairment with alteration of the retinal layers with cones and rods. No sign of motor dysfunction nor of iron deposition in the brain were evident, as assessed either by MRI and by histochemistry, even after 16 months of age or after backcrossing into the C57/BL6J strain. With the exception of retinal degeneration, these features were confirmed in another knock out mouse (Garcia et al., 2012). Tissues, form the latter model, were also investigated for CoA level, and no difference was observed in comparison to age-match littermates. Interestingly, these authors reported that the deletion of *Pank2* resulted to be lethal in the C57/BL6J strain. Brunetti et al. (2012, 2014) further analyzed the *Pank2*−/<sup>−</sup> mice and found a profound alteration of the mitochondria, both in cultured neurons and in the CNS and peripheral nervous system (PNS); these organelles were swollen with altered cristae and membrane potential and oxygen consumption was significantly reduced, albeit the function of respiratory chain complexes appeared to be normal. This was in perfected agreement with the observation of damaged mitochondria in the *fbl* hypomorph fly. When *Pank2*−/<sup>−</sup> mice were fed with a ketogenic diet for 2 months, they developed clear signs of motor and neurological impairment, with foot clasping, tail rigidity, and dystonic limb positioning. The histology revealed the presence of cytoplasmic, eosinophilic, PAS-positive inclusions in some neurons. The anti-ubiquitin staining decorated these inclusions but also smaller granules in apparently intact neurons and larger ones in degenerating neural cells. The electron microscopy analysis confirmed the presence of altered mitochondria in the basal ganglia of *Pank2*−/<sup>−</sup> mice; the phenomenon was dramatically worsened by exposure to the ketogenic diet. Damaged mitochondria were also found in the murine muscle and in the same tissue from a PKAN patient. All of these features were rescued when

pantethine was added to the diet, thus confirming the beneficial effect of the metabolite observed in the *Drosophila* PKAN model and indicating a possible therapeutic approach for PKAN patients.

The accumulation of iron in the GP is a constant feature of PKAN patients but no explanation of the phenomenon is available and very little has been done to understand it, also because the available animal models of the disease showed no sign of iron accumulation. A possible interpretation of the phenomenon was linked to the detection of increased glutathione–cysteine mixed disulfide together with reduced cysteine deoxygenase activity in the GP of PKAN patients (Perry et al., 1985). The accumulated cysteine would chelate iron and induce an abnormal production of ROS with consequent damage to lipids and neuronal membranes. Two works have studied cellular models to investigate possible biochemical mechanisms linking CoA and iron metabolism. Specific siRNA silencing of *Pank2* in different human cell lines induced a mark reduction in cell proliferation together with unexpected signs of iron deficiency, with decrease of Ft and increase of TfR1 and free protoporphyrin levels (Poli et al., 2010). The amount of aconitase, an iron-dependent enzyme, was also reduced, both in the cytosol and in the mitochondria.

Interestingly, the reduced level of Pank2 was associated to a remarkable increase of ferroportin expression, the sole cellular iron exporter, thus suggesting a possible linkage between Pank2 function and iron transport to the brain. The oxidative status and the response to iron supplementation were analyzed in fibroblasts from three PKAN patients and controls (Campanella et al., 2012). Sign of oxidative stress were detected in cells from patients already in basal conditions, and ROS production was increased in these cells after exposure to iron. Under these conditions the fibroblasts from PKAN patients were not able to modulate the binding of IRP1 to mRNA, which in turn resulted in defect in the regulation of ferritin and TfR1 and in higher amount of intracellular free iron (LIP; **Figure 4**). The data suggest that the increased production of ROS associated to Pank2 defects can perturb IRP1 activity and the control of LIP, thus generating the condition for further ROS production in a vicious cycle that could be extremely detrimental for neurons. Even though the biological and clinical meaning of iron accumulation in the brain of patients remains elusive, the administration of DFP, an iron chelating agent had beneficial effects and apparently reduced brain iron levels in two cases of idiopathic, adult-onset NBIA (Forni et al., 2008; Kwiatkowski et al.,

**FIGURE 4 | Schematic representation of alteration of iron homeostasis control in PKAN fibroblasts.** The scheme shows the different structural conformations of IRP1 (Apo, Fe–S, and mRNA-bound) in basal condition and after iron addition, in control **(A)** and in PKAN **(B)** cells. In basal condition, the amount of the regulatory functional form (mRNA-bound IRP1) is lower in

PKAN than in controls fibroblasts. Controls cells respond to iron addition, reducing the amount of mRNA-bound IRP1 and leading to up-regulation of ferritin (Ft) and down-regulation of transferrin receptor 1 (TfR1). This does not occur in PKAN cells where the levels of Ft and TfR1 do not change allowing free iron increase.

2012). More recently two clinical trials have tested the therapeutic potential of DFP; while the reduction of iron in the GP was consistent, the clinical benefit was absent in one case (Zorzi et al., 2011) and only partial in the other one (Abbruzzese et al., 2011). A study on a larger cohort of patients and for longer time is undergoing (http://clinicaltrials.gov/ct2/show/study/NCT017 41532).

## **PHOSPHOLIPASE 2, GROUP VI-ASSOCIATED NEURODEGENERATION**

*PLA2G6*-associated neurodegeneration (PLAN) is the second core NBIA syndrome (NBIA type II, OMIM 256600 and 610217) and is associated to mutations in the *PLA2G6* gene (Morgan et al., 2006). It can have two different age-related presentations, infantile neuroaxonal dystrophy (classic INAD) and atypical neuroaxonal dystrophy (atypical NAD). The former has an early onset presentation, usually with psychomotor regression and within 3 years of age and has a rapid progression, with neurological deterioration leading to loss of ambulation, four limb spasticity, truncal hypotonia, cerebellar ataxia, optic nerve atrophy. The atypical form has a later onset and a less homogeneous presentation with gait problems, ataxia or speech difficulties being the most common symptoms (Gregory et al., 2008). Dystonia and dysarthria together with neuropsychiatric features and visual defects characterize the progression of the disease that usually allows a longer life span than INAD. Rare forms of *PLA2G6*-associated dystoniaparkinsonism have been also identified, characterized by sub acute onset in early adulthood, frequently with neuropsychiatric changes and gait disturbance. The patients then develop dystonia and parkinsonism that may be accompanied by rapid cognitive decline.

In INAD patients the MRI consistently show cerebellar cortical atrophy often associated with signs of gliosis (Farina et al., 1999; Kurian et al., 2008; McNeill et al., 2008). The accumulation of iron is not a constant feature in patients. The metal deposition can be detected in the GP, substantia nigra, and dentate nuclei and usually increases with disease progression. Changes in the posterior corpus callosum and in the cerebral white matter are present. Prominent brain iron accumulation eventually with cerebellar atrophy can be the main neuroradiological sign in atypical INAD whereas non-specific changes such as cerebral atrophy are reported for individuals with *PLA2G6*-related dystonia-parkinsonism (Kurian and Hayflick, 2013).

In *PLA2G6*-associated neurodegeneration pathological changes are widely distributed throughout the CNS. Cerebral atrophy and sclerosis, with loss of neurons and gliosis, accumulation of lipid and degeneration of the optic pathway are evident (Cowen and Olmstead, 1963; Gregory et al., 2008). Depletion of cerebellar granular and Purkinje cells can be present. Numerous axonal swelling and spheroid bodies containing complex network of tubulovesicular membranes are usually detected in the CNS (Itoh et al., 1993; Gregory et al., 2008; Paisán-Ruiz et al., 2012). They are 30–100μm in size and can be often stained with antineurofilament antibodies. Less consistent is the same observation in biopsiesform skin, peripheral nerves, muscle and conjunctiva of INAD patients (Kurian et al., 2008). The analyses of brains form individuals with genetically proven PLAN (Gregory et al., 2008; Paisán-Ruiz et al., 2012) documented the presence of diffuse synuclein-positive Lewy

pathology, particularly evident in the neocortex. In some cases there was evidence of tau pathology, either as threads in the dendrites and axons or as neurofibrillary tangles in neural bodies. This suggests a close relationship between PLAN and Parkinson's disease.

Phospholipase 2, group VI-associated neurodegeneration is an autosomal recessive disease linked to mutations in *PLA2G6* gene (Morgan et al., 2006). To date different types of sequence variations have been documented, including non-sense and missense mutations, small exons deletions, splice-site, and copy number variations (Morgan et al., 2006; Wu et al., 2009a; Crompton et al., 2010). Even though no clear genotype/phenotype correlation exists, the early onset and more aggressive forms are associated with two null mutations, while the atypical presentation is more common in individuals carrying missense mutations. The dystonia-parkinsonism type is associated with mutation not affecting the catalytic domain of the enzyme (Engel et al., 2010).

The *PLA2G6* gene maps to chromosome 22q13.1 and contains 19 exons. Multiple transcript variants probably results from alternative splicing and are expressed at variable levels in most tissues. The two major isoforms (iPLA2β/iPLA2-VIA-1 and iPLA2γ/iPLA2-VIA-2) are quite similar and differ because of a lower number of ankyrin domains and a prolin-rich insertion at the N-terminus of the type 2 protein. They are both catalytically active and have a nucleotide binding-domain, a classical GSXSG lipase consensus sequence and a calmodulin-binding site at the C-terminus (**Figure 5**). The shorter isoforms (Ank-1 and Ank-2) lack the catalytic domain and the enzymatic activity. The active enzyme is a tetramer (Ackermann et al., 1994), probably enabled by interaction among ankyrin repeats. It is thus possible that the shorter inactive forms may act as dominant negative regulators/inhibitors as shown by co-transfection experiments (Larsson et al., 1998). The gene is ubiquitously expressed but seems to have a particularly relevant role in the CNS. It can be found in all regions of the mammalian brain and its activity is the most prominent among PLA2s in the rat brain (Yang et al., 1999; Balboa et al., 2002). The protein is usually considered to reside in the cytosol, but translocation to membrane compartments has

been documented as well as localization in the mitochondria (Williams and Gottlieb, 2002; Seleznev et al., 2006); the latter would be an important feature shared by different proteins coded by genes associated to forms of NBIA and would point to mitochondria as central actors in the development of such disorders. iPLA2β catalyzes the hydrolysis of ester bonds at the sn-2 position of glycerophospholipids, thus releasing free fatty acids and lysophospholipids (Balsinde and Dennis, 1997; Tang et al., 1997; Burke and Dennis, 2009; **Figure 5**); the thioester bond in acyl-CoAs is also a substrate of the enzyme that, in contrast to many other members of the PLA2 family, is active in the absence of calcium. The best recognized function of the enzyme is the homeostatic regulation of membranes topology, by modulating fatty acids recycling and phospholipids amount in the membranes (Balsinde and Dennis, 1997; Winstead et al., 2000); more recently the protein has been implicated in a variety of other cellular processes associated to the generation of arachidonic acid (AA; Green et al., 2008) a precursor of eicosanoids such as prostaglandins and leukotrienes, and lipid second messengers; the involvement in cell signaling processes is well documented by the role played in the control of cells proliferation (Balboa et al., 2008; Hooks and Cummings, 2008), apoptosis (Shinzawa and Tsujimoto, 2003; Pérez et al., 2006), insulin secretion (Zhang et al., 2006a), and store regulated-calcium influx (Smani et al., 2004; Strokin et al., 2012).

The direct connection between defects in iPLA2β and the development of the neurodegenerative process is not understood yet. On the basis of the relevant role of the enzyme in controlling membrane phospholipid turnover and mass it is possible to speculate that perturbation of membrane lipid homeostasis may lead to the structure abnormalities and the axonal pathology that characterizes the disease. This is probably an important but partial aspect in the pathogenesis and further studies are needed to understand the relevance of phospholipid metabolism and iPLA2 functioning in the brain and their involvement in PLAN and other neurodegenerative disease. In contrast to PKAN, mammalian animal disease modeling has been successful for PLAN and important advancement should come from the analysis of the existing mouse models. *PLA2G6* knock out mice are less fertile because of spermatozoa malfunctioning and show defects in pancreatic β-islet activity (Bao et al., 2004; Shinzawa et al., 2008). More importantly, they developed a progressive neurologic phenotype, with gait disturbances, defects in balance and climbing, poor performance at the rotarod and the hanging grip test (Malik et al., 2008; Shinzawa et al., 2008). The motor impairment was evident at about 1 year of age, worsened with age and was associated with a shorter lifespan (Shinzawa et al., 2008).The neuropathological assessment revealed features strikingly similar to those documented in INAD patients with numerous spheroids and vacuoles affecting axons and the neuropil throughout the brain and the PNS (sciatic nerve, mesenteric and celiac ganglia); these structures were often highlighted by the staining with anti-ubiquitin antibody and sometimes were also positive for αsynuclein; they often contained PAS-positive granules. Neither Lewy bodies nor iron accumulation were evident in the brain of *PLAG26*−/<sup>−</sup> mice. The ultrastructural examination evidenced a frequent presence of tubulovesicular structures, vacuoles, dense

granules, and amorphous material in the spheroids. A very similar phenotype was also observed other mouse models, either expressing a non-functional protein with an amino acid substitution in the ankyrin repeat domain (Wada et al., 2009) or expressing very low level of *PLA2G6* transcript because of a viral insertion upstream of the start codon (Strokin et al., 2012). In these mice the motor impairment was already evident at 2 months of age and progressively worsened. The mice died between 18 and 24 weeks of age. The pathology was characterized by the presence of numerous spheroids, very similar to those observed in INAD and containing tubulovesicular structures, vesicles, mitochondria, and amorphous material; they were distributed both in CNS and PNS, with particular involvement of the gracile and cuneate nuclei of the brainstem (Wada et al., 2009). Iron accumulation was not investigated in these mice. The striking temporal difference in the manifestation of the phenotype among different mouse models is not explained and raises the question whether the presence of the mutant protein may contribute to the development of the disease. Important information regarding the possible pathogenic pathway underpinning the disease have been recently provided by Beck et al. (2011), who performed an in depth morphological and biochemical analysis of the spinal cords and sciatic nerves of *PLA2G6*−/<sup>−</sup> mice at 15 (presymptomatic stage), 56 (early clinical stage), and 100 weeks (late clinical stage) of age. The presence of PAS-positive granules in the perinuclear space and proximal axons of neurons was the earliest abnormality detected in KO mice. The granules were decorated with antibodies for TOM20, a marker of the inner mitochondrial membrane. The ultrastructure analysis revealed that these granules were remnants of mitochondria filled with dense granules. At later stages swollen axons filled with granules and vacuoles appeared and progressively increased in number with age; many collapsed mitochondrial remnants as well as abnormal mitochondria with branched and tubular cristae and tubulovesicular structures were observed within the swollen axons, where perturbation of the axonal cytoskeleton was also evident. The plasma membranes at axon terminals were also degenerated. The lipid content of these tissues was analyzed by liquid chromatography/electron spray ionization tandem mass spectrometry and by imaging mass spectrometry. A severe perturbation in the relative amount of different types of phospholipids was documented, with increase of the phosphatidylcholine containing AA and docosahexaenoic acid (DHA), decrease of other phosphatidylcholines and increase of phosphatidylethanolamines and cardiolipins. Altogether these data suggest that the absence of iPLA2β enzymatic activity may lead to an insufficient remodeling of the presynaptic and the mitochondrial inner membranes with abnormal accumulation of DHA- and AA-containing phospholipids; these perturbation would cause a progressive degeneration of mitochondria and plasma membranes at synaptic terminals, associated to alteration of the cytoskeleton, and ultimately determining the formation of tubulovesicular structures and swollen axons, the main pathological features of INAD. Given the multifaceted role of iPLA2β in the cells, the existence of parallel pathways has to be taken into account, as demonstrated by the disturbance in Ca++ signaling observed in astrocytes derived from two different mouse models of INAD (Strokin et al., 2012). Exposure of astrocytes from mutant mouse strains to ATP revealed a shorter

duration of Ca++ responses, associated to reduced capacitative Ca++ entry. Similar results were obtained when wild type cells were exposed to a chemical inhibitor of iPLA2β. Defects in Ca++ responses could affect the neuron–astrocyte communication and potentially contribute to the development of the disease.

As to the role of iron in PLAN development no studies and data are available. In contrast to PKAN, the presence of iron is not a constant feature of PLAN patients; furthermore, even though the available animal models show a pathology remarkably similar to that observed in patients, they apparently do not show evidence of iron accumulation in the brain. Noteworthy, the analyses were usually limited to the histological detection by the Perls' staining. Altogether it may well be that the metal has not a relevant role in PLAN pathogenesis, but more careful studies could provide important information as to the possible connection between lipid and membrane homeostasis and iron metabolism in the brain.

#### **FATTY ACID HYDROXYLASE-ASSOCIATED NEURODEGENERATION**

Most of the cases of NBIA are associated to mutations in *PanK2* and in *PLA2G6* genes, but many other NBIA genes have been recently identified (**Table 1**). Interestingly the relevance of lipid metabolism and membrane remodeling is further highlighted by the association of mutations in the gene coding for the fatty acid-2-hydroxylase (FA2H) enzyme (Kruer et al., 2010) with another rare form of NBIA, named FAHN. The same gene is also involved in hereditary spastic paraplegia SPG35 (Dick et al., 2010) and in leukodystrophy (Edvardson et al., 2008). The clinical phenotype is quite similar to that of PLAN with spasticity, dystonia, ataxia and oculomotor disturbances. The MRI shows bilateral hypointensity in the GP and in SN, related to iron accumulation, cortical atrophy, and white matter lesions. The enzyme catalyzes the hydroxylation of fatty acids in sphingolipids, essential constituents of myelin sheaths (**Figure 3**). The relevance of this metabolic pathway for axonal functioning and maintenance is evidenced by the phenotype of *FA2H*−/<sup>−</sup> mice, showing evident signs of demyelination with axonal enlargement and loss in the CNS (Zöller et al., 2008; Potter et al., 2011).

## **MITOCHONDRIAL MEMBRANE PROTEIN-ASSOCIATED NEURODEGENERATION**

Mutations in the *C19orf12* gene are now associated with MPAN, an autosomal recessive disorder that represents between 5 and 30% of NBIA cases (Hartig et al., 2011; Panteghini et al., 2012; Hogarth et al., 2013). The evaluation of the genetically confirmed cases of MPAN allowed the identification of a distinctive clinical phenotype characterized by pyramidal and extrapyramidal signs, cognitive decline, neuropsychiatric changes, optic atrophy, and upper and lower motor neuron signs (Hartig et al., 2011; Hogarth et al., 2013). Great variability is present both in disease onset (between 3 and 30 years) and progression. Gait instability or visual impairment are often the initial symptoms, then followed by muscular weakness and atrophy, dystonia, and dysarthria. Almost constant is the cognitive decline leading to dementia as well as the appearance of neuropsychiatric modifications. Brain iron accumulation is the most significant sign at MRI. It usually

involves both the GP and the substantia nigra; it can be accompanied by cortical and cerebellar atrophy. The neuropathologic assessment has been performed in two cases, with very similar results and evidence of iron accumulation, axonal spheroids, tau and Lewy pathology that affected the basal ganglia, the archicortex, the neocortex, and the spinal cord. Iron was evident in the GP and the substantia nigra, inside neurons, astrocytes and macrophages with perivascular localization and associated with neuronal loss and gliosis. Numerous eosinophilic axonal spheroids were evident and were strongly immunoreactive for ubiquitin, and less positive for tau or APP staining. Extremely remarkable was the Lewy pathology with the presence of Lewy neurites and Lewy bodies positive for α-synuclein staining in many brain regions. In one case the overall burden was greater than that observed in cases of sporadic Lewy body disease (Hogarth et al., 2013). More recently mutations in *C19orf12* gene have been linked to cases of hereditary spastic paraplegia type 43 (Landouré et al., 2013) and pallido-pyramidal syndrome (Kruer et al., 2014), thus extending the clinical spectrum associated to *C19orf12* gene variants.

The *C19orf 12* gene consists of three exons and codes for two alternative mRNA isoforms (NM\_001031726.3 and NM\_031448.4) and proteins that differ for the presence of a stretch of 11 amino acids at the N-terminus of the longer form. Different types of mutation have been identified and many of them lead to truncated, non-functional proteins. Among the missense mutation, the c.32C > T is the most common and affects tyrosine 11, exclusively present in the longer form. The protein has a long hydrophobic domain (amino acid residues 42–75) that could represent a transmembrane region; the presence of two prolines in the middle of the domain suggests the possibility of an hairpin structure. The longer form is localized in the mitochondria (Hartig et al., 2011) and possibly in the ER (Landouré et al., 2013); its expression level is more significant in the brain, in blood cells and adipocytes and it appears to be co-regulated with genes involved in fatty acid biogenesis and valine, leucine, isoleucine degradation, that indicates a connection with CoA and lipid metabolism and mitochondria. No other information is available to explain the connection between the protein function, the neurodegenerative process, and iron accumulation in the brain, but very recently a *Drosophila* model with impaired expression of the two orthologs of human C19orf12 was obtained (Iuso et al., 2014); those flies showed reduced life span and climbing activity, signs of neurodegeneration (vacuoles) but not of iron accumulation (negative Prussian blue staining). The model will be an important tool to explore the mechanisms underlining the disorder.

## **OTHER FORMS OF NBIA**

## **β-PROPELLER PROTEIN-ASSOCIATED NEURODEGENERATION**

Very recently a subset of NBIA cases with homogenous clinical phenotype and disease history (known as SENDA) has been linked to mutations in the *WDR45* gene (Haack et al., 2012; Saitsu et al., 2013) and is now identified as BPAN. The clinical pattern is characterized by global developmental delay and intellectual deficiency in early childhood then followed by further neurological and cognitive regression in late adolescence or early adulthood, with parkinsonism, dystonia, and sometimes ocular defects and sleep perturbation (Hayflick et al., 2013). At the time of the clinical deterioration the MRI shows clear sign of iron accumulation involving the SN and the GP. A common and distinctive feature on T1-weighted-imaging is the presence of a thin line of hypointensity in the SN and cerebral peduncles, surrounded by a hyperintense halo. Cerebral and sometime cerebellar atrophy are present. The pathology from a single postmortem sample is available (Hayflick et al., 2013) and showed evident iron accumulation in the SN and to a lesser extent in the GP. This was associated with the presence of axonal spheroids, gliosis and neuronal loss. In the cerebellum there was a significant reduction of Purkinje cells. Neurofibrillary tangles were present in different brain regions, while no positivity for α-synuclein or APP deposition was evident. BPAN is due to "*de novo*" mutations in the WDR45 gene causing loss of function of the encoded protein; even though the gene is at the chromosome X, males and females present the same clinical phenotype, which seems to be due to somatic mosaicism or skewing of the X chromosome inactivation (Haack et al., 2012). The *WDR45* gene codes for a protein (WIPI4) with a seven-bladed beta-propeller structure and a phosphoinositide-binding motif for membrane interaction. It belongs to the large family of WD40 repeat protein and is one of the four mammalian homologs of yeast Atg18, an important regulator of autophagy. Its involvement in autophagy has been documented in yeast and mammalian cells (**Figure 3**), where it interacts with ATG2 (Behrends et al., 2010), and in *C. elegans*, where its deletion leads to accumulation of early autophagosome (Lu et al., 2011). The protein amount is clearly reduced in lymphoblast cells from BPAN patients and the autophagosome formation is hindered at an early stage as shown by the accumulation of LC3-II protein and its co-localization with ATG9a in enlarged membrane structures (Saitsu et al., 2013). Even though the relevance of the autophagy process in neurons and brain is well documented, this disorder represents the first direct link between the autophagy machinery and neurodegeneration; it will be of great interest to analyze the correlation with iron homeostasis.

## **KUFOR–RAKEB SYNDROME**

Also Kufor–Rakeb disease (KRS; OMIM 606693) is due to mutations in a gene, namely *ATP13A2*, encoding a protein involved in degradation processes. It is a rare autosomal disorder also identified as Parkinson's disease 9 and characterized by earlyonset parkinsonism, pyramidal signs, altered eye movements and dementia. Some cases show signs of bilateral iron deposition in the basal ganglia at the MRI, and they can be added to the group of NBIA (Schneider et al., 2010). Mutation in the same gene are also linked to cases of neuronal ceroid lipofuscinosis (NCL; Bras et al., 2012). Indeed, the accumulation of lipofuscin and α-synuclein was also observed in ATP13A2-null mice (Schultheis et al., 2013).

The gene codes for a type 5 P-type ATPase, a cation pump located in the lysosome whose function is not defined yet. Different studies in patients' primary cells and in *ATP13A2* nullmice have investigated the function of the protein suggesting its involvement in divalent cation handling, and lysosomal and mitochondrial functioning. The analysis of fibroblasts from patients carrying mutations in the *ATP13A2* gene evidenced severe perturbations of lysosomal function, with impaired degradation of substrates, reduced processing of lysosomal enzymes and decreased autophagosomes clearance (Dehay et al., 2012; Usenovic et al., 2012). The downregulation of *ATP13A2* expression in dopaminergic neurons or mouse primary cortical neurons induced the same phenotype and cell death, together with αsynuclein accumulation. Fibroblast from patients' also showed impaired maintenance of mitochondria, with network fragmentation, mitochondrial DNA alterations, reduced membrane potential, and ATP production (Grünewald et al., 2012). Similarly, silencing of *ATP13A2* in rat primary cortical neurons induced mitochondrial fragmentation either in basal condition or after exposure to cadmium (Ramonet et al., 2012). It is probable that these perturbations in organelle function and structure are linked to altered cation transport and/or homeostasis. Lysosomal Zn++ sequestration was shown to be impaired in patients' fibroblasts (Tsunemi and Krainc, 2013) and olfactory neurosphere cultures (Park et al., 2014) and this was associated with increased expression and accumulation of α-synuclein, elevation of ROS levels, ATP depletion and mitochondrial dysfunction, leading to cell death. The overexpression of ATP13A2 or Zn++ chelation could block Zn++ toxicity (Park et al., 2014). Furthermore KRS mutants were not able to protect cells form Mn++-dependent cytotoxicity (Tan et al., 2011). Even though no evidence suggests a direct role of the P-type ATPase in iron intracellular handling, the lysosomes, and acidic endosomes are fundamental for iron homeostasis and perturbation of their functioning may be linked to the accumulation of the metal observed in the brain of some KRS patients.

## **CONCLUSIONS AND FUTURE PERSPECTIVES**

While each of these NBIA types is associated to mutations in different genes, it is possible to underline common features that may point to shared pathogenic pathways. Aceruloplasminemia and neuroferritinopathy have in common iron mis-handling, which is the primer trigger of cellular oxidative damage, while the involvement of lipid metabolism and mitochondria is evident in many of the other diseases. It is clear for PKAN and PLAN in which mitochondrial network fragmentation, altered cristae and structural and functional abnormalities are well documented (Rana et al., 2010; Beck et al., 2011; Brunetti et al., 2012) and are probably linked to perturbation in the synthesis (PanK2) or the remodeling (iPLA2β) of membrane lipids. It will be of interest to analyze mitochondria from patients with mutations in *COASY*, that alters the same metabolic pathway of Pank2, and in C19orf12, that is linked to the mitochondrial outer membrane and is presumably connected with lipid metabolism (Hartig et al., 2011). Interestingly enough mitochondrial structure and function perturbations were found also in cells with defects in type 5 P-type ATPase, involved in KRS. This could be due to an impairment of lysosomal activity and ensuing defects in the clearance of damaged mitochondria (mitophagy). Perturbations in this process seem to play a major role in the development of PD, particularly in the genetic forms linked to mutations in Pink1, DJ-1, and Parkin, and this could explain the overlap between some forms of NBIA and PD. Mitochondria are also strategic organelles for the regulation

of iron metabolism and significant connection among iron and altered membrane structure (Patil et al., 2013) and mitophagy (Allen et al., 2013) have been described in the literature. In yeast defects in cardiolipin synthesis lead to diminished formation of iron–sulfur clusters (ISC) and iron accumulation in the organelles, a feature often observed in disorders with faulty ISC assembly, such as Friedreich ataxia or X-linked sideroblastic anemia with ataxia (XLSA-A; Rouault, 2012). Interestingly mitochondrial iron overload often brings up cytosolic iron deprivation and a vicious cycle with increased iron uptake and subsequent deposition in mitochondria (Huang et al., 2011). Such an iron misdistribution could be at the same time a strong trigger for the production of ROS, as occurs in aceruloplasminemia and neuroferritynopathy, and hinder mitochondrial renewal (Allen et al., 2013). The interplay among membrane homeostasis, mitochondria functioning, and iron metabolism may thus play a fundamental role in the development of different types of NBIA and further investigation in this field should lead to a better understanding of the biological mechanisms underpinning these neurodegenerative processes.

## **ACKNOWLEDGMENT**

The financial support of Telethon-Italia (grants no. GGP10099 to Sonia Levi and GGP11088 to Sonia Levi and Dario Finazzi) is gratefully acknowledged.

## **REFERENCES**


is an effective tool for the detection of novel intragenic PLA2G6 mutations: implications for molecular diagnosis. *Mol. Genet. Metab.* 100, 207–212. doi: 10.1016/j.ymgme.2010.02.009


degeneration and azoospermia. *Hum. Mol. Genet.* 14, 49–57. doi: 10.1093/hmg/ ddi005


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

*Received: 28 February 2014; paper pending published: 24 March 2014; accepted: 17 April 2014; published online: 07 May 2014.*

*Citation: Levi S and Finazzi D (2014) Neurodegeneration with brain iron accumulation: update on pathogenic mechanisms. Front. Pharmacol. 5:99. doi: 10.3389/fphar. 2014.00099*

*This article was submitted to Drug Metabolism and Transport, a section of the journal Frontiers in Pharmacology.*

*Copyright © 2014 Levi and Finazzi. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*

## The interplay between iron accumulation, mitochondrial dysfunction, and inflammation during the execution step of neurodegenerative disorders

## *Pamela J. Urrutia, Natalia P. Mena and Marco T. Núñez\**

Department of Biology and Research Ring on Oxidative Stress in the Nervous System, Faculty of Sciences, University of Chile, Santiago, Chile

#### *Edited by:*

Raffaella Gozzelino, Instituto Gulbenkian de Ciência, Portugal

#### *Reviewed by:*

Reinhard Gabathuler, Cydweli Consultants, Canada Jerome A. Roth, University at Buffalo, USA

#### *\*Correspondence:*

Marco T. Núñez, Department of Biology and Research Ring on Oxidative Stress in the Nervous System, Faculty of Sciences, University of Chile, Las Palmeras 3425, Santiago 7800024, Chile e-mail: mnunez@uchile.cl

A growing set of observations points to mitochondrial dysfunction, iron accumulation, oxidative damage and chronic inflammation as common pathognomonic signs of a number of neurodegenerative diseases that includes Alzheimer's disease, Huntington disease, amyotrophic lateral sclerosis, Friedrich's ataxia and Parkinson's disease. Particularly relevant for neurodegenerative processes is the relationship between mitochondria and iron. The mitochondrion upholds the synthesis of iron–sulfur clusters and heme, the most abundant iron-containing prosthetic groups in a large variety of proteins, so a fraction of incoming iron must go through this organelle before reaching its final destination. In turn, the mitochondrial respiratory chain is the source of reactive oxygen species (ROS) derived from leaks in the electron transport chain. The co-existence of both iron and ROS in the secluded space of the mitochondrion makes this organelle particularly prone to hydroxyl radical-mediated damage. In addition, a connection between the loss of iron homeostasis and inflammation is starting to emerge; thus, inflammatory cytokines like TNF-alpha and IL-6 induce the synthesis of the divalent metal transporter 1 and promote iron accumulation in neurons and microglia. Here, we review the recent literature on mitochondrial iron homeostasis and the role of inflammation on mitochondria dysfunction and iron accumulation on the neurodegenerative process that lead to cell death in Parkinson's disease. We also put forward the hypothesis that mitochondrial dysfunction, iron accumulation and inflammation are part of a synergistic self-feeding cycle that ends in apoptotic cell death, once the antioxidant cellular defense systems are finally overwhelmed.

**Keywords: inflammation, neurodegeneration, mitochondrial dysfunction, iron toxicity, Parkinson's disease**

## **INTRODUCTION**

Iron is an essential element necessary for the normal development of brain functions. Enzymes involved in neurotransmitter synthesis that possess iron as a prosthetic group are recognized targets of iron deficiency: monoamine oxidases A and B involved in dopamine catabolism, tryptophan hydroxylase, required for serotonin synthesis, tyrosine hydroxylase, required for dopamine and norepinephrine synthesis, glutamate decarboxylase involved in GABA synthesis and glutamate transaminase involved in L-glutamate synthesis, all belong to this group.

Abundant evidence shows that iron accumulation in particular areas of the brain is a hallmark of several neurodegenerative disorders (Zecca et al., 2004; Andersen et al., 2013), although it is uncertain whether iron accumulation is a primary cause of the disorder or a consequence of a previous dysfunction. Increased levels of iron promote cell death via hydroxyl radical formation, which enhances lipid peroxidation, protein aggregation, glutathione consumption, and nucleic acid modification. We recently put forward the hypothesis that iron accumulation, a process initiated by mitochondrial dysfunction, and the ensuing oxidative damage is part of the execution step, i.e., the death process of affected neurons (Núñez et al., 2012).

Mitochondrial dysfunction has long been associated with several neurodegenerative diseases that include Alzheimer's disease (AD), Huntington's disease (HD), Parkinson's disease (PD), amyotrophic lateral sclerosis (ALS), and Friedrich's Ataxia (FA; Schapira and Cooper, 1992; Moreira et al., 2010; Grubman et al., 2013). Mitochondrial dysfunction results in decreased ATP synthesis, as well as in decreased synthesis of iron–sulfur clusters (ISCs) and heme prosthetic groups. An association between mitochondrial dysfunction and mitochondrial iron accumulation has been found only in FA (Delatycki et al., 1999; Huang et al., 2009), although evidence for mitochondrial iron accumulation has been reported in experimental models of PD (Liang and Patel, 2004; Lee et al., 2009; Mena et al., 2011).

Inflammation in the central nervous system (CNS) is a condition strongly associated with neuronal death in several neurodegenerative disorders including PD and AD (Hirsch and Hunot, 2009). Inflammation is characterized by the occurrence of reactive microglia and a massive production of proinflammatory cytokines. These inflammatory processes trigger a chain of events including increased production of ROS and reactive nitrogen species (RNS), disruption of iron metabolism and mitochondrial dysfunction, finally leading to neurodegeneration.

#### **THE BASIS OF IRON TOXICITY**

The ability of iron to exchange readily one electron underlies its insertion in numerous catalytic processes found in living matter. The iron atom has octahedral coordination chemistry; therefore, it has six possible coordination bonds. Seminal work by Graf and associates demonstrated that iron is redox-inactive only if all its six coordination sites are stably bound. If one of the sites is free or loosely bound, iron is redox-active and competent of undertaking one-electron exchange reactions (Graf et al., 1984). It is noteworthy that Fe3<sup>+</sup> complexes with the chelators desferrioxamine, DTPA or phytate at 1:10 (mol:mol) ratio result in redox-inactive iron whereas Fe3<sup>+</sup> chelation with NTA, EDTA, EGTA, ATP, CDTA or bleomycin results in redox-active iron at the same 1:10 molar ratio (Graf et al., 1984).

Iron is a paramagnetic element with two stable oxidation states: 2<sup>+</sup> and 3+. As mentioned above, both Fe2<sup>+</sup> and Fe3<sup>+</sup> establish coordination complexes with a great variety of ligands. Iron complexes display a variety of reduction potentials, ranging from very positive to negative values because of a basic concept in coordination chemistry that establishes that the ligand modifies the electron cloud surrounding the metal, thus modifying its reduction potential. This versatility in reduction potential allows for fine-tuning between iron reduction potential and the electron transfer process it catalyzes. It is estimated that the predominant reduction potential for iron in the intracellular milieu of the cell is near zero V (Clark, 1960; Wood, 1988). Many *in vitro* experiments confirm iron-mediated production of the hydroxyl radical (•OH), which arises from the following reactions:

$$\begin{aligned} \text{1. } \mathrm{Fe^{2+}} + \mathrm{O\_{2}} &\leftrightarrow \mathrm{Fe^{3+}} + \mathrm{O\_{2}^{\bullet-}} \mathrm{E\_{0}:} - 0.43 \mathrm{V}; \quad \Delta\mathrm{G} = 41.5 \mathrm{K} / \mathrm{mol} \\\\ \text{2. } \mathrm{O\_{2}^{\bullet-}} + 2 \mathrm{H^{+}} &\rightarrow \mathrm{H\_{2}O\_{2} + O\_{2}} \\\\ \text{3. } \mathrm{Fe^{2+}} + \mathrm{H\_{2}O\_{2}} &\rightarrow \mathrm{Fe^{3+}} + \mathrm{OH} + ^{\bullet} \mathrm{OH} &\rightarrow \mathrm{\_{0}:0.10V}; \\\\ \Delta\mathrm{G} &= -9.7 \mathrm{K} / \mathrm{mol} \end{aligned}$$

The thermodynamic sum of reactions 1–3 gives reaction 4:

 $4.3\text{Fe}^{2+} + 2\text{O}\_2 + 2\text{H}^+ \rightarrow \text{Fe}^{3+} + \text{OH}^- + ^\bullet \text{OH}$ 
$$\Delta \text{G} = -21.2 \text{K} / \text{mol}$$

The intracellular environment provides abundant reducing power in the form of GSH (mM) and Asc (μM), which reduces Fe3<sup>+</sup> to Fe2+:

$$\begin{aligned} \text{(5. Fe}^{3+} + \text{GSH(Asc)} \rightarrow \text{Fe}^{2+} + \text{GSSH}(\text{Asc} \bullet) + \text{H}^+ \text{E}\_0 &: 0.262;\\ \Delta \text{G} &= -25.3 \text{K} / \text{mol} \end{aligned}$$

Changes in free energy were calculated applying the equation Δ*G* = −*nFE*<sup>0</sup> (Joule/mol), in which *n* is the number of electrons exchanged and *F* the Faraday constant. Reaction 1 values were from (Pierre and Fontecave, 1999); Reaction 2, the half-cell potential for H2O2 dismutation was considerer 0.45 V (Pierre and Fontecave, 1999) and the reduction potential of the Fe3+/Fe2<sup>+</sup>

half-cell was considered 0 V (Wood, 1988); Reaction 3 (Fenton reaction): *E*<sup>0</sup> half-cell values from (Buettner, 1993; Buettner and Schafer, 2000). Half-cell potentials for reaction 5 were obtained from (Millis et al., 1993; Pierre and Fontecave, 1999). GSH: reduced glutathione; GSSG: oxidized glutathione; Asc: ascorbate; Asc•: ascorbate free radical.

The hydroxyl radical is considered one of the most reactive species generated in biological systems, since its reaction rate is only limited by diffusion, with rate constants in the 109– 1012 Mol−<sup>1</sup> s <sup>−</sup><sup>1</sup> range (Davies, 2005). This molecule induces irreversible damage to DNA, RNA, proteins, and lipids. Indeed, the hydroxyl radical is believed to be the etiological agent for several diseases and may be involved in the natural process of aging (Lipinski, 2011).

The main components of cell iron homeostasis are the divalent metal transporter 1 (DMT1), a Fe2<sup>+</sup> transporter that brings iron into the cell, the transferrin receptor 1 (TfR1) that brings iron in through the endocytosis of Ferro-transferrin, the iron export transporter ferroportin 1 (FPN1) and the cytosolic iron storage protein ferritin. The expression of these proteins is transductionally regulated by the iron responsive element/iron regulatory protein (IRE/IRP) system, which is activated when cells have low iron levels, resulting in increased DMT1 and TfR1 levels and decreased FPN1 and ferritin expression (Muckenthaler et al., 2008).

In cells, iron in the 0.2–1.5 μM range is weakly complexed to low-molecular weight substrates such as citrate, carboxylates, amines, phosphate, nucleotides, GSH, and other molecules conforming the "cytosolic labile iron pool" (cLIP; Epsztejn et al., 1997; Kakhlon and Cabantchik, 2002; Petrat et al., 2002; Hider and Kong, 2011). Iron in this pool is redox-active, cycling between the Fe+<sup>2</sup> and Fe+<sup>3</sup> forms, with prevalence of the reduced form because of the reductive cytosol environment. This redox-active pool is suitable to experimental detection by the fluorophore calcein, which has higher affinity for Fe3<sup>+</sup> than for Fe2<sup>+</sup> but since the reduction potential for iron in the Fe-calcein complex is low, Fe3<sup>+</sup> bound to calcein is readily reduced in the intracellular environment, resulting in decreased calcein fluorescence (Petrat et al., 2002). In cultured neuroblastoma cells the LIP represents about 3% of total cellular iron under basal culture conditions, but this percentage increases 3–4 fold, to μM concentrations, after exposure of cells to high extracellular iron concentrations (Núñez-Millacura et al., 2002; Núñez et al., 2004). In cell models, iron overload generates increased lipid peroxidation, protein modifications and damage to DNA, consistent with the production of the hydroxyl radical (Mello-Filho and Meneghini, 1991; Núñez et al., 2001; Sochaski et al., 2002; Zoccarato et al., 2005).

