# NEURODEGENERATION: FROM GENETICS TO MOLECULES

EDITED BY: Victoria Campos-Peña, Marco Antonio Meraz-Ríos, Rosalinda Guevara-Guzmán and Karla Guadalupe Carvajal PUBLISHED IN: Frontiers in Cellular Neuroscience

#### *Frontiers Copyright Statement*

*© Copyright 2007-2016 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 978-2-88945-020-6 DOI 10.3389/978-2-88945-020-6

# 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

# **NEURODEGENERATION: FROM GENETICS TO MOLECULES**

#### Topic Editors:

**Victoria Campos-Peña,** Instituto Nacional de Neurología y Neurocirugía, Mexico **Marco Antonio Meraz-Ríos,** Centro de Investigación y de Estudios Avanzados, Mexico **Rosalinda Guevara-Guzmán,** Universidad Nacional Autónoma de México, Mexico **Karla Guadalupe Carvajal,** Instituto Nacional de Pediatría, Mexico

Bielschowsky Stain in alzheimer´s brain patient, neurofibrillary tangles are shown in dark brown. Image by Victoria Campos Peña

Chronic degenerative diseases are one of the major public health problems, particularly those affecting the nervous system. They are characterized by the degeneration of specific cell populations that include several pathologies which contribute significantly to morbidity and mortality in the elderly population. Therefore, in recent years, the study of neuroscience has gained significant importance.

Most of these neurodegenerative disorders are the result of a complex interaction between genetic and environmental factors that generate progression and can even determine its severity. The presence of mutations in genes as LRRK2, SNCA, PARK7, PARK2 or PINK1 is associated with Parkinson's disease. Mutations in genes such as APP, PS1 and PS2 are associated with familial Alzheimer's disease; while HTT gene mutations are the cause of Huntington's disease. In most cases, this condition is inherited in an autosomal

dominant pattern, which means one copy of the altered gene in each cell is sufficient to cause the disorder.

It is known that these mutations can also alter the proteins function; however, it has not yet been possible to fully understand how some genetic changes cause the disease or influence the risk of developing these disorders. Most symptoms seen in these conditions occurs when specific nerve cells are damaged or die generating a loss in brain communication. Also many of these mutations generate aggregation of intracellular or extracellular proteins affecting cell function and eventually causing neuronal death. It is unclear whether the presence of these aggregates play an important role in nerve cell death during the development of neurodegenerative diseases, or if they are simply part of the response of cells to the disease.

Other mutations affect the mitochondrial function generating alterations in energy production and promoting the formation of unstable molecules such as free radicals. Under normal conditions, the harmful effects caused by free radicals, are offset within the cell. However, in pathological conditions, the presence of mutations can alter this process by allowing the accumulation of radicals and damaging or killing cells.

On the other hand, we also know that these diseases may not have a direct genetic component, thus, the study of sporadic type neurodegenerative diseases is much more complex. Histopathological lesions as well as the cellular and molecular alterations are generally indistinguishable from familial cases. For this reason, it is important to understand the genetic and molecular mechanisms associated with this type of pathologies. In this sense, this issue aims to understand the molecular processes that occur in the brain, and how these are influenced by the environment, genetics and behavior.

**Citation:** Campos-Peña, V., Meraz-Ríos, M. A., Guevara-Guzmán, R., Carvajal, K. G., eds. (2016). Neurodegeneration: From Genetics to Molecules. Lausanne: Frontiers Media. doi: 10.3389/978-2-88945-020-6

# Table of Contents


Yash B. Joshi and Domenico Praticò

*91 Evaluation of inflammation-related genes polymorphisms in Mexican with Alzheimer's disease: a pilot study*

Danira Toral-Rios, Diana Franco-Bocanegra, Oscar Rosas-Carrasco, Francisco Mena-Barranco, Rosa Carvajal-García, Marco Antonio Meraz-Ríos and Victoria Campos-Peña

*101 Aged neuronal nitric oxide knockout mice show preserved olfactory learning in both social recognition and odor-conditioning tasks* Bronwen M. James, Qin Li, Lizhu Luo and Keith M. Kendrick


Bryan V. Phillips-Farfán, María del Carmen Rubio Osornio, Verónica Custodio Ramírez, Carlos Paz Tres and Karla G. Carvajal Aguilera

*135 Cerebellar transcriptional alterations with Purkinje cell dysfunction and loss in mice lacking PGC-1***a**

Elizabeth K. Lucas, Courtney S. Reid, Laura J. McMeekin, Sarah E. Dougherty, Candace L. Floyd and Rita M. Cowell


Marcio H. M. da Luz, Italo T. Peres, Tiago G. Santos, Vilma R. Martins, Marcelo Y. Icimoto and Kil S. Lee

*175 GABAergic alterations in neocortex of patients with pharmacoresistant temporal lobe epilepsy can explain the comorbidity of anxiety and depression: the potential impact of clinical factors*

Luisa Rocha, Mario Alonso-Vanegas, Iris E. Martínez-Juárez, Sandra Orozco-Suárez, David Escalante-Santiago, Iris Angélica Feria-Romero, Cecilia Zavala-Tecuapetla, José Miguel Cisneros-Franco, Ricardo Masao Buentello-García and Jesús Cienfuegos

*185 HDAC4 as a potential therapeutic target in neurodegenerative diseases: a summary of recent achievements*

Michal Mielcarek, Daniel Zielonka, Alisia Carnemolla, Jerzy T. Marcinkowski and Fabien Guidez

*194 Hydrogels as scaffolds and delivery systems to enhance axonal regeneration after injuries*

Oscar A. Carballo-Molina and Iván Velasco

*206 Identification of age- and disease-related alterations in circulating miRNAs in a mouse model of Alzheimer's disease*

Sylvia Garza-Manero, Clorinda Arias, Federico Bermúdez-Rattoni, Luis Vaca and Angélica Zepeda

*215 Identification of the antiepileptic racetam binding site in the synaptic vesicle protein 2A by molecular dynamics and docking simulations*

José Correa-Basurto, Roberto I. Cuevas-Hernández, Bryan V. Phillips-Farfán, Marlet Martínez-Archundia, Antonio Romo-Mancillas, Gema L. Ramírez-Salinas, Óscar A. Pérez-González, José Trujillo-Ferrara and Julieta G. Mendoza-Torreblanca

*227 Mitomycin-treated undifferentiated embryonic stem cells as a safe and effective therapeutic strategy in a mouse model of Parkinson's disease* Mariana Acquarone, Thiago M. de Melo, Fernanda Meireles, Jordano Brito-Moreira,

Gabriel Oliveira, Sergio T. Ferreira, Newton G. Castro, Fernanda Tovar-Moll, Jean-Christophe Houzel and Stevens K. Rehen

*239 Pleiotrophin as a central nervous system neuromodulator, evidences from the hippocampus*

Celia González-Castillo, Daniel Ortuño-Sahagún, Carolina Guzmán-Brambila, Mercè Pallàs and Argelia Esperanza Rojas-Mayorquín

*246 Protection against neurodegeneration with low-dose methylene blue and near-infrared light*

F. Gonzalez-Lima and Allison Auchter

*251 Retinal aging in the diurnal Chilean rodent (***Octodon degus***): histological, ultrastructural and neurochemical alterations of the vertical information processing pathway*

Krisztina Szabadfi, Cristina Estrada, Emiliano Fernandez-Villalba, Ernesto Tarragon, Gyorgy Setalo Jr., Virginia Izura, Dora Reglodi, Andrea Tamas, Robert Gabriel and Maria Trinidad Herrero

# Editorial: Neurodegeneration: from Genetics to Molecules

Marco A. Meraz-Ríos <sup>1</sup> , Rosalinda Guevara-Guzmán<sup>2</sup> , Karla G. Carvajal <sup>3</sup> and Victoria Campos-Peña<sup>4</sup> \*

 Departamento de Biomedicina Molecular, Centro de Investigación y de Estudios Avanzados, Mexico City, Mexico, Departamento de Fisiología, Facultad de Medicina, Universidad Nacional Autónoma de México, Mexico City, Mexico, Laboratorio de Nutrición Experimental, Instituto Nacional de Pediatría, Mexico City, Mexico, <sup>4</sup> Laboratorio Experimental de Enfermedades Neurodegenerativas, Instituto Nacional de Neurología y Neurocirugía, Mexico City, Mexico

Keywords: neurodegeneration, memory impairment, systemic inflammation and neurodegeneration, dopaminergic neurotransmission, caloric restriction, sirtuins, miRNAs, neuroprotective effects

#### **The Editorial on the Research Topic**

#### **Neurodegeneration: from Genetics to Molecules**

This research topic gives an overview of the current knowledge on some of the most common processes in neurodegenerative diseases such as Alzheimer, Parkinson and Temporal Lobe Epilepsy. Here we reviewed different aspects of each neurodegenerative process, including epigenetics, oxidative and inflammatory responses, genetic susceptibility, aging, metals and environmental exposure, autophagy, and micro RNA basic research.

Neurodegeneration: from genetics to molecules, is a multidisciplinary research topic aiming to be of interest to basic researchers in medical fields.

Many expected that the splendid and remarkable success of the human genome project introduced toward the end of the Twentieth century would halt the epidemic of neurodegenerative diseases. Yet, rather than diminished, it has risen to become a serious problem of morbidity, loss of useful life years, and mortality worldwide. Simultaneously, laboratory research has continued to unravel successive layers of the pathophysiology of neurodegenerative diseases and the mechanisms of its clinical complications, suggesting new routes to follow for the development of innovative therapeutic approaches.

Hence, the need for a compendium of articles that reviews the current state-of-the-art of neurodegeneration research spanning from a global perspective to fundamental molecular mechanisms. The editors intended this collection of articles in order to highlight the emerging challenges to many tenets of belief that have emerged from recent studies. We do so in the spirit of inspiring future research to attack the unresolved questions and to exploit the newer discoveries that have opened unanticipated horizons of understanding and raised novel questions and opportunities for therapies.

Calero et al. commence the research topic with an update of additional mechanisms conferring genetic susceptibility to Alzheimer's disease. They describe the analysis of mutations in the genes involved in the production of the amyloid related to EOFAD. In addition, they review the genes associated to LOAD and the list of genes associated with sporadic AD. Further, they analyze recent whole-exome studies looking for strong association to the disease, and integrate several studies dealing with epistasis as additional mechanisms conferring genetic susceptibility to AD, searching for networks rather than the contribution of specific genes (Calero et al).

Edited by: Egidio D'Angelo, University of Pavia, Italy

#### Reviewed by:

Fabio Blandini, Fondazione Istituto Neurologico Nazionale Casimiro Mondino (IRCCS), Italy Pier Giorgio Mastroberardino, Erasmus University Rotterdam, Netherlands

> \*Correspondence: Victoria Campos-Peña neurovcp@ymail.com

Received: 10 May 2016 Accepted: 19 July 2016 Published: 03 August 2016

#### Citation:

Meraz-Ríos MA, Guevara-Guzmán R, Carvajal KG and Campos-Peña V (2016) Editorial: Neurodegeneration: from Genetics to Molecules. Front. Cell. Neurosci. 10:187. doi: 10.3389/fncel.2016.00187

Landgrave-Gómez et al. reviewed the role of epigenetic mechanisms in the function and homeostasis of the central nervous system and their participation in a variety of neurological disorders. The evidence suggests that long-term changes in gene transcription could be associated with changes in chromatin structure and participate in a variety of neurological disorders (Landgrave-Gomez et al).

Chin-Chan et al. evaluated the environmental pollutants as risk factors for Alzheimer (AD) and Parkinson diseases (PD). They discuss the role of environmental factors such as lead, mercury, aluminum, cadmium and arsenic, as well as some pesticides and metal-based nanoparticles in the development of idiopathic AD and PD, and their mechanisms of action.

Ramos-Chávez et al. reviewed the neurological effects of inorganic arsenic (iAs) exposure as an important natural pollutant. They investigate the effects of gestational iAs exposure and evaluate the expression of the cysteine/glutamate transporters in cortex and hippocampus, demonstrating significant differences in spatial memory impairment in males while the effect was marginal in females (Chin-Chan et al).

Sankowski et al. examined the relationship of systemic inflammation and the brain, looking for novel roles of genetic, molecular, and environmental cues as drivers of neurodegeneration. The CNS responds to somatic and autonomic sensory information, also receiving input from the periphery about inflammation and infection. The hypothesis is that the resulting translocation of inflammatory mediators can interfere with neuronal and glial homeostasis. Here, they review recent genetic evidence suggesting an association between neurodegenerative disorders and persistent immune activation and evaluate their potential relevance in neurodegenerative disorders (Sankowski et al).

Joshi and Pratico reviewed the 5-lipoxygenase (5LO) and its downstream leukotriene metabolites as oxidative and inflammatory contributions to Alzheimer's disease. 5LO pathway and 5LO activating protein (FLAP) have been implicated in the molecular pathology of AD, including the metabolism of amyloid-β and tau. In this article you will find an overview of 5LO and FLAP, discussing their involvement in biochemical pathways relevant to AD pathogenesis and how targeting these proteins could lead to therapies relevant not only for AD, but also other related neurodegenerative conditions (Joshi and Pratico).

Toral-Rios et al. evaluated the relationship of polymorphisms of inflammation-related genes with Alzheimer's disease. Since it has been demonstrated that inflammation contributes to the process of neurodegeneration and therefore being a key factor in the development of AD, the authors evaluate the differences in frequencies between AD and controls by single-nucleotide polymorphism (SNP) of inflammation-related genes (Toral-Rios et al).

James et al. analyzed aged neuronal nitric oxide knockout mice. The animals show preserved olfactory learning suggesting that lack of NO releases protected animals against age-associated cognitive decline in memory tasks typically involving olfactory and hippocampal regions, but not against declines in reversal learning or locomotor activity (James et al).

Blanco Ayala et al. studied an alternative kynurenic acid synthesis route in the rat cerebellum. The Kynurenic acid (KYNA) is an endogenous antagonist of α7 nicotinic acetylcholine and excitatory amino acid receptors. Their results suggest that different mechanisms are involved in KYNA production in the rat cerebellum and that, specifically, DAAO and ROS can function as alternative routes for KYNA synthesis (Blanco Ayala et al).

Phillips-Farfán et al. demonstrated that caloric restriction (CR) has a protective effect against amygdala electrical kindling by inhibiting the mTOR signaling pathway. They investigated whether CR altered the levels of insulin and energy substrates. And found that CR decreases protein kinase B phosphorylation and ribosomal protein S6, suggesting an inhibition of the mTOR cascade as well as an increasing after-discharge threshold tending to reduce the after-discharge duration. It suggests an anticonvulsive action which could imply that CR has an anti-epileptic effect via inhibition of the mTOR pathway (Phillips-Farfan et al).

Lucas et al. found cerebellar transcriptional alterations with Purkinje cell dysfunction and loss in mice lacking the transcriptional coactivator peroxisome proliferator-activated receptor γ coactivator-1α (ppargc1a or PGC-1α). The authors found that mice lacking PGC-1α exhibit ataxia in addition to deficits in motor coordination. Their data suggest that dysfunction in multiple cell types contributes to motor deficits in the context of PGC-1α deficiency (Lucas et al).

Braidy et al. evaluated the differential expression of sirtuins in aging rat brains. They tested mRNA and protein expression levels of rat SIRT1-7, and levels of associated proteins in the brain using RT-PCR and western blotting. This article identifies important unknown roles for sirtuins in regulating cellular homeostasis and healthy aging (Braidy et al).

da Luz et al. found that dopamine induces the accumulation of insoluble prion protein and affects autophagic flux. Since dopamine metabolism generates several oxidative metabolites and can be highly reactive, the authors investigated whether these molecules can also affect the aggregation of cellular prion protein (PrPC). Their results bring new insight into the dopamine metabolism as a source of endogenous metabolites capable of altering PrPC solubility and its subcellular localization (da Luz et al).

Rocha et al. studied the GABAergic alterations in neocortex of patients with pharma-co-resistant temporal lobe epilepsy and explain the comorbidity of anxiety and depression. Their results showed a dysfunction of the GABAergic neurotransmission in temporal neocortex of patients with Temporal Lobe Epilepsy and comorbid anxiety and/or depression that could be also influenced by clinical factors such as seizure frequency and duration of illness (Rocha et al).

Mielcarek et al. evaluated histone deacetylase 4 as a potential therapeutic target in neurodegenerative diseases. HDAC4 is a member of the class IIa family of histone deacetylases and is the major player in synaptic. In this review the authors integrate the information regarding the biological role of HDAC4 in neurodegenerative processes (Mielcarek et al).

Carballo-Molina and Velasco evaluated the hydrogels as scaffolds and delivery systems to enhance axonal regeneration after injuries. As hydrogels are biodegradable, biocompatible and have the capacity to deliver a large range of molecules in situ, they have been used in several biomedical applications. This review discusses areas of opportunity where hydrogels can be applied, in order to promote axonal regeneration of the NS, by applying hydrogels in combination with cultured neural cells, forming three-dimensional structures, showing the formation of synapses and neuronal survival (Carballo-Molina and Velasco).

Garza-Manero et al. performed the identification of age- and disease-related alterations in circulating miRNAs in a mouse model of Alzheimer's disease. In this study the authors evaluated the levels of miRNA at different time-points in a transgenic mouse model of Alzheimer disease (3xTg-AD). They found agerelated significant changes in miRNA abundance for both WT and transgenic mice. Some of these were specific for the 3xTg-AD, suggesting that the age-dependent evolution of the ADlike pathology, rather than the presence and expression of the transgenes, modifies the circulating miRNA levels in the 3xTg-AD mice (Garza-Manero et al).

Correa-Basurto et al. using molecular dynamics and docking simulations, realized the identification of the antiepileptic racetam binding site in the synaptic vesicle protein 2A (SV2A). SV2A is the molecular target of the anti-epileptic drug levetiracetam and its racetam analogs. The importance of this study is that identifying the racetam binding site within SV2A will facilitate the synthesis of new radio-ligands to improve treatment response and evaluate epilepsy progression (Correa-Basurto et al).

Acquarone et al. studied in a mouse model of Parkinson's disease (PD), mitomycin-treated undifferentiated embryonic stem cells as a safe and effective cellular therapeutic strategy. The authors show that undifferentiated mouse embryonic stem cells (mESCs), pre-treated with mitomycin C (MMC) before transplantation, prevented tumorigenesis in a 12 week follow-up after mESCs were injected in nude mice and in 6-OH-dopaminelesioned mice. Intrastriatal injection of MMC-treated mESCs markedly improved motor function without tumor formation. These findings show novel strategies for cell therapies and particularly for treatment of PD (Acquarone et al).

González-Castillo et al. reviewed the role of pleiotrophin (PTN) as a central nervous system (CNS) neuromodulator. In this paper, the authors highlight and summarize the most recent advances and results that lead to proposing a PTN as a neuromodulatory molecule in the CNS, particularly in the hippocampus. Interesting data in CNS demonstrate that PTN exerts post-developmental neurotrophic and -protective effects, and recently has been involved in neurodegenerative diseases and neural disorders (Gonzalez-Castillo et al).

Gonzalez-Lima et al. provided an update on the cellular mechanisms mediating the neuroprotective effects of low doses of methylene blue and near-infrared light. In addition the authors demonstrate how the neurotherapeutic benefits of these two different interventions share the same cellular mechanism of action. Highlighting the similarities in energy transfer, low-dose hormetic dose-responses, and enhanced capacity for oxidative metabolic energy production (Gonzalez-Lima and Auchter).

Szabadfi et al. analyzed the retinal aging phenomena in the diurnal Chilean rodent (Octodon degus). Their results show overexpression of GFAP in Müller glial cells in aging retinas and a reduction in the number of rod bipolar cells and the ganglion cells while that of cone bipolar cells remained unchanged. Their results are very interesting adding novel data to understand the retinal aging process (Szabadfi et al).

Finally, we recognized that a single collection of articles cannot deal with the extremely large number of topics that characterize a complex multifactor condition such as neurodegeneration. The topics addressed, however, help developing a clear idea, not only of what has been obtained to date by previous studies but also of the unmet needs future research should focus on. We trust that the papers assembled in this research topic will prove useful in spurring and stimulating this future progress.

# AUTHOR CONTRIBUTIONS

The author MM and VC prepared the Editorial. All authors: VC, MM, RG, and KC contributed significantly to the review and editing of each manuscript accepted on this topic.

**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.

Copyright © 2016 Meraz-Ríos, Guevara-Guzmán, Carvajal and Campos-Peña. 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.

# Additional mechanisms conferring genetic susceptibility to Alzheimer's disease

*Miguel Calero1,2,3, Alberto Gómez-Ramos1,4, Olga Calero1,2, Eduardo Soriano1,5, Jesús Avila1,4\* and Miguel Medina1,3\**

*<sup>1</sup> Centro de Investigación Biomédica en Red sobre Enfermedades Neurodegenerativas, Madrid, Spain, <sup>2</sup> Chronic Disease Programme, Instituto de Salud Carlos III, Madrid, Spain, <sup>3</sup> Alzheimer Disease Research Unit, CIEN Foundation, Queen Sofia Foundation Alzheimer Center, Madrid, Spain, <sup>4</sup> Centro de Biología Molecular Severo Ochoa CSIC-UAM, Madrid, Spain, <sup>5</sup> University of Barcelona, Barcelona, Spain*

Familial Alzheimer's disease (AD), mostly associated with early onset, is caused by mutations in three genes (*APP*, *PSEN1*, and *PSEN2*) involved in the production of the amyloid β peptide. In contrast, the molecular mechanisms that trigger the most common late onset sporadic AD remain largely unknown. With the implementation of an increasing number of case-control studies and the upcoming of large-scale genomewide association studies there is a mounting list of genetic risk factors associated with common genetic variants that have been associated with sporadic AD. Besides apolipoprotein E, that presents a strong association with the disease (OR∼4), the rest of these genes have moderate or low degrees of association, with OR ranging from 0.88 to 1.23. Taking together, these genes may account only for a fraction of the attributable AD risk and therefore, rare variants and epistastic gene interactions should be taken into account in order to get the full picture of the genetic risks associated with AD. Here, we review recent whole-exome studies looking for rare variants, somatic brain mutations with a strong association to the disease, and several studies dealing with epistasis as additional mechanisms conferring genetic susceptibility to AD. Altogether, recent evidence underlines the importance of defining molecular and genetic pathways, and networks rather than the contribution of specific genes.

#### *Edited by:*

*Victoria Campos-Peña, Instituto Nacional de Neurologia y Neurocirugia, Mexico*

#### *Reviewed by:*

*Rafael Linden, Federal University of Rio de Janeiro, Brazil Tuck Wah Soong, National University of Singapore, Singapore*

#### *\*Correspondence:*

*Miguel Medina and Jesús Avila Centro de Investigación Biomédica en Red sobre Enfermedades Neurodegenerativas, Valderrebollo 5, 28031 Madrid, Spain mmedina@ciberned.es*

> *Received: 23 December 2014 Accepted: 23 March 2015 Published: 09 April 2015*

#### *Citation:*

*Calero M, Gómez-Ramos A, Calero O, Soriano E, Avila J and Medina M (2015) Additional mechanisms conferring genetic susceptibility to Alzheimer's disease. Front. Cell. Neurosci. 9:138. doi: 10.3389/fncel.2015.00138* Keywords: Alzheimer, epistasis, exome, GWAS, neurodegeneration, rare variants, risk factors, somatic mutations

# Introduction

Alzheimer's disease (AD) is characterized clinically by a gradual decline in memory and other cognitive functions and neuropathologically by gross brain atrophy and accumulation of extracellular amyloid plaques and intracellular neurofibrillary tangles. AD is the most common cause of dementia in the elderly, still without effective treatment. The disease has a strong genetic component and, in a small number of cases, AD segregates as an autosomal dominant trait in families. Although uncommon, the identification of these mutations in the last two or three decades has been very critical not only for diagnosing presymptomatic individuals from autosomal dominant families, but also for important advances in understanding AD pathobiology.

The inheritance of AD exhibits a dichotomous pattern. On one hand, early genetic linkage family studies led to the identification of dominantly inherited, rare mutations in the genes for the amyloid β precursor protein (*APP*)*,* presenilin 1 (*PSEN1*), and presenilin 2 (*PSEN2*) that are associated with early-onset AD (EOAD) with a penetrance close to 100%. Mutations in those three genes account for one third to one half of all autosomal dominant cases, which in turn represent less than 1% of total AD cases (reviewed in Guerreiro et al., 2012). On the other hand, the risk for late onset AD (LOAD), the most common form of the disease (*>*95 of total cases), is influenced by common variants such as the *APOE* haplotypes, as discussed below (see **Table 1**).

Over 30 dominant mutations in the *APP* gene (located at chromosome 21q21) account for about 15% of early-onset autosomal dominant cases of AD. Interestingly, two recessive APP mutations, A673V, and E693D, also reportedly cause EOAD (Tomiyama et al., 2008; Di Fede et al., 2009). *APP* encodes a ubiquitously expressed transmembrane protein and most mutations cluster around or within the Aβ domain. Duplications of *APP* and neighboring sequences are also linked to EOAD. Families carrying these duplications exhibit classic AD and cerebral amyloid angiopathy (reviewed in Guerreiro et al., 2012). These copy number mutations display different frequencies depending on the geographic population studied, being much more frequent in Japanese than in Europeans early-onset familial cases (Raux et al., 2005; Sleegers et al., 2006; Blom et al., 2008; Kasuga et al., 2009). Furthermore, patients with Down syndrome, which results from a chromosome 21 trisomy, develop AD neuropathology (Hartley et al., 2014).

Structurally similar integral membrane proteins *PSEN1* and *PSEN2* are part of the γ-secretase protein complex responsible for APP cleavage and Aβ generation (Medina and Dotti, 2003). About 80% of EOAD cases have been reported to carry dominant, pathogenic mutations in *PSEN1* (located at chromosome 14q24.3), whereas approximately 5% have been identified in *PSEN2*, which is located at chromosome 1q31–q42 (Bird, 1999) Mutations in *PSEN1* and *PSEN2* appears scattered throughout the protein, although some clustering can be observed in the transmembrane domains (Guerreiro et al., 2012).

On the other hand, in non-familial, sporadic cases of AD mostly associated with late onset (LOAD), the ε4 allele of the apolipoprotein E (*APOE*) gene has been identified as a major genetic risk factor contributing to the pathogenesis of AD (Corder et al., 1993; Strittmatter et al., 1993). The *APOE* gene is located at chromosome 19q13.2 and encodes a highly pleiotropic glycoprotein (Siest et al., 1995) involved in the transport of cholesterol and other lipids in the periphery and brain (Mahley, 1988). There are three polymorphic alleles (ε2, ε3, and ε4) encoding three isoforms that differ on two amino acid residues (112 and 158): ApoE2, ApoE3 (the most common form), and ApoE4. Presence of the ApoE4 allele increases risk in familial and sporadic EOAD and LOAD, increasing risk threefold when in heterozygosis and up to 15-fold when in homozygosis (Ashford, 2004). The ε4 allele has also a dose-dependent effect on the age at onset. Conversely, the allele ε2 appears to lowers the risk for LOAD and delays age at onset (Corder et al., 1994).

Overall, these four genes account for 30–50% of the inheritability of AD (Jonsson et al., 2012). The advent of genome-wide association studies (GWAS) in recent years has allowed the identification of novel genetic associations. Hence, over 20 genetic loci have been reported to associate with increased susceptibility for LOAD, essentially common variants with a small individual effect on risk (unlike *APOE*), but also recently identified rare susceptibility variants in LOAD with larger effects (see below).

Thus, besides *APOE*, successive GWAS analyses have generated replicable associations with LOAD for several genes: *CLU*, *PICALM*, *CR1*, *BIN1*, *CD33*, *MS4A* cluster, *CD2AP*, *EPHA1*, and *ABCA7* (reviewed in Karch et al., 2014). A recent meta-analysis of GWAS data from four large consortia confirmed these previous associations (except for *CD33*) and reported 12 new susceptibility loci for AD: *CASS4*, *CUGBP-CELF1*, *DSG2*, *FERMT2*, *HLA-DRB5-DRB1*, *INPP5D*, *MEF2C*, *NME8*, *PTK2B*, *SLC24A4- RIN3*, *SORL1*, *ZCWPW1* (Lambert et al., 2013; **Table 1**). These genes increase AD risk in a non-Mendelian fashion, but firstdegree relatives of LOAD patients have twice the expected risk of AD and LOAD is more frequent in monozygotic than in dizygotic co-twins (Reitz and Mayeux, 2014). However, the observed risk or protective effects of all the single nucleotide polymorphisms (SNPs) tagging these 21 loci are rather small, with odds ratio (OR) ranging from 1.22 to 0.77 (Karch et al., 2014), and the genetic effect attributable to each of these associated loci had population-attributable fractions or preventive fractions between 1.0 and 8.0% (Lambert et al., 2013). Although these data are of great value to delineate the fundamental physiopathological avenues of the disease, they only explain a small proportion of familial clustering (Manolio et al., 2009). Therefore, strong additional efforts in sequencing and post-GWAS analyses have to be put forward to find the remaining missing heritability in order to completely undercover the genetics of AD, utterly aiming at enabling effective prevention, prediction and treatment of the disease (Manolio et al., 2009; Lambert et al., 2013).

A number of recent excellent reviews have focused on the analysis of common genetic risk factors for LOAD and their role in pathogenesis (Guerreiro et al., 2013a; Lambert et al., 2013; Karch et al., 2014; Reitz and Mayeux, 2014). The genes identified can be classified in a few pathways, mainly lipid metabolism, immune response, endocytosis (Karch and Goate, 2015). Despite the identification of all the above-mentioned loci associated with AD, a large proportion of the genetic component of the disorder remains unexplained (Lord et al., 2014). Using alternative AD phenotypes may serve as a tool to unveil additional genes that could modify particular aspects of the disease (Karch and Goate, 2015). In here, we review additional mechanisms that may at least partially explain this missing heritability such as epistasis, rare variants, or the presence of somatic mutations. In this sense, this review does not intend to comprehend an exhaustive analysis of the literature on these subjects, but to put forward important concepts and the general evolution of the field. The analysis of epigenetic modifications or microRNA studies are out of the scope of the present review.

# Rare Variants

After a decade of intense efforts on GWAS analysis, lately followed by meta-analysis of studies performed on very large

#### TABLE 1 | AD-associated genes.


∗*GRN, MAPT, and PRNP mutations have been found in some individuals with clinical phenotypes indistinguishable from AD (Lee et al., 2014). Variants of genes associated with Alzheimer's disease (AD) can be classified according to their frequency as rare or common variants. Rare variants are commonly associated with early onset AD, while common variants are associated with late onset Alzheimer's disease (LOAD).*

cohorts, a number of common variants have been identified to have replicable, although small effects on LOAD with no clear functionality in some cases (Cruchaga et al., 2014); and it is unlikely that many common variants with moderate-large effects remain to be identified (Ridge et al., 2013). However, new rare functional variants, with larger effects are expected to be associated with AD. Thus, researchers from the AD Genetics Consortium argue that more than 25% of phenotypic variance remains unexplained by known markers, but it is tagged by common SNPs, hence suggesting that novel AD markers that account for large amounts of phenotypic variance are likely to be rare (Ridge et al., 2013). Thus, both rare and common variants contribute to AD risk (Guerreiro et al., 2013a; see **Table 1**). However, GWAS are not well suited for the discovery of these rare variants, and the use of new techniques such as exome analysis or whole genome sequencing will be required. Genome and exome sequencing studies in large data sets are likely to add new genes (with a moderate or low association). However, it remains to be seen whether additional pathways can be identified (Karch and Goate, 2015).

Besides the well-characterized mutations on *PSEN1*, *PSEN2,* and *APP* found in pedigrees of familiar EOAD, the advent of new powerful tools for genome analysis is already delivering a more comprehensive identification of rare variants potentially associated with the disease. For example, whole-genome sequence analysis from 1,795 Icelanders has led to the identification for the first time of a coding mutation (A673T) in *APP* that protects against AD and cognitive decline (Jonsson et al., 2012). In a different study, exome sequencing analysis has identified two novel pathogenic *PSEN1* mutations (p.L166V and p.S230R) in British EOAD (Sassi et al., 2014a), although a similar study did not identify novel variants in AD in an Asian population (Chung et al., 2014).

Additionally, rare variants in other genes previously not associated with familiar EOAD have been found to have a high to moderate effect size. Thus, two rare, highly penetrant mutations in *ADAM10* (p.Q170H and p.R181G) for LOAD have been reported (Kim et al., 2009), emphasizing the importance of whole genome or whole exome sequencing approaches to find rare variants causing LOAD, in addition to common variants (Jonsson et al., 2012).

Following whole-exome sequencing and whole-genome sequencing strategies, a rare variant in the *TREM2* gene (p.R47H) has also been associated with an increased risk of AD with an OR of 3.4 (Guerreiro et al., 2013b; Guerreiro and Hardy, 2013; Jonsson et al., 2013). Recently, two more *TREM2* variants have been associated with either increased (p.R62H) or decreased (p.S144G ) risk in AD (Benitez et al., 2014; Cuyvers et al., 2014). Interestingly, *TREM2* also appears to be associated with Parkinson's disease, frontotemporal dementia (Rayaprolu et al., 2013; Le Ber et al., 2014), and amyotrophic lateral sclerosis (Guerreiro and Hardy, 2013), although this remains controversial (Slattery et al., 2014).

Applying whole exome sequencing with a family-based design aimed at detecting novel AD risk genes, several rare, missense, and synonymous variants in phospholipase D3 (*PLD3*) have been reported to be associated with AD risk (Cruchaga et al., 2014). One variant in *PLD3* (p.V232M), which segregated with some LOAD families appears to increase two- to threefold the risk for AD, perhaps through its influence on APP processing (Cruchaga et al., 2014).

Whole exome sequencing analysis performed on seven African American AD cases, (Logue et al., 2014) has led to find two new rare variants in *AKAP9 (*A-kinase anchor protein nine gene) potentially associated with AD, that in the replica population from the AD Genetics Consortium showed a strong association (rs144662445, OR = 2.75; rs149979685, OR = 3.61).

Recently, the Ibero-American Alzheimer Disease Genetics Group Researchers analyzed the coding region and flanking sequences of *APP*, *PSEN1*, *PSEN2*, *MAPT,* and *GRN* by pooled-DNA exon sequencing in 167 clinical and five autopsy-confirmed, mainly EOAD cases (Jin et al., 2012). Interestingly, pathogenic mutations in *PSEN1*, *GRN,* and *MAPT* genes were found in 2.3% of the screened cases, suggesting that pathogenic mutations or risk variants in MAPT and in GRN are as frequent in clinical AD cases as mutations in *APP*, *PSEN1,* and *PSEN2* (Jin et al., 2012). These findings underscore the pleotropic role of genes such as *MAPT* and *GRN* that can influence both frontotemporal dementia and AD (Jin et al., 2012; Lee et al., 2014; **Table 1**). Likewise, missense or nonsense haplotypes in *PRNP* (p.I215V or p.Q160) further highlight how very similar genotypes in *PRNP* result in strikingly different clinical phenotypes such as prion diseases and AD (Muñoz-Nieto et al., 2013; Guerreiro et al., 2014).

In contrast, the role of rare coding variability in Mendelian inherited dementia genes (*APP*, *PSEN1*, *PSEN2*, *GRN*, *MAPT*, and *PRNP*) has also been investigated in LOAD. Results indicated that rare coding variability in *PSEN1* and *PSEN2* may influence the susceptibility for LOAD, while *GRN*, *MAPT*, and *PRNP* do not appear to be major contributors to LOAD (Sassi et al., 2014b).

# Somatic Mutations

Advanced age is the single most important risk factor for AD. Randomly acquired DNA damage in the nuclear genome is also associated with aging, and post-mitotic neuronal tissue is at special risk of DNA damage due to the elevated production of DNAdamaging reactive oxygen species (ROS) associated with their high metabolism (Barja, 2004; Kennedy et al., 2012). Moreover, the mutation rate of somatic cells is roughly an order of magnitude higher than mutation rates for germ-line cells (Lynch, 2010). On the other hand, evolution cannot put selective pressure on those deleterious mutations that produce defects long after the age of reproduction as it is the case of AD (Medawar, 1952; Kennedy et al., 2012). According to the so called somatic mutation theory of aging, the accumulation of mutations in the genetic material of somatic cells as a function of time results in a decrease in cellular function. A significant amount of research has shown that somatic mutations play an important role in aging and a number of age-related pathologies (Jeppesen et al., 2011; Kennedy et al., 2012). Additionally, some genetic traits may present antagonistic pleiotropy as in the case of *APOE* ε4 variant that appears to be associated with improved perinatal health and survival, favoring the selection of the ε4 allele despite its later life deleterious effects (reviewed in Eisenberg et al., 2010). With this perspective, it is not surprising that somatic cells, especially in the brain, may accumulate a much higher proportion of mutations in certain DNA hot spots (e.g., CpG dinucleotides) that correlate (at a lower scale) with the reported occurrence of pathogenic mutations in germinal-cell lines. The effect of these somatic mutations will depend on stochastic processes that allow the generation of cellular damage that may also propagate to other cells (Aguzzi and Rajendran, 2009). Therefore, in a wide sense, somatic mutations could be considered as rare variants locally generated.

Recent technological advances in the field of genomics and the emergence of next generation sequencing (NGS) techniques have aggravated the difficulties of interpreting GWAS to reveal the genetic basis of brain disorders. Despite a wealth of data on candidate genes affecting the susceptibility to AD and other neurological disorders, their inherent contribution to the pathogenesis and their relationship with non-genetic environmental risk factors is not yet fully understood. Evidence has accumulated showing that somatic variations can affect neuronal populations and may play a role in brain pathogenic processes. Thus, it has been suggested that, at least in some cases copy number variations (CNV) or SNP linked to major AD and other neurological disorders may lead to genetic dysregulation resulting in somatic mosaicism (Iourov et al., 2013). Moreover, environmental factors may affect cellular repair mechanisms and trigger a series of events that end up generating genomic instability.

Increasing evidence has also shown that, even at low levels of mosaicism, somatic mutations can cause neuropsychiatric, and pediatric disorders such as epilepsy, autism, lissencephaly, and intellectual disability. Interestingly, at least some of these cases appears to have been caused by somatic mutations limited to the brain (Poduri et al., 2013). In addition, a number of reports have shown that a significant population of neurons exhibit aneuploidy in both mice and humans and the level of aneuploidy could be as high as 1–3% in adult brain (Rehen et al., 2005). Furthermore, brain tissue from both AD and other disorders exhibits increased levels of aneuploidy when compared to unaffected brain (Iourov et al., 2009; Frade and López-Sánchez, 2010).

Furthermore, a recent study has investigated the presence of tissue-specific exonic single nucleotide variations (SNV), taking blood exome as a control (Gómez-Ramos et al., 2014). Interestingly, the number of SNV per chromosome was independent of chromosome size, but it was suggested to mainly relate to the number of protein-coding genes per chromosome. Although similar patterns of chromosomal distribution of tissue-specific SNVs were found, clear differences were also detected, supporting the notion that each individual tissue has a specific SNV exome signature.

Some neurodegenerative disorders have been associated with or can be affected by somatic mutations. Thus, EOAD has been attributed to a somatic mosaic *PSEN1* mutation present in the brain (Beck et al., 2004). Similarly, a case of sporadic Creutzfeldt– Jakob disease has been reported to be caused by an early embryonic somatic mutation in the *PRNP* gene (Alzualde et al., 2010).

Also, patients of disorders caused by the expansion of highly mutable repeats such as Friedrich's ataxia or Huntington's disease, can exhibit notable somatic heterogeneity in repeat lengths across different brain regions and tissues (Kahlem and Djian, 2000; Hellenbroich et al., 2001; Møllersen et al., 2010). Agerelated somatic mutations, known to play a role in cancer (Jacobs et al., 2012), have also been proposed to be involved in normal aging processes as well as in neurodegeneration (Kennedy et al., 2012). Interestingly, very recent whole-exome sequencing analyses of various tissues from sporadic AD patients have found a remarkably high number of brain-specific SNV in AD hippocampal samples when compared with blood (Parcerisas et al., 2014), suggesting that somatic genetic mosaicism and brainspecific genome reshaping may contribute to the pathogenesis of sporadic AD.

As NGS techniques become more efficient and affordable in the coming years, somatic mosaic mutations could be more readily detected, and somatic mutation rates could be systematically determined across different regions, cell types, and time points of the human brain development. Although the limitation of brain tissue will still limit studies on brain-specific somatic mutations, novel techniques, and improved bioinformatic analyses will allow us to address the role of somatic mosaicism in neurological conditions. Determining whether brain somatic mutation are at least partially responsible for AD and other neurodegenerative disorders will be one of the next major challenges in the field.

# Epistasis

The functional expression of certain genes may be determined by the interaction with other genes, as well as with environment factors. Therefore, some of the missing heritability may be conditioned by epistasis, a crucial factor to understand and interpret genetic pathways and their functional role in complex systems. Epistasis may refer to different phenomena, including the functional interaction between genes or the statistical deviation from additive gene action, among others (Phillips, 2008), yet in the context of this review, we will use the term epistasis as synonym of functional gene–gene interaction. However, GWAS, whole genome and exome sequencing analyses have been mainly focused on standard single-locus tests with different level of precision. With the increasing availability of genetic data delivered by the new genomic technologies, it is becoming clearer the need to address the problems from a holistic position, taking into account the interaction among different players at different levels.

The analysis of epistasis, even if limited to the interaction between two genes, is technically demanding and methodologically limited at present. Part of the challenge for epistasis analysis in GWAS is the magnitude of the search and the computational complexity associated with it (Ritchie, 2015). The methods and related software packages used to detect the interactions between genetic loci that contribute to human genetic disease have been recently re-examined (Cordell, 2009). Lately, several studies have reviewed novel advances in the methodology for detecting epistasis (Ma et al., 2013; Wei et al., 2014) and discussed its relevance in the context of GWAS (Wei et al., 2014). A protocol for exhaustive genome-wide association interaction analysis and applied to AD has been proposed, finding replicable epistasis between the *KHDRBS2* and *CRYL1* gene loci (Gusareva et al., 2014).

A powerful tool to disentangle the complexity of a disorder such as AD is the use of the concept of endophenotypes to define genetic association elements with more elementary phenomena rather than with the whole spectrum of complex diagnostic entities. Therefore, markers of disease (at any level) that correlate with a genetic trait may be useful to explore the pathophysiological mechanisms related to specific aspects of the syndrome (Cannon and Keller, 2006).

Synergy factor (SF) analysis has been used to assess over 100 claims of epistasis in sporadic LOAD, finding 27 gene–gene interactions that were significantly associated with AD (Combarros et al., 2009). Additionally, the authors demonstrated by metaanalysis the interaction of *APOE* ε4 with specific variation in four different genes, namely *ACT*, *BACE1*, *IL6*, and *BCHE* (Combarros et al., 2009). More recently, a systematic review of genetic studies published between 2009 and 2012 by the Genetics Core of the AD Neuroimaging Initiative (ADNI) focused on genetic associations with disease status or quantitative disease endophenotypes including structural and functional neuroimaging, fluid biomarker assays, and cognitive performance. In this review, the authors summarize the association of several AD risk genes with imaging, fluid and cognitive phenotypes, and indicated the association with multiple ADNI phenotypes of several other genes (i.e., *APOC1*, *FTO*, *GRIN2B*, *MAGI2*, and *TOMM40*). The authors suggested the combination of genetic data and phenotypes for targeting future studies employing NGS and convergent multi-omics approaches, and for clinical drug and biomarker development (Shen et al., 2014). Likewise, the use of neuroimaging measures to define powerful quantitative endophenotypes for exploring epistatic relationships in order to explain some of the missing heritability in AD has been proposed (Hohman et al., 2013). Moreover, the analysis of epistasis has also been shown to be a key factor for genomic prediction and its application in preventive strategies through the identification of population strata at increased risk of disease and clinical decision making (Wray et al., 2013).

Many studies are now focused on finding how genetic variants influence specific traits of AD pathology, such as amyloid burden (Kauwe et al., 2011, 2014; Bali et al., 2012; Hohman et al., 2013; Shen et al., 2014), neurofibrilar pathology (Kauwe et al., 2011; Shen et al., 2014), brain atrophy (Shen et al., 2014), cognitive decline (Pedraza et al., 2014), or inflammation (Kauwe et al., 2014). However, the complexity and heterogeneity of AD requires further analysis going beyond the study of single markers in isolation by taking into account interactions between genes. For example, many of the LOAD risk genes do not show single marker associations with amyloid pathology, but only through interaction with other genes (e.g., *BIN1* × *PICALM*; Hohman et al., 2013).

Therefore, it is important to note that epistasis may be a barrier to uncover the genetic basis of complex disorders, since the effects of quantitative trait loci can be masked by interactions with other loci (Phillips, 2008). As demonstrated for *GSTM3* gene and the *HHEX*/*IDE*/*KIF11* locus, the association with AD could be purely epistatic with neither polymorphism showing an independent effect (Bullock et al., 2013). The same group, involved in the Epistasis Project, has also reported an interaction between transferrin and *HFE* genes, confirming previous findings (Robson et al., 2004) and suggesting that iron overload may be involved in the development of AD (Lehmann et al., 2012).

Arguably, *APOE* is the main genetic risk factor in AD and it is likely to interact with other genes. The epistatic effect for *APOE* has been recognized both in familial EOAD and LOAD modifying the age at onset (Sorbi et al., 1995; Combarros et al., 2009; Hohman et al., 2013). Using whole-trascriptome analysis of brain gene expression, It has been demonstrated that *APOE* ε*4* carrier status was associated with a consistent transcriptomic shift that resembled the LOAD profile (Rhinn et al., 2013). However, unlike the genes underlying familial AD, LOAD susceptibility genes do not specifically alter the Aβ42/40 ratios, suggesting that these genes probably contribute to AD through distinct mechanisms (Bali et al., 2012). Interestingly, a recent study (Naj et al., 2014) has also demonstrated an association of *APOE* variants with age at onset among affected individuals with LOAD and observed novel associations of *CR1*, *BIN1*, and *PICALM* with age at onset.

Along the same lines, we have explored epistasis and pleiotropic effects of various genes and neurodegenerative disorders. AD and Creutzfeldt-Jakob disease (CJD) are now considered both as part of a wider group generically named as conformational disorders, and more specifically brain amyloidosis. Under the assumption that these two disorders share common pathophysiologic mechanisms involving protein aggregation in the brain leading to fatal degeneration, we investigated a possible genetic interaction between *APOE* ε4 allele and the polymorphic codon 129 of the *PRNP* gene in both AD and CJD populations compared to a common control population (Calero et al., 2011). Interestingly, we found a synergistic agedependent interaction between the two genes (*APOE* × *PRNP*) in both disorders (SF = 3.59, *p* = 0.027 for AD; and SF = 7.26, *p* = 0.005 for CJD). Our data suggest the involvement of common pathways involved in the generation, clearance, and neurotoxic signal transduction of Aβ peptides and PrP in AD and CJD. The finding of an age-dependent interaction underscores the importance of genetic risk analysis stratified according to other potential interacting/confounding factors such age, sex, or comorbidities.

Finally, the comprehensive analysis of high order interactions is limited by the exponential number of potential interactions and our technical capacity; therefore, we need to focus on particular subsets of the interaction space by using additional functional information (Phillips, 2008). However, the analysis of epistasis should not be limited only to candidate genes.