## **INFLAMMATORY CYTOKINES INDUCE THE PRODUCTION OF RNS, ROS AND IRON ACUMULATION**

Postmortem tissues from patients with AD, PD, HD, ALS or FA show oxidative damage in the affected brain regions (Nunomura et al., 1999; Barnham et al., 2004; Emerit et al., 2004). The association between inflammation and oxidative damage is mediated by the release of RNS and ROS during the inflammatory process. In particular, activated microglia have high levels of nitric oxide synthase (NOS) and NADPH oxidase (NOX), two enzyme systems that mediate the increase in the oxidative tone induced by inflammation.

Microglia, the brain-resident immune cells, are essential for the generation of the inflammatory response. They are activated by distress signals released from neighboring cells, initiating an innate response characterized by the production of pro-inflammatory cytokines and, incidentally, phagocytosis (McGeer et al., 1988; Colton and Wilcock, 2010). Indeed, many cases of AD and PD are accompanied by a dramatic proliferation of reactive amoeboid macrophages and activated microglia in the substantia nigra (SN) or frontal cortex (McGeer et al., 1988; Possel et al., 2000; Kiyota et al., 2009; Hewett and Hewett, 2012), together with high expression of pro-inflammatory cytokines (Bauer et al., 1991; Mogi et al., 1994; Muller et al., 1998; Nagatsu, 2002; Hewett and Hewett, 2012).

Inducible NOS (iNOS, also called NOS-2), which is scarcely expressed in the brain is induced during gliosis in pathological situations including AD (Aliev et al., 2009) and PD (Dawson and Dawson, 1998). Up-regulation of iNOS and of cyclo-oxygenase-1 and cyclo-oxygenase-2 in amoeboid microglia occurs in the SN of human PD patients (Knott et al., 2000). A study on the levels of iNOS mRNA in postmortem PD basal ganglia found a significant increase in iNOS expression in the dorsal two-thirds of the striatum and in the medial medullary lamina of the globus pallidus, accompanied by a reduction in iNOS mRNA expression in the putamen (Eve et al., 1998).

Inflammatory mediators, including LPS and some cytokines (TNF-α, IL-1β, and IFN-γ) induce the transcriptional activation of the iNOS gene in astrocytes and microglia via activation of the transcription factors STAT1 and NF-kB (Grzybicki et al., 1996; Possel et al., 2000; Hewett and Hewett, 2012). These factors translocate to the nucleus and bind to response elements present in the iNOS coding sequence. Upregulation of microglial iNOS expression is also observed after administration of 1-methyl-4 phenyl-1,2,3,6-tetrahydropyridine (MPTP; Liberatore et al., 1999; Tieu et al., 2003; Kokovay and Cunningham, 2005;Yokoyama et al., 2008). Interestingly, administration of MPTP produces significantly less neuronal loss in mice deficient in iNOS compared to their wild type counterparts (Dexter et al., 1986; Liberatore et al., 1999; Dehmer et al., 2000). In the 6-hydroxidopamine (6-OHDA) model, iNOS activity in the striatum induces neurodegeneration in rats. Pretreatment with the iNOS inhibitor L-NAME blocks amphetamine-induced rotations and significantly restores striatal dopamine (DA) levels in 6-OHDA treated rats (Barthwal et al., 2001). In neuroinflammatory models of PD, iNOS also participates in nigral neurodegeneration. Injection of LPS induces iNOS expression in the SN in a time- and dose-dependent manner; iNOS is present mainly in fully activated microglia with the characteristic amoeboid morphology. Furthermore, LPS-induced loss of dopaminergic neurons decreases significantly by administration of an iNOS inhibitor (Arimoto and Bing, 2003; Singh et al., 2005).

The iNOS enzyme is a relevant factor in the neurodegenerative process associated to AD. Early observations reported increased iNOS and nitrotyrosine protein modifications in AD brains, mainly in neurofibrillary tangle-bearing neurons and neuropil threads as well as in astrocytes (Vodovotz et al., 1996; Smith

et al., 1997; Wallace et al., 1997). Studies in transgenic mice overexpressing amyloid beta precursor protein (APP) demonstrated that several pathological changes such as vessel lesions, amyloid deposition and mitochondrial DNA deletions, are associated with the degree of NOS overexpression (Seyidova et al., 2004). Nevertheless, the APPsw/iNOS(−/−) mice, which express human APP mutations on an iNOS knockout background, show increased appearance of tau pathology, neuronal death, neuroinflammation and behavioral deficits compared with the parental APPsw mice (Colton et al., 2008). This evidence indicates that in AD, the production of NO can be protective or damaging, depending on the levels of NO production.

The phagocyte NOX is the main regulated source of ROS generation. The catalytic component of the NOX complex is composed by a family of multiple-pass transmembrane proteins, named NOX1–4. The most studied, NOX2, also known as gp91phox or phagocyte oxidase (PHOX), is highly expressed in innate immune cells including microglia and it is most likely the predominant NOX isoform expressed in astrocytes, while neurons express both NOX2 and NOX4 (Skalnik et al.,1991; Noh and Koh,2000; Lavigne et al., 2001; Abramov et al., 2004; Pawate et al., 2004). NOX2 forms a complex with p67phox, p47phox, p40phox, and p22phox subunits. Several stimuli induce NOX2 complex priming, including pro-inflammatory cytokines (TNF-α, IL-1β) and Toll-like receptor (TLR) agonists like LPS, peroxynitrite and proteases. The primed NOX2 complex requires yet additional activation to initiate substantial ROS production. PKC activators, growth factors, complement protein C5a and G protein-coupled receptor agonists generate afully active NOX complex (Yang et al.,2009,2013; Sareila et al., 2011).

Activation of NOX also occurs in experimental models of PD and AD. Treatment with MPTP results in increased synthesis of the proinflammatory cytokine IL-1β and increased membrane translocation of p67phox that is prevented by minocycline, a tetracycline derivative that exerts multiple anti-inflammatory effects (Wu et al., 2002). In addition, aging mice treated with MPTP display an increase in gp91phox and 3-nitrotyrosine (L'Episcopo et al., 2010; Huh et al., 2011). In agreement, gp91phox−/− mice display decreased degeneration of dopaminergic neurons induced by MPTP compared to wild type mice (Wu et al., 2003; Zhang et al., 2004). The unilateral injection of 6-OHDA into the right striatum of rats induces an increase of NOX1 and NOX2 both in the striatum and the SN. In concordance, dopaminergic neuronal and TNF-α and IFN-γ induction triggered by 6-OHDA are abrogated in the gp91phox−/− or minocycline treated mice (Hernandes et al., 2013). Additionally, striatal injection of 6-OHDA increases NOX1 expression in dopaminergic neurons in rat SN, and also increases 8-oxo-dG content, a marker of DNA oxidative damage. Moreover, NOX1 knockdown reduces 6-OHDA-induced oxidative DNA damage and dopaminergic neuronal degeneration (Choi et al., 2012).

Microglia of AD subjects display activated NOX2, resulting in the formation of ROS that are toxic to neighboring neurons (Shimohama et al., 2000). In conjunction, an increment in NOX1 and NOX3 mRNA levels in the frontal lobe tissue from AD brains was reported, suggesting the participation of other NOX family members in AD neuropathology (de la Monte and Wands, 2006). Recently, increased NOX-dependent ROS production in the superior/middle temporal gyri at the earliest clinical manifestations of disease, but not in late-stage AD, was reported (Bruce-Keller et al., 2010). Genetic inactivation of NOX2 in 12- to 15-month-old mice overexpressing the APPsw mutation (Tg2576 mice) results in reduced oxidative damage and rescues both the vascular and behavioral alterations observed in Tg2576 mice (Park et al., 2008). Studies done in cell cultures replicated the postmortem and animal findings on oxidative damage driven by NOX activation. Experiments using co-cultures of neuronal and glial cells found that Aβ acts preferentially on astrocytes but causes neuronal death (Abramov et al., 2004; Abramov and Duchen, 2005). The Aβ peptide causes transient increases in cytoplasmic calcium in astrocytes, associated with increased ROS generation, glutathione depletion and mitochondrial depolarization. Neuronal death after Aβ exposure was reduced both by NOX inhibitors and in the gp91phox knockout mice. These data are consistent with a sequence of events in which Aβ activates NOX in astrocytes by increasing cytoplasmic calcium, generating an oxidative burst that causes the death of neighboring neurons (Abramov et al., 2004; Abramov and Duchen, 2005; Park et al., 2008).

Inflammatory conditions such as those found in neurodegenerative diseases also affect iron homeostasis through transcriptional modification of iron transporters. In this context, the observation that the transcription factor NFκB induces DMT1 expression is highly relevantfor understanding the relationship between inflammation and iron homeostasis (Paradkar and Roth, 2006). We recently reported that the pro-inflammatory cytokines TNF-α, IL-6 and the TLR4 agonist LPS directly regulate DMT1mRNA and protein levels and induce a transient decrease in FPN1 protein, thus generating an increment of iron content in neurons and microglia (Urrutia et al., 2013). Supporting the results described above, a recent study using primary cultures of ventral mesencephalic neurons demonstrated that TNF-α or IL-1β induce an increment in DMT1 and TfR1 protein levels, together with a reduction of FPN1 levels, resulting in an increase in ferrous iron influx and decreased iron efflux in neurons (Wang et al.,2013). These findings were replicated in systemic tissues. Treatment of mouse splenocyte with LPS down-regulates the expression of FPN1 through a signaling mechanism mediated by TLR4 (Yang et al., 2002). Moreover, stimulation of macrophage cell lines with IFN-γ, TNF-α or LPS results in increased IRE-binding activity of IRP1 and IRP2, and increased DMT1 mRNA expression (Mulero and Brock, 1999; Wardrop and Richardson, 2000; Ludwiczek et al., 2003;Wang et al., 2005).

Considering that NFκB activation takes place downstream of TNF-α, IL-1 and LPS signaling pathways (Teeuwsen et al., 1991; Rothwell and Luheshi, 2000; Hanke and Kielian, 2011), inflammatory stimuli may induce DMT1 expression via NFκB activation. Indeed, TNF-α was detected in glial cells in the SN of PD patients but not in control subjects, together with immunoreactivity for TNF-α receptors in dopaminergic neurons of both control and PD patients (Boka et al., 1994). These findings are suggestive of a circuit in which activation of nigral microglia results in TNF-α secretion, which might increase iron uptake by dopaminergic neuron via NF-κB-induced DMT1 expression.

Indeed, an increase in the nuclear immunoreactivity of NFκB has been observed in PD brains or in animal models for this disease (Hunot et al., 1997), so it is possible that activation of NF-κB via inflammatory stimuli contributes to iron accumulation in PD. Accordingly, inflammation could induce the production of hydroxyl radical trough the activation of two parallel pathways: (i) through DMT1-mediated increase of intracellular iron levels and (ii) through increased hydrogen peroxide levels mediated by NOX activation.

A positive feedback loop can be established between ROS/RNS and inflammatory cytokines. ROS induce intracellular signaling pathways that result in the activation of transcriptional factors like NF-kB, AP-1 and Nrf-2, which regulate the expression of proinflammatory mediators such as Cox-2, MCP-1, IL-6, TNF-α, IL-1α, and IL-1β (Hensley et al., 2000; Thannickal and Fanburg, 2000; Ueda et al., 2002; Ridder and Schwaninger, 2009; Kitazawa et al., 2011; Guo et al., 2012; Kawamoto et al., 2012; Phani et al., 2012; Song et al., 2012; Zhang et al., 2012; Tobon-Velasco et al., 2013). These cytokines and chemokines, in turn, stimulate a cascade of events leading to increased oxidative stress via iNOS and NOX activation.

## **INFLAMMATORY CONDITIONS INDUCE MITOCHONDRIAL DYSFUNCTION**

The study of the relationship between inflammation and mitochondrial activity in the CNS is incipient. Intrastriatal injection of LPS induces mitochondrial dysfunction, microgliosis, iron accumulation and progressive degeneration of the dopamine nigro-striatal system (Zhang et al., 2005; Hunter et al., 2007, 2008; Choi et al., 2009), as observed in PD pathology. Similarly, cytokines such as IL-1β decrease mitochondrial activity through the production of NO in cardiomyocytes (Tatsumi et al., 2000).

Several reports indicate that TLRs regulate mitochondrial activity. Activation of TLR3 results in reduction of mitochondrial oxygen consumption mediated by opening of the permeability transition pore (Djafarzadeh et al., 2011). In co-cultures of cortical neurons with microglial cells, the TLR4 agonist LPS promotes decreased oxygen consumption and oxidative stress, with the subsequent nigral dopaminergic neuronal death in a rat model of inflammation (Xie et al., 2004; Hunter et al., 2007). Although these studies strongly suggest a link between TLRs and mitochondria dysfunction, further studies should clarify the molecular mechanisms involved and its relevance to particular neurodegenerative processes.

The production of ROS and RNS affects mitochondrial activity through destabilization of the ISCs (Cassina and Radi, 1996; Brown and Borutaite, 2004). The free radical superoxide damages and/or oxidizes 4Fe-4S clusters, which results in the formation of the "null" 3Fe-4S center form (Flint et al., 1993; Hausladen and Fridovich, 1994; Gardner et al., 1995; Bouton et al., 1996). Additionally, NO reacts with 4Fe-4S clusters generating [(NO)2Fe(SR)2] type complexes that inactivate several mitochondrial iron–sulfur enzymes including proteins which compose the electron transport chain (Drapier, 1997; see below). The above data are consistent with the notion that inflammation, ROS/RNS production, and mitochondrial dysfunction are linked processes.

Additionally, recent evidence shows that under certain conditions mitochondria can modulate the immune response. The mitochondrial protein MARCH5 (an ubiquitin E3 ligase constitutively expressed in the mitochondrion outer membrane) positively regulates TLR7 and TLR4 signaling, resulting in NFκB activation and expression of the NFκB-responsive genes IL-6 and TNF-α (Shi et al., 2011). In addition, activation of TLR1, TLR2 and TLR4 results in augmented mitochondrial ROS production by inducing translocation to mitochondria of TRAF6 (TLR signaling adaptor, tumor necrosis factor receptor-associated factor 6), which leads to the engagement and ubiquitination of ECSIT (evolutionarily conserved signaling intermediate in Toll pathways), a protein required for efficient assembly of mitochondrial complex I (West et al., 2011). It remains to be demonstrated whether this mechanism is operative in CNS cells.

Interestingly, mitochondrial ROS (mtROS) could arguably activate the inflammatory response. In vascular endothelium, mtROS act as intermediate signaling molecules to trigger production of IL-6 (Lee et al., 2010). In addition, patients with the autoinflammatory disorder TRAPS (tumor necrosis factor receptor-associated periodic syndrome), exhibit altered mitochondrial function with enhanced mtROS generation and increased production of IL-6, TNFα, and IL-1β; decreasing mtROS levels by the general antioxidant *N*-acetylcysteine effectively reduces inflammatory cytokine production after LPS stimulation (Bulua et al., 2011). These results point to novel pathways that link inflammation to mtROS production.

In summary, inflammation induces ROS production and mitochondrial dysfunction generating a self-feeding cycle that could lead to neurodegeneration in diseases where inflammation and oxidative damage are prevalent (**Figure 1**). In this cycle, [1] inflammation induces ROS and RNS generation by activation of the NOX and iNOS enzymes (Possel et al., 2000; Sareila et al., 2011; Hewett and Hewett, 2012); [2] in turn, ROS/RNS induce the

**FIGURE 1 | Inflammation causes ROS/RNS production, mitochondrial dysfunction, and iron accumulation.** Inflammation, oxidative damage, and mitochondrial dysfunction are common features of neurodegenerative diseases. A complex net of relationships connect these features, which through feedback mechanisms contribute to the evolvement of neuronal death (see text for details).

expression of inflammatory cytokines (Baeuerle and Henkel, 1994; Sen and Packer, 1996). [3] Additionally, inflammation induces mitochondrial dysfunction through activation of TLR signaling (Xie et al., 2004; Djafarzadeh et al., 2011). [4] ROS in turn induce mitochondrial dysfunction by destabilizing ISCs, which results in the inactivation of several mitochondrial iron–sulfur enzymes (Cassina and Radi, 1996; Brown and Borutaite, 2004). [5] Mitochondrial dysfunction leads to IRP1 activation and increased iron uptake (Lee et al., 2009; Mena et al., 2011). [6] Iron increases oxidative damage by transforming mild oxidative molecules like superoxide and hydrogen peroxide into the hydroxyl radical (Graf et al., 1984). [7] Electron transport chain inhibition increases ROS production by electron leak (Drose and Brandt, 2012), and arguably could modulate the innate immune response by TLR signaling regulation (Shi et al., 2011) [8]. Finally, [9] inflammation is likely to cause iron accumulation through induction of DMT1 expression and transient ferroportin decrease (Urrutia et al., 2013; Wang et al., 2013).

## **MITOCHONDRIAL DYSFUNCTION, INFLAMMATION AND IRON ACCUMULATION IN THE DEATH OF NEURONS IN PD**

Mitochondria have a key role in iron metabolism in association with the synthesis of ISCs and heme, prosthetic groups that are vital for cell function. Iron complexes are particularly relevant components of the electron transport chain: 12 proteins contain ISCs and eight proteins contain heme in their active centers (Rouault and Tong, 2005). Other proteins that have ISCs are the Krebs cycle enzymes aconitase and succinate dehydrogenase, ribonucleotide reductase, an enzyme that catalyzes the formation of deoxyribonucleotides from ribonucleotides, and ferrochelatase, involved in the addition of Fe to porphyrin IX during heme synthesis. We refer the reader to

Mitochondria have a redox-active iron pool (Petrat et al., 2001); an increase in this pool directly associates with an increase in oxidative damage and with calcium-dependent changes in the mitochondrial permeability transition pore (Pelizzoni et al., 2011; Kumfu et al., 2012; Zhang and Lemasters, 2013). Thus, cells must regulate tightly their mitochondrial Fe levels because an iron shortage affects numerous processes that have iron as a cofactor, including the electron transport chain, whereas an excess of redox-active iron promotes the generation of the noxious hydroxyl radical. How mitochondria regulate their iron content and what, if any, is the interplay between cytoplasmic and mitochondrial iron are incipient but highly relevant subjects to understand the mechanisms of mitochondrial dysfunction in neurodegenerative diseases.

There is increasing evidence that mitochondrial dysfunction plays an important role in the development of neurodegenerative diseases such as AD, HD, FA, and PD (Enns, 2003; Mandemakers et al., 2007; Sas et al., 2007; Gogvadze et al., 2009; Jellinger, 2009). Imbalances in ROS and ATP levels derived from mitochondrial dysfunction affect neurons particularly, given their dependence on ATP to propagate electrical signals, maintain ionic gradients, and facilitate anterograde and retrograde transport along axons (Su et al., 2013). The involvement of mitochondrial dysfunction in the pathophysiology of PD was noted very early in the study of the disease. Evidence of mitochondrial dysfunction in PD began in the eighties, when, after an intravenous injection of illicit drugs, four college students developed marked Parkinsonism. Analysis of the substances injected revealed the presence of MPTP, a compound metabolized by astrocytes into 1-methyl-4-phenylpyridinium (MPP+), which is then released into the extracellular space. MPP+ is taken up selectively by dopaminergic (DA) neurons where it inhibits mitochondrial complex I (Heikkila et al., 1984; Langston et al., 1984; Nicklas et al., 1985; Gautier et al., 2013). Further evidence showed that complex I activity and the number of complex I subunits are decreased in postmortem tissue of idiopathic PD patients (Bindoff et al., 1989; Mizuno et al., 1989; Schapira et al., 1989). These results strongly suggest that mitochondrial dysfunction is a pathognomonic sign in the pathophysiology of PD. Reduced complex-I activity and an increased susceptibility to MPP+ were also observed in cybrids containing mitochondrial DNA from PD patients (Swerdlow et al., 1996, 2001; Gu et al., 1998a), suggesting the presence of mitochondrial

DNA-encoded defects in PD (Chaturvedi and Flint Beal, 2013). Additionally, in the epidemiology field, the use in farming of the highly lipophilic pesticide rotenone, a potent inhibitor of mitochondrial complex I, has been linked to a higher incidence of PD in agricultural workers (Betarbet et al., 2000; Tanner et al., 2011; Pezzoli and Cereda, 2013).

Mitochondrial complex I is a major source of ROS. Complex I from mitochondria of PD patients contain 47% more protein carbonyls localized to catalytic subunits and a 34% decrease in complex I 8-kDa subunit. NADH-driven electron transfer rates through complex I inversely correlate with complex I protein oxidation status and with the reduction in the 8-kDa subunit protein levels (Keeney et al., 2006).

Knowledge on the mechanisms that associate mitochondrial dysfunction and iron dyshomeostasis in PD is incipient. Treatment of SH-SY5Y dopaminergic neuroblastoma cells with mitochondrial complex I inhibitors such as rotenone or MPP+ results in

**FIGURE 2 | A positive feedback loop in the death of neurons in PD.** Inhibition of mitochondrial complex I by endogenous or exogenous toxins or mutations in PD genes Parkin, Pink 1, Alpha-synuclein, DJ-1 or LRRK2 generates a multifactorial positive feedback loop. In this loop, complex I inhibition results in iron accumulation driven by decreased Fe-S cluster synthesis, IRP1 activation, increased DMT1 and TfR1 expression and decreased FPN1 expression, increased ROS levels and decreased

glutathione levels. Both increased oxidative stress and low GSH levels further inhibit complex I activity. Another input to this cycle is contributed by inflammatory cytokines that through self-feeding cycles induce mitochondrial dysfunction, increased ROS/RNS production and iron accumulation mediated by the transcriptional regulation of DMT1 and FPN1 (see text). The cumulative oxidative damage finally results in apoptotic death (see text for details).

ROS production and increased mitochondrial iron uptake (Lee et al., 2009; Mena et al., 2011). Moreover, inhibition of complex I by rotenone decreases the activity of three ISC-containing enzymes: mitochondrial and cytoplasmic aconitases and xanthine oxidase, and decreases the ISC content of glutamine phosphoribosyl pyrophosphate amidotransferase (Mena et al., 2011). The reduction in cytoplasmic aconitase activity is associated with an increase in iron regulatory IRP1 mRNA binding activity and with an increase in the mitochondrial labile iron pool (Mena et al., 2011). Since IRP1 activity post-transcriptionally regulates the expression of iron import proteins, ISC synthesis inhibition may result in a false iron deficiency signal with the ensuing iron accumulation.

Considering the evidence discussed, we propose that inhibition of mitochondrial complex I by endogenous and/or exogenous toxins or by inflammatory processes resulting from trauma or other causes, engage a vicious cycle of increased oxidative stress and increased iron accumulation (**Figure 2**). In this scheme, inhibition of mitochondrial complex I by endogenous or exogenous toxins, or because of mutations in PD genes Parkin, Pink 1, alphasynuclein, DJ-1 or LRRK2 (Langston and Ballard, 1983; Schapira et al., 1990; Hsu et al., 2000; Silvestri et al., 2005; Martin et al., 2006; Junn et al., 2009; Angeles et al., 2011; Mena et al., 2011), results in decreased electron transport chain activity [1] and the ensuing ATP synthesis decrease and ROS increase [2]. Decreased ATP levels impairs ISC synthesis that results in decreased activity of ISC-containing proteins and increased mRNA binding activity of the iron homeostasis protein IRP1. IRP1 activation leads to increased DMT1 and TfR1 expression (Lee et al., 2009; Mena et al., 2011) [3] and the ensuing iron accumulation (Asenjo, 1968; Dexter et al., 1987; Faucheux et al., 2003; Michaeli et al., 2007) [4]. Increased ROS and increased redox-active iron promotes the consumption of intracellular reductants such as GSH and ascorbate (Perry et al., 1982; Ehrhart and Zeevalk, 2003; Núñez et al., 2004; Jomova et al., 2010) [5], resulting in a further decrease in mitochondrial activity and ISC synthesis (Harley et al., 1993; Gu et al., 1998b; Jha et al., 2000; Chinta et al., 2007; Danielson et al., 2011). Another input to this cycle is contributed by inflammatory cytokines liberated by activated microglia and astrocytes (Mogi et al., 1994) [6], which enhance mitochondrial dysfunction (Tatsumi et al., 2000; Xie et al., 2004; Hunter et al., 2007; Djafarzadeh et al., 2011) [7], increase ROS production (Grzybicki et al., 1996) [8] and increase iron accumulation by modifying the expression of the iron transporters DMT1 and FPN1 (Urrutia et al., 2013; Wang et al., 2013) [9]. As discussed in the text, increased ROS back-feed the production of cytokines. Increased ROS levels, in particular increased hydroxyl radical generation, produces increased oxidative damage, which is counteracted by antioxidant defenses [10]. In time, the positive feedback loop of mitochondrial dysfunction, iron dyshomeostasis and inflammation results in alpha-synuclein aggregation, proteasomal dysfunction, changes in mitochondrial fission/fusion dynamics, opening of the mitochondrion PTP, increased cytoplasmic cytochrome c and activation of death pathways [11]. Debris and toxins from dying neurons enhance the activation of glial cells, which contributes to the inflammatory network (Zecca et al.,2008; Hirsch and Hunot,2009; Gao et al., 2011) [12].

In summary, because of the innate interconnectivity of mitochondrial complex I dysfunction, iron accumulation, oxidative stress, and inflammation, probably the initiation of any one of these factors will induce or enhance the others through the generation of a positive feedback loop that in time will end in apoptotic neuronal death. Still unanswered isthe question of why neurons of the SNc are so particularly prone to carry-on this cycle. On examination of this cycle, several therapeutic targets come to mind. Its intervention should result in prolonged life of the affected neurons.

## **ACKNOWLEDGMENTS**

This work was financed by grant 1130068 from FONDECYT, and grant ACT1114 from PIA-CONICYT.

## **REFERENCES**


Parkinson's disease using novel MRI contrasts. *Mov. Disord.* 22, 334–340. doi: 10.1002/mds.21227


low relative molecular mass complexes by macrophages. *Eur. J. Biochem.* 267, 6586–6593. doi: 10.1046/j.1432-1327.2000.01752.x


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

*Received: 09 January 2014; paper pending published: 29 January 2014; accepted: 19 February 2014; published online: 10 March 2014.*

*Citation: Urrutia PJ, Mena NP and Núñez MT (2014) The interplay between iron accumulation, mitochondrial dysfunction, and inflammation during the execution step of neurodegenerative disorders. Front. Pharmacol. 5:38. doi: 10.3389/fphar.2014.00038 This article was submitted to Drug Metabolism and Transport, a section of the journal Frontiers in Pharmacology.*

*Copyright © 2014 Urrutia, Mena and Núñez. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*

## Mitochondrial ferritin in the regulation of brain iron homeostasis and neurodegenerative diseases

## *Guofen Gao andYan-Zhong Chang\**

Laboratory of Molecular Iron Metabolism, College of Life Science, Hebei Normal University, Shijiazhuang, China

#### *Edited by:*

Raffaella Gozzelino, Instituto Gulbenkian de Ciência, Portugal

#### *Reviewed by:*

Sonia Levi, Vita-Salute San Raffaele University, Italy Ikuo Tooyama, Molecular Neuroscience Research Center–Shiga University of Medical Science, Japan

#### *\*Correspondence:*

Yan-Zhong Chang, Laboratory of Molecular Iron Metabolism, College of Life Science, Hebei Normal University, Shijiazhuang, Hebei Province 050016, China e-mail: chang7676@163.com

Mitochondrial ferritin (FtMt) is a novel iron-storage protein in mitochondria. Evidences have shown that FtMt is structurally and functionally similar to the cytosolic H-chain ferritin. It protects mitochondria from iron-induced oxidative damage presumably through sequestration of potentially harmful excess free iron. It also participates in the regulation of iron distribution between cytosol and mitochondrial contents. Unlike the ubiquitously expressed H-ferritin, FtMt is mainly expressed in testis and brain, which suggests its tissue-related roles. FtMt is involved in pathogenesis of neurodegenerative diseases, as its increased expression has been observed in Alzheimer's disease, restless legs syndrome and Friedreich's ataxia. Studies from our laboratory showed that in Alzheimer's disease, FtMt overexpression attenuated the β-amyloid induced neurotoxicity, which on the other hand increased significantly when FtMt expression was knocked down. It is also found that, by maintaining mitochondrial iron homeostasis, FtMt could prevent 6-hydroxydopamine induced dopaminergic cell damage in Parkinson's disease. These recent findings on FtMt regarding its functions in regulation of brain iron homeostasis and its protective role in pathogenesis of neurodegenerative diseases are summarized and reviewed.

**Keywords: mitochondrial ferritin, iron, brain, neurodegenerative diseases, oxidative damage**

## **INTRODUCTION**

Iron is an essential trace element for human health. In the brain, iron homeostasis is stringently regulated at three levels: organ, cellular, and subcellular, with different key regulatory molecules involved in each level. Dysregulation of brain iron homeostasis can lead to severe pathological changes in the neural system. For example, iron deficiency can slow down the development of neural system and cause language and motion disorders (Lozoff et al., 2000; Siddappa et al., 2004), while iron overload is closely related to neurodegenerative diseases (Zecca et al., 2004; Berg and Hochstrasser, 2006; Lee et al., 2006). The newly reported iron-storage protein, mitochondrial ferritin (FtMt) that locates in the mitochondria, possesses high homology to H-ferritin (Levi et al., 2001). It was reported that FtMt plays an important role in the regulation of cellular iron homeostasis (Corsi et al., 2002; Drysdale et al., 2002; Cazzola et al., 2003). Overexpression of FtMt affects iron homostasis and changes iron distribution between cytosol and mitochondria contents, and leads to cytosolic iron depletion (Corsi et al., 2002; Nie et al., 2005). In some neurodegenerative diseases characterized by iron overload, including Alzheimer's disease and Parkinson's disease (PD), increased expression of FtMt was observed (Shi et al., 2010; Wu et al., 2013). FtMt has a tissue-specific expression pattern and is rich in tissues with high metabolic activity, which is regarded as functionally important (Drysdale et al., 2002; Levi and Arosio, 2004). Evidences have shown that FtMt acts as a protective agent of neurons that maintains their normal functions and controls their apoptosis (Shi et al., 2010; Wang et al., 2011; Wu et al., 2013). Some of the molecular mechanisms underlying these protective functions were revealed recently, which

provided insights into the pathogenesis of neurodegenerative diseases and may help the development of new therapeutic strategies.

## **BRAIN IRON HOMEOSTASIS**

Human bodies contain 3–5 g of iron in average. Dietary iron is absorbed predominately in duodenum and enters blood circulation in small intestine. Once in blood circulation, iron binds to apotransferrin and forms transferrin (Tf). Tf is the major vehicle for iron transport in the body, and carries iron to other cells and tissues through the circulation. At the target cell, Tf binds to transferrin receptors (TfR) on the cell membrane, and the TfR-Tf-Fe complex is then endocytosed into the cell, where the iron is released. Free iron either enters mitochondrion for utilization in metabolic processes, such as synthesis of hemoglobin and Fe-S cluster, or is incorporated into the cytosolic iron-storage protein, ferritin, and serves as a cellular store of iron.

Iron needs to pass the blood-brain barrier in order to enter the brain. Tf-Fe in the blood circulation is uptaken at the surface of cerebral capillary endothelia, mainly through the classic TfR-mediated endocytosis (Bradbury, 1997). In addition, TfRindependent Tf-Fe uptake may also exist. As shown by Ueda et al. (1993), non-TfR bound iron was transported into the brain when the TfR-mediated iron transport was maximally inhibited by anti-TfR antibodies. Free iron can also enter the brain barriers by divalent metal transport-1 (DMT1), a proton driven transporter (Siddappa et al., 2002; Skjorringe et al., 2012). In endothelia, iron is released and transported across the abluminal membrane of the barriers into the cerebral compartment. This process likely involves iron exporter ferroportin (FPN) and DMT1 on the abluminal membrane, but the exact mechanism remains for further exploration (Moos et al., 2007; Mills et al., 2010; Zheng and Monnot, 2012). The elemental iron released into the brain interstitial fluid binds to brain Tf and becomes available for neurons and neuroglia expressing TfR (Han et al., 2003). The excess iron in neurons and neuroglia can be exported back to the brain interstitial fluid, and can be released into the cerebrospinal fluid in the brain ventricles through bulk flow (Bradbury, 1997; Zheng and Monnot, 2012). The apical microvilli of choroidal epithelia then capture the free iron by TfR or DMT1 and transport it back to the blood circulation (Mills et al., 2010).

Iron homeostasis in brain is precisely regulated. At the cellular level, iron homeostasis is mainly regulated by iron transporters TfR, DMT1, and FPN. It has been reported that the uneven distribution of TfR in cerebral endothelia is responsible for the differences of iron concentrations in different brain regions (Deane et al., 2004). Iron concentrations are high in the striatum and the hippocampus where higher TfR density and iron uptake rate are also observed (Deane et al., 2004), but are low in the cortex and the brain stem (Morris et al., 1992; Sugawara et al., 1992). Similar to the iron regulation at the peripheral, iron homeostasis in brain is tightly regulated by iron regulatory proteins (IRPs) IRP1 and IRP2 (Rouault, 2006). When the brain cellular iron concentration is low, the active center of IRPs binds to the stemloop structure of the iron-responsive element (IRE) located at the 3- -untranslated region (UTR) of TfR mRNA. This binding stabilizes TfR mRNA and increases its cellular expression level, thereby increasing iron uptake. When the iron concentration is high, the active center of IRP is occupied by four Fe-S, which blocks the binding of IRP to the IRE of TfR, resulting in low TfR translation level and reduced iron uptake (Leipuviene and Theil, 2007). The IRP/IRE system also regulates the stability of DMT1 with IRE (+IRE), FPN, and ferritin. However, binding of IRP to the IRE of FPN and ferritin decreases their stabilities, and causes lower protein expression (Rouault, 2006; Leipuviene and Theil, 2007). Thus, IRPs play a key role in the maintenance of cellular iron homeostasis. Studies of our laboratory and others have found that the IRP2−/<sup>−</sup> mice had significant misregulation of iron metabolism and developed neurodegeneration (Meyron-Holtz, 2004). Inside the cells, the iron storage level and the cellular liable iron level (LIP) are largely dependent upon the availability of the iron-storage protein, ferritin. Ferritin is a ubiquitous protein with an iron core that can accommodate up to 4500 iron atoms (Theil, 1987; Harrison and Arosio, 1996). It is a 24-mer globular protein complex that is made up of heart (H) and liver (L) subunits, the H-ferritin (21 kDa) and the L-ferritin (19 kDa), respectively (Ford et al., 1984; Theil, 1987). The ability of ferritin to sequester iron provides its dual functions, iron segregation in a non-toxic form and iron storage (Harrison and Arosio, 1996; Torti and Torti, 2002).

At the systematic level, brain iron homeostasis may involve the regulation of an peptide "hormone" hepcidin (Crichton et al., 2011). Hepcidin is mainly produced by hepatocytes in response to high iron concentration, inflammatory stimuli or hypoxia (Park et al., 2001; Nicolas et al., 2002). It binds to the extracellular loop of FPN and causes its internalization and degradation, and

thereby reduces cellular iron efflux (Ramey et al., 2010; Anderson and Wang, 2012). Several recent studies reported the identification of hepcidin producing cells in the brain and investigated hepcidin's functions under normal and pathological conditions (Zechel et al., 2006; Marques et al., 2009; Wang et al., 2010; Crichton et al., 2011). Zechel et al. (2006) showed that hepcidin is widely expressed in different brain areas, including the cortex, hippocampus, thalamus, cerebellum, spinal cord, and so on, in both neurons and in GFAP-positive glia cells. Increased hepcidin expression was detected in choroid plexus of the brain in response to peripheral inflammation (Marques et al., 2009). Studies in our lab also found that hepcidin mRNA levels in different brain regions increased with aging, and injection of hepcidin into the lateral cerebral ventricle decreased FPN levels and resulted in brain iron overload (Wang et al., 2010). These findings implied the important regulatory role of hepcidin on brain iron metabolism, though the cellular mechanisms remain to be elucidated.

## **MITOCHONDRIAL IRON HOMEOSTASIS AND MITOCHONDRIAL FERRITIN**

## **MITOCHONDRIAL IRON METABOLISM**

Although most iron is stored in the cytosol, the major flux of iron in many cells occurs in the mitochondria, where various metabolic activities occur. Fe-S clusters and heme biogenesis are the main events in which iron is utilized (Ponka, 1997). Iron transport into mitochondria is directly coupled with its uptake at the cell membrane (Pandolfo, 2002; Levi and Rovida, 2009). Several mechanisms have been proposed on the pathway of iron entry into mitochondria. One hypothesis proposed by Ponka (1997)suggests that iron is directly delivered to mitochondria by endosomes in a "kiss and run" paradigm. Another theory proposed by Shvartsman et al. (2007) suggests that no endosomal vesicle is involved in the transport of non-Tf-bound iron to mitochondria. This was supported by the observation that mitochondrial iron uptake was not hampered by the use of cellular compartment-specific iron chelators, and chaperones were bound to the incoming iron prior to its delivery to micochondria (Shvartsman et al., 2007). Researchers also investigated the involvement of iron transport proteins on the mitochondrial membrane, such as MRS3 and MRS4 identified in yeast (Foury and Roganti, 2002) and Mitoferrin 1 and Mitoferrin 2 found in zebra fish (Paradkar et al., 2009). Iron transport out from mitochondria may depend on adequate Fe-S synthesis (Pandolfo, 2002).

Iron flux in mitochondria must be precisely regulated because excess free iron can result in the production of damaging free reactive oxygen species (ROS) during electron transport (Eaton and Qian, 2002). Dysregulation of mitochondrial iron metabolism can severely affect the intracellular iron homeostasis, resulting in mitochondrial iron metabolism diseases, such as Friedreich ataxia (FRDA; Schmucker and Puccio, 2010). However, little is known about the regulatory mechanisms of iron trafficking and communication between cytosol and mitochondria. It has been reported that ferritins, under the influence of iron and oxygen metabolism, exert cellular protective roles against iron-mediated free radical damage (Arosio and Levi, 2002; Arosio et al., 2009). The newly identified H-ferritin-like protein

in mitochondria, FtMt, has been shown to modulate cellular iron metabolism and influence ROS level dramatically (Levi et al., 2001; Corsi et al., 2002; Nie et al., 2005). Studying the role of FtMt in mitochondria iron homeostasis may provide new insights into the treatment of diseases associated with abnormal iron homeostasis.

#### **FtMt SYNTHESIS AND DISTRIBUTION**

Mitochondrial ferritin was first identified in 2001 as a new human ferritin type that specifically locates in mitochondria (Levi et al., 2001). Other primates, mice, and rats also express this gene, which is highly homologous to human FtMt. The human FtMt gene is intronless and locates at chromosome 5q23.1. It encodes a ∼1 kb mRNA that translates to a 242 amino-acid FtMt precursor protein with a ∼60 amino-acid mitochondrial targeting signal sequence at the N-terminus. The sequence of the mature human FtMt has a 79% identity to the H-chain ferritin. The ferroxidase centers of FtMt and H-ferritin share a completely conserved sequence and a fully overlapped crystallographic structure (Langlois d- Estaintot et al., 2004), indicating their similar functions. Recombinant FtMt was proven to have iron incorporation activity *in vitro* that was as efficient as H-ferritin (Bou-Abdallah et al., 2005). However, unlike the cytosolic ferritins, FtMt mRNAs lack the IRE consensus sequences for iron-dependent translational regulation.

The ∼30 KDa human FtMt precursor protein is translocated to the mitochondria after synthesis, and is processed to become the ∼22 KDa mature protein as the subunit to form typical ferritin shells (Corsi et al., 2002). Unlike the ubiquitously expressed cytosolic H-ferritin, the expression of FtMt is tissue-specific, showing a high level of transcription in testis and brain. Immunohistochemistry analyses of mouse FtMt showed its expression in spermatids and interstitial cells, neuronal cells of brain and spinal cord, and some other tissues. But surprisingly no expression was detected in hepatocytes, splenocytes, or myocytes (Drysdale et al., 2002; Levi and Arosio, 2004; Santambrogio et al., 2007). This further suggests that FtMt expression is not related to the cellular iron level, and the expression pattern may reflect its tissue-related roles. It was also found that, in the pathological conditions associated with mitochondrial iron overload, such as Alzheimer's disease, PD, and sideroblastic anemia, the FtMt expression was largely induced (Cazzola et al., 2003; Shi et al., 2010; Wang et al., 2011; Wu et al., 2013; Yang et al., 2013).

## **ROLE OF FtMt IN MITOCHONDRIAL AND CYTOSOLIC IRON DISTRIBUTION**

As mentioned above, FtMt is structurally and functional similar to H-ferritin. The main biological function of FtMt is to incorporate excess free iron. It had a reduced ferroxidase activity as compared to H-ferritin, but the iron sequestering efficiency is as high (Corsi et al., 2002; Levi and Arosio, 2004). In addition to iron sequestration, FtMt was extensively studied on its function of maintaining intracellular iron homeostasis by modulating the traffick of iron in cytoplasm (Levi et al., 2001; Corsi et al., 2002; Nie et al., 2005). Corsi et al. (2002) found that overexpression of human FtMt in Hela cells resulted in decreased cytosolic ferritin and increased TfR levels and cytosolic iron deficiency. Using a stable cell line

transfected with mouse *FtMt* gene, Nie et al. (2005) also observed that FtMt dramatically affected intracellular iron metabolism. Overexpression of FtMt caused an increase in cellular iron uptake but a decreased cytosolic iron level associated with decreased cytosolic ferritin, suggesting that the increased iron influx was preferentially transferred into mitochondria and incorporated into FtMt rather than into cytosol (Nie et al., 2005). They also found that the expression of FtMt was associated with decreased mitochondrial and cytosolic aconitase activities, which was consistent with the increase in IRP-IRE mRNA binding activity (Nie et al., 2005). In addition, increased expression of FtMt was found in some genetic diseases associated with cellular iron deficiency and mitochondrial iron overload, such as the restless legs syndrome (RLS; Ondo, 2005; Snyder et al., 2009). Many detailed advances in the research of FtMt and related diseases are summarized below.

## **MITOCHONDRIAL FERRITIN IN THE PATHOPHYSIOLOGY OF NEURODEGENERATIVE DISEASES IRON, ROS AND CELL APOPTOSIS**

Excess iron in brain is known to cause neurodegeneration in adults (Zecca et al., 2004). Increased ferrous iron (Fe2+) levels can lead to the production of highly reactive hydroxyl radical via the Fenton reaction. Increased iron levels can also generate peroxyl/alkoxyl radicals due to Fe2+-dependent lipid peroxidation (Pollitt, 1999). These ROS can damage cellular macromolecules including proteins, lipids and DNA, and finally the oxidative stress will trigger apoptosis. Iron-induced oxidative stress can be very destructive because a positive-feedback loop can develop from the release of more free iron from the iron-containing proteins, such as ferritin, heme proteins, and Fe-S clusters. As a result, the toxic effect of brain iron overload is exacerbated.

#### **FtMt IN THE PATHOPHYSIOLOGY OF PARKINSON'S DISEASE**

Parkinson's disease is a common neurodegenerative disease characterized by the loss of dopaminergic neurons in the substantial nigra (SN) of the brain and the formation of filamentous intraneuronal inclusions (Parkinson, 2002). The pathogenesis of PD involves accumulation of non-heme iron in the SN and nigra, oxidative damages and dysfunctions of mitochondria (Halliwell, 1992; Zhang et al., 2000; Zecca et al., 2004; Lin and Beal, 2006). Studies in our lab have shown that FtMt maintains iron homeostasis and prevents neuronal damage in a 6-Hydroxydopamine (6-OHDA)-induced parkinsonian phenotype (Shi et al., 2010). In our studies, the neuroblastoma SH-SY5Y cells were stably transfected with FtMt gene, and the PD model was established by induction with the neurotoxin 6-OHDA. We found that overexpression of FtMt significantly protected neuronal cells from the 6-OHDA-induced cell death. A possible mechanism of this protection was proposed which involves the regulation of Bcl-2, Bax, and caspase-3 apoptotic pathways (**Figure 1**). FtMt attenuated ROS accumulation and lipid peroxidation, and inhibited mitochondrial damage induced by neurotoxin 6-OHDA. Moreover, FtMt strongly inhibited the elevation of iron levels and prevented the alteration of iron redistribution induced by 6-OHDA. These findings suggest that FtMt plays a neuroprotective role in PD by affecting the iron metabolism.

#### **FtMt IN THE PATHOPHYSIOLOGY OF ALZHEIMER'S DISEASE**

Alzheimer's disease (AD) is a common neurodegenerative disease in aged people. The brains of AD patients are characterized by extracellular plaques of amyloid-β (Aβ) and neurofibrillary tangles of tau protein (Selkoe, 1996). Aβ plays an important role in the pathophysiological mechanisms of AD, as it accumulates to abnormally high levels in the brains of AD patients and directly induces neuronal cell death (Selkoe, 2000, 2001). Although abnormal iron metabolism and impaired mitochondrial function have been reported in AD, little information is available about the role of FtMt in the pathogenesis of AD.

A recent study by Wang et al. (2011) investigated the expression and localization of FtMt in the temporal cortex and cerebellum of AD patients. By using RT-PCR, they found that the FtMt mRNA levels in the temporal cortex of AD patients were evidently increased as compared to the controls, but no significant differences of mRNA levels was found in the cerebellum. By *in situ* hybridization histochemistry, FtMt mRNAs were localized mainly in the neurons of the AD cortex. They also found that in human neuroblastoma cell IMR-32, FtMt expression was significantly induced by H2O2 treatment, and the increase in FtMt expression was dramatically accelerated when cells were treated with the combination of H2O2 and Aβ neurotoxin. Overexpression of FtMt in the IMR-32 cells also rescued the cell death induced by H2O2. These results indicated a neuroprotection effect of FtMt against oxidative stress and the involvement of FtMt in the pathological process of AD. However, the underlying molecular mechanisms of FtMt's action in AD and AD-like syndromes have not been fully elucidated.

To explore these mechanisms, our previous study by Wu et al. (2013) investigated the role of FtMt in Aβ25–35 treated rats. After the siRNA of FtMt was transfected into the hippocampus of the rats, we found that the FtMt down-regulated group released more cytochrome C, a sign of mitochondrial-dependent apoptosis, into the cytoplasm as compared to that of the control group. Increased number of apoptotic cells, decreased Bcl-2/Bax ratio and enhanced caspase-3 activation were observed, indicating a clear neuroprotectiove role FtMt plays *in vivo*. After treatment with Aβ25–35, knockdown of FtMt aggravated apoptosis in the hippocampus and oxidative damage to the tissue, as evidenced by increased levels of malonyl dialdehyde (MDA), protein carbonyls, and hydroxynonenal–histidine. The activities of the mitochondrial complex enzymes I–IV were also significantly decreased. To verify that the increased apoptosis was related to the low level of FtMt, we carried out further studies using SH-SY5Y cells that stably overexpressing FtMt. The results showed that FtMt overexpression reduced apoptosis in response to Aβ25–35 treatment and reduced the production of ROS as well. When FtMt was overexpressed in SH-SY5Y cells, the increase in caspase-3 protein and the reduction in the Bcl-2/Bax protein ratio following the Aβ25–35 treatment were largely neutralized. We further proposed that the direct neuroprotective effects of FtMt against Aβ25–35 toxicity could signal through the activation of the MAPK pathway in neurons, as the increase of extracellular signal regulated kinase (ErK) expression and the decrease of P38 level were observed.

Evidences accumulated thus far have shown that iron metabolism is closely related to the production of oxidative stress and the pathogenesis of neurodegenerative diseases. We further determined the correlation of iron with the mechanism in which FtMt reduces ROS levels in the Aβ25–35-treated cells. In our study, FtMt overexpression dramatically inhibited the elevation of LIP levels resulted by the Aβ25–35 treatment. To verify that the change of LIP is involved in the protective function of FtMt, we measured the levels of iron related proteins. We observed that overexpression of FtMt increased the TfR level and decreased the H-ferritin level in Aβ25–35-treated cells (Shi et al., 2010; Wu et al., 2013). Without FtMt overexpression, these levels were measured to go the reverse way. These findings suggested that FtMt redistributed iron from the cytosol to the mitochondria, resulting in a reduction of cytosolic iron levels. This in turn attenuated Aβ25– 35-induced neurotoxicity and reduced oxidative damage through the Erk/P38 kinase signaling. Our data also suggested that these effects were coordinately regulated by the intracellular LIP levels. Based on all these results, we proposed a possible neuroprotective mechanism of FtMt following the Aβ25–35 treatment, as shown in **Figure 2**.

## **FtMt IN THE PATHOPHYSIOLOGY OF FRIEDREICH ATAXIA**

Friedreich's ataxia is the most common genetic ataxia that caused by the deficiency of mitochondrial iron-binding protein frataxin (Schmucker and Puccio, 2010). The FRDA patients have severe mitochondrial iron overload, disruption of iron-sulfur cluster biosynthesis, and increased sensitivity to oxidative stress (Schmucker and Puccio, 2010). The protective role of FtMt in FRDA was first suggested by Campanella et al. (2004) in a study on frataxin-deficient yeast cells. FtMt expression rescued the respiratory deficiency caused by the loss of frataxin and

protected the activity of iron–sulfur enzymes in yeast. It also prevented yeast cells from developing mitochondrial iron overload, preserved the mitochondrial DNA integrity and increased resistance to H2O2. These data implied that FtMt could substitute most functions of frataxin in yeast, thus might play a protective role in FRDA. A follow-up study by Campanella et al. (2009) showed a similar function of FtMt in mammalian cells, including HeLa cells, and fibroblasts from FRDA patients. FtMt reduced the ROS level, increased the activity of mitochondrial Fe-S enzymes and the cell viability. Furthermore, FtMt expression reduced the LIP levels in both cytosol and mitochondria (Campanella et al., 2009). These results indicate that FtMt is involved in the regulation of iron distribution and availability in mitochondria and cytosol, thus controls ROS formation and protects cells characterized as defective in iron homeostasis and respiration.

## **FtMt IN RESTLESS LEGS SYNDROME**

Restless legs syndrome is a sensorimotor disorder. RLS patients are usually characterized as to have an urge to move the legs and to have abnormal sensations in the legs, especially in evenings and nights (Ondo, 2005). Unlike other neurodegenerative diseases, RLS was reported to have decreased cellular iron concentration in the brain and altered expression of iron metabolism-related proteins. Significant iron deficiency was observed in the neurons of SN in RLS patients (Connor et al., 2003; Schmidauer et al., 2005; Godau et al., 2007), and decreased ferritin and TfR and increased Tf were also observed, attesting the cellular iron deficient status (Connor et al., 2004). Considering the important role of iron in the redox reactions in mitochondria, Snyder et al. (2009) studied

the expression pattern of FtMt in the brain of RLS patients. The results showed that the staining of FtMt increased significantly in the RLS cases, and the neuromelanin-containing neurons in the SN were found to be the predominant cell type expressing FtMt. Since the numbers of mitochondria were also increased in the neurons, whether the increase of FtMt was a result of higher FtMt expression or from mitochondrial proliferation with normal amounts of FtMt could not be concluded. However, less cytosolic H-ferritin were observed in neurons of RLS cases, suggesting that the increased FtMt levels might contribute to the insufficient cytosolic iron levels in the SN neurons, thereby accelerating the pathogenesis of RLS (Snyder et al., 2009). Still, very little is known about the metabolic activity of SN and the role of FtMt in RLS, and further investigations are needed to understand more on the mechanisms.

## **SUMMARY**

Mitochondrial ferritin is a novel ferritin type that specifically locates in mitochondria. It is highly expressed in tissues with high metabolic activity and oxygen consumption, such as testis, brain, heart, and so on. This tissue specificity may correlate with its function. Studies so far suggest that FtMt plays a role in the protection of mitochondria from iron-dependent oxidative damage by sequestering the free excess iron. Current findings suggested important roles of FtMt in the pathogenesis of neurodegenerative diseases. The increased expression of FtMt in AD, PD, and other neurological disorders may relate to its neuroprotective role against iron overload and oxidative stress. But in RLS, its increased expression may link to the onset of disease rather than neuroprotection. Since FtMt lacks the IRE in its mRNA, which is different from otherferritins, it should not be regulated by iron directly. Further studies regarding the detailed mechanisms of the regulation of FtMt expression and the role FtMt plays in neurological disorders associated with abnormal iron metabolism are important topics that need to be explored in the future.

## **AUTHOR CONTRIBUTIONS**

Yan-Zhong Chang conceived the review and participated in design and discussion; Guofen Gao drafted the manuscript and participated in discussion. All authors read and approved the final manuscript.

## **ACKNOWLEDGMENTS**

This project was supported by the National Natural Science Foundation of China (31340064), the Natural Science Foundation of Hebei Province (C2012205082), and the Science and technology research Youth Fund Project of Hebei Colleges and Universities (Q2012036).

## **REFERENCES**


of antibodies against the transferrin receptor. *J. Neurochem.* 60, 106–113. doi: 10.1111/j.1471-4159.1993.tb05828.x


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

*Received: 14 January 2014; paper pending published: 26 January 2014; accepted: 30 January 2014; published online: 17 February 2014.*

*Citation: Gao G and Chang Y-Z (2014) Mitochondrial ferritin in the regulation of brain iron homeostasis and neurodegenerative diseases. Front. Pharmacol. 5:19. doi: 10.3389/fphar.2014.00019*

*This article was submitted to Drug Metabolism and Transport, a section of the journal Frontiers in Pharmacology.*

*Copyright © 2014 Gao and Chang. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*

## Mitochondrial iron–sulfur cluster dysfunction in neurodegenerative disease

## *Grazia Isaya\**

Department of Pediatric & Adolescent Medicine and Mayo Clinic Children's Center, Mayo Clinic, Rochester, MN, USA

#### *Edited by:*

Paolo Arosio, University of Brescia, Italy

#### *Reviewed by:*

Stanislav Yanev, Institute of Neurobiology – Bulgarian Academy of Sciences, Bulgaria Zvi Ioav Cabantchik, Hebrew University of Jerusalem, Israel Francesc Palau, Centro de Investigación Príncipe Felipe, Spain

#### *\*Correspondence:*

Grazia Isaya, Department of Pediatric & Adolescent Medicine and Mayo Clinic Children's Center, Mayo Clinic, 200 First Street SW, Stabile 7-52, Rochester, MN 55905, USA e-mail: isaya@mayo.edu

Growing evidence supports a role for mitochondrial iron metabolism in the pathophysiology of neurodegenerative disorders such as Friedreich ataxia (FRDA) and Parkinson disease (PD) as well as in the motor and cognitive decline associated with the aging process. Iron– sulfur enzyme deficits and regional iron accumulation have been observed in each of these conditions. In spite of significant etiological, clinical and pathological differences that exist between FRDA and PD, it is possible that defects in mitochondrial iron–sulfur clusters (ISCs) biogenesis represent a common underlying mechanism leading to abnormal intracellular iron distribution with mitochondrial iron accumulation, oxidative phosphorylation deficits and oxidative stress in susceptible cells and specific regions of the nervous system. Moreover, a similar mechanism may contribute to the age-dependent iron accumulation that occurs in certain brain regions such as the globus pallidus and the substantia nigra. Targeting chelatable iron and reactive oxygen species appear as possible therapeutic options for FRDA and PD, and possibly other age-related neurodegenerative conditions. However, new technology to interrogate ISC synthesis in humans is needed to (i) assess how defects in this pathway contribute to the natural history of neurodegenerative disorders and (ii) develop treatments to correct those defects early in the disease process, before they cause irreversible neuronal cell damage.

**Keywords: Friedreich ataxia, Parkinson disease, aging, mitochondria, iron–sulfur clusters, oxidative damage, antioxidants, iron-chelators**

## **THE NATURAL HISTORY OF NEURODEGENERATIVE DISEASE**

Although neurodegenerative disorders can present at different ages and with a broad variety of symptoms, their natural histories can be recapitulated by a few common steps (**Figure 1**): Normally developed and overall healthy neuronal cells are exposed to an insult that initiates the neurodegenerative process; this leads to progressive neuronal cell dysfunction and death, which ultimately leads to clinical signs and symptoms of neurological impairment. However, each of these common steps is influenced by different disease-specific factors. The initiating insult can be geneticallydetermined or environmentally-determined or result from a combination of both genetic and environmental factors. In addition, for many neurodegenerative disorders aging is a consistently important risk factor (Zecca et al., 2004b). The neurodegenerative process is often neuron-type specific–for example, motor neurons are exquisitely affected in amyotrophic lateral sclerosis, sensory neurons in Friedreich ataxia (FRDA), and dopaminergic neurons in Parkinson disease (PD) (Koeppen, 2011; Oshiro et al., 2011). In all cases, the neurodegenerative process is relentlessly progressive and there is a threshold for neurological manifestations, meaning that a critical mass of affected neuronal cells will have to become dysfunctional and/or die in order for clinical signs and symptoms to become apparent. The rate of progression and the threshold may vary depending on the cell type and respective brain region affected as well as additional factors that may influence disease severity (e.g., the presence of concomitant cardiac insufficiency and skeletal muscle weakness in FRDA). Importantly, before the threshold is reached there is a pre-symptomatic period during which the degenerative process is already advancing.

By plotting disease progression as a function of age with an arbitrary threshold for appearance of clinical signs and symptoms we can define three main modalities in the natural history of neurodegenerative disease (**Figure 2**). In most individuals there are slowly progressing, age-related degenerative changes that become clinically apparent only at an advanced age, manifesting as a decline in motor and cognitive performance after the 7th or 8th decade of life. In addition, there are conditions in which neurodegeneration occurs at a much faster pace with onset around the 6th,5th, or 4th decade or life, and others in which onset is much earlier, in the 2nd or even the 1st decade of life (**Figure 2**).

Below we will address the question as to whether and how defects in mitochondrial iron metabolism may contribute to these clinical scenarios. Indeed, there are at least two conditions, FRDA and PD, that provide paradigms for the link between mitochondrial iron dysfunction and, respectively, early-onset and late-onset neurodegeneration (Zecca et al., 2004a; Koeppen et al., 2007; Koeppen, 2011) [reviewed in (Horowitz and Greenamyre, 2010; Oshiro et al., 2011; Vaubel and Isaya, 2013)]. **Table 1** summarizes published evidence which together suggests that FRDA and PD share certain common aspects, also observed in the aging brain. These aspects include (i) regional iron accumulation in specific regions of the central and/or

**in the natural history of neurodegenerative disorders.** See text for details.

**FIGURE 2 | Graphic representation of the three main modalities in the natural history of neurodegenerative disease.** The rate of neurodegenerative disease progression is plotted as a function of age with an arbitrary threshold for appearance of clinical signs and symptoms. The plots show typical rates of progression for old age-related, late-onset and early-onset neurodegenerative disorders. The portion of each plot below the threshold represents the pre-symptomatic period during which the degenerative process is already active. FRDA, Friedreich ataxia; PD, Parkinson disease.

peripheral nervous systems (Zecca et al., 1996; Bartzokis et al., 2004; Boddaert et al., 2007; Oakley et al., 2007; Koeppen, 2011); (ii) cellular iron re-distribution within affected cell types that may result in mitochondrial iron accumulation and iron-catalyzed Fenton chemistry (Zecca et al., 2004a; Whitnall et al., 2008; Mastroberardino et al., 2009); and (iii) the presence of iron–sulfur enzyme deficits (Rotig et al., 1997; Longo et al., 1999; Betarbet et al., 2000) [reviewed in (Xu et al., 2010; Gille and Reichmann, 2011; Vaubel and Isaya, 2013; **Table 1**)]. This evidence supports the view that a common underlying mechanism involving iron metabolism–namely, defects in the biogenesis of mitochondrial iron–sulfur clusters (ISCs) and related enzymes–contributes to FRDA, PD and the aging process, in spite of these three conditions

being clinically, pathologically and etiologically very different from one another. Several excellent recent review articles recapitulate the current understanding of the molecules and mechanisms involved in ISC synthesis as well as the roles of iron dysregulation in the pathophysiology of FRDA, PD, and the aging process. Here we will review and discuss primarily the roles of ISC dysfunction in these conditions.

## **IRON–SULFUR CLUSTER SYNTHESIS AND NEURODEGENERATIVE DISEASE**

Mitochondria across eukaryotes contain a machinery that is responsible for the biogenesis of ISC inside mitochondria (Schilke et al., 1999; Muhlenhoff et al., 2002), and that also somehow regulates the assembly of ISC in other cellular compartments (Kispal et al., 1999; Gerber et al., 2004; Pondarre et al., 2006; Martelli et al., 2007). ISC-dependent enzymes are present in the mitochondria, the cytoplasm and the nucleus where they participate in such processes as the citric acid cycle and the electron transport chain, ribosome biogenesis, and nuclear DNA synthesis and repair among others [reviewed in (Ye and Rouault,2010; Lill et al.,2012)]. The core machinery that catalyzes the initial step in ISC assembly in the mitochondrial matrix consists of cysteine desulfurase (NFS1), a cysteine desulfurase that generates elemental sulfur; frataxin, an iron-binding protein that provides elemental iron and also stimulates NFS1 activity; and scaffold protein (ISCU), a scaffold protein upon which [2Fe–2S] and [4Fe–4S] clusters are initially assembled before being transferred to the appropriate enzymes [reviewed in (Ye and Rouault, 2010; Lill et al., 2012; Vaubel and Isaya, 2013)]. When this process functions normally, vital enzyme activities are maintained throughout the cell, iron-catalyzed oxidative damage is limited, and there is a balance between cellular iron uptake and mitochondrial iron utilization (**Figure 3**).

The consequences of defects in ISC synthesis have been extensively studied in a number of *S. cerevisiae* mutants in which this process was genetically impaired [although only partially since a complete loss of ISC synthesis is not compatible with life across eukaryotes (Cossee et al., 2000; Lill and Kispal, 2000; Kispal et al., 2005)]. These mutants have consistently shown a series of key mitochondrial and cellular features. Within mitochondria, reduced formation of ISC leads to an increase in the fraction of labile iron that leads to higher rates of Fenton chemistry resulting in loss of mitochondrial DNA integrity and overall loss of oxidative phosphorylation (Knight et al., 1998; Li et al., 1999; Karthikeyan et al., 2003). Simultaneously, the cell responds to reduced mitochondrial ISC synthesis with a rapid increase in cellular iron uptake and intracellular re-distribution of iron that is depleted in the cytoplasm but continues to accumulate in mitochondria until it precipitates out of solution as an amorphous mineral (Babcock et al., 1997; Knight et al., 1998; Li et al., 1999; Chen et al., 2004). These features hold true in multicellular organisms including humans as we will see in the specific cases of FRDA and PD.

## **FRIEDREICH ATAXIA**

FRDA is an autosomal recessive disease and the most common genetically-determined ataxia that affects approximately 1:40,000 individuals in the Caucasian population [for recent reviews see


**Table 1 | Indicators of mitochondrial Fe–S cluster dysfunction in FRDA, PD, and aging.**

\*(Rotig et al., 1997; Koeppen, 2011); \*\*(Horowitz and Greenamyre, 2010); \*\*\*(Bota et al., 2002; Xu et al., 2010).

**mitochondrial ISC synthesis and their biological roles.** When mitochondrial ISC synthesis functions normally, vital enzyme activities are maintained throughout the cell, iron-catalyzed oxidative damage is limited, and there is a balance between iron uptake and iron utilization. See text for additional details. NFS1, cysteine desulfurase; ISD11, adaptor protein required for NFS1 stability; ISCU, scaffold protein; I, II, III, respiratory chain complexes I, II, and III; Aco, mitochondrial aconitase; mtDNA, mitochondrial DNA; nDNA, nuclear DNA.

(Pandolfo, 2009; Koeppen, 2011)]. Patients are healthy at birth and remain largely asymptomatic for the first 5–10 years of life but then begin to present progressive neurological impairment and additional problems including cardiac disease, muscle weakness, skeletal deformities, vision and hearing loss, and diabetes. Patients eventually become wheelchair-bound and most often die of cardiac failure in the 2nd or 3rd decade of life. Certain regions of the central (cerebellum and spinal cord) and peripheral (dorsal root ganglia and their nerves) nervous systems as well as the heart, skeletal muscles, skeleton, and endocrine pancreas are affected (Koeppen, 2011). This early-onset, very dramatic clinical and pathological progression results in most patients from reduced levels of a mitochondrial iron-binding protein called frataxin (Campuzano et al., 1996). The biochemical properties of frataxin include the ability to bind iron, the ability to donate iron to other iron-binding proteins, and the ability to oligomerize, store iron and control iron redox chemistry [reviewed in (Bencze et al., 2006)]. Through these properties, frataxin plays key roles in different iron-dependent pathways (primarily, although not exclusively, ISC synthesis) and is therefore critical for mitochondrial iron

metabolism and overall cellular iron homeostasis and antioxidant protection [reviewed in (Wilson, 2006; Vaubel and Isaya, 2013)].

## **MITOCHONDRIAL AND OTHER CELLULAR CONSEQUECES OF FRATAXIN DEFICIENCY**

Reduced levels of frataxin in FRDA mouse models and human patients result in defects in ISC enzymes even before mitochondrial iron accumulation becomes detectable (Puccio et al., 2001; Stehling et al., 2004). However, iron dysregulation is an early effect of frataxin depletion, which increases the fraction of labile redox-active iron inside mitochondria (Wong et al., 1999) leading to progressive accumulation of oxidative damage (Whitnall et al., 2012). In addition, in mouse heart the lack of frataxin results in gene expression changes leading to down-regulation of proteins involved in mitochondrial ISC synthesis, heme synthesis, and iron storage as well as proteins involved in cellular and mitochondrial iron uptake, which collectively lead to intracellular iron redistrbution and progressive mitochondrial iron accumulation (Huang et al., 2009). All of these effects ultimately result in progressive impairment of energy metabolism and accumulation of oxidative damage [reviewed in (Pandolfo, 2006; Wilson, 2006)]. There are also additional changes outside of mitochondria affecting pathways involved in antioxidant, metabolic, and inflammatory responses, which are believed to contribute to disease progression [(Pianese et al., 2002; Coppola et al., 2009; Lu et al., 2009; Paupe et al., 2009; Sparaco et al., 2009; Wagner et al., 2012); reviewed in (Pandolfo, 2012)].

## **PARKINSON DISEASE**

PD is a common neurodegenerative disease affecting ∼1:800 individuals between the 5th and 7th decade of life with a median survival from diagnosis of about 15 years. The degenerative process exquisitely affects dopaminergic neurons in the substantia nigra and clinically leads to progressive motor deficits including bradykinesia, rigidity, resting tremor, and postural instability as well as a variety of non-motor symptoms in the late stages of the disease. Age is the main risk factor for PD although susceptibility genes and environmental toxins are also implicated in the pathophysiology [reviewed in (Lees et al., 2009)].

#### **IRON DYSREGULATION CONTRIBUTES TO PD**

A significant body of data [reviewed in (Mastroberardino et al., 2009)] supports a role for a defect in ISC synthesis in PD: (i) mitochondrial impairment and oxidative stress are involved in the pathophysiology of PD (Greenamyre and Hastings, 2004); (ii) partial inhibition of Complex I of the respiratory chain (which includes several ISC-containing subunits) recapitulates many features of PD (Betarbet et al., 2000; Przedborski et al., 2001); (iii) iron levels are increased in the substantia nigra and within substantia nigra dopaminergic neurons of PD patients (Berg and Hochstrasser, 2006; Oakley et al., 2007); (iv) iron chelation protects substantia nigra neurons in animal models of PD (Ben-Shachar et al., 1991; Kaur et al., 2003) as well as PD patients (Devos et al., 2013). Mastroberardino et al. have described a novel pathway of iron transport to mitochondria of substantia nigra dopaminergic neurons involving transferrin and the transferrin receptor 2 (Mastroberardino et al., 2009). This pathway normally delivers transferrin-bound iron to mitochondria and to Complex I of the respiratory chain. In PD, however, there is an induction of transferrin receptor 2 expression and accumulation of oxidized transferrin inside mitochondria, which results in the release of labile ferrous iron from transferrin and the generation of hydroxyl radicals via Fenton chemistry (Mastroberardino et al., 2009; Horowitz and Greenamyre, 2010). A concomitant reduction in mitochondrial ISC synthesis has been postulated (Horowitz and Greenamyre, 2010), and could probably account for the cascade of events described above in at least two ways (**Figure 4**). Early on, a specific defect in the synthesis and/or delivery of ISC cofactors needed for Complex I activity could induce the transferring receptor 2 pathway, thereby triggering the increase in cellular iron uptake and the ensuing mitochondrial iron accumulation with iron-catalyzed oxidative damage. Under these conditions, the need to handle redox-active iron could likely make the ISC assembly machinery especially susceptible to oxidative damagefrom radicals generated via Fenton chemistry. Thus, progressive accumulation of oxidative damage could later lead to a more generalized ISC synthesis defect further enhancing cellular iron dysregulation in dopaminergic neurons (**Figure 4**).

## **IRON ACCUMULATION IN THE AGING BRAIN**

So far we have reviewed FRDA, where a defect in ISC synthesis is clearly associated with disease pathogenesis early on in the disease process, and PD, where a defect in ISC synthesis may be involved at both an early and a late stage of disease progression. Could a reduction in ISC synthesis also contribute to age-dependent iron accumulation in the brain? Iron accumulates in specific regions of the brain (primarily the globus pallidus and the substantia nigra) in an age-dependent manner, and iron-induced oxidative damage is implicated in age-dependent neuronal loss (Bartzokis et al., 2004; Zecca et al., 2004a,b; Xu et al., 2010). Moreover, loss of mitochondrial aconitase activity, an enzyme of the citric acid cycle that requires a [4Fe–4S] cluster for function and stability (Bulteau et al., 2003), is a marker of aging (Longo et al., 1999; Bota et al., 2002; Bulteau et al., 2006). Thus, it is possible to once again envision how the need to handle redox-active iron could subject the ISC assembly machinery to progressive accumulation of oxidative damage. This could lead to an age-dependent decline in ISC synthesis and loss of critical ISC enzyme activities, with progressive cellular and mitochondrial iron dysregulation, similar to what we have discussed above for FRDA and PD. **Figure 5** shows a possible unifying model whereby a continuum of clinical phenotypes

**FIGURE 4 | Proposed model for a role of ISC synthesis defects in PD progression.** See text for details. Tf, transferrin; TfR2, transferrin receptor; other abbreviations are as in the legend for **Figure 3**.

are associated with ISC synthesis defects of varying degree. We hypothesize that genetically-determined or acquired (environmental or age-dependent) defects in ISC synthesis contribute more prominently to human disease than it is currently appreciated. Indeed, age-dependent accumulation of iron and iron-catalyzed oxidative damage have been implicated not only in FRDA and PD but also Alzheimer's disease (Zecca et al., 2004b), atherosclerosis (Zacharski et al., 2000), and heart disease (Wood, 2004). As the aging population continues to expand, it is likely that the prevalence of conditions associated with age-dependent iron accumulation will also increase.

#### **IRON CHELATION FOR TREATMENTS OF FRDA AND PD**

As we have seen, eukaryotic cells require mitochondrial ISC synthesis not only to ensure the biogenesis and function of a large number of ISC-dependent enzymes but also to maintain a balance between iron uptake and iron utilization (**Figure 3**).

While there are no approaches as yet available to correct defects in ISC synthesis directly, targeting chelatable iron represents a reasonable approach to limit the consequences of these defects, primarily mitochondrial iron overload and iron-catalyzed Fenton chemistry. Iron chelation is widely used to reduce iron deposition in the organs of patients affected by disorders characterized by global iron overload (e.g., hemochromatosis; Barton, 2007; Flaten et al., 2012). However, in the case of FRDA or PD, the regional nature of the iron accumulation and the abnormal iron distribution within the affected cell types have suggested the need for agents able to appropriately "relocate" iron while at the same time limiting its participation in radical-generating reactions (Boddaert et al., 2007). Deferiprone, a chelator used for treating iron overload, has been shown to possess this iron "relocating" ability by scavenging labile iron from mitochondria and delivering it to cytoplasmic and extracellular apotransferrin. In cells derived from FRDA patients, deferiprone decreased the levels of mitochondrial labile iron and also limited oxidative damage (Kakhlon et al. 2008, 2010), presumably by limiting iron-catalyzed Fenton chemistry (Kontoghiorghes, 2009). A six-month open-label single-arm study with deferiprone, administered together with the antioxidant idebenone, was initially conducted in nine adolescents affected by FRDA. By use of magnetic resonance imaging, selective iron removal was observed in the nucleus dentatus of the cerebellum in all patients, while improved neuromotor function was observed in the youngest patients (Boddaert et al., 2007). An 11 month open-labeled study in 20 FRDA patients confirmed that deferiprone and idebenone combined could improve iron deposits in the dentate nucleus with a stabilizing effect on certain neurological parameters (in addition to significantly improved heart hypertrophy parameters; Velasco-Sanchez et al., 2011). However, a subsequent doubleblind, randomized placebo-controlled phase 2 trial showed a worsening of ataxia with doses of deferiprone ≥40 mg/Kg/day. In the same study, there were no significant changes in ataxia with lower doses of deferiprone, though improvements in posture, gait, and kinetic function were noted in some patients, and a significant decrease in left ventricular mass in most patients [reviewed in (Wilson, 2012; Pandolfo and Hausmann, 2013)]. These dose-dependent effects may be explained by a report that concentrations of deferiprone ≥50 μM in cultured cells induced iron depletion rather than redistribution, with deleterious effects on ISC enzyme activities and even on frataxin levels (Pandolfo and Hausmann, 2013). A moderate iron chelation regimen that did not alter systemic iron levels was recently tested in a pilot clinical study of deferiprone in early-stage PD patients; this regimen was shown to improve iron deposits in the substantia nigra as well as motor parameters of disease progression (Devos et al., 2013). These studies underscore the importance of identifying iron chelation modalities that can correct iron distribution within affected neuronal cells without inducing deleterious global changes in iron metabolism, including changes in the expression of frataxin and other mitochondrial proteins involved in ISC synthesis. Better understanding of disease natural history and ability to intervene in the presymptomatic phase may also be critical to achieve effective treatments.

### **CONCLUSION AND FUTURE DIRECTIONS**

The best characterized human condition linked to abnormal mitochondrial ISC synthesis is FRDA. Tissue-specific defects in mitochondrial ISC synthesis have more recently been identified in patients with isolated myopathy (Mochel et al., 2008; Olsson et al., 2008) or sideroblastic anemia (Camaschella et al., 2007), which are linked to mutations in the scaffold ISCU and the enzyme glutaredoxin 5, respectively. The combination of genetic and clinical heterogeneity (as illustrated by these three disorders) can make inherited defects in ISC synthesis difficult to recognize, and it is likely that their combined prevalence is underestimated. It is also possible that tissue-specific decline in ISC synthesis is implicated in the iron accumulation that occurs not only in PD and the aging brain (Zecca et al., 2004b) but also the vasculature (Zacharski et al., 2000) and the heart (Wood, 2004) during aging. Another possible link to aging is suggested by a recent report that impaired ISC biogenesis leads to nuclear genome instability (Veatch et al., 2009). These interesting links remain largely unexplored due to the lack of suitable technology. The activities of natural (e.g., mitochondrial aconitase or succinate dehydrogenase; Foury, 1999; Mochel et al., 2008) or artificial (Hoff et al., 2009a,b) ISC-containing enzymes are currently used as measures of ISC synthesis. One limitation is that these activities lay downstream of the initial and rate-limiting step in ISC synthesis (i.e., the assembly of [2Fe–2S] clusters on ISC scaffolds). In addition, they can be influenced by factors (e.g., oxidative stress) independent of the actual rate of ISC synthesis. In essence, technology to interrogate ISC synthesis directly *in vivo*, for translational and clinical research studies, is currently lacking. Development of such technology will open the possibility to (i) assess the extent to which this pathway contributes to neurodegenerative disease and age-related disorders, and (ii) develop strategies to correct primary or secondary functional deficits. Given the role played by ISC synthesis in the maintenance of many vital enzymes and iron homeostasis, as well as the prevention of ironcatalyzed oxidative stress, we predict that this technology will have important biomedical applications.

#### **ACKNOWLEDGMENT**

Grazia Isaya is supported by a grant from the National Institutes of Health/National Institute on Aging (AG15709).

#### **REFERENCES**


Betarbet, R., Sherer, T. B., MacKenzie, G., Garcia-Osuna, M., Panov, A. V., Greenamyre, J. T., et al. (2000). Chronic systemic pesticide exposure reproduces features of Parkinson's disease. *Nat. Neurosci.* 3, 1301–1306. doi: 10.1038/81834

Boddaert, N., Le Quan Sang, K. H., Rotig, A., Leroy-Willig, A., Gallet, S., Brunelle, F., et al. (2007). Selective iron chelation in Friedreich ataxia: biologic and clinical implications. *Blood* 110, 401–408. doi: 10.1182/blood-2006-12-065433


feedback inhibition of the SIRT3 deacetylase. *Hum. Mol. Genet*. 21, 2688–2697. doi: 10.1093/hmg/dds095


**Conflict of Interest Statement:** Mayo Clinic has a financial interest associated with technology used in the author's research, which has been licensed to a commercial entity. Mayo Clinic, but not the author, has received royalties of less than the federal threshold for significant financial interest.

*Received: 15 January 2014; paper pending published: 13 February 2014; accepted: 15 February 2014; published online: 03 March 2014.*

*Citation: Isaya G (2014) Mitochondrial iron–sulfur cluster dysfunction in neurodegenerative disease. Front. Pharmacol. 5:29. doi: 10.3389/fphar.2014.00029*

*This article was submitted to Drug Metabolism and Transport, a section of the journal Frontiers in Pharmacology.*

*Copyright © 2014 Isaya. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*

## Dysregulation of cellular iron metabolism in Friedreich ataxia: from primary iron-sulfur cluster deficit to mitochondrial iron accumulation

## *Alain Martelli 1,2,3,4,5\* and Hélène Puccio1,2,3,4,5\**

*<sup>1</sup> Department of Translational Medecine and Neurogenetics, Institut de Génétique et de Biologie Moléculaire et Cellulaire, Illkirch, France*

*<sup>2</sup> INSERM, U596, Illkirch, France*

*<sup>3</sup> CNRS, UMR7104, Illkirch, France*

*<sup>4</sup> Université de Strasbourg, Strasbourg, France*

*<sup>5</sup> Chaire de Génétique Humaine, Collège de France, Illkirch, France*

#### *Edited by:*

*Paolo Arosio, University of Brescia, Italy*

#### *Reviewed by:*

*Stanislav Yanev, Bulgarian Academy of Sciences, Bulgaria Andrei Adrian Tica, University of Medicine Craiova Romania, Romania*

#### *\*Correspondence:*

*Alain Martelli and Hélène Puccio, Department of Translational Medecine and Neurogenetics, Institut de Génétique et de Biologie Moléculaire et Cellulaire, BP 10142, 67404 Illkirch, France e-mail: martelli@igbmc.fr; hpuccio@igbmc.fr*

Friedreich ataxia (FRDA) is the most common recessive ataxia in the Caucasian population and is characterized by a mixed spinocerebellar and sensory ataxia frequently associating cardiomyopathy. The disease results from decreased expression of the FXN gene coding for the mitochondrial protein frataxin. Early histological and biochemical study of the pathophysiology in patient's samples revealed that dysregulation of iron metabolism is a key feature of the disease, mainly characterized by mitochondrial iron accumulation and by decreased activity of iron-sulfur cluster enzymes. In the recent past years, considerable progress in understanding the function of frataxin has been provided through cellular and biochemical approaches, pointing to the primary role of frataxin in iron-sulfur cluster biogenesis. However, why and how the impact of frataxin deficiency on this essential biosynthetic pathway leads to mitochondrial iron accumulation is still poorly understood. Herein, we review data on both the primary function of frataxin and the nature of the iron metabolism dysregulation in FRDA. To date, the pathophysiological implication of the mitochondrial iron overload in FRDA remains to be clarified.

**Keywords: Friedreich ataxia, frataxin, iron metabolism, iron-sulfur cluster, mitochondria, iron metabolism disorders**

## **INTRODUCTION**

Ataxias are a heterogeneous group of disorders characterized by loss of coordination due to the degeneration of the neuronal networks closely linked to cerebellar function. Friedreich's ataxia (FRDA) is the most prevalent form of hereditary ataxia in Caucasians, accounting for 75% of ataxia with onset prior to 25 years of age (Cossee et al., 1997). FRDA is characterized by progressive spinocerebellar and sensory ataxia (Harding, 1981). The symptoms associated with the disease include the absence of deep tendon reflexes, dysarthria, pyramidal signs, muscular weakness, and positive extensor plantar response (Harding, 1981; Pandolfo, 2009). The neurological symptoms result from progressive degeneration of large sensory neurons in the dorsal root ganglia (DRG) and their axonal projection in the posterior columns, as well as from degeneration of the spinocerebellar and corticospinal tracts of the spinal cord (Koeppen and Mazurkiewicz, 2013). The dentate nucleus of the cerebellum is also affected and accounts for the cerebellar phenotype (Koeppen, 2011). FRDA is also characterized by primary non-neurological manifestations, in particular hypertrophic cardiomyopathy and increased incidence of diabetes (Harding and Hewer, 1983). The cardiomyopathy associated with FRDA is due to the natural transition from hypertrophy to dilation. The latter promotes cardiomyocytes death and replacement of contractile cells by fibrotic tissue leading to severe systolic and diastolic dysfunction (Tsou et al., 2011; Payne and Wagner, 2012; Weidemann et al., 2012). Lethal congestive heart failure and supraventrivular arrhythmias is the primary mode of death in ∼60% of patients with FRDA (Harding, 1981; Tsou et al., 2011; Weidemann et al., 2012).