# Conclusion

Over 20 loci have been associated with LOAD, defining three main routes altered in AD: lipid metabolism, immune response, and endocytosis (see **Table 1**). Altogether, these association analyses to different genes by GWAS underscore the importance of defining pathways and networks rather the contribution of specific genes. Future research should be focused on defining shared and specific mechanisms among distinct neurodegenerative and other chronic disorders including cardiovascular disease, metabolic disorders, or even cancer. Further insights into complex disorders such as AD are expected from the integration of different -omics with detailed high-quality clinico-physiological characterization of cases and controls allowing the development of new disease biomarkers and therapeutic avenues, and will enable the implementation of personalized medicine (Ramanan and Saykin, 2013). The identification by GWAS of multiple disease-associated loci of small effect size emphasizes the polygenic nature of the heritability of complex traits and common disorders such as AD, giving rise to approach the disorder as of quantitative dimensions instead of as a qualitative disorder (Plomin et al., 2009).

However, a large proportion of genetic component of the disorder remains unexplained (Lord et al., 2014). As reviewed above, epistasis, rare variants, and the presence of somatic mutations are additional genetic mechanism that may explain in part this missing heritability. Future research should emphasize: (i) the

# References


complete characterization of the genetic risk factors including common-low effect size and rare-high effect size variants, (ii) the use of endophenotypes to associate specific traits of the disease to genetic factors and use this information for risk prediction and definition of treatment response groups, (iii) the analysis of epistasis among different risk genes and the interaction with other factors such as age and sex, and (iv) the potential pleiotropism associated with certain loci that may confer either risk or protection to AD and may transversally regulate different neurodegenerative disorders. From a practical point of view, it is important to acknowledge that most AD patients usually present a combined pathology with other chronic disorders including cardiovascular disease or metabolic disorders such as diabetes mellitus type 2; and therefore we should investigate how these pathologies interact with each other, in order to properly treat each individual patient.

# Funding

This work was supported by CIBERNED, CIEN and Reina Sofia Foundations and the Carlos III Health Institute (PI12/00045).


**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.

*Copyright © 2015 Calero, Gómez-Ramos, Calero, Soriano, Avila and Medina. 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.*

# Epigenetic mechanisms in neurological and neurodegenerative diseases

#### **Jorge Landgrave-Gómez , Octavio Mercado-Gómez and Rosalinda Guevara-Guzmán\***

Facultad de Medicina, Departamento de Fisiología, Universidad Nacional Autónoma de México, México, D.F., México

#### **Edited by:**

Victoria Campos-Peña, Instituto Nacional De Neurologia Y Neurocirugia, Mexico

#### **Reviewed by:**

Alexander K. Murashov, East Carolina University, USA Jose F. Maya-Vetencourt, Italian Institute of Technology, Italy

#### **\*Correspondence:**

Rosalinda Guevara-Guzmán, Facultad de Medicina, Departamento de Fisiología, Universidad Nacional Autónoma de México, Ave. Universidad # 3000 Col. UNAM Delegación Coyoacán, Apartado Postal No. 70250, 04510 México, D.F., México e-mail: rguevara@unam.mx

The role of epigenetic mechanisms in the function and homeostasis of the central nervous system (CNS) and its regulation in diseases is one of the most interesting processes of contemporary neuroscience. In the last decade, a growing body of literature suggests that long-term changes in gene transcription associated with CNS's regulation and neurological disorders are mediated via modulation of chromatin structure. "Epigenetics", introduced for the first time by Waddington in the early 1940s, has been traditionally referred to a variety of mechanisms that allow heritable changes in gene expression even in the absence of DNA mutation. However, new definitions acknowledge that many of these mechanisms used to perpetuate epigenetic traits in dividing cells are used by neurons to control a variety of functions dependent on gene expression. Indeed, in the recent years these mechanisms have shown their importance in the maintenance of a healthy CNS. Moreover, environmental inputs that have shown effects in CNS diseases, such as nutrition, that can modulate the concentration of a variety of metabolites such as acetyl-coenzyme A (acetyl-coA), nicotinamide adenine dinucleotide (NAD+) and beta hydroxybutyrate (β-HB), regulates some of these epigenetic modifications, linking in a precise way environment with gene expression. This manuscript will portray what is currently understood about the role of epigenetic mechanisms in the function and homeostasis of the CNS and their participation in a variety of neurological disorders. We will discuss how the machinery that controls these modifications plays an important role in processes involved in neurological disorders such as neurogenesis and cell growth. Moreover, we will discuss how environmental inputs modulate these modifications producing metabolic and physiological alterations that could exert beneficial effects on neurological diseases. Finally, we will highlight possible future directions in the field of epigenetics and neurological disorders.

**Keywords: epigenetics, neurodegeneration, DNA methylation, postranslational modification, Parkinson disease, epilepsy**

#### **EPIGENETICS**

The term epigenetics is derived from the theoretical and experimental work of Conrad Waddington. He coined the term to describe a conceptual solution to a phenomenon that arises as a fundamental consideration of developmental biology (Waddington, 1942). All of the different cells in the body of one individual have exactly the same genome, that is, exactly the same DNA nucleotide sequence, with only a few exceptions in the reproductive, immune and nervous systems. Thus, in the vast majority of instances, one's liver cells have exactly the same DNA as neurons. However, those two types of cells are clearly vastly different in terms of the gene products that they produce. Some level of mechanism must exist, Waddington reasoned, that is "above" the levels of genes encoded by the DNA sequence, which controls the DNA readout. For this reason, he defined the term *epigenetics* in the early 1940s as "the branch of biology which studies the causal interactions between genes and their products which bring the phenotype into being" (Waddington, 1968). In the original sense of this definition, epigenetics is referred to all molecular pathways modulating the expression of a genotype into a particular phenotype.

However, and with the fast expansion in this field, epigenetics has been redefined and accepted today as "the study of changes in gene function that are mitotically and/or meiotically heritable and that does not entail a change in DNA sequence." In this way, recent advances have evolved our understanding of classical epigenetic mechanisms and the broader landscape of molecular interactions and cellular functions that are inextricably linked to these processes. The current view of epigenetics includes the dynamic nature of DNA methylation, active mechanisms for DNA demethylation, differential functions of 5-methylcytosine and its oxidized derivatives, the intricate regulatory logic of histone post-translational modifications, the incorporation of histone variants into chromatin, nucleosome occupancy and dynamics. Nevertheless, of all these modifications, the mechanisms better described in literature generally comprise histone variants, posttranslational modifications of amino acids on the aminoterminal tail of histones, and covalent modifications of DNA bases.

In this chapter, we will discuss some of these epigenetic modifications and how these modifications are associated with neurologic homeostasis and diseases.

#### **LINKING THE ENVIRONMENT, NUTRITION AND EPIGENETIC MODIFICATIONS**

Although many aspects of nutrition and different kinds of lifestyles influence metabolic status and disease trajectory throughout our life, emerging findings suggest that changing our metabolism with exercise or different dietary regimens such as ketogenic diets, low-carbohydrate diets, intermittent fasting or physical exercise can alter the concentration of a variety of metabolites, some of them capable of modulating the activity of proteins that elicit epigenetic modifications (**Figure 1**; Shimazu et al., 2013; Shyh-Chang et al., 2013).

These epigenetic modifications seem to regulate important networks of genes mediating physiological processes associated with the beneficial effect of these diets, providing a rationale and simple way to prevent or even treat these diseases. Some reports have shown the efficacy of exercise and diet in cancer; cardiovascular disease, diabetes, obesity, rheumatoid arthritis and even in some neurological/neurodegenerative diseases such as Alzheimer and epilepsy (Müller et al., 2001; Ahmet et al., 2005;

Belkacemi et al., 2012; Kroeger et al., 2012; Lee et al., 2012; Varady et al., 2013; Colman et al., 2014).

Consistently, some reports have shown that aging it's a process that may be altered through some diets, such as calorie restriction (Colman et al., 2014). The precise mechanisms of how environment mediates epigenetic modifications are not clearly understood, however in this manuscript we will portray some studies that aim to epitomize the relationship between environment, metabolism, epigenetics and neuro -logical/neurodegenerative diseases.

#### **EPIGENETIC MODIFICATIONS**

Within cell nucleus, the fundamental units of chromatin are named nucleosomes. Each nucleosome is formed by 147 DNA base pairs wrapped tightly around an octamer of histone proteins, which is assembled by two copies of each of the four core histones (H2A, H2B, H3 and H4). The linker histone H1 binds to the DNA between the nucleosomal core particles, and their function is to stabilize higher order chromatin structures. Moreover, each histone protein consists of a central globular domain and N-terminal tail that contains multiple sites for potential modifications (Wang et al., 2013).

In this regard, a variety of different modifications on amino acid residues of histones have been described. Histone posttranslational modifications include acetylation, methylation, phosphorylation, ubiquitination and sumoylation (**Table 1**; Sassone-Corsi, 2013).

The principal residues that are substrates of these modifications are lysine, arginine, serine and threonine amino acids (Rothbart and Strahl, 2014). These modifications have been associated to repression or activation of gene transcription depending on the site of the modification, strongly suggesting the existence of a histone code. This hypothesis proposes that specific modifications of histones induce to the interaction with proteins associated with the chromatin, producing a differential regulatory response of gene expression (Strahl and Allis, 2000; **Table 1** and **Figure 2**). These modifications are dynamic in the way that they are actively added and removed by histone-modifying enzymes in a site-specific manner, which is essential for coordinated transcriptional control.



charges of DNA. **(C)** On the other hand different specific marks of methylation of histone 3 are associated with both transcriptional activation **(C)** and repression **(D)**. Also a specific mark of phosphorylation on the (S10) amino acid of histone 3 has been associated with transcriptional activation **(E)** so the lack of this mark may be associated with transcriptional repression **(F)**.

# **HISTONE ACETYLATION**

The acetylation of histones is a modification associated generally to transcriptional activity that indicates access of the transcription machinery to the genes and thus active mechanisms (Strahl and Allis, 2000; Balazs, 2014). This effect could be explained by the chemistry of this modification in which an acetyl group (−COCH3) is incorporated to an amino terminal residue and thus, the positive charge of histones is reduced, inducing a minor interaction with DNA resulting in a decrease of the chromatin compaction (**Figures 2A,B**; Shahbazian and Grunstein, 2007).

# **HISTONE METHYLATION**

Histone methylation is currently associated with multiple processes such as transcriptional activation and repression, depending on the modified amino acid residue (**Figures 2C,D**). This modification occurs mainly on arginine and lysine residues. Additionally, these residues could be methylated multiple times giving different signals depending on how many times the residue is methylated, making its analysis difficult. In this regard, current literature has shown that lysine residues can be methylated even three times; meanwhile, arginine residues can only be methylated twice (Strahl and Allis, 2000). Furthermore, there have been some studies associating some processes with these types of modifications for example H3K4, H3K36 and H3K79 are associated with chromatin aperture. Nevertheless, the methylation of these residues has been also associated with other specific functions. On the other hand, H3K4 trimethylation has been associated with promoter regions. The monomethylathion of this same residue recruits regulatory elements that potentiate the promoter activity; such elements are known as *enhancers*. Dimethylation of H3K36 has been related to RNA POL II elongation during transcription (Li et al., 2007). Also, the dimethylation of H3K79 is particular of promoter regions stimulating a permissive chromatin for local transcription (Jacinto et al., 2009). Conversely, the modifications associated with transcriptional repression are performed on H3K9 and H3K27 residues (Baylin and Jones, 2011).

# **DNA METHYLATION**

In mammalians, DNA methylation is the covalent union of methyl groups of cytosines that are found mainly in the context of dinucleotide 5<sup>0</sup> -CpG-3<sup>0</sup> (**Figure 3A**; Klose and Bird, 2006). The addition of methyl groups protrudes above the major groove and when DNA is symmetrically methylated, the methyl groups promote a conformational change of DNA structure. The main consequence of methyl modification is that a variety of transcription factors cannot recognize the DNA and thus induce repressional transcription (Prokhortchouk and Defossez, 2008).

DNA methylation generates patterns that are established during embryonic development and such patterns are maintained by a mechanism when DNA replicates (**Figure 3**). Interestingly, these patterns change over time, principally due to environmental factors (i.e., nutrition, metabolites, exercise, chemical agents) (Fraga et al., 2005). The mechanism of DNA methylation is carried out by a set of proteins named DNA methyltransferases (DNMTs). There are two groups of these proteins; (1) one for *de novo* methylation; and (2), one for methylation maintenance. Both enzymes differ depending on the DNA substrate: for example, maintenance of DNA methylation is accomplished by DNA methyl transferase 1 (DNMT1). These proteins add methyl groups to pre-existing methyl patterns on a new strand of DNA during replication (**Figure 3B**; Jeltsch, 2006). On the other hand, *de novo* DNA methylation are carried out by DNMT3a and DNMT3b. Such proteins are responsible for the addition of new methyl groups to cytosines that have not been methylated previously (Goll and Bestor, 2005).

maintenance of DNA methylation is accomplished by DNA methyl-transferases (DNMT1) when DNA replication occurs **(B)**.

#### **HISTONE VARIANTS**

Histone variants such as H2A and H3.3 have been known since several decades ago and recently, a lot of evidence has been accumulated about their role in their participation on the differential structure of chromatin (Henikoff et al., 2004). Among them, H2A.Z has been located on DNA regions associated with transcriptional activation, mainly, on promoter regions. This variant is important because it induces a less stable structure of chromatin compared with that of the canonical histone H2 (Draker and Cheung, 2009). Another histone variant associated with promoter regions is H3.3. This variant as well as H2A.Z, is mainly found on promoter regions suggesting that their structure promotes the formation of a more permissive chromatin (Jin et al., 2009).

#### **NEUROEPIGENETICS AND THEIR ROLE IN NEURONAL FUNCTION**

Over the last two decades, the field of epigenetics, particularly the emerging field of neuroepigenetics, has begun to have a great impact in different areas such as the study of the CNS development, learned behavior, neurotoxicology, cognition, addiction and lately neurological and neurodegenerative pathology (Sweatt, 2013). In this regard, epigenetics has undergone an exponential expansion. A quick search of the PubMed database reveals that about 98% of all the research work on epigenetics was published within the last 15 years (Sweatt, 2013). Thanks to these studies, nowadays we know that either maternal behavior, environmental toxins, nutrition, physociological or physical stress, learning, drug exposure or psychotrauma, leads to active regulation of the chemical and three-dimensional structure of DNA and thus, regulates epigenetics modifications in the CNS linking environmental *stimuli* and gene expression regulation (Tsankova et al., 2007; Borrelli et al., 2008; Renthal and Nestler, 2008; Champagne and Curley, 2009; Day and Sweatt, 2010; Dulac, 2010).

These epigenenomic changes allow perpetual alterations in gene readout in cells in the CNS affecting neuronal function and physiology. For example, a central regulator of homeostasis in the brain, the brain-derived neurotrophic factor (BDNF), a member of the neurotrophin family of proteins that plays crucial roles in the development, maintenance, and plasticity of the CNS (Chao et al., 2006) have been demonstrated to play an important role on different psychiatric disorders associated with early-life adversity, including depression; schizophrenia, bipolar disorder and autism. Even when the underlying mechanisms of the alterations over the expression of BDNF are unknown in these conditions, epigenetic modifications seem as a plausible candidate, as early-life exposures, chronic emotional *stimuli*, or even emotional behavior, disrupts epigenetic programming in the brain with lasting consequences for gene expression and behavior (Renthal et al., 2007; LaPlant et al., 2010; Kundakovic et al., 2014).

However, epigenetics is such a new field of science that in most of the cases, its impact has not been fully demonstrated. Even though, it is now clear that there is a dynamic interplay between genes and experience, a clearly delineated and biochemically driven mechanistic interface between genes and environment, this interface is epigenetics (Sweatt, 2013).

#### **ALZHEIMER'S DISEASE AND EPIGENETICS**

Alzheimer's disease (AD) is an age-related and slowly neurodegenerative disorder of the brain and the most common form of dementia in the elderly (Sezgin and Dincer, 2014). The disease is clinically characterized by progressive memory loss and cognitive impairment. Moreover, the histopathological features of AD are senile plaques composed of amyloid beta (Aβ) fibrils and neurofibrillary tangles composed of microtubule-associated protein tau, combined with massive cholinergic neuronal loss, mainly in the hippocampus and association regions of neocortex (Hardy, 2006; Ballatore et al., 2007). This disease currently affects approximately 2% of the population in industrialized countries and its incidence will increase dramatically over the time (Sezgin and Dincer, 2014).

AD is a multifactorial disease involving; genetic, metabolic, nutritional, environmental and social factors that are associated with onset and progression of the pathology. For this reason, and considering that the main risk factor of this disorder is aging, it is reasonable to think that life history such hypertension, diabetes, inflammation, obesity or head injury are closely related with AD (Marques et al., 2011). However, how these factors induce epigenetic changes that mediate the network genes involved in this disease is a question that remains to be answered.

At present, studies of epigenetic changes in AD are starting to emerge. As we mentioned before, aging is the most important risk factor for AD an epigenetic changes have been observed in aging tissues. Recently, it has been observed that environmental factors even transient ones in early life can induce AD-like pathogenesis in association with aging (Wu et al., 2008a). Furthermore, a difference in DNA methylation patterns typical of brain region and aging has been identified in this context (Balazs, 2014). In this regard, a recent study by Hernandez et al. examined the DNA methylation patterns in >27,000 CpG sites from donors ranging in age 4 months to 102 years and a strong relationship was found between DNA methylation and aging. Moreover, in the temporal and frontal cortices pons and cerebellum regions, more than 1,000 associations were found between DNA methylation at CpG sites and age and some associations were significant in all four regions. Interestingly, the majority of the association sites were in CpG islands and the pattern was similar in the frontal cortex, temporal cortex and pons, but different in cerebellum. These results suggest that and age-dependent increase in DNA methylation may be important for maintaining gene expression with age (Hernandez et al., 2011).

As it has been reported in many studies, memory can be compromised during aging. Preclinical and basic studies have shown that epigenetic mechanisms are involved in formation and maintenance of memory (for reviews, see Levenson and Sweatt, 2005; Zovkic et al., 2013; Jarome et al., 2014). For example, inhibition of DNA methylation has deleterious effects on neuronal plasticity together with histone modifications (Day and Sweatt, 2011; Zovkic et al., 2013). Moreover, it has been observed that associative learning was impaired in 16-monthold mice compared with that of 3-month-old mice which was associated with specific reduction in acetylation of H4K12 (Peleg et al., 2010).

Until now, most of the studies have analyzed DNA methylation in the brain of AD patients (Balazs, 2014). In this regard, a variety of studies suggest a genome-wide decrease in DNA methylation present in aging and AD patients (**Table 2**; Mastroeni et al., 2011). Interestingly, the folate/methionine metabolism is critically linked with DNA methylation mechanisms, consistently with this fact; studies show that folate and S-adenosyl methionine are significantly decreased in AD (Bottiglieri et al., 1990; Morrison et al., 1996). All this data indicates that AD patients produce a hypomethylation across the DNA genome. Recently, Bakulski et al. provided a semi-unbiased, quantitative, genome-wide localization of DNA epigenetic differences in frontal cortex of control and AD cases. These authors determined DNA methylation of 27, 587 CpG sites spanning 14,475 genes. Interestingly, they found that in control samples, the methylation state is markedly affected by age, with about the same number of sites being hypermethylated as hypomethylated with age. Compared with controls, 6% of genes featured on the array were differentially methylated in AD samples, but the mean difference was relatively modest (2.9%). Gene ontology analysis revealed a relationship between the main disease-specific methylation loci and several molecular **Table 2 | Epigenetic modifications implicated in Alzheimer's disease**.


functions and biological processes, including hypermethylation of genes involved in transcription and DNA replication, while membrane transporters were hypomethylated (Bakulski et al., 2012).

Also, some reports have focused on research DNA methylation at the 5<sup>0</sup> promoter regions of candidate genes according to the basis of hypothesis concerning the molecular mechanisms of AD as microtubule-associated protein tau, amyloid precursor protein (APP) and preseniline-1 genes in the frontal cortex and hippocampus of both control and AD cases at different Braak stages. Interestingly, there wasn't any significant difference on CpG methylation between the control and AD samples (Barrachina and Ferrer, 2009). Other studies have reported hypomethylation of APP in the promoter region of normal 70 year-old human brain (Tohgi et al., 1999). However, as mentioned above, no difference was found in methylation of selected regions of the APP gene in various stages of AD progression (Barrachina and Ferrer, 2009). Also, it has been found that the change in methylation status differed among transcription factor binding sites of tau promoter (Wang et al., 2013).

Additionally to DNA methylation, histone modifications have been studied in recent years. Francis et al. investigated histone acetylation in mouse models of AD. In APP/presenilin1 double mutant transgenic mice, associative learning was impaired and this was linked to a marked reduction in H4K14 histone acetylation (Francis et al., 2009). Furthermore, studies *in vitro* have shown that exposure of cortical and hippocampal cultures to Aβ oligomers resulted in increased levels of acetylated H3K14 and a loss of dendritic spines, which was prevented by inhibition of histone acetyl transferase. Also, in young pre-plaque AD transgenic mice, these authors observed markedly increased levels of H3K14 and H3K9me2 compared with those of wildtype non-transgenic mice. Most importantly, similar changes were observed in histone transcription activating and repressive marks in the occipital cortex of AD samples (Lithner et al., 2013).

Although there are now treatments against AD, these are only palliatives and the pathology is currently incurable, whereby, there is an intense interest in the development of new potential therapies. Epigenetic therapies have achieved some progress in the field of cancer, thus, several inhibitors of HDACs and DNA methylation are approved for hematological malignances by the US Food and Drug Administration and have been in clinical use for several years (Wu et al., 2008a). HDAC inhibitors (HDACIs) are the most thoroughly studied and have shown acceptable results in AD models. The inhibitors widely used in clinical research include trichostatin A (TSA), valproic acid (VPA), sodium 4-phenylbutyrate (4-PBA) and vorinostat (SAHA) (Wang et al., 2013).

In a study conducted by Su et al., VPA showed to inhibit Aβ production in HEK293 cell transfected with a plasmid carrying the Swedish APP751 mutation. Interestingly, using the APPV717F transgenic model of AD, VPA was able to inhibit Aβ production in the brain of mice at biologically relevant doses of 400 mg/kg (Su et al., 2004). In another study, VPA showed to decrease Aβ production and alleviate behavioral deficits by inhibiting GSK-3β-mediated γ-secretase cleavage of APP in APP23 transgenic mice (Qing et al., 2008). These results give us the idea about the possible contribution of epigenetic modifications in AD, which suggests that the drugs targeting epigenetic process may be of future therapeutic value (Wang et al., 2013).

As mentioned widely in scientific literature, the interaction between diet and epigenetics is the best documented in cancer pathology (Ho et al., 2009; Shu et al., 2010). Furthermore, based on evidence in support of epigenomics in regulating gene expression in stress-mediated AD risk factors, and the pathophysiology of AD, there has been growing interest in examining whether diet and nutraceuticals targeting epigenomics may prevent, delay, or reverse the course of AD (Chiu et al., 2014). In this regard, the Mediterranean diet rich in vegetables, fruits and nuts, legumes, olive oil and fish with relative low intakes of red meat has been suggested to reduce the risk for AD onset (Scarmeas et al., 2009; Frisardi et al., 2010). Other studies appoint that anti-oxidant-rich diets and consumption of dietary phytochemical such as caffeic acid, epigallocatechin-3-gallate, *Gingko biloba*, resveratrol and phenolic compounds present in red wine slowed down disease progression by inhibiting Aβ production or amyloid aggregation in animal models (Kolosova et al., 2006).

It is well known that DNA methylation occurs within folate/methionine/homocysteine (HCY) metabolism which uses micronutrients such as folate, methionine, choline and betaine enzyme's cofactors (Chouliaras et al., 2010; Wang et al., 2013). Diverse reactions occur and methionine is converted to S-adenosyl-methionine (SAM) and then converted to S-adenosyl-homocysteine (SAH), which in turn is converted to HYC in a reversible reaction. Most important, SAM is the common methyl donor for DNA methylation that regulates gene expression and determines the chromosome conformation (Sezgin and Dincer, 2014). An early study showed that SAM levels have been found to be decreased in post-mortem AD patients (Morrison et al., 1996). Also, lower bioavailability of SAM causes changes in the expression of genes involved in APP metabolism because this metabolite maintains the appropriate methylation of genes involved in APP processing (Sezgin and Dincer, 2014). Fuso

et al. recently reported that reduction of folate and Vitamin B12 in culture medium of neuroblastoma cell lines cause a reduction in SAM levels resulting in an increase of PSEN1 and BACE levels together with Aβ production. Conversely, the simultaneous administration of SAM to the deficient medium restored the normal gene expression and reduced the Aβ levels (Fuso et al., 2007). Interestingly, the same group demonstrated that Vitamin B deficient-animals have shown that SAM inhibits the increase in progression of Alzheimer-like features (Fuso et al., 2012). This data suggests that folate or Vitamin B12-rich diets could be beneficial as therapy for AD patients; however, more studies are needed.

#### **PARKINSON'S DISEASE AND EPIGENETICS**

Parkinson's disease (PD) is the second most common neurodegenerative disorder after AD affecting approximately 1–2% of the population over the age of 65 and reaching a prevalence of almost 4% in those aged above 85. Resting tremor, bradykinesia, rigidity, and postural instability are the main clinical symptoms of the disease often accompanied by nonmotor symptoms including autonomic insufficiency, cognitive impairment, and sleep disorders (Thomas and Beal, 2011; Coppedè, 2014). The brain of PD individuals is pathologically characterized by a progressive loss of neuromelanin containing dopaminergic neurons in the *substantia nigra* with the presence of eosinophilic, intracytoplasmic inclusions termed as Lewy bodies (structures containing aggregates of α-synuclein as well as other substances) and Lewy neurites in surviving neurons. Unfortunately, only some improvements of the symptoms are offered by current treatments based on levodopa and dopaminergic therapy, but there is no currently available treatment to avoid the progression of the disease (Thomas and Beal, 2011; Coppedè, 2014).

The vast majority of PD cases are idiopathic forms, likely resulting from a combination of polygenic inheritance, environmental exposures, and complex gene-environment interactions imposed on slow and sustained neuronal dysfunction due to aging (Migliore and Coppedè, 2009). In a minority of the cases, PD is inherited as Mendelian trail, and studies in PD families allowed the identification of at least 15 PD loci (PARK1-15) and several causative genes (Nuytemans et al., 2010). In addition, there are genes such as LRRK2, SNCA, MAPT and GBA that are associated with sporadic PD without family history (**Table 3;** Coppedè, 2012).

Most of the studies evaluating the role of epigenetic in pathogenesis have focused on the analysis of promoter methylation of causative PD genes in post-mortem brains and peripheral blood; however, the role of DNA methylation and its links to PD pathogenesis is currently unclear (Coppedè, 2012). Recent studies have shown that methylation of SNCA gene (the gene coding for α-synuclein) may be involved in disease via structural changes or overexpression of the protein, leading to protein aggregation or via impaired gene expression (Ammal Kaidery et al., 2013). In this regard, methylation of SNCA intron 1 has been demonstrated to be associated with decreased SNCA transcription, whereas reduced methylation at this site was found to be decreased in several brain regions, including

#### **Table 3 | Epigenetic modifications of Parkinson's disease related genes**.


the *subtancia nigra* of sporadic patients, causing the increased expression of the SNCA gene (Jowaed et al., 2010). These results raise the possibility that the increased α-synuclein production that is associated with PD may result from increased SNCA expression, as a consequence of a decreased methylation state of its gene (Ammal Kaidery et al., 2013). Additionally, it has been demonstrated that α-synuclein sequesters DNMT1 in the cytoplasm, leading to global DNA hypomethylation in PD and dementia with Lewy body in post-mortem brains, as well as in transgenic mouse models (Desplats et al., 2011). Conversely, the overexpression DNMT1 in both transgenic mouse models and cellular cultures restore the nuclear level of the enzyme (Ammal Kaidery et al., 2013).

The regulation of SNCA by epigenetic histone modifications is yet to be studied in human PD brains. Studies in cell cultures and animal models of the disease, such as those induced by mitochondrial toxins, including 1-methyl-4-phenylpyridinium (MPP+), paraquat, rotenone, or those overexpressing human α-synuclein, have revealed that α-synuclein translocates into the nucleus interacting with histones and inhibiting histone acetylation (Goers et al., 2003). Furthermore, in Drosophila models, nuclear-targeted α-synuclein has been shown to bind to histones and reduce histone 3 acetylation through its association with HDAC1 and SIRT2 (Kontopoulos et al., 2006).

In recent years, there has been considerable progress in the development of epigenetic-based drugs for the treatment of neurodegenerative disorders such as PD. Such inhibitors of HDACs and DNMTs are currently approved and available for clinical investigation (Xu et al., 2012). In this regard, the targeted downregulation of SIRT2 has been shown to ameliorate α-synuclein toxicity and dopaminergic loss in flies and in primary mesencephalic culture. Moreover, toxicity associated with nuclear-targeted α-synuclein in both SH-SY5Y neuroblastoma cells and flies can be rescued by using HDACIs (Outeiro et al., 2007), thus, HDACIs have been theorized to be efficacious in neurodegenerative diseases (Harrison and Dexter, 2013). In this regard, Wu et al. demonstrated that trichostain A (a well-known HDAC inhibitor), protects dopaminergic neurons from MPP<sup>+</sup> toxicity in primary neuron-glia co-cultures in a dose dependent manner (Wu et al., 2008b). Moreover, Kid and Schneider demonstrated that vorinostat (another HDAC inhibitor) protected two different dopaminergic neuronal cell lines from apoptosis induced by MPP<sup>+</sup> (Kidd and Schneider, 2010), thus, the above results give us an idea about the alternative therapy by inhibiting HDACs in PD patients.

Although the etiology of PD is still unknown, multiple lines of evidence support oxidative stress and mitochondrial dysfunction as part of the pathogenic cascade. It would be interesting to know whether antioxidants-rich diets that have a helpful effect in other degenerative disease such as AD (Kolosova et al., 2006), could have the same effect in PD patients. To this regard, therapy focusing on nutrition, neutraceutical and antioxidants as part of a healthy lifestyle might protect against cell death and thus delay or halt disease progress; however, clinical and basic studies are needed to prove such hypothesis (Bega et al., 2014).

#### **EPILEPSY AND EPIGENETICS**

Epilepsy is the third most common chronic brain disorder affecting 50 million of people worldwide (Aroniadou-Anderjaska et al., 2008). In this disorder, a variety of structures of the central nervous system such as the hippocampus, the amygdala and the piriform cortex are susceptible to trigger electrical discharges that contribute to brain damage and to the epileptogenic mechanism (Houser, 1990; Blümcke et al., 1999). These discharges promote some morphological changes in the hippocampus such as, cellular death in the CA1 and mossy fiber sprouting and dispersion of the granule cell layer, alterations that are thought to be involved in the formation of recurrent excitatory circuits that contributes to seizure susceptibility (Heck et al., 2004).

In this regard, it is well known that seizures can give rise to enduring changes that reflect alterations in gene expression patterns, contributing in this way to the hallmarks of epilepsy (Roopra et al., 2012). Moreover, some studies suggest that these long-term changes mediated by seizures are mediated via modulation of chromatin structure. One transcription factor in particular, the repressor element 1-silencing transcription factor (REST/NRSF) has received a lot of attention due to its association with a great sub-set of genes associated with important processes involved in neuronal homeostasis and because it may seem to recruit a variety of proteins that elicit epigenetic modifications such as histone deacetylases and histone methyltransferases (Bruce et al., 2004; Ballas and Mandel, 2005; Ballas et al., 2005; Johnson et al., 2006; Pozzi et al., 2013). Some reports have shown that the induction of seizures in animal models induce an overexpression in both REST/NRSF protein and mRNA levels (Formisano et al., 2007; Noh et al., 2012), suggesting that seizures may cause an unbalance in the epigenetic modifications that control important processes of neuronal homeostasis. In contrast, recent studies have shown that REST/NRSF is induced in the aging human brain regulating a network of genes associated with stress resistance (Lu et al., 2014). This evidence suggests that REST/NRSF regulates important processes in embryonic and adult neuronal homeostasis and that the dysregulation of this transcription factor may impair epigenetic modifications that regulate precisely an important network of genes contributing to distinct neurological/ neurodegenerative disorders such as epilepsy or AD.

From a public health perspective, an alternative for the treatment of epilepsy is a change of lifestyle or diet. These methods have probably been used for over 2000 years and actually metabolic regulation of neuronal excitability is increasingly recognized as a factor in seizure pathologies and control (Stafstrom et al., 2008; Yuen and Sander, 2014). In this way, approximately half of the pharmacoresistant patients that have tried metabolism based therapies experience seizure control, opening the possibility of a strong link between the environments, in this case nutrition, with this pathology (Greene et al., 2003; Bough et al., 2006; Marsh et al., 2006; Patel et al., 2010).

These studies suggest that metabolism-based therapies such as ketogenic diets, calorie restriction or intermittent fasting leads to a range of biochemical and metabolic changes that induce a metabolic shift in pathways such as glycolysis, ketogenesis or beta oxidation, modifications that have been shown to increase seizure thresholds and to decrease epileptogenesis in animal models (Marsh et al., 2006; Patel et al., 2010).

Moreover, recent studies have shown that environmental inputs such as nutrition or exercise modulates cell metabolism, and critical links between metabolism and epigenetic control are beginning to emerge (Sassone-Corsi, 2013). For example, the availability of specific metabolites such as acetyl-coenzyme A (acetyl-coA) and nicotinamide adenine dinucleotide (NAD+) dictates the efficacy of histone deacetylases (Katada et al., 2012).

In this regard, it has been shown that beta hydroxybutyrate (β-HB), a ketone body that rises with ketogenic diets, during strenuous exercise or during fasting (Newman and Verdin, 2014), acts as an endogenous inhibitor of histone deacetylases linking in a precise way metabolism, epigenetics and epilepsy (Shimazu et al., 2013). Thus, these studies strongly suggest that the neuroprotective effects exerted by these kinds of therapies are not only mediated via metabolism alterations but also by epigenetic modifications that may be involved in the expression of an unknown sub-set of genes related to epilepsy.

Other interesting epigenetic modifications involved in epilepsy are methylation of DNA. In this field, Kobow et al. using Methyl-seq, mapped for the first time the global DNA methylation patterns in chronic epileptic rats; they showed that chronic epilepsy in animal models is characterized for a global hypermethylation on DNA. Moreover, this group shows that ketogenic diets diminish this increase of DNA methylation, suggesting that these kinds of therapies exert their effect not only modulating metabolism, but also via epigenetic modifications (Kobow et al., 2013). More importantly, it opened the possibility for the development of new metabolism based therapies designed to regulate these epigenetic modifications contributing to the inhibition of the seizure threshold in epilepsy.

#### **THE ROLE OF REST/NRSF IN NEUROLOGICAL DISORDERS**

A growing body of literature suggests that long-term changes in gene transcription associated with a lot of neurological disorders are mediated via modulation of chromatin structure. One transcription factor in particular, REST/NRSF (repressor element 1-silencing transcription factor) (**Figure 4**), has received

a lot of attention due to the possibility that it may control the expression of approximately 1,300 genes (Bruce et al., 2004; Johnson et al., 2006) that could be associated with a variety of processes that are important for neuronal homeostasis such as; synaptic transmission, synaptogenesis, excitability or even neurogenesis (Ballas and Mandel, 2005; D'Alessandro et al., 2009). REST modulates these genes in the nervous system recruiting protein complexes that elicit different epigenetic modifications (**Figure 4**; Roopra et al., 2012). Now it has been shown that REST is upregulated in pyramidal and dentate gyrus neurons after *status epilepticus* induced by kainate (Palm et al., 1998) or even by ischaemic insults (Formisano et al., 2007; Noh et al., 2012). Therefore, the upregulation of REST has been previously considered as harmful in mature neurons. In contrast, recent studies have shown that induced expression of REST/NRSF in mature hippocampal neurons is a protective mechanism that modulates the inhibitory homeostatic control of intrinsic excitability (Pozzi et al., 2013). Moreover, it has been shown that REST/NRSF protects neurons from age-related toxic insults in AD and surprisingly these levels seems to be associated with preservation of cognitive function and increased longevity (Lu et al., 2014). These findings suggest that basal levels of REST/NRSF are necessary for a normal physiological condition in the adult brain and that elevated levels of REST/NRSF, characteristic of epilepsy, may not be an epileptogenic factor, rather it seems to be a homeostatic mechanism triggered by repeated hyper-excitability *stimuli*. This is an open issue that needs further investigation.

#### **CONCLUDING REMARKS**

As we state in this manuscript, one of the main factors that contributes to a variety of the most common diseases is the environment. Many epigenetic enzymes are potentially susceptible to changes in the levels of a variety of metabolites, and are, hence, poised to respond to changes on environment. In this sense, it has been demonstrated that changing our lifestyle could mediate great beneficial effects regulating a network of genes via the modulation of chromatin structure, providing new alternatives for the prevention of many diseases.

Different questions remain to be answered including which epigenetic modifications are implicated in neurological disorders, how does the environment mediate these changes, could pharmacological inhibitors of these modifications provide an alternative for treating disease, and so on. Increasing evidence on this field had taught us that these modifications are capable of regulating great networks of genes that can influence a variety of physiological processes important for overall homeostasis and that the disruption of this balance can increase the risk of disease.

From a public health perspective, we need to better understand which alterations in metabolism and in chromatin structure cause disease and, maybe, it will be possible to design rationale metabolism-based therapies that could function as alternative treatments of these kinds of disorders.

#### **ACKNOWLEDGMENTS**

This work was supported by grants from Programa de Apoyo a Proyectos de Investigación e Innovación Tecnológica, DGAPA-PAPIIT (IN211913) and Consejo Nacional de Ciencia y Tecnología, CONACyT (239594). The authors want to thank Mrs. Josefina Bolado for editing this English-language text chapter.

#### **REFERENCES**


decline: possible underlying mechanisms. *J. Alzheimers Dis.* 22, 715–740. doi: 10. 3233/JAD-2010-100942


of neurodegenerative diseases. *Neurobiol. Aging*, pii:S0197-4580(14)00497-7. doi: 10.1016/j.neurobiolaging.2014.07.028. [Epub ahead of print].


**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 October 2014; accepted: 06 February 2015; published online: 27 February 2015*.

*Citation: Landgrave-Gómez J, Mercado-Gómez O and Guevara-Guzmán R (2015) Epigenetic mechanisms in neurological and neurodegenerative diseases. Front. Cell. Neurosci. 9:58. doi: 10.3389/fncel.2015.00058*

*This article was submitted to the journal Frontiers in Cellular Neuroscience*.

*Copyright © 2015 Landgrave-Gómez, Mercado-Gómez and Guevara-Guzmán. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution and 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*.

# Environmental pollutants as risk factors for neurodegenerative disorders: Alzheimer and Parkinson diseases

#### Miguel Chin-Chan† , Juliana Navarro-Yepes † and Betzabet Quintanilla-Vega \*

Department of Toxicology, CINVESTAV-IPN, Mexico City, Mexico

Neurodegenerative diseases including Alzheimer (AD) and Parkinson (PD) have attracted attention in last decades due to their high incidence worldwide. The etiology of these diseases is still unclear; however the role of the environment as a putative risk factor has gained importance. More worryingly is the evidence that pre- and post-natal exposures to environmental factors predispose to the onset of neurodegenerative diseases in later life. Neurotoxic metals such as lead, mercury, aluminum, cadmium and arsenic, as well as some pesticides and metal-based nanoparticles have been involved in AD due to their ability to increase beta-amyloid (Aβ) peptide and the phosphorylation of Tau protein (P-Tau), causing senile/amyloid plaques and neurofibrillary tangles (NFTs) characteristic of AD. The exposure to lead, manganese, solvents and some pesticides has been related to hallmarks of PD such as mitochondrial dysfunction, alterations in metal homeostasis and aggregation of proteins such as α-synuclein (α-syn), which is a key constituent of Lewy bodies (LB), a crucial factor in PD pathogenesis. Common mechanisms of environmental pollutants to increase Aβ, P-Tau, α-syn and neuronal death have been reported, including the oxidative stress mainly involved in the increase of Aβ and α-syn, and the reduced activity/protein levels of Aβ degrading enzyme (IDE)s such as neprilysin or insulin IDE. In addition, epigenetic mechanisms by maternal nutrient supplementation and exposure to heavy metals and pesticides have been proposed to lead phenotypic diversity and susceptibility to neurodegenerative diseases. This review discusses data from epidemiological and experimental studies about the role of environmental factors in the development of idiopathic AD and PD, and their mechanisms of action.

Keywords: Alzheimer's diseases, Parkinson's disease, neurodegenerative disorders, beta-amyloid, tau protein, alpha-synuclein, metals, pesticides

#### Edited by:

Victoria Campos-Peña, Instituto Nacional De Neurologia Y Neurocirugia, Mexico

#### Reviewed by:

Robert Weissert, University of Regensburg, Germany Zhihong Chen, Cleveland Clinic, USA

#### \*Correspondence:

Betzabet Quintanilla-Vega, Department of Toxicology, CINVESTAV-IPN, Ave. IPN 2508, Mexico City, 07360 Mexico Tel: 52+55-57473310, Fax: 52+55-57473395 mquintan@cinvestav.mx

†These authors have contributed equally to this work.

> Received: 10 January 2015 Accepted: 17 March 2015 Published: 10 April 2015

#### Citation:

Chin-Chan M, Navarro-Yepes J and Quintanilla-Vega B (2015) Environmental pollutants as risk factors for neurodegenerative disorders: Alzheimer and Parkinson diseases. Front. Cell. Neurosci. 9:124. doi: 10.3389/fncel.2015.00124

**Abbreviations:** Aβ, amyloid beta peptide; AICD, amyloid intracellular domain; AD, Alzheimer's disease; Al, aluminum; AGE, advanced glycation end products; ApoE, apolipoprotein E; APP, amyloid precursor protein; As, arsenic; α-syn, alpha-synuclein; BACE1, beta-site APP cleaving enzyme 1; Cd, cadmium; CSF, cerebrospinal fluid; EOAD, early-onset Alzheimer' disease; Hg, mercury; HR, hazard ratio; IDE, insulin-like degrading enzyme; LB, Lewy bodies; LOAD, late-onset Alzheimer' disease; LRP1, low density lipoprotein receptor-related protein 1; Me-Hg, methyl mercury; Mn, manganese; MTs, microtubules; NEP, neprilysin; NFTs, neurofibrillary tangles; NPs, nanoparticles; OP, organophosphates; OCl, organochlorines; PHFs, paired helical filaments; Pb, lead; PD, Parkinson's disease; PQ, paraquat; PS, presenilin; ROS, reactive oxygen species; RR, relative risk; sAPPβ, soluble APP beta; SNpc, substantia nigra pars compacta; SNpr, substantia nigra pars reticulate; Tau, tau protein; P-Tau, phosphorylated Tau; TCE, trichloroethylene; Zn, zinc.

#### Chin-Chan et al. Environmental factors in neurodegeneration

# Introduction

Life expectancy has increased in last decades and health care improvements have contributed to people living longer. However, this has also contributed to increase the number of people with chronic disabling diseases such as Alzheimer (AD) and Parkinson (PD). Genesis of both neurodegenerative diseases has not been elucidated and several endogenous (genetic) and exogenous (environment) factors contribute to the onset and/or development of these illnesses, which highlights the necessity to expand the research on identifying the environmental risk factors that predispose to the development of these neurodegenerative diseases.

It is known that the etiology of neurodegenerative diseases is multifactorial, and there is evidence that potential external factors including lifestyle and chemical exposures are linked with the risk of the onset of these diseases. Since the vast majority of AD and PD cases are observed in elderly populations, yet the exposure to risk factors occurred years or decades before the diagnosis, the assessment of chronic exposures is difficult to perform in retrospective studies to associate them with the onset/development of the disease. Therefore more research for better definition of exposure, as well as for the identification of early specific biomarkers for the diagnosis of these diseases is needed. Attention is now focused on environmental factors that potentially damage the developing nervous system through epigenetic mechanisms, resulting in neurodegenerative diseases later in life. In this review we briefly examined the evidence of environmental etiologies related to two of the most common neurodegenerative diseases, AD and PD, from epidemiological as well as experimental studies.

# Alzheimer's Disease

Alzheimer's disease is the major form of dementia in elderly and possibly contributes to 60–70% of cases. It is a progressive, disabling and irreversible disease (Goedert and Spillantini, 2006). There are two recognized forms of AD. The first one is named familial or of early onset (EOAD), which is directly related to specific gene mutations in the amyloid precursor protein (APP) and presenilin (PSEN) 1 and 2 genes, both related to the amyloidbeta (Aβ) peptide synthesis (Piaceri et al., 2013). The EOAD begins at early age, less than 65 years, and only explains 5% of all cases. The second one, the late-onset or sporadic AD (LOAD) is the most common form of AD with 95% of all cases. This form of AD is not caused by punctual mutations, but some genetic risk factors have also been described such as polymorphisms in ApoE (encoding for apolipoprotein E), SORL1 (encoding for neuronal receptor of ApoE), and GSK3 (encoding for glycogen synthase kinase 3 beta) genes. The ApoE gene is the strongest genetic risk factor for LOAD, although it is not sufficient to explain the occurrence of the disease (Godfrey et al., 2003). Therefore, the etiology of LOAD remains unclear, and it is suggested that it has a multifactorial etiology.

Two hypotheses have been most studied for AD development. One is related to the overproduction of the Aβ peptide. According to this, neurofibrillary tangles (NFTs) result from the onset of amyloid deposits as Aβ plaques. While the second hypothesis suggests that the hyperphosphorylation of the Tau protein and its subsequent deposition as NFTs is the ultimate responsible for the disease. The amyloid cascade hypothesis establishes that Aβ aggregation initiates the brain damage leading to memory loss and to AD (Hardy and Higgins, 1992). Aβ is normally produced during aging, mediated by the proteolytic processing of the APP by the amyloidogenic enzymatic pathway. In this pathway, APP is processed by β- and γ-secretase complexes producing Aβ, soluble APPβ (sAPPβ) and the amyloid intracellular domain (AICD). Alternatively, APP can be processed by the non-amyloidogenic pathway leading to the production of AICD, sAPPα but not Aβ (Thinakaran and Koo, 2008). Thus Aβ increased levels in the brain of LOAD patients could be mediated by: (i) an increase in APP expression; (ii) an increase in the amyloidogenic pathway; or (iii) a reduction in the non-amyloidogenic pathway. It is stablished that the increase of a member of the β secretase complex, BACE1 (betasite APP cleaving enzyme 1) produces high brain Aβ levels (Sun et al., 2012). On the other hand, the reduction on the activity of ADAM10 (a desintegrin and metalloproteinase domaincontaining protein 10) could also lead to the overproduction of Aβ (Kojro and Fahrenholz, 2005). Furthermore, some mutations such as those in PSEN 1 or 2, a catalytic member of the γ-secretase complex, can also increase the production of Aβ (Piaceri et al., 2013).