The mutated gene in FRDA is localized on the long arm of chromosome 9 (9q21.11) and codes for a small mitochondrial protein called frataxin (FXN) (Campuzano et al., 1996, 1997; Koutnikova et al., 1997). All FRDA patients carry at least one allele with an expansion of a GAA-triplet repeat in the first intron of the FXN gene. Most patients are homozygous for this mutation, but a few patients (4%) are compound heterozygous for the GAA expansion and a classical mutation (nonsense, missense, deletions, insertions) leading to loss of FXN function (Campuzano et al., 1996; Cossee et al., 1999; Gellera et al., 2007). Normal chromosomes contain up to 40 GAA repeats, whereas diseaseassociated alleles contain 100 to more than 1500 GAA repeats, most commonly ∼600–900. This GAA expansion leads to transcriptional silencing of FXN through a mechanism involving modifications of the chromatin structure of the locus, resulting in expression of a structurally and functionally normal frataxin but at levels that are estimated at ∼5–30% of normal (reviewed in Gottesfeld, 2007; Schmucker and Puccio, 2010). As demonstrated in knockout animals, complete absence of frataxin leads to early embryonic death (Cossee et al., 2000). The rare non-GAA mutations in FXN that have been associated with FRDA lead to production of non-functional or partially functional proteins (Correia et al., 2008). In most cases, compound heterozygous patients are clinically indistinguishable from patients that are homozygous for the GAA expansions, but a few missense mutations (e.g., G130V, D122Y, R165P, L106S) in compound heterozygous patients cause atypical or milder clinical presentations (Cossee et al., 1999; Gellera et al., 2007).

The genetic basis of FRDA in humans raises challenges for modeling the disease in other species. Despite the difficulty in generating perfect FRDA models, a multitude of complementary models have been generated enabling significant advances in understanding the function of frataxin, the pathophysiology of the disease and some of the mechanisms implicated in GAAbased silencing (reviewed in Martelli et al., 2012b; Perdomini et al., 2013). Due to its high evolutionary conservation, the effect of FXN depletion has been modeled in diverse organisms, including yeast (Babcock et al., 1997; Foury and Cazzalini, 1997), invertebrates such as *C. elegans* (Vazquez-Manrique et al., 2006; Ventura et al., 2006; Zarse et al., 2007) and Drosophila (Anderson et al., 2005; Llorens et al., 2007), and in mice (Puccio et al., 2001; Miranda et al., 2002; Simon et al., 2004; Al-Mahdawi et al., 2006; Martelli et al., 2012a). However, due to the complexity of the clinical phenotype of individuals with FRDA and the species specificity in regulation of certain fundamental pathways, in particular iron metabolism, mouse models or mammalian cell culture models are probably better suited to understand the pathophysiological mechanisms involved in the disease.

## **IRON DYSREGULATION IN FRIEDREICH ATAXIA**

Early characterization of the pathophysiology in individuals with FRDA provided evidence of a link between frataxin deficiency and cellular iron metabolism dysregulation. Indeed, Lamarche and colleagues were the first to report the presence of granular iron deposits in cardiomyocytes of FRDA patients (Lamarche et al., 1980). After the discovery of the disease-causing gene, the generation of the yeast strain deficient for the yeast frataxin homolog, Yfh1, (-Yfh1) showed that iron could accumulate in large amount within mitochondria (Babcock et al., 1997; Foury and Cazzalini, 1997). In mammals, mitochondrial iron accumulation and deposits were observed in the conditional mouse model reproducing the cardiac phenotype (MCK mouse) (Puccio et al., 2001). Iron metabolism dysregulation was also observed in heart autopsies of individuals with FRDA (Michael et al., 2006; Ramirez et al., 2012). Biochemical studies of heart biopsies also demonstrated a deficit in mitochondrial iron-sulfur (Fe-S) cluster-containing enzymes (aconitase and respiratory chain complexes I-III) (Rotig et al., 1997). Finally, the presence of markers of oxidative damage in blood and urine samples was reported (Emond et al., 2000; Schulz et al., 2000; Bradley et al., 2004), although contradictory results from patient data have been reported (Di Prospero et al., 2007; Myers et al., 2008; Schulz et al., 2009). Altogether, these observations led to the early assumption of a pathophysiological implication of iron-dependent pathways in FRDA.

The presence of mitochondrial iron accumulation in FRDAaffected neurons is however less clear. Both dentate nucleus and dorsal root ganglions (DRGs) of individuals with FRDA have been studied to investigate iron dysregulation. Dentate nucleus is an iron-rich cerebellar structure that shows signs of neurodegeneration in patients with FRDA (Koeppen, 2011). Despite a report of difference in the MRI signals that suggests an overall increase of iron in the dentate nucleus of individuals with FRDA (Boddaert et al., 2007), no difference in iron concentrations was measured using autopsies (Koeppen et al., 2007). However, modification of the expression of iron-related proteins such as transferrin receptor 1 (TFR1), ferritins (FRTs) and ferroportin (FPN) were observed, thereby suggesting a change in iron metabolism (Koeppen et al., 2007). Further investigations using X-ray fluorescence (XRF) suggested that iron was relocating from dying neurons to microglia of dentate nucleus (Koeppen et al., 2012). Similarly, DRGs from individuals with FRDA do not show overall iron concentrations above normal (Koeppen et al., 2009). However, the expression of FRTs, the iron-storage proteins, increases as satellite cells surrounding affected DRG neurons proliferate (Koeppen et al., 2009, 2013), thus suggesting again a redistribution of iron from dying neurons to satellite cells.

Although these observations suggest that iron is released during neuronal degeneration and then stored by surrounding glial cells, they do not give any indication on the primary involvement of iron dysregulation in the neuropathophysiology. In particular, mitochondrial iron deposits have never been reported in neurons from FRDA individuals and were not observed in an inducible conditional mouse model reproducing the neuronal phenotype (Prp mice) (Simon et al., 2004).

To understand the role and pathophysiological implication of iron in the disease, it is therefore essential to understand the function of frataxin and how its impairment can lead to cellular iron dysregulation.

## **IRON AND FRATAXIN FUNCTION**

## **FRATAXIN AS AN IRON-BINDING PROTEIN**

Frataxin is a highly conserved protein present from gramnegative bacteria to eukaryotes, including yeast and mammals (**Figure 1A**). Frataxin is localized within the eukaryotic mitochondria and is ubiquitously expressed in mammals. The structure of frataxin is unique and conserved in between species: frataxin is a small globular acidic protein composed of a long Nterminal alpha helix and a short C-terminal alpha helix that both interact with a central beta-sheet structure (**Figure 1B**) (Musco et al., 2000).

Although bacterial (CyaY), yeast (Yfh1), and mammalian (FXN) frataxins all exist as soluble monomers, early *in vitro* studies of bacterial CyaY and yeast Yfh1 showed that the proteins are able to form oligomeric spheroidal structures in the presence of excess iron (Adamec et al., 2000; Gakh et al., 2002; Layer et al., 2006; Adinolfi et al., 2009). These oligomeric structures can capture up to 50–75 atoms of iron, in a similar way as ferritin. Due to its property in scavenging iron, oligomeric frataxin was initially proposed to act as ferritins by providing bio-available iron within mitochondria (Adamec et al., 2000; Cavadini et al., 2002). This hypothesis was further sustained by the capacity of mitochondrial ferritin to complement for frataxin deficiency in the -Yfh1 yeast strain and in HeLa cells (Campanella et al., 2004; Zanella et al., 2008). However, further experiments in yeast showed that

expression of human mitochondrial ferritin only partially rescues the -Yfh1 strain through a mechanism that does not overlap with frataxin function (Sutak et al., 2012). In addition, modulating the expression of Yfh1 in a yeast mutant strain that accumulates iron in mitochondria, in a similar way as the -Yfh1 strains, does not modify iron bio-availability (Seguin et al., 2010). The relevance of the *in vivo* function of oligomeric frataxin is also questioned by *in vitro* data showing that bacterial CyaY forms iron-rich oligomeric structures only under aerobic conditions and high ionic strengths (Adinolfi et al., 2002; Layer et al., 2006). Furthermore, yeast Yfh1 bearing a point mutation that prevents oligomerization can rescue the -Yfh1 strain (Aloria et al., 2004), therefore indicating that oligomerization is not required to fulfill the main function of frataxin *in vivo*.

In higher eukaryotes, the oligomerization process does not appear to be fully conserved. Frataxin is encoded by a nuclear gene and synthesized as a precursor protein (FXN1−210) that is then matured in two steps within the mitochondrial matrix to give an intermediate form (FXN42−210) and the major mature form (FXN81−210) (Condo et al., 2007; Schmucker et al., 2008) (**Figure 1A**). Only the precursor and intermediate forms of FXN can form oligomers in an iron-independent way, whereas mature human FXN is not prone to oligomerization (O'neill et al., 2005; Prischi et al., 2009). Furthermore, *in vivo* experiments using mouse fibroblasts deleted for the endogenous murine frataxin showed that the expression of the mature human FXN81−<sup>210</sup> is sufficient to promote cell survival (Schmucker et al., 2011), thus indicating that oligomerization is not a process required for the primary and essential function of mammalian frataxin *in vivo*.

Although the functional relevance of an iron-rich oligomeric frataxin is questionable, there is clear evidence that monomeric frataxin can also bind iron *in vitro*. Several iron-binding sites have been characterized depending on the oxidative state of iron (Fe2<sup>+</sup> or Fe3+) and the origin of the frataxin proteins (CyaY, Yfh1, or human FXN) (Yoon and Cowan, 2003; Bou-Abdallah et al., 2004; Cook et al., 2006; Yoon et al., 2007; Huang et al., 2008). A primary iron-binding site appears however to be conserved and involves residues of the acidic ridge localized within the first alpha helix of frataxin (**Figure 1**). The site binds Fe2<sup>+</sup> with a dissociation constant (Kd) within the micromolar range (3–55μM) (Yoon and Cowan, 2003; Nair et al., 2004; Cook et al., 2006) but seems to be poorly specific as other cations were shown to also bind CyaY (Pastore et al., 2007).

### **FUNCTION OF FRATAXIN IN Fe-S CLUSTER BIOGENESIS**

The capacity of frataxin to bind iron and the evidence of an iron metabolism dysregulation in individuals with FRDA and in -Yfh1 yeast strains led to the assumption that frataxin plays a key role in the mitochondrial iron metabolism. Further biochemical and interaction studies provided several hypotheses. Interactions with mitochondrial aconitase, ferrochelatase and proteins of the mitochondrial Fe-S cluster machinery were reported (Gerber et al., 2003; Bulteau et al., 2004; Yoon and Cowan, 2004; Bencze et al., 2007), and the hypothesis of frataxin being an iron provider to various iron-dependent mitochondrial pathways was brought forward. However, interactions with aconitase and ferrochelatase are still poorly characterized and were reported not to be reproducible (Schmucker et al., 2011). To date, only the interaction of frataxin with proteins involved in the mitochondrial Fe-S biogenesis have been extensively and convincingly characterized.

Fe-S clusters are inorganic redox-active protein cofactors that are present in almost all living organisms. They play cardinal roles in various functions throughout the cell, including electron transport in the respiratory complexes and DNA repair or metabolism. Although Fe-S clusters can adopt different configurations, [Fe2S2] and [Fe4S4] clusters are the most frequent Fe-S clusters in eukaryotes. *De novo* biosynthesis of Fe-S clusters occurs within mitochondria (reviewed in Lill, 2009; Beilschmidt and Puccio, 2014). The first step involves the assembly of a Fe-S cluster on a scaffold protein ISCU (Isu in yeast) from inorganic iron and sulfur. A cysteine desulfurase complex NFS1/ISD11 provides the sulfur through a persulfide intermediate. ISCU and NFS1/ISD11 interact and form a ternary ISCU/NFS1/ISD11 complex with a most likely α2β2γ4 stoichiometry (Schmucker et al., 2011; Colin et al., 2013). Once the cluster is assembled on ISCU, it is transferred to acceptor proteins with the help of additional components of the mitochondrial Fe-S cluster machinery, such as the HSCB/HSPA9 chaperone system or proteins (e.g., ISCA1/2) that provide Fe-S cluster to a subset of mitochondrial proteins (**Figure 2**). Alternatively, a still uncharacterized intermediate provided by the early Fe-S cluster machinery is exported from the mitochondria to the cytosol via the ABCB7 transporter where it is used by the cytosolic Fe-S cluster assembly machinery (CIA machinery) to generate Fe-S clusters for cytosolic and nuclear acceptor proteins (reviewed in Lill, 2009) (**Figure 2**).

The first hints for the primary involvement of frataxin in Fe-S cluster biogenesis came from the characterization of the cardiac mouse model mimicking the FRDA cardiomyopathy, as Fe-S cluster dependent enzymes were affected prior to the appearance of the heart dysfunction and the mitochondrial iron accumulation (Puccio et al., 2001; Seznec et al., 2004; Martelli et al., 2007) (**Figure 3A**). In parallel, early phylogenetic studies predicted a role of frataxin in Fe-S cluster metabolism (Huynen et al., 2001). The implication of frataxin in Fe-S cluster biogenesis was later confirmed in yeast depleted for Yfh1 (Duby et al., 2002; Muhlenhoff et al., 2002). Furthermore, an iron-dependent interaction of Yfh1 with Nfs1 and Isu1 was reported in yeast (Gerber et al., 2003), while *in vitro* reconstitution experiments

**FIGURE 2 | Schematic view of the Fe-S cluster machinery and the IRP-mediated cellular iron regulation.** *De novo* Fe-S cluster biogenesis occurs within mitochondria and involves assembly of inorganic sulfur and iron on a scaffold protein ISCU. Iron is imported into the mitochondria by mitoferrins (MFRN). The process of Fe-S cluster assembly occurs within a complex consisting of NFS1-ISD11, the cysteine desulfurase providing the sulfur, ISCU and eventually frataxin (FXN), which regulates the NFS1 activity and the entry of iron within the complex (Colin et al., 2013). The process also needs electrons (e–) that may be provided by a mitochondrial ferredoxin (FDX2). Once assembled the cluster on ISCU is transferred to acceptor proteins with the help of additional proteins, such as the chaperones HSCB and HSPA9, ISCA proteins, IBA57, NFU1, BOLA3, and NUBPL (reviewed in Lill, 2009; Beilschmidt and Puccio, 2014). Alternatively, a still uncharacterized compound (X) provided by the mitochondrial machinery is exported to the cytosol via ABCB7 and is

used by the cytosolic Fe-S cluster assembly machinery (CIA machinery) to assemble Fe-S clusters for cytosolic and nuclear acceptor proteins. Among the cytosolic Fe-S cluster acceptors, IRP1 is a regulator of cellular iron metabolism. In normal conditions, IRP1 binds a Fe-S cluster to become an aconitase devoid of regulatory capacity. IRP2 exists only as an apoprotein and is regulated through proteasomal degradation mediated by the iron sensor protein FBXL5 (Salahudeen et al., 2009; Vashisht et al., 2009). Both IRPs can regulate the expression of key genes involved in iron metabolism, such as transferrin receptor 1 (TFR1), ferritins (FRTs), and the iron exporter ferroportin (FPN), by binding a specific mRNA motif called IRE. Depending on the location of the IRE compared to the open reading frame (ORF), IRPs can increase (+) or decrease (−) protein expression, thus controlling cellular iron import and storage (reviewed in Anderson et al., 2012). DMT1: divalent metal transporter involved in iron import.

**FIGURE 3 | MCK mouse model. (A)** Phenotypic characteristics of the MCK mouse model: MCK mice develop progressive hypertrophic cardiomyopathy characterized by progressive left ventricule (LV) dysfunction and hypertrophy starting around 5 weeks of age. The cardiomyopathy leads to cell death and fibrosis. MCK mice prematurely dye around 11 weeks. Fe-S cluster deficit is a primary feature in the mouse pathology with significant differences observed in 4 weeks old mice. Mitochondrial iron (mit-Fe) accumulation is observed in the later stage of the disease. **(B)** Electron microscopy picture obtained from a heart sample of a 7 weeks old MCK mouse showing mitochondrial abnormalities, in particular collapse cristae and electron-dense deposits (arrows) corresponding to mitochondrial iron deposits. mf, myofiber. **(C)** Semi-nested PCR on cDNA from heart samples of 8 weeks old control and deleted MCK mice was performed as described (Santambrogio et al., 2007) to assess mitochondrial ferritin (FTMT) expression. Total RNA from tissue was extracted using Trizol® reagent (Life Technologies) and submitted (+) or not (−) to reverse transcription (RT). Testis cDNA was used as positive control for *Ftmt* expression, and a classical PCR to amplify mitochondrial aconitase (Aco2) cDNA was carried out as a control for reverse transcription and loading. Samples without reverse transcription (RT–) were used as control for specific PCR amplification.

showed that human FXN could transfer iron to ISCU (Yoon and Cowan, 2003) and that bacterial CyaY could provide iron for Fe-S cluster formation (Layer et al., 2006). Altogether, these results suggested that frataxin might be the iron donor for the assembly of the Fe-S cluster *in vivo*. However, data from mammals, yeast, and bacteria were quite controversial as to the direct frataxin protein partner in Fe-S biogenesis (Gerber et al., 2003; Layer et al., 2006; Shan et al., 2007; Li et al., 2009). Recently, these results were reconciled by independent work using mammalian recombinant proteins showing that frataxin interacts with a preformed complex composed of NFS1, ISCU, and ISD11 (Tsai and Barondeau, 2010; Schmucker et al., 2011). A similar complex was also reported in bacteria (Prischi et al., 2010). In line with results obtained with the bacterial CyaY suggesting that frataxin is a regulator of the Fe-S clusters synthesis (Adinolfi et al., 2009), the binding of frataxin in the mammalian system was shown to stabilize the complex and to activate the cysteine desulfurase activity (Tsai and Barondeau, 2010; Colin et al., 2013). Moreover, although the formation of the complex was shown to be iron-independent (Schmucker et al., 2011), frataxin appears to concomitantly activate the cysteine desulfurase activity and to control iron entry within the complex (Colin et al., 2013). More recently, the biochemical characterization of the successive steps of the cysteine desulfurase activity in yeast provided evidence that frataxin triggers a conformational change that modifies the substrate-binding site of the enzyme (Pandey et al., 2013). All these results therefore indicate that frataxin, by controlling both iron entry and sulfide production, is essential in the process of Fe-S cluster assembly during the initial stage of the biogenesis. However, how frataxin controls iron entry within the complex still need to be determined. In particular, residues that define the primary iron-binding site of the protein *in vitro* were also shown to be involved in the interaction with the cysteine desulfurase and/or the ISCU/NFS1/ISD11 complex (Prischi et al., 2010; Schmucker et al., 2011). Hence, the *in vivo* implication of this iron-binding site needs to be further investigated.

## **FROM IRON-SULFUR CLUSTER DEFICIT TO MITOCHONDRIAL IRON ACCUMULATION**

Although the recent advances point to a primary role of frataxin in Fe-S cluster biogenesis, the cellular mechanism that links frataxin deficiency to mitochondrial iron overload remains elusive. However, mitochondrial iron accumulation is not specific to frataxin deficiency, but rather appears as a general hallmark of primary Fe-S deficiency, as it has been observed in various yeast strains deleted for different genes involved in Fe-S cluster biogenesis (Kispal et al., 1997; Garland et al., 1999; Schilke et al., 1999; Lange et al., 2000; Voisine et al., 2001). Furthermore, mutations in human genes implicated in Fe-S cluster biogenesis have recently been identified as disease-causing genes (reviewed in Beilschmidt and Puccio, 2014), and some of the associated disorders are also characterized by mitochondrial iron accumulation. Mutations in the scaffold protein ISCU lead to myopathy with lactic acidosis with different severity, also known as Swedish myopathy (Mochel et al., 2008; Olsson et al., 2008; Kollberg et al., 2009). The major mutation, due to a founder effect in Sweden, induces a muscle-specific cryptic splice site that leads to a truncated protein (Mochel et al., 2008; Olsson et al., 2008). In muscle biopsies, iron labeling (Perl's staining) showed accumulation of iron within mitochondria (Mochel et al., 2008; Kollberg et al., 2009). Mutations in ABCB7 are associated with X-linked sideroblastic anemia with ataxia, a condition that is characterized by the presence of iron-rich perinuclear mitochondria within erythroblasts (sideroblasts) (Allikmets et al., 1999; Bekri et al., 2000). Similarly, a mutation in GLRX5, a protein linked to Fe-S cluster biogenesis, although its function is still unclear (Rodriguez-Manzaneque et al., 2002), was identified in a patient presenting sideroblastic anemia (Camaschella et al., 2007). Interestingly, mutations linked to human disease in proteins involved in the delivery of Fe-S cluster to a subset of mitochondrial proteins (e.g., NFU1, IBA57, BOLA3—see **Figure 2**) are not associated with iron accumulation (reviewed in Beilschmidt and Puccio, 2014).

It is most likely that the pathways leading to iron dysregulation and mitochondrial iron accumulation are shared among the different disease linked to primary Fe-S cluster biogenesis. Although the link between Fe-S cluster deficit and iron metabolism has been observed in both yeast and mammals, it is unlikely that the mechanisms involved in mitochondrial iron accumulation are strictly conserved between the two species, as cellular iron homeostasis involves different modes of regulation in yeast and higher eukaryotes.

## **CELLULAR IRON REGULATION IN MAMMALS**

In mammals, the Iron Regulatory Proteins (IRP) 1 and 2 largely regulates cellular iron homeostasis. IRP1 and IRP2 are cytosolic translational regulators that control the expression of proteins involved in iron handling and distribution (**Figure 2**), as well as targeting other transcripts that are not directly involved in iron metabolism such as HIF2α and mitochondrial aconitase (reviewed in Hentze et al., 2010; Anderson et al., 2012). IRP1 and 2 can bind specific mRNA motifs, called Iron Responsive Elements (IRE), thereby influencing protein expression by regulating either protein translation or mRNA metabolism. Indeed, when the IRE is located in the 5 UTR, binding of IRPs blocks translation, whereas the formation of an IRP/IRE complex in the 3 UTR leads to an increase half-life of the mRNA, therefore increasing translation. Transferrin receptor 1 (TFR1), implicated in cellular iron import, and the ferritins (FRTs), involved in cytosolic iron storage, are key proteins regulated by IRPs (**Figure 2**). The mRNA of TFR1 contains several IRE motifs within the 3 UTR, whereas mRNAs coding for FRTs contain an IRE motif in the 5 UTR. Although IRP1 and IRP2 exhibit some functional redundancy as both proteins can control TFR1 and FRTs expressions, some IRE sequence specificities have been reported (Ke et al., 1998; Anderson et al., 2013). However, the activity of IRP1 and IRP2 are mostly regulated differently. When cytosolic iron concentration increases, the iron-binding protein FBXL5 targets IRP2 to ubiquitination and proteasomal degradation (**Figure 2**) (Salahudeen et al., 2009; Vashisht et al., 2009). However, although IRP1 can also be targeted by FBXL5, its IRE-binding activity is mainly negatively regulated through the insertion of a [Fe4-S4] cluster leading to a protein with cytosolic aconitase activity devoid of IRE-binding activity (**Figure 2**) (Haile et al., 1992a,b).

IRP1 and IRP2 have been shown to have overlapping functions as observed in knockout mouse models (Galy et al., 2005; Anderson et al., 2013; Ghosh et al., 2013), however, IRP2 is considered as the main iron regulator under normal physiological conditions as IRP1 exists mainly as an aconitase (Meyron-Holtz et al., 2004; Moroishi et al., 2011) (**Figure 2**).

## **THE NATURE OF THE MITOCHONDRIAL ACCUMULATED IRON**

To date, no data provide a clear answer on the nature of the iron that is accumulated in affected mitochondria, but few interesting hints are available. The analysis of heart tissues from MCK mice by electron microscopy showed the presence of electrondense particles within the mitochondrial matrix (**Figure 3B**) that correlated with iron accumulation (Puccio et al., 2001). Similar structures were observed in mitochondria of heart tissue from individuals with FRDA (Michael et al., 2006) and in the liver conditional knockout mouse (ALB mouse) (Martelli et al., 2012a). In FRDA patients' samples, histological analysis suggested that mitochondrial ferritin (FTMT) might be involved in the formation of the iron-rich structures (Michael et al., 2006). However, recent data obtained in MCK mice suggested that FTMT is not involved since iron was reported to be mostly present as mineral nonferritin aggregates (Whitnall et al., 2012). In addition, despite a similar pattern of iron deposits as in patients (**Figure 3B**), no Ftmt mRNA could be detected in both the heart of MCK mice (**Figure 3C**) or the liver of ALB mice (AM and HP, unpublished results) using a semi-nested PCR protocol developed to specifically assess Ftmt expression (Santambrogio et al., 2007). These results further question the potential role of FTMT in the molecular pathophysiology.

Interestingly, the iron-rich aggregates observed in mouse models and patient samples are reminiscent of the mitochondrial phosphate-iron nano-particles that were identified in -Yfh1 yeasts (Lesuisse et al., 2003), as well as in the yeast strains deleted for Yah1 and Atm1 (the homologs of ferredoxin or ABCB7, respectively) (Miao et al., 2008, 2009). The formation of these aggregates in yeast mitochondria lead to a decrease of available iron that affects heme biosynthesis (Lesuisse et al., 2003; Seguin et al., 2010).

## **MODIFICATIONS OF IRON-RELATED PROTEIN AND GENE EXPRESSION IN THE CARDIAC MOUSE MODEL**

Although the characterization of the cardiac MCK mouse model and cellular models deficient in frataxin have provided several clues on the nature of iron dysregulation occurring after frataxin deficiency (Seznec et al., 2005; Whitnall et al., 2008, 2012; Huang et al., 2009), available data are sometimes contradictory.

The variations in activity, protein and mRNA expressions of key genes implicated in iron regulation and distribution that have been reported in MCK mice are shown in **Table 1**. IRP1 was shown to be activated into its IRE-binding form in MCK mice (Seznec et al., 2005), in agreement with the primary role of frataxin in Fe-S cluster biogenesis. This was also observed in frataxin knockdown experiments using HeLa cells (Stehling et al., 2004). Accordingly, similar observations were made in knockdown experiments or knockout animals targeting other proteins of the Fe-S cluster machinery (Biederbick et al., 2006; Fosset et al., 2006; Pondarre et al., 2006; Wang et al., 2007). Furthermore, ferritin L (FRTL) mRNA and protein levels were shown to progressively increase in MCK mice (Seznec et al., 2005). More recently, an increase of transferrin receptor 1 (TFR1) protein expression concomitant to a decrease of expression of ferroportin (FPN), the cellular iron exporter, was reported in FXN-deficient mice (Whitnall et al., 2008; Huang et al., 2009), thus suggesting an overall increase of the iron import capacity. This modification of iron import was confirmed by 59Fe import experiments (Whitnall et al., 2008). However, in contradiction with the reported increase of FRTL levels (Seznec et al., 2005), both ferritin L and H displayed decreased protein levels when compared to control animals, and most strikingly, no difference in IRP1 IRE-binding activity could be observed (Whitnall et al., 2008, 2012). As iron metabolism is a tightly regulated pathway, caution in comparing animal data raised in different laboratories have to be taken [differential animal food, circadian rhythm, and experimental condition before sacrifice (diet intake)]. Notably, in the latter reports, the mobility shift assays show that IRP1 is almost fully activated into its IRE-binding

#### **Table 1 | Modifications in iron-related gene expression and activity in MCK mice.**


*"UP" indicates increase of expression or activity in Fxn-deleted mice compared to control mice, "DOWN" means decrease and "Similar," no detectable change. Ages (in weeks) corresponding to the observations are indicated in brackets.*

#*IRP1 almost fully activated in both control and deleted MCK mice.*

*\*Mitochip array data; n.a: data not available.*

£*No difference was observed at 4 weeks.*

form in both control and deleted mice (Whitnall et al., 2008, 2012), in contradiction with IRP1 being essentially an aconitase in normal physiological conditions. Furthermore, the authors also provided evidence that IRP2 is more active in FXN-deleted animals (Whitnall et al., 2008, 2012), thus indicating a depletion of iron within the cytosol leading to reduced proteasomal degradation of IRP2. Cytosolic iron depletion was confirmed by iron concentration measurements after tissue fractionation (Whitnall et al., 2008), although these measurements do not discriminate between the available cytosolic iron pool and the one trapped within ferritins, which were shown under the same experimental conditions to be decreased (Whitnall et al., 2008). More interestingly, the mRNA level of mitoferrin-2 (MFRN2), the mitochondrial iron transporter, is significantly increased in MCK mice (Huang et al., 2009). A similar increase of MFRN2 mRNA has been observed in skeletal muscle biopsies from ISCU myopathy patients (Crooks et al., 2014). These results suggest the existence of a Fe-S cluster-dependent regulation of mitochondrial iron import, whether direct or indirect, that may control mitochondrial iron overload. However, whether up-regulation of MFRN2 is sufficient to explain mitochondrial iron accumulation in FXN-deficient mice is not known.

## **CONCLUSION AND PERSPECTIVES**

To understand the consequence of mitochondrial iron overload on the pathophysiology of FRDA is of particular interest in the context of therapeutic approaches for FRDA. Early report suggested that the iron accumulation generated toxic free radicals through Fenton reaction, therefore implying iron chelators as possible therapeutic agents. Although it is most likely that reactive oxygen species play a role in FRDA, the primary involvement as well as the importance of reactive oxygen species in the pathophysiology are still a matter of debate in the field.

Recently, Deferiprone, an iron chelator that may cross the blood brain barrier, has been used in preclinical and clinical studies for FRDA, but the results were somehow puzzling as different doses showed opposite effects, if any (reviewed in Pandolfo and Hausmann, 2013), therefore further questioning the rationale behind the use of chelation therapy in FRDA. In line with these results, the data obtained with the MCK mouse model further indicates that both cellular and mitochondrial iron imports are increased in the absence of frataxin. Does it mean that cells, and in particular mitochondria, are in iron deprivation rather than facing toxic iron accumulation? This question may seem counterintuitive when total mitochondrial iron is measured in FRDA models, but the characterization of the accumulated iron in frataxin-deficient cells provide further evidence that iron may not be biologically available within mitochondria. Hence, the role of chelation therapies should be to target this non-available iron to make it available again for biological processes, rather than depleting iron from the cell as it is expected in other disorders of iron overload.

The primary function of frataxin in Fe-S cluster biogenesis is now on the way of being fully elucidated through the biochemical characterization of the complex in which it is involved. Attention is also brought to the understanding of the cellular consequences of frataxin deficiency. In particular, the mechanism leading to mitochondrial iron accumulation, and most importantly, the consequences of this accumulation on the pathophysiology is under investigation. All together, these data will be valuable for the evaluation and design of new therapeutic approaches that may (or not) use iron chelators. The recent identification of several other human genetic disorders linked to primary Fe-S cluster deficiency and displaying mitochondrial iron accumulation, as well as the development of the corresponding cellular and animal models will clearly be an asset to address these questions.

## **ACKNOWLEDGMENTS**

We would like to thank all Friedreich Ataxia patients and their families, as well as patients' associations throughout the world for their role in supporting research. Our work on FRDA is and has been founded in the past years by the Friedreich Ataxia Research Alliance (FARA), l'Association Française de l'Ataxie de Friedreich (AFAF), la Fondation pour la Recherche Médicale (FRM), l'Association Française contre les Myopathies (AFM), l'Agence Nationale pour la Recherche (ANR) and by the European Community under the European Research Council [206634/ISCATAXIA] and the 7th Framework Program [242193/EFACTS].

## **REFERENCES**


clinical mutants. *FEBS J.* 275, 3680–3690. doi: 10.1111/j.1742-4658.2008. 06512.x


disease severity, and correlation with neurological symptoms. *Circulation* 125, 1626–1634. doi: 10.1161/CIRCULATIONAHA.111.059477

Whitnall, M., Suryo Rahmanto, Y., Huang, M. L., Saletta, F., Lok, H. C., Gutierrez, L., et al. (2012). Identification of nonferritin mitochondrial iron deposits in a mouse model of Friedreich ataxia. *Proc. Natl. Acad. Sci. U.S.A.* 109, 20590–20595. doi: 10.1073/pnas.1215349109

Whitnall, M., Suryo Rahmanto, Y., Sutak, R., Xu, X., Becker, E. M., Mikhael, M. R., et al. (2008). The MCK mouse heart model of Friedreich's ataxia: Alterations in iron-regulated proteins and cardiac hypertrophy are limited by iron chelation. *Proc. Natl. Acad. Sci. U.S.A.* 105, 9757–9762. doi: 10.1073/pnas.0804261105


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

*Received: 16 April 2014; accepted: 14 May 2014; published online: 03 June 2014. Citation: Martelli A and Puccio H (2014) Dysregulation of cellular iron metabolism in Friedreich ataxia: from primary iron-sulfur cluster deficit to mitochondrial iron accumulation. Front. Pharmacol. 5:130. doi: 10.3389/fphar.2014.00130*

*This article was submitted to Drug Metabolism and Transport, a section of the journal Frontiers in Pharmacology.*

*Copyright © 2014 Martelli and Puccio. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*

## HFE gene variants, iron, and lipids: a novel connection in Alzheimer's disease

## *Fatima Ali-Rahmani 1,2 , Cara-Lynne Schengrund2 and James R. Connor 1\**

<sup>1</sup> Departments of Neurosurgery, Neural and Behavioral Sciences and Pediatrics, Center for Aging and Neurodegenerative Diseases, Penn State Hershey Medical Center, Hershey, PA, USA

<sup>2</sup> Departments of Biochemistry and Molecular Biology, The Pennsylvania State University College of Medicine, Hershey, PA, USA

#### *Edited by:*

Paolo Arosio, University of Brescia, Italy

*Reviewed by:* Torben Moos, Aalborg University, Denmark James Duce, University of Leeds, UK

#### *\*Correspondence:*

James R. Connor, Departments of Neurosurgery, Neural and Behavioral Sciences and Pediatrics, Center for Aging and Neurodegenerative Diseases, Penn State Hershey Medical Center, Hershey, PA 17033, USA e-mail: jconnor@hmc.psu.edu

Iron accumulation and associated oxidative stress in the brain have been consistently found in several neurodegenerative diseases. Multiple genetic studies have been undertaken to try to identify a cause of neurodegenerative diseases but direct connections have been rare. In the iron field, variants in the HFE gene that give rise to a protein involved in cellular iron regulation, are associated with iron accumulation in multiple organs including the brain. There is also substantial epidemiological, genetic, and molecular evidence of disruption of cholesterol homeostasis in several neurodegenerative diseases, in particular Alzheimer's disease (AD). Despite the efforts that have been made to identify factors that can trigger the pathological events associated with neurodegenerative diseases they remain mostly unknown. Because molecular phenotypes such as oxidative stress, synaptic failure, neuronal loss, and cognitive decline, characteristics associated with AD, have been shown to result from disruption of a number of pathways, one can easily argue that the phenotype seen may not arise from a linear sequence of events. Therefore, a multi-targeted approach is needed to understand a complex disorder like AD. This can be achieved only when knowledge about interactions between the different pathways and the potential influence of environmental factors on them becomes available. Toward this end, this review discusses what is known about the roles and interactions of iron and cholesterol in neurodegenerative diseases. It highlights the effects of gene variants of HFE (H63D- and C282Y-HFE) on iron and cholesterol metabolism and how they may contribute to understanding the etiology of complex neurodegenerative diseases.

**Keywords: iron, cholesterol, sphingolipids, Alzheimer disease, HFE, H63D, brain**

#### **INTRODUCTION**

The average increase in life expectancy has been accompanied by an increase in the number of people with dementia, a problem expected to affect half of those living to be 85 or older. The most prevalent form of dementia is Alzheimer's disease (AD) and its incidence is expected to triple by 2050 (Hebert et al., 2013). The observations that more than 40% of AD patients carried the ApoE4 allele (Cedazo-Minguez and Cowburn, 2001) and that those carrying both the ApoE4 allele and expressing the H63D variant of the hemochromatosis protein HFE were prone to earlier onset of AD (Moalem et al., 2000; Percy et al., 2008) support the hypothesis that disruption of the normal metabolism of both iron and cholesterol contribute to AD. More specifically, the need for a discussion of the role of iron and cholesterol in neurodegenerative disease stems from our observation that disruption of normal iron metabolism in H63D-HFE-expressing human neuroblastoma cells resulted in altered cholesterol metabolism as well as our findings that mice expressing the orthologous H67D-HFE had alterations in brain iron and cholesterol metabolism and a reduction in brain volume that correlated with poorer recognition and spatial memory, symptoms associated with AD (Ali-Rahmani et al., 2014a).

#### **IRON IN NEURODEGENERATIVE DISEASES**

Numerous studies have implicated metals such as iron, copper, zinc, and aluminum in the pathogenesis of AD (Sayre et al., 2000; Connor et al., 2001; Bush, 2003; Maynard et al., 2005; Connor and Lee, 2006; Liu et al., 2006; Roberts et al., 2012). Evidence indicates that oxidative stress induced by excess iron contributes to neurodegeneration (Castellani et al., 2007). Because iron is the most abundant transition metal in the body and is readily available from several dietary options, it is important to understand how its regulation in the body is influenced by other factors and how its elevation or depletion affects cellular processes that could lead to AD pathogenesis. To understand the role of iron in neurodegeneration, it is necessary to understand its normal function and regulation.

#### **ROLE OF IRON IN BRAIN**

Iron is essential for a number of cellular processes needed for survival. It is a required cofactor for a number of enzymes involved in cell functions such as energy production (mitochondrial electron transport chain), DNA synthesis and repair, ribosome biogenesis, neurotransmitter synthesis, myelin synthesis and lipid metabolism, and cell cycle regulation. Iron is also

needed for heme production and formation of iron-sulfur clusters that are essential for electron transport and DNA repair (Rouault and Tong, 2005; Tong and Rouault, 2006). Though these reactions occur in every organ, the role of iron in the brain is particularly important. Brain has the highest demand for oxygen of all organs and iron-containing neuroglobin is essential to meet this need. In addition, to maintain the complex communication network and to establish plasticity, the brain constantly remodels cell contacts and synapses. These processes rely on protein synthesis by ribosomes which, in turn, depend on [4Fe–4S] cluster-containing proteins (Kispal et al., 2005). The iron-sulfur clusters are also essential for DNA synthesis (Netz et al., 2012). Moreover, iron is an essential trophic factor needed for myelination. This is because key enzymes such as 3-hydroxy-3-methylglutaryl-coenzyme A reductase (HMGCoAR), squalene synthase, and glucose-6-phosphate dehydrogenase needed for the biosynthesis of myelin cholesterol and lipids require iron as a cofactor (Lange and Que, 1998; Todorich and Connor, 2004). The role of iron in these processes is thought to reflect its ability to (1) donate electrons needed for redox reactions, (2) transfer electrons in mitochondria, and (3) to bind oxygen in heme. But if iron accumulates in an unchelated form, its ability to readily exchange electrons can result in formation of reactive free radicals by the Fenton reaction (Lloyd et al., 1997). In turn, this can lead to oxidative stress that can negatively impact a number of cellular processes and disrupt cell membrane integrity. Because oxidative stress is considered a causative agent in the etiology of AD (Castellani et al., 2007; Cai et al., 2011), the role of iron and disorders resulting in iron overload have been under intense investigation for their potential role as a risk factor for AD.

## **IRON IN AD**

Several studies have reported that brain iron content and expression of iron-regulating proteins such as ferritin (Frt; high expression indicates more iron) increase with age, but the increase in iron is greater than ferritin thus suggesting the increased iron is not stored properly (Hallgren and Sourander, 1958; Milton et al., 1991; Connor et al., 1992a,b, 1995; Bartzokis et al., 1994, 2007; Martin et al., 1998; Hirose et al., 2003). The finding that iron accumulation is significantly higher in brains of patients with AD or mild cognitive impairment (MCI) than in those of age-matched non-demented controls (Ding et al., 2009) supports the proposal that the increase in iron underlies age-related cognitive decline (Bartzokis, 2009). Disturbances in iron metabolism have been found in post-mortem brain tissue from patients with AD, in cerebrospinal fluid (CSF) and *in vivo* using magnetic resonance imaging (MRI). Regions where iron deposition was found include the hippocampus, basal ganglia and cortex (Loeffler et al., 1995; Bartzokis et al., 2000; Ding et al., 2009) as well as senile plaques and neurofibrillary tangles and cells surrounding them (Connor et al., 1992a). Moreover, higher brain iron content was also found to correlate with the severity of cognitive impairment (Zhu et al., 2009). Thus, use of MRI to monitor brain iron has been proposed as a method for assessing disease progression. Excess unchelated iron is considered a major cause of oxidative stress that can lead to modification of proteins, lipids, DNA, and RNA, thereby, inducing severalfeatures associated withAD (Markesbery, 1997; Ferrari, 2000). These alterations are toxic to cells because they result in activation of cell apoptosis pathways and eventually cell death. In fact, markers of protein oxidation and lipid peroxidation were consistently found to be elevated in the brains and CSF samples from AD patients and those with MCI who subsequently developed AD (Markesbery, 1997; Pratico et al., 2002; Brys et al., 2009). Moreover, markers of lipid peroxidation in ventricular fluid were reported to correlate with cortical atrophy, reduced brain weight and severity of AD (Montine et al., 1999). Proteins involved in maintaining iron homeostasis include the transferrin receptor (TfR; iron uptake), transferrin (Tf; iron transport), ferritin (Frt; iron storage), HFE (iron uptake), ceruloplasmin (iron transport; feroxidase- conversion of Fe2<sup>+</sup> to Fe3+), DMT1 (iron export), and ferroportin (iron export). Indeed, alterations in the expression pattern of Tf, Frt, and HFE were found in brains of AD patients and are discussed below. Moreover, genetic mutations in *Tf* (C2 variant; rs1049296) and two variants of *HFE,* C282Y (rs1800562) and H63D (rs1799945), were shown to affect body iron status (Fleming et al., 2000; Bartzokis et al., 2010; Kauwe et al., 2010), and oxidative stress (Loeffler et al., 1995; Lleo et al., 2002). Where H63D has been associated with increased risk of AD, C282Y has been found to be protective for AD. There is substantial evidence for the role of iron overload in AD that is further supported by epidemiological evidence implicating variants of iron management proteins in AD risk. The AlzGene meta-analysis of the Tf C2 allele (Bertram et al., 2007; www.alzgene.org/) currently shows a significant, although low, odds ratio of AD: 1.2 (95% confidence interval, 1.06–1.3; June 2010), with a similar pattern in Caucasians and east Asians.