Another mechanism to increase brain Aβ levels is through a reduction in the Aβ degradation. There are proteins collectively known as Aβ-degrading enzyme (IDE)s that have the ability to degrade Aβ, including insulin-like IDE, angiotensin-converting enzyme (ACE), endothelin-converting enzyme (ECE), plasmin, cathepsin B, aminopeptidase A, matrix metalloproteinase (MMP) 2 and 9, neprilysin (NEP, neutral endopeptidase) and others. These enzymes have been suggested as viable therapeutic targets for AD treatment (Nalivaeva et al., 2012). Finally, a reduced brain clearance of Aβ can be another pathway for the brain accumulation of Aβ. Some cholesterol transporters such as the low density lipoprotein receptor-related protein 1 (LRP1) are involved in the Aβ export from the brain to the cerebrospinal fluid (CSF). This receptor links the imbalance of cholesterol homeostasis with AD pathogenesis (Zlokovic et al., 2010).

On the other hand, aggregates of the microtubule (MT)-associated protein Tau observed in cell bodies and apical dendrites as NFTs cause neurofibrillary lesions associated with AD. Tau is a phosphoprotein mainly localized in the axon of neurons for the stabilization of MTs; it contains a high number of serine and threonine residues, and is therefore a substrate of many kinases (Goedert et al., 1988). The abnormal aggregation of Tau into insoluble paired helical filaments (PHFs), which are the major component of NFTs found in cell bodies and apical dendrites of neurons are lesions associated with AD (Friedhoff et al., 2000). Under pathological conditions, Tau is hyperphosphorylated at ''pathological'' sites leading to MTs depolymerization, axonal transport disruption and aggregation (Götz, 2001). It has been proposed that repeat domains (RD) of the MT-binding domain (MBD) in the C-terminal structure of Tau can rapidly form PHFs compared with the complete protein, suggesting that RDs are indispensable for its aggregation (Wille et al., 1992), and for Tau filament formation (Tokimasa et al., 2005).

There is no cure for AD, and therapeutic treatments are basically to ameliorate the symptoms. Therefore, an early and opportune diagnosis is indispensable to slow the progression of the disease. Currently, the determination of Tau and Aβ levels in blood and CSF are broadly used for the diagnosis of AD, and several medical tools are also used to confirm the diagnosis including the medical history, mental status tests, and evaluations of the brain structure and function with neuroimaging techniques (Lewczuk et al., 2014). However, these biomarkers are not sensitive nor specific for AD. Interestingly, an emerging body of evidence suggests that micro RNAs (miRNAs, small non-coding RNAs involved in the post-transcriptional regulation of gene expression) could be putative biomarkers for detecting neurodegeneration Thus, recent reviews have shown that some miRNAs are differentially associated with AD by modulating the expression of important genes involved in Aβ production (e.g., BACE1) or inflammation (Goodall et al., 2013; Van den Hove et al., 2014).

metals, pesticides, nanoparticles, and diet can affect the two targets of AD such as Aβ generation and Tau phosphorylation. The figure depicts the molecular

or the hyperphosphorylation of Tau protein and its subsequent deposition as neurofibrillary tangles (NFTs) (lower part). For more detail see the text.

Aβ Homeostasis as a Target of Environmental Factors Environmental factors such as diet (fat-rich), heavy metals, biogenic metals and pesticides have been involved in AD development due to their ability to disrupt metabolic pathways involved in the homeostasis of Aβ. In addition, factors such as lifestyle (antioxidants and exercise) can prevent AD development (**Figure 1**). Many of these environmental factors are oxidative agents acting through different mechanisms as discussed later. The brain is particularly vulnerable to oxidative stress do to its high glucose-based metabolic rate, low levels of antioxidants, high levels of polyunsaturated fatty acids, and high enzymatic activities related to transition metals that catalyze the formation of free radicals (Halliwell et al., 1992). In addition, micromolar concentrations of Aβ induce the formation of H2O<sup>2</sup> in culture cells leading to neurotoxicity, and the presence of some antioxidant enzymes prevents the toxicity of the peptide (Butterfield et al., 2001). The mechanism by which Aβ generates free radicals is not known, and other endogenous factors also generate reactive oxygen species (ROS) in AD. For instance, the ion Fe3+, which is at high concentrations in NFTs and Aβaggregates, catalyzes the formation of reactive species such as H2O2, as well as advanced glycation end products (AGE) that are related to neurodegeneration (Smith et al., 1997b). On the other hand, activated microglia that surrounds the senile plaques is a source of NO and O<sup>2</sup> (Cras et al., 1990), which can react to form the peroxinitrite radical (ONOO-) (Smith et al., 1997a). Likewise, inflammation has gained importance in AD pathogenesis (Tuppo and Arias, 2005). The central nervous system is considered a privileged site with its own immune system and microglia and astrocytes are the principal cells involved in the inflammatory response. It is accepted the microglial chemotaxis of Aβ and the phagocytosis of amyloid fibrils, effects that produce an increase in the secretion of pro-inflammatory cytokines and ROS, which in consequence produces neuronal loss (Rogers and Lue, 2001). In agreement, astrocytes are also recruited into amyloid plaques for Aβ degradation (Wyss-Coray et al., 2003), and it is possible that the activation of microglia and astrocytes is a consequence of Aβ aggregation. The role of inflammatory processes in AD is supported by the use of non-steroidal anti-inflammatory drugs (NSAIDs) to reduce the Aβ levels (Weggen et al., 2001), and the risk of AD (Etminan et al., 2003).

# **Metals**

Lead (Pb) is a heavy metal well known by its neurological toxic effects, although a direct association with AD development has not been reported. Pb affects cognitive abilities, intelligence, memory, speed processing and motor functions in children (Mason et al., 2014), while studies in elderly are limited. A cohort study reported that bone Pb levels were associated with poor cognitive performance scores in old workers, suggesting that past Pb exposure can contribute to late cognitive deterioration (Dorsey et al., 2006). However, a recent study reported no association between serum Pb levels and AD (Park et al., 2014). Despite the few epidemiological studies relating Pb exposure with AD, the evidence is more solid in experimental studies. The influence of Pb in AD was initially suggested from results in rats early exposed (from postnatal day 1–20) or late exposed (at 18–20 months of age) to Pb (200 ppm, drinking water). Results showed an increase in the APP mRNA expression late in life after the neonatal exposure, but not in rats exposed as adults (Basha et al., 2005). Similarly, a study performed in non-human primates (Macaca fascicularis) exposed to Pb (1.5 mg/Kg/day, from birth to 400 days) showed that monkeys exposed at a young age had an increased number of amyloid plaques late in life (at 23 years old) (Wu et al., 2008). The increased Aβ levels appear to be mediated by an augmented expression of APP and BACE1 (Wu et al., 2008). These effects were also observed in differentiated SH-SY5Y cells incubated with Pb (5–100 µM/48 h) and analyzed 6 days later (Bihaqi and Zawia, 2012). Another study performed in differentiated SH-SY5Y cells showed an increase in Aβ secretion and APP expression, as well as reduced expression and protein levels of NEP (an Aβ IDE), suggesting that Pb can target both the synthesis and degradation of Aβ (Huang et al., 2011). However, in a recent work conducted in our laboratory, Pb did not show changes in NEP expression in differentiated SH-SY5Y cells exposed to 50 µM Pb, but an increase in APP levels (Chin-Chan et al., 2015). Another mechanism by which Pb increases Aβ levels is by reducing the brain Aβ clearance. A recent study showed that acute Pb exposure (27 mg/Kg, i.p.) to APP transgenic mice (V717F) reduced the expression of LRP1, resulting in the accumulation of Aβ in the hippocampus and cortex of treated mice (Gu et al., 2011). Studies from this group support that Pb can disrupt the brain export of Aβ leading to its accumulation and plaques formation (Behl et al., 2009, 2010).

Exposure to Pb during development is a good example of an environmental contaminant as a risk factor to promote neurodegenerative diseases such as AD, supporting the hypothesis that many adult diseases have a fetal origin (FeBAD) (Basha et al., 2005). The group of Zawia has extensively worked on latent responses to prenatal and early postnatal exposures to Pb. Authors exposed male neonatal rats to Pb (200 ppm, drinking water) from postnatal day 1–20, or to aging animals (18–20 months of age), and observed an increase in the APP mRNA expression as well as in the activity of the transcription factor Sp1 (one of the regulators of APP) in the cortex of neonates, and after 20 months of Pb exposure had ceased. They observed a concomitant increase in Aβ levels in old animals exposed to Pb at birth. Interestingly, APP and Aβ protein levels did not respond to Pb exposure at old age (Basha et al., 2005). Similarly, a study conducted in cynomolgus monkeys exposed to Pb (1.5 mg/Kg/day, via infant formula) from birth to 400 days of age, and terminated 23 years later showed increased mRNA levels of APP and Sp1 in the frontal cortex compared with control animals, and high levels of the biomarker of oxidative DNA damage, 8-oxo-dG, suggesting an oxidative mechanism (Wu et al., 2008). Aβ1–42 and Aβ-1–40 levels in aged monkeys were also increased, as well as the intracellular Aβ staining and dense-plaques compared with age-matched controls (Wu et al., 2008). In addition, a study reported that gestational exposure to Pb (0.1, 0.5 and 1%, drinking water) in mice led to increased brain levels of Aβ and worst spatial memory performance, as well as increased levels of pro-inflammatory agents such as interleukin-1 (IL-1) and tumor necrosis factor alpha (TNF-α; Li et al., 2014a).

Mercury (Hg) is a heavy metal with a high potential to cause neurotoxicity. Early studies about Minamata and Iraq disasters led us to understand the neurotoxicity of this metal. It is widely accepted that Hg disrupts the brain development and produces cognitive and motor disabilities (Johansson et al., 2007), and in adults, Hg exposure produces memory loss and cognitive alterations (Wojcik et al., 2006; Chang et al., 2008). An early study suggested a link between Hg exposure and AD. Authors reported increased levels of Hg (in microsomes) and bromide (in the whole brain), and reduced levels of rubidium (in the whole brain, microsomes, and nuclei), selenium (Se; in microsomes) and zinc (Zn; in nuclei) in AD brains compared with controls (Wenstrup et al., 1990). A subsequent work reported more than a 2-fold increase in blood Hg levels in AD patients (n = 33) compared with a control group (n = 65), as well as a positive correlation between blood Hg concentration and CSF levels of Aβ (n = 15, Pearson's correlation r = 0.7440, p = 0.0015) (Hock et al., 1998). More recently, several metals, including manganese (Mn), nickel, cadmium (Cd), Pb and Hg were determined in plasma and CSF of AD patients (n = 173) and healthy controls (n = 54), however only plasma Mn and Hg concentrations were significantly higher in AD patients (Gerhardsson et al., 2008). On the other hand, ApoE4 is a risk factor for AD, probably because this protein does not have sulfhydryl (SH) groups to scavenge heavy metals like Hg, whereas ApoE2 has four SH groups and then the ability to reduce the metal toxicity in the brain (Mutter et al., 2004). Therefore, ApoEε2 is considered a protective genotype for AD development (Suri et al., 2013). Similarly, Godfrey et al. (2003) observed a shift toward the risky ApoEε4 allele in patients with presumptive Hg-related neuropsychiatric symptoms with an elevated Hg body burden (n = 400; Godfrey et al., 2003).

The ability of Hg to increase Aβ levels has been studied in vitro and in vivo, and the suggested mechanisms include an increased production, a reduced degradation and/or a diminished brain clearance of the peptide. Olivieri et al. (2000) reported an increased secretion of both Aβ-40 and Aβ-42 when neuroblastoma cells were exposed to 50 µg/dL of inorganic Hg concomitant with ROS overproduction (Olivieri et al., 2000), and a study conducted in aggregating brain-cell cultures of fetal rat telencephalon showed that MeHg (noncytotoxic concentrations/10–50 days) produced increased APP levels accompanied by ROS production and glia activation (Monnet-Tschudi et al., 2006). Rat pheochromocytoma (PC12) cells exposed to both inorganic and organic (MeHg) Hg (10–1000 nM) also showed a dose-dependent overproduction of Aβ-40 probably by an increase in APP levels as well as to a reduction in Aβ degradation by NEP (Song and Choi, 2013). However, an increase in Aβ-42 levels was observed in differentiated SH-SY5Y cells exposed to Hg (10 and 20 µM) but not in the APP expression, and rather a reduced activity of the Aβ-degrading enzyme, NEP was observed (Chin-Chan et al., 2015). Negative effects of Hg on Aβ aggregation have also been published, for example, Atwood et al. (1998) studied the role of the pH and the ability of various metals to aggregate Aβ; Zn, Cu and Fe showed the highest potential on Aβ aggregation, and Hg did not show an important effect (Atwood et al., 1998). Regarding in vivo studies, oral administration of 20–2000 µg/Kg/day/4 weeks of MeHg produced a dose-dependent increase in Aβ-42 in the hippocampus of male rats, but not significant changes in APP or NEP protein levels (Kim et al., 2014). Interestingly, authors observed a reduced brain expression of the LRP1 receptor, which was positively correlated with increased Aβ levels in the hippocampus and reduced levels in the CSF, suggesting a reduced clearance of the pathogenic peptide from the brain (Kim et al., 2014).

Inorganic arsenic (As) is a known neurotoxic metalloid with adverse effects in both the neurodevelopment and cognitive function (Tyler and Allan, 2014), although its effects in elderly have been less studied. There are few papers evaluating the role of As exposure as a risk factor for AD development. A study conducted in rural-dwelling adults and elders in Texas, US (FRONTIER project) reported that long-term exposure to low As levels (3–15 µg/L As in water) correlated (after adjustment by confounders, including the ApoEε4 presence) with a poor score of cognitive abilities and memory, which reflects the earliest manifestations of AD (O'Bryant et al., 2011). Similarly, a positive correlation was observed between serum As levels and the cognitive ability in AD patients from Hong Kong (n = 44, Pearson's correlation coefficient r = 0.55, p < 0.0001) compared with matched controls (n = 41), lower serum Zn concentrations were observed in AD patients as well (Baum et al., 2010). On the other hand, the study conducted in several European countries showed a higher prevalence of AD and other dementias in those countries with As levels in topsoils about 18 ppm such as Italy, Switzerland, Spain, France, Belgium and Norway, compared with countries with lower As levels (in the range of 9 ppm), including Luxemburg, Denmark, Finland, UK and Nertherlands (Dani, 2010). In experimental studies, the administration of inorganic As (20 mg/L, drinking water during the gestation and early postnatal life) to mice produced a significant loss of spatial memory (Ramos-Chávez et al., 2015).

A plausible mechanism for the cognitive and memory alterations induced by As exposure is by alterating the amyloid pathway. Zarazúa et al. (2011) reported that cholinergic SN56.B5.G4 cells incubated with sodium arsenite or the organic form dimethylarsinic acid (DMA) (5–10 µM/12–24 h) showed an increase in APP and sAPPβ levels, and consequently an increase in Aβ only with DMA. Similar effects were observed in neurons from Tg2576 mice (a murine model that overexpresses a mutant form of APP most used in AD). They suggest that DMA-induced effects may be due by an increased Aβ anabolism (enhanced APP expression), although authors did not discard an alteration in the Aβ degradation pathway (Zarazúa et al., 2011). The mechanism by which As causes Aβ overproduction has not been determined, but As exposure has been associated with brain inflammatory responses and oxidative stress, which is in agreement with the inflammatory and oxidative hypotheses of AD (Gong and O'Bryant, 2010).

Cadmium is another toxic heavy metal associated with neurological alterations including memory loss and mental retardation (Wang and Du, 2013). An early study observed higher plasma levels of various metals including Cd, aluminum (Al), As, and Se in AD patients (n = 24) compared with healthy volunteers (n = 28) (Basun et al., 1991). Also, the liver from autopsied AD patients (n = 17) had significant higher Cd levels compared with age- and sex-matched control subjects (n = 17) (Lui et al., 1990). However, Gerhardsson et al. (2008) did not observe significant differences in Cd concentrations in plasma or CSF in patients with AD (n = 173) compared with healthy control subjects (n = 54) (Gerhardsson et al., 2008). There is evidence linking Cd exposure with Aβ overproduction. Li et al. (2012) observed cognitive alterations accompanied by an increased production of Aβ-42 and increased size and number of senile plaques in the cerebral cortex and hippocampus from APP/PSEN1 transgenic mice treated with Cd (2.5 mg/Kg/4 days, drinking water). These effects were attributed to a reduced expression of ADAM10, sAPPα and NEP proteins, suggesting that the non-amyloidogenic pathway as well as Aβ degradation are target of Cd exposure (Li et al., 2012). Additionally, authors reported an increase in free-Zn levels, suggesting that Cd displaces Zn from its native enzymes, including NEP.

Recently, Ashok et al. (2015) investigated the role of the exposure to individual metals (As, Pb and Cd) and their combination in the AD-amyloid pathway in male rats exposed from gestational day 5 to postnatal day 80 through drinking water. They reported that metals activated the synthesis of Aβ in the frontal cortex and/or hippocampus, mediated by an increase in APP, and in APP-processing enzymes such as beta secretases BACE1 and PSEN at postnatal days 24 (post-weaning) and 90 (adulthood). Pb was the most potent metal to induce Aβ, followed by Cd, and As had the smallest effect, however all did increase the APP production. Interestingly, they demonstrated a synergic effect of metals mainly due to As, the exposure to these three cations produced a dramatic increase in Aβ, PSEN1, BACE1 and APP, suggesting an enhanced amyloidogenic processing (Ashok et al., 2015). Authors also observed (Ashok et al., 2015) increased levels of malondialdehyde (MDA), reduced activity of antioxidant enzymes, and the induction of 1L-1α and IL-1β in the frontal cortex and hippocampus of rats exposed to As + Pb + Cd mixture. Authors suggest that ROS-induced IL-1 overproduction was responsible for the APP expression. This is supported by the fact that the 5'ÚTR region of the mRNA of APP has a responsive element for IL-1 (Rogers et al., 1999; Ashok et al., 2015).

Aluminum is a neurotoxic element involved in the etiology of neurodegenerative disorders such as AD; however, there is no consistent evidence. The incident of Al pollution in Cornwall, UK (1998) gave evidence of Al potential neurotoxicity. Similar brain pathological characteristic found in AD patients were observed in subjects exposed to Al in this region (Exley and Esiri, 2006), as well as alterations in brain functions (Altmann et al., 1999). A recent study conducted in China reported a marginal positive association between Al levels in soil and the mortality caused by AD (Shen et al., 2014); while other studies reported no association. Experimental evidence appears to be more consistent. Chronic oral Al administration in rats (20 g/day in the food/twice weekly from 6 months of age to the end of their lives) increases the Aβ production by raising the levels of APP in hippocampal and cortical tissues (brain regions important for the memory process) (Walton and Wang, 2009). Cultured rat cortical neurons exposed to Al (50 µM/48 days) showed an accumulation of Aβ; furthermore, Al induced conformational changes in Aβ and enhanced its aggregation forming fibrillar deposits on the surface of cultured neurons. The aggregated Aβ was dissolved by the addition of desferroxamine, a chelator of Al (Exley et al., 1993; Kawahara et al., 2001). The administration of Al plus Dgalactose (an animal model for AD) produced the impairment of memory and increased the production of Aβ in the cortex and hippocampus. Additionally, an augmented expression of BACE1 and a reduction of NEP were observed in this cotreatment (Luo et al., 2009). Another study showed that Al reduced the Aβ degradation by decreasing the activity of cathepsin B, suggesting the activation of the amyloidogenic pathway and a reduction of the catabolism of Aβ (Sakamoto et al., 2006). In addition, a reduced expression of LRP1 was also observed in these mice administered with Al plus D-galactose, indicating a possible reduction of the clearance of Aβ as well (Luo et al., 2009). Transgenic mice (Tg2576) fed with Al (2 mg/Kg, in the diet/9 months) showed an increased production of Aβ and proteins involved in its anabolism; the accumulation of amyloid plaques were reversed by the treatment with vitamin E, suggesting the contribution of Al-induced oxidative mechanism (Praticò et al., 2002).

As mentioned earlier, miRNA can be biomarkers of early diagnosis of AD, however few studies have reported the involvement of pollutants in the miRNAs homeostasis (Ray et al., 2014). Interestingly, Tg2576 mice exposed to Al (2 mg/Kg/9 months, through the diet) showed an increased expression of miRNAs (e.g., miR146a and miR125b) involved in a proinflammatory response similar to that observed in AD brains (Bhattacharjee et al., 2014; Zhao et al., 2014), and treatment of primary human astroglial (HAG) cells with 100 nM of Al + Fe increased the expression of NFκB-induced miR-125b and miR-146a (Pogue et al., 2011); these miRNAs are reported in AD pathology (Lukiw, 2012). Further studies are necessary to look for a possible relation between xenobiotic exposures and the deregulation of miRNA expression involved in neurodegeneration.

#### **Pesticides**

The association between chronic pesticide exposure and the prevalence of dementias, including AD has not been as well studied as with other environmental risk factors, and results are often inconsistent. This is mainly because the difficulty on getting adequate data on the levels of exposure of individual pesticides, which is often indirectly evaluated by structure questionnaires (exposure index). Some of the studies with positive associations include the one performed in the agricultural Cache County, Utah, US in about 3000 occupationally pesticides-exposed participants who were followed-up for 3, 7 and 10 years. The hazard ratio for developing AD was slightly higher for organophosphate (OP) pesticides exposure (HR = 1.53, 95% CI, 1.05–2.23) than to organochlorines (OCl) (HR = 1.49, 95% CI, 0.99–2.24), after adjusting for some variables, including ApoE genotype (Hayden et al., 2010). Similarly, a recent case-control study observed a 3.8-fold increase in the OCl metabolite DDE in serum from AD patients (n = 79) compared with control participants (n = 86); in addition, authors reported that the highest tertile of DDE levels was associated with an increased risk for AD development (odds ratio-OR = 4.18, 95% CI, 2.54–5.82), and carriers of an ApoEε4 allele may be more susceptible (Richardson et al., 2014). Baldi et al. (2003) evaluated the association between lifelong cumulative exposure to pesticides and neurodegenerative diseases in a subsample from a cohort of elderly people (aged 65 years or older) (PAQUID study) in southwestern France. Authors analyzed 96 incident cases of AD (71 women and 25 men) in a 5-year follow-up approach, and observed a significant association between AD and occupational exposure to pesticides in men with a relative risk of 2.4 (95% CI, 1.0–5.6) after adjusting by education and smoking. Results were not significant in women (Baldi et al., 2003).

The role of pesticides in alterations observed in cognitive functions and AD has been suggested based on epidemiological studies, but the mechanisms have been poorly explored. In vitro studies performed in differentiated SH-SY5Y cells incubated with OCl pesticides such as DDE and its parent compound DDT (1 µM/48 h) showed increased APP protein levels, although authors did not evaluate Aβ levels (Richardson et al., 2014). A recent study reported that DDT augmented Aβ levels by increasing APP and BACE1 levels in human neuroglioma H4-AβPPswe cells, as well as by reducing the clearance and degradation of the peptide by targeting the Aβ-degrading enzyme, IDE and the ATP-binding cassette transporter A1 (ABC1; Li et al., 2015). Regarding in vivo data, chlorpyrifos (CPF), an OP insecticide associated with cognitive impairment, oxidative stress and neuronal damage caused a significant increase in Aβ levels in the cortex and hippocampus, as well as increased memory loss and reduced motor activity in Tg2576 mice 6 months after an acute subcutaneous administration of 50 mg/Kg of CPF (Salazar et al., 2011). However, Peris-Sampedro et al. (2014) did not find increased Aβ levels neither significant changes in memory acquisition in Tg2576 mice treated with CPF (25 mg/Kg/twice weekly/4 weeks, intragastric) and analyzed 6 months later (Peris-Sampedro et al., 2014). More studies are needed to better understand the mechanisms by which OCl, OP and other insecticides are linked to AD.

Paraquat (PQ), is a common used herbicide that has been suggested to be involved in AD development. A recent study showed that treatment of wild type and APP transgenic (Tg2576) mice with PQ (10 mg/Kg/twice a week/3 weeks) produced a significant increase in Aβ levels in transgenic mice that was associated with mitochondrial oxidative damage in cerebral cortex leading to the impairment of learning and memory. Interestingly, the overexpression of peroxiredoxin 3, a mitochondrial antioxidant defense enzyme produced an improvement in cognitive functions and a significant reduction in Aβ levels in APP transgenic mice exposed to PQ (Chen et al., 2012a), suggesting that pro-oxidant xenobiotics like PQ can contribute to AD.

# **Nanoparticles**

As the synthesis of NPs for different applications, including drug delivery strategies in the treatment of AD is growing, it is necessary to study the potential toxic effects on proteins related to AD development.

There are not epidemiological studies associating the exposure to NPs with AD. However, there is increasing experimental evidence suggesting the potential role of NPs in brain damage. A recent study reported that nasal administration of TiO2-NPs (2.5–10 mg/Kg/90 days) to mice caused neuronal death in the hippocampus, oxidative stress and gliosis, and microarray analysis revealed a decline of genes associated with memory and cognition (Ze et al., 2014). Similarly, rats exposed to CuO-NPs (0.5 mg/Kg/day/14 days, i.p.) showed worst spatial cognition and a reduction in electrophysiological endpoints such as long-term potentiation, which matched with augmented levels of ROS and lipid peroxidation products (MDA and 4-hydroxinonenal-HNE), and reduced levels of antioxidants enzymes (An et al., 2012). Studies of NPs of Al, Cu and Ag administered at different doses and routes in rats and mice showed that they produce brain alterations such as motor, sensory and cognitive deteriorations (Sharma et al., 2009; Sharma and Sharma, 2012). However a recent study did not observe memory loss in adult mice administered with Ag-NPs (10, 25, and 50 mg/Kg/7 days) (Liu et al., 2013). Regarding in vitro studies, the exposure of human (SK-N-SH) and mouse (Neuro-2a) neuroblastoma cells to silica NPs (SiNPs) (10 µg/mL/24 h) raised the intracellular content of Aβ in both cell lines, which was associated with increased APP and reduced NEP protein levels. These effects may be mediated by ROS production, since SiNPs increased the production of intracellular ROS (Yang et al., 2014). Likewise, treatment of Neuro-2a cells to silver NPs (AgNPs, 12.5 µg/mL/24 h) showed the deposition of Aβ plaques and an increased expression of APP, while NEP and LPR1 (or LDLR) expression and protein levels were reduced, suggesting that AgNPs can induce AD by altering the amyloidogenic pathway: Aβ synthesis, degradation or clearance (Huang et al., 2015). Interestingly, authors also reported an increased expression of genes involved in the inflammatory response such as IL-1, C-X-C motif chemokine 13 (CXCL13), macrophage receptor with collagenous structure (MARCO), and glutathione synthetase (GSS) (Huang et al., 2015).

#### Tau Hyperphosphorylation by Environmental Factors

Several environmental factors have shown to mediate AD development through alterations on Tau phosphorylation and/or aggregation (**Figure 1**).

#### **Metals**

In vivo and in vitro studies have suggested the potential of Hg to induce P-Tau. Fujimura et al. (2009) reported an increased neuronal death and more migrating astrocytes in cerebral cortex of male mice exposed to MeHg (30 ppm, drinking water), as well as increased levels of P-Tau mediated by c-jun N-terminal kinase (JNK; Fujimura et al., 2009). An in vitro study showed that inorganic Hg (50 µg/dL/30 min) increased P-Tau in SH-SY5Y cells by a ROS-dependent mechanism, which was reverted by the co-treatment with the antioxidant melatonin (Olivieri et al., 2000). Another study demonstrated that Hg ions coordinate with Cys291 of the second repeated (R2) of the MT-binding domain of Tau increasing the heparin-induced aggregation, and a conformational change in Tau demonstrated by circular dicroism (CD) analysis (Yang et al., 2010). On the other hand, Cd appears to play a role in Tau hypothesis since it promotes the aggregation of this protein. It was shown that Cd (II) accelerates heparin-induced aggregation of the third repeated (R3) of Tau. The binding of Cd (II) to the dimeric R3 produces changes on its conformation demonstrated by CD (Jiang et al., 2007). Subchronic As administration to rats (NaAsO<sup>2</sup> at 3 and 10 mg/Kg/day/4–12 weeks, intragastric) induced P-Tau, suggesting that As-destabilization and disruption of the cytoskeletal framework may lead to axonal degeneration (Vahidnia et al., 2008). Regarding Pb, it was reported that infantile Pb exposure in cynomolgus monkeys elevated mRNA and protein levels of Tau as well as its transcriptional regulators (Sp1 and Sp3) in aged primates (23 years old). An increase in P-Tau phosphorylation and mRNA and protein levels of cyclin dependent kinase 5 (cdk5, a kinase that phosphorylates Tau) were also observed (Bihaqi and Zawia, 2013). Other studies also reported that maternal (Li et al., 2010a) and early postnatal exposures (Liu et al., 2014) to Pb produced significant increased P-Tau levels and cognitive impairment in mice. Finally, chronic Al exposure caused Tau aggregation, and it was suggested that Al is bound to P-Tau in the Al-NFTs lesions (Singer et al., 1997; Shin et al., 2003). Also, a study showed that Al is able to confer resistance to the degradation of PHFs both in vivo and in vitro (Shin et al., 1994), and it can inhibit the activity of the protein phosphatase 2 (PP2), which is involved in P-Tau de-phosphorylation (Yamamoto et al., 1990).

#### **Pesticides**

There is some evidence suggesting that pesticide exposure can disrupt Tau function. A recent study showed that the administration of the insecticides deltamethrin (pyethroid) and carbofuran (carbamate) to rats (daily administration by gavage/28 days) produced neuronal death in the cortex and hippocampus and a dysfunction in the spatial memory and learning. These alterations were attributed to a reduced expression of synaptic proteins involved in the memory consolidation. Additionally, P-Tau and activation of p-GSK3β (a major kinase that phosphorylates Tau) were observed (Chen et al., 2012b). Similarly, Wills et al. (2012) showed P-Tau in the striatum, through the activation of p-GSK3β, as well as hyperacetylation of α-tubulin in mice treated with PQ (10 mg/Kg, i.p., twice weekly/6 weeks), suggesting a cytoskeleton remodeling (Wills et al., 2012).

### **Nanoparticles**

The effect of NPs on Tau phosphorylation has not been extensively studied. Silica NPs (siNPs) used in medicine are also able to increase P-Tau at Ser262 and Ser396, two phosphorylation sites characteristic of AD. It was demonstrated that this effect was dependent on the activation of the kinase GSK3β in human SK-N-SH and mouse Neuro-2a cells by a mechanism probably mediated by oxidative stress, since ROS were increased in cells exposed to these NPs (Yang et al., 2014).

# Parkinson's Disease

Parkinson Disease is a chronic and progressive neurological disorder characterized by the selective loss of dopaminergic neurons of the substantia nigra pars compacta (SNpc). The cardinal features of the syndrome are related to motor dysfunction including tremor at rest, rigidity, akinesia (or bradykinesia), and postural instability. The motor symptoms appear when at least 60% of dopaminergic neurons are lost and 80–85% of dopamine content in the striatum is depleted (Jankovic, 2008; Wirdefeldt et al., 2011). Additional to the neuronal loss, the main neuropathological hallmark of PD is the presence of Lewy bodies (LB) in the surviving neurons, which are eosinophilic cytoplasmic inclusions containing aggregates of protein such as α-synuclein (α-syn) (Gibb and Lees, 1988; Spillantini et al., 1998). PD is the second most common neurodegenerative disorder after AD. Due to the lack of specific/differential diagnostic biomarkers, the diagnosis of PD is based on clinical criteria of specific cardinal motor signs of the disease and on the response to levodopa. PD diagnosis is confirmed by the depletion of brain stem pigmented neurons and the presence of LB at necropsy, this is the reason of the misclassification of PD cases (about 10–15%) (Schrag et al., 2002; Jankovic, 2008). There is no cure for PD, and the existing therapies only provide brief relief of motor symptoms through improving the dopamine deficit or by surgical methods. This highlights the need of research on early specific/differential biomarkers to have more accurate diagnosis of neurodegenerative disorders, as well as biomarkers for the identification of populations at risk to implement neuroprotective therapies (Jankovic, 2008).

As in the case of AD, circulating miRNAs are being studied as differential biomarkers for PD. Some reviews have recently addressed this topic, showing the association of specific miRNAs for some genes involved in PD, such as SNCA and LRRK2 (encoding for leucine–rich repeat kinase 2) with PD development (Goodall et al., 2013; Maciotta et al., 2013). Some studies have reported differentially expressed miRNAs in serum of PD patients not observed in control subjects or in other diseases. For example, Vallelunga et al. (2014) reported two differentially expressed miRNAs (miR-30c and miR-148b) in Italian PD patients (n = 25 vs. 25 healthy controls) (Vallelunga et al., 2014), and another study found that serum levels of miR-29c, miR-29a, and miR-19b were down-regulated in PD patients (n = 65 vs. 65 healthy controls) from Barcelona, Spain (Botta-Orfila et al., 2014). Also, a reduced expression of miR-34b and miR-34c in several brain areas including the substantia nigra of PD patients (n = 11 vs. 6 healthy controls) was detected; interestingly the misregulation of miR-34b/c was observed in patients in premotor stages of the disease. Additionally, these miRNAs were deregulated in differentiated SH-SY5Y dopaminergic neuronal cells, which was associated with altered mitochondrial function, oxidative stress and ATP depletion, as well as decreased protein levels of DJ1 (a mitochondrial peroxidase) and Parkin (an E3 ubiquitin ligase) that are associated with the familial form of PD (Miñones-Moyano et al., 2011).

Although the research on PD has rapidly advanced, the molecular mechanisms involved are still unclear and its etiology is complex. Several molecular mechanisms of neuronal death in PD pathogenesis have been described including mitochondrial dysfunction, impairment of protein quality pathways, oxidative/nitrative stress, microglia activation and inflammation. These mechanisms converge and are consistent with a major role of oxidative stress in PD, which damage organelles and proteins leading to increased protein aggregates (e.g., α-syn), that in turn overwhelms the degradation systems leading to a self-perpetuating cycle and further oxidative stress (Wirdefeldt et al., 2011; Goldman, 2014). The evidence in postmortem PD brains supports these mechanisms, as well as a decreased in reduced GSH levels, α-syn aggregation, proteasome impairment and autophagy dysfunction (review in Navarro-Yepes et al., 2014).

A fraction of PD occurrence has a clear familial inheritance and it is related to mutations in at least 6 genes that have been associated with PD onset. The identification of genes such as SNCA or PARK1 encoding for α-syn (maybe involved in the regulation of dopamine release and transport), LRRK2 or PARK8 encoding for LRRK2 (or Dardarin), PARK7 encoding for DJ1, PARK6 or PINK1 encoding for PTEN-induced putative kinase 1 (PINK1, a mitochondrial kinase), and PARK2 encoding for Parkin have provided clues about the molecular mechanisms involved in its pathogenesis (Corti et al., 2011; Cookson, 2012). However, 90% of PD cases are sporadic and cannot be attributed only to genetic factors, which suggests that PD have a multifactorial etiology (Goldman, 2014). In addition to the aging, which is the main risk factor for PD (Tanner and Goldman, 1996), epidemiological evidence suggests that the exposure to environmental toxicants, mainly pesticides, metals and solvents could increase the risk of developing PD, and factors such as tobacco consumption can protect against PD development (**Figure 2**; Hatcher et al., 2008; Gao and Hong, 2011).

# Metals

It has been proposed that chronic exposure to heavy metals such as iron, Pb and Mn and their combinations can be associated with an increased risk of developing PD, since they accumulate in the substantia nigra and generate oxidative stress. However, epidemiological evidence is controversial (Lai et al., 2002). The epidemiological evidence of Pb association with PD is more consistent because the accumulative lifetime exposure can be estimated through Pb concentration in bone that has a half–life of years to decades. Initially, Kuhn et al. (1998) reported that 7 out of 9 postal workers exposed to leadsulfate batteries for up to 30 years developed parkinsonian symptoms, suggesting that Pb intoxication may play a role in the occurrence of these symptoms (Kuhn et al., 1998). Coon et al. (2006) evaluated this association in 121 PD patients vs. 414 controls and found that chronic Pb exposure (evaluated by Pb concentrations in tibial and calcaneal bones) increased 2–fold the risk of PD (OR = 2.27, 95% CI, 1.13–4.55) for individuals in the highest quartile of lifetime Pb exposure relative to those in the lowest quartile (Coon et al., 2006). In the same way, it was reported that the cumulative exposure to Pb increases the risk of PD (OR = 3.21, 95% CI, 1.17–8.83) in 330 PD patients (vs. 308 controls) recruited from 4 clinics for movement disorders in Boston, MA area (Weisskopf et al., 2010b), and the exposure to Pb for more than 20 years showed a stronger association with PD risk in a health system populationbased case-control study (144 cases vs. 464 controls) from the metropolitan Detroit area (Gorell et al., 2004). At the molecular level, Pb exposure significantly decreases the dopamine release and the dopamine D1 receptor sensitivity post-synaptically in microdialysate samples from rats subchronically exposed to Pb (50 ppm/90 days) (Kala and Jadhav, 1995), and in rats treated with 250 ppm of Pb for 3–6 weeks through drinking water (Tavakoli-Nezhad and Pitts, 2005). Furthermore, it increases the lipid peroxidation and reduces the antioxidant cell capacity (Sandhir et al., 1994), and causes fibrillation and aggregation of α-syn (Yamin et al., 2003), which induces hippocampal injury and decreases the ability of learning and memory in rats exposed to 0–300 ppm of Pb (Yamin et al., 2003; Zhang et al., 2012).

Manganese is an essential element with important physiological functions for cellular homeostasis. The epidemiologic evidence does not provide sufficient support for an association between Mn exposure and PD risk (Wirdefeldt et al., 2011; Mortimer et al., 2012). Only one case–control study (144 cases vs. 464 controls) in a population from the metropolitan Detroit area reported an increase of PD risk when the exposure to Mn was over 20 years (OR = 10.63, 95% CI, 1.07–105.99) (Gorell et al., 2004). However, occupational or environmental exposures to Mn have been associated with a neurological syndrome that include cognitive deficits, neuropsychological abnormalities and Parkinsonism (Guilarte, 2013). Mn was related to PD since 1837, when it was noted that high Mn exposures caused a severe and debilitating disorder known as ''manganism'' or manganese–induced Parkinsonism, which consists on an extra pyramidal syndrome that resembles the dystonic movements associated with parkinsonian symptoms (Couper, 1837; Jankovic, 2005), but it is clinically distinct from PD since patients do not respond to dopamine replacement therapies (Cook et al., 1974; Huang et al., 1993; Lu et al., 1994). Several cases of Mn–induced Parkinsonism have been reported in individuals whose professions involve prolonged contact with high atmospheric levels of Mn such as welders, miners and smelters (Rodier, 1955; Wang et al., 1989; Lee, 2000). Several investigations have shown that sustained exposure to low-concentrations (below the current US standard of 5.0 mg/m<sup>3</sup> ) is consistent with early manganism, suggesting that Mn is a neurotoxic chemical (Park, 2013). Patients with manganism and primates experimentally intoxicated with Mn consistently show damage to the globus pallidus, which is in contrast with PD where there is a preferential degeneration of dopamine neurons in the SNpc and preservation of the pallidum (Perl and Olanow, 2007). Likewise, it was observed microglia activation in the substantia nigra pars reticulate (SNpr) and SNpc in Cynomolgus macaques exposed to Mn (5–6.7 mg/Kg/week/10 months) (Verina et al., 2011). In vitro,

it has been observed that Mn treatment (50–300 µM MnCl2/ 3–48 h) induces cytochrome C release and activation of caspases 9 and 3, as well as protein aggregation in N27 dopaminergic neuronal cells that stably express α-syn (Harischandra et al., 2015).

Iron is an essential element transported into the brain through the transferrin receptor and divalent metal transporter 1 (DMT1; Zheng and Monnot, 2012). It has been evaluated in relation to the risk of PD in few epidemiological studies without convincing evidence (Rybicki et al., 1993; Logroscino et al., 2008; Miyake et al., 2011; Abbott et al., 2012). However, iron and its deregulated homeostasis have been proposed to have a role in the pathogenesis of PD because of its prooxidants characteristics that may lead to ROS generation via Fenton and Haber–Weiss reactions (Stohs and Bagchi, 1995; Sian-Hulsmann et al., 2011). The substantia nigra has the highest levels of iron in the human brain, probably due to the presence of neuromelanin in pigmented SNpc dopaminergic neurons that have an impressive capacity of chelating metals, iron in particular; however, this may be a dual-edged sword that may increase their vulnerability since iron may react with ROS produced from dopamine metabolism and promote the further generation of highly toxic radicals (Zecca et al., 2002, 2004). Alterations in iron distribution have been observed in the substantia nigra of PD postmortem brains (Dexter et al., 1987, 1991; Hirsch et al., 1991). On the other hand, it was observed in postmortem samples that although the total iron concentration in the whole substantia nigra was not significantly different between parkinsonian and control samples, there was an increase in the free-iron concentration and a decrease in iron–binding ferritin levels, ferritin sequestrates the excess of iron under physiological conditions (Wypijewska et al., 2010). Likewise, it was reported that free-iron induces fibrillation and aggregation of α-syn in a dose- and timedependent way in SK-N-SH cells incubated with ferric iron (1–10 mmol/L/24–48 h) (Li et al., 2010b). Mice administered with iron (120 µg/g of carbonyl iron, oral gavage) at a dose equivalent to that found in iron-fortified human infant formula (12 mg/L of iron) from days 10 to 17 post-partum (an equivalent period to the first human year of life) showed a progressive midbrain neurodegeneration and enhanced vulnerability to toxic injury at 12 and 24 months of age (Kaur et al., 2007).

#### Pesticides

The hypothesis that pesticide exposures may be related to PD development was prompted by the discovery that intravenous injection of 1-methy l-4pheny l-1, 2, 3, 6-tetrahydropyridine (MPTP), a byproduct of the synthesis of heroin, developed a Parkinson syndrome clinically indistinguishable from PD (Langston et al., 1983); subsequent findings showed that MPTP selectively damaged dopaminergic neurons in the substantia nigra (Langston and Ballard, 1983; Langston et al., 1984). Since then, environmental factors with similar toxicological profiles have received attention as potential risk factors for PD.

A meta-analysis conducted in 2000 evaluated the association between pesticide exposures and PD in 19 case–control studies published between 1989 and 1999. Authors showed that most studies found an elevated risk of PD with the exposure to pesticides, the calculated combined OR was 1.94 (95% CI, 1.49–2.53); similar ORs were observed in studies conducted in United States, Asia, Europa and Canada. Additionally, it was observed that the risk of PD increased with longer exposure times, with an OR of 5.81 (95% CI, 1.99–16.97) for ≥10 years of exposure; however, specific types of pesticides were not identified (Priyadarshi et al., 2000). Subsequently, Brown et al. (2006) reviewed 31 case–control studies published until 2003, and found that about half of them reported significant associations between pesticide exposure and PD risk with ORs from 1.6–7. Interestingly, in most studies, authors observed a positive association between the exposure to herbicides and insecticides and PD risk, but not with the exposure to fungicides alone (Brown et al., 2006).

In line with this, a recent review by Freire and Koifman (2012) analyzed the epidemiological evidence published between 2000 and 2011, including ecological, cross–sectional, prospective and case–control studies. They found that 7 out of the 8 prospective (cohort) studies provided evidence of an association between pesticide exposure and PD, reporting risk estimates of 2-fold or higher. Among 23 case–control studies, 13 studies reported a significant increased risk of PD for the professional use of pesticides in comparison with unexposed controls, with ORs ranging from 1.1 to 1.4, which is in agreement with the review of Priyadarshi in the 1990's (Freire and Koifman, 2012). Furthermore, van der Mark et al. (2012) performed a systematic review and calculated the summary risk ratio (sRR) from 39 case–control studies, 4 cohort studies and 3 cross–sectional studies. When a job–exposure matrix was constructed, a higher sRR (2.5, 95% CI, 1.5–4.1) was observed compared with self–reported exposure evaluation (1.5, 95% CI, 1.3–1.8). This meta–analysis found a positive association between PD and insecticides (sRR = 1.50, 95% CI, 1.07–2.11), and herbicides (sRR = 1.40, 95% C, 1.08–1.81), but not with fungicides (sRR = 0.99, 95% CI, 0.71–1.40) (van der Mark et al., 2012), in agreement with Brown et al. (2006) and Freire and Koifman (2012). Other factors related to pesticide exposure such as well–water consumption, farming, and rural living have been associated with an increased PD risk. The meta–analysis of Priyadarshi et al. (2001) found a combined OR of 1.56 (95% CI, 1.18–2.07) for rural living, 1.42 (95% CI, 1.05–1.91) for farming and 1.26 (95% CI, 0.97–1.64) for well–water consumption. However, whether of these factors are independent risk factors or correlated with pesticide exposure could not be determined (Priyadarshi et al., 2001).

In support to epidemiological evidence, increased levels of some pesticides have been quantified in postmortem brains from PD patients. High concentrations of some OCl pesticides have been observed in PD cases compared with controls, including dieldrin, lindane, and p-p-DDE (Fleming et al., 1994; Corrigan et al., 1998, 2000). In the same way, 2 epidemiologic studies reported a significant association (OR ranging from 1.3 to 1.8) between dieldrin use and PD in farmers participants in the Agricultural Health Study (AHS; Kamel et al., 2007; Tanner et al., 2011). Another nested case–control study within the Finnish Mobile Clinic Health Examination Survey in Finland, with serum samples collected during 1968–1972, observed that increasing serum concentrations of dieldrin were associated with an increased PD risk (OR = 1.95, 95% CI, 1.26–3.02) in 68 cases vs. 183 controls restricted to never smokers, while no other OCl pesticide showed an association (Weisskopf et al., 2010a).

The epidemiologic evidence that dieldrin exposure may be associated with PD is supported by toxicological data at molecular level. Dieldrin may cross the blood–brain barrier and remains in lipid-rich tissues such as the brain (Kanthasamy et al., 2005), and it has been shown that it is selectively toxic to dopaminergic neurons and could induce several of the pathologic mechanisms of PD including the depletion of brain dopamine levels, increased ROS in nigral dopaminergic neurons, inhibition of mitochondrial oxidative phosphorylation that lead to a reduction of cellular ATP production, alteration of the mitochondrial membrane potential and cytochrome C release in animal models such as rats and mice chronically exposed to dieldrin (0.3–3 mg/Kg/day in the diet) (Bergen, 1971; Wagner and Greene, 1978; Purkerson-Parker et al., 2001; Hatcher et al., 2007), and in primary mesencephalic cultures or dopaminergic cell lines (0.01–300 µM) (Sanchez-Ramos et al., 1998; Kitazawa et al., 2001, 2003; Kanthasamy et al., 2005). Aggregation of α–syn, ubiquitin–proteasome impairment function (Uversky et al., 2001; Sun et al., 2005) and microglia activation (Mao and Liu, 2008) have also been observed.

Paraquat is a quaternary nitrogen herbicide used worldwide. Due to its structural similarity to MPP (the active metabolite of MPTP), it was thought to be toxic to dopaminergic neurons and thus might be related to PD. The possible association between PQ and PD received attention from the study of Liou et al. (1997) performed in PD patients (120 patients and 240 controls) in Taiwan, in which the pesticide use was associated with an increased risk of developing PD, being higher for those individuals who reported using PQ (Liou et al., 1997). Likewise, Tanner et al. (2011) reported a significant association between PD and the use of oxidative pesticides, including PQ (OR = 2.5, 95% CI, 1.4–4.7) in professional pesticide applicators (110 cases and 358 controls) (Tanner et al., 2011). Similarly, other epidemiologic studies have associated the exposure to PQ with PD (Hertzman et al., 1990; Ascherio et al., 2006; Kamel et al., 2007; Wang et al., 2011).

Paraquat is taken up into dopaminergic terminals by the dopamine transport and organic cation transporter 3 (Rappold et al., 2011), and causes cellular toxicity by oxidative stress through the cellular redox cycling generating superoxide radical by the oxidation of NADPH, which in turn impairs the restauration of GSH levels and thus the activity of several antioxidant systems (Berry et al., 2010; Franco et al., 2010). It has been observed that repeated administrations of PQ to adult mice and rats (5–10 mg/Kg/ week/at least 3 weeks, i.p.) increase ROS levels in the striatal homogenate, induce a dose-dependent decrease in dopaminergic neurons from the substantia nigra, a decline in striatal dopamine nerve terminal density, and a neurobehavioral syndrome characterized by reduced ambulatory activity (Brooks et al., 1999; McCormack et al., 2002; Kuter et al., 2010). PQ also reproduces other biochemical and neuropathological characteristics of human Parkinsonism such as microglia activation (Wu et al., 2005; Purisai et al., 2007), α-syn up-regulation and fibrillation (Uversky et al., 2001; Manning-Bog et al., 2002), increases lipid peroxidation (increase of 4-hydroxynonenals) (McCormack et al., 2005), alters parkin solubility promoting its intracellular aggregation (Wang et al., 2005), induces a proteasome dysfunction in SH-SY5Y cells (Ding and Keller, 2001; Yang and Tiffany-Castiglioni, 2007), as well as in homogenates from postmortem PD brains (McNaught and Jenner, 2001; McNaught et al., 2002), impairs mitochondrial function at the level of complex III to generate ROS (Castello et al., 2007; Drechsel and Patel, 2009), promotes cytochrome C release (González-Polo et al., 2004; Fei et al., 2008), induces GSH depletion (Schmuck et al., 2002; Kang et al., 2009), and causes cell injury leading to apoptotic cell death (Berry et al., 2010; Franco et al., 2010). PQ has been used as a toxicological model for PD that has permitted getting important information about the mechanisms involved in the neurodegeneration associated with PD (Gao and Hong, 2011).

Rotenone, an OP insecticide has also been associated with an increased risk of PD. Two epidemiological studies found an association between rotenone exposure and PD risk, reporting an increased risk of 10–fold (OR = 10.0, 95% CI, 2.9–34.3) in East Texas farmers (Dhillon et al., 2008), and 2.5-fold (OR = 2.5, 95% CI, 1.3–4.7) in PD cases (n = 110) compared with controls (n = 358) from professional pesticide applicators participants in the AHS (Tanner et al., 2011). Rotenone can freely cross the blood–brain barrier and is a well-established mitochondrial toxin that specifically inhibits the complex I (NADH–dehydrogenase) of the electron transport chain leading to ATP depletion, energy failure and mitochondrial ROS production, which in turn induces cytochrome C release and apoptotic cell death (Clayton et al., 2005; Radad et al., 2006; Sherer et al., 2007). It has been shown that, like MPTP, rotenone treatment in animal models (1.5–3 mg/Kg/day/up to 3 weeks) reproduces features of PD such as bradykinesia, postural instability and/or rigidity, reduces the tyrosine hydroxylase-positive neurons in the substance nigra, induces a loss of striatal dopamine, and the accumulation of α-syn and poly-ubiquitin positive aggregates in remaining dopaminergic neurons (Betarbet et al., 2000; Sherer et al., 2003; Cannon et al., 2009). Likewise, Betarbet et al. (2006) observed that chronic administration of 3.0 mg/Kg/day of rotenone for up to 5 weeks to male rats caused the oxidation of DJ-1, accumulation of α-syn, and proteasomal impairment (Betarbet et al., 2006). These effects were also observed in neuroblastoma SK-N-MC cells treated with rotenone (5 nM/4 weeks), as well as a loss of GSH, oxidative DNA and protein damage and caspase-dependent death (Sherer et al., 2002; Betarbet et al., 2006). Rotenone has also the capacity to activate microglia (Sherer et al., 2003); Gao et al. (2002) demonstrated that the addition of microglia to primary neuronenriched cultures (neuron/glia cultures) markedly increased the dopaminergic neurodegeneration induced by rotenone (1 nM/8 days), and this neurotoxicity was attenuated by the inhibition of NADPH oxidase or scavenging the superoxide radical that is liberated from the microglia (Gao et al., 2002). Since rotenone recapitulates several mechanisms of PD pathogenesis, this pesticide is currently used as a toxicological model to study the underlying mechanisms on the PD development.

Despite the widespread use of OP insecticides such as malathion, methyl parathion, chlorpyriphos and diazinon, not many studies have evaluated the association between specific OP and PD risk. Dhillon et al. (2008) found a 2–fold increase (OR = 2.0, 95% CI, 1.02–3.8) in the risk of PD in Texan agricultural workers exposed to chlorpyriphos (cases = 100, controls = 84) (Dhillon et al., 2008). An increased risk of PD was also observed in rural residents from California possibly exposed to high levels of chlorpyriphos (OR = 1.87, 95% CI, 1.05–3.31) and diazinon (OR = 1.75, 95% CI, 1.12–2.76) through the consumption of contaminated well–water (Gatto et al., 2009). One study conducted in a population from the Group Health Cooperative (GHC) in Western Washington State occupationally exposed to methyl parathion found a high risk of PD (OR = 8.08, 95% CI, 0.92–70.85), although the association was not statistically significant (Firestone et al., 2005). This is particularly relevant, because parkinsonian effects have been reported in cases of patients intoxicated with OP (Bhatt et al., 1999).

# Solvents

Solvents are widespread used due to their commercial applications, including metal degreasing, dry cleaning, and as ingredients of paint thinners and detergents. Some solvents are lipophilic and thus easily absorbed by the central and peripheral nervous system tissues (Lock et al., 2013). There are isolated cases of acute Parkinsonism associated with large solvent exposures such as in workers exposed to n-hexane (Pezzoli et al., 1989), and toluene (Papageorgiou et al., 2009), among others. There is no consistent evidence of the association of solvent exposure and PD (Wirdefeldt et al., 2011). One case-control study based on a questionnaire reported an increased risk of PD by the exposure to organic solvents (OR = 2.78, 95% CI, 1.23–6.26) in 86 PD patients and 86 controls from the Emilia-Romagna region in Italy (Smargiassi et al., 1998). Another case–control study reported an increased risk of PD when the exposure to solvents was above 20 years (OR = 3.59, 95% CI, 1.26–19.26) in 182 cases (vs. 422 controls) identified through death certificates of the Rolls-Royce PLC national pension fund archive from employees of five manufacturing locations in United Kingdom who had any mention of PD (McDonnell et al., 2003).

Trichloroethylene (TCE) is one of the specific solvents that has been investigated in detail (Goldman, 2014). Some clinical case reports have reported the onset of PD in workers exposed to TCE through chronic inhalation and dermal exposure by handling TCE, suggesting a potential link between the exposure to TCE and PD (Kochen et al., 2003; Gash et al., 2008). More recently, an epidemiologic study in 99 twin pairs discordant for PD showed that the exposure to TCE was associated with a 6–fold increased risk of PD (OR = 6.1, 95% CI, 1.2–33) (Goldman, 2014). In animal models, TCE may recapitulate several key pathological features of PD. The systemic exposure of adult rats to TCE (1000 mg/Kg/day/5 days a week/2 and 6 weeks, oral gavage) inhibits mitochondrial complex I enzyme activity, increases oxidative stress markers, activates the microglia, induces nigral αsyn accumulation and a significant loss of dopaminergic neurons on the SNpc in a dose-dependent manner, as well as defects in the rotarod behavior test (Liu et al., 2010). In a similar way, the administration of n-hexane and its metabolite 2, 5-hexanedione (400 mg/Kg/day/5 days a week/6 weeks, i.p.) to mice caused that both chemicals reduced the striatal dopamine concentration by 38 and 33%, respectively, but neuronal cell loss was not confirmed (Pezzoli et al., 1990). On the other hand, there is no evidence that acute or subchronic exposure to toluene promotes the degeneration of the nigrostriatal dopamine system (Lock et al., 2013).

# Nanoparticles and PD

Nanoparticles are an important alternative in the development of treatment strategies for neurodegenerative diseases due to their small particle size, large surface and high drug loading efficiency, which allow them to cross the blood-brain barrier and efficiently release specific drugs (Li et al., 2014b; Leyva-Gómez et al., 2015).

Although some NPs are being used in therapies for PD, no epidemiological studies are available associating them with PD risk. However, there is evidence suggesting that they could contribute to alter the molecular mechanisms involved in the pathogenesis of PD. Thus, it was reported that intranasal instillation of SiO2-NPs (20 µg/day/1–7 days) to rats resulted in their presence in the striatum, the induction of oxidative damage, an inflammatory response, and depleted dopamine concentration and tyrosine hydroxylase levels, suggesting that these NPs have a negative impact on striatal dopaminergic neurons (Wu et al., 2011). Another report in adult zebrafish exposed to SiO2-NPs (300 and 1000 µg/mL; 15 and 50-nm of size) showed alterations in neurobehavioral parameters (general, cognitive behavior and locomotive activity), with the most significant effects observed with the smallest NPs, similar to those observed in neurodegenerative diseases (Li et al., 2014b). In vitro studies also support the potential contribution of NPs in PD development. The exposure of dopaminergic neurons (PC12 cells) to SiO2-NPs (25–200 µg/mL/24 h) triggered an oxidative stress, disturbed the cell cycle, induced apoptosis, and activated the p53-mediated signaling pathway (Wu et al., 2011); while the exposure of these cells to TiO2-NPs (50, 100 and 200 mg/mL/24 h) induced a dose–dependent increase in the expression and aggregation of α-syn, as well as a reduction of the expressions of Parkin (E3 ligase), and the ubiquitin C-terminal hydrolase (UCH-L1), these events were associated with increased oxidative stress (Wu and Xie, 2014). Also, the exposure PC12 cells to iron oxide (Fe2O3-NPs; 0.15–15 mM) decreased the neurite growth in response to the nerve growth factor (NGF) (Pisanic et al., 2007). Likewise, citratecapped gold nanoparticles (Au-NPs; 0.3–32 nM, 10–22 nm) produced a dose-dependent aggregation of purified α-syn, being strongest for the smallest NPs (Alvarez et al., 2013). In contrast, the administration of Neurotensin (NTS)-polyplex NPs (8.5 nmol/Kg, i.v), a nanocarrier gene with a potential for nanomedicine-based applications for PD treatment, to BALB/c mice does not produce systemic inflammatory (up to 24 h after treatment) nor hepatic cytotoxicity (at 24 and 96 h after treatment), supporting the safety of these NTS-polyplex NPs as a potential therapeutic approach (Hernandez et al., 2014).

# Early Exposure to Environmental Factors and AD or PD Development: Epigenetic Evidence

Epigenetic DNA modifications include DNA methylation, histone post-translational modifications (mainly acetylation) and miRNAs (Holliday, 2006). DNA methylation is one of the most studied epigenetic modifications that influence the gene expression. It involves the addition of methyl groups to cytosine bases located at cytosine–phosphate–guanine (CpG) sites by the action of DNA methyltransferases (DNMTs). Alterations in DNA methylation on the promoter regions of genes regulate the gene expression of important processes such as embryonic development, cellular differentiation and aging (Bird, 2002). Increasing evidence suggests that epigenetic changes in the developing embryo that may play important roles in the susceptibility to diseases in later life (imprinted disease phenotypes) result from maternal exposures to environmental stimuli at critical periods of development. This suggests that a short exposure to chemicals could be memorized through epigenetic mechanisms long after the chemical trigger has gone (Jang and Serra, 2014), and recent studies have suggested that an epigenetic component could be involved in neurodegenerative diseases related to environmental factors (Marques et al., 2011).

The latent brain expression of genes observed in animals developmentally exposed to an environmental contaminant may be mediated through epigenetic pathways that are regulated via the DNA methylation. While the conditions leading to early life hypo- or hyper-methylation of specific genes are not known, both can induce oxidative DNA damage; for instance the hypo-methylation of APP gene increases its expression driving the overproduction of APP and Aβ levels, which in turn facilitate the ROS production damaging the DNA, and producing neuronal loss. While the hyper-methylation affects the gene transcription and DNA repair pathways. Therefore, both changes in DNA methylation can impact gene expression and imprint susceptibility to oxidative DNA damage in the aged brain (Zawia et al., 2009). Thus, it is suggested that Pb interferes with the DNA methylating capacity, thus altering the expression of AD-related genes. The study performed in aged monkeys developmentally exposed to Pb revealed a reduced activity of brain Dnmt, and the exposure of mouse primary cells from the cerebral cortex to Pb (0.1 µM) resulted in a similar effect on Dnmt1 activity a week after 24 h-treatment (Wu et al., 2008). Also, Bihaqi and Zawia (2012) showed a significant latent increase in AD biomarkers an a reduction in the protein and mRNA levels of DNA methylating enzymes Dnmt1 and Dnmt3a, and methyl CpG binding protein 2 (MeCP2) in differentiated SH-SY5Y cells treated with Pb (5–100 µM/48 h) and analyzed 6 days later (Bihaqi and Zawia, 2012). Aberrant CpG methylation in APP, Tau and GSK3β genes was reported in post-mortem brains (Iwata et al., 2014). In addition, it suggested that reduced levels of CpG methylation in the promoter of APP could be mediated by the oxidation of guanine (8-oxdG) (Zawia et al., 2009); this is because the oxidation of guanine in CpG dinucleotides inhibits adjacent cytosine methylation (Weitzman et al., 1994). On the other hand, Cd, another metal involved in AD pathology, reduces the enzymatic activity of Dnmt in rat liver cell cultures (Poirier and Vlasova, 2002), but this effect has not been evaluated in cerebral cells. While a study showed that subchronic As exposure (3 and 36 ppm/from gestation until 4 months of age) altered the methylation of genes involved in neuronal plasticity, including reelin (RELN) and protein phosphatase 1 (PP1), which was associated with memory deficits (Martínez et al., 2011). Regarding other compounds, the perinatal exposure to permethrin (34 mg/Kg/daily, by gavage from postnatal day 6–21) to mice showed altered brain functions including biomarkers of maintenance of dopaminergic neurons, and impairment of spatial memory at 6 months of age (Nasuti et al., 2013).

The relation between epigenetic modifications and PD has been less studied; however, a potential role of DNA methylation in the promoter of α-syn encoding gene (SNCA) in the neuropathogenesis of PD has been suggested, considering that αsyn is a fundamental component of LB, the main hallmark of PD (Lu et al., 2013). A DNA hypomethylation of SNCA was reported in the substantia nigra of sporadic PD patients, suggesting that it might contribute to the dysregulation of SNCA expression in PD (Jowaed et al., 2010; Matsumoto et al., 2010). In addition, increased SNCA mRNA levels were observed in SNpc of PD (Chiba-Falek et al., 2006), and reduced levels of Dnmt1 have been observed in postmortem brains from PD and dementia with LB (DLB) patients, as well as in brains of α-syn transgenic mice; authors suggest that this effect could be a novel mechanism of epigenetic dysregulation in LB-related diseases such as PD (Desplats et al., 2011). Finally, a lesser degree of methyation of the TNFα promoter, a key inflammatory cytokine associated with dopaminergic cell death was observed in the SNpc from PD patients, predisposing to an increase neuronal vulnerability to inflammatory reactions (Mogi et al., 1996; Pieper et al., 2008).

Environmental factors associated with an increased risk of PD such as pesticides can alter the expression of genes by epigenetic mechanisms (Kwok, 2010). It was reported that pre-treatment with 5-aza-2'deoxycytidine (5'-aza-dC, a DNMT inhibitor) exacerbated the dopaminergic neuron damage induced by PQ, MPP+, 6-hydroxydopamine (6-OHDA) and rotenone treatment, and induced oxidative stress, the transcriptional up-regulation of α-syn, and demethylation of the α-syn promoter (Wang et al., 2013). Likewise, the folate deficiency sensitizes mice to MPTP-induced PD-like pathology and motor dysfunction (Duan et al., 2002); it is well known that folate deficiency alters the development of human nervous system (Greenblatt et al., 1994).

On the other hand, it was reported that the exposure to environmental neurotoxicants associated with PD during early life or pregnancy can determine the progressive damage of the substantia nigra years before the onset of clinical parkinsonism, as well as to increase the vulnerability to effects of a second environmental factor (two–hit model) (Logroscino, 2005). A study in C57BL/6 mice daily treated with pQ (0.3 mg/Kg) or maneb (1 mg/Kg) or PQ + maneb from postnatal day 5–19 and then re-exposed as adults to PQ (10 mg/Kg) or maneb (30 mg/Kg) or PQ + maneb (twice a week/3 weeks) showed that dopaminergic cell loss and decreased dopamine levels were amplified by the adult re-challenge to the pesticides, suggesting that the developmental exposure to neurotoxins enhanced the adult susceptibility to a new toxic insult (Thiruchelvam et al., 2002). Similarly, prenatal exposure of pregnant C57BL/6J mice to PQ (0.3 mg/Kg) or maneb (1 mg/Kg) altered the development of the nigrostriatal system and enhanced its vulnerability to neurotoxins later in life, which could contribute with the development of PD during aging (Barlow et al., 2004).

Although there is no direct evidence linking early exposure to environmental pollutants and epigenetic changes with increased susceptibility to LOPD, there is a plausible association based on the following considerations: (1) epigenetic alterations have been observed in PD brains; (2) the exposure to environmental factors is associated with an increased risk of LOPD development and factors such as pesticides and metals can alter mechanisms of epigenetic regulation such as DNA methylation; and (3) early exposure to environmental pollutants might be associated with LOPD later in life. Further studies are needed to confirm this hypothesis in this promising research field to understand the mechanisms underlying the long-term effects of the environment on the PD development.

# Concluding Remarks

The emerging association between exposures to several toxic compounds with neurodegenerative diseases is of considerable public health importance, given the increasing dementia prevalence, the negative social and economic consequences of neurodegeneration-related disabilities, and the increasing environmental pollution in some geographic areas worldwide. Some of the epidemiological studies show not consistent results on getting significant estimates of hazard risk for AD or PD, mainly due to some limitations that include the difficulty on accurate diagnosis for AD or PD cases due to the lack of specific biomarkers, the deficiency to accurately assess chronic exposures, and/or the lack of inclusion of important confounding variables such as co-exposure to toxic compounds, genetic variants and lifestyle among others. Nevertheless, epidemiological studies along with experimental data have led to highlight the potential risk to develop these degenerative diseases due to the exposure to environmental pollutants such as metals, NPs and pesticides, among others. Interestingly, these pollutants show similar mechanisms of toxicity, which converge in a generalized mechanism based on the generation of oxidative stress that leads to common hallmarks of both neurodegenerative disorders. For example, the generation of oxidative stress by increasing the production of ROS and/or deregulating the antioxidant enzymes promotes the formation

# References


of protein aggregates such as Tau, Aβ or α-syn. This in turn overwhelms the degradation systems, and produces the activation of the glia inducing neuroinflammation, a process that per se increases the generation of further oxidative stress leading to a self-perpetuating cycle, and finally to neuronal loss of specific brain region such as the hippocampus and cerebral cortex in AD and substantia nigra in PD. The oxidative stress induced by these neurotoxicants activates/inhibits signaling pathways leading to augmented/diminished activity of enzymes that promote the accumulation of toxic materials in neural cells such as damaged/aberrant proteins, Aβ in AD or α–syn in PD and oxidative byproducts, or the oxidation of DNA that can alter genetic or epigenetic regulation. Furthermore, the link between early life exposure to environmental factors and the origin of neurodegenerative diseases is getting attention and can help to clarify the role of the environment on the development of these degenerative diseases. On the other hand, the lack of specific/differential biomarkers for AD or PD limits the early diagnosis and then the timely treatment. In this regard, specific circulating miRNAs have been associated with pathological processes such as AD and PD, therefore they are promising non-invasive biomarkers for these neurological diseases. Additionally, the identification of biomarkers to determine the past exposure to environmental pollutants is of vital importance for a better and opportune management of these diseases. Thus, as we have more knowledge of the risk from the exposure to environmental pollutants, more well-designed epidemiological studies (controlling for as many variables as possible and with high sample sizes) are necessary to improve the quality of life of elderly and to prevent the development of neurodegenerative diseases worldwide.

induced by conditions representing physiological acidosis. J. Biol. Chem. 273, 12817–12826. doi: 10.1074/jbc.273.21.12817


density lipoprotein receptor protein-1 and protein kinase C. Neurotoxicology 31, 524–532. doi: 10.1016/j.neuro.2010.05.004


pesticide paraquat exposure in rats. A comparison with peripheral tissues. Neurochem. Res. 35, 1121–1130. doi: 10.1007/s11064-010-0163-x


parkinsonism and Parkinson's disease. Neurosci. Lett. 211, 13–16. doi: 10. 1016/0304-3940(96)12706-3


metal sulfate-stressed human astroglial (HAG) primary cell cultures. J. Inorg. Biochem. 105, 1434–1437. doi: 10.1016/j.jinorgbio.2011.05.012


enzymes. Pharmacol. Toxicol. 74, 66–71. doi: 10.1111/j.1600-0773.1994. tb01077.x


in alpha-synuclein overexpressing dopaminergic neuronal cells and enhances susceptibility to apoptotic cell death. J. Pharmacol. Exp. Ther. 315, 69–79. doi: 10.1124/jpet.105.084632


**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.

Copyright © 2015 Chin-Chan, Navarro-Yepes and Quintanilla-Vega. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution and 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.

# Neurological effects of inorganic arsenic exposure: altered cysteine/glutamate transport, NMDA expression and spatial memory impairment

#### **Lucio A. Ramos-Chávez <sup>1</sup> , Christian R. R. Rendón-López <sup>1</sup> , Angélica Zepeda<sup>1</sup> , Daniela Silva-Adaya<sup>2</sup> , Luz M. Del Razo<sup>3</sup> and María E. Gonsebatt <sup>1</sup>\***

<sup>1</sup> Departamento de Medicina Genómica, Instituto de Investigaciones Biomédicas, Universidad Nacional Autónoma de México, Mexico, DF, Mexico

<sup>2</sup> Laboratorio Experimental de Enfermedades Neurodegenerativas, Instituto Nacional de Neurología y Neurocirugía, Mexico, DF, Mexico

<sup>3</sup> Departamento de Toxicología, Centro de Investigación y Estudios Avanzados, Mexico, DF, Mexico

#### **Edited by:**

Victoria Campos, Instituto Nacional de Neurologia y Neurocirugia, Mexico

#### **Reviewed by:**

Annalisa Scimemi, SUNY Albany, USA Takashi Tominaga, Tokushima Bunri University, Japan

#### **\*Correspondence:**

María E. Gonsebatt, Departamento de Medicina Genómica, Instituto de Investigaciones Biomédicas, Universidad Nacional Autónoma de México, A.P: 70228, Ciudad Universitaria, Mexico, DF 04510, Mexico e-mail: margen@unam.mx

Inorganic arsenic (iAs) is an important natural pollutant. Millions of individuals worldwide drink water with high levels of iAs. Chronic exposure to iAs has been associated with lower IQ and learning disabilities as well as memory impairment. iAs is methylated in tissues such as the brain generating mono and dimethylated species. iAs methylation requires cellular glutathione (GSH), which is the main antioxidant in the central nervous system (CNS). In humans, As species cross the placenta and are found in cord blood. A CD1 mouse model was used to investigate effects of gestational iAs exposure which can lead to oxidative damage, disrupted cysteine/glutamate transport and its putative impact in learning and memory. On postnatal days (PNDs) 1, 15 and 90, the expression of membrane transporters related to GSH synthesis and glutamate transport and toxicity, such as xCT, EAAC1, GLAST and GLT1, as well as LAT1, were analyzed. Also, the expression of the glutamate receptor N-methyl-D-aspartate (NMDAR) subunits NR2A and B as well as the presence of As species in cortex and hippocampus were investigated. On PND 90, an object location task was performed to associate exposure with memory impairment. Gestational exposure to iAs affected the expression of cysteine/glutamate transporters in cortex and hippocampus and induced a negative modulation of NMDAR NR2B subunit in the hippocampus. Behavioral tasks showed significant spatial memory impairment in males while the effect was marginal in females.

**Keywords: arsenic, gestational, neurological effects, xCT, EAAC1, GLT1, NMDAR**

#### **INTRODUCTION**

Experimental, as well as epidemiological studies, provide evidence suggesting that both environment and genetics are important components in the development of neuropathologies at early age or later in life. Diet components and chronic exposure to heavy metals and metalloids have been associated with the manifestation of neurological impairments, particularly when exposure occurs during the development and maturation of the nervous system (Winneke, 2011).

Inorganic arsenic (iAs) is an ubiquitous metalloid that is used in wood preservation, as a pesticide, in electronic devices due to its semiconductor capacities and also as a chemotherapeutic agent (ATSDR, 2007). This metalloid which is considered an epidemiologically important natural pollutant can be found in arsenic-containing minerals, ores and groundwater. Globally, more than 200 million of individuals drink water with levels of iAs above the World Health Organization reference value of 10 µg/L. Increased concentrations of iAs have been found in groundwaters in Argentina, Chile, China, India, Mexico, Taiwan and the USA where people are chronically exposed to iAs by drinking water from contaminated wells as a result of geothermal activities, mineral dissolution or deposition and weathering of atmospheric volcanic particles.

Deficits in cognitive functions as evidenced by decreased intelligence, verbal coefficients (Calderón et al., 2001) and impairments in learning and memory (Tsai et al., 2003; Rosado et al., 2007; von Ehrenstein et al., 2007; Asadullah and Chaudhury, 2008; Wasserman et al., 2014) have been associated with chronic exposure to iAs. The neurological and cognitive dysfunctions seem to be dependent on the concentration, timing and duration of exposure (Tyler and Allan, 2014).

In human and in many mammalian species iAs is reduced, methylated into trivalent and pentavalent methylated species and conjugated with glutathione (GSH, L-γ-glutamyl-L-cysteinylglycine, Thomas et al., 2004; Kumagai and Sumi, 2007). These events are associated with the generation of oxidative stress (Kumagai and Sumi, 2007). The presence of iAs and its methylated metabolites have been reported in umbilical cord blood studies from populations at risk (Concha et al., 1998; Parajuli et al., 2013) suggesting that they can cross the placenta and reach the developing fetus. An increasing number of epidemiological and animal model studies have shown that iAs exposure has harmful effects on brain function (Vahter, 2008; Parajuli et al., 2013; Tyler and Allan, 2014). However, there is little evidence of the neurotoxic effects during gestation a crucial development stage that may impact on normal adult life.

Studies in murine models have demonstrated that iAs crosses the blood-brain barrier (BBB) and is methylated in different brain regions that express the arsenic 3 methyltransferase (AS3MT) enzyme (Rodríguez et al., 2005; Sánchez-Peña et al., 2010). Transplacental transfer of As species from pregnant mice to fetus has been documented (Devesa et al., 2006; Jin et al., 2010). Moreover, AS3MT mRNA has been detected in mouse fetuses and embryos suggesting that As could be methylated in fetal tissues (Devesa et al., 2006). iAs methylation requires the presence of S-Adenosyl methionine as the methyl donor and cellular reductants such as thiorredoxin and GSH (Thomas et al., 2004). Thus, the metabolism of iAs consumes GSH, which is the main antioxidant in the central nervous system (CNS; Dringen, 2000). Inadequate GSH availability may modulate iAs biotransformation and determine disease susceptibility.

GSH penetrates the BBB poorly, therefore, CNS GSH levels depend on *de novo* synthesis which is limited by the intracellular availability of the sulfhydryl amino acid L-cysteine (L-cys; Valdovinos-Flores and Gonsebatt, 2012). Under aerobic conditions, L-cys autooxidizes to its disulfide form cystine (Lcys2), which is the predominant form of the aminoacid in plasma (Valdovinos-Flores and Gonsebatt, 2012). Specific membrane transporters such as xCT (SLC7A11)/4F2hc (SLC3A2), also known as the x<sup>−</sup> c L-cys2/glutamate (L-glu) antiporter system, participates in the influx of L-cys for GSH synthesis (Valdovinos-Flores and Gonsebatt, 2012). xCT is widely expressed in both mouse and human brain (Burdo et al., 2006). The xc- system is also an important source of extracellular glutamate and is related to oxidative protection (Shih et al., 2006). However, because xCT uptakes L-cys<sup>2</sup> in exchange for the excitatory Lglu, increased activity of this transporter could be deleterious and lead to excitotoxicity (Lau and Tymianski, 2010). The removal of extracellular L-glu involves EAAT3/EAAC1 (SLC1A1), part of the x-AG system in neurons and GLAST and GLT1 in glia. EAAC1 is also an important transporter for L-cys uptake in neurons (Valdovinos-Flores and Gonsebatt, 2012). Another important amino acid transporter system with wider substrate selectivity than x<sup>−</sup> c or x-AG is the L system. LAT1 (SLC7A5) and LAT2 (SLC7A8) are the catalytic subunits of these amino acid transporters and are linked by a disulfide bridge to the heavy chain 4F2hc. *In vitro* and *in vivo* studies have provided evidence of L-cys transport by both LAT1 and LAT2 (Killian and Chikhale, 2001; Meier et al., 2002).

Glutamate is the most abundant excitatory neurotransmitter in the CNS. Its effects are mediated by ionotropic and metabotropic receptors. The ionotropic receptors (named after the agonists that activate them): α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid (AMPAR) and N-methyl-D-aspartate (NMDAR) are widely expressed in the CNS. A distinct property of NMDAR is that allows the entry of Ca2+, in addition to the passage of K<sup>+</sup> and Na+. Thus, excitatory postsynaptic potentials can increase Ca2<sup>+</sup> levels in the postsynaptic neuron which can potentially act as a second messenger initiating signaling cascades. The activation of NMDAR is also voltage-dependent due to the extracellular blockage by Mg2<sup>+</sup> or Zn2+. Then, the passage of cations (mostly Ca2+) occurs when the blockage is removed by a large number of excitatory inputs or the repetitive firing of the presynaptic cell or both. These properties are considered the bases of synaptic plasticity, learning and memory storage processes.

NMDAR is formed by several protein subunits producing a number of receptor isoforms. The expression of NMDAR subunits is differentially regulated during development and in response to synaptic activity. NR1/NR2A containing NMDAR receptors predominate at synaptic sites in the adult nervous system whereas NR1/NR2B receptors predominate during development and tend to be concentrated at extrasynaptic sites (Paoletti et al., 2013). NR2B subunits modulate the pharmacological and functional properties of the NMDA receptor (Mony et al., 2009). Consequently, NR2B has been implicated in modulating the synaptic function in activities such as learning, memory processing, and feeding behaviors, as well as being involved in a number of human disorders (Mehta et al., 2013). Additionally, the results of some experimental models suggest that exposure to xenobiotics might interfere with the expression of NMDAR subunits NR2A and NR2B during brain development (Olney et al., 2000; Li et al., 2012).

Studies using C3H and CD1 mice show that iAs crosses the placenta modifying gene expression that could lead to aberrant gene expression later in life (Shen et al., 2007; Waalkes et al., 2007). We hypothesized that the gestational exposure to iAs would up-regulate GSH *de novo* synthesis and L-cys<sup>2</sup> influx via xCT and EAAC1 in brain cells. This condition could lead to increased levels of extracellular L-glu and to the modulation of NMDAR expression in brain regions such as cortex and hippocampus, where this receptor participates in learning and memory. Adult CD-1 male and female mice received 20 mg/L of iAs in drinking water for 1 month before mating. Pregnant females received continued exposure during gestation and lactation. At weaning, 50% of the pups continued drinking water with iAs while the rest drank deionized water similar to control animals. Results suggest that arsenic exposure disrupts L-cys and L-glu transport in the hippocampus by the up-regulation of xCT and EAAC1 and down-regulation of GLT1. This altered L-cys and L-glu transport was associated to the negative regulation of NR2B subunits and to impaired spatial memory.

# **MATERIALS AND METHODS**

#### **CHEMICALS**

All chemicals were purchased from Sigma-Aldrich (St Louis, MO, USA) unless otherwise indicated. For western blots, primary rabbit antibodies against xCT, EAAC1, GLAST or GLT1 (ab37185, ab124802, ab416 or ab41621 respectively) were obtained from Abcam, Cambridge, MA, USA. Anti-LAT1 (sc-34554) from Santa Cruz Biotechnology, Santa Cruz, CA, USA. Rabbit anti-NR2A, anti-NR2B or mouse anti-GAPDH (AB1555P or AB1557P, MAB374 respectively) from Millipore, Bedford, MA, USA. Rabbit anti mouse-β-tubulin (T4026) from Sigma-Aldrich. Secondary goat anti-rabbit antibodies were obtained from Cell Signaling Technology (Danvers, MA, USA). For immunofluorescence staining chicken anti-MAP2 (ab5392) from Abcam. Anti-rabbit Alexa Fluor 594 (A11039) and anti-chicken Alexa 488 (A21207) secondary antibodies were obtained from Life Technologies, Carlsbad, CA, USA.

#### **ANIMALS AND TREATMENT**

Seven- to eight- week-old CD-1 mice were obtained from the Animal Care Facility at the Instituto de Investigaciones Biomédicas, UNAM, and were maintained at 23–25◦C under a 12 h light/dark cycle and a relative humidity of 50–60%. Animals had free access to food (Harlan 2018S Diet; Harlan, Indianapolis, IN, USA) and water. Mice were housed in groups of 4 animals per plastic cage and separated by sex. One group of randomly selected mice (12 male and 12 female) received 20 mg/L of (iAs) daily as sodium arsenite via their drinking water for 30 days (exposed group). The same number of animals were assigned to the control group and received drinking water without iAs. The dose of treatment was chosen taking into consideration reports on iAs reproductive toxicity (Golub et al., 1998). Sodium arsenite solutions were prepared freshly daily in deionized water. After 30 days of treatment, each male was mated with one (single) female. Initiation of gestation was estimated by vaginal plug formation. Then, the males were removed, and the female mice were housed individually. Throughout the experiment, water consumption was recorded daily. Body weight was recorded every 6 days during the 30 days prior to mating in both sexes, and on days 0, 7, 14 and 18 of gestation in females. The exposed females continued to receive water with 20 mg/L of iAs during the gestation and lactation period. On postnatal days (PNDs) 1 and 15 randomly selected iAs exposed and control pups from each litter were sacrificed. On PND 1 whole brains were removed. On PND 15 the brain regions could be identified and were dissected on ice to isolate cortex and hippocampus. Tissue samples were immediately frozen by immersion in liquid nitrogen and maintained at −70◦C until processed. Sex differentiation in the offspring was performed based on anogenital length (Suckow et al., 2000). On PND 15 the exposed litter was divided: 50% continued to receive drinking water with iAs while the rest received deionized water similar to control animals until PND 90. These 3 groups, including controls (Control), the group exposed to iAs only during gestation and lactation (iAs-PND 15) and the group exposed to iAs during gestation, lactation and for the first 90 days (iAs-PND 90) were used for the behavioral tests. Control litters continued to drink water without iAs. At weaning, on PND 21, mice in each litter were separated from the mothers.

The experiments were performed following the guidelines stated in the "Principles of Laboratory Animal Care" (NIH publication #85-23, revised 1985) and "Especificaciones técnicas para la producción, cuidado y uso de los animales de laboratorio (Clave NOM-062-ZOO-1999)" of the "Norma Oficial Mexicana de la Secretaría de Agricultura, Ganadería, Desarrollo Rural, Pesca y Alimentación (SAGARPA)" (published August, 2001).

#### **WESTERN BLOTS**

Membrane enriched fractions were obtained from frozen tissue samples as described previously (Schindler et al., 2006) for western blot determination of xCT, EAAC1, LAT1, GLAST, GLT1 and the NMDA receptor subunits NR2A and NR2B. Briefly, frozen tissues were homogenized in 10 volumes of extraction buffer containing 10 mM HEPES, 10 mM NaCl, 1 mM KH2PO4, 5 mM NaHCO3, 5 mM EDTA, 1 mM CaCl2, 0.5 mM MgCl2, 1 mM PMSF, and 10 mg/ml aprotinin and leupeptin. The homogenates were centrifuged at 6,300 g for 10 min at 4◦C. Supernatants were recovered and centrifuged at 100,000 g for 30 min at 4◦C. The pellets were finally suspended in 40 mM Tris-HCl, pH 9.5, 8 M urea and 4% (w/v) Triton X-100. Protein concentrations were determined using a Pierce BCA Protein Assay kit (Thermo Scientific, Meridian Rd, Rockford, USA). The samples (5–40 µg protein per well) were subjected to SDS-PAGE and transferred into nitrocellulose membranes (Bio-Rad Laboratories, Germany). The membranes were blocked with TBS containing 5% Blotto and 0.1% Tween-20 and incubated with the respective primary antibodies. The blots were probed with mouse anti-β-tubulin or anti-GAPDH after stripping, which were used as loading controls. The protein bands were visualized with appropriated HRP-linked secondary antibodies using the ECL Prime western blotting detection reagent (GE Healthcare Bio-Sciences, Pittsburgh, PA); images were captured and densitometric analysis was performed with Image J software version 1.46r software (U. S. National Institutes of Health, Bethesda, Maryland, USA).

#### **IMMUNOFLUORESCENCE**

On PND 15, mice were transcardially perfused with ice-cold 0.9% saline followed by ice-cold 4% paraformaldehyde in phosphate buffer (PB), pH 7.4. Brains were removed, postfixed at 4◦C and successively immersed in 20% and 30% sucrose cryoprotection solutions. Sections (22 µm) were collected in 24 well culture plates filled with 0.9% phosphate buffered saline (PBS), pH 7.4. After 3 washes with PBS + 0.3% Triton X-100 (PBST), the sections were incubated overnight at 4◦C in rabbit anti-xCT (1:100) or rabbit anti-EAAC1 (1:300) and chicken anti-MAP2 (1:800) primary antibodies with a 2% normal horse serum in PBST solution. After 3 washes in PB solution sections were incubated with anti-rabbit Alexa Fluor 594 and anti-chicken Alexa 488 secondary antibodies (1:300) diluted in PB solution for 2 h. Finally, sections were mounted with Vectashield mounting medium with DAPI (Vector Laboratories, Burlingame, CA, USA) and analyzed under the microscope. Photomicrographs were acquired with an Olympus BX51WI DSU confocal microscope (Olympus, Center Valley, PA, USA) coupled to a Hamamatsu EM-CCD C9100 camera (Hamamatsu, Hamamatsu, Japan).

#### **GSH AND GSSG LEVEL DETERMINATION**

Tissue GSH and GSSG levels were measured in whole brain, the cortex and hippocampus using a microplate-adapted fluorometric *o*-phthalaldehyde (OPA) method (Senft et al., 2000). Fluorescence was determined with 365 nm excitation and 430 nm emission filters in a DTX 800/880 Multimode Detector (Beckman Coulter, Fullerton, CA, USA).

#### **METHYLATED ARSENIC SPECIES DETERMINATION**

Concentrations of arsenic species were determined in urine, whole brain, the cortex and hippocampus by hydride-generation atomic absorption spectrometry using cryotrapping (HG-CT-AAS) as described previously (Hernández-Zavala et al., 2008). The quantification was performed using independent calibration curves of the arsenic species. Arsenic acid disodium salt (99% pure), and dimethyl arsinic acid (DMAV; 98% pure) were obtained from Sigma-Aldrich. Methylarsonic acid (MMAV) disodium salt (99% pure) was obtained from Ventron (Danvers, MA, USA). Sodium borohydride was obtained from EM Science (Gibbstown, NJ, USA). Prior to analysis, tissue samples were digested with 2 X ultrapure grade phosphoric acid (J. T. Baker) as described by Hughes et al. (2003).

#### **BEHAVIORAL TEST "PLACE RECOGNITION TASK"**

The object location task was conducted using the methodology described by Mumby et al. (2002), which was adapted to mice. Behavioral testing was performed in a 30 cm H × 60 cm W × 60 cm L acrylic open field box with white walls. Two objects of identical shape, size and color (white, black or red) were used in each trial. The objects and box were cleaned with a mixture of 10% ethanol, 10% dextran in destilled water, prior to each trial to eliminate any odor cues. Throughout the experimental period, the distal environmental cues were kept constant, and dark conditions were maintained in the experimental room. For three consecutive days, mice were habituated individually to the context and experimenter, by performing one session per day for 10 min. On day 4, the test was conducted in two phases: the exploration phase for recognition of object location, and the test phase, for discrimination of the location change. In the exploration phase, the two objects were placed in identical orientations with respect to two opposite corners of the box (10 cm from the corners) and the animals were allowed to explore the objects for 5 min. The animals were then returned to their homecage for 15 min and in the meantime the cage was cleaned and one of the objects was moved to a new location (opposite corner, at half the initial distance from the corner but maintaining an identical orientation with respect to the corner of the box). The animal was returned to the arena and let explore the objects for 5 min. The sessions were recorded with a video camera and analyzed by a trained observer. The behavior measurements included the frequency and cumulated time in seconds in which an animal approached (touch with body or vibrissae) and contacted each object with the paws during exploration. The discrimination index (DI) was calculated based on the formula DI = time spent exploring the re-placed object/time spent exploring both objects. Animals that explored the unmoved objects for less than 10 s during the test phase were not included in the study. DI was also determined in the exploration phase for each object, to verify object or place preference.

#### **STATISTICAL ANALYSIS**

The data are expressed as the means ± SE. The number of animals tested is indicated in each case. Student's *t*-test or one way analysis of variance (ANOVA) were used to assess statistical significance followed by Dunnett's multiple comparison test or Tukey's *post hoc* test. A *p* value <0.05 was considered statistically significant in all cases.

#### **RESULTS ARSENIC EXPOSURE**

Parental exposure to 20 mg/L of iAs during 30 days was well tolerated and did not alter body weight, water consumption or mating behavior (data not shown). No differences in body weight between control and exposed females were observed during gestation. A significant decrease in water consumption was observed in exposed female mice during lactation and until weaning (**Figure 1A**). The litter size between exposed and unexposed groups was also similar (data not shown). The exposed litter did not show signs of overt toxicity, i.e., ataxia, redness, swelling, fetal malformations or death at birth and throughout the experiment. (**Figure 1A**). The estimated average intake of iAs for males was 2.69 ± 0.69 mg/kg per day during the 30 days prior to mating, while females ingested 2.72 ± 0.88 mg/kg/day. During gestation, the amount of iAs ingested was 2.92 ± 1.17 mg/kg/day (20–21 days), and during lactation the amount increased to 10.73 ± 1.9 mg/kg/day. A significant decrease in water intake was observed in the exposed group.

#### **ARSENIC SPECIES IN THE URINE OF THE MOTHER AND IN THE WHOLE BRAIN AND REGIONS OF THE OFFSPRING WITH GESTATIONAL EXPOSURE ON PND 1 AND PND 15**

As species levels were determined in the urine of pregnant females between days 15–18 of gestation. The main As species in exposed female urine was DMA followed by MMA and iAs. Total As levels among exposed females was 1500 times higher than that in controls (**Figure 1B**). On PND 1, the whole brain of the exposed litter showed DMA (58.9%) and iAs (41%) as the main As species. The levels of As species in control litters were 6.7 times lower than those of the exposed offspring (**Figure 1C**). During lactation (PND 15), the levels of As species in the male and female cortex and hippocampal regions were not different between controls and exposed mice (**Figures 1D,E**). As species accumulation was observed in the livers of exposed mice (**Figures 1D,E**).

#### **EFFECTS OF GESTATIONAL EXPOSURE ON GSH LEVELS IN WHOLE BRAIN ON PND 1, AND IN THE CORTEX AND HIPPOCAMPUS ON PND 15 MICE**

We hypothesized that the gestational exposure to iAs would modulate GSH levels in brain cells. On PND 1, the gestationally exposed litter showed significantly increased levels of oxidized GSH (GSSG) in the whole brain (**Figure 2A**). At this stage, sex or brain regions could not be clearly differentiated. On PND 15, no changes in GSH or GSSG levels were observed between the cortex and hippocampal regions from exposed and control male mice (**Figure 2B**). However, female mouse hippocampal regions showed a significant increase in GSH levels contents (**Figure 2C**).

#### **CHANGES IN GSH LEVELS INDUCED BY iAS EXPOSURE ARE ASSOCIATED WITH CHANGES IN THE EXPRESSION OF xCT, EAAC1 AND LAT1 TRANSPORTERS AND MODULATION OF NMDAR SUBUNITS IN DIFFERENT BRAIN REGIONS**

Changes in GSH could be due to the modulation of cysteine and glutamate transporters. Western blot analysis was performed to

explore the expression of xCT, EAAC1 and LAT1 transporters in the whole brain on PND 1 and in the cortex and hippocampus on PND 15 and 90. The expression of xCT, EAAC1 and LAT1 transporters (**Figures 3A–C** respectively) was significantly increased in the brains of pups at PND 1. This up-regulation continued at PND 15 for xCT and EAAC1 (**Figures 4C,D**) but not for LAT1 (data not shown) in gestationally exposed male and female mice. The expression of xCT and EAAC1 was observed mainly in hippocampal neurons (**Figures 4A,B**). Increased extracellular levels of glutamate have been associated with modulation of NMDAR subunits. At the same time, iAs exposure down-regulated the NR2A NMDA receptor subunits in the male hippocampus in PND 15 pups (**Figure 4E**), while the NR2B subunit was down-regulated in both the male and female cortex and hippocampus (**Figure 4F**).

On PND 15, gestationally exposed offspring were divided, and one half received drinking water with 20 mg/L of sodium arsenite while the other half received water without arsenic until PND 90. The place recognition task was performed at this time-point. The animals were then sacrificed and the hippocampal regions were examined for transporter expression. Exposed male showed increased xCT expression, although continued exposure (iAs-PND 90) seemed to significantly diminish the levels of xCT as compared with the animals that were only exposed during gestation and until PND 15 (iAs-PND 15; **Figure 5A**). No change in EAAC1 expression was observed in any of the groups (**Figure 5B**). However, the expression of GLT1 was down-regulated in the animals exposed during gestation and until PND 15 (iAs-PND 15), and no changes were observed in GLAST (**Figures 5C,D**). With respect to the NMDA receptor subunits, the expression of both NR2A and NR2B was significantly down-regulated in those animals exposed only during the gestation period and until PND 15 (iAs-PND 15; **Figures 5C,D**). Female mouse hippocampal regions showed marginal up-regulation of xCT expression in both conditions (iAs-PND 15 and -PND 90; *p* < 0.08) and non-statistically significant changes in the expression of GLAST, GLT1 or NMDAR subunits (data not shown).