## **IRON, HFE, AND AD**

Iron accumulation in the brain is accompanied by an increase in oxidative stress that is consistently observed in AD (Connor et al., 1992a; Markesbery, 1997; Liu et al., 2005; Maynard et al., 2005). The high iron (*HFE*) gene product primarily regulates iron uptake into cells by interacting with the TfR to restrict Tf binding (Feder et al., 1998). Genetic variants of the *HFE* gene are unable to maintain iron homeostasis, and in particular, the H63D variant has been under investigation as a potential risk factor for neurodegenerative diseases. Several studies have reported an increased frequency of the H63D mutation in AD patients (Moalem et al., 2000; Sampietro et al., 2001; Combarros et al., 2003; Pulliam et al., 2003; Connor and Lee, 2006). Pulliam et al. (2003) found increased levels of markers of oxidative stress in individuals with HFE mutations compared to controls. The H63D mutation was expressed ∼5 times more frequently in AD patients younger than 70 years compared to patients over 80 (Sampietro et al., 2001). Indeed, MRI studies revealed increased accumulation of iron in carriers of HFE mutations (Nielsen et al., 1995; Berg et al., 2000). Studies have shown that iron staining is most dense in the proximity of amyloid beta (Aβ) plaques and in cells associated with plaques (Connor et al., 1992b) and that iron promotes deposition of Aβ (Rottkamp et al., 2001). Interestingly, a similar pattern was observed with respect to HFE expression in AD brains, suggesting HFE expression may be increased in cells in the vicinity of the plaque. A normal response of HFE is to limit iron uptake thus these cells may

be trying to limit iron uptake; a function that would be compromised if the mutant protein is expressed (Connor and Lee, 2006).

## **HFE GENE AND VARIANTS**

*HFE* (high iron), originally called HLA-H, is a major histocompatibility complex (MHC) class I-like gene. It is located on the short arm of chromosome 6 and was first identified by Simon et al. (1976) in association with an iron overload disorder called hereditary hemochromatosis (HH). The HFE gene encodes a 49 kDa membrane glycoprotein that resembles the MHC-class1 family of proteins. The three most common single nucleotide polymorphisms (SNPs) in the HFE gene are H63D (*rs1799945* in exon 2), C282Y *(rs1800562* in exon 4), and S65C (Merryweather-Clarke et al., 2000; Le Gac et al., 2001). Examination of the frequency and global distribution of HFE mutations indicated that HFE variants are most common in Caucasian populations with afrequency of up to 25% for H63D, 12% for C282Y, and 1.6–5.5% for S65C. S65C is not as common as the other two variants (Camaschella et al., 2002) and has not been associated with physiological changes or diseases (Le Gac et al., 2001).

Not only are the H63D and C282Y variants prevalent, epidemiological evidence associates them with several neurodegenerative, metabolic diseases, and malignancies. Though C282Y is less prevalent in the general population than H63D, it is strongly associated with HH and recently was identified as a risk factor for prostate, breast, colorectal, and brain tumors (Shaheen et al., 2003; Syrjakoski et al., 2006; Osborne et al., 2010; Gannon et al., 2011). It has also been found to protect against neurodegenerative diseases (Buchanan et al., 2002; Correia et al., 2009). In contrast, the H63D variant was identified as a risk factor for several diseases such as AD (Moalem et al., 2000; Sampietro et al., 2001; Combarros et al., 2003; Pulliam et al., 2003; Percy et al., 2008), amyotrophic lateral sclerosis (ALS; Wang et al., 2004; Goodall et al., 2005), and stroke (Ellervik et al., 2007).

## **FUNCTION OF HFE**

The best characterized function of HFE is its role in regulation of cellular iron uptake which it mediates by binding to the TfR. Binding affinity of HFE for TfR is comparable to the binding affinity of iron-loaded transferrin (holotransferrin; Feder et al., 1998). When a cell has enough iron, the HFE competitively binds to the TfR at the same site as iron-loaded Tf molecules thereby preventing iron uptake via endocytosis. Formation of the HFE-TfR complex is pH-dependent, with a neutral pH allowing tight binding and an acidic pH inhibiting it (Lebron et al., 1998, 1999). HFE can prevent endocytosis of TfR alone by inducing its phosphorylation. This results in higher expression of cell surface TfR than is found intracellularly (Salter-Cid et al., 2000).

The function of HFE in iron regulation is disrupted by mutations causing formation of the H63D- and C282Y-HFE variants. It has been shown that while WT-HFE bound to TfR allows only one iron bound Tf (Fe-Tf) to bind per TfR and be taken up by cells (Lebron et al., 1998, 1999), H63D-HFE allows uptake of more than one molecule of Fe-Tf because it does not reduce the affinity of Fe-Tf for TfR. Because C282Y-HFE is retained in the trans-golgi complex due to its inability to bind β2M and be transferred to the cell surface, it does not interact with TfR, thus resulting in more iron uptake than WT-HFE (Feder et al., 1997; Waheed et al., 1999, 2002). In sum, both variants are associated with cell and tissue iron-overload. For more details see Connor and Lee (2006).

In addition to the increased iron accumulation seen in cells expressing H63D-HFE (Lee et al., 2007; Mitchell et al., 2009b), disruption of mitochondrial potential (Lee et al., 2007), increased influx of intracellular Ca2<sup>+</sup> (Mitchell et al., 2009b), increased glutamate uptake (Mitchell et al., 2009b), increased secretion of monocyte chemoattractant protein-1 (MCP1) that has a role in neuroinflammation (Mitchell et al., 2009a), increased ER stress (Liu et al., 2011), increased oxidative stress (Lee et al., 2007), increased toxicity to Aβ (Mairuae et al., 2010), and decreased Pin1 activity (Hall et al., 2010) that contributes to increased Tau phosphorylation (Hall et al., 2011) are also found. These findings demonstrate that the expression of H63D-HFE creates a permissive milieu for processes that can influence other pathways in neuronal cells such as lipid homeostasis, neurotransmission, and myelination that may ultimately lead to AD.

## **LIPID DYSHOMEOSTASIS IN NEURODEGENERATIVE DISEASES**

## **CHOLESTEROL METABOLISM IN THE BRAIN**

Cholesterol is an essential component of all cell membranes and is needed for maintaining their structure and fluidity. It is required for growth and replication of mammalian cells, exo- and endocytosis, and is a precursor of steroid hormones and bile acids. Cholesterol is synthesized in the liver through a series of reactions with the rate-limiting step catalyzed by HMGCoAR which catalyzes conversion of HMG-CoA to mevalonic acid. Cholesterol is transported to organs other than the brain in the form of lipoproteins. In the brain, almost all of its cholesterol is synthesized *in situ*, with little to none taken up from the circulation. The insulation of brain from changes in circulating cholesterol is achieved by the blood–brain barrier (BBB). The highest rate of cholesterol synthesis in the CNS occurs during early stages of development especially when myelination is occurring (Jurevics and Morell, 1995, 1997). The mature brain continues to synthesize cholesterol but at lower levels (Mcmillan et al., 1972; Boyles et al., 1985). The finding that the half-life of cholesterol in the adult human brain is approximately 5 years, indicates that it must have an efficient cholesterol recycling mechanism (Bjorkhem et al., 1998).

In the brain, oligodendrocytes, astrocytes and neurons are capable of synthesizing cholesterol (Saito et al., 1987; Suzuki et al., 2007). Most of the cholesterol synthesized by oligodendrocytes is found in myelin. Neurons have a high need for cholesterol due to membrane turnover during synaptic transmission. Instead of synthesizing their own, neurons obtain most of their cholesterol from astrocytes (Srivastava et al., 1997; Stone et al., 1997) which when packaged with ApoE, also synthesized by astrocytes (Pitas et al., 1987a,b; Schmechel, 1993; LaDu, 1998; Dolev and Michaelson, 2004), is delivered to neurons where it is bound by cell surface lipoprotein receptors such as low density lipoprotein receptor (LDLR) and LDLRrelated protein 1 (LRP1) that recognize ApoE prior to endocytosis. Though other apolipoproteins are also found in the CNS, i.e., Apo AI, ApoJ, ApoD, evidence suggests that ApoE is highly abundant in the brain and is responsible for the shuttling of cholesterol between astrocytes and neurons (Bjorkhem and Meaney, 2004). A schematic depicting cholesterol metabolism in the CNS is shown in **Figure 1**.

In addition to recycling of cholesterol, there is need for removal of excess cholesterol from the brain. Due to its hydrophobic nature, cholesterol cannot cross the BBB. Two mechanisms have been proposed for its removal; cholesterol hydroxylation and ApoEdependent efflux. The presence of ApoE-bound cholesterol in the CSF indicates that some excretion of brain cholesterol occurs this way but the mechanism is not understood. Based on the rate of CSF renewal and the amount of cholesterol found in the CSF, it has been estimated that 1–2 mg cholesterol may be eliminated from the brain via the CSF each day (Pitas et al., 1987a). The major mechanism of cholesterol clearance from the brain is its conversion by the enzyme cholesterol 24-hydroxylase (CYP46A1) to 24Shydroxycholesterol (24S-HC), a metabolite that can traverse the BBB (Lutjohann et al., 1996; Bjorkhem et al., 1997, 1998). Introduction of the hydroxyl group in the side chain of the cholesterol molecule causes a rearrangement of membrane phospholipids that allows excretion of the oxysterol in an energetically favorable manner (Kessel et al., 2001). The direction of movement is mediated by the concentration gradient (Bjorkhem and Meaney, 2004). Because almost all of the 24S-HC originates from the brain, it has been suggested that it can be used as a marker of brain cholesterol homeostasis. Consistent with this view, a number of studies have shown altered levels of 24S-HC in the CSF or plasma of patients with neurological diseases such as AD and multiple sclerosis (MS; Bretillon et al., 2000a; Lutjohann et al., 2000; Leoni et al., 2002). These studies also highlight the importance of a normal flux of cholesterol across the CNS that for healthy humans is approximately 0.09 mg/day per kg (Saito et al., 1987) and 1.4 mg/day per kg in mice (Panzenboeck et al.,2002). In people with AD cholesterol efflux across the CNS is elevated and is proportional to the severity of dementia (Pfrieger, 2003a). A similar trend was observed in a mouse model of Niemann Pick C1 disease (NPC1), where the increase in cholesterol efflux from the brain

was associated with increased neurodegeneration (Panzenboeck et al., 2002).

#### **ROLE OF CHOLESTEROL IN THE BRAIN**

Two major functions for cholesterol in the nervous system are in myelination and synaptic transmission. In addition, it is an essential constituent of cellular membranes and of detergent insoluble lipid rafts, areas enriched in sphingolipids and proteins involved in signal transduction, where it can influence signal transduction and cellular processes. The importance of cholesterol in the brain is evident from the fact that it has ∼25% of the body's cholesterol (Bjorkhem andMeaney,2004). In the brain, cholesterol is enriched in myelin (i.e., oligodendroglia) and is an essential component of membranes of all cells. Notably, myelin consists of ∼70% lipids [mostly cholesterol, phospholipids, and glycosphingolipids (GSLs), i.e., galactocerebroside in molar ratios of ∼4:4:2] and 30% proteins (Bjorkhem and Meaney, 2004). The high lipid content of myelin is consistent with the membrane properties required for its role in supporting saltatory conduction. Myelin is produced by oligodendrocytes (Miller, 2002) and cholesterol enrichment in myelin reduces permeability to ions allowing the action potential to move down the axon without diffusing across the membrane (Bjorkhem and Meaney, 2004). It has been shown that a high cholesterol content is needed for proper myelination (Saher et al., 2005). A deficiency in oligodendrocyte cholesterol synthesis was shown to delay proper myelination (Marcus and Popko, 2002; Saher et al., 2005) and to be crucial for the development and maintenance of myelin membranes (Colognato et al., 2002; Gielen et al., 2006; Debruin and Harauz, 2007). Low cholesterol levels in an aging brain could contribute to the loss of myelination that Bartzokis (2011) postulated was an underlying cause of multiple degenerative brain disorders.

Cholesterol content of a membrane regulates its fluidity with greater amounts of cholesterol making them more rigid and lesser amounts making them more fluid and allowing more permeability to ions (Barenholz, 2002). Therefore, disruption of cholesterol distribution in the membrane can affect its function in maintaining normal cell activity. Consistent with this view, altered dendritic

morphology was observed upon depletion of membrane cholesterol in cultured neurons (Holtzman et al., 1995). Disruption of lipid homeostasis significantly alters CNS structure and function possibly by affecting the composition of lipid rafts found in neurons, astrocytes, and oligodendrocytes (Tsui-Pierchala et al., 2002; Gielen et al., 2006; Debruin and Harauz, 2007). Changes in cholesterol concentration can also affect cell surface availability of the carbohydrate moieties of GSLs (Novak et al., 2013) which in turn might affect a cell's ability to respond to GSL-carbohydrate binding proteins.

Several studies have shown that cholesterol is needed for endoand exocytosis and plays a crucial role in synapse structure and function (Pfrieger, 2003b; Takamori et al., 2006). Consistent with this, cholesterol is enriched in presynaptic terminals and its pharmacological depletion reduces synaptic transmission (Suzuki et al., 2004). Several key synaptic proteins such as synaptophysin and the soluble NSF attachment protein receptor (SNARE) proteins are either predominantly found in lipid rafts or must be recruited into them in order to effectively interact with other proteins to promote neurotransmitter release and transmission. These effects are thought to be mediated by direct interaction of cholesterol with synaptophysin (Thiele et al., 2000), and formation of the SNARE complex is cholesterol-dependent (Lang et al., 2001; Mitter et al., 2003). Moreover, Suzuki et al. (2007) showed that an increase in the cholesterol content of lipid rafts, induced by treatment with brain derived neurotropic factor (BDNF), resulted in increased expression of raft-associated presynaptic proteins, changes associated with synapse development. Collectively, these studies highlight the importance of cholesterol in regulating expression of synaptic proteins and for synaptic transmission.

## **BRAIN vs. PLASMA CHOLESTEROL IN AD**

Findings regarding cholesterol content in the plasma of aging and AD individuals are inconsistent with regards to whether elevated or lower serum cholesterol is a risk factor for AD. A consistent observation in human studies is that high serum cholesterol in mid-life can increase the risk of developing AD later in life, but is associated with improved cognition in elderly individuals (Schreurs, 2010). Similarly, lower total serum cholesterol (Kim et al., 2002) and lower high density lipoprotein (HDL) cholesterol were found to be associated with poor cognition in the elderly (Atzmon et al., 2002). However, status of brain cholesterol could be a better predictor of AD pathogenesis. With regard to the exchangeability between serum and brain cholesterol, the idea that under normal conditions cholesterol from serum does not cross the BBB into the brain is well-established. Support for this concept was provided by studies showing no incorporation of label into lipids of the brain or spinal cord after administration of D2O to adult rats (Waelsch et al., 1940). These findings have been replicated by several investigators (Jurevics and Morell, 1995; Dietschy and Turley, 2001). Intravenous injections of 14C-labeled cholesterol into healthy volunteers and pregnant women resulted in no label being found in the brains of the healthy volunteers (Meaney et al., 2001) or fetal brain tissue (Plotz et al., 1968), emphasizing that almost all cholesterol in the brain comes from *de novo* synthesis with little to none from plasma under normal physiological conditions. Surprisingly, brain endothelial cells have been shown to take up

a small amount of LDL cholesterol via LDL receptors expressed on their luminal surface (Dehouck et al., 1994, 1997). It has been hypothesized that uptake of small amounts of plasma lipoproteins into the CNS could occur at levels too low to be detected by current methods (Dietschy and Turley, 2001). However, significant uptake of cholesterol from plasma could occur under pathological conditions where the integrity of the BBB is compromised or in cases of chronic hypercholesterolemia. This possibility is supported by studies showing a number of metabolic changes in the brains of individuals/experimental animals fed diets high in cholesterol. Currently, knowledge is limited in this respect and further studies are needed to elucidate the effects of dietary cholesterol on brain cholesterol, especially in cases of neurological disease.

## **CHOLESTEROL HOMEOSTASIS AND AD PATHOLOGY**

There is substantial epidemiological, genetic, and molecular evidence of disruption of cholesterol homeostasis in AD (Shobab et al., 2005; Sjogren et al., 2006). Mutations in several genes involved in cholesterol uptake, such as LRP and the very-lowdensity lipoprotein (VLDL) receptor (Zerbinatti and Bu, 2005), as well as in enzymes that regulate cholesterol catabolism such as Cyp46A1 (Wolozin, 2003; Vaya and Schipper, 2007), have been associated with increased risk of AD. In addition epidemiological studies revealed increased susceptibility to AD in patients with elevated plasma cholesterol levels in mid-life (Jarvik et al., 1995; Notkola et al., 1998; Roher et al., 1999). However, studies of cholesterol in brains obtained from animals with AD symptoms and AD human brain autopsy tissue have yielded inconclusive results regarding whether high or low brain cholesterol is a risk factor for AD. The lack of agreement between studies could be due to analysis of cholesterol content from different regions/domains, i.e., total brain cholesterol vs. membrane cholesterol vs. cholesterol in lipid raft domains. It has been reported that while the asymmetric distribution of cholesterol in the plasma membrane of aged mice was reduced relative to that of younger mice, total membrane cholesterol was unaltered (Igbavboa et al., 1996). However, cholesterol content of lipid rafts from brains of older mice expressing either human ApoE3 or ApoE4 was greater than in those from brains of younger animals (Igbavboa et al., 2005). It has also been reported that brain cholesterol increases in certain conditions with age, i.e., NPC1 (Vance, 2006) while studies of other conditions provided evidence for a decline in total brain cholesterol, i.e., AD (Ledesma, 2003) and Huntington's disease (Valenza et al., 2005). In addition to total brain cholesterol, maintenance of cellular cholesterol content is crucial in mediating APP processing via alpha and gamma secretases (Bogdanovic et al., 2002; Cam and Bu, 2006) and therefore has physiological relevance to AD. Ginsberg et al. (1993a,b) have shown that in AD brains cell membranes are less stable due to altered lipid composition. In sum, experimental studies have shown that either abnormally high or low cholesterol can be associated with pathogenic manifestations seen in neurological diseases. Excessive loss of brain cholesterol would be expected to be particularly devastating because the rate of cholesterol synthesis in the brain decreases with age and under normal conditions there is no influx of cholesterol from the plasma into the brain.

ApoE is a major apolipoprotein and cholesterol carrier in the brain (Mahley, 1988). It is synthesized and secreted primarily from astrocytes as part of small dense cholesterol containing lipoproteins (Pitas et al., 1987a,b; Schmechel, 1993; LaDu, 1998; Dolev and Michaelson, 2004). Neurons take up astrocyte-released ApoE via LRP1 mediated endocytosis (Herz, 1988; Jeon and Blacklow, 2005). In humans, ApoE exists as three isoforms (E2, E3, E4; Mahley, 1988). The association of ApoE4 as a strong risk factor for AD has been well established (Mahley, 1988; Corder, 1993; Mahley et al., 2006). Despite advances in understanding how ApoE functions, the molecular mechanisms by which ApoE4 contributes to AD are not completely understood. In the general population, ApoE3 is the most common isoform (allele frequency 77–78%), followed by ApoE4 (15%; Mahley, 1988). However, ApoE4 is present in ∼40% of AD patients (Corder, 1993). Isoform-specific effects in cholesterol transport were observed with ApoE4 being less efficient thanApoE3 (Michikawa et al.,2000; Gong,2002; Rapp et al., 2006). There is also evidence that ApoE3-expressing astrocytes can supply more cholesterol to neurons than those expressing ApoE4 (Gong, 2002). Recent reports about the ability of neurons to recycle ApoE and retain ApoE inside cells (Fazio et al., 2000) have led to the hypothesis that ApoE may play additional roles in mediating cell signaling and intracellular-homeostasis. One proposed role, that of an antioxidant, is based on its ability to bind metals such as copper, iron, and zinc, with its highest binding affinity being for iron (Miyata and Smith, 1996). Interestingly, the antioxidant and metal binding activities of ApoE were found to be allele-specific, with ApoE4 less efficient in binding metals (Mutter et al., 2004) and reducing oxidative stress than ApoE3 (Miyata and Smith, 1996).

## **SPHINGOLIPIDS IN NEURODEGENERATIVE DISEASES**

Sphingolipids contain sphingosine as their basic building block and are enriched in lipid rafts. When substituted with a fatty acid in an amide linkage on C2, ceramide is produced that when linked to phosphorylcholine yields sphingomyelin and when glycosylated yields a GSL. Sphingolipids function in cell–cell recognition, signaling cascades that result in cell proliferation, apoptosis, stress responses, inflammation, differentiation, and axon growth (Venable et al., 1995; Hannun and Obeid, 2002; Spiegel and Milstien, 2002; Lavieu et al., 2006; Snider et al., 2010). Unequivocal evidence for the need for appropriate synthesis of sialylated GSLs was provided by the finding that brains of children who lacked the ability to synthesize GM3 failed to develop normally (Simpson et al., 2004). Dysregulation of GSL metabolism has also been implicated in a number of metabolic and neurological diseases such as Fabry disease, Krabbe disease, Gaucher disease, Tay-Sachs disease, Metachromatic leukodystrophy, Niemann-Pick disease, AD (Haughey et al., 2010; He et al., 2010; Ryland et al., 2011).

Several studies have shown abnormalities in the lipid content and expression of enzymes regulating their content in AD brains. For example, the total phospholipid and sulfatide content in AD brains was decreased compared to that in controls (Skinner et al., 1989; Soderberg et al., 1992; Gottfries et al., 1996; Pettegrew et al., 2001; Han et al., 2002; Cheng et al., 2003). On the other hand, ceramide, a pro-apoptotic lipid was found to be elevated in the brains (Cutler et al.,2004) and CSF of patients withAD (Satoi et al., 2005). Ceramide can be produced by hydrolysis of sphingomyelin via sphingomyelinases, or by *de novo* synthesis from fatty acyl CoA and sphingosine. It has been proposed that oxidative stress and other age-related factors contribute to the age-related accumulation of ceramide and induction of apoptotic signaling in neurons and other cells (Kolesnick and Kronke, 1998; Cutler et al., 2004; Costantini et al.,2005; Perez et al., 2005). He et al. (2010)replicated the findings of a reduction in sphingomyelin and an elevation of ceramide in AD brains made by Satoi et al. (2005). In addition, they also found reduced levels of sphingosine-1-phosphate (S1P), a pro-survival metabolite, which correlated significantly with the levels of Aβ peptide and hyperphosphorylated tau protein (He et al., 2010).

In addition to the sphingolipids mentioned above, the relationship of gangliosides, a sub-class of GSLs, to AD pathogenesis has been investigated. Gangliosides are found in their highest concentration in the gray matter of the brain, with GM1, GD1a, GD1b, and GT1b accounting for 65–85% of them (Schengrund, 2010). Interestingly, total ganglioside content was found to be reduced in most regions of brains from early onset or familial cases of AD,while in cases of late-onset or sporadic AD, ganglioside reduction was observed only in the temporal cortex, hippocampus and frontal white matter (Svennerholm and Gottfries, 1994). These observations suggest an age-dependent and/or regionspecific pattern of distribution that could be differentially altered in AD.

Though reduction of total brain ganglioside levels in AD has been reported, significant elevation of certain ganglioside species has also been consistently reported. For example, GM1, one of the major gangliosides in brain that is enriched in lipid rafts, was found to be significantly elevated in brains of AD patients (Svennerholm and Gottfries, 1994) and, in lipid rafts isolated from their frontal and temporal cortices (Molander-Melin et al., 2005) A similar observation was made in a neuroblastoma cells expressing the H63D-HFE variant (Ali-Rahmani et al., 2014a) suggesting these changes are driven by HFE genotype and that a model for determining the mechanism for HFE impact on lipid rafts exists

Additional support for the role of gangliosides in AD was provided by the observation that GM1 stimulated production of Aβ by increasing γ-secretase activity (Zha et al., 2004), that it binds to Aβ with high affinity (Ariga et al., 2001), and could serve as a "seed" for Aβ aggregation by promoting formation of toxic Aβ fibrils (Choo-Smith et al., 1997; Yanagisawa, 2005; Kimura and Yanagisawa, 2007; Okada et al., 2007; Yamamoto et al., 2007). Indeed, GM1 was found bound to Aβ in amyloid plaques in AD brains (Hayashi et al., 2004). It has also been shown to influence AD pathogenesis by altering calcium homeostasis. It has been shown that plasma membrane-associated GM1 can enhance intracellular Ca2<sup>+</sup> levels by increasing influx of extracellular Ca2<sup>+</sup> (Ledeen and Wu, 2002; Wu et al., 2007) which might affect calcium-mediated phosphorylation of tau and APP (Schengrund, 2010). In addition to affecting APP and Ca2<sup>+</sup> homeostasis, GM1 has been found to be associated with altered content of other lipids such as cholesterol. Molander-Melin et al. (2005) reported that an elevation of GM1 and GM2 in lipid rafts isolated from AD brains was associated with a concomitant decrease in their cholesterol (Molander-Melin et al., 2005). Similarly, pharmacological depletion of cholesterol in murine neuroblastoma cells resulted in a significant elevation

in the recognition of GM1 (Petro and Schengrund, 2007), suggesting an inverse relationship between GM1 and cholesterol. The fact that both lipids are present in rafts supports the hypothesis that significant changes in either or both could affect the ability of rafts to modulate a number of cellular processes affecting neuronal function. Collectively, these studies show that sphingolipids play important roles in AD pathogenesis. Further elucidation of factors that can influence sphingolipid composition is needed. Factors inducing oxidative stress such as iron are potential targets. Future investigation in this area will further our understanding of the connection between iron and lipid metabolism and will help in development of better treatment strategies.

### **LIPID RAFTS IN AD**

Given the abundance of cholesterol and sphingolipids in the brain, lipid rafts have been proposed as critical for regulating processes such as neuronal signaling, neuronal cell adhesion, axon guidance and neurotransmission required for normal brain function. Rafts have been shown to facilitate intramembrane proteolysis to activate or inactivate several transmembrane proteins such as APP and receptors involved in neurotrophin signaling (Landman and Kim, 2004; Vetrivel et al., 2005). In addition, ionotropic receptors, pre-synaptic SNARE complex proteins (i.e., synaptophysin), as well as, post synaptic receptors [such as *N*-methyl-D-aspartic acid (NMDA) and gamma-aminobutyric acid (GABA)], and key proteins involved in myelination (i.e., CNPase, MBP) are known to localize in lipid rafts and modulation of lipid raft composition has been shown to affect their function (Marta et al., 2004; Stetzkowski-Marden et al., 2006; Willmann et al., 2006; Besshoh et al., 2007; Huang et al., 2007). Changes in the composition of rafts by depletion of cholesterol or sphingolipids can affect the structure (assembly and disassembly) and function of lipid rafts and ultimately, cellular phenotype (Keller et al., 2004; Veatch and Keller, 2005). Evidence is also accumulating to support the concept that there is a dynamic equilibrium between cholesterol and sphingolipids within the rafts and that when levels of one of these components is altered it affects the concentration of the other (Ali-Rahmani et al., 2011).

The roles of lipid rafts in the brain have been under intense investigation, particularly in the context of neurodegenerative diseases. Relevant to AD are studies showing loss of lipid rafts in the temporal cortex (Molander-Melin et al., 2005) and reduced cholesterol in rafts isolated from hippocampi of AD patients (Ledesma, 2003; Abad-Rodriguez et al., 2004). These observations are consistent with findings of significant loss of total brain cholesterol in advanced stage AD patients (Abad-Rodriguez et al., 2004). Studies focused on investigating the role of lipid rafts and their cholesterol content in amyloidogenesis have yielded conflicting results. Analyses of the effect of changes in the cholesterol content of lipid rafts, sites where APP and presenilin−1 (enzyme that cleaves APP to generate toxic Aβ) localize or are recruited to in order to promote amyloidogenesis, indicated that a high cholesterol content favored it (Bouillot et al., 1996; Lee et al., 1998; Morishima-Kawashima and Ihara, 1998; Hayashi et al., 2000). In contrast studies in which cholesterol was depleted from neuronal cells expressing human APP resulted in colocalization of presenilin-1 and APP in nonraft domains and increased Aβ production (Abad-Rodriguez et al., 2004). These findings along with the evidence of low total brain cholesterol in AD patients support the hypothesis that therapeutic approaches designed to lower brain cholesterol could be detrimental to normal brain function. In fact, accumulating evidence indicates that loss of cholesterol can promote neurodegeneration as lower serum/total brain/membrane cholesterol have been implicated in increased risk of Parkinson's (Du et al., 2012), ALS (Albert, 2008; Chio et al., 2009), Smith–Lemli–Opitz syndrome (Jira et al., 2003), Huntington's (Valenza et al., 2005), Alzheimer's (Ledesma, 2003; Wellington, 2004; Wolozin, 2004), and Niemann–Pick Type C (Vance, 2006) diseases. In summary, lipid rafts modulate a number of cellular processes that affect neuronal function and alterations in their composition (cholesterol and sphingolipid content) can influence cellular phenotype. Investigation of lipid rafts in the context of iron related disorders such as HH, AD, and iron deficiency anemia are exciting new areas of research which should be further explored to unravel potential signaling pathways underlying iron related disorders that might be treated by modulation of the lipid composition of rafts.

## **IS THERE A LINK BETWEEN IRON AND CHOLESTEROL METABOLISM?**

Contribution of iron to neurodegeneration in AD has also been proposed to occur based on its ability to influence four pathways implicated in AD: (1) APP metabolism (Mantyh et al., 1993; Bodovitz et al., 1995; Schubert and Chevion, 1995; Rottkamp et al., 2001; Rogers et al., 2002; Kuperstein and Yavin, 2003), (2), the loss of calcium homeostasis (Hidalgo and Nunez, 2007), (3) the degradation of a subset of microglia (Lopes et al., 2008), and (4) oxidative stress (Honda et al., 2004; Liu et al., 2006; Smith, 2006; Castellani et al., 2007, 2012). Here we propose that the effect of iron on lipid metabolism is also a potential contributor to AD.

Despite the efforts that have been made to identify factors that can trigger the pathological events associated with AD the sequence of events is still not clear. Accumulation of Aβ has been thought to be the triggering event, preceding NFT formation, and ultimately culminating in synaptic failure and memory impairment. However, other evidence indicates that tau pathology may precede formation of amyloid plaques. The evidence of memory decline accompanied by symptoms characteristic of AD, in the absence of amyloid plaques and/or NFTs, supports questioning whether these pathological markers are the only causative agents. Since molecular phenotypes such as oxidative stress, synaptic failure, neuronal loss, and cognitive decline, characteristics associated with AD, have been shown to result from disruption of a number of the pathways discussed above, one can easily argue that there is no linear sequence of events that causes AD. A good example of an interconnection between pathways that may have relevance to AD lies at the intersection between iron and cholesterol metabolism. Both iron and cholesterol metabolism have been independently implicated in the etiology of AD as described above. A large number of studies have established that disruption of cholesterol metabolism is associated with multiple aspects of AD, including APP metabolism, Tau phosphorylation, synaptic integrity and transmission, and cognitive function. Iron dyshomeostasis has also been shown to contribute to the above-mentioned aspects of AD, suggesting an interactive cross-talk between the two pathways that may result in synergistic deleterious effects contributing to the development of AD. Evidence is accumulating for a potential link between iron and cholesterol metabolism in the context of atherosclerosis, yet this link and its relevance to AD remains largely unexplored. There is some evidence to support this association. (1) Iron is a required cofactor for a number of enzymes involved in cholesterol metabolism so it is logical to speculate that alteration in iron content could influence the activity of those enzymes. (2) Iron, ApoE, and cholesterol have all been found in association with extracellular amyloid plaques and intracellular neurofibrillary tangles, hallmark features of AD. (3) Iron is a potent source of oxidative radicals that have been consistently reported to affect membrane lipids and thereby influence lipid homeostasis. (4) Rabbits fed a cholesterol-rich diet were found to accumulate iron and Aβ deposits in the brain, and to have increased mortality (Ghribi et al., 2006). Finally, and perhaps most relevant are the recent findings that carriers of both the ApoE4 (cholesterol transporter) allele and H63D-HFE variant (iron accumulation, cholesterol decrease) had increased risk for and 5.5 year earlier onset of AD (Moalem et al., 2000; Sampietro et al., 2001; Combarros et al., 2003; Percy et al., 2008). Collectively, these studies highlight an area of investigation that could lead to a better understanding of the underlying causes for AD and impact the 25% of the Caucasian population that carries at least one H63D allele.

The finding that carriers of both the H63D-HFE and ApoE4 alleles have an earlier onset of AD (Percy et al., 2008) indicates that there is a connection between iron, cholesterol, and AD; however, the interaction of iron and cholesterol and its relevance to AD had not been specifically investigated prior to our studies of the effect of H63D-HFE on both (Ali-Rahmani et al., 2014a). While the H63D-HFE-expressing neuroblastoma cells had increased iron content and iron is a required cofactor for a number of enzymes involved in cholesterol synthesis (e.g., HMGCoAR and squalene synthase), the cells expressed less HMGCoAR and had less total cholesterol than WT-HFE-expressing controls. They also had an increased expression of CYP46A1 (Ali-Rahmani et al., 2014a). These observations were recapitulated in studies of the brains of H67D-HFE mice. Brain cholesterol content of H67D-HFE mice was significantly higher in 6 month old mice, however as these mice aged their brain cholesterol content continued to decline much more than in WT-HFE mice (**Figure 2**). Strikingly, a significant negative correlation was found between brain iron and cholesterol content for H67D-HFE mice ages 12–24 months (Ali-Rahmani et al., 2014a). These mice had higher total brain iron content and lower brain cholesterol content than their wild-type littermates. These results are similar to the association found between increased brain (nigrostriatal) iron content and low serum cholesterol in patients with Parkinson's disease (Du et al.,2012). Though it is believed that cholesterol content of serum is not necessarily reflective of that of brain, it is important to note that in another study it was shown that patients with high serum iron and low serum cholesterol had the highest risk of PD (Powers et al., 2009), indicating the relevance of the interplay between iron and cholesterol metabolism in a neurodegenerative disease.

An inverse relationship between iron and cholesterol content was also observed in several additional studies. For example,

**FIGURE 2 |Total brain cholesterol content of mice expressingWT- or H67D-HFE.** Brains were dissected from mice at 6, 12, 18, and 24 months. After homogenization, protein concentration was determined by Bradford assay (n = 7–17 per group including both male and female mice). Lipids were extracted from 100 μl of brain homogenate. Cholesterol content was measured by following the manufacturer's protocol (Biovision), and is shown as μg of cholesterol per mg of protein (\*\*\*p < 0.001). Data were analyzed using a two-way ANOVA followed by Bonferroni tests to compare effects of age and genotype on age-related changes in brain cholesterol of HFE variant mice using Graphpad (Prism) software. Error bars represent the standard error of the mean.

Brunet et al. (1997) found that treatment of rats with an ironsalicylate complex resulted in an elevation of iron, products of lipid peroxidation, and reduced total serum cholesterol. Treatment of rats with an iron dextran complex resulted in an elevation of iron and a decrease in total serum cholesterol regardless of whether the rats were fed a normal diet or one high in cholesterol (Turbino-Ribeiro et al., 2003). Similar results were found in earlier studies in which rabbits treated with an iron dextran complex and fed a normal diet were found to have reduced serum cholesterol (Dabbagh et al., 1997). In fact, diet-induced iron overload resulted in reduced expression of liver HMGCoAR and CYP7A1 (Brunet et al., 1999).

One possible explanation for the effects of a high iron diet is that they are mediated by iron-induced oxidative stress (Dabbagh et al., 1994) and lipid peroxidation (Britton et al., 1987). Depletion of energy (ATP and NADPH) due to oxidative stress and lipid-peroxidation mediated membrane damage were shown to cause disruption of lipid synthesis and transport (Kehrer, 1993). These observations coupled with the inverse relationship between iron and cholesterol described above support the hypothesis that the changes induced in iron and cholesterol by expression of H63D-HFE and possibly additional mediators disrupt the normal neuronal function. One possible molecule involved in these interactions is Aβ. The presence of the iron response element (IRE) in the promoter region of the APP gene and the observation of an iron-concentration dependent increase in APP mRNA provides compelling evidence for the role of iron in APP expression and possibly Aβ generation (Bodovitz et al., 1995; Rogers et al., 2002; Rogers and Lahiri, 2004). Interestingly, there is also evidence that Aβ can, in turn, modulate cholesterol and sphingolipid metabolism by regulating the activities of key lipid enzymes (Zinser et al., 2007; Grimm et al., 2012). Specifically, Aβ has been shown to inhibit cholesterol synthesis by inhibiting HMGCoAR (Grimm et al., 2012), suggesting that a feedback mechanism exists between cellular cholesterol and Aβ content. Furthermore, GM1 has been shown to serve as a "seed" for Aβ aggregation into toxic fibrils (Hayashi et al., 2004; Yanagisawa, 2005; Okada et al., 2007). Intriguingly, there is evidence that an increase in GM1, or a change in its accessibility to GM1-binding proteins can occur upon reduction of cellular cholesterol (Lingwood et al., 2011). A link for a relationship between iron, cholesterol, and GM1, was provided by our finding that H63D-HFE human neuroblastoma cells contained more iron (Lee et al., 2007), expressed more cholera toxin binding GM1 (Ali-Rahmani et al., 2011) and had less cholesterol than cells expressing WT-HFE (Ali-Rahmani et al., 2014a). Taken together, our findings and the studies of the role of APP metabolism in AD and its relationship to either iron or cholesterol, provide the basis for the proposed mechanism (**Figure 3**) by which effects induced by expression of H63D-HFE may lead to the symptoms associated with AD.

With the exception of Aβ content, all of the pathological effects indicated in this model have been observed in either cells expressing H63D-HFE or in the brains of mice expressing H67D-HFE. Cellular studies indicated no difference in the levels of total APP or Aβ between WT- and H63D-HFE-expressing cells, but increased sensitivity to Aβ treatment was observed in the latter (Mairuae et al., 2010). Further study of the age-related effects of expression of the H63D variant on APP metabolism in an *in vivo* model should further define the interrelationships of HFE,APP, iron, and cholesterol metabolism, and the mechanisms by which they may contribute to dementia.

As important as it is to study the impact of iron overload on lipid metabolism, the question of what impact high serum and/brain cholesterol levels might have on iron metabolism in the brain should also be investigated. Such a study might help to explain the (1) established link between hypercholesterolemia and atherosclerosis and the fact that they both are significant risk factors for AD, (2) connection between disruption of iron metabolism and atherosclerosis, and (3) impact of a cholesterol enriched diet on iron management. Results from studies in rabbits that were fed a diet high in cholesterol showed an increase in iron accumulation in the brain (Ong and Halliwell, 2004; Ong et al., 2004; Ghribi et al., 2006) and an elevation of lipid peroxidation products (De La Cruz et al., 2000; Gokkusu and Mostafazadeh, 2003). Though passage of dietary cholesterol into the brain is prohibited by the BBB, some metabolites, small molecules, and inflammatory cytokines produced by chronic hypercholesterolemia may be able

**to AD.** We propose that (1) the increase in brain iron induced by the expression of H63D-HFE is initially compensated for by an elevation in brain cholesterol. (2) The elevated iron and cholesterol then synergistically enhance Aβ accumulation and together induce oxidative stress and lipid peroxidation. As the cycle of increased iron, increased Aβ and inhibition of cholesterol synthesis continues, the levels of oxidative radicals supersede the cell's antioxidant capacity. This affects mitochondrial function, enhances secretion of inflammatory cytokines and other toxic radicals, increases tau

the concomitant changes it induces occur they start to alter expression of enzymes involved in cholesterol metabolism thereby reducing its concentration. (4) In an effort to survive, cells adapt to a state of low cholesterol content in an effort to limit Aβ production. (5) Over time, due to decreased cholesterol and damage induced by oxidative radicals and other changes, myelin and eventually synapses disintegrate leading to cell death. The resultant neuronal loss would then impair memory function culminating in AD.

to cross from the circulation into the BBB and eventually alter the permeability of the BBB to other harmful substances. The observation of iron accumulation in the brain after long-term feeding of a high-cholesterol diet provides support for this argument and raises the question of whether and/or how hypercholesterolemia allows increased passage of iron into the brain parenchyma. The next logical question is then how would individuals with H63D-HFE be affected by these changes? We know that individuals with H63D-HFE have elevated iron levels in liver as well as in the brain (Waheed et al., 1997; Adams et al., 2005; Bartzokis et al., 2010). If these individuals develop hypercholesterolemia, they could have even more passage of iron into the brain, more oxidative stress and other disease-associated consequences. Interestingly, it has been shown that individuals with the ApoE4 allele have high serum cholesterol and an increased incidence of hypercholesterolemia. This observation may provide an explanation for why individuals with both ApoE4 and H63D-HFE have an increased risk and earlier onset of AD. Further elucidation of this mechanism would provide an explanation for why atherosclerosis and hypercholesterolemia, conditions prevalent in many aging individuals, may contribute to the development of AD. In relation to that, studies aimed at teasing out the effects of diet (high-iron or high cholesterol, or both) on mice expressing WT-HFE (have normal iron; able to protect cells from iron overload) should yield results indicating the extent to which diet-induced changes affect iron and cholesterol metabolism in the brain.