#### **ARSENIC GESTATIONAL EXPOSURE DISRUPTS THE PLACE RECOGNITION TASK PERFORMANCE "SPATIAL MEMORY"**

The place recognition task was used in 90 day old animals to determine if iAs exposure had an impact on the spatial memory,

**FIGURE 2 | GSH and GSSG levels in mouse brain on PNDs 1 and 15**. **(A)** On PND 1, As species were determined in the whole brain. **(B,C)** On PND 15 cortex and hippocampus were dissected from males and females and

processed for GSH and GSSG determination. Data are expressed in mean nmol/g of fresh tissue ± SE. Data were analyzed using Student's t-test. (\*) Significantly different from controls, P < 0.05, n = 9 per group.

a function in which the hippocampal formation is strongly involved. The cognitive discrimination ability when an object changes location was evaluated in the different experimental groups of male and female mice including the control group, the iAs-PND 15 and iAs-PND 90 groups. During the recognition phase, no preference was observed for the location of the object or the object itself, which was measured as the time devoted to exploring each object (**Figure 6A**). In the test phase, iAs-PND 15 and iAs-PND 90 males showed significantly decreased recognition of the object location (**Figure 6B**). In contrast, this effect was marginally significant in iAs-PND 15 females and not significant in those with longer exposure (iAs-PND 90; **Figure 6B**).

#### **DISCUSSION**

iAs is an ubiquitous metalloid, present in man-made products as well as in food and water. The presence of iAs in drinking water causes health detrimental effects worldwide. Chronic exposure usually occurs through generations, however, few studies have investigated *in utero* developmental effects of iAs exposure (Vahter, 2008) while the impact at a molecular levels remains less understood. In exposed populations, As species can be present in cord blood (Concha et al., 1998; Hall et al., 2007) indicating that arsenic is transferred to the fetus. Moreover, the presence of As in cord blood has been inversely associated with neurodevelopmental indicators (Parajuli et al., 2013) and correlate with cognitive deficits that include alterations in pattern memory, functional memory, full scale IQ and verbal IQ (Tyler and Allan, 2014).

NMDA glutamate receptors participate in learning and memory. Overstimulation of these receptors by excess of glutamate can cause cell death during glucose and oxygen deprivation (Jung et al., 2012) and memory deficits due to xenobiotic exposure (Olney et al., 2000; Li et al., 2012). Lcys2/ L-glu, L-cys and L-glu transporters participate in GSH

synthesis in the CNS and are modulated by oxidative stress. The hypothesis of this work was that exposure to iAs during gestation would increase oxidative stress up-regulating xCT. This condition might increase the efflux of L-glu that could lead to the activation of transporters to remove the extracellular glutamate. At the same time, oxidative stress and excess of glutamate might

**FIGURE 5 | Cystine, cysteine and glutamate transporters in male hippocampus on PND 90**. **(A)** xCT; **(B)** EAAC1; **(C)** GLT1; **(D)** GLAST; **(E)**NR2A; **(F)** NR2B. Densitometric evaluation of the blot images was performed using β-tubulin as loading control. Bars represent mean ± SE relative to control values. Data were analyzed using one-way

ANOVA and Tukey's post hoc test. (\*) Significantly different from controls, P < 0.05., n = 3–5. Control: Controls, iAs-PND 15: exposure to arsenic during gestation and up to day 15th during lactation, iAs-PND 90: exposure to iAs during gestation, lactation and until day 90th.

**FIGURE 6 | The effect of As exposure on the spatial working memory in the recognition place task in male and female mice on PND 90**. Discrimination index (DI) = (novel place exploration time/total exploration time). **(A)** male and female DI of location in the recognition phase and **(B)** during test phase in controls, animals exposed during

gestation and lactation (iAs-PND 15) and animals exposed during gestation, lactation and until PND 90 (iAs-PND 90). Each bar represents the mean ± SE (n = 8). Data were analyzed using an ANOVA. (\*) Significantly different from controls after Dunnett's post hoc test, P < 0.05. # P = 0.0625.

modulate NMDAR subunit expression (Scimemi et al., 2009) in the developing brain which might be reflected later as memory impairment.

Mice metabolize iAs and clear iAs metabolites from tissues more efficiently than humans (ATSDR, 2007). Results of the present study show that whole brain and liver tissues of mice exposed to 20 mg/L contained on average concentrations of total As species of 28 and 260 ngAs/g, respectively (**Figures 1D,E**). Similar concentrations have been found in liver samples from residents of an arsenic endemic area who drink water containing between 0.22–2 mg/L of iAs. In this case, human liver samples showed concentrations of 100–1,200 ngAs/g of liver tissue, respectively (Mazumder, 2005).

The levels of iAs in drinking water (20 mg/L) did not cause any visible signs of toxicity to mice before mating or during gestation in CD1 mouse. Control and exposed litters were similar in size (number of individuals), weight and sex composition. In contrast, in a similar study using the FVB/NJ mouse strain reduced fertility was observed (He et al., 2007) most likely due to different strain sensitivity. iAs exposed lactating females, significantly diminished water consumption (**Figure 1A**). To verify the iAs toxicokinetics, As species were determined in pregnant female urine and in the brain and liver tissues from the offspring. DMA was the main As species observed in exposed female urine during late gestation (**Figure 1B**) similar to results documented in human studies (Concha et al., 1998). In the newborn brains, iAs and DMA were the predominant species (**Figure 1C**). On PND 15, brain regions could be identified and the cortex and hippocampus were isolated. At this time, the levels of As species in exposed nursing males or females were not different from control males or females (**Figures 1D,E**). This was not the case for the liver where the main species was DMA. The As species accumulation after gestational exposure observed in this study was similar to that reported by Jin et al. (2010) for albino mice. It has been shown that As accumulation in milk is very low in humans (Concha et al., 1998). Thus, these results indicate that the presence of As in the newborn brains resulted from absorption through the placenta and was eliminated from the brain but not from the liver during lactation. This could reflect differences in the kinetics at the organ level that affect the disposition of As metabolites (Devesa et al., 2006).

The presence of iAs and DMA in the brain of newborn exposed litters suggests that the placenta does not limit the passage of As species and could be associated with the higher levels of oxidized GSH (GSSG; **Figure 2A**). At the same time, the transporter systems xc- (xCT), x-AG (EAAC1) and LAT1 were up-regulated (**Figures 3A–C** respectively) suggesting an increased in L-cys transport for GSH synthesis due to the elevation of GSSG. In lactating litters, (PND 15) the levels of GSH or GSSG in the cortex and hippocampus were not different from those determined in control animals (**Figures 2B,C**) most likely due to the lower levels of As species (**Figures 1D,E**). However, the up-regulation of xCT and EAAC1 transporters continued (**Figures 4A,B**) and was also observed in CA1 hippocampal cells by immunofluorescence (**Figures 4A,B**). Increased expression of both xCT and EAAC1 could protect against glutamate toxicity (Lewerenz et al., 2006) reducing NMDAR activation (Scimemi et al., 2009). At the same time, a significant down-regulation of the NR2B subunit in both sexes and NR2A subunit in the hippocampus of exposed PND 15 males but not females, was observed. This observation suggests that the alteration of Lcys and glutamate transport may modify NR2B expression in the hippocampus due to gestational exposure. Prenatal stress (Zhao et al., 2013), ethanol (Brady et al., 2013), high fat diets (Page et al., 2014) and nicotine (Wang et al., 2011) have been associated with down-regulation of NR2B or the NR2B/NR2A ratio. The reduced expression of NR2B in the NMDAR has been observed in connection with cognitive impairments and neuropathologies (Paoletti et al., 2013). On PND 90, exposed litters showed significantly impaired place recognition performance compared to controls. Males were more affected than females (**Figure 6B**), suggesting that hippocampal neurons were affected by As exposure. Animals were sacrificed to investigate transporters and NMDAR subunit expression on PND 90 hippocampus. xCT expression remained to be upregulated (**Figure 5A**) especially in those animals exposed to iAs during gestation and until PND 15. However, on PND 90, EAAC1 expression was not different between control and exposed litters. GLT1, another transporter that participates in the removal of glutamate, was also down regulated in animals exposed during gestation. These results suggest that glutamate levels might be increased in the hippocampus of these mice, leading to NR2A and NR2B down-regulation (**Figures 5C,D,F**) and memory deficits (**Figure 6B**). These observations are consistent with findings where extracellular glutamate increase initiates adaptive responses that involve a gradual down-regulation of the expression of NMDA receptors in response to environmental toxics or a glutamate transport blocker in neuronal models (Cebers et al., 2001; Win-Shwe et al., 2009). PND 90 males exposed during gestation and that continued to drink water with arsenic also showed memory deficits, increased expression of xCT and marginal down-regulation of NR2B but not NR2A, suggesting that synaptic efficiency could be affected. Further research is needed to clarify these observations. Taken together, our data may support the idea of GSH as an important neuronal reservoir to prevent excitotoxicity (Koga et al., 2011).

Our results indicate that gestational exposure to iAs impairs NMDAR subunits expression in the hippocampus affecting spatial memory. This impairment is associated with increased oxidative damage at birth and altered L-cys2/ glutamate and L-cys transport, which might in turn down-regulate NR2B expression (Paoletti et al., 2013). NMDAR subunit expression changes during brain development. NR2B is more abundant during the second week of postnatal development as neurons mature and become enriched at extrasynaptic sites (Roullet et al., 2010; Qiu et al., 2011). Gestational and postnatal exposure disrupts this pattern down-regulating NR2B especially on PND15 when this subunit is more predominant. Mouse hippocampus EAAC1, does not alter the activation of receptors at the synaptic cleft but reduces the recruitment of NR2B-containing NMDAR in perisynaptic/extrasynaptic sites (Scimemi et al., 2009). Then, the increased expression of EAAC1 observed in iAs exposed mice, could further alter the glutamate lifetime in the extracellular space impairing NMDAR activation and the induction of long term potentiation.

Cognitive impairment was observed in the rat after realgar (a mineral drug containing arsenic) exposure. Excess of extracellular glutamate was observed in hippocampus. Glutamate accumulation in the synaptic cleft was related to decreased expression of NR1 and up-regulation of NR2A subunit leading to calcium overload, down-regulation of GLT1 and ultrastructural changes in hippocampal neurons (Tao-guang et al., 2014). Interestingly, increased activity of xCT transport is accompanied with an increase in glutamate levels and neuronal death via overstimulation NMDAR (Jackman et al., 2012). Also, xCT−/ <sup>−</sup> mice show significantly lower extracellular hippocampal glutamate concentrations and optimal spatial working memory (De Bundel et al., 2011) suggesting that xCT constitutes a source for non-vesicular glutamate release. Additionally, downregulation of NR2B can occur through ubiquitination if NMDAR agonists are increased (Ehlers, 2003). Then, disrupted glutamate transport by increased xCT activity might be responsible for the down-regulation of NR2B. *In vitro* cultures of hippocampal neurons have shown that both NR2B and NR2A show endocytosis trafficking through endosomes (Scott et al., 2004). This event could be disrupted by prenatal As exposure. Also, adult rats exposed to different concentrations of sodium arsenite during 3 months after weaning, showed cognitive impairments and a dose-dependent down-regulation of NR2A in both mRNA and protein levels in hippocampus (Luo et al., 2009, 2012). Similarly, NR2A up-regulation was observed on PND 90 in male mouse hippocampus (**Figure 5E**). The NMDAR NR2A subunit in adult rat is sensitive to arsenic induced neurotoxicity (Luo et al., 2009, 2012).

Additionally, the persistent lower expression of NR2B subunit in males might be due to epigenetic changes. iAs methylation consumes GSH and SAM which affects DNA methylation (Reichard and Puga, 2010; Tyler and Allan, 2014). According to Reichard and Puga (2010) the epigenetic modifications observed during mouse gestational exposure to iAs suggest target genespecific methylation changes, some of which are hypomethylated while others suffer hypermethylation. Thus, hypermethylation of NR2B promoter and/or hypomethylation of NR2B repressors could lead to the down-regulation of NR2B. In this respect, rats exposed during gestation to 3 and 36 ppm of sodium arsenite in drinking water showed changes in the methylation status of genes involved in neuronal plasticity in cortex and hipocampus (Martínez et al., 2011). Histone modifications have been also observed in mice prenatally exposed to 100 µg/L of sodium arsenite which could be associated with altered learning in adults (Cronican et al., 2013). Also, there are evidences that the changes in the activity-dependent NR2B expression (Lee et al., 2008) or NR2B expression during chronic intermitent ethanol treatment (Qiang et al., 2010) are due to epigenetic modifications.

The altered activity/expression of the NR2B subunit due to post-transcriptional modifications (Qiu et al., 2011), ubiquitinization (Ehlers, 2003) or epigenetic modifications (Lee et al., 2008; Qiang et al., 2010; Tyler and Allan, 2014) have been implicated in the modulation of learning and memory processing, pain perception, feeding behavior as well as being involved in neurological disorders. Our results show that NMDAR subunit expression by prenatal exposure to iAs is affected, which may in turn alter memory early in life, which is in line with what has been reported in some human populations (Tyler and Allan, 2014). Learning and memory are complex processes involving several brain regions and neuronal networks. This work shows how As species might disrupt the expression of key components that could lead to behavioral alterations and the development of neuropathologies later in life. It remains important to identify the environmental agents that might impair neural development, maturation and physiology by interfering with the biochemistry of crucial neurotransmiters and aminoacids such as glutamate and cysteine.

#### **ACKNOWLEDGMENTS**

We thank M. Sc. Luz del Carmen Sanchez-Peña for arsenic species analysis, M. Sc. Tzipe Govezensky for supervision in the statistical analyses, Pavel Petrosyan, Ph.D, MVZ Diana Hernández-Loranca and MVZ Oscar Hernández Campos for their help handling mice and Miguel Tapia-Rodríguez, Ph. D. for his assistance during the confocal analysis of brain slices. Lucio A. Ramos-Chávez received a scholar fellowship from CONACYT No. 211718 and DGEP UNAM. This work was supported by grant from PAPIIT (UNAM) IN 207611 and CONACYT 102287 to María E. Gonsebatt. This study was performed in partial fulfillment of the requirements for the Ph.D degree in the posgrado en Ciencias Biológicas at the Universidad Nacional Autónoma de México.

# **REFERENCES**


promotor CpG-island methylation of genes involved in neuronal plasticity. *Neurochem. Int.* 58, 574–581. doi: 10.1016/j.neuint.2011.01.020


**Conflict of Interest Statement**: The Guest Associate Editor Victoria Campos declares that, despite being affiliated to the same institution as author Daniela Silva-Adaya, the review process was handled objectively and no conflict of interest exists. 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 October 2014; accepted: 13 January 2015; published online: 09 February 2015*.

*Citation: Ramos-Chávez LA, Rendón-López CRR, Zepeda A, Silva-Adaya D, Del Razo LM and Gonsebatt ME (2015) Neurological effects of inorganic arsenic exposure: altered cysteine/glutamate transport, NMDA expression and spatial memory impairment. Front. Cell. Neurosci. 9:21. doi: 10.3389/fncel.2015.00021*

*This article was submitted to the journal Frontiers in Cellular Neuroscience*.

*Copyright © 2015 Ramos-Chávez, Rendón-López, Zepeda, Silva-Adaya, Del Razo and Gonsebatt. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution and 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*.

# Systemic inflammation and the brain: novel roles of genetic, molecular, and environmental cues as drivers of neurodegeneration

#### **Roman Sankowski 1,2\* † , Simone Mader <sup>2</sup>\* † and Sergio Iván Valdés-Ferrer 1,2,3\* †**

<sup>1</sup> Elmezzi Graduate School of Molecular Medicine, Manhasset, NY, USA

<sup>2</sup> Feinstein Institute for Medical Research, Manhasset, NY, USA

<sup>3</sup> Department of Neurology, Instituto Nacional de Ciencias Médicas y Nutrición Salvador Zubirán, México City, Mexico

#### **Edited by:**

Victoria Campos, Instituto Nacional de Neurologia y Neurocirugia, Mexico

#### **Reviewed by:**

Rafael Linden, Federal University of Rio de Janeiro, Brazil Muzamil Ahmad, Indian Institute of Integrative Medicine, India Claudia Verderio, Istituto di Neuroscienze, Italy

#### **\*Correspondence:**

Roman Sankowski, Simone Mader, and Sergio Iván Valdés-Ferrer, Feinstein Institute for Medical Research, 350 Community Dr., Manhasset, NY 11030, USA e-mail: rsankowski@nshs.edu; smader@nshs.edu; svaldesfer@ nshs.edu, sivaldes@gmail.com

†Roman Sankowski, Simone Mader and Sergio Iván Valdés-Ferrer have contributed equally to this work.

The nervous and immune systems have evolved in parallel from the early bilaterians, in which innate immunity and a central nervous system (CNS) coexisted for the first time, to jawed vertebrates and the appearance of adaptive immunity. The CNS feeds from, and integrates efferent signals in response to, somatic and autonomic sensory information. The CNS receives input also from the periphery about inflammation and infection. Cytokines, chemokines, and damage-associated soluble mediators of systemic inflammation can also gain access to the CNS via blood flow. In response to systemic inflammation, those soluble mediators can access directly through the circumventricular organs, as well as open the blood–brain barrier. The resulting translocation of inflammatory mediators can interfere with neuronal and glial well-being, leading to a break of balance in brain homeostasis. This in turn results in cognitive and behavioral manifestations commonly present during acute infections – including anorexia, malaise, depression, and decreased physical activity – collectively known as the sickness behavior (SB). While SB manifestations are transient and self-limited, under states of persistent systemic inflammatory response the cognitive and behavioral changes can become permanent. For example, cognitive decline is almost universal in sepsis survivors, and a common finding in patients with systemic lupus erythematosus. Here, we review recent genetic evidence suggesting an association between neurodegenerative disorders and persistent immune activation; clinical and experimental evidence indicating previously unidentified immune-mediated pathways of neurodegeneration; and novel immunomodulatory targets and their potential relevance for neurodegenerative disorders.

**Keywords: neurodegeneration, systemic inflammation and sepsis, autoimmune disorders, anti-brain antibodies, TNF, HMGB1, connectome**

#### **INTRODUCTION: AN EVOLUTIONARY PERSPECTIVE**

All multicellular organisms have germ-line encoded surveillance systems destined to detect potentially dangerous, perhaps lethal, invaders (Medzhitov and Janeway, 1997). This extremely effective response, the innate immunity, depends on detection of specific molecular patterns present in invaders, but not expressed by self-tissues. Innate immunity aroused approximately 600 million years ago, and is evolutionarily conserved across species and kingdoms. The main components of innate immunity – from tolllike receptors (TLR) to inflammasomes – are highly conserved from plants to mammals (Jones and Dangl, 2006). Adaptive immunity appeared much later in evolution, perhaps 500 million years ago, and is only present in vertebrates (Cooper and Alder, 2006; Moresco et al., 2011). Unlike innate immunity, adaptive immunity depends on one individual's experience, in which adaptive responses form with high efficiency only against antigens that have been recognized and presented by the innate immunity. Although the origin of the nervous system remains controversial, the first centralized nervous system probably appeared with

the first bilateral organisms (organisms with two relatively identical halves, called *bilaterians*) during the Ediacaran period around 550 million years ago (Knoll et al., 2004). Ever since, both nervous and immune systems have co-evolved and are in constant communication.

The nervous system of the evolutionarily ancient nematode *Caenorhabditis elegans* has the ability to regulate innate immune responses (Andersson and Tracey, 2012), and aid in decisionmaking regarding finding bacteria that can be used as food and avoiding pathogenic bacteria (Reddy et al., 2009). In mammals, the nervous system also has the ability to sense inflammatory stimuli directly, thus allowing to recognize a potential source of damage through generation of pain, and to modulate the response to infection (Mina-Osorio et al., 2012; Chiu et al., 2013a). Although the afferent pathways and integration of immune information in the brain are areas of active research, there is evidence that central muscarinic signaling modulates inflammation in experimental sepsis (Pavlov et al., 2009; Rosas-Ballina et al., 2015), obesity (Satapathy et al., 2011), and inflammatory colitis (Ji et al., 2014). The efferent axis of neuroimmune control is better understood after the cholinergic anti-inflammatory pathway (CAP) (Borovikova et al., 2000), a cholinergic reflex system that regulates inflammation via the vagus nerve that stimulates the splenic nerve to release noradrenaline. Noradrenaline in turn stimulates a subset of acetylcholine (ACh)-producing splenic T-cells (CD4+CD44hiCD62Llo) to release ACh, which binds to α7 nichotinic receptors on the surface of macrophages, resulting in down-regulation of TNF by blocking the nuclear translocation of nuclear factor kappa B (NFκB) (Rosas-Ballina et al., 2011). Thus far, this is a unique scenario in which an immune cell acts as interneuron in a reflex system. Electrical as well as chemical stimulation of the CAP have been shown to decrease the inflammatory burden and increase survival of experimental sepsis (Borovikova et al., 2000; Bernik et al., 2002).

Neuroimmune modulation is a flexible phenomenon that relies on environmental cues from a continuously changing milieu. People undergoing persistent stress have been found to display abnormal immune responses. For instance, caregivers of Alzheimer disease (AD) patients have higher levels of anxiety and depression than age-matched controls; otherwise, these healthy caregivers also have reduced total T-cells and T helper cells in peripheral circulation, as well as higher titers of anti-Epstein–Barr virus antibodies (Kiecolt-Glaser et al., 1987). Moreover, the observed immune and behavioral changes are increased in proportion to disease progression. T-cell response to mitogenic stimuli progressively decreases, and while all subjects have the same number of infectious episodes, the number of days unable to perform activities of daily living, as well as the number of doctor visits are significantly increased in AD caregivers (Kiecolt-Glaser et al., 1991). One mechanism for the immune dysfunction in response to chronic stress is a reduction of telomeres and telomerase activity, potentially leading to early immune senescence (Epel et al., 2004). Mice with naturally elevated anxiety levels have increased activated microglia and perivascular macrophages in the brain, than less anxious strains (Li et al., 2014), suggesting that anxiety can increase the inflammatory background in the brain.

The acute effects of systemic inflammation upon cognition and behavior are not limited to the elderly or the critically ill. As we have witnessed in ourselves and those near us, even a minor and self-limited common cold induces a transient syndrome known as *sickness behavior* (SB) marked by fatigue, depression, lack of drive, malaise, sleep disturbances, decreased physical activity, and social interactions, as well as cognitive impairment (Capuron et al., 1999, 2001). Healthy volunteers develop anxiety, depression, and memory impairment in response to a low dose of lipopolysaccharide (LPS), and the development of such clinical scenario correlates with TNF secretion (Reichenberg et al., 2001). Some chronic infections may go unrecognized for long periods, as is the case in tuberculosis, human immunodeficiency virus (HIV), hepatitis B virus (HVB), or hepatitis C virus (HCV). Unlike septic patients, patients with chronic infections have an organized and targeted immune response (versus a massive and diffuse one in sepsis), even if the response is ineffective to clear the infection. Those patients, however, have increased cognitive and behavioral problems. For instance, patients with HCV infection have increased rates of fatigue and depression, as well as evidence of metabolic brain dysfunction in the absence of acute hepatitis (Forton et al., 2001, 2002; Hilsabeck et al., 2002). Patients with chronic HCV that have been treated with pegylated interferon (IFN)-α develop significantly higher incidence of depression when compared to the patients' state before IFN-α administration (Reichenberg et al., 2005). This supports the role of large loads of inflammatory cytokines in inducing and sustaining brain dysfunction. Experimentally, NADPH oxidative activity and nitric oxide synthase (iNOS) are induced in the brain shortly after systemic inflammation (Wong et al., 1996; Yokoo et al., 2012), potentially leading to NMDA-dependent neurotoxicity (Dawson et al., 1993). Evidence derived from post-mortem studies indicates that iNOS is also induced in the brain of patients dying of severe sepsis, but the role of iNOS as inductor of neurotoxicity in response to systemic inflammation has not been assessed in patients surviving severe sepsis (Sharshar et al., 2003). Experimentally, preemptive administration of the free radical scavenger endarvone before sepsis induction resulted in reduced neuronal damage and blood–brain barrier (BBB) permeability (Yokoo et al., 2012). Administration of the antioxidants *N*-acetylcysteine and deferoxamine shortly after murine sepsis induction has shown long-term neuroprotective effects (Barichello et al., 2007).

## **IMMUNITY IN SEPSIS**

The immune system is in a constant state of surveillance against potential pathogens and self-generated molecules indicative of damage. Under normal conditions, inflammation is a wellorchestrated response with constant fine-tuning. Once microorganisms have breached the skin and mucosal barriers, innate immunity is critical in preventing further invasion by launching inflammation. After the infection source has been cleared, the inflammatory response also plays an important role in tissue repair and functional healing. When the source of damage has been controlled, the same mechanisms that initiated and regulated inflammation will dampen the response. Large loads of pathogens, or infection by highly virulent pathogens, can trigger an *en-masse* systemic response that leads to sepsis and multiple organ failure (Deutschman and Tracey, 2014). Moreover, during sepsis the inflammatory response does not resolve for several days or weeks (Valdés-Ferrer, 2014), leading to persistent release of high-mobility group box-1 (HMGB1) (Valdés-Ferrer et al., 2013b). HMGB1 is a highly conserved nuclear protein that can be passively released by stressed cells, or actively secreted by monocytes and other immune cells in response to inflammatory signals, playing a critical role in the innate immune response to inflammatory and sterile injury alike (Andersson and Tracey, 2011). HMGB1 in turn primes resident monocytes toward an inflammatory, cytokine producing phenotype (Valdés-Ferrer et al., 2013a). Severe trauma, as well as surgery can lead to large loads of endogenous pro-inflammatory molecules (damage-associated molecular patterns (DAMPs) being released. A few DAMPs have been shown to induce brain dysfunction *in vivo*. Of those, TNF and IL-1 can mediate long-standing cognitive and behavioral changes and, in experimental settings, interfering with the effect of TNF reduces the effect of trauma in the formation of contextual memory (Terrando et al., 2010b). A number of genes regulating immune responses are also closely related to neurodegenerative diseases (**Table 1**).




AD, Alzheimer's disease; AGS, Aicardi–Goutières syndrome; CR1, complement receptor 1; fALS, familial amyotropic lateral sclerosis; FTD, frontotemporal dementia; GRN, granulin gene; PD, Parkinson disease; NPSLE, neuropsychiatric SLE; HDLS, hereditary diffuse leukoencephalopathy with spheroids; SOD, super oxide dismutase; TREM, triggering receptor expressed on myeloid cells; TYROBP, TYRO protein tyrosine kinase-binding protein.

#### **CEREBRAL CONSEQUENCES OF SYSTEMIC INFLAMMATION: WHAT WE HAVE LEARNED FROM SEPSIS AND OTHER INFLAMMATORY CONDITIONS**

The nervous system is particularly vulnerable to damage in response to systemic inflammation. Inflammation-induced infiltration of immune cells and mediators into the brain leads to profound structural and functional changes (**Figure 1**). As a consequence, up to 81% of septic patients develop sepsis-associated delirium (SAD) (Ely et al., 2001, 2004), with elderly patients being at particularly high risk (McNicoll et al., 2003; Iwashyna et al., 2012). In the elderly, severe sepsis is sufficient to trigger new cognitive decline of sufficient importance as to profoundly interfere with quality of life (Iwashyna et al.,2010). SAD is not only common in septic patients but also is a reliable indicator of bad prognosis (Eidelman et al., 1996). Six-month mortality among septic patients is twice as high in patients who develop SAD at any point in time during hospitalization (Ely et al., 2004). Altered mental status is

more common in septic patients with bacteremia than in those with negative blood cultures (Eidelman et al., 1996). Magnetic resonance imaging (MRI) in patients with septic shock has shown that new white matter lesions are common (Sharshar et al., 2007). Neonatal sepsis is also marked by abnormalities of the white matter (66% of infants in one cohort), and white matter lesions correlate to poorer mental and psychomotor development at 2 years (Shah et al., 2008). The burden of white matter damage worsens with duration of shock, indicating that sepsis interferes with brain connectivity (**Figure 2**). Gray matter damage in response to sepsis is far less clear. Imaging and post-mortem studies from patients, as well as data derived from experimental models, indicate that gray matter damage is the norm. In comparison to subjects dying of other causes, brain histology from severe sepsis patients shows a higher number of CD68<sup>+</sup> and major histocompatibility complex (MHC)-class II microglia in the cerebral cortex and the white matter. Moreover, in septic brains ameboid-shaped activated microglia

can be found in gray and white matter alike (Lemstra et al., 2007). An MRI study of premature infants showed reduced volume of deep gray matter structures, and although the findings were consistent in the small subset of septic infants, those were not different from premature non-septic infants (Boardman et al., 2006).

In experimental sepsis, persistent cognitive impairment has been observed in rats completely recovered and with negative blood cultures. This indicates that clearing the trigger of sepsis does not prevent the appearance of persistent brain damage (Barichello et al., 2005). Brain mitochondria become dysfunctional during experimental sepsis, showing increased proton permeability, inadequate membrane potential recovery, and reduced oxidative phosphorylation (d'Avila et al., 2008). In mice, the BBB becomes leaky within 24 h following sepsis induction (Yokoo et al., 2012). While the number of hippocampal neurons is not reduced in mice surviving abdominal sepsis, there is neuronal degeneration (Yokoo et al., 2012), and the spine density of CA1 neurons is significantly diminished, and this finding correlates with circulating HMGB1 (Chavan et al., 2012). Anti-HMGB1 neutralizing monoclonal antibodies improve the memory deficit observed in experimental sepsis, although the mechanism of neuroprotection behind this protective effect is still under investigation. In response to systemic LPS, cortical mRNA expression of IL-1β, IL-6, TLR2, TLR4, Scavenger A, and glial fibrillary acidic protein (GFAP) are upregulated within 4 h, indicating that cortical inflammation and glial activation occur in parallel or shortly after systemic inflammation ensues (Noh et al., 2014; Silverman et al., 2014). LPS also increased brain mitochondrial complex II/III activity, and reduced

brain glutathion levels within 1 h after systemic administration, supporting the role of systemic inflammation in cerebral mitochondrial dysfunction in sepsis (Noh et al., 2014). Interestingly, even a single and relatively low dose of LPS (that is sufficient to cause around 10% mortality) has been shown to have long-term behavioral and cognitive consequences. In one study, 30-day evaluation showed increased anhedonia and anxiety, altered working memory, as well as reduced exploratory behavior (Anderson et al., 2015). Consistent with that, in a model of endotoxemia in aged rats, a single systemic injection of LPS induced brain inflammation that lasted for at least 30 days. Interestingly, those rats had only transient increase of circulating TNF and IL-1 lasting <6 h in response to the injection (Fu et al., 2014). The hippocampus and dentate gyrus were particularly affected, showing astrogliosis and increased TNF, IL-1, and NF-κB mRNA and protein levels. This suggests that even transient bouts of systemic inflammation of only limited significance can cause sustained brain damage.

Delirium and brain dysfunction can be induced directly by soluble mediators such as TNF, IL-1, or HMGB1, as well as indirectly through activation of microglia and astrocytes. LPS has been shown to induce microglial activation and memory impairment in young mice through a mechanism that is at least partially dependent on interleukin (IL)-1, and HMGB1, both occurring within 24 h of LPS administration (Terrando et al., 2010a). However, while the rapid effect of LPS upon brain homeostasis is clear, a single sublethal injection of LPS (5 mg/kg) is sufficient to impair behavior and memory, mediated by a reduction in neural stem cell proliferation in the dentate gyrus, as well as inducing microglia invasion and activation to the hippocampus, all lasting for at least 30 days (Anderson et al., 2015). Minocycline, a tetracycline derivative used commonly for acne and other infections, inhibits activation and proliferation of microglia (Tikka et al., 2001). Intracerebroventricular (ICV) administration of minocycline immediately after sepsis induction by CLP reduced cerebral inflammation, decreased BBB permeability, and protected against memory impairment observed in untreated septic mice (Michels et al., 2015).

#### **SUSCEPTIBLE BRAIN: PREDISPOSING FACTORS DRIVE NEURODEGENERATION**

While the nervous system is susceptible to peripheral challenges, one question under active research concerns to the role of a

*susceptible* nervous system. Genes and molecular factors rendering rodents and humans prone to neurodegeneration have been identified recently. Genes predisposing to neurodegeneration are summarized in **Table 1**. The mouse strain DBA/2J has a natural anxious behavior. In response to a systemic challenge with lowdose LPS (1 mg/kg), DBA/2J mice develop increased anxiety, and increased expression of hypothalamic mRNA expression of the inflammatory genes*Il1b*,*Il6*,*tnf*,and*Nos2* in comparison to behaviorally normal strains (Li et al., 2014), suggesting that emotional stress has a role in magnifying a systemic inflammatory stimulus.

Sepsis-induced cognitive decline can be exacerbated in individuals with a susceptible brain (Iwashyna et al., 2010). Interestingly, in community-based patients with AD, even mild inflammatory conditions can lead to cognitive decline and disease progression (Holmes et al., 2009). Similarly, in a mouse model of AD, within 24 h after the systemic administration of LPS, cerebral IL-1β and IL-6 were induced, followed by changes in β-amyloid precursor peptide (APP) isoforms similar to those observed in AD patients (Brugg et al., 1995). In the past few years, possible mechanisms of damage have been proposed based on rodent models. In contrast to healthy control mice, scrapie-infected mice – a model of prion disease – show markedly enhanced susceptibility to LPS, leading to altered working memory, synaptic loss, enhanced expression of inflammatory receptors, and microglial activation (Murray et al., 2012). Cognitive and motor coordination are more severely impaired in mice that have been scrapie-infected for longer time periods, and one-time inflammation seems to accelerate disease features of neurodegeneration (Cunningham et al., 2009). Moreover, in prion disease neurodegeneration occurring in response to systemic inflammation is proportional to the burden of preexisting neurodegeneration. In a rat model of toxin-induced Parkinson disease (PD) induced by 6-hydroxydopamine, persistent systemic inflammation induced by IL-1β resulted in a reduction in the number of neurons, as well as an increase in activated microglia in the substantia nigra (SN) (Pott Godoy et al., 2008). The damage to SN in response to inflammation is not limited to susceptible animals. In response to systemic LPS, 8-month-old wild type mice showed a rapid reduction of tyrosine hydroxylase (dopaminergic) neurons, with a reciprocal expansion in activated microglia in the SN occurring shortly thereafter (Reinert et al., 2014). Amyotrophic lateral sclerosis (ALS) is a neurodegenerative disorder characterized by progressive, irreversible, and highly lethal motor neuron damage. In transgenic mice expressing a mutant form of superoxide dismutase 1 (SOD1), an experimental model of ALS, inflammation in the anterior horns of the spinal cord is a hallmark. Inflammation is driven by tissue-resident microglia displaying a neurodegeneration-specific pattern of gradual increase in insulinlike growth factor 1 (*Igf1*) and the tyrosine kinase receptor AXL (*Axl)* transcription that closely correlate with disease progression (Chiu et al., 2013b). The activated microglia induce motor neuron death through induction of NF-κB (Frakes et al., 2014). The peripheral nervous system of SOD1 mutant mice also displays inflammation that is proportional to disease progression; in this case, there is invasion of the anterior (motor) roots by circulation-derived active macrophages (Chiu et al., 2009). While spinal cord inflammation may be common in SOD1 mutant mice, those mice display an increased susceptibility to chronic systemic

inflammation, leading to disease progression, induction of TLR2, early axonal damage, and accelerated mortality (Nguyen et al., 2004). Although the mechanism for increased susceptibility to systemic inflammation in SOD1 mutant mice is unknown, it is possible that persistent systemic inflammation can further activate microglia and induce the translocation of NF-κB and activation of downstream inflammatory mediators.

Altogether, current evidence indicates that cognitive impairment is common in sepsis; that a susceptible brain is fertile ground for systemic inflammatory-induced dysfunction; and that cognition and behavior are persistently challenged, even after resolution of acute sepsis.

#### **SYSTEMIC LUPUS ERYTHEMATOSUS: SYSTEMIC AUTOIMMUNE INFLAMMATION AS DRIVER OF NEURODEGENERATION**

Systemic lupus erythematosus (SLE) is a chronic relapsing– remitting autoimmune disease that affects multiple organs and often involves the central nervous system (CNS). It has a high female preponderance and occurs mainly in women of childbearing age. Immunologically, it is characterized by a loss of tolerance to self-antigens and abnormal B- and T-cell responses. Immunoglobulin complexes can deposit in tissue and can cause systemic inflammation. Anti-nuclear antibodies can be found in up to 98% of patients, yet can also be detected in other autoimmune conditions. Neuropsychiatric SLE (NPSLE) is an incompletely understood medical problem, and its clinical picture diverse. Depending on the study, the percentage of patients with neurological and psychiatric involvement can vary from 12 to 95% (Ainiala et al., 2001). CNS involvement indicates a more severe clinical presentation of SLE (Mak et al., 2012). Due to the wide range of motor, sensory, cognitive, and behavioral symptoms, diagnosis of neuropsychiatric symptoms is challenging (Jeltsch-David and Muller, 2014). Memory impairment is linked to neuronal cell death in the hippocampus, and might be caused by antibody mediated cytotoxicity following binding to neuronal cell surface receptors, such as the NMDA receptor (Faust et al., 2010) or neuronal surface P antigen (NSPA) (Bravo-Zehnder et al., 2015). Cognitive impairment in NPSLE has been associated with a higher frequency of hippocampal atrophy (Appenzeller et al., 2006). There is a great set of data demonstrating that neuropsychiatric symptoms could be caused by autoantibodies, cytokines or microvasculopathy, and thrombosis. The most common brain abnormalities in patients include microvasculopathy, which can be caused by antiphospholipid antibodies that bind to clotting factors and endothelial cells (Belmont et al., 1996). In addition to vasculopathy, post-mortem brain studies of patients show infarcts and hemorrhages, cortical atrophy, ischemic demyelination, as well as CNS demyelination (Hanly et al., 1992). Currently, it is not known whether CNS involvement develops independently or occurs as a consequence of systemic organ dysfunction or both. Neurological involvement might be related to treatment, infection, or metabolic disorders or may be part of a coexisting disease. Increased BBB permeability leads to CNS penetration of immunoglobulin, pro-inflammatory cytokines, and albumin. A limited number of studies addressed the genetics of NPSLE, and found an increased susceptibility in patients with brain involvement compared to SLE

sparing CNS involvement. Particularly TREX 1 (DNase III) gene mutations have been reported in NPSLE (**Table 1**; Lee-Kirsch et al., 2007b; de Vries et al., 2010), a mutation that is also found in other brain diseases and is involved in apoptosis and oxidative stress. TREX 1 knockout mice develop severe inflammatory myocarditis,resulting in reduced survival rates (Morita et al.,2004) due to accumulation of single stranded DNA fragments, which facilitates the production of type 1 IFN (Miner and Diamond, 2014). Larger genome wide association studies (GWAS) comparing patients with and without neuropsychiatric involvement are needed to further understand NPSLE. Despite improved imaging and the availability of potential biomarkers (autoantibodies, cytokines, chemokines), NPSLE remains a diagnostic dilemma. So far, no specific treatment is available for NPSLE; however, the CD20 B-cell targeted treatment rituximab was shown to be promising in a small cohort of refractory NPSLE patients (Tokunaga et al., 2007). In addition, peptide mimotopes in patients with anti-brain reactive antibody responses may hold promise (Bloom et al., 2011).

#### **MULTIPLE SCLEROSIS: BRAIN INFLAMMATION AS DRIVER OF NEURODEGENERATION**

Multiple sclerosis (MS) is the most common inflammatory demyelinating diseases in young adults with a high risk of long-term disability affecting over 2.5 million people worldwide (Compston and Coles, 2008). Despite advances in diagnosis and treatment, the cause of MS remains unknown. While effective treatment for the relapsing–remitting form has improved significantly, treatment for the progressive disease course is still of limited utility (Lassmann et al., 2007). MS is presumably caused by an exogenous trigger in genetically predisposed individuals (Sospedra and Martin,2005b),resulting in inflammation,demyelination,and neurodegeneration. MS is initiated by autoreactive T-cells against a yet unknown CNS antigen (Sospedra and Martin, 2005a). In addition to T-cells, in the last years, an important role of B-cells and antibodies has re-emerged as major contributors to the disease (Hauser et al., 2008). Experimental autoimmune encephalitis (EAE), the most widely used animal model for studying MS (Friese et al., 2006) is either induced by injection of myelin derived proteins together with adjuvant or by adoptive transfer of CD4<sup>+</sup> encephalitogenic T-cells. It has been increasingly recognized that this model does not accurately represent the full spectrum of the human disease. MS is a heterogeneous disease, with patients experiencing a broad range of motor, cognitive, and neuropsychiatric impairment (Chiaravalloti and DeLuca, 2008). Approximately 80% of patients develop relapsing–remitting MS (RRMS), where patients experience relapses that are followed by partial or complete remission (Sospedra and Martin, 2005a). In an advanced disease stage, the majority of RRMS patients convert to secondary progressive MS (SPMS), with a steady disease progression in the absence of relapses and remissions (Sospedra and Martin, 2005b). A minority of patients (15%) suffer from an early onset progressive neurological decline, defined as primary progressive MS (PPMS) (Sospedra and Martin, 2005a,b; Miller and Leary, 2007). PPMS presentation is similar to that of SPMS patients (Confavreux and Vukusic, 2006). Since the progressive disease is not associated with the number of relapses or the extend of inflammation (Friese et al., 2014), it remains unknown what drives neurodegeneration in MS.

So far, the interplay, timing, and localization of inflammation, breakdown of the BBB, demyelination, axonal dysfunction, neurodegeneration, gliosis, atrophy, and repair mechanisms are incompletely understood (**Figure 1**) (Noseworthy et al., 2000; Bruck, 2005). The pathological hallmark of MS is CNS *plaques*, which are areas of focal demyelination in the white matter (Popescu and Lucchinetti, 2012). Whereas glial scarring was described as a characteristic feature of demyelinating plaques, most studies used to emphasize an initial preservation of axons. During RRMS, focal lesions, which are disseminated in space and time (McFarland and Martin, 2007), are associated with a BBB breakdown and an infiltration and activation of immune cells (Sospedra and Martin, 2005b). Lesions can be completely or partly resolved due to remyelination and resolution of inflammation. In contrast, the progressive disease stage results in irreversible deficits and is pathologically characterized by chronic axonal degeneration and gliosis. Although demyelination is described as primary event in MS, and neurodegeneration is believed to occur as a secondary event, it is now widely accepted that axonal loss occurs already in acute white matter plaques (Bruck and Stadelmann, 2003). MS was believed to be primarily a white matter disease, yet cortical gray matter lesions and atrophy in the brain and spinal cord can be observed at different time points of the disease, even as early as at the time of diagnosis (Peterson et al., 2001; Bø et al., 2003; Kutzelnigg et al., 2005, 2007; Geurts et al., 2007; Barkhof et al., 2009). In addition, the marker for neuronal integrity *N*-actetylaspartate (NAA) is decreased throughout the CNS at an early disease stage (De Stefano et al., 2001). NAA levels are also decreased in the normal appearing white matter and in the cortical gray matter. Increased concentrations of the neurofilament protein in the cerebrospinal fluid (CSF) can be observed at all stages of the disease, which additionally reflects early ongoing neurodegeneration (Kuhle et al., 2011).

While there has been major progress in treating RRMS with anti-inflammatory and immunomodulatory treatment, there are currently no effective treatment options for the progressive disease. The inefficiency of anti-inflammatory treatment in PPMS and SPMS was explained by the lack of inflammation during neurodegeneration. However, this could be due to impaired drug delivery to the brain due to a widely restored BBB integrity in the progressive disease stage (Lassmann et al., 2012). In contrast to the traditional view that neurodegeneration in MS occurs in the presence of limited or almost absent inflammation, it has been shown that inflammation is present at all stages of the disease (Frischer et al., 2009). In fact, there is a massive infiltration of microglia and immune cells into gray matter lesions, including the deep gray matter (Lucchinetti et al., 2011). However, chronic lesions in the gray matter show less extend of immune cell infiltration compared to chronic white matter lesions (Popescu and Lucchinetti, 2012). Cortical lesions can be observed even before detection of white matter lesions. Most patients have the highest prevalence of cortical lesions in the progressive disease stage, where diffuse meningeal inflammation correlates with the disease severity, suggesting that cytokines, chemokines, reactive oxygen species (ROS), and glutamate released by meningeal infiltrating immune cells contribute to neurodegeneration by disturbing the neuronal metabolic pathway (Friese et al., 2014). A redistribution of ion channels could be a compensatory effect of the inflamed axon but finally accelerates neurodegeneration.

Overall, the cause and mechanism of neurodegeneration in MS is unresolved, yet it is highly likely that there is a large variability within individual subgroups of MS patients. In some patients, inflammation can be caused by primary neurodegeneration, whereas demyelination can be the primary cause in others, causing axonal transection and thereby initiating Wallerian and retrograde neurodegeneration (Friese et al., 2014). It has been shown that demyelination can lead to axonal degeneration and neuronal loss by limited trophic support of oligodendrocytes (Fünfschilling et al., 2012; Lee et al., 2012). Neurodegeneration in MS can also occur, at least partly, in the absence of demyelination since axonal and neuronal injury in the normal appearing white matter is rather associated with global CNS inflammation but not with white matter demyelination (Friese et al., 2014). The continuous worsening of the disease can be explained by progressive neuronal loss, which cannot be compensated over time by protective repair mechanisms. However, the total lesion load does not correlate with the extent of neurodegeneration, suggesting that neurodegeneration is the result of inflammatory processes in nonlesioned white matter. Alternatively, MS can be a primary neurodegenerative disease rather than an autoimmune disease (Trapp and Nave, 2008). However, genetic findings show no correlation of genetic risk alleles in MS with other neurodegenerative diseases. A role of genetic heritability in MS was supported by early findings, which observed aggregations of MS cases in some families as well as an increased prevalence of MS in monozygotic twins compared to dizygotic twins (Ebers et al., 1995). Genetic susceptibility genes have been described but the disease does not follow a Mendelian inheritance pattern but a rather complex pattern of interaction. The most striking findings regarding genetic susceptibility in MS comes from studies obtained in the 1970s showing an association of MS with alleles of the MHC, particularly HLA-DRB1\*15, of the class II gene HLA-DRB1, which is the most important risk allele in MS. GWAS and meta analysis enabled the discovery of many single nucleotide polymorphisms (SNIP) in MS. To date, around 350 MHC and non-MHC loci have been identified (Wang et al., 2011), which are mainly involved in immunological processes (Wang et al., 2011). Some of those risk genes overlap with other autoimmune diseases, whereas other risk genes are unique to MS. Based on the current studies, the genetic risk for MS is not linked to a single gene mutation but rather caused by a complex interplay of many SNIP, which amplify and result in small or moderate risk effects. MS is probably caused by a multifactorial pattern of inheritance, which needs further investigation in individual larger patients groups. Once GWAS studies identify SNIPs in MS patients, there is a great interest to correlate these findings with functional data. Prime examples are the findings of SNIPS in the *TNFRSF1A* gene that have been associated with a worse clinical outcome in MS patients. This mutation results in a novel soluble splice form of the TNF receptor (TNFR1) that, in contrast to the membrane bound form, lacks NF-κB activity and apoptotic activity but can block the function of TNF and thus mimics anti TNF therapies, which exacerbate MS (Gregory et al., 2012).