## **HOW DOES H63D-HFE AFFECT INTERACTIONS BETWEEN GLIAL AND NEURONAL CELLS?**

Experimental results indicate that H63D-HFE creates a permissive milieu for pathogenic processes (Lee et al., 2007; Mitchell et al., 2009a,b; Hall et al., 2010, 2011; Mairuae et al., 2010; Ali-Rahmani et al., 2011; Liu et al., 2011). These observations raise the question of how the effects of iron accumulation and resulting oxidative stress, elevation of extracellular glutamate, increased secretion of MCP-1 and 24S-HC in one cell type impacts other cells in the brain and *vice versa*. The importance of this question is highlighted by the characteristic iron-deficient phenotype of macrophages in HH patients, a puzzle that has not yet been solved. It has been shown that although HH patients have a significant iron overload in their livers and other organs, their macrophages appear to be iron-depleted (Cardoso and de Sousa, 2003). Consistent with this observation, cell studies have shown that the normal function of WT-HFE is not only to manage iron uptake but to prevent its efflux from cells into the extracellular space (Drakesmith et al., 2002). Because this function is impaired in H63D-HFE cells more iron is released from macrophages (Drakesmith et al., 2002). Considering that microglia in brain function similarly to macrophages in the circulation, expression of H63D-HFE may affect them analogously. If it does, it is likely that H63D-HFE expressing microglia would release more iron than normal and this could affect other cells in the CNS. Support for this is provided by results from a number of studies that examined the effects of iron on cells in the CNS. Zhang et al. (2007) showed that conditioned media from iron-loaded microglia increased survival of oligodendrocytes by delivering iron in ferritin, which could contribute to the proliferative effects on oligodendrocytes. However, lipopolysaccharide

(LPS) activation of iron-loaded microglia attenuated the proliferative effects on oligodendrocytes and caused them to release inflammatory cytokines (Zhang et al., 2007). It has also been shown that microglial activation and secretion of inflammatory cytokines can decrease ferritin synthesis via the modulatory effects of nitric oxide on the iron regulatory system (Chenais et al., 2002). It is logical to argue that a decrease in ferritin along with a preexisting iron-overload and/or increased iron uptake would result in more intracellular free iron that could exacerbate the oxidative damage often seen in neuronal and glial cells in neurological disorders (Rotig et al., 1997; Hirsch and Faucheux, 1998; Smith et al., 1998). So the question is what causes microglial activation? Iron accumulation has been shown to influence glial activation either directly or through induction of other factors such as the MCP-1, a chemokine (Rollins, 1997) shown to cause glial activation (Vrotsos et al., 2009) as well as induction of pro-apoptotic genes and cell death (Zhou et al., 2006). Of particular interest is the finding that cells expressing H63D-HFE have increased secretion of MCP-1 (Mitchell et al., 2009a). Therefore, it is likely that iron accumulation within brain cells affects microglia and this in turn, influences the function of neurons and glia, thereby creating a vicious cycle of events that potentiate stress and damage within the CNS. In such a scenario, endothelial cells, gatekeepers of the BBB, could also be influenced by these changes. In fact, experimental studies have shown that iron accumulation damages endothelial cells by induction of apoptotic pathways (Carlini et al., 2006). Therefore, maintaining iron homeostasis is essential for preserving endothelial function needed to maintain integrity of the BBB. This is crucial for protecting the brain from the harmful effects of chemicals that can be present in the circulation and for maintaining a stable flux of ions into the brain for proper impulse generation and propagation by neurons. The changes in iron and cholesterol induced by expression of H63D-HFE have major effects on the CNS and as such their effects on neuro-glial interactions should be further investigated. For such investigations, the H67D-HFE mouse model could be useful since we found that the animals expressed symptoms associated with human AD (Ali-Rahmani et al., 2014a). Schematic in **Figure 3** summarizes findings from *in vitro* and *in vivo* studies with H63D variant of HFE. The H67D-HFE mouse model could also be used to study the effects of changes in dietary iron or cholesterol, as well as the effects of statins and iron chelators on learning and memory. Results of such studies could provide insights into how geneenvironmental interactions may contribute to the pathogenesis of AD.

Because endothelial cells of brain microvasculature come in direct contact with the red blood cells (RBCs) and exchange material such as ions, nutrients, etc., with them, it was postulated that characterization of RBCs may provide useful information about the brain metabolism. The changes in the membranes of RBCs could reflect alterations in the brain and such alterations could be used as biomarkers to diagnose and monitor progression of neurological diseases. In fact, morphology of RBCs and the protein composition of RBC membranes from RBCs isolated from AD subjects were shown to be altered (Sabolovic et al., 1997; Mohanty et al., 2010). Furthermore, oxidative stress associated changes in the phospholipid content of RBC membranes

isolated from RBCs from AD patients were observed (Oma et al., 2012) and subsequently a panel of ten plasma phospholipids was shown to predict occurrence of memory impairment in aging individuals (Mapstone et al., 2014). These findings provide support for the hypothesis that RBC lipid dynamics could be important determinants in neurological diseases. Further support for this concept was provided by the observation that RBC membranes from autistic children tended to have reduced cholesterol and increased GM1 relative to those from healthy controls (Schengrund et al., 2012). Based on the finding of similar lipid changes in neuroblastoma cells expressing H63D-HFE, a logical future avenue of investigation would be to compare the lipid and protein composition of RBC membranes from individuals expressing H63D-HFE with those from both healthy people and those with neurological disorders. If indeed a biochemical profile is identified in RBCs from carriers of HFE variants (control, MCI, AD) that is associated with disease pathology, it could be evaluated for its effectiveness as a biomarker to study disease progression. Identification of such a biomarker would significantly aid in future investigations and possibly treatment. Furthermore, the fact that decreased brain cholesterol is associated with disruption of myelin and synapse management as well as cognitive function (Ali-Rahmani et al., 2014a), provides compelling evidence to ascertain whether H63D-HFE is associated with increased risk of developing autism.

## **POTENTIAL BIOMARKERS FOR AD PROGRESSION**

It is accepted that lipids are crucial for maintenance of normal brain function. However, it is difficult to analyze its lipid content in living individuals. Due to the magnetic properties of iron, brain iron status can be accessed, at least indirectly, by MRI (Bartzokis et al., 1999). However, no such tool is available to monitor the status of brain cholesterol. Owing to the high lipid content of myelin, certain MRI techniques that measure myelination by way of quantifying white matter tracts may provide an idea of lipid status, but no direct conclusions can be made. Because of the difficulty of directly monitoring changes in the brains of living individuals, researchers are searching for potential biomarkers that can be used to identify onset and follow the progression, and response to treatments of neurodegenerative diseases. With respect to brain cholesterol homeostasis, one candidate proposed as an effective biomarker is 24S-HC. Almost all 24S-HC found in plasma and CSF originates from the action of a brain specific enzyme, CYP46A1 which catalyzes the first step in the clearance of cholesterol from the brain (Bjorkhem et al., 1998). Results from a number of studies have shown an elevation of 24S-HC either in the CSF or in the plasma of patients suffering from traumatic brain injury or neurodegenerative diseases (Bretillon et al., 2000b; Cartagena et al., 2010). More specifically, it was found that during the early stages of AD or vascular dementia (Lutjohann et al., 2000) and active periods of MS, high levels of 24S-HC were present in either the CSF or circulation (Leoni et al., 2002), possibly due to ongoing demyelination (Bjorkhem and Meaney, 2004). However, in late stage AD patients' CSF levels of 24S-HC were found to be reduced (Bretillon et al., 2000a; Leoni et al., 2002, 2006; Papassotiropoulos et al., 2002); they were also reduced in the brains of patients who died from AD

(Heverin et al., 2004). The reduction may reflect the extensive neuronal loss seen in late stages of AD. Consistent with these findings, was our observation that expression of the brain-specific enzyme CYP46A1 is elevated in both human neuroblastoma cells expressing H63D-HFE and in the brains of H67D-HFE mice (Ali-Rahmani et al., 2014a). Though the amount of 24S-HC was not measured, the observed concomitant decrease in cellular and total brain cholesterol indicated that it is probable that the expression of H63D/H67D would be associated with higher levels of 24S-HC in serum or CSF or both. It would be useful to confirm this idea using the H67D mouse model. These findings could then provide a rationale for measuring the levels of 24S-HC in the plasma and CSF of individuals carrying H63D-HFE. If a relationship were established between levels of plasma and CSF 24S-HC and incidence or severity of AD, it could be used as a biomarker.

It is also possible that brain volume as determined by MRI could be used as a biomarker for monitoring relative changes in the impact of HFE genotype with normal aging and in disease. This idea is based on the early reports of elevated brain iron in the carriers of H63D (Bartzokis et al., 2010), as well as on reduced volume of brain regions involved in memory in aged H67D-HFE expressing mice, a finding that correlated significantly with the concentrations of brain cholesterol and iron, as well as measures of learning and memory (Ali-Rahmani et al., 2014a). Therefore, these findings suggest that progression of memory decline with normal aging and in patients diagnosed with AD could be monitored by using volumetric MRI measurements. These tools could also be useful for assessing the efficacy of treatment. In summary, the foregoing discussion supports the hypothesis that 24S-HC and brain volume could be useful for monitoring disease progression and possibly treatment efficacy in carriers of H63D-HFE with a neurodegenerative disease.

## **IS IT IRON ACCUMULATION OR EXPRESSION OF AN HFE VARIANT THAT IS THE CULPRIT?**

H63D and C282Y are the most common variants of HFE and expression of either results in iron overload. Additionally, a number of other cellular processes are also altered in cells expressing these variants. This raises the question of whether the observed changes result from iron accumulation or occur due to a gain-offunction of the mutation. Epidemiological studies indicate that H63D-HFE is a risk factor for AD and C282Y-HFE for certain cancers. Results in **Table 1** summarize previous findings. The fact that both variants of HFE, H63D, and C282Y, are associated with intracellular iron accumulation, yet result in opposing phenotypes, H63D-HFE promoting apoptosis and C282Y-HFE promoting cell survival, indicates that these mutations by themselves could be playing a role independent of their effect on cellular iron status. Supporting this argument are the findings that their subcellular distribution differs. Both WT- and H63D-HFE are found in the plasma membrane, while C282Y-HFE is retained in the ER/golgi. This striking difference is important because the subcellular localization could affect protein-protein interactions and cell signaling. Of potential importance with regard to cell signaling is our observation that while WT-HFE and H63D-HFE were localized in fractions containing lipid rafts, C282Y-HFE was not


**Table 1 | Comparison of** *in vivo* **and** *in vitro* **findings of H63D- and C282Y-HFE variants.**

(Ali-Rahmani et al., 2011). Because lipid rafts provide a platform for a number of cell signaling pathways, an intriguing possibility is that expression of C282Y-HFE alters at least some cell signaling thereby disrupting normal homeostasis. Here it is important to note that although H63D-HFE is present in lipid rafts, its interaction with TfR is altered such that it loses the ability to prevent Tf binding to TfR and iron uptake. Therefore, interactions of both of these mutants with other proteins may be affected. Currently, other binding partners of WT and mutant HFE proteins are not known and should be interrogated in order to elucidate its role in cell signaling. In response to the question of whether the variant itself is a culprit, current evidence indicates that the variant *per se* as well as the resulting iron overload contribute to the observed effects.

Another potential areaforfuture investigation is to identify proteins that act as molecular switches in the cell and whose expression and/or activity are directly regulated by cellular iron content. This would help to dissect the molecular mechanism(s) underlying a wide array of processes and diseases such as AD and cancer in which iron is implicated. One possible candidate is peptidylprolyl cis-trans isomerase (Pin1). It is known to play a key role in regulation of signal transduction pathways (Wulf et al., 2005), particularly in cell proliferation (Lu et al., 1996). Phosphorylation of Pin1 causes it to become inactive. Pin1 modulates the folding, activity, and stability of target proteins by causing their isomerization. The findings that Pin1 is downregulated in degenerating neurons from AD patients (Liou et al., 2003) but its expression and activity is elevated in many cancers including those of the breast, prostate, brain, lung, and colon (Ryo et al., 2001; Wulf et al., 2001; Ayala et al., 2003; Bao et al., 2004), provides support for the argument that Pin1 can serve as a molecular switch. Interestingly, varying concentrations of iron were shown to alter Pin1 phosphorylation and activity in WT- and H63D-HFE cells, indicating that iron can modulate Pin1 activity, but the change in expression is HFE genotype dependent (Hall et al., 2010); presenting another example were HFE genotype could impact cellular basis of disease.

## **THERAPEUTIC IMPLICATIONS: IS STATIN THERAPY APPROPRIATE FOR PATIENTS WITH AD AND THE HFE-H63D GENE VARIANT?**

Because AD is a multifactorial disease with evidence for the involvement of environmental factors, multiple pathways and genetic mutations in genes with diverse functions, it would be beneficial to identify subpopulations responsive to specific treatments based on their genotype. Identification of genetic and environmental factors, knowledge about their interactions, and resulting pathological markers should allow us to improve therapeutic outcomes. The studies described regarding the effects of a mutation affecting iron uptake on cholesterol metabolism and their relationship to neurodegeneration support our position that eventually medical interventions must take HFE genotype into consideration. For example, the observation that cells carrying the H63D-HFE allele have lower baseline levels of total cholesterol and exhibit slower growth relative to those expressing WT-HFE (Ali-Rahmani et al., 2014a) supports the idea that H63D-HFE positive patients with AD will differ from those with WT-HFE in their response to statin therapy. Our findings that treatment of H63D-HFE cells with a statin that can cross the BBB resulted in decreased cell survival and that statin treatment of the ALS mouse model that also carries the HFE-H63D gene variant worsens the disease (unpublished data) supports this proposal. These data imply that lowering CNS cholesterol could be deleterious to neuronal function, and more so in the carriers of H63D-HFE. Further support for this proposal is provided by a recent clinical trial that showed a reduction in right hippocampal volume after 1 year of atorvastatin therapy in AD patients (Sparks et al., 2008). Though in this study they didn't stratify based on HFE genotype, such an investigation will be valuable. This finding suggests that in addition to lowering plasma cholesterol levels in AD patients, statin treatment may have reduced brain cholesterol possibly resulting in neuronal loss and atrophy of certain brain regions. Thus, our data are suggesting that use of statins that can cross the BBB, particularly in the presence of HFE gene variants should be evaluated clinically.

In conclusion, the research discussed points out a connection between iron, cholesterol, and neuronal function. This review also provides a synopsis of some of the many questions about the roles that iron, cholesterol, and sphingolipids may have in the evolution of dementing diseases in the brain and introduces a new concept, namely that HFE gene variants dramatically influence cholesterol metabolism.

## **REFERENCES**


damage in rabbits fed with long-term cholesterol-enriched diets. *J. Neurochem.* 99, 438–449. doi: 10.1111/j.1471-4159.2006.04079.x


Rollins, B. J. (1997). Chemokines. *Blood* 90, 909–928.


Schubert, D., and Chevion, M. (1995). The role of iron in β-amyloid toxicity. *Biochem. Biophys. Res. Commun.* 216, 702–707. doi: 10.1006/bbrc.1995.2678


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

*Received: 21 April 2014; paper pending published: 09 June 2014; accepted: 24 June 2014; published online: 08 July 2014.*

*Citation: Ali-Rahmani F, Schengrund C-L and Connor JR (2014) HFE gene variants, iron, and lipids: a novel connection in Alzheimer's disease. Front. Pharmacol. 5:165. doi: 10.3389/fphar.2014.00165*

*This article was submitted to Drug Metabolism and Transport, a section of the journal Frontiers in Pharmacology.*

*Copyright © 2014 Ali-Rahmani, Schengrund and Connor. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited andthatthe original publication inthis journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*

## The role of iron in neurodegenerative disorders: insights and opportunities with synchrotron light

## *Joanna F. Collingwood1,2 \* and Mark R. Davidson2,3*

<sup>1</sup> Warwick Engineering in Biomedicine, School of Engineering, University of Warwick, Coventry, UK

<sup>2</sup> Materials Science and Engineering, University of Florida, Gainesville, FL, USA

<sup>3</sup> The Tech Toybox, Gainesville, FL, USA

#### *Edited by:*

Raffaella Gozzelino, Instituto Gulbenkian de Ciência, Portugal

#### *Reviewed by:*

James Duce, University of Leeds, UK Bohic Sylvain, Institut National de la Santé et de la Recherche Médicale, France

#### *\*Correspondence:*

Joanna F. Collingwood, Warwick Engineering in Biomedicine, School of Engineering, University of Warwick, Library Road, Coventry CV4 7AL, UK e-mail: j.f.collingwood@warwick.ac.uk

There is evidence for iron dysregulation in many forms of disease, including a broad spectrum of neurodegenerative disorders. In order to advance our understanding of the pathophysiological role of iron, it is helpful to be able to determine in detail the distribution of iron as it relates to metabolites, proteins, cells, and tissues, the chemical state and local environment of iron, and its relationship with other metal elements. Synchrotron light sources, providing primarily X-ray beams accompanied by access to longer wavelengths such as infra-red, are an outstanding tool for multi-modal non-destructive analysis of iron in these systems.The micro- and nano-focused X-ray beams that are generated at synchrotron facilities enable measurement of iron and other transition metal elements to be performed with outstanding analytic sensitivity and specificity. Recent developments have increased the scope for methods such as X-ray fluorescence mapping to be used quantitatively rather than semi-quantitatively. Burgeoning interest, coupled with technical advances and beamline development at synchrotron facilities, has led to substantial improvements in resources and methodologies in the field over the past decade. In this paper we will consider how the field has evolved with regard to the study of iron in proteins, cells, and brain tissue, and identify challenges in sample preparation and analysis. Selected examples will be used to illustrate the contribution, and future potential, of synchrotron X-ray analysis for the characterization of iron in model systems exhibiting iron dysregulation, and for human cases of neurodegenerative disorders including Alzheimer's disease, Parkinson's disease, Friedreich's ataxia, and amyotrophic lateral sclerosis.

**Keywords: iron, synchrotron X-rays, human brain, Alzheimer's disease, Parkinson's disease, neuromelanin, amyloid aggregation, magnetic resonance imaging**

## **INTRODUCTION**

The study of iron in neurodegenerative disorders emerged a century ago, arguably catalyzed by developments in histological methods to demonstrate iron in tissue, and by observations of abundant iron deposition in the human brain in health and disease. Lhermitte et al. (1924) describe a Parkinson's disease (PD) case with diminished intracellular iron and iron-rich deposits ("globules") in the globus pallidus (GP), and with iron in the substantia nigra (SN) apparently unaltered. Yet on the 90th anniversary of this paper, as we see evidence that iron chelation can modify brain iron with beneficial effects for PD patients (Devos et al., 2014), there remain open questions and continuing debate about the precise nature and consequence of iron dysregulation in PD (Gerlach et al., 1994; Galazka-Friedman et al., 1996; Oakley et al., 2007; Friedman et al., 2009; Crichton et al., 2011; Visanji et al., 2013). Similar questions about the pathophysiology of iron arise in many neurodegenerative disorders including multiple system atrophy (MSA; Visanji et al., 2013), Alzheimer's disease (AD; Smith et al., 2010), Huntington's disease, Friedreich's Ataxia, and motor neuron disease (MND) including amyotrophic lateral sclerosis (ALS). The extensive evidence for poorly liganded iron contributing to the pathophysiology

of neurodegenerative disorders is reviewed elsewhere (Kell, 2010).

Disruptions to normal iron homeostasis can exacerbate deleterious excess formation of radical species (Kell, 2010), and changes in regional or cellular iron concentration, or in the proteins responsible for tightly regulating iron metabolism, may indicate vulnerability to oxidative stress damage that is observed in the pathophysiology of neurodegenerative disorders (Smith and Perry, 1995; Castellani et al., 2007; Rouault, 2013). Aberrant peptide aggregation leads to the formation of pathological hallmarks of neurodegenerative disorders such as the beta amyloid (Aβ)-rich plaques in AD, and the alpha (α-)synuclein-rich Lewy bodies in PD. Iron has long been implicated in mechanisms of toxicity associated with aberrant peptide aggregation (Rottkamp et al., 2001), although work to determine a precise role for iron is ongoing (House et al., 2004; Everett et al., 2014a). The reductase behavior of the Aβ and α-synuclein peptides has been shown *in vitro* (Khan et al., 2006; Davies et al., 2011). The chemical reduction of ferrihydrite (iron oxide) particles by Aβ<sup>42</sup> has also now been demonstrated *in vitro* (Everett et al., 2014a). The contribution of synchrotron X-ray analysis to such studies, and to understanding fundamental structural

and conformational properties of the peptides, will be considered here.

The use of X-rays to study iron in human tissue, and proteins responsible for iron homeostasis, is well-established. For example, X-ray fluorescence (XRF) spectroscopy was used in 1968 to evaluate iron content in formalin-fixed tissues (Earle, 1968). The use of X-rays for metallomics-related research has been advanced by the design and development of synchrotron facilities: particle accelerators with a cyclical path that generate intense beams of light primarily in the X-ray region of the electromagnetic spectrum. The first synchrotrons were designed in the 1940s, and subsequently built as large-scale facilities that over several generations of development have enabled exceptionally diverse and cutting-edge research. In the past two decades there has been significant progress in improving beam focusing and detectors, facilitating chemical element imaging and analysis at micro- and nanometer spatial resolution using techniques such as synchrotron XRF (SXRF) and X-ray absorption near edge spectroscopy (XANES). There has also been progress in developing techniques to analyze proteins and protein–metal interactions including high-throughput crystal structure analysis of proteins, advancing methods to observe conformational changes to proteins as a function of temperature and pH, for example, and techniques permitting time-resolved analysis of these processes (Duke and Johnson, 2010).

Perspectives on developments over the past decade are given in several reviews of methods for the spatial analysis of metals in biological tissues: these consider both stand-alone laboratory instruments and large shared facilities including neutron spallation and synchrotron light sources. Lobinski et al. (2006) reviewed various methods for the analysis of trace metals in biological environments, noting how technical advances such as nanoflow chromatography enabled work with significantly smaller sample volumes, and how improvements in the delivery of focused synchrotron beams significantly increased scope to analyze the local chemical environment of metal ions with X-ray absorption spectroscopy (XAS). Synchrotron methods including Fourier transform infrared (FTIR) micro-spectroscopy and imaging (Miller and Dumas, 2006), scanning transmission X-ray microscopy (STXM; Obst and Schmid, 2014), and SXRF microscopy (Ralle and Lutsenko, 2009) have been considered, while Jackson reviewed the stand-alone technique of laser ablation inductively coupled plasma mass spectrometry (LA-ICP-MS) for transition metal element analysis in rodent brain (Jackson et al., 2006), and Qin et al. (2011) compared SXRF microscopy with LA-ICP-MS and secondary ion mass spectrometry (SIMS). Notably, McRae et al. (2009) produced an exceptionally broad and thorough review of techniques for *in situ* imaging of metals in cells and tissues.

Synchrotron access is necessarily limited, and experiment leadtime from proposal to beamtime is typically six to nine months. The techniques provide opportunities to advance presently intractable questions, and many of the analytical techniques available are highly complementary, require minimal sample preparation, and may be performed with little or no significant damage to the sample. For technical resources and news of developments at individual synchrotrons, readers are referred to the most recent articles cited in this review, and to the website www.lightsources.org/. Here, we provide some perspective on developments in the field over the past twenty years, drawing on examples to show how pioneering experiments in synchrotron X-ray research have contributed to our understanding of the role of iron neurobiology, and by extension its role in various neurodegenerative diseases.

## **DETERMINING IRON DISTRIBUTION AND FORM IN TISSUES AND CELLS**

Iron is the most abundant of the transition metals in human brain, but to gain a full understanding of how iron in cells and tissues is compartmentalized and bound requires exceptional analytical sensitivity and specificity. While chemical and immunohistochemical methods have been successfully used to progress understanding of iron and other metal elements at regional, cellular, and sub-cellular level, they have limitations which have prompted investigators to develop alternative techniques with greater sensitivity, specificity, and spatial resolution for the analysis of metal ions in tissues. The majority of systems are laboratorybased, including particle-induced X-ray emission (PIXE), electron probe X-ray microanalysis (EPMA), and LA-ICP-MS, reviewed in depth elsewhere (McRae et al., 2009). Within the bio-iron community, arguably the best-known synchrotron X-ray technique for iron analysis in tissues is SXRF, which has been demonstrated at increasingly high spatial resolution and rapid analytical rates to map cells and tissues in recent years (Collingwood et al., 2005a; Tomik et al., 2006; Ortega et al., 2007; Popescu et al., 2009a; Leskovjan et al., 2011; Ugarte et al., 2012).

## **SXRF MAPPING OF IRON IN TISSUES AND CELLS**

When an X-ray beam interacts with a sample, it excites natural fluorescence from the chemical elements within the tissue for which the fluorescence excitation energy is at or below the incident beam energy. The fluorescent X-rays have element-specific wavelengths, and their intensity determines the relative abundance of each element. XRF was initially used with laboratory systems to capture spectra from homogenized samples, but the beam intensity and fine optics control delivered by synchrotron facilities has enabled SXRF to evolve into a technique for highresolution mapping of samples. The peak intensity for specific elements, or the full fluorescence spectrum, may be obtained in each map pixel (**Figure 1**). SXRF emission is directly proportional to atomic abundance, so in principle SXRF can be used for quantitative analysis. SXRF mapping in biological samples has mainly been semi-quantitative to date, where challenges with matrix-matching, and localized foci of specific elements causing heterogeneous patterns of self-absorption, has made true quantification impractical. The penetration depth of the beam is typically ≤1 mm, so that for the majority of cell or tissue samples the beam will penetrate the full thickness and beyond; an exception to this being the use of thick samples in rapid SXRF (Popescu et al., 2009a). In the focused beam configuration it has become standard to map with resolution better than 100 nm at some beamlines; for XANES it continues to be typical to work at micron-scale spatial resolution.

## **X-RAY ABSORPTION SPECTROSCOPY ANALYSIS WITH XANES AND EXAFS**

fluorescence spectrum is obtained at each point, and used to generate

Determining the oxidation state and structural environment of iron in biological materials is a significant analytical challenge; more so if sample microstructure is to be preserved. Synchrotron XAS is an attractive candidate to address questions about the chemical state and mineralized form of iron in cells and tissues, because of the exceptional sensitivity and spatially resolved precision with which microfocus XAS can be performed. XANES provides chemical state information, and can be used to identify the primary forms of iron oxide and/or iron-binding metalloproteins in the sample volume from which the spectrum is obtained (**Figure 1**). For high-quality spectra (from, for example, purified samples of iron-binding proteins), the energy spectrum

may be collected over an extended range, to obtain extended X-ray absorption fine structure (EXAFS), and the data utilized to extract fundamental structural information about the local environment. EXAFS can reveal the identities and structural positions of atoms surrounding the scattering iron atoms: for example, the neighboring shells of oxygen and iron atoms to determine the types of iron oxide found in ferritin, hemosiderin, and neuromelanin.

and the XANES spectrum recorded.

The regions of the fluorescence energy spectrum used in SXRF mapping, XANES, and EXAFS are illustrated graphically in a review of the analysis of metalloproteins in cells (Cook et al., 2008), and below we consider the application of SXRF imaging and XANES to analyze iron in various neurodegenerative disorders.

## **X-RAY FLUORESCENCE ANALYSIS OF IRON IN NEURODEGENERATIVE DISORDERS**

X-ray fluorescence was used to analyze iron in neurodegenerative disorders as early as 1968, using a laboratory X-ray source (Earle, 1968). XRF spectra showing elemental intensity as a function of angle, determined by fluorescence emission energy, were obtained from the brain tissues of 11 PD cases and an unknown number of controls with no known neurological conditions. Emission peak intensities were compared to obtain preliminary indications about the concentration of iron and other elements in regions of the brain. Detection limits were judged to range, for light to heavy elements, respectively, from approximately 100 to 10 μg element/g dried tissue. This enabled specific and simultaneous analysis of physiologically important elements including the transition metals iron, zinc, and copper. Dissection and pelleting of samples prior to analysis meant that information about iron distribution in the tissue microstructure was not preserved. However, the heterogeneous and reproducible patterns of brain iron deposition enabled comparisons of regional iron levels for health and disease. Earle concluded that iron was consistently elevated in PD brain tissue compared for control. As Earle found elevated brain iron in a case of Pick's disease, as well as in an elderly man with a recent infarct, the brain iron changes were not attributed as being specific to PD. Earle acknowledged the limitations of using archived formalin-fixed tissues (see Section "Approaches to Study Design and Sample Preparation"), some of which may have been in storage since 1862, and proposed that the experiments be verified with unfixed tissue. This was done by others in due course, including Dexter et al. (1989) and Jenner (1989).

## **SXRF: THE EMERGENCE OF SYNCHROTRON X-RAY FLUORESCENCE MAPPING**

Synchrotron XRF provides orders-of-magnitude higher flux than can be achieved with bench-top X-ray systems. This, combined with developments in optics to focus high intensity X-rays into micrometer or nanometer diameter beams, significantly improved the signal that could be obtained from small sample volumes. It became practical to obtain SXRF spectra from multiple points in a sample, and to automate this process to obtain maps of cells and tissues, with each pixel containing a complete metal-ion spectrum (**Figure 1**). Synchrotron beamlines utilizing soft X-rays now routinely achieve sub-micron, and in some cases ∼10 nm spatial resolution. This energy range is better-suited to light elements, and for combined iron-fluorescence and absorption analyses it is usually more efficient to work at beamlines with access to the K-edge for iron.

Ektessabi et al. (1999) collected SXRF spectra from dopaminergic neurons in single cases of healthy control and PD brain, to investigate elemental constituents, concentrations, distributions, and chemical states of iron and other elements in the individual neurons. Mapping resolution was 6 μm × 8 μm, and incident beam energy 13.5 keV. Areas ∼ 100 μm × 100 μm were mapped in formalin-fixed tissue sectioned at 8 μm thickness prior to mounting on Mylar film. Consistent with prior XRF observations in pelleted samples (Earle, 1968), they observed iron signal associated with the neuromelanin granules to be approximately twice

as intense in PD tissue compared to the control. Neuromelaninassociated iron co-localized with other metals including calcium, zinc, and copper. Efforts were made to determine the iron oxidation state, but limited flux in the microfocus configuration prevented useful XANES being obtained from such dilute samples (Ektessabi et al., 1999). Comparisons of iron signal within and without the neuromelanin-rich neurons led to a reported ratio of 11:1; this result in formalin-fixed tissue is very different to results from a subsequent independent study where we observed a ratio ∼ 3:2 for neuron versus neuropil analysis in unfixed specimens measured by EPMA, which then became ∼1:1 when iron counts were normalized to sulfur counts (Oakley et al., 2007). One interpretation of this difference is that extra-cellular (or weakly bound) iron may have been selectively leached from the formalinfixed tissue; however, the evidence in Ektessabi's paper for raised intra-neuronal iron in PD compared to healthy control is in good agreement with our work in the unfixed tissues which confirmed raised iron for PD versus control (significant at *p* < 0.0001) in individual dopaminergic neurons using chemically unfixed specimens from 16 PD cases and 14 controls without neurologic disease (Oakley et al., 2007).

Subsequent to Ektessabi's study, Szczerbowska-Boruchowska et al. (2004) performed some of the earliest SXRF analysis on unfixed tissues from neurodegenerative disorders, comparing iron distributions in SN and spinal cord tissue from single cases of PD and ALS and a control with no neurological disorder. They worked with fresh-frozen 20 μm cryosections that were freeze-dried following mounting onto a plastic (AP-1) foil. Adjacent tissue sections were used for supporting histology. Areas 500 μm × 500 μm were mapped at 10 μm × 5 μm resolution, with higher resolution mapping (5 μm × 2 μm) of pigmented neuronal cell bodies in the SN. With a dwell time of 3 s per point, detailed maps could be obtained in a matter of hours. Relative concentrations may be well-determined by SXRF, where sample preparation and measurement conditions (such as detector position) are identical, and data are normalized to incoming beam. In the present study, standards were used to obtain apparent elemental concentrations, and detection limits were determined using the following expression developed in keeping with the original framework set out by Currie (Currie, 1968):

$$DL\_i = \Im \beta \frac{C\_i}{Y\_i} \sqrt{B\_i}$$

where the detection limit *DLi* depends on the mass per unit area *Ci* for element *i*, the net peak area of element *Yi* for element *i*, and the background *Bi*.

Szczerbowska-Boruchowska et al. (2004) observed that iron, along with Zn, S, and Cl, was concentrated in the SN neuronal cell bodies regardless of disease state. They observed, consistent with the prior study in fixed tissue (Ektessabi et al., 1999), elevation of Fe (and other elements) in the PD SN compared to the healthy control, seeing iron elevation in both nigral neuronal cell bodies (perikarya) and white matter. It was noted that iron was "*strongly accumulated inside neuron perikaryal parts and, additionally in the case of PD, in structures that were not identified histopathologically*" (Szczerbowska-Boruchowska et al., 2004). The latter may

have included extracelluar deposits (perhaps of neuromelanin) or iron-rich glia. Neurons in the spinal cord, as compared to the SN, did not carry the same typical elevation of Fe, but – in addition to other elements including Ca and Zn – the apparent Fe concentration increased in ALS spinal cord white matter compared to control.

While this study only included single cases, it is reportedly the first micron-level-precision spatial analysis of Fe focii in SN neurons to be performed in chemically unfixed tissue; an important step after the confirmation by ICP-MS (in chemically unfixed bulk tissue samples) of elevated iron in PD SN compared to healthy controls (Dexter et al., 1989). Subsequent examination of spinal cord using a slightly expanded number of ALS (*n* = 3) and control (*n* = 5) cases found variable levels of Fe in the spinal cord neuron perikarya and in the white matter. No significant differences were observed at whole tissue level (Tomik et al., 2006). The heterogenity of metal ion distribution in many atomical structures, and the associated length scales, means that SXRF mapping is not always an ideal tool for whole tissue comparisons unless anatomical regions are very well matched: this is equally a challenge for large-area rapid SXRF (Popescu et al., 2009a) as it is for microfocus studies. Here, prior observations had been made of elevated iron and selenium in lumbar spinal cord from 38 MND cases compared with 22 controls using bulk samples for neutron activation analysis (NAA). The marked elevation of Fe and Se, coupled with the absence of correlation with disease stage or motor neuron counts, prompted the conclusion that the Fe and Se elevation is an early stage event in the disease process and likely a contributing factor (Ince et al., 1994; Markesbery et al., 1995).

A strength of SXRF mapping is its ability to sensitively detemine relative spatial distributions of elements at a wide range of resolutions ranging from small anatomical regions and cell layers to the sub-cellular. Ortega et al. (2007) showcased sub-micron SXRF by achieving 90 nm resolution mapping with an extended version of the Kirkpatrick-Baez (KB) focussing geometry utilizing the hard X-ray energy range, as shown in **Figure 2**. This study of PC12 rat cells was designed to extend studies of iron in the SN by looking at the relationship between iron and dopamine. The PC12 pheocromocytoma cell line served as an *in vitro* model of dopaminergic neurons; cells were differentiated with nerve growth factor prior to exposure to sub-cytotoxic concentrations of iron and/or AMT (alpha-Methyltyrosine), an inhibitor of dopamine synthesis. The cells were freeze-dried prior to measurement, and SXRF sampling times were 1/3–1 s per point with repeat sampling, rather than extended dwell times, to achieve good signal:noise. This is consistent with recommendations for optimal preservation of metal ion information in microfocus SXRF analysis of biological materials (Bacquart et al., 2007). By utilizing a combination of epi-fluorescence to visualize dopamine, and SXRF mapping for iron, they observed the co-localization of iron and dopamine in dopamine vesicles, and demonstrated the impact of dopamine synthesis inhibition on iron distribution. Use of sub-micron resolution SXRF allowed detailed visualization of neuronal processes in addition to the perikarya, revealing that blocking dopamine synthesis led to lowering of iron content specifically in dopamine vesicles, predominantly in the neuronal processes.

**FIGURE 2 | SXRF mapping with hard X-rays at sub-cellular resolution.** Dopamine-producing cells are chemically mapped with sub-90 nm beam, revealing potassium and iron distribution. In panel **(A)** the intensity distribution in the focal plane is shown; panel **(B)** illustrates the dopamine producing cells exposed in vitro to 300 mM FeSO4 for 24 h. Chemical element distributions of potassium (K) and iron (Fe) were recorded on distinct cellular areas including cell bodies **(C)** and distal ends **(D)**. Iron was found in 200 nm structures in the cytosol, neurite outgrowths, and distal ends, but not in the nucleus. Units for the min-max intensity bar are arbitrary. Scale bars = 1 mm. Adapted from Ortega et al. (2007).

#### **THE ADVENT OF QUANTITATIVE SXRF**

The above study illustrated a significant advance in synchrotron hard X-ray capabilities, with a factor of 10 improvement in SXRF imaging resolution, and a reported detection limit of 10−<sup>18</sup> g of Fe within a 100 nm diameter structure (Ortega et al., 2007). Although SXRF is unique amongst the microprobes for its combined sensitivity and non-destructive potential, it has historically been recognized as a semi-quantitative method that demonstrates consistent but relative (not absolute) metal ion distribution when compared with quantitative microprobe techniques (Ugarte et al., 2012; O'Reilly et al., 2014). The practical challenges of SXRF analysis, including methods to analyse images and draw meaningful comparisons between samples, are explored carefully in a elemental analysis of breast tissue (Geraki et al., 2002); the study also considers challenges in SXRF calibration. In principle, quantitative analysis requires matrix-matched standards that have comparable bulk composition, density, and thickness of the sample. Due to recent efforts, SXRF mapping is now being recognized for its quantitative potential in the analysis of biological materials (Qin et al., 2011). Approximate concentrations, sufficient to permit a degree of sample comparison, have previously been obtained by use of standards or supporting analysis of the sample materials. In Ortega's analysis of the PC12 cells, hundreds of additional cells were analyzed by microPIXE, to perform an approximate calibration for the SXRF nanoimaging via mass normalization of the X-ray emission (Ortega et al., 2007). Although the concentrations achieved are not absolute, this is a good example of how complementary analytical approaches may be used to overcome the limitations of a given technique. The PC12 cell model was subsequently analyzed using the same SXRF

and microPIXE approach to investigate competition between iron and manganese (Carmona et al., 2010). Manganese toxicity can result in a parkinsonism-like disorder in humans and it has been postulated that Fe and Mn compete for serum iron binding protein transferrin, and subsequent iron transport protein divalent metal transporter (DMT), but there is conflicting evidence as to whether iron uptake is promoted or diminished by Mn exposure in PC12 cells (Carmona et al., 2010). In addition to demonstrating how Mn localizes to the Golgi apparatus, a decrease in intracellular Fe was observed after the cells were exposed to Mn, consistent with the possibility that Mn competes with Fe for binding sites and transport mechanisms (Carmona et al., 2010). An alternative approach to calibrate SXRF for elemental mapping in cells incorporates the morphological data that can be obtained from atomic force microscopy (AFM) and STXM, to correct for the self-absorption effects that can strongly distort analysis of the lighter elements such as magnesium (Malucelli et al., 2013). Meanwhile, Kosior et al. (2012) have proposed what is reportedly the first method to achieve absolute quantitative analysis of iron and other metal elements by SXRF, using quantitative phase contrast imaging to obtain the projected mass for each sample, and thereby perform the mass correction pixel-by-pixel for the SXRF maps. This is designed to address the challenges for quantification posed by factors such as the variations at the cellular level in density and cell thickness. Most SXRF analysis presently provides relative, rather than absolute, concentration maps of elements, as the additional information required to convert SXRF maps to perform the mass correction (Kosior et al., 2012) is not yet routinely acquired.