So far, most genetic studies focus on genes responsible for the initiation of the disease rather than on genes influencing the severity of the disease (Friese et al., 2014). Thus, future studies are needed to identify genes that trigger the progression of the disease in the presence of inflammation. One study showed that meningeal inflammation is associated with small fiber axonal loss in the spinal cord of patients that were HLA-DRB1\*15 positive, correlating neurodegeneration with increased genetic susceptibility (DeLuca et al., 2013). Another study demonstrated higher glutamate levels in the brain of MS patients harboring certain risk alleles for genes involved in the glutamate pathway (Baranzini et al., 2010). A polymorphism of the inositol polyphosphate-4 phosphatase, type II (Inpp4b), which was described for MS results in decreased nerve conduction velocity, which could aggravate the disease (Lemcke et al., 2014). More GWAS studies are needed to confirm these results and there should be a particular focus on polymorphisms associated with the progressive disease stage. In addition to genetic susceptibility genes, several environmental factors have been suggested to trigger the disease, such as vitamin D, smoking, Epstein–Barr virus infection, and geographical location in relation to latitude gradient (Ascherio and Marrie, 2012).

#### **MECHANISMS OF NEURODEGENERATION: SYSTEMIC INFLAMMATION DRIVES DISRUPTION OF BRAIN NETWORKS**

Biological systems, such as the neuronal network of the human brain (Sporns et al.,2005;Achard et al.,2006;Bullmore and Sporns, 2009) have "small-world" properties (Watts and Strogatz, 1998). Small-world networks have two levels of organization (**Figure 2**). On the local level, groups of neurons specialized in a specific task form functional modules with high short intramodular connectivity. On the global level, different modules are connected through long intermodular connections. The advantage of the latter type of connections is enhanced computational efficiency through parallel processing of information (Watts and Strogatz, 1998; Latora and Marchiori, 2001; Bullmore and Sporns, 2009). Anatomically, long intermodular connections are formed by axonal fiber tracts in the white matter (Bullmore and Sporns, 2009; Toga et al., 2012). Long fibers are characterized by high energetic "wiring costs" (Bullmore and Sporns, 2012). To provide the energy for the maintenance of these long fibers the brain is relying on a constant energy supply. Recent findings have elegantly identified oligodendrocyte-derived lactate as the main energetic substrates for axonal maintenance (Fünfschilling et al., 2012). Consistently, disruption of this oligodendrocyte-neuronal metabolic coupling triggered neurodegeneration (Lee et al., 2012). Systemic inflammation poses dramatic challenges to the energetic supply of the brain.

To cover its *wiring costs* the brain is highly reliant on a constant nutrient supply. Nutrient supply through blood vessels can be compromised through vascular pathologies associated with systemic inflammation. During severe sepsis, disseminated intravascular coagulation leads to diffuse intravascular formation of thrombi and hemorrhages due to depletion of coagulation factors. The consequences are diffuse ischemic foci throughout the body and dysfunction of affected organs. Infarctions or hemorrhages occurring in the course of long connecting tracts in the brain lead to disconnection of distant brain regions and reduced efficiency of neuronal networks. These changes might contribute to white matter lesions observed in MRI studies of acute sepsis cases (Sharshar et al., 2007) and sepsis survivors (Morandi et al., 2012; Semmler et al., 2013). Mitochondrial dysfunction is another complication of sepsis affecting brain function. Despite normal or even increased tissue oxygen availability (Boekstegers et al., 1991), oxygen utilization is drastically reduced in critically ill patients leading to multi-organ failure and mortality (Brealey et al., 2002). Brain mitochondrial dysfunction was shown in rodent models of sepsis (d'Avila et al., 2008). This hibernation-like metabolic state was proposed as a physiological protective tissue reaction (Singer et al., 2004). However, mitochondrial dysfunction might persist in sepsis survivors (Comim et al., 2011). The cause for mitochondrial dysfunction might be a combination between induction of ROS during sepsis (Wong et al., 1996) and impaired mitochondrial turnover (Singer, 2014). Taken together, the complicated interrelation between focal hypoperfusion and reduced oxygen utilization lead to a complex acute and chronic phenotype referred to as sepsis-associated encephalopathy (Gofton and Young, 2012; Sonneville et al., 2013).

Autoimmune disorders have a chronic course of vascular pathology with acute flares. The most common vascular pathology is the autoantibody-associated antiphospholipid syndrome (Ben Salem, 2013; Giannakopoulos and Krilis, 2013;Million and Raoult, 2013). Patients with antiphospholipid syndrome display cognitive deficits (Gómez-Puerta et al., 2005; Tektonidou et al., 2006). MRI studies found diffuse infarctions and white matter lesions in these patients (Gómez-Puerta et al., 2005; Tektonidou et al., 2006; Valdés-Ferrer et al., 2008). The role of mitochondrial dysfunction in autoimmune diseases is not clear. In SLE, mitochondrial dysfunction and ATP depletion were shown as triggers of lymphocytic cell death (Gergely et al., 2002). An interesting view of mitochondrial dysfunction in the context of MS was recently proposed (Lassmann, 2011). In line with the concept of high "wiring costs" imposed on the brain by long intermodular connections (Bullmore and Sporns, 2012), Hans Lassmann argues that inflammation in MS causes mitochondrial damage and inability of the brain to maintain neuronal processes. The source of mitochondrial damage is radicals formed as a consequence of inflammation in MS (Lassmann, 2011; Fischer et al., 2012). Disruption of neuron–glia metabolic coupling might be another potential mechanism causing neurodegeneration and network disruption (Allaman et al., 2011; Lee et al., 2012).

Taken together these findings indicate that systemic inflammation leads to an energy crisis of the brain that reduces its connectivity. Oxidative stress might be the main mediator of this pathology. Thus, inflammation-induced changes in the brain resemble hallmarks of the aged brain where oxidative damage leads to decreased expression of genes associated with synaptic plasticity and increased expression of stress-response genes (Lu et al., 2004). Likewise, the brain during systemic inflammation shows hallmarks of neurodegenerative diseases where oxidative stress and mitochondrial damage have consistently been found (Lin and Beal, 2006). Studies with bigger sample sizes are needed to identify common mechanisms between systemic inflammation, brain aging, and neurodegeneration.

#### **SYSTEMIC INFLAMMATION-ASSOCIATED IMMUNOPATHOLOGY IRREPARABLY DAMAGES THE ARCHITECTURE OF THE BRAIN**

With approximately 90 billion neuronal and non-neuronal cells, respectively (Azevedo et al., 2009), the human brain is characterized by high architectural complexity (Braitenberg and Schüz, 1991). Due to this complexity, the repair capacity is limited rendering the human brain highly susceptible to tissue damage. As primary preventive measures the brain is protected from different modes of tissue damage; for example, by the skull bone from mechanical damage or by the BBB from blood-borne pathogens. The BBB is mainly formed by endothelial cells and astrocytes (Abbott et al., 2006; Weiss et al., 2009); the formation of tight junctions between endothelial cells forms a highly selective barrier that becomes more permeable during systemic inflammation (McColl et al., 2008; Weiss et al., 2009). A third kind potential source of brain tissue damage is the immune system itself. As pathogen defense is invariably associated with host tissue damage (Graham et al., 2005) an anti-inflammatory milieu preserves the brain from aberrant immune activation. Under physiological conditions, astrocytes and neurons actively modulate the activation of brain immune cells (Neumann, 2001; Tian et al., 2012). Through this cross-talk brain cells can actively recruit immune cells for purposes of brain homeostasis such as synaptic plasticity, induction of inflammation, clearance of debris, and resolution of inflammation.

Brain-resident microglia and peripheral immune cells maintain immune surveillance of brain parenchyma, CSF, and perivascular space for infectious agents or damage-associated milieu changes (Ousman and Kubes, 2012; Ransohoff and Engelhardt, 2012). In the case of brain infection, complete eradication of some invading pathogens can only be achieved at the cost of irreparable damage to brain tissue. To prevent such damage, the immune system has established active mechanisms of pathogen tolerance (Medzhitov et al., 2012). Examples for coexistence-prone pathogens are herpes simplex virus type I (Khanna et al., 2004) or *Cryptococcus gattii* (Cheng et al., 2009). A growing body of evidence indicates that not only immune tolerance but also resolution of neuroinflammation is a tightly regulated active immunological process (Schwartz and Baruch, 2014). Taken together, anti-inflammatory brain milieu, pathogen tolerance, and resolution of neuroinflammation require a balanced action between different branches of the immune system. An imbalance within the immune system leading to systemic inflammation may be a driver of neurodegeneration through the mechanisms discussed below.

#### **CASPASES AS MEDIATORS OF INFLAMMATION AND NEURODEGENERATION**

Apoptosis is one of the main drivers of neurodegeneration. Apoptosis and cell death constantly occur under physiological conditions throughout the human body and cell debris is cleared by immune cells mostly without induction of chronic inflammation (Green et al., 2009). However, during systemic inflammation, apoptosis of stressed cells might further exacerbate the underlying pathology (Zitvogel et al., 2010). Activators of apoptosis lead to direct or indirect activation of caspases. Interestingly, caspases are not only classical executors of apoptosis (Friedlander, 2003)

but also inflammatory caspases are crucial for the activation of the innate immune system through the inflammasome (Martinon and Tschopp, 2004). Caspase-1, as the cleaving enzyme for IL-1β and constituent of the inflammasome, is the prototypic representative of the latter class of caspases (Martinon et al., 2002).

Activation of the innate immune system through the inflammasome is a driver of pathology in age-associated and autoimmune neurodegenerative disorders. In AD, the NLRP3 inflammasome was described a sensor of β-amyloid (Heneka et al., 2013). Of interest, deletion of Caspase-1 or NLRP3 rescued the phenotype in APP/PS1 AD transgenic mice (Heneka et al., 2013). Consistently with that the deletion of the inflammasome scavenger receptor CD36 (Sheedy et al., 2013) ameliorated pathology in the Tg2576 AD mouse model (Park et al., 2013). MS-like lesions were found in humans with mutations of proteins associated with the inflammasome (Compeyrot-Lacassagne et al., 2009). Additionally, NLRP3 and IL-1 knockout mice had decreased pathology following EAE (Matsuki et al., 2006; Gris et al., 2010). Approved clinical treatment options of MS such as IFN β (Guarda et al., 2011) or glatiramer acetate (Burger et al., 2009) have been shown to decrease IL-1β levels, the main cytokine processed by the inflammasome. Additionally, novel evidence extends the functions of the classical apoptotic caspases linking neuroinflammation to neurodegeneration. Activation of microglial caspase-8, -3, and -7 can drive neurodegeneration (Burguillos et al., 2011). Finally, recent evidence from *C. elegans* has described protective effects through mediators of the intrinsic apoptosis pathway against ROS (Yee et al., 2014). Taken together, these finding show an intricate relationship between inflammation and activation of apoptosis. However, the definite role of these mediators in the brain remains to be characterized.

#### **MICROGLIA- AND MACROPHAGE-DERIVED MICROVESICLES AS INDUCERS OF NEURODEGENERATION**

Cellular components of innate immunity can pack and secrete inflammatory messengers in microvesicles (MVs). Peripheral macrophages, as well as brain microglia can secrete inflammasome components (caspase-1, IL-1β, and IL-18) in MVs, and the presence of extravesicular inflammatory inducers (e.g., astrocitic ATP) is sufficient to induce the neurotoxicity by the inflammatory load of MVs (Bianco et al., 2005; Sarkar et al., 2009; Gulinelli et al., 2012). Although MVs are visible in the CSF in healthy controls, the load of MVs in RRMS, neuromyelitis optica (NMO), brain infections, and brain tumors is significantly increased and, in MS, correlates with disease activity (Verderio et al., 2012). Recent evidence suggest that MVs play a critical role in the spectrum of AD as well. MVs released by activated microglia participate in the neurodegenerative process of AD by promoting the generation of highly neurotoxic soluble forms of β-amyloid (Joshi et al., 2014). Based on this collective evidence, it is now clear that EVs produced by peripheral myeloid cells, as well as immune brain cells, are novel and potentially critical biomarkers for neuroinflammatory conditions by providing a link between inflammation and neurodegeneration.

#### **IMMUNE CELLS AND IMMUNE MEDIATORS AS DRIVERS OF NEURODEGENERATION**

Various triggers of apoptosis have been described with respect to the brain. Neuronal apoptosis can be directly induced by ROS, pro-inflammatory cytokines or activated immune cells. The consequences of ROS-induced mitochondrial damage on brain metabolism have been discussed above. Additionally, damaged mitochondria are a major source of ROS (Rego and Oliveira, 2003) and mediators of apoptosis (Green and Kroemer, 2004). Conversely, inactivation of ROS has anti-apoptotic effects (Hockenbery et al., 1993; Greenlund et al., 1995). The inflammatory cytokine TNFα (Tamatani et al., 1999;Kaur et al., 2014) and tumor necrosis factor-related apoptosis-inducing ligand (TRAIL) (Aktas et al., 2005) directly induce neuronal apoptosis (**Figure 3**). Additionally, intracerebroventricularly injected TNFα was shown to induce depression-like symptoms (Kaster et al., 2012). Cytokine mediated induction of apoptosis was also observed by IL-1β (Wang et al., 2005; Kaur et al., 2014). Sources of cytokines under systemic inflammation are brain resident, paravascular or peripheral immune cells (Dantzer et al., 2008). Furthermore, activated immune cells can directly induce neuronal cell death (**Figures 1** and **3**). Brain-resident microglia convey neuronal toxicity through various mechanisms including secretion of neurotoxic factors (Liu and Hong, 2003; Block et al., 2007), as well as through activation of cyclooxygenase/prostaglandin E2 (COX/PGE2) pathways (Montine et al., 1999; Liang et al., 2005). In fact, blocking the COX/PGE2 pathway by experimentally deleting the prostaglandine receptor EP2 increases mitochondrial degradation of β-amyloid, potentially opening a new therapeutic avenue for AD (Johansson et al., 2015).

Peripheral immune cells can penetrate the BBB under conditions of systemic inflammation (Schmitt et al., 2012) and contribute to brain pathology (Rezai-Zadeh et al., 2009) (**Figure 3**). Cytotoxic T-cells were shown to be directly neurotoxic in autoimmune and aging-associated neurodegenerative disorders of the CNS (Neumann et al., 2002). Co-localization of T-cells with neurons and neuron-specific cytotoxicity of T-cells was shown *in vivo* and *in vitro* (Giuliani et al., 2003; Nitsch et al., 2004). The pivotal role of pathogenic T-cells in MS is suggested by the fact that adoptive transfer of myelin reactive T-cells is sufficient to induce EAE in rodents. Finally, B-cells have multiple roles in MS (Krumbholz et al., 2012). On the one hand, B-cells can serve as antigen presenting cells for T-cells and B-cell derived cytokines can activate pathogenic T-cells (Schneider et al., 2011). On the other hand, intrathecal IgG synthesis and persistent oligoclonal bands in the CSF are a hallmark finding in MS. Furthermore, antibodies and complement deposition is present in acute MS lesions (Lucchinetti et al., 2000). B-cells were shown to mature in the draining cervical lymph nodes and migrate to the brain (Stern et al., 2014). With advancing pathology, B-cells can be found in serum, brain parenchyma, and meninges of patients (Serafini et al., 2004; Kuerten et al., 2014); and B-cell depletion was shown to be beneficial in a subgroup of MS patients (Hauser et al., 2008).

#### **ANTI-BRAIN ANTIBODIES AS DRIVERS OF NEURODEGENERATION**

B-cell-derived anti-brain antibodies have been identified as drivers of brain pathology in various diseases. In the last decade, an increasing number of anti-brain antibodies has been detected that can affect cognition and behavior (Diamond et al., 2013).

For many of these newly discovered antibodies their frequency in disease and involvement in pathogenesis has not yet been determined. Although anti-brain antibodies can be present in around 5% of healthy individuals (Diamond et al., 2013), an intact BBB restricts entry of antibodies into the brain. Under pathological conditions, antibodies may penetrate the BBB through different mechanisms including local and systemic inflammation, or antigen mediated endocytosis (Diamond et al., 2013). In addition, fenestrated endothelial cells of the circumventricular organs, which lack the tight junction of the BBB, may facilitate entry of antibodies into the brain parenchyma (Popescu et al., 2011); or antibodies may be produced intrathecally by B-cells, which migrate into the CNS.

So far, few antibodies have been confirmed to be directly neurotoxic. The most striking evidence is provided by tumorassociated autoantibodies causing paraneoplastic neurological disorders. These antibodies are cross-reactive to neuronal antigens expressed on cancer cells; neuronal pathology is induced through neuronal cell death after antibody binding to neuronal antigens. Clinically, patients present with severe neuropsychiatric symptoms. Rapid removal of paraneoplastic antibodies and surgical removal of the tumor can prevent further neuronal cell death und may reverse neurological pathology. Patients that harbor an antibody response to intracellular antigens have a worse response to treatment compared to patients with antibodies to extracellular neuronal autoantigens. This could be due to an involvement of pathogenic T-cells, which cause irreversible neuronal cell death (Lancaster and Dalmau, 2012).

Anti-NMDA receptor encephalitis is the most common paraneoplastic disorder, with ovarian teratomas as the underlying malignancy in approximately 50% of the patients (Dalmau et al., 2011). These antibodies bind the extracellular domain of the NMDA receptor subunit 1 (GluN1) and cause psychiatric symptoms such as anxiety, memory loss, and psychosis (Dalmau et al., 2007). Binding of antibodies results in reduced levels of NMDA receptor *in vitro* and *in vivo* through antibody mediated capping, crosslinking, and internalization of the receptor (Hughes et al., 2010), which can be reversed upon removal of the antibodies (Seki et al., 2008). Interestingly, patients with high titers of NMDA receptor antibodies have lower levels of the NMDA receptors in postsynaptic dendrites (Dalmau et al., 2008). Direct injection of the antibodies in the cortex of rodents results in excitotoxicity by increased glutamate release (Hughes et al., 2010). A recent study found NMDA receptor antibodies in patients with demyelinating disorders indicating that these antibodies might cause neuropsychiatric symptoms in these patients (Titulaer et al., 2014).

Furthermore, NMDA-receptor-specific antibodies to the subunit 2 (GluN2) have been found in a subset of SLE patients with neuropsychiatric symptoms. These antibodies are crossreactive to DNA and bind the consensus sequence D/E W D/E Y S/G (DWEYS) present in the extracellular domains of GluN2A and GluN2B subunits of the NMDA receptor (Gaynor et al., 1997). DNA–NMDA receptor antibodies preferentially bind the open configuration of the NMDA receptor and augment NMDA receptor-mediated excitatory postsynaptic potentials (Faust et al., 2010). DNA–NMDA receptor antibodies have been found in serum, CSF, and brain tissue of SLE patients (DeGiorgio et al., 2001; Kowal et al., 2006). Notably, patient derived DNA–NMDA receptor antibodies cause hippocampal neuronal loss and persistent memory impairment in rodents (Kowal et al., 2006). These findings indicate that DNA–NMDA receptor antibodies are directly involved in neurodegenerative processes in SLE. Hippocampal brain abnormalities in NPSLE patients further support this notion (Appenzeller et al., 2006). Depending on the antibody concentrations, DNA–NMDA receptor antibodies can cause either neuronal dysfunction by transiently enhancing excitatory postsynaptic potentials or can result in neuronal cell death (Kowal et al., 2006; Faust et al., 2010). This evidence could be of high relevance in terms of reversibility of symptoms. Furthermore, anti-brain antibodies were also shown to induce neuropsychiatric symptoms in patients with other autoimmune disorders such as celiac disease (Alaedini et al., 2007; Hadjivassiliou et al., 2013) or inflammatory bowel diseases (Häuser et al., 2011; Papathanasiou et al., 2014). Taken together, anti-brain antibodies were shown to cause neuropsychiatric pathology in different diseases presenting novel therapeutic options.

These promising findings of anti-brain antibodies in systemic inflammatory disorders encouraged the search for anti-brain antibodies in inflammatory brain disorders. MS is characterized by oligoclonal bands in the CSF and antibodies in acute MS lesions (Lucchinetti et al., 2000). Histopathological studies confirm antibody mediated demyelination (Storch et al., 1998). Despite these findings all attempts to identify pathogenic antibodies to CNS antigens and infectious agents in the context of MS were rather unsatisfactory. In contrast to MS, highly specific anti-brain antibodies were discovered in NMO (Lennon et al., 2005), a disease which can closely resemble MS but requires different treatment. Antibodies to the water channel protein aquaporin-4 (AQP4) are detectable in around 80–90% of patients with NMO (Mader et al., 2010; Waters et al., 2012). AQP4 is localized on astrocytic endfeet forming the BBB. NMO is characterized by optic neuritis and longitudinal extensive transverse myelitis over three or more vertebral segments, which can lead to blindness and paralysis of patients (Wingerchuk et al., 1999; Cree, 2008). AQP4 antibody seropositivity is highly predictive for the disease (Matiello et al., 2008) and enables early treatment of patients. This is particularly important since IFN β, commonly used for treating MS, can worsen the disease outcome of NMO patients (Palace et al., 2010). Pathological findings show immunoglobulin and complement deposition around blood vessels with AQP4 specific loss in brain and spinal cord lesions (Lucchinetti et al., 2002). Rodents develop NMOlike lesions in the brain and spinal cord upon injection of patient derived AQP4 antibodies in an EAE animal model (Bennett et al., 2009; Bradl et al., 2009; Kinoshita et al., 2009). In addition, one study showed loss of astrocytes and demyelinating lesions after injection of AQP4 IgG into the brain together with human complement (Saadoun et al., 2010). AQP4 antibodies might require the help of encephalitogenic T-cells to breach the BBB (Saadoun et al., 2010; Pohl et al., 2013) and in addition CNS specific T-cells may require a local inflammatory environment for the antibodies at the lesion site (Pohl et al., 2013). Antibodies to AQP4 bind to astrocytes and lead to complement depend cytotoxicity as well as antibody

dependent cytotoxicity, resulting in astrocytosis (Papadopoulos et al., 2014). Demyelination and neuronal cell death could occur as a secondary inflammatory response, due to secretion of toxic compounds such as nitrogen species, reactive oxygen or glutamate by activated astrocytes (Brosnan and Raine, 2013), which could lead to limited trophic support to myelin (Levy, 2014). Recently, cognitive impairment has been reported in NMO (Saji et al., 2013), yet this finding needs to be replicated in larger cohorts. The pathological role of AQP4 antibodies in cortical neuronal loss at areas of high AQP4 expression is not well understood. Although NMO is a rather rare disease, the findings obtained from AQP4 IgG and its contribution in glial injury, demyelination and neurodegeneration could be of direct relevance for MS and related inflammatory diseases, and will help to better understand the role of astrocytes in inflammation and neurodegeneration.

#### **IMMUNE-MEDIATED DISRUPTION OF THE NEUROGENIC NICHE MAY CONTRIBUTE TO NEURODEGENERATION**

A final mechanism potentially connecting systemic inflammation and neurodegeneration is impairment of neurogenesis. Neurogenesis is a central mechanism required for neuronal maintenance and adaptive plasticity in the healthy and diseased brain (Jin et al., 2006a). Inflammatory mediators have various effects on neurogenesis (Whitney et al., 2009). Impairment of neurogenesis was shown in neurodegenerative diseases such as AD (Lazarov and Marr, 2010) and neuropsychiatric disorders such as depression (Sahay and Hen, 2007). Interestingly, approved AD drugs (Jin et al., 2006b; Kotani et al., 2008) and chronic antidepressant treatment (Malberg et al., 2000; Santarelli et al., 2003) induce neurogenesis. Inflammation and microglial activation is detrimental for neurogenesis that can be restored by anti-inflammatory treatment (Ekdahl et al., 2003; Monje et al., 2003). Moreover, microglia are not only involved in the maintenance of the neurogenic niche (Sierra et al., 2010) but also in synaptic maintenance (Stevens et al., 2007; Parkhurst et al., 2013). Of interest, systemic immune cells were shown to be involved in regulation of neurogenesis. CD4<sup>+</sup> T-cells were shown to promote (Wolf et al., 2009) while CD8<sup>+</sup> T-cells impair proliferation of neural progenitor cells (Hu et al., 2014). An effect of B-cells on neurogenesis was not observed (Wolf et al., 2009). These findings need to be independently replicated. However, one may speculate that neuropsychiatric symptoms elicited by chronic inflammation may be driven by detrimental changes of neuronal homeostasis. Thus, specific immune modulatory treatment might be beneficial.

#### **CONCLUDING REMARKS**

The immune and nervous systems have co-evolved from early invertebrates to higher mammals, creating an intricate bidirectional modulating dialog. The CAP is a good example of the influence of the nervous system upon the immune response. In the opposite direction, the activation of innate immunity in response to serious, as well as non-life threatening, infections induces the maturation and release of TNF, IL-1β, and other inflammatory cytokines that in turn cause transient anorexia, malaise, depression, and other features of the sickness syndrome (**Figure 3**). This is not surprising if we take into consideration that glia constitutes no less than half of the cells in a mammalian brain.

Sustained systemic inflammation is a common feature of many autoimmune disorders, and is present in most sepsis survivors. Cognitive impairment is common in sepsis survivors, as well as patients suffering from chronic inflammatory conditions. Cognition and behavior are persistently challenged, even after apparent resolution of acute sepsis. Moreover, systemic inflammation occurring in a susceptible brain (e.g., patients with AD) may lead to even further disruption in quality of life and activities of daily living. Up to 95% of patients with SLE develop neuropsychiatric dysfunction. In SLE, part of the repertoire of DNA-binding autoantibodies cross-react with hippocampal NMDA receptors, and – through a leaky BBB – gain access to the brain, inducing cognitive decline and other neuropsychiatric manifestations. In patients with rheumatoid arthritis, the baseline vagal tone of is persistently low, suggesting a possible mechanism for persistent inflammation. Those examples indicate that the normal neuroimmune cross-talk in health can become deleterious during disease, particularly in a primed brain – one with preexistent damage. Recently, cellular, molecular, environmental, and genetic components have been linked to the persistent brain disfunction of systemic inflammation. Here, we have discussed mechanistic evidence for the intricate interrelation between inflammation and neurodegeneration. Identification of druggable targets derived from these mechanisms holds the promise to prevent long-term disability and improve the quality of life in patients with chronic inflammatory conditions.

#### **ACKNOWLEDGMENTS**

Simone Mader work was supported by S.L.E. Lupus Foundation. The authors want to thank Benjamin Obholzer for the image design.

#### **REFERENCES**


a meta-analysis of genome-wide association studies. *Lancet* 377, 641–649. doi:10.1016/S0140-6736(10)62345-8


lateral sclerosis-associated mutations in copper-zinc superoxide dismutase. *Proc. Natl. Acad. Sci. U.S.A.* 95, 15763–15768. doi:10.1073/pnas.95.26.15763


G and human complement produces neuromyelitis optica lesions in mice. *Brain* 133(Pt 2), 349–361. doi:10.1093/brain/awp309


Zitvogel, L., Kepp, O., and Kroemer, G. (2010). Decoding cell death signals in inflammation and immunity. *Cell* 140, 798–804. doi:10.1016/j.cell.2010.02.015

**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: 29 November 2014; accepted: 15 January 2015; published online: 02 February 2015.*

*Citation: Sankowski R, Mader S and Valdés-Ferrer SI (2015) Systemic inflammation and the brain: novel roles of genetic, molecular, and environmental cues as drivers of neurodegeneration. Front. Cell. Neurosci. 9:28. doi: 10.3389/fncel.2015.00028 This article was submitted to the journal Frontiers in Cellular Neuroscience. Copyright © 2015 Sankowski, Mader and Valdés-Ferrer. 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.*

# **Yash B. Joshi and Domenico Praticò\***

Department of Pharmacology and Center for Translational Medicine, Temple University School of Medicine, Philadelphia, PA, USA

#### **Edited by:**

Victoria Campos, Instituto Nacional de Neurologia y Neurocirugia, Mexico

#### **Reviewed by:**

Muzamil Ahmad, Indian Institute of Integrative Medicine, India Luigi Iuliano, Sapienza University of Rome, Italy

#### **\*Correspondence:**

Domenico Praticò, Department of Pharmacology and Center for Translational Medicine, Temple University School of Medicine, 947 Medical Education and Research Building, 3500 North Broad Street, Philadelphia, PA 19140, USA e-mail: praticod@temple.edu

Alzheimer's disease (AD) is the most common, and, arguably, one of the most-well studied, neurodegenerative conditions. Several decades of investigation have revealed that amyloid-β and tau proteins are critical pathological players in this condition. Genetic analyses have revealed specific mutations in the cellular machinery that produces amyloid-β, but these mutations are found in only a small fraction of patients with the earlyonset variant of AD. In addition to development of amyloid-β and tau pathology, oxidative damage and inflammation are consistently found in the brains of these patients. The 5 lipoxygenase protein enzyme (5LO) and its downstream leukotriene metabolites have long been known to be important modulators of oxidation and inflammation in other disease states. Recent in vivo evidence using murine knock-out models has implicated the 5LO pathway, which also requires the 5LO activating protein (FLAP), in the molecular pathology of AD, including the metabolism of amyloid-β and tau. In this manuscript, we will provide an overview of 5LO and FLAP, discussing their involvement in biochemical pathways relevant to AD pathogenesis. We will also discuss how the 5LO pathway contributes to the molecular and behavioral insults seen in AD and provide an assessment of how targeting these proteins could lead to therapeutics relevant not only for AD, but also other related neurodegenerative conditions.

**Keywords: Alzheimer's disease, amyloid beta, tau, synapse, memory, 5-lipoxygenase, oxidative stress, neuroinflammation**

#### **ALZHEIMER's DISEASE: BACKGROUND**

Alzheimer's disease (AD) is the most common aging-associated neurodegenerative condition with dementia, marked by profound and irreversible memory impairment and cognitive deficits. The total number of individuals with AD worldwide is estimated to be over 35 million with a predicted estimated annual economic burden of \$600 billion USD (Alzheimer's Association, 2014). The vast majority of AD cases are sporadic, without a clear genetic component, and symptoms of AD typically declare themselves after the age of 65, with 11% of those 65 and older and 32% of 85 and older showing signs of AD (source: Alzheimer's Association). Current population demographics suggest that those aged 65 and older will increase from 13% now to 20% of total population in 2030, making the future burden of AD a tremendous public health challenge. In stark contrast to the looming public health challenge of AD, current therapeutic options for AD are limited. Although a plethora of agents are currently being investigated in phase II and phase III trials, currently approved medications include several acetylcholinesterase inhibitors and N-methyl Daspartate (NMDA) antagonists, which do little to modify disease course (Caraci et al., 2013; Tan et al., 2014). Given the confluence of an increased burden of AD in the near-future to health systems globally and a lack of approved therapeutic targets, investigation of targets that address multiple different facets of AD pathophysiology must be actively sought to help address this problem. Below we will give an overview of some of the molecular insults associated with AD, and discuss how the 5-lipoxygenase (5LO) enzyme presents a novel molecular pathway that is an attractive target for AD therapy.

#### **A**β **AND tau IN ALZHEIMER's DISEASE**

Through extensive clinical and molecular work over the past several decades, the biochemical pathways that lead to AD pathology have been well characterized. The cardinal pathologies observed in AD are the extracellular deposits of amyloid-β protein (Aβ) known as Aβ plaques, and intracellular accumulations of the hyperphosphorylated microtubule-associated tau protein known as neurofibrillary tangles (Iqbal et al., 2010; Holtzman et al., 2011). Current dogma presumes Aβ as the upstream molecular initiator in AD based on evidence that mutations in the Aβ precursor protein (APP) and presenilins, main components of the pathways that cleave it to produce Aβ peptides, are found in early-onset, familial variants of AD. Additionally, patients with Down's syndrome, in which there is an additional chromosome 21, the locus of the APP gene, have significantly increased rates of AD when compared with the general population (Wilcock and Griffin, 2013). However, more recent clinical data have also found mutations in the APP gene that are protective and reduce AD risk (Jonsson et al., 2012, 2013).

Amyloid-β peptide is formed by the sequential cleavage of APP by the β-secretase (β APP cleavage enzyme, BACE 1) and the γ-secretase complex (composed of the nicastrin, presenillin, PS1), anterior-pharynx defective-1 protein [APH-1], and presenillin enhancer protein [Pen-2], as shown in **Figure 1.** While APP may be cleaved by α-secretase and then γ-secretase to produce nonamyloidogenic products, the Aβ producing pathway is thought to be priviledged in AD. Generation of higher amounts and subsequent aggregation of Aβ peptide through the sequential βand γ-secretase cleavages is thought to lead to soluble oligomers, followed by longer fibrils, and finally insoluble plaques, which are found abundantly in the vast majority of AD patients on autopsy. Although insoluble plaques have been found in the brains of patients without AD, current thinking is that low-*n* Aβ oligomers perpetuate the brunt of molecular insults in AD rather than insoluble plaques *per se* (Ono and Yamada, 2011).

The hyperphosphorylation of the microtubule-associated tau protein also contributes to the molecular damage in AD. Tau is thought to be important in neuronal ultrastructure and axonal transport, both critical to overall neuron function and signaling (Iqbal et al., 2010). Upon hyperphosphorylation, tau loses affinity for microtubules, dissociating from them, and begins to aggregate, eventually precipitating inside neuronal cells, as shown in **Figure 2.** While Aβ is hypothesized to be the initiating event, cortical burden of neurofibrillary tau tangles correlates with dementia severity much more robustly (Oddo et al., 2006; Nelson et al.,

2007). Normal tau protein phosphorylation status is generally thought to be maintained by the relative balance of tau-specific kinases(s), which would add phosphate, and phosphatase(s), which would remove phosphate. At present, cyclin-dependent kinase 5 and glycogen synthase kinase 3 beta represent two such tau kinases that have been found to be abnormally functional in the brains of AD patients, and therefore of functional importance (Hanger et al., 1992; Baumann et al., 1993; Pei et al., 1999).

Although the specific mechanisms from disruption of normal functioning of both Aβ and tau to AD symptomatology remains unclear, both have been associated with oxidative stress and inflammation found in the brains of AD patients.

# **OXIDATIVE STRESS AND INFLAMMATION IN ALZHEIMER's DISEASE**

Balance of oxidation and reduction is critical to appropriate cellar function and results from the interplay of mechanisms that produce pro-oxidant molecules and those processes that detoxify them. The brain receives an overwhelming proportion of total body blood flow (i.e., oxygen) and glucose when adjusted for its weight, and in neurons, this relatively high oxygen and glucose requirement is directed towards energy generation (i.e., ATP production) and oxidative phosphorylation in mitochondria (Scheinberg and Stead, 1949; Smith et al., 2007). The vast majority of oxidative species (predominantly reactive oxygen species such as superoxide anions, hydroxyl radicals and hydrogen peroxide) produced by this metabolism is detoxified by antioxidant vitamins (e.g., vitamin C, and E) as well as superoxide dismutase, catalase and glutathione peroxidase (Adibhatla and Hatcher, 2010). A basal level of oxidants contributes to appropriate cellular signaling and modulates cell function but elevated levels of reactive oxygen species leads to oxidation and irreversible modification of proteins, lipids and nucleic acids which ultimately result in the disruption of their regular function (Ray et al., 2012). Interestingly, consistent evidence supports the hypothesis that a progressive accumulation of oxidative stress damage to important cellular molecules is a fundamental mechanism involved in the process of aging, which is the strongest risk factor for developing sporadic AD (Jacob et al., 2013).

In AD, oxidative modifications of proteins, lipids, DNA in the brain have been repeatedly found. Thus several studies have shown the presence of DNA and RNA oxidation products such as 8-oxo-2<sup>0</sup> -deoxyguanosine (8-oxo-dG) and 8-dihydro-2<sup>0</sup> guanosine (8OH) in AD brains, and lipid peroxidation products such as F2-isoprostanes and various reactive aldehydes have been reported in not only in AD brains but also in patient cerebrospinal fluid of patients with a clinical diagnosis of AD (de Leon et al., 2007; Sutherland et al., 2013). Although the precise mechanism of oxidation stress in AD remains elusive, the sulfur atom of methionine 35 in Aβ peptide has been shown to produce free radicals (Butterfield et al., 2013). Replacement of methionine to cysteine reduced oxidative damage in model organisms such as *C. elegans*, and substitution of sulfur with a methylene moiety reduces oxidation *in vitro* of the Aβ peptide (Yatin et al., 1999; Dai et al., 2007). Although clinical trials have not yet found a meaningful disease-modifying effect of antioxidant agents in the

treatment of AD, many studies have founds significant association between diets rich in antioxidants and lower risk of AD-risk (Pocernich et al., 2011).

Besides oxidative stress damage, altered inflammatory reactions are strongly associated with AD pathology and cognitive dysfunction. The dysregulation of inflammatory cytokines as well as immune cells (i.e., microglia and astrocytes) activation in AD brains has been well-documented. Microglia have some ability to clear Aβ, but are unable to effectively phagocytate high concentrations (or insoluble conformations) of it resulting in aberrantly activated microglia that associate with both Aβ plaques and neurofibrillary tangles (Hickman et al., 2008; Johnston et al., 2011; D'Andrea et al., 2004; Krabbe et al., 2013; Morales et al., 2013). Astrocytes can be directly stimulated by Aβ to secrete pro-inflammatory molecules, and evidence is growing that they have the ability to produce Aβ peptides themselves (Blasko et al., 2000; Wang et al., 2011a; Jo et al., 2014). Microgliosis and astrocytosis follow deposition of Aβ plaques and neurofibrillary tangles as shown by immunohistochemistry, and have been shown to precede neuronal loss (Sheng et al., 1997a,b; Sheffield et al., 2000; Wright et al., 2013). Resulting directly from immune cell activation, AD brains also have higher tissue levels of cytokines, including various interleukins (IL) such as IL-1β, tumor necrosis factor α (TNFα), and interferon γ (IFNγ), all independently-linked to increased production of Aβ and tau phosphorylation (Zilka et al., 2012).

Despite this compelling evidence, prospective clinical trials targeting oxidative stress and inflammation have previously not found clear and incontrovertible disease-modifying effects on the progression of AD. With regard to oxidation, the largest prospective clinical trials have tested combinations of the monoamine oxidase inhibitor, selegiline and alpha-tocopherol (i.e., vitamin E) or the cholinesterase inhibitor, donepezil and alpha-tocopherol, either in patients with severe disease or in the prodromal stage, in those with mild cognitive impairment (Sano et al., 1997; Petersen et al., 2005). In these studies, no difference was found between antioxidant-treated groups and controls in terms of disease progression or on cognitive assessment at the end of intervention. Other studies using a variety of antioxidants such as resveratrol and curcumin, among others, have noted either no benefit, or benefit with limited effect sizes in small cohorts (for an excellent review see Mecocci and Polidori, 2012). Recently, the Trial of Vitamin E and Memantine in Alzheimer's Disease (TEAM-AD) has reported reduction in functional cognitive decline in those receiving alpha-tocopherol compared to placebo (Dysken et al., 2014). However, these results have been criticized due to the relatively high dose of vitamin E used, and the fact memantine alone or in combination with vitamin E did not produce similar protective effects (Corbett and Ballard, 2014).

The earliest anti-inflammatory strategies in AD resulted from data showing reduced AD incidence in patients with rheumatoid arthritis, who have a high exposure to non-steroidal Joshi and Praticò 5-lipoxygenase and Alzheimer's disease

anti-inflammatory agents (McGeer and Rogers, 1992; Stewart et al., 1997). This finding was reproduced in many population studies (for a review, see McGeer and McGeer, 2013). However, placebo-controlled trials for AD using anti-inflammatory agents showed little benefit and significant adverse effects leading to subject dropout, although it is to be noted that the majority of these trials used a relatively short treatment window before trial termination or cessation (McGeer et al., 1996). Despite these trial failures, new data indicating mutations in Triggering receptor expressed on myeloid cells 2 gene (*TREM2*), which encodes a membrane protein found on immune cells, confers significant risk in the development of AD, has reinvigorated interest in an anti-inflammatory strategy in AD (Jonsson et al., 2013).

In recent years, mounting evidence has indicated that Aβ and tau pathologies begin depositing and impair neuronal function long before symptoms are manifest. Give this insight it is unlikely that recent and previously conducted clinical trials can adequately address the true effectiveness of anti-oxidant and antiinflammatory agents. Additionally, the multifactorial nature of the molecular insults in AD make it likely that a strategy that addresses not only pathological Aβ and tau accumulation, but also oxidation and inflammation would have the best chance of success.

#### **THE 5-LIPOXYGENASE PATHWAY**

The 5LO inserts molecular oxygen into the 5th carbon of free or esterified fatty acids, most notably arachidonic acid. In order to carry out the reaction, 5LO also requires 5LO activating protein, FLAP, which presents the substrate for enzymatic action (for a complete review on 5LO biology, see Rådmark and Samuelsson, 2010). Immediate products of 5LO include unstable 5-hydroperoxyeicosatetraenoic acid which is either reduced to 5-hydroxyeicosatetraenoic acid or leukotriene A<sup>4</sup> (LTA4). Depending on the cellular milieu, LTA4 can be metabolized either to leukotriene B<sup>4</sup> or C<sup>4</sup> (LTB<sup>4</sup> or LTC4), with LTC<sup>4</sup> further being metabolized to LTD<sup>4</sup> and LTE<sup>4</sup> (Bishayee and Khuda-Bukhsh, 2013), as shown in **Figure 3**.

5-lipoxygenase is found in vasculature, in endothelial cells, as well as throughout the central nervous system, in both neuron and glia (Chu and Praticò, 2009). Interestingly, 5LO expression and its metabolites increase with age in both animal models as well as human subjects. 5-lipoxygenase protein levels are particularly enriched in both cortex as well as hippocampus, areas known to be particularly vulnerable in neurodegeneration, and in AD in particular (Lammers et al., 1996; Chinnici et al., 2007). 5-lipoxygenase promotes lipid peroxidation *in vitro* as well as in brain tissue (Czubowicz et al., 2010; Czapski et al., 2012). Leukotrienes, metabolic products of 5LO activation, initiate immune cell chemotaxis and are critical molecular players in the inflammatory pathophysiology of asthma and allergy (Kanaoka and Boyce, 2014).

#### **5LO, FLAP AND THE ALZHEIMER's DISEASE PHENOTYPE**

Post-mortem studies have initially shown that 5LO is increased in AD (Firuzi et al., 2008; Ikonomovic et al., 2008). A small pilot study in humans has linked 5LO gene polymorphisms to

early- and late-onset AD, although large-scale population studies are yet to confirm these findings (Qu et al., 2001). In neuro2A cells harboring the Swedish APP mutation, knockout of 5LO reduces reduces their ability to produce and release soluble Aβ levels (Chu and Praticò, 2011a). This effect is due to reduced activity of γ-secretase, with 5LO knockout lowering steady-state expression of nicastrin, PS1, APH-1 and Pen-2 proteins without affecting either APP, the β-secretase or α-secretase pathway. On the other hand, overexpression of 5LO produces the opposite effect *in vitro*: Aβ levels are elevated, and associated with increased protein and mRNA levels of all γ-secretase proteins. While several molecules have been linked to the regulation of γ-secretase mRNA level, 5LO modulation occurs through the phosphorylation of cyclic adenosine response element-binding protein, CREB. Pharmacological inhibition of 5LO also reproduces 5LO knockout effects (Chu and Praticò, 2011b). Although γ-secretase produces Aβ peptides, which are pathologic in AD, it is also involved in the processing of Notch, a protein critical for neuronal functioning and differentiation (Imbimbo and Giardina, 2011). Clinical trials investigating γ-secretase inhibitors for use in AD have to date been largely unsuccessful because they are hypothesized to have altered Notch signaling as well as Aβ production (Doody et al., 2013). Fortunately, with 5LO modulation, γ secretase-dependent Notch production is unperturbed, making any future and potential use of this class of drug a feasible alternative to calssical γ-secretase inhibitors. By directly controlling the substrate availability for 5LO, FLAP modulation similarly affects Aβ production with FLAP knockout and pharmacological inhibition reducing Aβ through a CREB-mediated γ-secretase down-regulation (Chu and Praticò, 2012). *In vitro* data in neuronal cells have been reproduced in mouse models of AD-like amyloidosis with similar effects of 5LO and FLAP on Aβ, which includes immunohistochemical evidence of plaque reduction upon 5LO or FLAP knockout (Giannopoulos et al., 2013, 2014). Recently, these data have also been reproduced in TgCRND8 animals, with reduction found in amyloid-associated angiopathy upon administration of MK886, an inhibitor of FLAP (Hawkes et al., 2014). Intriguingly, leukotriene metabolites of 5LO have also been reported to increase β- and γ-secretase-mediated generation (Wang et al., 2013).

In addition to Aβ, 5LO and FLAP also modulate tau phosphorylation. In cells overexpressing 5LO, tau is hyperphosphorylated, generating both early-stage, as well as advanced-stage tau phosphoepitopes (Chu et al., 2013). Beyond tau phosphorylation, 5LO overexpression results in production of paired helical filaments of tau, precursors to insoluble tau deposition and neurofibrillary tangle formation. Knockout or inhibition of 5LO or FLAP produces results in amelioration of tau hyperphosphorylation (Chu and Praticò, 2013; Chu et al., 2013). Both *in vivo* in murine models and *in vitro*, 5LO pathway changes are mediated by 5LO and FLAP influence on cyclin-dependent kinase 5 activity (Chu and Praticò, 2013). As mentioned previously, abnormalities in Aβ metabolism are presumed to be upstream of tau phosphorylation but 5LO effects on tau phosphorylation seem to be independent of Aβ. In a series of *in vitro* experiments, γ-secretase blockade in the presence of 5LO overexpression did not prevent 5LO-mediated hyperphosphorylation of tau (Chu et al., 2013). These *in vitro* results have also been corroborated in not only a mouse model of amyloidosis such as the Tg2576 mouse, but also a model with both Aβ and tau pathology such as the 3xTg mouse (Chu et al., 2012b, 2013).

Beyond neuropathology, knockout of 5LO or FLAP mitigates age-dependent AD learning and memory insults seen in two models of AD. In both Tg2576 and 3xTg mice, 5LO knockout or inhibition is associated with improvements in learning and memory over baseline in fear conditioning paradigms while overexpression of 5LO is associated with greater cognitive insult (Chu and Praticò, 2011a; Chu et al., 2012a,b; Giannopoulos et al., 2013, 2014). Knockout or inhibition of 5LO and FLAP also ameliorates markers of synaptic protein pathology and restores hippocampal long-term potentiation to wild-type levels (Giannopoulos et al., 2013, 2014). Even in aged transgenic mice, when significant AD pathology has been deposited, targeting the 5LO pathway appears to improve the overall AD-like phenotype (Di Meco et al., 2014).

# **5LO, FLAP AND ALZHEIMER's DISEASE-ASSOCIATED OXIDATION AND INFLAMMATION**

While 5LO and FLAP appear to reduce the cardinal pathologies in AD, they also act on AD pathology-induced oxidative and inflammatory insult.

In cultured rat hippocampal neurons, 5LO pathway inhibition results in reduced Aβ-induced reactive oxygen species generation and subsequent calcium dysregulation in a concentrationdependent manner (Goodman et al., 1994). 5-lipoxygenase pathway inhibition also protects against glutamate-induced excitotoxicity *in vivo* in rats, particularly in aged animals (Uz et al., 1998). In other *in vitro* systems, 5LO overexpression does not by itself lead to oxidative damage, but when overexpressed in the presence of Aβ peptides, reduces glutathione peroxidase and catalase levels (Wang et al., 2011b). 5-lipoxygenase pathway inhibition by the pyrazole CNB-001, a 5LO-specific inhibitor, protects against endoplasmic reticulum dysfunction and proteasome toxicity induced by Aβ both in cultured neurons as well as *in vivo* (Valera et al., 2013). Interestingly, inhibition of FLAP is not sufficient to protect against Aβ oxidative toxicity *in vitro*, which suggests that even when 5LO is decoupled from Aβ metabolism, its ability to insert molecular oxygen is preserved, retaining pro-oxidative properties in a leukotriene-independent fashion. However, this phenomenon is otherwise not well described or replicated in neurons and requires further scrutiny and exploration. On the other hand, leukotriene blockade has been linked to reduction in cognitive deficits induced by traumatic brain injury in rats, a source of oxidative stress, and a well-known risk factor for AD (Corser-Jensen et al., 2014).

Besides oxidation, both 5LO and FLAP are active players in the neuroinflammation found in AD. Disruption of the 5LO pathway, either genetically or pharmacologically, reduces not only microglia, but also astrocytosis in the brains of AD animals, seen on immunohistochemical analyses (Chu and Praticò, 2012). Moreover, with 5LO/FLAP disruption there is an associated reduction in pro-inflammatory cytokine levels. An argument can be made that reduction in Aβ and tau pathology caused by 5LO/FLAP inhibition independently predisposes AD transgenic animals to have reduced neuroinflammation at baseline. While this may be true to some extent, chronic lipopolysaccharide administration in AD animals lacking 5LO increases steady-state levels of γ-secretase machinery and tau phosphorylation but does not change baseline microgliosis, astrocytosis, or brain levels of inflammatory cytokines (Joshi et al., 2014). This line of data suggests that the 5LO system's contribution to the neuroinflammation does not depend exclusively on Aβ or tau.

# **IMPLICATIONS FOR ALZHEIMER's DISEASE THERAPY AND BEYOND**

Although several clinical trials have targeted Aβ specifically, increasingly, their failures have lead commentators to reevaluate not only the time course of intervention in AD, but also the selected therapeutic targets. While tau is increasingly being recognized as a viable target, and anti-oxidant and anti-inflammatory strategies are frequently employed in studies to modify disease course, no agent to date has been employed that independently modulates Aβ metabolism, tau phosphorylation, oxidation and inflammation. Although more work must be done to dissect the mechanisms of 5LO action in model systems more advanced than mice before considering it as a viable target, 5LO pathway suppression has the added benefit of being safe in other chronic conditions such as asthma (i.e., Zileuton, Monteleukast). Beyond AD, 5LO pathway targeting may also be useful in other dementias and related conditions including: non-AD amyloidoses (i.e., cerebral amyloid angiopathy), tauopathies (e.g., frontotemporal dementia, chronic traumatic encephalopathy, progressive supranuclear palsy), oxidation-linked nervous system disorders (i.e., amyotrophic lateral sclerosis) and inflammation-centered disease processes (i.e., multiple sclerosis). Further work exploring the 5LO system would undoubtedly shed more light on the molecular pathways common to neurodegeneration and represent a novel and promising platform for future drug development.

### **ACKNOWLEDGMENTS**

The work form the author's lab described in the present article has been in part supported by grants for the National Institute of Health, the Alzheimer's Association and the Alzheimer's Art Quilt Initiative.

#### **REFERENCES**


model involvement of gamma-secretase. *Am. J. Pathol.* 178, 1762–1769. doi: 10. 1016/j.ajpath.2010.12.032


Zilka, N., Kazmerova, Z., Jadhav, S., Neradil, P., Madari, A., Obetkova, D., et al. (2012). Who fans the flames of Alzheimer's disease brains? Misfolded tau on the crossroad of neurodegenerative and inflammatory pathways. *J. Neuroinflammation* 9:47. doi: 10.1186/1742-2094-9-47

**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: 27 October 2014; accepted: 02 December 2014; published online: 14 January 2015*.

*Citation: Joshi YB and Praticò D (2015) The 5-lipoxygenase pathway: oxidative and inflammatory contributions to the Alzheimer's disease phenotype. Front. Cell. Neurosci. 8:436. doi: 10.3389/fncel.2014.00436*

*This article was submitted to the journal Frontiers in Cellular Neuroscience*.

*Copyright © 2015 Joshi and Praticò. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution and 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*.

# Evaluation of inflammation-related genes polymorphisms in Mexican with Alzheimer's disease: a pilot study

Danira Toral-Rios 1† , Diana Franco-Bocanegra2† , Oscar Rosas-Carrasco<sup>3</sup> , Francisco Mena-Barranco<sup>4</sup> , Rosa Carvajal-García<sup>5</sup> , Marco Antonio Meraz-Ríos <sup>6</sup> and Victoria Campos-Peña<sup>7</sup> \*

<sup>1</sup> Departamento de Fisiología Biofísica y Neurociencias, Centro de Investigación y de Estudios Avanzados, Mexico City, Mexico, <sup>2</sup> Posgrado en Ciencias Biológicas, Universidad Nacional Autónoma de México, Mexico City, Mexico, <sup>3</sup> Instituto Nacional de Geriatría, Mexico City, Mexico, <sup>4</sup> Hospital Regional de Alta Especialidad de Ixtapaluca, Estado de México, Mexico, <sup>5</sup> Centro Geriátrico SINANK'AY, Querétaro, Mexico, <sup>6</sup> Departamento de Biomedicina Molecular, Centro de Investigación y de Estudios Avanzados, Mexico City, Mexico, <sup>7</sup> Laboratorio Experimental de Enfermedades Neurodegenerativas, Instituto Nacional de Neurología y Neurocirugía Manuel Velasco Suárez, Mexico City, Mexico

#### Edited by:

Karla Guadalupe Carvajal, National Institute of Peadiatrics, Mexico

#### Reviewed by:

Changiz Geula, Northwestern University, USA Diego Albani, Istituto di Ricerche Farmacologiche Mario Negri-IRCCS, Italy

#### \*Correspondence:

Victoria Campos-Peña, Laboratorio Experimental de Enfermedades Neurodegenerativas, Instituto Nacional de Neurología y Neurocirugía Manuel Velasco Suárez, Mexico City, Insurgentes Sur 3877, ZIP 14269, Mexico neurovcp@ymail.com †These authors have contributed equally to this work.

> Received: 10 January 2015 Accepted: 31 March 2015 Published: 18 May 2015

#### Citation:

Toral-Rios D, Franco-Bocanegra D, Rosas-Carrasco O, Mena-Barranco F, Carvajal-García R, Meraz-Ríos MA and Campos-Peña V (2015) Evaluation of inflammation-related genes polymorphisms in Mexican with Alzheimer's disease: a pilot study. Front. Cell. Neurosci. 9:148. doi: 10.3389/fncel.2015.00148 Amyloid peptide is able to promote the activation of microglia and astrocytes in Alzheimer's disease (AD), and this stimulates the production of pro-inflammatory cytokines. Inflammation contributes to the process of neurodegeneration and therefore is a key factor in the development of AD. Some of the most important proteins involved in AD inflammation are: clusterin (CLU), complement receptor 1 (CR1), C reactive protein (CRP), tumor necrosis factor α (TNF-α), the interleukins 1α (IL-1α), 6 (IL-6), 10 (IL-10) and cyclooxygenase 2 (COX-2). In particular, COX-2 is encoded by the prostaglandin-endoperoxide synthase 2 gene (PTGS2). Since variations in the genes that encode these proteins may modify gene expression or function, it is important to investigate whether these variations may change the developing AD. The aim of this study was to determine whether the presence of polymorphisms in the genes encoding the aforementioned proteins is associated in Mexican patients with AD. Fourteen polymorphisms were genotyped in 96 subjects with AD and 100 controls; the differences in allele, genotype and haplotype frequencies were analyzed. Additionally, an ancestry analysis was conducted to exclude differences in genetic ancestry among groups as a confounding factor in the study. Significant differences in frequencies between AD and controls were found for the single-nucleotide polymorphism (SNP) rs20417 within the PTGS2 gene. Ancestry analysis revealed no significant differences in the ancestry of the compared groups, and the association was significant even after adjustment for ancestry and correction for multiple testing, which strengthens the validity of the results. We conclude that this polymorphism plays an important role in the development of the AD pathology and further studies are required, including their proteins.

Keywords: Alzheimer's disease, inflammation, prostaglandin-endoperoxide synthase, cyclooxygenase 2, genetic ancestry

# Introduction

Alzheimer's disease (AD) is a neurodegenerative disease which major symptom is the impairment of memory and other cognitive functions (Collette et al., 1999; Di Paola et al., 2007). It has been estimated that the global prevalence of dementia is around 4.5% and that AD accounts for 65% of the total number of dementia Alzheimer's type (DAT; Kalaria et al., 2008; Rizzi et al., 2014), which places AD as the major global cause of dementia. The prevalence of dementia varies considerably around the world. There are few studies particularly in Mexico, however it has been determined that in Mexican population above 20 years old the prevalence of dementia is 6.1–7.9%, and the country was reported to be the fifth with the highest prevalence of the disease in Latin America (Llibre Rodriguez et al., 2008; Prince et al., 2008; Mejia-Arango and Gutierrez, 2011; Ramírez-Díaz et al., 2015). AD is a complex disease which etiology relies both in environmental and genetic factors, but its exact causes are unknown to date. The definitive diagnose of AD can be only performed by post mortem histological analysis, in which certain distinctive lesions must be found. These lesions consist of protein aggregates known as neuritic plaques (NPs), composed of the amyloid-β peptide (Wong et al., 1985), and neurofibrillary tangles (NFTs), composed of hyperphosphorylated tau protein (Kosik et al., 1986).

In AD the neuroinflammation is an early and continuous feature of the disease (Hensley, 2010; McGeer and McGeer, 2010; Zhang and Jiang, 2015). It has been reported that the activation of the immune system, which leads to a general inflammatory state in the brains, one of the major and most constant characteristics of AD, as well as other neurodegenerative diseases (Meraz-Ríos et al., 2013). This response involves cellular and molecular changes, the recruitment of peripheral immune cells (Rezai-Zadeh et al., 2011), and the release of inflammatory mediators in the brain (Heneka et al., 2010).

Several studies in animal models have confirmed that the presence of Aβ in the brain leads to the activation of microglial cells and astrocytes (Frautschy et al., 1992; Hanzel et al., 2014). The high levels of chemokines and chemokine receptors in brain regions surrounding NPs suggest that there is a chemotactic migration of microglia towards Aβ peptides (Walker et al., 2006). The activation of astrocytes and microglia leads to an increased secretion of pro-inflammatory proteins, such as cyclooxygenases, complement proteins and their receptors, acute phase proteins, adhesion molecules, chemokines, and cytokines (Liao et al., 2004; Ramesh et al., 2013). The chronic increased secretion of this proteins leads to increased oxidative stress and enhances cell death, which leads to neurodegeneration in the central nervous system (CNS; Meraz-Ríos et al., 2013).

It has been proposed that sequence variations in the genes that code pro-inflammatory and anti-inflammatory proteins might play a role changing the function or expression rate of the proteins and in this way modifying the inflammatory response in the brain; this could have an effect in the risk of developing AD. We selected 14 single-nucleotide polymorphisms (SNPs), according to the Alzgene Top results. Numerous studies have examined the presence of SNPs in genes of proinflammatoy cytokines such as Interleukin-1α (IL-1α; Combarros et al., 2002), Interleukin 6 (IL-6; Chen et al., 2012), Tumor Necrosis Factor α (TNF-α; Laws et al., 2005; Ardebili et al., 2011) and in the anti-inflammatory Interleukin 10 (IL-10; Bagnoli et al., 2007), whose production have found altered in CSF and peripheral blood in AD patients (Blum-Degen et al., 1995; Swardfager et al., 2010). The promoter region SNPs of TNF gene, rs1800629 and rs1799724, have been studied for a possible involvement with a functional alteration in the production of this proinflammatory cytokine, in particular the presence of the rs1799724 in Caucasians with a diagnosis of probable AD, correlated with altered levels of Aβ42 in CSF (Laws et al., 2005). IL1A is another candidate gene associated with AD (Combarros et al., 2002); in this case we selected the rs17561 variant, which has been previously studied in Americans and Japanese populations (Minster et al., 2000; Yucesoy et al., 2006). As for the anti-inflammatory cytokines, the most studied polymorphisms are rs1800871, rs1800896, rs1800872; they are located in the promoter region of IL10 gene and have been related with an alteration of transcriptional activation with a genedosage related effect and have replicated an association in Caucasian populations (Bagnoli et al., 2007; Combarros et al., 2009). Also, a possible interaction within these SNPs and the variants on the IL6 gene was reported (Combarros et al., 2009).

In the same way, some polymorphisms on Clusterin (CLU) and Complement Receptor 1 (CR1) genes have shown a consistent association with AD in genome-wide association studies (GWAS; Harold et al., 2009; Lambert et al., 2009; Jun et al., 2010; Hu et al., 2011). Another protein that has been linked with the inflammatory process in AD is the c-reactive protein (CRP). This acute phase reactant has been found in association with plaques and NFTs (Yasojima et al., 2000). Also, elevated tissue levels of CRP have been related with an increased risk for developing AD; however when the disease has established, the CRP levels decrease (Schmidt et al., 2002; O'Bryant et al., 2013). For these reasons, this protein has been proposed as an early biomarker for AD. Moreover, SNPs on CRP gene like rs1130864 and rs1800947 have been involved with altered CRP levels (Kok et al., 2011) and with a possible association with AD in Caucasians (Flex et al., 2004; van Oijen et al., 2007). Additionally, the cyclooxygenase-2 (COX-2) is encoded by the prostaglandin-endoperoxide synthase 2 gene (PTGS2); this enzyme involved in neuroinflammation, is responsible for prostaglandins (PG's) synthesis. It has been documented that COX-2 expression increases in the hippocampus of AD brain, which also correlates to the severity of the pathology (Ho et al., 2001; Meraz-Ríos et al., 2013). Due to this fact, polymorphisms in the PSTG2 gene have been studied; mainly, the rs20417 located in the promoter region has shown an association with decreased risk of AD (Abdullah et al., 2006).

Recently, publications related to the presence of polymorphisms associated with the development of AD have increased; however, numerous lines of evidence have demonstrated discrepant results among populations. These findings suggest that it is necessary to reduce the confounding factors and focus on identifying the cause (Jiang et al., 2013). The vast majority of these studies have not evaluated the role of individual differences in ancestry as a confounding factor in their study population. Mexican and other Latin American populations are the result of mixed breeding among three ancestral populations: Amerindian, Caucasian and African. Intra or inter-group differences in genetic ancestry might represent a source of spurious associations in case-control studies if they are not appropriately regarded.

Considering all of the above mentioned, the aim of this study was to evaluate a set of candidate polymorphisms in inflammation-related genes, and to determine whether they are associated with the risk of developing AD in the Mexican population, considering the genetic ancestry of the study subjects.

# Materials and Methods

### Study Population

The subjects were patients with a clinical diagnose of dementia of the Alzheimer-type (DAT), or cognitively healthy subjects (controls). All subjects were Mexican, with Mexican ancestry back to the third generation, and at least 60 years old at the sampling date. This work was carried out according with the ethical standards of the Committee on Human Experimentation of the institution (Instituto Nacional de Neurología y Neurocirugía Number 100/07) in accord with the Helsinki Declaration of 1975. A Student's t test was done comparing the parameters of mean age between cases and controls.

Ninety four DAT patients, previously diagnosed by a group of Geriatricians and Neurologists, were included according to the NINCDS-ADRDA (Mckhann et al., 1984) criteria. The DAT patients were interviewed at their scheduled visits at Geriatric Clinic of the General Hospital Mocel in Mexico City and Geriatric Center in Querétaro.

The inclusion criteria were: (1) patients with DAT were required to be able to read and write; (2) patients have to be 60 years old and over; and (3) they had to hand in an informed consent sheet signed by both the informant and the elderly person. The exclusion criteria were as follows: (1) acute and/or exacerbated chronic disease present within 30 days before the interview that could affect the quality of response to questionnaires, according to the medical staff of the study; (2) decreased alertness (for any cause); (3) severe aphasia; (4) visual and hearing impairment making it difficult for the patient/caregiver to fill out the questionnaires; (5) suffering other neurological diseases that could have influenced the diagnosis of dementia; and (6) live in a nursing home.

The inclusion criteria for controls were: (1) being an adult of 60 years of age or over; (2) not having memory complaint reported by either the informant or the elderly person; (3) completing an Mini Mental State Examination (MMSE) score of ≥24; (4) having the ability to read and write; and (5) handing in an informed consent sheet signed by both the informant and the older person. The exclusion criteria for this group were: (1) suffering from any acute or severe chronic illness; and (2) being less alert or suffering from severe aphasia, impaired vision and/or hearing, which would make it difficult for the older person to answer any of the questionnaires.

# Analyzed Polymorphisms

The selection of the polymorphisms was performed searching in the Alzforum database. The search was directed in order to find SNPs, located in genes that code inflammation-related proteins, and which had previous reports of association with AD in other populations.

Fourteen SNPs were selected, which were distributed among eight genes that code inflammation-related proteins. The selected genes were: PTGS2 (Ma et al., 2008), CLU (Lambert et al., 2009), CR1 (Zhang et al., 2010), CRP (Eriksson et al., 2011), TNF (Ardebili et al., 2011), the pro-inflammatory interleukins 1α (IL-1α, Combarros et al., 2002) and 6 (IL-6, Chen et al., 2012), and the anti-inflammatory interleukin 10 (IL-10, Bagnoli et al., 2007). The selected SNPs are listed in **Table 1**.

Additionally, in order to perform the ancestry analysis of the study population, a set of 10 ancestry informative markers (AIMs) were selected (rs4884, rs2695, rs17203, rs2862, rs3340, rs722098, rs223830, rs1800498 and rs2814778). These AIMs are SNPs which frequencies significantly vary between Amerindian, Caucasian and African populations. These AIMs have already been tested to accurately estimate ancestry in populations of Latin American origin (Salari et al., 2005; Choudhry et al., 2006; Ziv et al., 2006).

# DNA Extraction and Genotyping

Peripheral blood was extracted from all subjects, and was stored in Vacutainer® tubes with EDTA. Genomic DNA was extracted from blood using the extraction kit Midi Kit QIAamp® DNA Blood (Qiagen). The extracted DNA was used to perform realtime PCR for the 14 inflammation related SNPs and the 10 AIMs. PCR was performed in a 7500 Fast Real Time PCR System® (Applied Biosystems). SNPs were genotyped using TaqMan® probes (Applied Biosystems) with real-time PCR detection.

# Statistical Analysis Ancestry Analysis

The genotypes obtained by the genotyping of the AIMs were analyzed in order to produce estimates of the proportion in which each ancestral population is represented in the study population. To achieve this, LEADMIX (Wang, 2003) and STRUCTURE (Pritchard et al., 2000) software were required.

For the analysis with LEADMIX, the 10 genotyped AIMs were used, while the information of the frequency of these AIMs in ancestral populations was obtained from the database dbSNPs,


#### TABLE 1 | SNPs analyzed in the present study.

available in the National Institute for Health website. Fisher's exact test was used to compare proportions between cases and controls.

For the analysis with STRUCTURE, only seven of the genotyped AIMs were used (rs4884, rs2695, rs17203, rs2862, rs3340, rs1800498 and rs2814778). Because the analysis with STRUCTURE requires individual genotype information from subjects belonging to the ancestral populations (reference subjects), a database composed of data collected by Salari et al. (2005), data from the 1000 Genomes Project (Abecasis et al., 2010), and data from the study population was used. Reference subjects for the Caucasian population were Utah residents of European ancestry, Finnish, English, Spanish, Italians and Germans. Reference subjects from the African population were people from Nigeria, Sierra Leone and the Central African Republic. Because the 1000 Genomes Project lacks information from the Amerindian population, reference subjects for this population were taken exclusively from Salari et al. (2005) and they were Mayan and Native Americans from the Pima, Cheyenne and Pueblo ethnic groups.

#### Association Testing

Allele and genotype frequencies were determined using the SNPStats software (Solé et al., 2006). Departure from Hardy–Weinberg equilibrium (HWE) was assessed by χ 2 test, using the GenAlEx 6.5 software (Peakall and Smouse, 2012).

For the association testing, estimates of the individual ancestry proportions obtained from STRUCTURE were included along with age and gender in the regression model for odds ratio calculation in SNPStats. For this analysis, only the values for Amerindian and African ancestry were included to avoid collinearity in the model. In order to account for multiple testing, the Bonferroni correction was applied while interpreting the significance of the associations found.

#### Linkage Disequilibrium Analysis

Additionally the level of linkage disequilibrium (LD) among the selected SNPs was calculated. For this purpose, the Haploview v. 4.2 software was used.

#### Results

#### Age and Gender in the Study Population

A total of 194 subjects were genotyped, 94 DAT and 100 controls. **Table 2** presents the distribution of gender among each study group as well as the mean age for each gender and group. No significant differences were observed in the distribution of gender between study groups (p = 0.543). Results from the Student's t test comparing the mean age between DAT and controls not showed significant differences (p = 0.111). Also, we didn't find differences when we performed a comparison testing the mean age of the different genders independently (women p = 0.071, men p = 0.899).

#### Ancestry Analysis of the Study Population Ancestry Proportion Estimation in Study Groups

The contribution of each ancestral population in each study group was calculated using the Bertorelle and Excoffier (1998) estimator with LEADMIX. Results are shown in **Figure 1**. As it can be noted, Amerindian ancestry was represented in a higher proportion in controls as well as in DAT, followed closely by Caucasian ancestry, and finally by African ancestry in a lower proportion. The results of the Fisher's exact test showed that there are no significant differences (p = 0.706) in



ancestry proportions between DAT and controls. This result was corroborated with the analysis in STRUCTURE, which yielded the ancestry proportion estimates for each study subject; the estimates were subsequently grouped into a cluster (**Figure 2**). Derived from these analyses we conclude that both groups are genetically homogeneous in ancestry.

#### Analyzed SNPs and AD Risk

Allele and genotype frequencies of the 14 SNPs were analyzed (**Table 3**). Genotypes were assessed for departure from HWE. Four polymorphisms presented a significant departure in the

study groups. The triangle obtained from the STRUCTURE software (Pritchard et al., 2000) allows to observe that the distribution of the individuals of each study group between the three ancestral clusters is similar.

control group; they were rs6656401 (p = 0.012), rs1799724 (p = 0.013), rs1800795 (p = 0.043) and rs1800871 (p = 0.000). Because of the deviations, those SNPs were excluded from our replication study.

In order to adjust for multiple testing, the Bonferroni correction was applied; according to the calculations, a corrected p value less than 0.00357 was considered significant. The association test showed important differences in the rs20417 located in the PTGS2 gene. The presence of GG genotype was significantly more frequent in the DAT group (OR = 2.89, 95% CI = 1.41–5.92, p = 0.0028), this could suggest an association with AD. No significant differences were found between DAT and controls in the allele frequencies of the rest of the SNPs analyzed (**Table 3**).

#### LD Calculation

The software Haploview was used to measure LD among pairs of loci, and the r 2 an LD parameter was calculated.

**Figure 3** shows that the highest values of r 2 can be found between the SNPs rs1800796 and rs1524107 within the IL-6 gene, and between the SNPs rs1800871 and rs1800872 within the IL-10 gene.

# Discussion

This study is focused on the search of genetic risk factors associated with AD in Mexican population, considering the ancestry of the population studied and analyzing some of the main genes related to the inflammatory process observed in AD patients. Previous studies have identified AD associated variants in the genes analyzed in this study; nevertheless, until now these studies have been often limited to the evaluation of variants within a single gene, or a few genes.

<sup>1</sup>http://www.ncbi.nlm.nih.gov/SNP

#### TABLE 3 | Allelic and genotyping analysis.


\*\* After Bonferroni correction (p = 0.00357) only the rs20417, was significant. + Excluded because of significant departure from Hardy-Weinberg in controls.

HWE has been proposed as a routine evaluation in case-control genetic association studies. Genotyping errors, population stratification, a small sample size and non-random mating generally contribute to departures on this equilibrium and favor false associations (Salanti et al., 2005; Wang and Shete, 2012). Based in this fact, we omit four SNPs from the association test, because they present a HWE departure in control group. Another resource that we used for reducing the probability of obtaining false positives in the statistical analysis of the association test was the correction for multiple testing.

In this study, the homozygous genotype GG in the SNP rs20417, was overrepresented in AD patients and underrepresented in controls in the PTGS2 gene. These results essentially match with the results obtained by Abdullah et al. (2006) and Listì et al. (2010), who found the C allele is linked to a decrease in the risk of developing the disease. Our data suggest that GG polymorphism in PTGS2 could be a risk factor for AD in Mexican population.

PTGS2 is expressed in neurons of the neocortex and hippocampus under normal physiological conditions, suggesting its essential role in normal neuronal function. COX-2 catalyzes the formation of PG's, which are well-known inflammatory mediators. In fact, during some pathological conditions it can be expressed in microglia and astrocytes. Several authors have noted that COX2 is involved in several neurological diseases such as Parkinson's disease (Teismann et al., 2003),

amyotrophic lateral sclerosis (Almer et al., 2001), schizophrenia (Müller et al., 2004) and AD (Pasinetti and Aisen, 1998; Hoozemans and O'Banion, 2005). In 1999, Yasojima et al. reported that COX-2 was substantially upregulated in affected areas of AD brain, this dominant upregulation in neurons could reflect a defensive response of the neurons to stress (Yasojima et al., 1999). However, the chronic expression of COX-2 may contribute to aggravate stress through the generation of free radicals causing oxidative stress and neurotoxicity. COX-2 in the CNS may enhance neuritic plaque formation by means of the production of PGE2, which can induce the expression of the Aβ precursor protein (APP). In turn, the presence of NPs can increase the expression of COX-2 (Nogawa et al., 2003).

Therefore, the GG genotype in the SNP rs20417 that displayed association with AD in this study might have an effect over the gene expression rate, resulting in an alteration of protein concentrations. This might exert an influence over the mechanisms of Aβ synthesis, thus contributing to modify the risk of developing the disease. Conversely, the C allele has a reduced promoter activity, suggesting that lower promoter activity of COX-2 gene might be protective for AD. In the rest of the analyzed SNPs no significant differences were found when comparing the allele or genotype frequencies between AD patients and controls. These results contradict previous reports in which an association between these SNPs and the risk of developing AD was found (Bonafè et al., 2001; Brull et al., 2001; Combarros et al., 2002, 2009; Flex et al., 2004; Laws et al., 2005; Bagnoli et al., 2007; van Oijen et al., 2007; Harold et al., 2009; Lambert et al., 2009; Jun et al., 2010; Zhang et al., 2010; Hu et al., 2011; Chen et al., 2012; Shen et al., 2014).

We also found two regions of high LD, one in the IL-6 gene and another in the IL-10 gene. These results and the HWE deviations presented in both genes, suggest that future genetic association studies focused on those regions should be done using a larger sample size in order to clarify a possible association with AD in the Mexican population as was observed principally in Caucasian populations (Bonafè et al., 2001; Brull et al., 2001; Bagnoli et al., 2007; Combarros et al., 2009; Chen et al., 2012).

Most inconsistencies in genetic association studies, are probably due to ethnic differences among analyzed populations (Dixon et al., 2011), because populations with differences in genetic ancestry should not be extrapolated. In order to validate our results, we focused in the stratification of our samples, which is a confounding factor that could partly explain the relative lack of replication across populations (Thomas and Witte, 2002). To avoid the effects of stratification we estimate the ancestry proportion in each study group and all the groups were included with age and gender in the regression model for odds ratio calculation in SNPStats. The inclusion of the adjustment for ancestry in the association testing ensures that the contribution of each ancestral population is homogeneous between AD patients and controls and it does not play a confounding role in the analysis (Epstein et al., 2007). The ancestry proportion estimates obtained in this study are comparable with previous reports in the Mexican population, observed by Juárez-Cedillo et al. (2008) using 15 short tandem repeats (STRs), and by Rangel-Villalobos et al. (2009), using 12 Y-chromosome STRs. Therefore these results corroborate the validity of the AIMs selected.

It is important to mention that this is the first study in Mexican population that considers the analysis of ancestry in AD patients in which it was possible to find that the presence of the GG genotype in the SNP rs20417 displayed an association with AD patients in our population. We consider the association of rs20417 genotype GG was adequate based on confidence interval (1.47–6.11). In the same way after the correction by ancestry, another polymorphism (rs1130864) showed a significant p value (0.0008), however this association is questionable because the IC = 0.77–2.51. Although this could be a pilot study, it is essential to increase the sample size to support statistical power. Also, it is important to conduct more extensive studies involving genes related to the COX2 metabolism. This will establish possible mechanisms involved in the pathological process of the disease.

# Conclusions

In summary, this is the first study in Mexican population that considers the analysis of ancestry in AD patients. Our results showed that the rs20417 SNP in the PTGS2 gene presented differences in its allele and genotype frequencies pointing at the G allele and the GG genotype, and it could be considered as a risk factor for AD in Mexican patients. These results manifest the relevance of the role that COX-2 is playing in the pathological mechanisms that result in the development of AD. Ancestry analysis showed that there were not significant differences in AD patients and controls with respect to genetic ancestry, corroborating the validity of the associations found.

# Acknowledgments

This work was supported by a grant from SEP-CONACYT no. 157548. We appreciate the support provided by Esteban Gonzalez Burchard and María del Mar Del Pino Yanes for

# References


his advice in the analysis. We appreciate the support provided by the Instituto Nacional de las Personas Adultas Mayores (INAPAM) for the recruitment of controls subjects; particularly thank Aracely Escalante-Jasso, Directora General-INAPAM; Dr. Sergio Valdes y Rojas, Director de Atención Gerontológica-INAPAM; Lic. Joel Clímaco Toledo, Director de Programas Estatales; and Rosalba Juárez Soria, Jefa del Departamento de Desarrollo Comunitario.

interaction between the genes for IL-6 and IL-10 in the risk of Alzheimer's disease. J. Neuroinflammation 6:22. doi: 10.1186/1742-2094-6-22


estimations in the Mexican Mestizo population from Mexico City using 15 STR polymorphic markers. Forensic Sci. Int. Genet. 2, e37–e39. doi: 10.1016/j.fsigen. 2007.08.017


community-based sample: the Honolulu-Asia aging study. Neurobiol. Aging 27, 211–217. doi: 10.1016/j.neurobiolaging.2005.01.013


**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.

Copyright © 2015 Toral-Rios, Franco-Bocanegra, Rosas-Carrasco, Mena-Barranco, Carvajal-García, Meraz-Ríos and Campos-Peña. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution and 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.

# Aged neuronal nitric oxide knockout mice show preserved olfactory learning in both social recognition and odor-conditioning tasks

Bronwen M. James 1,2 , Qin Li <sup>1</sup> , Lizhu Luo<sup>1</sup> and Keith M. Kendrick <sup>1</sup> \*

<sup>1</sup> Key Laboratory for NeuroInformation of Ministry of Education, Center for Information in Medicine, University of Electronic Science and Technology of China, Chengdu, Sichuan, China, <sup>2</sup> Department of Medicine, St Bernard's Hospital, Gibraltar, UK

There is evidence for both neurotoxic and neuroprotective roles of nitric oxide (NO) in the brain and changes in the expression of the neuronal isoform of NO synthase (nNOS) gene occur during aging. The current studies have investigated potential support for either a neurotoxic or neuroprotective role of NO derived from nNOS in the context of aging by comparing olfactory learning and locomotor function in young compared to old nNOS knockout (nNOS<sup>−</sup>/<sup>−</sup>) and wildtype control mice. Tasks involving social recognition and olfactory conditioning paradigms showed that old nNOS<sup>−</sup>/<sup>−</sup> animals had improved retention of learning compared to similar aged wildtype controls. Young nNOS<sup>−</sup>/<sup>−</sup> animals showed superior reversal learning to wildtypes in a conditioned learning task, although their performance was weakened with age. Interestingly, whereas young nNOS<sup>−</sup>/<sup>−</sup> animals were impaired in long term memory for social odors compared to wildtype controls, in old animals this pattern was reversed, possibly indicating beneficial compensatory changes influencing olfactory memory may occur during aging in nNOS<sup>−</sup>/<sup>−</sup> animals. Possibly such compensatory changes may have involved increased NO from other NOS isoforms since the memory deficit in young nNOS<sup>−</sup>/<sup>−</sup> animals could be rescued by the NO-donor, molsidomine. Both nNOS<sup>−</sup>/<sup>−</sup> and wildtype animals showed an age-associated decline in locomotor activity although young nNOS<sup>−</sup>/<sup>−</sup> animals were significantly more active than wildtypes, possibly due to an increased interest in novelty. Overall our findings suggest that lack of NO release via nNOS may protect animals to some extent against age-associated cognitive decline in memory tasks typically involving olfactory and hippocampal regions, but not against declines in reversal learning or locomotor activity.

Keywords: aging neuroscience, cognitive decline, nitric oxide, neurodegeneration, neuronal nitric oxide synthase gene, olfactory memory

# Introduction

Many neuronal networks are involved in the decline in cognitive functioning observed during physiological aging and there is evidence that impaired glutamatergic neurotransmission may play a central role. Binding studies have demonstrated an agerelated loss of NMDA receptors in different brain regions of rodents and primates

#### Edited by:

Rosalinda Guevara-Guzmán, Universidad Nacional Autónoma De México, Mexico

#### Reviewed by:

Ulkan Kilic, Istanbul Medipol University, Turkey Gabriela Sanchez-Andrade, Sanger Institute, UK

#### \*Correspondence:

Keith M. Kendrick, Key Laboratory for NeuroInformation of Ministry of Education, Center for Information in Medicine, University of Electronic Science and Technology of China, No.4, Section 2, North Jianshe Road, Chengdu, Sichuan 610054, China k.kendrick.uestc@gmail.com

> Received: 07 January 2015 Accepted: 09 March 2015 Published: 27 March 2015

#### Citation:

James BM, Li Q, Luo L and Kendrick KM (2015) Aged neuronal nitric oxide knockout mice show preserved olfactory learning in both social recognition and odor-conditioning tasks. Front. Cell. Neurosci. 9:105. doi: 10.3389/fncel.2015.00105 (Gonzales et al., 1991; Wenk et al., 1991; Pittaluga et al., 1993). In addition, age-related declines in the amount of mRNA for specific subunits of the NMDA receptor in cortical regions, dentate gyrus and other brain regions have also been reported (Magnusson, 2000; Magnusson et al., 2010). Functional experiments have revealed that the ability of NMDA receptors to stimulate noradrenaline release from hippocampal nerve endings significantly decreases during senescence (Pittaluga et al., 1993) and long-term potentiation (LTP) is attenuated in aged rodents (Barnes and McNaughton, 1985; Deupree et al., 1993).

Along with a role for altered glutamatergic neurotransmission in functional impairments during aging, an involvement of altered nitric oxide (NO) signaling has also been suggested. NO is a diffusible gas molecule in the brain which acts both as a regulator of neurotransmission, synaptic plasticity, neurogenesis, gene expression, neurotoxicity and neuroprotection (Guevara-Guzman et al., 1994; Kendrick et al., 1997; Dawson and Dawson, 1998; Estrada and Murillo-Carretero, 2005). There is much evidence for a dual role of NO in protecting and/or contributing to oxidative damage as a result of aging. Many reports have linked aging with an increase in NO production and its activity as a free radical (McCann et al., 1998; Calabrese et al., 2002; Floyd and Hensley, 2002). While NO is a relatively unreactive radical, it is able to form other reactive intermediates (e.g., NO reacts with the free radical superoxide (O<sup>2</sup> <sup>−</sup>) producing the powerful oxidant peroxynitrite anion (ONOO−)). These reactive intermediates can trigger nitrosative damage which in turn may lead to age-related diseases due to the structural alteration of proteins, inhibition of enzymatic activity and interference with regulatory function (Drew and Leeuwenburgh, 2002; Steinert et al., 2010). In support of this, nNOS knockout (nNOS−/−) mice are protected from neural damage in a number of models of hypoxia, ischaemia and neurotoxic damage (Morikawa et al., 1992; Kuluz et al., 1993; Ayata et al., 1997; Zaharchuk et al., 1997; Itzhak et al., 1998a,b; Shimizu-Sasamata et al., 1998).

On the other hand, there are several reports suggesting that NO may also possess neuroprotective properties, for example NOS containing neurons are particularly resistant to neurodegeneration in Huntington's Disease (Ferrante et al., 1985) and excitatory amino acid neurotoxicity (Koh et al., 1996). There is much evidence supporting the cytoprotective actions of NO (Kendrick et al., 1996; Pantazis et al., 1998; Gidday et al., 1999; Teng et al., 1999). However, whether NO is neurotoxic or neuroprotective appears to depend on a number of factors including the concentration of NO (Canals et al., 2001; Calabrese et al., 2007), the isoform involved, the type of cells in which NO is produced and the overall redox state of the environment (Lipton, 1999; Calabrese et al., 2007). Current work is divided over which of the dual roles (neuroprotector/neurotoxic) that NO can play is most important during cerebral aging.

A number of immunocytochemical studies have suggested an important role for NO during aging, however they point to a reduction in NOS containing neurons in several brain regions of the aging rat brain (Cha et al., 1998), particularly the cingulate cortex, parietal area 1, temporal areas 1, 2 and 3, the medial part of occipital cortex area 2, the monocular and binocular part of occipital cortex area 1, the entorhinal cortex, the hippocampus CA1-4, the dentate gyrus and the subiculum. In addition, the number of dendritic branches is decreased (possibly leading to a reduction in released NO) in aged rats and NOS-immunoreactive neurons also tend to become shorter (Cha et al., 1998). On the other hand another study has reported increased brain mitochondrial nNOS expression during aging in rats resulting in a dysfunctional pattern of mitochondrial protein nitration which might contribute to increased apoptosis (Lam et al., 2009). The significance of these age-related changes is not completely established however, although they strongly suggest that alterations in amount and pattern of nNOS expression in the aging brain may be of functional importance.

A reduction in the number of NADPH diaphorase reactive cells (a marker for NOS activity) has been reported in the striatum and olfactory cortex (consisting of the anterior olfactory nucleus and the piriform, the peri-amygdaloid and the entorhinal cortices) in aged rats (Necchi et al., 2002). This reduction is accompanied by a concomitant decrease in the number of nNOS positive cells and the total expression of nNOS protein revealed by immunocytochemistry and Western blotting respectively (Necchi et al., 2002).

Decline in other components of the NO signaling pathway during aging has been investigated using confocal laser scanning microscopy. Circuit-specific alterations of NMDA receptor subunit 1 have been observed in the dentate gyrus of aged monkeys and a reduced number of NADPH diaphorase positive neurones have been detected in the cerebral cortex and striatum (Yamada and Nabeshima, 1998). No age-related changes in NADPH diaphorase activity have been detected in the hippocampus but in situ hybridization studies have shown an increase in hippocampal nNOS mRNA expression (Yamada and Nabeshima, 1998).

NO is known to stimulate soluble guanylyl cyclase leading to an elevation of cGMP (cyclic guanosine 3':5'-cyclic monophosphate). Basal levels of cGMP are maintained by endogenous nitrergic tone (Vallebuona and Raiteri, 1994; Fedele et al., 1996), thus the reduction in activity of nNOS with senescence may contribute to the two-fold reduction in levels of cGMP observed in the hippocampus of rats aged 12 and 24-months old (Vallebuona and Raiteri, 1995). In addition, the activity of soluble guanylyl cyclase (sGC) demonstrates a form of reduced activity in the hippocampus during aging, since hippocampal soluble guanylate cyclase is 30% less responsive to exogenous NO in aged rats when compared to younger controls (Vallebuona and Raiteri, 1995).

The effect of aging and the NOS system has been studied behaviorally using rats in the Morris water maze (Law et al., 2002) where a deficit in spatial memory was observed in some (but not all) rats aged 28-months. In the rats exhibiting the deficit, hippocampal nNOS protein expression was greatly decreased compared to younger rats and the cognitively unimpaired aged rats although their nNOS mRNA expression was increased (Law et al., 2002). It was suggested that the changes in transcriptional activation in older animals might be a compensatory attempt by aged neurones to maintain sufficient neuronal communication and NO balance in the face of a declining NOS-containing neuron population (Law et al., 2002).

A number of studies have used nNOS−/<sup>−</sup> mice to investigate the role of NO derived specifically from nNOS in terms of neurodegeneration, neuroprotection, neural plasticity and cognitive as well as many other behavioral functions. In the first instance there is strong evidence that nNOS−/<sup>−</sup> mice, or mice treated with NOS inhibitors, are significantly protected against neurotoxic and ischaemic damage in the brain (Morikawa et al., 1992; Kuluz et al., 1993; Itzhak et al., 1998a,b; Shimizu-Sasamata et al., 1998). Thus it is possible that age-related neurodegenerative changes would be reduced in nNOS−/<sup>−</sup> leading to reduced cognitive decline. On the other hand a number of experiments in young nNOS−/<sup>−</sup> mice have found evidence for reduced hippocampal LTP (O'Dell et al., 1994) and for impairments in spatial memory (Kirchner et al., 2004; Tanda et al., 2009; Walton et al., 2013), working memory (Tanda et al., 2009; Zoubovsky et al., 2011) and contextual fear conditioning (Kelley et al., 2009). Thus, it is possible that age-associated cognitive dysfunction in nNOS−/<sup>−</sup> animals could even be increased compared to control animals, although alternatively reduced neurodegenerative changes might result in a more stable cognitive phenotype during the course of aging.

The current study has therefore investigated the significance of an altered nNOS neuronal signaling system on agerelated cognitive decline. There is substantial evidence for the involvement of the NMDA-nNOS-NO-soluble guanylate cyclase signaling cascade in synaptic plasticity associated with olfactory learning (Kendrick et al., 1997; Sanchez-Andrade et al., 2005; Sanchez-Andrade and Kendrick, 2009). NO has also been reported to influence neurogenesis in both olfactory bulb and hippocampus which are important for learning. In mice pharmacological reductions of NO impair both social recognition learning and the social transmission of food preference, although these target NO production from all three NOS isoforms (Sanchez-Andrade et al., 2005). Less is known about the effects of NO derived from nNOS per se and particularly in associative learning paradigms involving non-social olfactory cues and the hippocampus (Bunsey and Eichenbaum, 1995). Given evidence in both humans and rodents for age-associated changes in olfactory perception and memory (Doty et al., 1984; Doty, 1989; Mobley et al., 2014) the olfactory system is also a particularly appropriate model for studying interactions between aging and specific genes.

We have therefore utilized olfactory memory and associated reversal learning paradigms in young and old nNOS−/<sup>−</sup> and wildtype control mice to investigate whether lack of NO derived from nNOS might contribute to age-associated cognitive decline. We chose to use two different olfactory learning tasks; a social recognition memory task (Sánchez-Andrade and Kendrick, 2011) and also a non-social olfactory conditioning task (Brennan et al., 1998). Performance on both of these tasks is impaired by NMDA and AMPA receptor antagonists (James, 2003; Sanchez-Andrade et al., 2005) and involves both the olfactory bulb and hippocampus (Eichenbaum et al., 1989; Pittaluga et al., 1993; Bunsey and Eichenbaum, 1995; Brennan et al., 1998; Sanchez-Andrade et al., 2005; Jüch et al., 2009). There is some evidence that performance on social recognition memory declines with age in rodents (Prediger et al., 2006; Markham and Juraska, 2007), and a previous study has reported only a relatively minor impairment on social recognition memory in young nNOS−/<sup>−</sup> mice in terms of a reduced memory duration (Jüch et al., 2009). The olfactory conditioning task also allowed us additionally to investigate aging effects on reversal learning. Reversal learning is well established as being sensitive to aging (Schoenbaum et al., 2002; Brushfield et al., 2008) and primarily involves frontal cortical regions (Clark et al., 2004; Mizoguchi et al., 2010). Finally, to investigate more general effects of aging, the locomotor activity of young and old nNOS−/<sup>−</sup> mice and wildtype controls was assessed using activity boxes.

# Methods

# Animals

The nNOS−/<sup>−</sup> mice used were a line originally generated by homologous recombination on a mixed B6/129S genetic background and where the targeted deletion of the α-subunit of nNOS resulted in >95% reduction in brain nNOS catalytic activity (Huang et al., 1993). The nNOS−/<sup>−</sup> animals were originally supplied by Dr Ted Dawson (John's Hopkins University, Baltimore, USA) and bred in a specific pathogen free barrier unit at the Babraham Institute in Cambridge (UK) together with appropriate genetic background wildtype counterparts (F2 129Sv × C57BL/6J). Four different cohorts of male nNOS−/<sup>−</sup> and wildtype control animals were used in the experiments: (1) a YOUNG cohort (aged 3–5 months, n = 12 wildtype and n = 9 nNOS−/−); (2) an OLD cohort (aged 18–24 months, n = 13 wildtype and n = 11, nNOS−/−). These first two cohorts were used in both of the olfactory learning tasks and for activity monitoring. Another; and (3) cohort of YOUNG wildtype (n = 12) and nNOS−/<sup>−</sup> (n = 12) mice was used in a separate control test for learning speed in the olfactory conditioning paradigm; and (4) a final cohort of YOUNG wildtype (n = 24) and nNOS−/<sup>−</sup> (n = 24) mice were used in an experiment investigating effects of the NOdonor molsidomine on the olfactory habituation and social recognition task. Mice were housed in groups of 2–5 under temperature controlled conditions and a 12:12 h light: dark cycle (lights on 07:30). In all cases, housing and rearing conditions were tightly controlled and comparisons between nNOS−/<sup>−</sup> and wildtype animals were made using age/cohort matched controls. All animals were first raised in the specific pathogen free barrier unit and then moved to a dedicated behavioral testing unit for experimental work. All animals had food (standard REM rodent diet) and water available ad libitum (other than during a period of restricted feeding prior to the olfactory conditioning task). They were handled daily for at least 5 days before behavioral tests were carried out and at least weekly for the rest of the time. The breeding of nNOS−/<sup>−</sup> animals and all experiments were carried out in strict accordance with UK Home Office guidelines and specifically licensed under the Animals (Scientific Procedures) Act 1986.

#### Genotyping Using PCR

Tail biopsies were taken from all experimental animals postmortem. The tail tissues were placed in 500 µL of tail lysis buffer (100 mM EDTA, pH 8.0, 50 mM Tris-HCL pH 8.0, 0.5% SDS (w/v), 50 µg/ml proteinase K (Promega)) and digestions were then incubated at 55◦C for 12-h. Genotyping confirmed that the nNOS/neomycin junction was present in all nNOS−/<sup>−</sup> mice and absent in all wildtype mice. Genotyping using PCR required two independent reactions. The wildtype allele was identified with a forward primer: that annealed to part of the nNOS gene; and a reverse primer that annealed to nNOS itself, (the primer pair generated a 500 bp PCR product):


The mutant allele was detected using a separate primer pair: a forward and reverse primer both directed to the nNOS/Neomycin junction and which generated a 700 bp PCR product:


To amplify sequences relating to each genotype the same reaction conditions were set up in Stratagene thin walled tube strips. Separate were run for each primer set thus there were two separate reactions for each DNA sample. The DNA samples from the tails samples were re-suspended in 100 µl of TE buffer (10 mM Tris (pH 8.0) and 1.0 mM ethylenediamine tetra-acetic acid). Once the PCR reaction was complete a 1.5% agarose gel was made by dissolving agarose (iGI or GIBCO BRL) in 1 × buffer containing 45 mM Tris/Borate, 2 M EDTA pH 8.0 and heating in a microwave until molten. Ethidium Bromide was added to a final concentration of 0.5 µg/ml mixed by swirling and left to cool. Gels were poured when the temperature of the molten agarose had fallen below 60◦C and left to solidify. Gels were submerged in 1 × TBE buffer and the samples were loaded along with 10 µl dye per well and a DNA ladder and then were run at 100 V/cm. The nucleic acids were then visualized under UV illumination.