## **SXRF MAPPING: DEVELOPMENTS IN THE CONTEXT OF ALZHEIMER'S DISEASE**

In synchrotron studies to evaluate iron in neurodegenerative disorders, the primary target regions have been those previously demonstrated to accumulate iron, to be especially prone to degeneration, and/or to exhibit iron associated with pathological hallmarks of a given disease. Some of the earliest SXRF mapping experiments were therefore of the cortex and hippocampus, which undergo significant atrophy in AD. For example Ishihara et al. (2002) used small (3 mm × 4 mm) blocks of frozen tissue from the superior temporal gyrus of two confirmed AD cases, briefly fixed in 4% paraformaldehyde and 2% glutaraldehyde for 2 h prior to cryosectioning at 5 μm thickness and mounting on polyester (Mylar®) film. Retrospectively, hematoxylin, and eosin staining was used to determine tissue structure (Ishihara et al., 2002). The small areas and absolute numbers of cells sampled, in combination with the absence of a control for the AD cases, limit the conclusions that can be drawn from this study. The fixation step also raises the possibility that loosely bound metal ions may have been mobilized and displaced. However, this cellularresolution multi-elemental mapping of neuronal perikarya in the cortex is an early demonstration of how SXRF was introduced to study transition metal ion distributions in brain regions exhibiting significant pathology. SXRF was used during this same period to compare elements including iron, zinc, and copper in frozen unfixed rat brain tissue, as part of a study on iodine deficiency (Zhang et al., 2002). Relative concentrations for hippocampus

versus cerebral cortex were plotted, but the structures within cortex and hippocampus were not spatially resolved. As observed for iron using histochemical methods (Morris et al., 1994), and more recently for a range of metals using microbeam analysis, the layers within the hippocampus exhibit distinct patterns of iron and zinc distribution. The zinc elevation in the dentate gyrus has been illustrated by SXRF in rat hippocampus (Flinn et al., 2005), and the iron-rich layers in the layers of the surrounding cornu Ammonis have been shown by SXRF in human hippocampus (Antharam et al., 2012). It is of interest to determine these spatial distributions of transition metals in the hippocampus, which is especially prone to atrophy in AD, as analytical studies of bulk tissue iron content in unfixed human hippocampus have produced conflicting results as to whether iron concentration in the AD hippocampus is elevated compared with healthy controls (Schrag et al., 2011).

## **RAPID SCANNING SXRF**

As the spatial resolution limits for hard X-ray SXRF of single cells have been pushed in recent years (Ortega et al., 2007), so have the rates at which large sample areas can be imaged (Popescu et al., 2009a,c). The time required to map a sample is primarily dependent on matrix size and the dwell time (or effective dwell time, for raster scanning) per point in the matrix. It would usually be desirable to map at a minimal resolution ∼50 μm, preferably higher, to permit delineation of primary cell layers if not individual cells. For a tissue section cut with thickness ∼20 μm, and with typical elemental concentrations requiring ∼1 s per point (including overhead) to obtain useful fluorescence signal, it would therefore take a little over 10 h to map a 10 mm × 10 mm area. Use of synchrotron experiment time (beamtime) is often optimized by minimizing the area required for scanning, and working at the lowest resolution that permits the experiment aims to be achieved. However, Nichol and colleagues have looked to overcome the limited sample area that can be analyzed by SXRF by developing an approach using very thick (conventional autopsy) samples. This maximizes the opportunity for the incoming beam to interact with the sample and thereby maximizes fluorescence yield from each anatomical region of interest, subject to any selfabsorption effects. As Popescu et al. (2009a) observe, the escape depth is element-dependent, so although under most conditions they can obtain relative metal ion distribution maps across a sample area, the penetration depth for each map will differ as a function of the element mapped. Their approach, utilizing thick tissue slices and rapid SXRF scanning, has enabled transition metal maps to be generated for selected regions of the central nervous system in cases of PD, spinocerebellar ataxia, and healthy control (Popescu et al., 2009a,b,c). This is an advance on histochemical staining in that the elemental maps are achieved simultaneously from a given sample, and there is excellent specificity (if spectra are fitted, rather than simply gated), for elements in addition to iron such as copper and zinc, so long as the element-specific differences in sampling depth are not critical to the experiment design. Of the transition metals in brain tissue, Fe is the most abundant on average, followed by Zn, and then Cu. One of the sacrifices of rapid SXRF (which also arises in more conventional microfocus SXRF measurements) is that in order to minimize scan

time, the signal:noise for the individual pixel spectra are usually insufficient to gather robust data about other more dilute metal ions present in the tissue such as Mn, as typically occurs when working with an effective time of <6 ms per 40 μm pixel. This can be addressed by reducing the scan rate to increase the effective dwell time per pixel (although this must be limited above a certain value to avoid beam damage), or by repeating scans (Bacquart et al., 2007). Although high effective in-plane resolution can be achieved with rapid SXRF, the beam penetrates comparatively deeply into the sample at each point in the matrix, so that a single pixel represents a small surface area but a significant depth [1/e attenuation of the Fe signal is quoted as 310 μm for brain tissue (Hopp et al., 2010)]. The angle between the incoming beam path and the fluorescence detector is 90◦, and each is, respectively, at 45◦ to the sample (see general illustration in **Figure 1**). This path must be considered if trying to define boundaries in order to undertake any high resolution multi-model analysis of SXRF maps from thick samples. Each SXRF approach presents constraints for sample preparation; for the rapid SXRF which uses large-area thick sections, samples have been fixed through immersion in formalin, which others have reported provides a vector for redistribution or loss of loosely bound metal ions (e.g., Szczerbowska-Boruchowska et al., 2004; Schrag et al., 2011). Variations in effective sampling depth occur as a function of element, but also if there are significant variations in the matrix (for example, due to calcifications, intense foci of the element being mapped, or variable hydration states). Fixed tissues have been protected from dehydration by spraying the 1 mm-thick slices with buffered formalin and heat-sealing under thin Mylar® prior to mapping (Hopp et al., 2010).

The challenge of sample matching, in order to make comparisons between cases, is present for both rapid and microfocus SXRF studies, albeit at different length scales. Where practical, advance sectioning to ensure equivalent anatomical levels can significantly improve the comparisons that can be made. A rapid SXRF study comparing a case of PD (male, 70 years old) and healthy control brain (female, 80 years old) reveals many shared structures, although the levels in the coronal sections are slightly offset as described in the study and evidenced in hippocampus profiles (Popescu et al., 2009a). The PD midbrain exhibits a higher concentration of iron in the SN than for the control (Popescu et al., 2009a), consistent with prior observations. The midbrain section is taken at the level of the inferior colliculi so the iron-rich red nucleus is not observed. The extent to which offsets in level permit or preclude comparison will depend both on the degree of heterogeneity of iron deposition in a given brain region, and on the magnitude of differences between the study groups.

## **VALIDATING MRI MEASUREMENT OF BRAIN IRON**

As interest grows in the capacity of MRI to clinically analyse brain iron (Drayer et al., 1986; Schenck and Zimmerman, 2004), there is increasing need for direct validation of the indirect evidence for iron contrast in MRI magnitude and phase data. Correlation of MRI transverse relaxation with tissue iron concentration from bulk tissue samples has previously been achieved for AD and control brains (House et al., 2008), but a number of factors, including myelin, compete with iron to influence fundamental MRI parameters. As there is significant heterogeneity in the microstructural distribution and bound states of iron [as observed in hippocampus (Antharam et al., 2012) and SN (Blazejewska et al., 2013)], there is value in being able to directly describe the spatial relationship between iron and MRI. Given the limited sensitivity of routine iron histochemistry for this purpose [recently noted in postmortem analysis of iron and MRI in the SN (Blazejewska et al., 2013)], there have been growing efforts to correlate SXRF microscopy maps with the corresponding MRI data from post-mortem human tissue (Collingwood et al., 2008a; Hopp et al., 2010; Antharam et al., 2012). This has included studies in formalin-fixed tissue to achieve approximate correlations between susceptibility-weighted imaging (SWI) and iron maps for multiple anatomical regions in each case via the rapid SXRF approach (Hopp et al., 2010), and for the first time in unfixed tissues, working at high spatial resolution in thin sections to match SXRF iron maps with R2 and R2 ∗ maps, permitting detailed analysis of the hippocampus in AD cases and age-matched controls and demonstration of a positive correlation between iron and R2, R2 ∗ (Antharam et al., 2012).

## **INVESTIGATING EVIDENCE OF OXIDATIVE STRESS IN NEURODEGENERATIVE DISEASE WITH SXRF**

The role of iron in neurodegenerative disorders is often considered in the framework of oxidative stress (Kell, 2010; Smith et al., 2010). Analysis of selenium is important in this context, as it is an essential element found in the cofactor for glutathione peroxidase, which is involved in regulating oxygen free radicals. Selenium is likely very important in protecting against free radical damage from poorly liganded iron in neurodegenerative disorders. SXRF is especially useful for sensitive and specific analysis of selenium in tissue (Schulmann-Choron et al., 2000), where Schulmann-Choron and colleagues obtained SXRF spectra from acid-digested rat brain tissue, and reported a minimum detection limit (MDL) of 20 ppb for selenium, using the full beam (focused to approximately 1 mm2), and counting for 500 s per acquisition. Ince et al. (1994) had reported elevated iron and selenium in lumbar spinal cord from MND cases compared with controls using NAA, and in spatial SXRF analysis of SN and spinal cord in PD and ALS cases, selenium was selectively observed in the neuronal bodies of the SN (Szczerbowska-Boruchowska et al., 2004). Subsequently, the SXRF investigation of neuromelanin in normal SN led to the observation that selenium concentration in neuromelanin appears to increase with age, and may indicate increased requirements for protection against oxidative stress as a function of aging (Bohic et al., 2008).

## **XAS: LOOKING AT THE OXIDATION STATE OF IRON AND ITS LOCAL ENVIRONMENT**

Determining absolute concentrations of tissue iron, and the proportion of iron to which MRI sequences are sensitive, is important to help with recognizing disruptions to iron homeostasis and to identify potential clinical markers of disease. However, the questions can become more subtle when we consider underlying mechanisms of iron-mediated toxicity. Some instances of tissue iron overload may be a consequence of impaired bioavailability (Rouault, 2013; Visanji et al., 2013), and under certain circumstances the colocalization of iron and species with, for example, reductase potential like Aβ<sup>42</sup> (Khan et al., 2006) or α-synuclein (Davies et al., 2011), may be a more significant contributing factor to oxidative stress damage than tissue iron concentration in its own right (Gallagher et al., 2012; Everett et al., 2014a). As SXRF may be used non-destructively, there is plenty of scope to correlate SXRF iron maps with other parameters relevant to iron-mediated toxicity, and to probe the oxidation state and mineral form of iron within SXRF maps by techniques such as XANES (Collingwood et al., 2005a; Bacquart et al., 2007).

Two decades ago, De Stasio et al. (1995) used X-ray secondaryemission microscopy (XSEM) to spatially evaluate the chemical distribution of iron and other metals in isolated rat cerebellar granule cells after they had been exposed to iron in solution. The iron absorption spectrum was measured between 50 and 60 eV, and a difference image at the iron absorption edge between the 55 and 54 eV energies was calculated to gain a map of iron distribution. The spatial distribution of the elemental analysis was 0.5 μm or better. This particular experiment, perhaps due to some limitation in the protocol, did not enable the association of iron with particular cells or cellular structures to be identified; rather, iron was found throughout the specimens. However, in principle the absorption profile can contain a great deal of information about the local chemical environment of the absorbing species, and below we consider how XAS, specifically EXAFS and XANES, have been applied to the study of iron in the brain.

Ascone and Strange have reviewed in detail the contribution of XAS to the analysis of metalloproteins (Ascone and Strange, 2009), using the descriptor "*metalloproteomics*" defined as the "*structural and functional characterization of metal-binding proteins.*" They acknowledge that XAS in biological materials requires a synchrotron X-ray source to achieve the necessary flux, stability of the beam and associated optics, and the ultrasensitive fluorescence detectors, along with a variety of highly regulated sample environments. Concerning XANES of biological materials, or "bio-XANES," the importance of the pre-edge spectrum, the developments in gaining quantiative structural information from bio-XANES, and some limitations of "fingerprinting" with XANES are discussed. XANES and EXAFS continue to be excellent tools for analysis of the electronic structure for metal elements in biological materials, and with an expanding cross-disciplinary user base, there have been improvements in both the sophistication and accessibility of purposedesigned (and open source) software tools for analysis of the spectra (Ascone and Strange, 2009). More complete studies can be undertaken by combining synchrotron methods such as EXAFS and X-ray diffraction for structural determination of iron sites in multi-metal-bearing proteins (Einsle et al., 2007) for metalloproteins such as ceruloplasmin (Ascone and Strange, 2009), or to evaluate trafficking of metal ions between proteins. Improvements in beamline instrumentation continue to improve technical scope and measurement efficiency, and rapid acqusition of high-quality XANES has enabled XANES "mapping" at some beamlines with XANES obtained at each pixel in the sample map.

## **X-RAY ABSORPTION SPECTROSCOPY TO STUDY IRON AND NEUROMELANIN IN THE SUBSTANTIA NIGRA**

Many experiments have now been performed to compare iron in the SN in the healthy and PD brain. These experiments have primarily been used to image or probe the distribution of iron in tissue sections by various qualitative and quantitative methods, or iron concentrations have been determined for bulk samples. Spatial analysis of iron in unfixed tissues of the SN has shown some evidence for site-specific elevation of iron in PD dopaminergic neurons compared to healthy controls (Szczerbowska-Boruchowska et al., 2004; Oakley et al., 2007), but elevated iron concentration alone does not provide information about the chemical form of the iron, or indicate whether excess iron is bound in a state that differs from the iron normally present. Here, we review XAS studies that have attempted to address this question in the context of PD.

The dopaminergic neurons of the SN are selectively vulnerable in PD and confirmed to accumulate iron in the disease state (Dexter et al., 1989; Oakley et al., 2007; Crichton et al., 2011). These neurons are normally pigmented with neuromelanin, which has a strong affinity for metal ions including iron (Double et al., 2003). X-ray microanalysis was performed on an electron microscope to determine Fe2<sup>+</sup> and Fe3<sup>+</sup> associated with various forms of neuromelanin, and from this it was deduced that the primary form of iron in dopaminergic neurons is neuromelanin-bound iron, and that neuromelanin preferentially binds Fe3<sup>+</sup> (Jellinger et al., 1992). X-ray analysis at various synchrotron facilities has subsequently been used to further explore the nature of iron storage in neuromelanin, which is considered important to understand if we are to fully explain the contribution of iron-mediated toxicity in PD.

Kropf et al. (1998) used EXAFS to look at the structure of human neuromelanin and its analogs. They observed that both human and synthetic neuromelanin have a common iron center within a sixfold distorted oxygen octahedron, with a distance ∼ 2 Å. Earlier studies with Mössbauer spectroscopy had reported evidence for superparamagnetic behavior in both synthetic neuromelanin and purified human neuromelanin from the SN; in the latter study, the parameters were observed to be similar to those found for hemosiderin and ferritin (Gerlach et al., 1995). Kropf et al. (1998) investigated the possibility that there are two binding sites in neuromelanin, by looking with EXAFS at the nearneighborhood environment of iron in both extracted human and synthetic neuromelanin. The study included samples that were either fully or partially (30%) saturated with iron, where for the latter the most easily chelatable iron had been removed. As the partially saturated human neuromelanin spectrum was virtually indistinguishable from the saturated version, it was concluded that there were no differences in the affinities of the binding sites (**Figure 3**). Melanin aggregation had formerly been described using a fractal construct, implying a sponge-like structure, and Kropf and colleagues reported their EXAFS observations as being consistent with this model, suggesting that instead of two iron binding sites, iron would simply be removed more rapidly from outer than inner sites for a neuromelanin particle. The structural information from the EXAFS indicated that a superparamagnetic structure would be unlikely. Mössbauer spectroscopy was then

**X-ray absorption spectroscopy analysis of iron.** Panel **(A)** shows EXAFS spectra of iron bound to synthetic, natural 100% saturated, and natural 30% saturated (depleted) neuromelanin, where the traces appear similar. Panel **(B)** shows the processed EXAFS data reveal marked differences between the synthetic and natural neuromelanin. Panels **(C** ,**D)** show XMCD spectra collected after 144 h for **(C)** ferrihydrite, and **(D)** ferrihydrite incubated with Aβ42, where the positive and negative peaks (A,B,C,D) indicate an antiferromagnetic structure, and the dramatically increased amplitude of peaks (A,B) in panel **(D)** is consistent with a significant increase in the proportion of Fe 2 + in the antiferromagnetically ordered mineral. Panels **(A,B)** were published in Kropf et al. (1998), Copyright Elsevier (1998). Panels **(C,D)** are adapted with permission from Everett et al. (2014a). Copyright 2014 American Chemical Society.

used to demonstrate that the neuromelanin exhibited only paramagnetic behavior (Kropf et al., 1998), and it was concluded that the iron in neuromelanin does not follow the ferritin model in this regard.

In a parallel study, EXAFS was used to analyze iron in whole tissue samples that had been gently homogenized prior to analysis (Griffiths et al., 1999). The study included samples from the SN and GP regions in a group of PD cases and healthy control brains, with *n* = 6 in each group. For the tissue, it was concluded that ferritin-like parameters provided the most appropriate fit for both anatomical regions. This result for SN is not necessarily in contradiction of the former neuromelanin analysis (Kropf et al., 1998), which was performed on isolated neuromelanin and would not have included iron from other sources such as astrocytes, oligodendrocytes, and activated microglia in the vicinity of extracellular neuromelanin. Indeed, it has been suggested that neuromelanin-bound iron would not exceed ∼15% iron in the SN (Galazka-Friedman et al., 1996). Given the heterogeneous nature of iron and neuromelanin distribution throughout the subfields and cell types in the SN (Morris and Edwardson, 1994), the exact region dissected for analysis will influence the proportion of tissue iron that is neuromelanin-bound; it has also been suggested that the strong affinity of neuromelanin for iron leads to additional iron from other sources (such as glia) binding neuromelanin during the extraction and purification process, as neuromelanin is not normally saturated (Double et al., 2003). Based on the number of parameters required to fit clusters of electron-dense cores in these same samples (observed by electron microscopy), it was concluded that a greater degree of sub-cellular clustering occurred within PD tissue than in controls (Griffiths et al., 1999). While this was interpreted as a change in sub-cellular clustering of ferritin, it is possible that for SN this observation included neuromelanin clusters, consistent with former histopathological observations and a concurrent study from Ektessabi et al. (1999) using SXRF imaging of PD and control SN where they noted "*various sizes of melanin granules released from dying nigral neurons scattered in a more condensed form (neuromelanin aggregates) than those in the nigral neurons of the control subject.* "

Ektessabi et al. (1999) had originally been unable to obtain useful XANES data from the formalin-fixed SN tissues that they mapped by SXRF, but at the 39XU Spring-8 Japanese synchrotron they mapped a small area at 7.2 keV, just above the iron K-edge, prior to obtaining XANES in the fluorescence configuration with

a beam spot limited with a pinhole to approximately 10 μm in the sample plane (Yoshida et al., 2001). Reference standards of FeO and Fe2O3 were used in order to determine the position of the Fe2<sup>+</sup> and Fe3<sup>+</sup> edges, respectively, and to thereby compare relative proportions of Fe2<sup>+</sup> and Fe3<sup>+</sup> at sites of interest throughout the SN. It is reported that with a 0.5 eV energy step, that dwell times ranged from 20 to 100 s per step (Yoshida et al., 2001). While comparisons between neuromelanin-bound iron in intact and dead neurons were made (which suggested an increased fraction of Fe3<sup>+</sup> in neuromelanin associated with degenerating neurons), the use of a single case, and the prior fixation and paraffin embedding of the tissue means that this study is more significant as a demonstration of potential for iron XANES from individual cells in sections of human brain tissue.

MicroXANES in unfixed SN tissue was performed by Chwiej et al. (2007), where a single PD case and six controls were compared in freeze-dried tissue. Regions of pigmented neuromelanin, consistent with those observed in dopaminergic neurons, were selected for XANES analysis, and no significant difference between the PD and control samples was detected. In contrast with the prior XANES reported from fixed tissue (Yoshida et al., 2001), the only detectable chemical state of the iron in these unfixed tissues was Fe3<sup>+</sup> (Chwiej et al., 2007). This observation, while it contradicts prior reports of Fe2<sup>+</sup> in normal PD SN, is strongly supported by an earlier Mössbauer analysis of SN from PD and control cases, where only Fe3<sup>+</sup> was detected in bulk tissue samples (Galazka-Friedman et al., 1996).

The XANES study by Baquart el al, analysing PC12 cells as a model system for the SN (Bacquart et al., 2007), is especially thorough in assessing the impact of sample preparation and beam exposure on the scope for radiation damage in the form of mass loss (Williams et al., 1993) and photoreduction (Yano et al., 2005). This XANES study was concurrent with members from the same team reporting high resolution SXRF analysis of iron and dopamine vesicles in the same cell model (Ortega et al., 2007). XANES was selected for its scope for measurement with minimal preparation of the cells, avoiding processes that might modify the chemical species. Beam focus of 1.5 μm × 4 μm was achieved with K–B mirrors, and iron and arsenic were analyzed by XANES captured in fluorescence mode (which is optimal for dilute specimens). Regions sampled within the PC12 cells included cytosol, the mitochondrial network, and nucleus. It was observed that cells measured in the frozen hydrated state (maintained at –100◦C, in a liquid nitrogen cryostream) were generally preferable to freeze-dried cells at room temperature, as this increased repeatability, sensitivity, led to no noticable shift in oxidation state, and minimized beam damage. Interestingly, however, it was observed that for iron, any change in chemical state between the frozen hydrated and freeze dried cells was insignificant in the XANES, and that the technically more straightforward method of working with the freeze dried cells at room temperature, facilitating observation with the videomicroscope during measurement, made this the preferred option (Bacquart et al., 2007). To achieve good signal to noise, with minimal beam damage to the sample, it was noted that short repeated energy scans were preferable to a single scan with longer dwell times at each energy. Iron XANES at the K-edge (7.112 keV) were

obtained over the energy range 7.037–7.200 keV, collecting for 5 s/step, with a step size of 1.0 eV pre- and post-edge, and 0.5 eV through the edge region (7.082–7.142 keV), giving a scan time of approximately 18 min. (Scan times longer than this led to significant signal decay for XANES of arsenic in the freeze dried cells at room temperature, indicating mass loss of this element. Mass loss was tracked by decreasing white line intensity; notably, the dose observed to cause mass loss in the hard X-ray region at room temperature did not have the same effect with soft Xrays at 113 eV, or under cryoconditions, where both protected against mass loss. Minimizing the dose was also important to reduce scope for photoreduction by the X-rays.) The microfocussed achromatic beam was very stable; beam movement on the sample corresponded to a displacement of approximately 0.1 μm over 100 eV. While the short XANES scans could be repeated and summed to improve signal quality without causing detectable sample damage, the iron concentration in the PC12 cells was sufficient that good XANES data were acquired in a single acquisition. The "limit of speciation" for elements in the atomic mass range of 20–40 was judged to be approximately 13 μg/g (Bacquart et al., 2007). Progress in synchrotron technology is resulting in samples being exposed to higher photon flux, so the conditions required for sample preservation will require ongoing evaluation (George et al., 2012).

## **X-RAY ABSORPTION SPECTROSCOPY TO STUDY IRON IN FRIEDREICH'S ATAXIA**

In order to study mitochondrial iron chemistry in the context of Friedreich's Ataxia, Popescu et al. (2007) analyzed the heavy mitochondrial fraction isolated from primary fibroblasts taken from individuals with and without Friedreich's Ataxia, and analyzed them by XANES at the iron K-edge. In order to determine the mineral state of the iron, they fitted the spectra with a library of 22 iron compounds, and concluded from the fitting results (with supporting data from Western blotting) that the mineralized iron in the Friedreich's Ataxia patients is more highly organized than in the unaffected individuals, and that the Friedreich's Ataxia patients mineralize a significant fraction of cellular iron in mitochondrial ferritin (MtFt; Popescu et al., 2007). MtFt has been postulated as having a protective role for mitochondria in cells which have high levels of metabolic activity and oxygen consumption, rather than being associated with iron storage (Levi et al., 2001), and since Popescu's XANES analysis, this potentially protective role has been explored specifically in the context of Friedreich's Ataxia (Campanella et al., 2009).

## **COMBINING SYNCHROTRON X-RAY FLUORESCENCE MAPPING AND ABSORPTION SPECTROSCOPY**

As the microfocus SXRF and XANES techniques mature, there has been increased recognition of their complementarity, and many microfocus beamlines are now well-equipped to measure both during the course of an experiment (**Figure 1**). Practical measures such as simultaneous acquisition of CCD data in transmission mode to determine crystalline parameters of biominerals, and introducing custom lithographic finder grids to aid location and retrospective analysis of sites within a sample, have previously been proposed (Collingwood et al., 2005b).

#### **INVESTIGATING IRON BIOMINERALIZATION IN ALZHEIMER'S DISEASE**

One of the early applications of the combined SXRF and XANES analysis was to demonstrate how isolated microscale iron focii in large (≥1 cm2) areas of tissue could be efficiently identified and characterized in a way that is impractical or impossible with other microprobe techniques. The motivation for this work was the detection of tiny deposits of magnetite, a mixed valence iron oxide, in human brain tissue, via analysis of bulk tissue samples and isolated material (Kirschvink et al., 1992; Dunn et al., 1995). There were early indications that levels of magnetite were elevated in AD tissue compared for healthy controls (Hautot et al., 2003), although the extent of this, and the mechanism for its formation *in vivo*, was relatively unexplored. Our preliminary work with avian tissue (Mikhaylova et al.,2005), known to contain particulate magnetite, established the protocols for the first demonstration of magnetite *in situ* in human brain, using AD amyloid-plaquerich tissue from the superior frontal gyrus (Collingwood et al., 2005a). SXRF mapping at resolutions ranging from 100 to 5 μm were used to detect and precisely locate iron focii, and XANES analysis was performed with the 5 μm beam spot to determine the forms of iron present at the sites, which primarily included ferritin-like ferrihydrite, and/or magnetite. The combination of SXRF and XANES has since been used to study the amyloidβ precursor protein (AβPP)/presenilin 1 (PS1) mouse model of AD (Gallagher et al., 2012). This model is engineered to overexpress the Swedish mutation of human AβPP, and mutant human PS1, and is known to exhibit amyloid deposition in the brain from approximately 6 months. SXRF and XANES were used to demonstrate magnetite deposits in amyloid-rich regions of the AβPP/PS1 transgenic mouse brain (Gallagher et al., 2012), and in this study we suggested that the magnetite deposits are an indication that iron dysregulation is an early event in AD-related pathology. Subsequently, in work led by the Telling group, we suggest a mechanism, demonstrated *in vitro* by synchrotron methods including XAS and X-ray magnetic dichroism (XMCD), by which the reductase behavior of Aβ<sup>42</sup> may lead to the observation of magnetite in these plaque-rich tissues (Everett et al., 2014a).

#### **UNDERSTANDING NEUROMELANIN**

Another area in which the combination of SXRF and XANES has been utilized, is in the study of neuromelanin in human tissues (Bohic et al., 2008). SN tissue was obtained from formalin-fixed paraffin embedded samples from seven human brains ranging in age from 24 weeks to late adulthood, and the cases were selected on the basis of having no significant neuropathology, SN related pathology, or dopamine-related disorders. Pigmented neurons were SXRF mapped, and then XANES obtained at the K-edge in fluorescence mode in 1 eV increments over the range 7.07–7.37 keV. The authors report increased levels of cellular neuromelanin, followed by a darkening of the pigment, as a function of aging. From the trace metal mapping it was established at a high level of spatial resolution that iron and selenium are closely associated with neuromelanin from an early stage of development, and that other metals such as calcium, copper, and zinc become associated with it at a later stage. The levels of neuromelanin-bound iron in the tissue were below the expected saturation value for neuromelanin, consistent with prior work (Double et al., 2003). The intensity of the SXRF spectra indicate that neuromelanin-bound iron increases as a function of age, although the microXANES spectra were consistent with ferritin regardless of age. It should be noted that the chemical form of iron in the SN was shown by Mössbauer to be demonstrably different in formalin-fixed archive samples compared for fresh-frozen (Galazka-Friedman et al., 1996), but here the XANES conclusions are generally consistent with those previously obtained in unfixed sections (Chwiej et al., 2007). Bohic et al. (2008) conclude, contrary to prior EXAFS observations in extracted human neuromelanin (Kropf et al., 1998), that neuromelanin is likely to have variability in its metal binding domains that suggest different functional roles. There is, however, unanimous agreement between all four studies cited here that the iron bound to neuromelanin is in the Fe3<sup>+</sup> form within the limits of experimental error. A subsequent study of extracted neuromelanin granules was conducted by Tribl et al. (2009), where a variety of proteins were shown to be associated with neuromelanin, including L-ferritin. This observation may in due course contribute to explaining the multiple binding affinities reported for neuromelanin, and arguably there is an opportunity here to combine immunohistochemistry with sub-micron SXRF and XANES to determine if and how L-ferritin colocalises with neuromelanin in intact pigmented cells.

## **IRON-BINDING PROTEINS AND AGGREGATION**

The main emphasis of the previous sections in this review has been on the use of synchrotron techniques to locate and characterize iron, including that which is tightly coordinated to metalloproteins. Synchrotron analysis also has the scope to provide insights where metalloprotein structure and conformation is concerned, which is an essential aspect of understanding iron regulation (and by extension dysregulation) in neurodegenerative disorders.

Synchrotron techniques have been used for analysis of a variety of iron-binding metalloproteins, including transferrin, hemoglobin, and ferritin, where the latter has been investigated in many synchrotron experiments to further understand both the protein structure, and the nature of the iron oxide core formed within (Kim et al., 2011). The powerful combination of high resolution X-ray crystallography and EXAFS has been especially useful in determining the structure of metalloproteins (Hasnain, 2004; Duke and Johnson, 2010). One of the advantages of synchrotron X-ray crystallography is the rapidity with which diffraction patterns can be obtained from highly complex structures. Several groups have used time-resolved Laue diffraction studies to investigate structure and conformational properties of metalloproteins (Duke and Johnson, 2010). For example, Bourgeois et al. (2003) reported the application of nanosecond Laue crystallography to determine protein structure using myoglobin as model system. This study with 150 ps X-ray pulses enabled room temperature observation of the dynamics of protein conformational changes.

#### **SMALL ANGLE X-RAY SCATTERING**

A widely used synchrotron technique for the conformational study of metalloproteins is small angle X-ray scattering (SAXS), which has been used to look at proteins intimately involved in the uptake and transport of iron. As with crystallography, time resolution can be a useful feature in SAXS measurements, enabling transient aggregates to be distinguished from folding intermediates (Segel et al., 1999). Castellano et al. (1993) used SAXS to look at the conformational changes in aggregates of human transferrin. They looked specifically at transferrin from human serum (serotransferrin), an 80 kDa metal binding protein which is structurally modified on binding to metal ions, and which has sites for two Fe ions. This investigation was motivated by a desire to understand the potential role of transferrin in binding Cu and Al in addition to Fe. Castellano's study included observations of apotransferrin and monoferric transferrin, and they concluded that the conformational changes brought about by the iron binding process were consistent with prior observations made with both X-ray and neutron scattering methods. The transferrin was considered in a fractal framework, permitting particle and cluster sizes to be distinguished (Castellano et al., 1993). More recently, it has been observed that transferrin can form fibrils under certain conditions *in vitro*, and this led to an investigation to determine whether transferrin has amyloid-like properties. Synchrotron X-ray circular dichroism (XCD) was used to test whether transferrin in solution has amyloid-like properties, and this helped show that transferrin aggregation in solution does not appear to involve major structural changes to the protein, or formation of beta-pleated sheets (Booyjzsen et al., 2012).

Small angle X-ray scattering is ideal for studying protein (mis)folding and aggregation problems, which are a common feature in neurodegenerative disorders. In turn, iron and other metal elements are associated with modifying or exacerbating protein aggregation in the majority, if not all, of these disorders (Kell, 2010). SAXS permits the size and shape of soluble aggregates to be characterized, and can provide information about the order of aggregation (i.e., whether it is a dimer, trimer, or higher order). One protein that has been studied in detail by SAXS is α-synuclein, which forms insoluble fibrils in neurodegenerative disorders classed as the "synucleinopathies," including PD, dementia with Lewy bodies (DLB), and MSA. Uversky et al. (2002, 2005) conducted a series of studies incorporating SAXS, which investigated the conformation, shape, and association of the α-synucleins in solution. Typical experiments were performed at room temperature, and involved the α-synuclein passing through a flow cell along a 1.3 mm path, with 25 μm mica windows (Li et al., 2001). SAXS permitted wild-type, and familial PD point mutations (A30P and A53T) to be distinguished, determining the radius of gyration, *R*g, and the confirmation and globularity (indicating packing density) of the protein. It was observed that the wild-type and mutated forms of α-synuclein have identical *R*<sup>g</sup> of 40 Å at neutral pH, and that this value decreases as the pH is lowered, indicating compacting of the protein giving rise to a reduction in volume. The homogeneity of the proteins, the absence of aggregation under these conditions, and the configuration approximating to a random coil at neutral pH is also demonstrated through the SAXS data (Li et al., 2001). Subsequent studies considered in more detail the monomeric and fibrillar forms of α-synuclein, including exploring factors that promote and inhibit aggregation (Uversky et al., 2002, 2005). Recent advances in time-resolved XAS include work from Lima et al.

(2011)to enable MHz-rate data acquisition with picosecond lasers to enable studies at physiologically relevant concentrations for biological systems.

## **SYNCHROTRON FOURIER-TRANSFORM INFRA-RED ANALYSIS**

To study proteins such as α-synuclein, Aβ, and prion protein in intact tissues, another technique may be used: FTIR spectroscopy, where the infrared spectrum is obtained from a given material. FTIR is routinely performed using laboratory sources, but for certain measurement configurations, synchrotron light sources have capacity to achieve brightness that is orders of magnitude brighter than standard sources. A detailed review of the principles of the instrumentation, and of approaches to biological specimen preparation for FTIR, is provided by Miller and Dumas (2006), who consider FTIR micospectroscopy (FTIRM), and imaging (FTIRI). Where the beam spot size at the sample is determined by a pinhole, synchrotron FTIR is especially advantageous for microscale analysis with resolution ∼ 10 μm. The spatial resolution of FTIRM is limited by the wavelength of the infrared, from approproximately 1.7 μm at 4000 cm−<sup>1</sup> to 13 μm at 500 cm−1. Spectra from biological samples are obtained in transmission mode with thin samples (typically 5–30 μm), or in reflection mode for highly reflective or unsectionable samples (Miller and Dumas, 2006).

Fourier transform infrared may be used to study protein folding dynamics on the microsecond time scale, at good spatial resolution and with a small volume of sample. While slower than laser techniques, the white beam ensures the complete spectra can be obtained (Miller and Dumas, 2006). However, one of the most interesting applications of FTIR is the subcellular chemical mapping that can be achieved, which is highly complementary to the analysis of metal and mineral structures outlined previously in this review. FTIR gives insight into the nucleic acid, protein, and lipid content of particular structures, and does not significantly heat the sample, so that prolonged studies of individual living cells may be performed. Multimodal imaging, such as combining SXRF and FTIRM with epifluorescence microscopy to look at trace metals and AD senile plaques (Miller et al., 2006), can be done with fluorescent tag concentrations at levels that do not interfere with the infrared, and there is growing interest in combining synchrotron FTIR, SXRF, and XAS (Miller et al., 2006; Kastyak et al., 2010). For example, Kastyak et al. (2010) used synchrotron FTIR to analyze crystalline deposits of creatine detected in SXRF-mapped unfixed tissue from spinal cord, brain stem, and motor neuron cortex in cases of ALS. Intense focii ("hot spots") of calcium, zinc, iron, and copper were also noted in the ALS tissues. Miller et al. (2006) initially demonstrated the combination of SXRF microfocus mapping and FTIR to unambiguously associate iron, copper, and zinc with Aβ deposits in unfixed air-dried human brain tissue from AD cases. FTIRM and SXRF were mapped with a resolution of 5–10 μm, and the Amide I absorbance was used to confirm the presence of Aβ. Tissue sections were mounted on quartz after the slides had been coated with a 200 nm sputtered coating of aluminum. The latter was necessary for infrared reflectivity, but the spatially heterogeneous trace contamination of iron in even this thin aluminum layer place some constraints on the tissue iron analysis that could be performed in this study. The copper and zinc distributions (as demonstrated by SXRF) showed good

correlation with regions containing Aβ deposits (as demonstrated by FTIR). In a subsequent study, using Ultralene® as an ultra-clean SXRF and FTIR compatible substrate for the tissue samples, the correlation between transition metal ions and amyloid deposition was investigated for the AβPP/PS1 mouse model of AD (Leskovjan et al., 2011). Sections were cut at 30 μm thickness, air-dried, and kept in a dessicator prior to imaging. Approximate quantification of elemental concentrations was achieved using thin film standard reference materials. Amyloid was visualized with a simple Thioflavin-S staining protocol, where they observed no change in the iron, copper, or zinc content of the map subsequent to staining, and FTIRM was used (in transmission mode) to determine tissue protein density using a spatial mapping resolution of 4 μm. Normalization of metal ion content to protein content for each selected plaque was used to determine whether iron, copper, and zinc accumulate in large dense areas of amyloid deposition in the AβPP/PS1 model of AD; it should be noted that only large plaques (30–50 μm) were included in the quantitative correlations, to ensure they occupied the full thickness of the section. Analysis of regional iron, copper, and zinc distribution was achieved by a combination of standard segmentation (e.g., of the hippocampus) and by cluster analysis. Interestingly, with SXRF microscopy, an elevation (∼1/3) of cortical iron was observed in the AD model compared to the wild type, which is consistent with observations from Smith et al. (2010) in human AD and the iron was not directly associated with the large regions of amyloid deposition that were included in the analysis. Zinc, however, became associated with the amyloid deposits at the endstage of the study (56 weeks), having not demonstrated elevation at 24 and 40 weeks. Leskovjan et al. (2011) acknowledge that there may possibly be a relationship between brain iron and the amyloid pathology, but given that iron was not specifically elevated at the amyloid-plaque sites, when normalized to protein content, there was clearly some doubt. Their study highlights an important question about the manifestation of iron dysregulation in AD: does tissue iron have to be significantly elevated above normal concentrations in order for iron to contribute to pathophysiological processes?

## **DISCUSSION**

## **DOES IRON-MEDIATED TOXICITY DEPEND ON IRON CONCENTRATION?**

Historically, attention has been paid to brain cells, tissues, and even whole brain regions exhibiting marked changes in iron concentration. These indicate candidate sites vulnerable to iron-mediated toxicity, but analyses that focus purely on iron concentration in tissue have been challenged (Friedman et al., 2009), and lead in many cases to conflicting observations (Friedman et al., 2009; Schrag et al., 2011), and the contribution of iron overload to pathology is unclear (Rouault, 2013). Kell considers the problem from the perspective of poorly liganded iron (Kell, 2010), placing emphasis on the local chemical environment, and the availability of iron to participate in reactions that stimulate the overproduction of damaging radical species. We suggest here that significantly elevated concentrations of iron are not a prerequisit for this scenario.

We will take Aβ aggregation as a case study, where the interactions between Aβ<sup>42</sup> and iron *in vitro*, and the relationship between Aβ and iron deposition in humans and AβPP mouse models, have been characterized by synchrotron methods (Collingwood et al., 2005a; Gallagher et al., 2012; Everett et al., 2014b). Many studies have demonstrated that iron is associated with insoluble deposits of Aβ in human brain, even in the pre-clinical stage of AD (Smith et al., 2010), and it is understood that iron provides much of the natural contrast that makes amyloid plaques observable in high resolution MRI (Meadowcroft et al., 2009; Petiet et al., 2011). However, observed accumulation of iron in amyloid plaques is reportedly lower in mouse models of AD than in human tissue (Leskovjan et al., 2009; Meadowcroft et al., 2009), and tissue iron concentration has been shown not to correlate with plaque burden in human AD cases (House et al., 2008). This observation was reinforced in study of 60 aged human brains, where it was demonstrated that there is no relationship between tissue iron concentration and congophilic amyloid angiopathy or senile plaque burden (Exley et al., 2012). The formation of spherulites from Aβ<sup>42</sup> has been demonstrated in the presence of copper (House et al., 2009), and it is postulated that these structures, which form *in vitro* and are also observed in human brain tissue, may correspond to the senile plaques with fibrillar structure routinely observed with transmission electron microscopy. SXRF microfocus analysis of the spherulites in unfixed human AD hippocampus revealed direct association of the spherulites with copper, not iron, despite iron being abundant in the tissue (Exley et al., 2010). Subsequent Perls staining in formalin-fixed human hippocampus from another individual also showed no direct positive correlation between iron and spherulite distribution (House et al., 2011). In our SXRF analysis of AD and control hippocampus, where microfocus-resolution iron maps were correlated with MRI microscopy, again no disease-dependent change in total regional iron level was observed, although the AD tissue exhibited a higher proportion of R2 and R2 ∗ hyperintensity artifacts consistent with iron-rich deposits within the hippocampus (Antharam et al., 2012). It is therefore necessary to consider not only concentration of iron and its co-localization (or otherwise) with Aβ, but also the chemical and bound state of iron in tissues exhibiting amyloid aggregation. Synchrotron facilities provide exceptionally sensitive and unambiguous techniques for this purpose. Our study with SXRF and XANES to investigate the mineral form of iron deposits in human AD plaque-rich cortical tissue demonstrated the presence of ferrihydrite (Fe3+, consistent with normal ferritin cores) and magnetite (alternating lattices of Fe2<sup>+</sup> and Fe3+) within the tissue (Collingwood et al., 2005a); an observation subsequently supported by the detailed characterization of ferritin-core-sized magnetite/maghemite iron oxide particulates in extracted amyloid plaque core materials (Collingwood et al., 2008b). The subsequent SXRF and XANES analysis in the AβPP/PS1 model produced essentially the same observation concerning magnetite formation, despite quantitative analysis of the same brains showing no overall difference in brain iron concentration between wild-type and the AD mouse model (Gallagher et al., 2012).

What is the origin of the mixed valence oxides observed in the human and mouse model tissues, and is their association with regions exhibiting Aβ deposition a coincidence? The reductase behavior of Aβ<sup>42</sup> peptide, a primary component of amyloid pathology in AD, has been demonstrated *in vitro*, where it has the capacity to reduce iron from its more stable ferric (Fe3+) form to the more reactive bio-available ferrous (Fe2+) form at physiological pH (Khan et al., 2006; Everett et al., 2014b). Using a combination of XAS (using the total electron yield method), and XMCD by obtaining two XAS spectra with opposed magnetic fields oriented in the direction of the X-ray beam (**Figure 3**), we have now shown that Aβ<sup>42</sup> has the capacity to chemically reduce iron in ferrihydrite particles (Everett et al., 2014a), and the results from this *in vitro* study support the possibility that the Fe2<sup>+</sup> oxide that forms in the presence of Aβ1−<sup>42</sup> is a precursor to magnetite formation (Everett et al., 2014a). These results are further supported by chemical imaging using STXM (**Figure 4**). Cellular ferrireductase behavior by α-synuclein, a peptide which

is a key constituent of protein aggregates in PD, DLB, and MSA, has also been observed (Davies et al., 2011). These results indicate that co-localization of trace levels of labile or poorly liganded iron with aggregating Aβ1−<sup>42</sup> or α-synuclein has the potential to lead to significant overproduction of radical species in various neurodegenerative disorders.

Taken together, these synchrotron observations by FTIR and SXRF (Leskovjan et al., 2011), SXRF and XANES (Collingwood et al., 2005a; Antharam et al., 2012; Gallagher et al., 2012), and XANES and XMCD (Everett et al., 2014a,b), support the suggestion that mineralized iron deposits may prove a useful clinical marker of amyloid-mediated iron deposition (Meadowcroft et al., 2009; Petiet et al., 2011; Antharam et al., 2012), and they provide strong support for the hypothesis that interaction mechanisms

**FIGURE 4 | Illustration of STXM chemical mapping of iron associated with aggregated Aβ<sup>4</sup>** <sup>2</sup>**.** Panel **(A)** shows bright-field transmission electron microscopy (TEM) images and panel **(B)** shows scanning transmission X-ray microscopy (STXM) images of fibrillar amyloid structures formed following the incubation of Aβ42 with ferrihydrite. TEM pictures show Aβ structures present

after 0.5, 48, and 144 h of incubation. STXM images show the carbon (C) and iron (Fe) content of an Aβ/ferrihydrite aggregate, along with a carbon/iron (C + Fe) composite image of the same aggregate. Reprinted (adapted) with permission from Everett et al. (2014a). Copyright 2014 American Chemical Society.

between iron and aggregating Aβ, rather than necessarily measures of total iron concentration or co-localization of iron and amyloid, are critical to our understanding of Aβ-mediated toxicity in AD.

### **APPROACHES TO STUDY DESIGN AND SAMPLE PREPARATION**

As technical developments at synchrotron facilities have enabled and encouraged a growth in metallomics research over the past twenty years, some excellent progress has been made in understanding aspects of iron distribution and storage in the brain. However, this is a highly multidisciplinary area, requiring understanding of the biology, chemistry, and physics of the sample preparation, measurement techniques, and research question, in addition to appreciation of the neuroscience and (patho)physiology. Our aim in writing this review has been not only to showcase synchrotron insights into iron in neurodegenerative disorders, but also to provide a resource, through the cited literature, for investigators to draw upon when they are designing future synchrotron experiments.

Investigators often work with limited quantities of donated tissue specimens, with restricted information about the archive conditions or control over sample preparation. It is practical to work with small volumes of material because a high degree of sensitivity can be achieved with synchrotron experiments, but it is important to ensure that samples are properly selected, characterized, and matched in advance of the synchrotron experiment, and that all reasonable steps are taken to protect the integrity of the property being measured (e.g., avoiding steps that introduce contamination or alter the oxidation or mineral state of the iron being studied).

In the original XRF analysis of PD tissues by Earle, his concern that potassium may have leached from the tissue during storage (Earle, 1968) highlights a fundamental challenge in the analysis of trace metals in tissues. Sample storage and preparation may partially or completely alter the property to be analyzed, and careful experiment design is required to evaluate and minimize the scope for this. For example, fixation by prolonged immersion in a chemical solution such as formalin may be necessary to preserve components of the tissue, but it can have a detrimental effect on metal ions of interest where they are mobilized in fixative, and the extent of leaching varies depending on many factors including sample size, archival period, composition and pH of the fixing solution, and the bound states of the individual elements (ranging from tightly encapsulated within proteins or mineral deposits, to labile ions within the tissue). These factors will also determine the extent to which relevant chemical and mineral states are preserved. A consequence is that making comparisons between studies where some have used fixed tissue, and others have used fresh or frozen tissues, is not always viable. As metal ions are bound in a myriad number of ways, their propensity to migrate, to be washed out of samples, or altogether lost from tissues, can be expected to vary significantly depending on the experiment protocol, the microstructure of the tissue, and whether proportions of unbound or loosely bound metal ions are influenced by the disease condition being studied. The extent of iron leaching from archived chemically fixed tissue might be expected to depend on how strongly iron is bound in the tissue, raising the interesting question as to whether iron in healthy

and diseased brains is equally prone to loss. We anticipate that the loosely bound iron implicated in neurodegenerative disorders (Kell, 2010) will be disproportionately lost in chemically fixed tissues.

Multi-modal imaging is especially valuable in addressing complex questions, and the sensitive site-specific analysis that can be performed in tissues with synchrotron techniques is of most value when observations can be directly related to broader histological findings in the same tissue. The non-destructive nature of complementary methods such as FTIR and SXRF are a great advantage, permitting multi-modal and repeated/multi-resolution analysis of sample regions, and one of the great advantages of synchrotron studies is the minimal sample preparation and handling required for most experiments. For example, cells can be grown directly onto the substrates to be imaged in air, and tissue may be cryosectioned and dried onto suitable support media with no further preparation other than to cover the section in a manner that complies with local health and safety guidelines.

Sample substrates are of great importance, as measurement sensitivity (such as in SXRF) precludes the use of standard microscope slides which contain iron in quantities significantly above SXRF detection limits. Slide coatings on any substrate can be problematic, as illustrated by Miller et al. (2006), unless they are of very high purity. Synthetic quartz slides are a practical option for many experiments if a rigid substrate is required. They do not permit X-ray transmission data to be obtained, but their chemical resistance facilitates subsequent histological staining. Thin polymeric films such as Ultralene® have been used for many experiments due to their low scattering background, and they are especially useful where both transmission and reflection data are required. Silicon nitride membranes are a versatile alternative, particularly for work with cultured cells, and facilitate multi-modal imaging ranging from the hard X-ray region to the infra-red (Carter et al., 2010).

## **CONCLUSION**

In this review we have explored several ways in which synchrotron techniques have contributed to our understanding of iron and iron-binding metalloproteins in neurodegenerative disorders. This has included neuromelanin in PD, Aβ in AD, and the regional, cellular, and sub-cellular distribution of iron and other elements in health and disease. There is untapped scope to exploit synchrotron techniques in this field, especially multimodal analysis where complementary methods (such as FTIR and SXRF) can obtain correlations between proteins, metabolites, and elemental distributions at high spatial resolution. Analysis of concentration change alone cannot usually explain mechnisms or sufficiently inform treatment strategies, and insights into the inorganic chemistry aspects of the "iron question" in neurodegenerative disorders may yet be provided by synchrotron techniques that are still largely the preserve of chemists and physicists, such as XMCD (Everett et al., 2014a). The majority of studies to date have not included steps to preserve sample chemistry using cryogenic or anoxic conditions, but this may be necessary to address fundamental questions in neurodegenerative disorders. Additionally, many temporal aspects of iron trafficking within cells and tissues remain to be understood. The time-resolved experiments

that can be achieved at synchrotrons may offer uniquely sensitive non-destructive opportunities to investigate real-time processes in living cells and tissues.

Synchrotron techniques offer tremendous potential to advance our understanding of iron dysmetabolism in neurodegenerative diseases, with unparalleled sensitivity, specificity, and temporal resolution making it practical to work with tiny analyte volumes, in highly dilute systems, performing unambigous speciation, and gaining a measure of many properties in a single or combined experiment at a given facility. The simplicity of the sample preparation techniques also offers excellent scope for complementary imaging. For example, SXRF has advanced our understanding of the relationship between iron and other elemental distributions in various regions of the brain in health and disease, and permitted identification of iron-rich structures in tissue so that they might subsequently be analyzed to determine oxidation states and biomineral form. Sub-cellular analysis has permitted direct investigation of the relationship between iron, other metal elements, dopamine, and neuromelanin, whereas lower-resolution spatial analysis has enabled direct correlation between iron maps and magnetic resonance imaging parameters to assist in the interpretation of clinical MRI measurement of brain iron.

Considering the work reported to date, the scale of which we have sought to capture in this review, it is evident that while certain questions have been studied in detail, some neurodegenerative disorders and/or brain regions have been largely overlooked, or only analyzed by a sub-set of the synchrotron techniques available. With this opportunity to learn more about the role of iron in neurodegenerative disorders, emphasis should be placed on experiment design, given the limited access to synchrotrons and to donations of human tissue for post-mortem analysis. This will ensure that with the burgeoning opportunities and growing interest in synchrotron research, carefully formed and precise questions are addressed to ensure effective advancement of our understanding of, and capacity to diagnose and treat, neurodegenerative disorders.

## **LIST OF TERMS**

The primary technical terms and acronyms used in this article are summarized below. A broader overview of techniques and terms is given in the review by McRae et al. (2009), and suggested guidelines for terminology in this field are provided elsewhere (Lobinski et al., 2010).


## **AUTHOR CONTRIBUTIONS**

Concept, first draft, and preparation of figures: Joanna F. Collingwood. Review and critical comment: Mark R. Davidson.

## **ACKNOWLEDGMENTS**

We thank our many collaborators involved with synchrotron analysis of brain iron, performed primarily at the MRCAT ID-10 beamline, Argonne National Laboratory Advanced Photon Source (USA), and the I18 beamline, Diamond Light Source (UK). Their input has shaped many of the observations presented here. Especial thanks to Dr. Albina Mikhailova, Professor Jon Dobson, Professor Christopher Batich, Professor Jeff Terry, Dr. Soma Chattopadhyay, Dr. Raul Barrea, Professor Fred Mosselmans, Dr. Paul Quinn, Dr. Kalotina Geraki, Dr. Mary Finnegan, Professor Christopher Exley, Dr. Neil Telling, and Dr. James Everett. This review was written with support from EPSRC grant EP/K035193/1 (Joanna F. Collingwood).

## **REFERENCES**


ferrihydrite and the Alzheimer's disease peptide beta-amyloid. *Inorg. Chem.* 53, 2803–2809. doi: 10.1021/ic402406g


approaches (IUPAC Technical Report). *Pure Appl. Chem.* 82, 493–504. doi: 10.1351/PAC-REP-09-03-04


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

*Received: 18 May 2014; paper pending published: 17 June 2014; accepted: 25 July 2014; published online: 19 August 2014.*

*Citation: Collingwood JF and Davidson MR (2014) The role of iron in neurodegenerative disorders: insights and opportunities with synchrotron light. Front. Pharmacol. 5:191. doi: 10.3389/fphar.2014.00191*

*This article was submitted to Drug Metabolism and Transport, a section of the journal Frontiers in Pharmacology.*

*Copyright © 2014 Collingwood and Davidson. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*

#### *Zvi Ioav Cabantchik1 \*, Arnold Munnich2, Moussa B. Youdim3 and David Devos <sup>4</sup>*

*<sup>1</sup> Department of Biological Chemistry, Adelina and Massimo Della Pergola Chair, Alexander Silberman Institute of Life Sciences, Hebrew University of Jerusalem, Givat Ram, Jerusalem, Israel*

*<sup>2</sup> Clinical Research Unit, Medical Genetic Clinic and Research Unit INSERM 781, Hôpital Necker-Enfants Malades and Université Paris V René Descartes, Paris, France*

*<sup>3</sup> Technion-Rappaport Family Faculty of Medicine, Eve Topf Center of Excellence, Haifa, Israel*

*<sup>4</sup> Department of Medical Pharmacology, EA1046, Faculty of Medicine, Lille Nord de France University and Lille University Medical Center, Lille, France*

#### *Edited by:*

*Raffaella Gozzelino, Instituto Gulbenkian de Ciência, Portugal*

#### *Reviewed by:*

*Guenter Weiss, Medical University of Innsbruck, Austria Gian L. Forni, Ospedale Galliera Genova, Italy*

#### *\*Correspondence:*

*Zvi Ioav Cabantchik, Department of Biological Chemistry, Adelina and Massimo Della Pergola Chair, Alexander Silberman Institute of Life Sciences, Hebrew University of Jerusalem, Givat Ram, 91904 Jerusalem, Israel e-mail: ioav@cc.huji.ac.il*

The traditional role of iron chelation therapy has been to reduce body iron burden via chelation of excess metal from organs and fluids and its excretion via biliary-fecal and/or urinary routes. In their present use for hemosiderosis, chelation regimens might not be suitable for treating disorders of iron maldistribution, as those are characterized by toxic islands of siderosis appearing in a background of normal or subnormal iron levels (e.g., sideroblastic anemias, neuro- and cardio-siderosis in Friedreich ataxia- and neurosiderosis in Parkinson's disease). We aimed at clearing local siderosis from aberrant labile metal that promotes oxidative damage, without interfering with essential local functions or with hematological iron-associated properties. For this purpose we introduced a conservative mode of iron chelation of dual activity, one based on scavenging labile metal but also redeploying it to cell acceptors or to physiological transferrin. The "scavenging and redeployment" mode of action was designed both for correcting aberrant iron distribution and also for minimizing/preventing systemic loss of chelated metal. We first examine cell models that recapitulate iron maldistribution and associated dysfunctions identified with Friedreich ataxia and Parkinson's disease and use them to explore the ability of the double-acting agent deferiprone, an orally active chelator, to mediate iron scavenging and redeployment and thereby causing functional improvement. We subsequently evaluate the concept in translational models of disease and finally assess its therapeutic potential in prospective double-blind pilot clinical trials. We claim that any chelator applied to diseases of regional siderosis, cardiac, neuronal or endocrine ought to preserve both systemic and regional iron levels. The proposed deferiprone-based therapy has provided a paradigm for treating regional types of siderosis without affecting hematological parameters and systemic functions.

**Keywords: iron, chelators, sideroblastic anemia, neurodegeneration, Parkinson's disease, Friedereich ataxia**

## **INTRODUCTION**

Iron homeostasis relies on the orchestration of a network of systemic and cellular mechanisms for the acquisition, internal distribution and use of iron (Cairo and Recalcati, 2007; Camaschella, 2009; Fleming and Ponka, 2012). Disruption of links in the metabolic network can lead to excessive iron accumulation in particular cell compartments or tissues (causing localized siderosis and thus damage) and also to impaired reuse of iron, generating a "deficiency in the midst of abundance" or vice versa (Breuer and Cabantchik, 2009; Rouault, 2012, 2013). This impairment translates into acquired or inherited anaemias (Camaschella, 2009; Fleming and Ponka, 2012; Rouault, 2012)and other disorders that display regional siderosis in organs such as heart, (Wood, 2007; Pennell et al., 2011; Fleming and Ponka, 2012; Rouault, 2012) brain (Berg et al., 2002; Zecca et al., 2004; Benarroch, 2009; Li et al., 2011; Sian-Hulsmann et al., 2011; Rouault, 2013)and other tissues(Pietrangelo, 2007). In all these disorders, the simultaneous dearth and surplus of iron pose new challenges in drug therapy, with the need to (i) detoxify discrete siderotic foci without affecting essential iron-dependent functions elsewhere and, conversely, (ii) replenish iron-deprived regions without further overloading those already in surplus. Ultimately, addressing the upstream factors that lead to iron maldistribution might lead to more effective therapies than those attempting to correct end-organ dysfunctions. (Camaschella, 2013) However, current tools for clinical intervention are limited to agents designed to treat systemic siderosis (Berdoukas et al., 2011; Ma et al., 2012; Camaschella, 2013) or iron deficiency (Weiss and Goodnough, 2005) but not discrete islands of siderosis that appear in a background of normal or subnormal iron levels (Pietrangelo, 2007; Breuer and Cabantchik, 2009).

We describe here a novel chelation concept designed to scavenge excess iron from regional foci of siderosis but also render the chelated metal metabolically reusable by redeploying it to cellular machineries or systemic acceptors such as transferrin.

## **IRON CHELATION FOR THE TREATMENT OF SYSTEMIC SIDEROSIS**

## **CAUSES OF TISSUE SIDEROSIS AND TREATMENT**

This branch of pharmacology and medicinal chemistry has traditionally focused on the design, synthesis and application of drugs for relieving multi-organ iron accumulation (Fleming and Ponka, 2012) - particularly in patients with primary or secondary hemosiderosis (**Figure 1**) (Ma et al., 2012; Camaschella, 2013). In this context, excessive iron accumulation results from (i) enhanced erythrophagocytosis [which leads to increased iron deposition in hepatic and splenic reticuloendothelial system (RES) cells] (Weiss and Goodnough, 2005; Cairo and Recalcati, 2007; Fleming and Ponka, 2012) and/or (ii) the excessive release of iron into plasma by iron-rich RES cells or by an hyperabsorptive gut, resulting in high plasma iron levels to override transferrin's iron binding capacity (TIBC) and prompting the generation of non-physiological, non-transferrin-bound iron (NTBI) (Brissot et al., 2012; Cabantchik et al., 2013). Ultimately, chemically labile forms of plasma NTBI (referred to generically as labile plasma iron, LPI) infiltrate cells and attain therein toxic levels (Brissot et al., 2012; Cabantchik et al., 2013). (Comment: the term NTBI should be used with caution, as generically it is an apophasis-from the Greek απóϕασις-, namely something that is defined by what it is not).

In hemosiderosis, the toxicity that is associated with iron depends on a numbers of factors: (i) the chemical forms and concentrations of the NTBI/LPI, (ii) the LPI's membrane agencies and entry routes in particular cells; (iii) the cell's ability to handle "surplus" iron by sequestering it within "safe" ferritin sanctuaries and (iv) the cell's enzymatic and non-enzymatic means of counteracting the metal-catalyzed formation of reactive oxygen species (ROSs) involved in oxidative damage(Halliwell, 2006). *Although excessive iron ingress into cells can be deleterious, it is not the hyper-accumulated metal that is causatively associated with oxidative damage but rather a minor fraction referred to as labile cell iron (LCI)*(Halliwell, 2006; Cabantchik et al., 2013). The LCI is characterized by the: (i) redox activity, which largely determines not only its bio-catalytic roles but also its propensity to generate noxious ROSs from reactive oxygen intermediates of the respiratory chain and other oxygen-dependent reactions; (ii) ability to transfer the metal between natural ligands (including putative chaperons) and between compartments, which in turn determines the intracellular metal distribution, and (iii) the metal's amenability to chelation therapy via sequestration into a stable and potentially unreactive iron-chelates. Both LCI and LPI are direct pharmacological targets of chelators that act directly on plasma NTBI-LPI (Cabantchik et al., 2013) and/or permeate into cells and dissipate intracellular iron agglomerates (De Domenico et al., 2008).

In all forms of hemosiderosis, the therapeutic challenge is to attain a balance between iron detoxification and the maintenance of essential, iron-dependent functions. That requires the monitoring of indicators of systemic iron status in the plasma and of regional iron status in selected organs. Historically, (Pietrangelo, 2007; Berdoukas et al., 2011) the diagnosis and treatment of hemosiderosis were followed by (i) direct analysis of liver iron concentration in biopsies taken from what is the major iron

**FIGURE 1 | Chelation modalities as treatments for siderosis.** Systemic siderosis (e.g., hemosiderosis) is characterized by elevated plasma iron levels (hyperferremia) that leads to organ iron accumulation and is generally characterized by elevated plasma ferritin (i.e., hyperferritinemia) that other than in a context of inflammation reflects iron stores. Chelation is designed to detoxify organs from surplus iron and dispose the latter via biliary or urinary secretion. Regional siderosis covers a wide spectrum of inherited disorders (e.g., Friedreich ataxia, sideroblastic anemias, iron-refractory-iron-deficiency anemia-IRIDA), and acquired disorders (e.g., anemia of chronic disease-ACD or cancer) that are characterized by a maldistribution of iron within cells of particular organs (e.g., cardiomyocytes or neurons or blast cells) or at the level of the organism (e.g., liver and spleen versus plasma). The plasma iron levels can span from subnormal (as in functional iron deficient anemias) to supranormal (as in sideroblastic anemias) and likewise those of plasma (or serum) ferritin levels. Chelation is designed here not merely to detoxify a siderotic region but where applicable render the chelated iron available for reuse. The new chelation modalities comprise (i) targeted detoxification, whereby a prochelator is activated at the target site by specific resident activators, as found in some brain areas (Sohn et al., 2008; Zheng et al., 2009) and (ii) iron redeployment, whereby a chelator that detoxifies cells from surplus iron and/or also scavenges essential iron reintegrates the metal into the erythron or specific tissues (Breuer and Cabantchik, 2009; Sohn et al., 2011).

storage organ and (ii) assays of serum levels of ferritin, which (in the absence of inflammation) is a surrogate (long-term) marker of liver-accumulated metal. Analytical methods for estimating urinary and fecal iron secretion have been applied in studies of body iron balance, with a view to directly assessing the efficacy of chelators in reducing the body's iron burden. (Berdoukas et al., 2011) The introduction of non-invasive methods for assessing tissue iron agglomerates was a major step forward in assessing longterm organ siderosis and the efficacy of iron chelation therapy. (Berdoukas et al., 2011) These assessments were initially performed in organs such as the liver [by using superconducting quantum interference device (SQUID) magnetometry] and, more recently, in the liver, heart and endocrine glands with T2 and T2∗ nuclear magnetic resonance (NMR) spin relaxation methods (Wood, 2007; Berdoukas et al., 2011; Pennell et al., 2011; Camaschella, 2013). The most prominent example is the case of cardiac siderosis that results from transfusional hemosiderosis, in which the T2∗ NMR relaxation time was validated (i) retrospectively for its correlation with cardiac iron measured in tissue biopsies and (ii) prospectively for its correlation with impaired cardiac function (Wood, 2007; Pennell et al., 2011).

Today's arsenal of iron chelators in clinical use comprises three agents approved for the treatment of transfusional hemosiderosis, (Berdoukas et al., 2011; Ma et al., 2012; Camaschella, 2013). The parenterally administered hexadentate deferrioxamine and the two orally active, bidentate deferiprone (DFP) and the tridentate deferasirox (DFX). These agents are used either alone or in various combinations designed to optimize the reduction of the body's iron burden by extracting excess iron from fluids and tissues and with time maintain NTBI-LPI at basal (i.e., sub-toxic) levels. (Berdoukas et al., 2011; Cabantchik et al., 2013).

## **IRON CHELATION FOR THE TREATMENT OF REGIONAL SIDEROSIS**

## **CAUSES AND TARGETS FOR TREATMENT**

Unlike hemosiderosis (in which plasma iron infiltrates cells in several different organs), regional siderosis mostly results from a disruption in cell homeostasis that causes iron to be abnormally distributed in certain cell types or specific organs (Fleming and Ponka, 2012; Rouault, 2012). In various types of sideroblastic anemia (Camaschella, 2009; Fleming and Ponka, 2012; Rouault, 2012) iron accumulates in mitochondria due to faulty use of the metal in the synthesis or secretion of iron sulfur clusters (ISCs) or heme (Rouault, 2012, 2013). A concomitant cytosolic iron depletion (Kakhlon et al., 2008; Rouault, 2012) generates a vicious cycle in which ever more incoming iron ends up in mitochondrial iron precipitates (Kakhlon et al., 2010; Whitnall et al., 2012). The damage caused by abnormal intracellular metal distribution has two major components: (i) toxic mitochondrial iron accumulation *per se* (mostly as labile/toxic iron-oxides) and (ii) a metabolic decline in cell iron-dependent activities (due to faulty synthesis of essential hemoproteins or ISC proteins) (Kakhlon et al., 2008; Rouault, 2012). Abnormal iron distribution is also observed in systemic functional iron deficiencies (Camaschella, 2009; Fleming and Ponka, 2012; Rouault, 2012). In acquired anemia of chronic disease (ACD) (Weiss and Goodnough, 2005) and in inherited ironrefractory-iron-deficiency-anemia (IRIDA), (Camaschella, 2009; Fleming and Ponka, 2012) macrophages of the RES system (in the spleen and liver) accumulate iron. The end results is plasma iron depletion and ensuing systemic iron deprivation. Moreover, in patients with chronic kidney disease and functional iron deficiency, intensive treatment with intravenous iron supplements can also cause the iatrogenic accumulation of macromolecular iron-sugar aggregates in spleen and liver. This can also be identified by T2∗ NMR as iron overload (Ghoti et al., 2012). However, iron accumulation in the RES is *per se* not pathological, since apparently no labile iron is generated as long as the metal is safely shielded within shells of protein (e.g., ferritin) or saccharide (e.g., sucrose or dextran).

Iron maldistribution is also found in various neurodegenerative disorders(Zecca et al., 2004; Weinreb et al., 2010; Li et al., 2011; Pichler et al., 2013; Rouault, 2013).In the inherited movement disorder Friedreich's ataxia (FRDA), siderosis manifests itself first in the brain's dentate nuclei and ultimately in the heart (Rouault, 2013). In Parkinson's disease (PD), the dopaminergic neurons of the substantia nigra pars compacta progressively succumb to oxidative stress (Zecca et al., 2004; Halliwell, 2006; Li et al., 2011; Sian-Hulsmann et al., 2011; Pichler et al., 2013).

However, the contribution of brain siderosis to the etiology of PD is subject to debate—largely because opposing hypotheses have mostly relied on correlations between functional or biophysical indices of PD and indicators of systemic, rather than local and/or labile, iron levels (Zecca et al., 2004; Hider et al., 2011; Li et al., 2011) Those studies included trials of chelators that affect systemic iron but not necessarily brain iron, (Li et al., 2011; Mounsey and Teismann, 2012) or genetic factors that affect systemic iron levels [e.g., HFE and TMPRSS6 gene mutations (Camaschella, 2009, 2013; Pichler et al., 2013)] that in turn might be inversely correlated with PD severity (Pichler et al., 2013) and possibly with iron accumulated in the substantia nigra. Moreover, in regional siderosis (whether inherited or acquired), detection of iron agglomerates by T2∗ MRI or histochemical staining is not necessarily indicative of an existing or impending disease state, just as failure to detect regional iron agglomerates does not rule out an existing or impending disease state.

By analogy with cardiac siderosis, we hypothesized that a causal relationship between iron and disease would be revealed if the course of disease is modified by selectively modifying labile iron levels in regions of siderotic damage (Breuer and Cabantchik, 2009; Kakhlon et al., 2010) We further reasoned that removal of the toxic labile component by "selective" chelation might suffice for relieving iron maldistribution disorders in which siderosis is the major pathological factor (i.e., by prevention and/or restoration of otherwise affected functions) (**Figure 1**). However, for disorders in which siderosis results from an inherited inability to use the metal for the production of essential ISCs or heme (as in sideroblastic anemia and FRDA), selective detoxification by targeted iron chelation (Zheng et al., 2009, 2010) might have only limited benefits. Moreover, in the absence of a safety mechanism that keeps chelated iron (resulting from systemic, non-specific chelation) in the body (**Figure 1**), the risk of iatrogenic iron deprivation might outweigh the benefits of regional detoxification (Breuer and Cabantchik, 2009). And conversely, the risk that systemic (parenteral) supply of polymeric iron supplements might pose to hypoferremic/ hyperritinemic patients that retain iron polymers their RES iron stores due to chronic inflammation for extended time periods (Ghoti et al., 2012) might outweigh modest changes in erythropoiesis evoked by increased transferrin saturation (Weiss and Goodnough, 2005). Prospective studies dealing with those issues are critically missing not only in clinical practice, but also in validated analysis of chemical stability of iron formulations as part of their shelf life and especially once delivered to patients via parenteral routes.

## **NOVEL THERAPEUTIC STRATEGIES FOR REGIONAL SIDEROSIS**

#### **IRON REDEPLOYMENT (FIGURE 2)**

This approach seeks to conserve regionally or systemically chelated iron by redeploying it to sites of use (i.e., to ISC or heme biosynthetic machineries or the physiological acceptor apotransferrin) (Sohn et al., 2008; Breuer and Cabantchik, 2009; Kakhlon et al., 2010). Redeployment (which is essentially applicable to any disorder of iron maldistribution, whether cellular or systemic) was initially assessed in cell-based models, (Kakhlon et al., 2008; Sohn et al., 2008, 2011; Moreau et al., 2013) corroborated in an animal model of regional siderosis (Moreau et al., 2013; Devos et al., 2013) and then translated to a clinical setting (Vreugdenhil et al., 1989; Boddaert et al., 2007; Wood, 2007; Abbruzzese et al., 2011; Kwiatkowski et al., 2012; Devos et al., 2013; Moreau et al., 2013). The membrane-permeant, bidentate DFP (1,2-dimethyl-3-hydroxy-pyridine-4-one) is the prototypic redeployment agent. It has a moderate ability to (i) rescue iron-overloaded cells by scavenging labile iron [especially from mitochondria, (Kakhlon et al., 2008, 2010; Devos et al., 2013; Moreau et al., 2013) the organelles most affected by cell iron accumulation and the ensuing labile-iron mediated oxidative damage (Halliwell, 2006)] and (ii) deliver chelated metal to extracellular apotransferrin (Sohn et al., 2008, 2011) (so that systemic iron losses are essentially avoided) and (iii) translocate across membranes (including the blood brain barrier) and gain access to foci of iron accumulation (thus conferring a degree of protection against metal-initiated oxidative damage) (Kakhlon et al., 2008; Sohn et al., 2011).

The clinical assessment of DFP as a potential iron redeployment agent has been facilitated by (i) the drug's good safety profile in the treatment of hemosiderosis (Berdoukas et al., 2011) (ii) its proven chelating (and thus life-saving) effect in patients with severe cardiac siderosis (Wood, 2007; Berdoukas et al., 2011; Pennell et al., 2011) and (iii) the ability to support hemoglobin production in patients with ACD due to rheumatoid arthritis (Vreugdenhil et al., 1989). As a major goal was to redeploy iron and thus recycle the chelated metal, (Kakhlon et al., 2008; Breuer and Cabantchik, 2009) dose optimization required the siderotic region to be monitored for tangible indications of chelation and systemic iron-parameters (via the assessment of plasma transferrin saturation, complete blood counts and urinary iron) (Vreugdenhil et al., 1989; Boddaert et al., 2007; Wood, 2007; Abbruzzese et al., 2011; Kwiatkowski et al., 2012; Devos et al., 2013; Moreau et al., 2013).

The first attempts to assess the clinical safety and efficacy of up to 12 months of treatment with a moderate dose of DFP (20–30 mg/kg/d, initially combined with the antioxidant idebenone) were performed in adolescent patients with FRDA (Boddaert et al., 2007; Velasco-Sánchez et al., 2011). With the exception of a few cases of neutropenia (which resolved spontaneously after withdrawal of these patients from the study), DFP treatment modified neither plasma iron parameters nor the CBC but did reverse the cardiomyopathic hypertrophy (Velasco-Sánchez et al., 2011). The observed reduction in R2∗ NMR values in the cerebellar dentate nuclei essentially proved that DFP probably acted in these brain areas by gaining direct access to LCI and (indirectly) dissipating large iron agglomerates (Boddaert et al., 2007). Similar hematological responses were observed following long-term DFP treatment of neurodegeneration with brain iron accumulation (NBIA) (Abbruzzese et al., 2011; Kwiatkowski et al., 2012) and, more recently, of PD patients on stable dopamine/dopaminergic regimens (Moreau et al., 2013) In the latter population, a 12 month oral DFP regimen was associated with clear reductions in the amount of agglomerated iron in the substantia nigra and attenuation of disease progression (according to the Unified Parkinson's Disease Rating Scale). The drug appeared to act by (i) cell iron detoxification and thereby a reduction in oxidative damage and (ii) reduction in enzymatic dopamine catabolism and/or nonenzymatic oxidation of naturally produced dopamine or administered dopamine agonists. At this stage we are not in a position to assess to what extent the protective effects afforded by chelation can be attributed solely to direct chelation of LCI and in turn reduction of metal-driven oxidations and ensuing cell damage. Chelation of labile iron at the level of prolyl-hydroxylases can lead to protective mechanisms involving HIF-dependent pathways, (Weinreb et al., 2010) particularly those leading to the local synthesis of erythropoietin, a hormone implicated in neuroprotection (Sturm et al., 2010; Matilla-Dueñas et al., 2013).

## **TARGETED (SITE-DIRECTED) CHELATION (FIGURE 3)**

Targeted chelation has the inherent advantage of regional metal detoxification—particularly when it is based on a chelator that can be specifically activated *in situ* in the target region and can spare the organism from the risk of non-specific chelation and thus iron deprivation. At present, the best studied experimental approach is based on pro-chelators that are activated in the tissue, cell or organelle (i.e., at or close to the site of damage). This activation is achieved by an intrinsic (local) activation factor (e.g., an enzyme) that is only present or active in the target region (Sohn et al., 2008; Zheng et al., 2009) This concept is exemplified by the carbamylated hydroxyquinoline (HQ) pro-chelators that are converted by brain cholinesterases (ChEs) into active HQ chelators. As the ChE enzymes can be chemically modified and thereby inhibited by the leaving carbamate group (**Figure 3**), the approach has the potential benefit of addressing the dementia components of PD and Alzheimer's disease (AD). Moreover, the incorporation of N-propargylamine R groups confers to the HQ prochelators the additional capacity of monoamine oxidase (MAO) inhibition for the symptomatic treatment of PD (Sohn et al., 2008; Zheng et al., 2009). Several prochelators with these single- or double-agent features have been tested in animal models of neurodegenerative disorders (such as amyotrophic lateral sclerosis (ALS), AD and PD). They have both neuroprotective and neurorestorative effects (Sohn et al., 2008; Zheng et al., 2009; Weinreb et al., 2010) and do not appear to interfere (in the short term) with systemic iron and other HQ-chelatable biometals (such as copper and zinc). However, remaining to be assessed is to what extent the MAO-B versus MAO-A selectivity is retained over the long term treatment of chronic neurodegenerative condition.

## **CONCLUSIONS AND PERSPECTIVES**

Until recently, iron chelation therapy has focused primarily on the traditional role of reducing total iron burden that characterizes systemic siderosis (as in hemosiderosis) by chelating excess metal and excreting it via urinary and/or biliary-fecal routes (Berdoukas et al., 2011; Brissot et al., 2012; Ma et al., 2012; Camaschella, 2013). In the design of chelators, the major goal of systemic iron detoxification from excess iron, has hitherto mostly relied on high binding affinity properties and specificity of the chelating units for sequestering labile iron but also for generating nontoxic iron-chelates that ultimately should be disposed from the organism. The treatment of regional siderosis demanded a novel modality of chelation, one that can be used for clearing siderotic regions comprised of labile (toxic) metal but without interfering with essential local functions, thus sparing the patient's systemic iron pools. That selectivity property is of cardinal importance, since increased serum iron levels have recently been suggested by genome-wide meta-analysis (GWAS) to be correlated with a decreased risk of developing PD (Pichler et al., 2013), although, as we referred earlier, a direct role of the implicated genes (HFE and TMPRSS) in brain iron metabolism in general and on neurodegeneration in specific, still need to be properly evaluated, particularly in the context of PD.

The proposed chelation modalities designed for selective treatment of local siderosis have been tested in both cell-based and animal models. Those modalities are referred as: (i) **site-directed (targeted) chelation**, which relies on pro-chelating agents that are enzymatically converted into active chelators in various brain areas and act locally as metal-detoxifying agents (Sohn et al., 2008; Zheng et al., 2009) and (ii) **iron redeployment,** which aims at correcting foci of iron maldistribution by reutilizing tissue chelated metal and thus maintaining it within the body (Breuer and Cabantchik, 2009; Kakhlon et al., 2010).

Although several novel, brain-targeted pro-chelators have already displayed neuroprotective and neurorestorative properties in preclinical models of neurodegeneration, (Sohn et al., 2008; Zheng et al., 2009; Weinreb et al., 2010) their long-term clinical safety, specificity and efficacy have yet to be assessed. An iron redeployment agent that has already shown clinical benefits in the early stages of degenerative neurosiderosis is DFP, a moderately active, oral chelator with a proven record in the treatment of transfusional and hereditary cardiac siderosis (Wood, 2007; Berdoukas et al., 2011; Pennell et al., 2011). To date, most published placebo-controlled clinical studies have been carried out with DFP in combination with either idebenone (in FRDA) or with dopamine/dopaminergic drugs (in PD). Administration of the chelator was associated with significant benefits - possibly via disease-modifying mechanisms. Large, prospective, randomized clinical trials of DFP (alone and in combination with symptomatic drugs) are now needed, in order to properly appreciate the full therapeutic potential of new chelation modalities in regional forms of siderosis. This applies to not only neurosiderosis (as exemplified by FRDA, NBIA, and PD) but also various forms of hereditary and acquired sideroblastic anemia (Camaschella, 2009; Fleming and Ponka, 2012; Rouault, 2012) and other hallmark iron maldistribution disorders (Pietrangelo, 2007; Kakhlon et al., 2010; Rouault, 2012). The main challenge is to optimize existing treatments and develop novel specific, efficacious and safer drugs—possibly by combining a site-directed targeting component with a chelating moiety that is unmasked at the target sites and, like DFP, has metal-redeployment features (**Figure 3**).

### **AUTHOR CONTRIBUTIONS**

Writing of manuscript: a, first draft writing; b, review and critical comment; c, final draft writing. Zvi Ioav Cabantchik: a, b, c; Arnold Munnich: b; Moussa B. Youdim: c; David Devos: b,c.

#### **REFERENCES**


are restored by drug-mediated iron relocation. *Blood* 112, 5219–5227. doi: 10.1182/blood-2008-06-161919


**Conflict of Interest Statement:** The authors have no financial disclosures to make or potential conflicts of interest to report in relation to this review. (The clinical trial on PD was funded by the French Ministry of Health PHRC grant Projet Hospitalier Recherche Clinique Protocol ID: 2008-006842-25), whereas Apopharma provided the compound deferiprone and instructions on its use. Zvi Ioav Cabantchik was supported by the A&M Della Pergola Chair in Life Sciences, served on the Scientific Advisory Board of Novartis ESB, received in the past research grants from NIH, EEC (5 and 6 framework), Israel Science Foundation, AFIRNE (Paris), Novartis, Shire and Apopharma (also lecture honoraria from the last two) and is presently a consultant for Aferrix, Ltd and Hinoman, Ltd, Israel. David Devos served on the Scientific Advisory Board for Novartis and Aguettant and has received PHRC grants from the French Ministry of Health and research funding from the ARSLA charity and honoraria from pharmaceutical companies for consultancy and lectures. Arnold Munnich is the head of the Pediatric Genetic Unit at Hospital Necker, Paris France. He has served as an advisor to Sanofi Genzyme until May 2013. He has presently no other appointment with commercial entities. Moussa B. Youdim receives Royalties from Teva Pharmaceutical Co. Israel and is Chief Scientific Officer co-owner and shareholder of Abital Pharma Pipeline Ltd, Israel, and of Avraham Pharmaceutical Co. Israel.

*Received: 21 November 2013; paper pending published: 09 December 2013; accepted: 15 December 2013; published online: 31 December 2013.*

*Citation: Cabantchik ZI, Munnich A, Youdim MB and Devos D (2013) Regional siderosis: a new challenge for iron chelation therapy. Front. Pharmacol. 4:167. doi: 10.3389/fphar.2013.00167*

*This article was submitted to Drug Metabolism and Transport, a section of the journal Frontiers in Pharmacology.*

*Copyright © 2013 Cabantchik, Munnich, Youdim and Devos. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*

## ADVANTAGES OF PUBLISHING IN FRONTIERS

FAST PUBLICATION Average 90 days from submission to publication

#### COLLABORATIVE PEER-REVIEW

Designed to be rigorous – yet also collaborative, fair and constructive

RESEARCH NETWORK Our network increases readership for your article

## OPEN ACCESS

Articles are free to read, for greatest visibility

## TRANSPARENT

Editors and reviewers acknowledged by name on published articles

## GLOBAL SPREAD Six million monthly page views worldwide

## COPYRIGHT TO AUTHORS

No limit to article distribution and re-use

IMPACT METRICS Advanced metrics track your article's impact

## SUPPORT By our Swiss-based editorial team

EPFL Innovation Park · Building I · 1015 Lausanne · Switzerland T +41 21 510 17 00 · info@frontiersin.org · frontiersin.org