# Olfactory Habituation and Social Recognition Memory

In this task YOUNG and OLD nNOS−/<sup>−</sup> and wildtype animals were repeatedly exposed to the same adult stimulus animal (male C57/Bl6 × 129sv) to establish if they showed a progressive reduction in investigation time indicative of their both being able to detect its odor and habituate to repeated exposure to it. The adult male stimulus animals were anesthetized (Hypnorm/Hypnovel 6 mg/kg i.p.) so as to remove any influence of their behavior on the test animals. This was of particular importance in the context of the current experiments in view of previous reports of altered aggression (Nelson et al., 1995), social motivation and stress behaviors in nNOS−/<sup>−</sup> animals (Tanda et al., 2009; Walton et al., 2013). Animals were then returned to their home cages overnight before being tested for their ability to remember the same stimulus mouse compared with a novel one 24 h later, as indexed by a reduced investigation time of the familiar compared with the unfamiliar stimulus animal.

In this paradigm experimental animals were first habituated to a testing arena (A Perspex round bottomed bowl, Bioanalytical Systems, USA: diameter = 90 cm at top and 45 cm at base and height = 60 cm) for 10-min before being presented with a sedated adult stimulus mouse for 1-min. After a further 10-min, the test mice were presented with the same stimulus mouse for another 1-min trial. The same procedure occurred for two more trials making a total of four trials with the same stimulus animal. Ten minutes after the fourth trial the test animal was returned to its home cage overnight. In between trials the sedated stimulus mouse was replaced into a single cage on a warming plate. At the end of the day, and after a full recovery, the sedated mouse was returned to its home cage.

The following day the test mouse was habituated to the arena for 10 min before being presented with two sedated stimulus mice for a test time of 2-min: one being the familiar mouse presented the previous day on trials 1–4 and the other a novel mouse (not used as a stimulus mouse in any previous trial). The time spent investigating each mouse was recorded using a handheld stopwatch. The behaviors were all videotaped to allow blind scoring of the investigation times.

A further experiment using this paradigm was carried out on the fourth cohort of young (3–5 months old) nNOS−/<sup>−</sup> (n = 24) and wildtype (n = 24) animals. Here half the animals in each group (i.e., n = 12) were either given the NO donor molsidomine (10 mg/kg, i.p. Sigma UK, dissolved in saline) or saline (1 ml/kg, i.p.) as a control. Molsidomine is a prodrug, and upon metabolism releases NO and an active metabolite (3-morpholinosydonimine, SIN-1), used clinically in the treatment of angina pectoris (Rosenkranz et al., 1996). Animals were injected 30-min prior to the first exposure to the stimulus mouse in the habituation trials on day 1. The timing and dose of molsidomine given were based on results from several previous studies demonstrating its ability to reverse effects of NOS inhibitors in object recognition memory tasks (Meyer et al., 1998; Pitsikas et al., 2002, 2003a). All mice were given saline injections (1 ml/kg, i.p.) for 3 days prior to commencement of testing in order to habituate them to any effects of injection stress.

#### Olfactory Conditioning Paradigm

An olfactory conditioning procedure was used to promote recognition memory for artificial odors. For this, lemon or peppermint food essence was combined with a sugar reward similar to the paradigm reported by Brennan et al. (1998). Mice in both the YOUNG and OLD groups were initially group housed and handled daily. They were then singly housed for 4 days in a procedure room on a 12 h light/dark cycle (lights on at 07:00) with free access to food and water. Animals were handled and weighed each day so that a stable baseline weight could be calculated. They were then placed on a restricted feeding regime in which 2 g of standard rodent chow was given each day at 17:00 in order to maintain their body weight at 85% free-feeding weight. The training began 3 days after the initiation of the food restriction.

Half the mice in each group were trained to associate the sugar reward with lemon odor (conditioned odor, CS+; SuperCook, Sherburn-in-Elmet, Leeds, U.K.). Peppermint odor (SuperCook, Sherburn-in-Elmet, Leeds, U.K.) served as the nonconditioned stimulus (CS−) presented in the absence of sugar. For the other half of the mice, peppermint was presented as the CS+ and lemon as the CS−. The odors were presented by sprinkling 50 µl of the food essence over clean sawdust in a plastic petri dish (Sterilin, 10 cm diameter). In the dish containing the conditioned odor, small fragments of sugar were placed beneath the sawdust during training.

The conditioning trials took place in cages identical to the home cage but with no sawdust/food/water. A trial consisted of placement of a dish containing either CS+ or CS− in the cage for a period of 10-min. At the end of each trial, the mice were returned to their home cages and the dishes washed thoroughly in hot water and Hibiscrub/70% alcohol. All mice received two training sessions per day over 2–3 days with two CS+ and two CS− trials per session (i.e., 16–24 trials in total). Individual training sessions were given 2 h apart and the order of the trials was chosen pseudo-randomly over the eight daily trials with a maximum of three CS+ or CS− trials permitted in a row.

To determine whether the mice had learned to differentiate between the CS+ and CS− odors a preference test was conducted on the day following the final conditioning session using two black Perspex compartments (height: 30 cm × width: 30 cm × depth: 30 cm) linked by a connecting passage (height: 30 cm × width: 10 cm × depth: 10 cm). Dishes containing sawdust were placed in each chamber. Prior to preference testing the mice were habituated to the apparatus for 5 min by placing them in the central connecting passage and allowing them to investigate freely. The dishes were then removed briefly in order to add 50 µl of lemon odor to one dish and 50 µl peppermint to the other in the absence of sugar. The mouse was replaced in the central section and its behavior recorded for 5-min by an observer using a hand held stopwatch. The total time spent in each compartment and the amount of time digging in each dish was recorded. In order to remove any odor trails the apparatus was thoroughly cleaned and wiped dry using 2% acetic acid between training sessions and between testing.

The paradigm was extended further to establish an index of reversal learning; such that the following day mice were re-trained using reversed contingencies (i.e., 1 day of reversal training with 4 trials, 2 × new CS+ and 2 × new CS−) and were then re-tested the subsequent day. Using this approach we could establish whether animals persevered with their original learning (i.e., continued to show a significant preference for the original CS+ odor), or showed extinction of the original learning indicating progress towards a reversal (i.e., showed no significant preference for the new CS+ compared to the new CS− odor), or showed a complete reversal (i.e., showed a significant preference for the new CS+ odor).

#### Locomotor Activity Tests

Locomotor activity was assessed using a battery of clear Perspex photo-beam boxes (width = 21 cm × height = 20 cm × depth = 36 cm; Technix, Babraham Institute, U.K). These contained two transverse infrared beams 10 mm from the base spaced equally along the length of the box. A computer recorded beam breaks and runs (where the front and rear beams are broken in close succession) in 5 min time bins. Animals were placed in the boxes (which were cleaned between animals using 2% acetic acid) for 2 h on three consecutive days, always at the same time of day (each testing session began at the same time each day between 08:00 and 16:00). This design allowed assessments of novelty reactivity (including sensitivity and habituation) as well as basic locomotor competence.

# Statistics

For the olfactory habituation and social recognition task investigation times of the four successive habituation trials were compared using a 2-way repeated measures ANOVA, with factors (GENOTYPE (wildtype and nNOS−/−) and TRIAL (1–4)). The investigation times for the 24 h memory test (familiar compared with unfamiliar) were compared within groups using a paired t-test. For the experiment investigating effects of molsidomine in YOUNG animals a 2-way ANOVA was carried out with factors (GENOTYPE (wildtype or nNOS−/<sup>−</sup> and TREATMENT (molsidomine or saline)) for the habituation phase and factors (TREATMENT and FAMILIARITY (familiar and unfamiliar stimulus animals)) for the 24 h social recognition memory phase. For the olfactory conditioning experiment compartment preference and digging time data were subjected to 2-way ANOVAs with factors (GENOTYPE × ODOR (CS+ or CS−)). In order to compare the effect of age and genotype directly a preference index for both the compartment and digging preference data was calculated: (Compartment preference index = time spent in the compartment containing the CS+ minus time spent in compartment containing CS−)/(time spent in CS+ compartment plus time spent in CS− compartment; Digging preference index = calculated the same way, but using time spent digging in CS+ and CS− odor dishes). To investigate aging effects a 2-way ANOVA was then performed with factors AGE and GENOTYPE. In order to analyze the reversal data the preference indices were compared using an unpaired t-test. For the locomotor activity experiments data was collected and analyzed by comparing the total number of beam breaks and runs between groups using a 2-way ANOVA with factors (TIME (5 min time bins) and SESSION (Days 1–3)). To test for age differences a 2-way ANOVA with factors (GENOTYPE × AGE) and post hoc Tukey test was performed using mean activity and beam break data.

# Results

# Olfactory Habituation and Social Recognition Memory in Young and Old nNOS−/<sup>−</sup> and Wildtype Animals

For the YOUNG groups a two way ANOVA with GENOTYPE and TRIAL as factors revealed no significant effect of GENOTYPE (F1,80 = 1.97, P = 0.164) and a significant effect of TRIAL (F3,80 = 9.97, P < 0.001), but no GENOTYPE × TRIAL interaction (P > 0.1). This indicates that overall both nNOS and wildtype animals displayed equivalent amounts of investigation and a similar reduction due to habituation across trials. However, an exploratory analysis using a t-test did show that nNOS−/<sup>−</sup> mice spent significantly more time investigating the stimulus animal on Trial 1 (P < 0.05—see **Figure 1**). YOUNG nNOS−/<sup>−</sup> mice were impaired in their ability to discriminate between the familiar and unfamiliar stimulus mice at the 24 h test (t<sup>11</sup> = 0.2876, P = 0.7790) whereas the wildtype controls showed a clear discrimination spending significantly more time investigating the unfamiliar stimulus animal (t<sup>9</sup> = 2.689, P = 0.0248).

A similar analysis of the OLD groups also revealed a significant main effect of TRIAL (F3,76 = 5.92; P = 0.001) indicating that both OLD nNOS−/<sup>−</sup> and wildtype groups of mice habituated similarly to the presence of the stimulus mouse by showing reduced investigation times over trials. This suggests that their ability to detect olfactory cues from the stimulus animal was unimpaired. However, there was also a main effect of GENOTYPE (F1,76 = 18.15, P < 0.001) due to wildtype control mice spending significantly longer investigating the stimulus animal. While there was no GENOTYPE × TRIAL interaction (F3,76 = 0.85, P = 0.471), exploratory t-tests revealed significant differences between investigation times in the two groups in trials 1, 2 and 3 (see **Figure 1**). While OLD wildtype mice did not show a significant discrimination between familiar and unfamiliar stimulus mice at the 24 h test (P > 0.1), nNOS−/<sup>−</sup> mice did (P < 0.05—see **Figure 1**).

# Effects of Molsidomine Treatment on Olfactory Habituation and Social Recognition Memory in Young Animals

In order to test whether memory deficits in YOUNG nNOS animals in the social recognition task were due to a lack of NO produced by other NOS isoforms we performed an additional experiment where we tried to rescue their deficit by treatment with the NO-donor molsidomine. A two-factor ANOVA showed there was no effect of molsidomine on the behavior of the YOUNG wildtype or nNOS−/<sup>−</sup> mice during the habituation phase (TREATMENT: F1,188 = 1.97, P = 0.163; GENOTYPE: F1,188 = 2.63, P = 0.106; TREATMENT × GENOTYPE interaction: (F1,188 = 0.004, P = 0.991). Molsidomine did not appear to influence the presence of a normal recognition memory at 24-h (evidenced by increased investigation time of the novel stimulus animal) in wildtype control mice (saline treated: t<sup>11</sup> = 2.352, P = 0.0384; molsidomine treated: t<sup>11</sup> = 2.827, P = 0.0165). However, while in saline treated nNOS−/<sup>−</sup> mice there was no evidence for recognition memory at 24 h (t<sup>11</sup> = 1.45, P = 0.1739) in molsidomine treated animals there was (t<sup>11</sup> = 3.388, P = 0.0061) (**Figure 2**).

A further analysis of treatment effects in the two genotypes (wildtype and nNOS−/−) was performed using 2-way ANOVAs

FIGURE 1 | Performance of YOUNG (A) and OLD (B) nNOS and wildtype mice in the social recognition task. Graphs show mean ± sem habituation of olfactory investigation times across 4 × 1 min trials separated by 10 min where test animals are exposed to an anesthetized adult stimulus. Histograms show mean ± sem investigation times of a familiar vs. unfamiliar stimulus mouse (presented together) in a 2 min test given 24 h after the habituation trials.

Both YOUNG and OLD mice from the two groups show significant habituation of investigation across the 4 trials (see text) but only YOUNG wildtype and OLD nNOS−/<sup>−</sup> animals show significantly greater investigation time for the unfamiliar vs. familiar stimulus mice indicative of the formation of a long term recognition memory. \*P < 0.05 two-tailed t-test unfamiliar vs. familiar; ##P < 0.01, #P < 0.05 two-tailed t-tests wildtype vs. nNOS−/−.

with factors FAMILIARITY (i.e., unfamiliar and familiar stimulus animals) and TREATMENT. In wildtype animals there was a significant effect of FAMILIARITY (F1,44 = 21.60, P < 0.001) but no effect of TREATMENT (F1,44 = 0.15, P = 0.697) or FAMILIARITY × TREATMENT interaction (F1,44 = 0.22, P = 0.642). Thus saline and molsidomine treated animals showed a similar degree of preference for investigating the novel stimulus animals. In nNOS−/<sup>−</sup> mice on the other hand while there were no main effects of FAMILIARITY or TREATMENT there was a significant FAMILIARITY × TREATMENT interaction (F1,44 = 4.77, P = 0.034) showing that molsidomine treatment produced a significant change compared to saline and rescued the recognition memory deficit.

# Olfactory Conditioning and Reversal Learning in Young and Old nNOS−/<sup>−</sup> and Wildtype Animals

Analyses of the data within YOUNG groups of animals revealed that both nNOS−/<sup>−</sup> and wildtype control mice spent significantly longer in the compartment containing the CS+ odor (nNOS−/−: t<sup>10</sup> = 4.844, P = 0.0007; wildtype: t<sup>12</sup> = 2.589, P = 0.0237—see **Figure 3**). This result was paralleled in the digging preference data which revealed that both groups spent significantly longer digging in the dish containing the CS+ odor (nNOS−/−: t<sup>7</sup> = 3.441, P = 0.0108; wildtype: t<sup>8</sup> = 4.477, P = 0.0021). When a 2-way ANOVA was performed with factors GENOTYPE and ODOR, both nNOS−/<sup>−</sup> and wildtype controls showed a significant preference for the compartment containing the conditioned odor (ODOR: F1,44 = 47.71, P < 0.001), and no differences were observed between groups in the amount of time spent in each compartment (GROUP: F1,44 = 0.45, P = 0.501). There was a marginal GENOTYPE × ODOR interaction (F1,44 = 3.95, P = 0.053). The digging preference data revealed that both groups spent significantly more time digging in the dish containing the CS+ odor (ODOR: F1,30 = 18.86, P < 0.001). However, a significant effect of GENOTYPE was also observed

(F1,30 = 4.31, P = 0.047) and a GENOTYPE × ODOR interaction (F1,30 = 5.12, P = 0.031) showing that the nNOS−/<sup>−</sup> mice spent significantly more time digging in the dish containing

mice and also the amount of time spent digging in the shavings with the

the CS+ odor. In order to control for the possibility that the YOUNG nNOS−/<sup>−</sup> mice might be learning the conditioning task more quickly than wildtype controls (resulting in an apparent advantage in reversal learning after 1 day of training), two other cohorts of YOUNG mice naïve to the test (n = 12 nNOS−/<sup>−</sup> and n = 12 wildtype) were trained on the original task for only 1 day (i.e., 4 trials) and then tested on the subsequent day. Neither the wildtype controls nor the nNOS−/<sup>−</sup> mice showed a preference for the compartment containing the conditioned odor after a single day of training (wildtype: t<sup>11</sup> = 0.3530, P = 0.7307 and nNOS−/−: t<sup>11</sup> = 1.384, P = 0.1938) and further analyses using a 2-way ANOVA revealed no significant effect of GENOTYPE (F1,44 = 0.21, P = 0.648) or ODOR (F1,44 = 2.65, P = 0.111). There was also no GENOTYPE × ODOR interaction (F1,44 = 0.77, P = 0.386). However, the digging preference data did reveal a significant effect of ODOR (F1,42 = 15.38, P < 0.001), although both wildtype and nNOS−/<sup>−</sup> mice showed a significant preference for the conditioned odor. There was

CS+ and CS− contingency and also following reversal learning (r). There is no reversal learning data for old wildtype animals since they did not learn the original contingency. \*P < 0.05, \*\*P < 0.01, #P < 0.05 one-tailed CS+ vs. CS− (t-test).

no significant effect of GENOTYPE (F1,42 = 0.62, P = 0.434) or GENOTYPE × ODOR interaction (F1,42 = 0.19, P = 0.666). This control experiment therefore provided no evidence that YOUNG nNOS−/<sup>−</sup> mice could learn the task better than wildtypes after 1 day of training and so the advantages we found for nNOS−/<sup>−</sup> animals in reversal learning are unlikely to have been due to a superior learning speed to that in wildtype controls.

Analyses of the compartment preference data using a 2-way ANOVA in the OLD animals revealed no significant effect of GENOTYPE on the amount of time spent in each compartment (F1,38 = 0.03, P = 0.861) but a significant effect of ODOR (F1,38 = 11.68, P = 0.002). There was also a significant GENOTYPE × ODOR interaction (F1,38 = 12.88, P < 0.001) confirming that the nNOS−/<sup>−</sup> mice demonstrated a significant preference for the CS+ containing compartment whereas the wildtype controls did not. Analyses of the digging preference data revealed significant effects of GENOTYPE (F1,38 = 13.19, P < 0.001) and ODOR (F1,38 = 15.07, P < 0.001) and a GENOTYPE × ODOR interaction (F1,38 = 11.53, P = 0.002), which supported the compartment preference finding (see **Figure 3**).

In order to test for an interaction with age and genotype a 2-way ANOVA was performed with factors AGE and GENOTYPE using the preference indices. For the compartment preference data the main effect of AGE was marginally significant (F1,41 = 3.28, P = 0.078) and a significant main effect of GENOTYPE was observed (F1,41 = 7.50, P = 0.009). The lack of an age effect could be related to the fact that the mice tended to sit in the connecting channel between the two compartments. No AGE × GENOTYPE interaction was found (F1,41 = 0.90, P = 0.349). When the digging preference data were analyzed a significant effect of both AGE (F1,34 = 6.23, P = 0.018) and GENOTYPE (F1,34 = 7.17, P = 0.011) were observed but there was no AGE × GENOTYPE interaction (F1,34 = 0.11, P = 0.745).

When the contingencies were reversed, (i.e., mice previously conditioned to associate lemon odor with a sugar reward were then trained to associate peppermint with the sugar reward or vice versa), the YOUNG wildtype mice spent significantly more time in the compartment containing the odor to which they were conditioned originally (t<sup>12</sup> = 2.540, P = 0.0260); i.e., they persevered with the original learning. However, the YOUNG nNOS−/<sup>−</sup> mice showed a clear preference for the compartment containing the odor to which they had been most recently trained (t<sup>9</sup> = 2.674, P = 0.0254; i.e., they showed reversal of original learning). The digging preference data revealed that nNOS−/<sup>−</sup> mice appeared to spend more time digging in the dish containing the new conditioned odor, however this did not quite achieve significance (t<sup>9</sup> = 1.847, P = 0.0979). The wildtype control mice spent approximately equal time digging in each dish (t<sup>11</sup> = 0.7259, P = 0.4831—see **Figure 3**).

When the OLD nNOS−/<sup>−</sup> mice were tested for learning a reversal of task contingencies, in compartment preference or digging preference indices (t<sup>8</sup> = 0.8695, P = 0.4099 and t<sup>7</sup> = 0.2128, P = 0.8375). However, analysis of the side preference index data for reversal learning using an unpaired t-test revealed a trend towards a difference between the YOUNG and OLD nNOS−/<sup>−</sup> mice (t<sup>16</sup> = 1.869, P = 0.0800). There was a similar non-significant trend for digging preference (t<sup>15</sup> = 1.727, P = 0.1048—see **Figure 3**).

# Locomotor Activity in Young and Old nNOS−/<sup>−</sup> and Wildtype Animals

YOUNG nNOS−/<sup>−</sup> mice were significantly more active than their wildtype controls over the three activity sessions in terms of the total number of runs and number of beam breaks in each 2 h activity session (see **Figure 4**). Further analyses of the 5 min time bin data using a 2-way ANOVA revealed significant main effects of GENOTYPE and TIME BIN for both the run and the beam break data over the 3 days (day one run data: effect of GENOTYPE F1,456 = 35.47, P < 0.001, and TIME BIN F23,456 = 20.51, P < 0.001). The increased activity levels were apparent during the first 25–30 min of each test session, after which there appeared to be no difference between groups (see **Figure 5**). There was a decrease in activity, common to both groups, that occurred from test day one to test day three (effect of SESSION on total number of runs: F2,1338 = 40.79, P < 0.001).

No significant differences were observed in the activity of the OLD nNOS−/<sup>−</sup> and the wildtype control mice aged 18–24 months over the 3 days of testing, when the overall number of runs and beam breaks in each session were compared between groups (see **Figure 4**). There was however a decrease in activity, common to both groups that occurred from test day one to test day three (effect of SESSION on total number of runs: F2,1722 = 49.12, P < 0.001).

Detailed analyses of the activity data using a two-way ANOVA with factors AGE and GENOTYPE revealed a significant effect of age on day one with the older mice making fewer numbers of runs and beam breaks in each session than their younger counterparts (F1,92 = 6.71; P = 0.011—see **Figure 4**). This was not evident on days two and three. In addition, there was a significant effect of GENOTYPE with the nNOS−/<sup>−</sup> mice making more runs and beam breaks in each session compared to their wildtype controls (F1,92 = 5.97, P = 0.016). However, there was no significant interaction between AGE and GENOTYPE (F1,92 = 1.94; P = 0.167) indicating that the decline

in activity observed was not exclusive to either the nNOS−/<sup>−</sup> or wildtype mice.

# Discussion

The current studies using nNOS−/<sup>−</sup> mice have provided support for the hypothesis that NO derived from nNOS may contribute to age-associated neurodegenerative changes that lead to cognitive decline using two different olfactory learning paradigms. However, there appeared to be no evidence for a similar protection against age-associated decline in reversal learning or locomotor activity. Tasks involving social recognition and olfactory conditioning paradigms showed that old aged (18–24 months) nNOS−/<sup>−</sup> animals showed improved retention of learning compared to similar aged wildtype controls. Young (3–5 months) nNOS−/<sup>−</sup> animals showed superior reversal learning to wildtypes in the conditioned learning task, the later showing perseveration of learning (i.e., they failed to learn the reversal contingency). On the other hand old aged nNOS−/<sup>−</sup> animals failed to show complete reversal learning during the same 1-day time-period. Whereas young nNOS animals were impaired in long term memory for social odors compared to wildtype controls, in old aged animals this pattern was reversed, possibly indicating beneficial compensatory changes influencing olfactory memory may occur during aging in nNOS animals. Both nNOS−/<sup>−</sup> and wildtype animals showed an age-associated decline in locomotor activity although young nNOS−/<sup>−</sup> animals were significantly more active than wildtypes. Overall our findings suggest that lack of NO release via nNOS may protect animals to some extent against age-associated cognitive decline in memory tasks typically involving olfactory and hippocampal regions, but not against declines in reversal learning or locomotor activity.

The current study provided evidence for an olfactory social recognition memory impairment in young nNOS−/<sup>−</sup> animals since they failed to show significantly differential investigation times between familiar and unfamiliar stimulus animals after 24 h. This is in agreement with a previous study using a juvenile recognition version of this paradigm and which reported a similar memory impairments in nNOS−/<sup>−</sup> animals at 24 h although not after 6 h or less (Jüch et al., 2009). However, while old wildtype control animals lost their ability to exhibit a recognition memory after 24 h, old nNOS−/<sup>−</sup> animals actually gained this ability, having been unable to do this at a younger age. Thus in this context not only do nNOS−/<sup>−</sup> animals appear to be protected against age-associated cognitive decline but actually displayed a gain of function suggestive of positive compensatory changes over the course of aging.

Another recent study has reported that nNOS−/<sup>−</sup> animals do not show a preference for social novelty in a three chamber preference test where time spent with a socially familiar or unfamiliar stimulus animal is recorded (Walton et al., 2013). It is therefore possible that this might have contributed to the absence of recognition memory we found in young nNOS−/<sup>−</sup> animals. However, Jüch et al. (2009) reported a preference for a novel juvenile stimulus animal in a social recognition test in young nNOS−/<sup>−</sup> animals after 6 h and we also obtained similar findings after 24 h in old nNOS−/<sup>−</sup> animals. We also found similar high investigation times of novel stimulus animals during the habituation trials in young nNOS−/<sup>−</sup> animals indicating a similar heightened interest in investigating them as in wildtype control animals. Differences in findings between our study as well as that of Jüch et al. (2009) and those reported by Walton et al. (2013) may therefore have reflected the different paradigms used. In the Walton et al., study the test animal was exposed continuously to the ''familiar'' stimulus animal for 10 min, whereas in our study only for 4 × 1 min periods. Also, during the preference test familiar and novel animals were located in separate compartments in the Walton et al. study whereas in ours and the Jüch et al. study they were presented together to the test animal.

Our finding that the NO donor molsidomine could rescue the social recognition memory deficit in young nNOS−/<sup>−</sup> mice suggests that the deficit seen in these animals may be due to insufficient NO being released from other NOS isoforms in the brain. This might therefore imply that in older nNOS animals some compensatory changes had occurred to increase NO production from other isoforms resulting in an apparent gain of function with age. Indeed, molsidomine treatment has been reported to reduce age-associated decline in passive avoidance and object recognition learning in rats (Pitsikas et al., 2005). Clearly in view of our present results it would be of interest to investigate whether molsidomine would reverse the age-related decline in olfactory learning. There are two different stages of protein synthesis upon which consolidation of memories at intermediate and long-term time points depend, with the latter starting at around 6 h post initial learning (Crow et al., 2003; Richter et al., 2005; Wanisch et al., 2008; Engelmann, 2009). A number of differences in expression of proteins involved in the regulation of mRNA trafficking, stability and translation have been reported in both the hippocampus (Kirchner et al., 2004) and olfactory bulb (Jüch et al., 2009) of young nNOS−/<sup>−</sup> mice which may be associated with late stage memory consolidation impairments in these, or other proteins, which underlie improved performance in old nNOS−/<sup>−</sup> animals. These different compensation possibilities clearly require further investigation.

We also found a clear age-related impairment in learning of the non-social olfactory conditioning task we used. In contrast to younger mice, wildtype control mice aged 18–24 months did not show a preference for the CS+ odor, whereas both young and old nNOS−/<sup>−</sup> mice did. The age-associated impairment in this task is supported by previous studies reporting similar findings using a simpler version of the paradigm (Frick et al., 2000) and other odor-reward associations (Roman et al., 1996). As with the social recognition memory paradigm learning performance on this non-social task is likely to involve plasticity changes in both olfactory bulb and hippocampus (Bunsey and Eichenbaum, 1995; Brennan et al., 1998). Thus in two rather different olfactory memory tasks we have found clear evidence for maintained or even improved performance in old aged nNOS−/<sup>−</sup> mice whereas wildtype control animals show a marked age-associated decline.

Although there is evidence for age-associated decline in olfactory sensitivity in humans, and some reports of this in rodents (Doty et al., 1984; Doty, 1989; Mobley et al., 2014), the present study did not find clear evidence for this in either wildtype of nNOS−/<sup>−</sup> mice. While the old wildtype mice did show impaired learning of both social and non-social odors, they displayed normal habituation to social odors across repeated trials in the social recognition task. However, they did show longer overall investigation times of the stimulus animals which might have been as a result of compensation for impaired sensitivity, although it could equally well have been as a result of a greater motivation for social interaction compared to the old nNOS−/<sup>−</sup> animals The old wildtype mice were also clearly able to locate the buried sugar during the training phase of the olfactory conditioning task, even if they could not learn to discriminate between the CS+ and CS− odors. This maintained ability to find buried food rewards in aged mice despite impaired use of odor cues to guide choice of location where to dig is in agreement with a previous study (Frick et al., 2000). Thus overall, it is unlikely that olfactory memory deficits in the wildtype animals were due to impaired olfactory perception, although we cannot rule out the possibility that this might have partially contributed.

Our findings on reversal learning in the olfactory conditioning task showed that in young animals whereas wildtype controls persevered with the original CS+ and CS− contingency after 1 day of training (4 trials) the young nNOS−/<sup>−</sup> animals displayed a successful reversal. This could have been due to stronger learning of the original contingency in the wildtypes, although we found no evidence that they could learn this task more quickly than their young nNOS−/<sup>−</sup> counterparts. Thus it is possible that nNOS−/<sup>−</sup> mice exhibit greater behavioral flexibility in terms of adapting to new priorities. Since the old wildtype animals failed to learn to discriminate between the conditioned odors we could not test whether they had impaired reversal learning on this task, however the old nNOS−/<sup>−</sup> animals failed to show a complete reversal after 1 day of trials. Although this indicated an age-associated decline in reversal learning in nNOS−/<sup>−</sup> animals they showed equivalent investigation times for the conditioned odors, suggesting that they were in the process of learning the new contingency. Thus, although this would need to be confirmed using other reversal learning paradigms, it is possible that old nNOS−/<sup>−</sup> animals might be at least partially protected against the normal expected age-associated deficit in this behavior. There is some evidence for up-regulation of dopaminergic (D1 receptor) signaling in nNOS−/<sup>−</sup> mice (Tanda et al., 2009) and impairments in reversal learning during aging in rats are associated with altered dopaminergic signaling in orbitofrontal cortex, with behavioral deficits being reversed by local infusion of a D1 receptor agonist (Mizoguchi et al., 2010).

Although the nNOS−/<sup>−</sup> mice are protected from ageassociated decline in these olfactory learning tasks it is not possible at this stage to conclude that this is entirely due to reduced NO from nNOS without supporting pharmacological evidence. Other compensatory factors as a result of long-term physiological or behavioral differences and the impact of nNOS deletion throughout development could not be assessed in these experiments. For example nNOS−/<sup>−</sup> mice have increased expression of the GABA transporter GAT2 and of the GABA<sup>B</sup> receptor (Wultsch et al., 2007) and NO may interact with GABA<sup>B</sup> to influence recognition memory (Pitsikas et al., 2003b).

We have shown previously that learning in both the social recognition memory and olfactory conditioning paradigms used in the current study are impaired by pharmacological inhibition of all NOS isoforms using L-nitroarginine (James, 2003; Sanchez-Andrade et al., 2005). Thus, as already discussed above in relation to molsidomine effects on young nNOS−/<sup>−</sup> mice in the social recognition task, NO derived from other NOS isoforms might be responsible for protection against age-associated cognitive decline. There is evidence that NO derived from endothelial NOS (eNOS) plays a role in promoting both neural plasticity and learning, and combined nNOS/eNOS knockout animals show more severe hippocampal LTP deficits (O'Dell et al., 1994). The expression of eNOS, like nNOS, is also reduced during aging (Morley et al., 1996) and impaired spatial memory has been reported in late middle-aged (14–15 month old) eNOS knockout mice (Austin et al., 2013). Although there are no reports of compensatory increased expression of eNOS in young nNOS−/<sup>−</sup> animals, it is possible that normal ageassociated declines in eNOS might be reduced leading to preserved cognitive function in old nNOS−/<sup>−</sup> animals. Further, since the nNOS knockout used in our study targets only the α variant of the gene and leaves the β and γ splice variants intact (Eliasson et al., 1997), it is possibile that these latter splice variants are contributing to NO release and preserved cognitive function in aged animals.The increased activity we observed in young nNOS−/<sup>−</sup> animals using activity boxes supports previous findings using open field and elevated plus maze paradigms (Tanda et al., 2009; Walton et al., 2013). This increased activity in nNOS−/<sup>−</sup> animals was primarily contributed to in the first 25–30 min after they were placed in the activity boxes and as such may have been due to an enhanced response to novelty. In agreement with a previous study (Tanda et al., 2009) we also found some evidence for increased investigation times of novel animals by young, but not old, nNOS−/<sup>−</sup> animals in the

# References


social recognition task. Other studies have generally failed to show any consistent evidence for increased responses to novelty in nNOS knockouts (see Wultsch et al., 2007). Nevertheless, importantly in the context of the current experiments both old wildtype and nNOS−/<sup>−</sup> animals showed significant reductions in locomotor activity indicating that that lack of nNOS expression does not protect against age-associated declines in general motor activity.

In summary we have provided the first evidence for preserved cognitive function in aged animals with deletion of nNOS (targeting the α rather than β or γ splice variants). This suggests that NO derived from nNOS may contribute to age-associated neurodegenerative changes in olfactory and hippocampal regions involved the two different olfactory memory tasks we used. There was however still an age-associated decline in reversal learning in nNOS−/<sup>−</sup> animals demonstrating a reduced protection from neurodegenerative changes in frontal cortical regions. Age-related reductions in locomotor behavior were also not prevented. Further studies will be required to investigate the precise signaling pathways involved in maintenance of cognitive function in aged nNOS−/<sup>−</sup> mice. Since both olfactory learning tasks used in the current study are NMDA receptor dependent (James, 2003; Sanchez-Andrade et al., 2005) it is likely that ageassociated decline in performance in wildtype animals reflects impaired NMDA-NO-dependent neural plasticity, possibly as a result of reduced numbers of NMDA-receptors (Gonzales et al., 1991; Wenk et al., 1991; Pittaluga et al., 1993) or reduced efficiency of glutamate-NMDA-NO signaling triggering protein synthesis-dependent plasticity changes involved in memory consolidation (Crow et al., 2003; Richter et al., 2005; Wanisch et al., 2008; Engelmann, 2009). Thus in aged nNOS−/<sup>−</sup> animals compensatory mechanisms may have somehow prevented these changes from occurring. Additionally, as discussed above, it is possible that compensatory changes might also involve altered interactions between NO and GABAergic and dopaminergic signaling.

# Acknowledgments

BMJ was supported by a BBSRC Studentship at the Babraham Institute, Cambridge, UK.


**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.

Copyright © 2015 James, Li, Luo and Kendrick. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution and 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.

# Alternative kynurenic acid synthesis routes studied in the rat cerebellum

Tonali Blanco Ayala<sup>1</sup> , Rafael Lugo Huitrón<sup>1</sup> , Liliana Carmona Aparicio<sup>2</sup> , Daniela Ramírez Ortega<sup>1</sup> , Dinora González Esquivel <sup>1</sup> , José Pedraza Chaverrí <sup>3</sup> , Gonzalo Pérez de la Cruz <sup>4</sup> , Camilo Ríos <sup>1</sup> , Robert Schwarcz <sup>5</sup> and Verónica Pérez de la Cruz <sup>1</sup> \*

<sup>1</sup> Departamento de Neuroquímica, Instituto Nacional de Neurología y Neurocirugía Manuel Velasco Suárez, S.S.A., México D.F., Mexico, <sup>2</sup> Laboratorio de Neuroquímica, Instituto Nacional de Pediatría, S.S.A., México D.F., Mexico, <sup>3</sup> Facultad de Química, Departamento de Biología, Universidad Nacional Autónoma de México, México D.F., Mexico, <sup>4</sup> Facultad de Ciencias, Departmento de Matemáticas, Universidad Nacional Autónoma de México, México D.F., Mexico, <sup>5</sup> Maryland Psychiatric Research Center, Department of Psychiatry, University of Maryland School of Medicine, Baltimore, MD, USA

#### Edited by:

Arianna Maffei, The State University of New York at Stony Brook, Stony Brook, USA

#### Reviewed by:

De-Lai Qiu, Yanbian University, China Xin Wang, Stanford University and Howard Hughes Medical Institute, USA

#### \*Correspondence:

Verónica Pérez de la Cruz, Departamento de Neuroquímica, Instituto Nacional de Neurología y Neurocirugía Manuel Velasco Suárez, S.S.A., Insurgentes Sur 3877, México D.F. 14269, Mexico veped@yahoo.com.mx

> Received: 28 December 2014 Accepted: 24 April 2015 Published: 18 May 2015

#### Citation:

Blanco Ayala T, Lugo Huitrón R, Carmona Aparicio L, Ramírez Ortega D, González Esquivel D, Pedraza Chaverrí J, Pérez de la Cruz G, Ríos C, Schwarcz R and Pérez de la Cruz V (2015) Alternative kynurenic acid synthesis routes studied in the rat cerebellum. Front. Cell. Neurosci. 9:178. doi: 10.3389/fncel.2015.00178 Kynurenic acid (KYNA), an astrocyte-derived, endogenous antagonist of α7 nicotinic acetylcholine and excitatory amino acid receptors, regulates glutamatergic, GABAergic, cholinergic and dopaminergic neurotransmission in several regions of the rodent brain. Synthesis of KYNA in the brain and elsewhere is generally attributed to the enzymatic conversion of L-kynurenine (L-KYN) by kynurenine aminotransferases (KATs). However, alternative routes, including KYNA formation from D-kynurenine (D-KYN) by D-amino acid oxidase (DAAO) and the direct transformation of kynurenine to KYNA by reactive oxygen species (ROS), have been demonstrated in the rat brain. Using the rat cerebellum, a region of low KAT activity and high DAAO activity, the present experiments were designed to examine KYNA production from L-KYN or D-KYN by KAT and DAAO, respectively, and to investigate the effect of ROS on KYNA synthesis. In chemical combinatorial systems, both L-KYN and D-KYN interacted directly with peroxynitrite (ONOO<sup>−</sup>) and hydroxyl radicals (OH ), resulting in the formation of KYNA. In • tissue homogenates, the non-specific KAT inhibitor aminooxyacetic acid (AOAA; 1 mM) reduced KYNA production from L-KYN and D-KYN by 85.1 ± 1.7% and 27.1 ± 4.5%, respectively. Addition of DAAO inhibitors (benzoic acid, kojic acid or 3-methylpyrazole-5-carboxylic acid; 5 µM each) attenuated KYNA formation from L-KYN and D-KYN by ∼35% and ∼66%, respectively. ONOO<sup>−</sup> (25 µM) potentiated KYNA production from both L-KYN and D-KYN, and these effects were reduced by DAAO inhibition. AOAA attenuated KYNA production from L-KYN + ONOO<sup>−</sup> but not from D-KYN + ONOO<sup>−</sup>. In vivo, extracellular KYNA levels increased rapidly after perfusion of ONOO<sup>−</sup> and, more prominently, after subsequent perfusion with L-KYN or D-KYN (100 µM). Taken together, these results suggest that different mechanisms are involved in KYNA production in the rat cerebellum, and that, specifically, DAAO and ROS can function as alternative routes for KYNA production.

#### Keywords: D-amino acid oxidase, kynurenine, microdialysis, oxidative stress, reactive oxygen species

**Abbreviations:** AOAA, Aminooxyacetic acid; D-KYN, D-Kynurenine; DAAO, D-amino acid oxidase; DMSO, Dimethylsulfoxide; D-Trp, D-tryptophan; 3-HK, 3-Hydroxykynurenine; L-KYN, L-Kynurenine; KYNA, Kynurenic acid; KP, Kynurenine pathway; KATs, Kynurenine aminotransferases; MPC, 3-Methylpyrazole-5-carboxylic acid; NMDA, N-methyl-D-aspartate; ONOO−, peroxynitrite; OH•, Hydroxyl radical; PFC, Prefrontal cortex; ROS, Reactive oxygen species.

# Introduction

In the mammalian brain, the tryptophan metabolite kynurenic acid (KYNA) functions as an endogenous antagonist of the α7 nicotinic acetylcholine receptor (α7nAChR; Hilmas et al., 2001) and the N-methyl-D-aspartate receptor (NMDAR; Kessler et al., 1989; Alkondon et al., 2011). KYNA, which is also a ligand of the G protein-coupled receptor GPR35 (Wang et al., 2006) and can activate the aryl hydrocarbon receptor (DiNatale et al., 2010), is considered a neuromodulator since fluctuations in its endogenous levels bi-directionally influence extracellular concentrations of glutamate, dopamine and γ-aminobutyric acid (GABA) levels in the rat brain (Carpenedo et al., 2001; Rassoulpour et al., 2005; Amori et al., 2009; Wu et al., 2010; Pocivavsek et al., 2011; Beggiato et al., 2014), and reductions in KYNA formation result in increased levels of extracellular acetylcholine (Zmarowski et al., 2009). Notably, increases in cerebral KYNA levels, which are seen in the aged brain (Moroni et al., 1988; Gramsbergen et al., 1992; Heyes et al., 1992; Kepplinger et al., 2005) and in several major neurological and psychiatric diseases (Baran et al., 1999; Schwarcz et al., 2001; Guidetti et al., 2004; Kepplinger et al., 2005; Sathyasaikumar et al., 2011), have been suggested to be causally related to cognitive impairments (Wonodi and Schwarcz, 2010; Pocivavsek et al., 2012, 2014).

In the brain as elsewhere, KYNA synthesis is attributed to several distinct kynurenine aminotransferases (KATs), which catalyze the irreversible transamination of L-kynurenine (L-KYN) to KYNA (Okuno et al., 1991; Guidetti et al., 2007a; Han et al., 2010). Of these enzymes, KAT II, which is preferentially contained in astrocytes (Guidetti et al., 2007b), has received most attention since it appears to be responsible for the rapid mobilization of newly produced KYNA (Schwarcz et al., 2012). However, alternative routes of KYNA production exist under physiological conditions. For example, KYNA can be formed from D-kynurenine (D-KYN) through oxidative deamination by D-amino acid oxidase (DAAO; Loh and Berg, 1971; Ishii et al., 2010), as demonstrated in the brain and in peripheral tissues of mice, rats and rabbits (Mason and Berg, 1952; Loh and Berg, 1971; Fukushima et al., 2009; Wang et al., 2012) and, recently, in human brain tissue (Pérez-de la Cruz et al., 2012). However, KATs also recognize D-KYN as a substrate and can catalyze the de novo formation of KYNA from D-KYN in the brain in vivo (Pérez-de la Cruz et al., 2012).

Neosynthesis of KYNA can also involve the transamination of L-tryptophan by tryptophan-2-oxoglutarate aminotransferase (Hardeland, 2008). Thus, the enolic form of the primary reaction product, indole-3-pyruvic acid, is highly susceptible to reactive oxygen species (ROS) and readily undergoes pyrrole ring cleavage by interaction with oxygen intermediaries. The transiently formed product then spontaneously cyclizes to generate KYNA. L-KYN, too, is easily oxidized and can be converted to KYNA in the presence of hydrogen peroxide (H2O2), a process that is substantially enhanced by horseradish peroxidase (Zsizsik and Hardeland, 2001b). In biological systems, too, KYNA formation can result from direct reactions of either indole-3-pyruvic acid or KYN with ROS. Examples include KYNA synthesis in several rat organs after incubation with indole-3-pyruvic acid under conditions that are conducive to the generation of free radicals (ascorbate/Fe/H2O2) (Politi et al., 1991), and KYNA production from L-KYN in homogenates of Lingulodinium polyedrum exposed to light and various ROS generators (Zsizsik and Hardeland, 2001a, 2002).

The present study was designed to examine the various routes of KYNA neosynthesis from L-KYN and D-KYN in parallel. Using the rat cerebellum, which was selected on the basis of its high DAAO content and relatively low KAT activity (Baran and Schwarcz, 1993; Horiike et al., 1994; Moreno et al., 1999; Verrall et al., 2007), we also compared KYNA formation in the presence or absence of ROS. Our results demonstrate that de novo KYNA formation can involve different mechanisms, and that ROS should be considered a viable alternative for KYNA production from both L-KYN and D-KYN under physiological and, possibly, pathological conditions.

# Materials and Methods

# Animals

Adult, male Wistar rats (280–320 g), obtained from the vivarium of the National Autonomous University of Mexico (Mexico City), were used for this study. The animals were housed five per cage in acrylic cages and provided with a standard commercial rat diet (Laboratory rodent diet 5001, PMI Feeds Inc., Richmond, IN, USA) and water ad libitum. All rats were housed in the same room under identical environmental conditions, i.e., temperature (25 ± 3 ◦C), humidity (50 ± 10%) and lighting (12 h light/dark cycles).

Animals were killed by decapitation, and their tissues were immediately dissected out on ice and preserved at −70◦C. All procedures with animals were carried out according to the National Institutes of Health Guide for the Care and Use of Laboratory Animals, and the local guidelines on the ethical use of animals from the Health Ministry of Mexico. All efforts were made to minimize animal suffering during the study.

#### Materials

KYNA, L-KYN, D-KYN, dimethylsulfoxide (DMSO), DLpenicillamine, diethylenetriaminepentaacetic acid (DTPA), H2O2, ethylenediaminetetraacetic acid (EDTA), 3-methylpyrazole-5-carboxylic acid (MPC), kojic acid and benzoic acid were obtained from Sigma Aldrich Company (St. Louis, MO, USA). All other chemicals were of the highest commercially available purity. Solutions were prepared using deionized water obtained from a Milli-RQ (Millipore) purifier system.

# ONOO<sup>−</sup> Synthesis

ONOO<sup>−</sup> was synthesized as previously described (Beckman et al., 1994). Five ml of an acidic solution (0.6 M HCl) of H2O<sup>2</sup> (0.7 M) were briefly mixed with 5 ml of 0.6 M KNO<sup>2</sup> in an ice bath, and the reaction was quenched with 5 ml of ice-cold 1.2 M NaOH. Residual H2O<sup>2</sup> was removed using granular MnO<sup>2</sup> pre-washed with 1.2 M NaOH, and the reaction mixture was then left overnight at −20◦C. The resulting yellow liquid layer on top of the frozen mixture was collected for the experiment immediately before use, and adjusted to a final concentration of 50 µM using Ringer solution (144 mM NaCl, 4.8 mM KCl, 1.2 mM MgSO<sup>4</sup> and 1.7 mM CaCl2, pH 7.2). Concentrations of ONOO<sup>−</sup> were determined in quartz cuvettes using a molar extinction coefficient of 302 nm = 1670 M−<sup>1</sup> cm−<sup>1</sup> (Hughes and Nicklin, 1970).

#### Chemical Combinatorial Assays

The ability of OH• to produce KYNA in combination with L-KYN or D-KYN was examined using the Fe3+-EDTA-H2O<sup>2</sup> system (Halliwell et al., 1987; Floriano-Sánchez et al., 2006). The system contained L-KYN or D-KYN (20 µM each), 0.2 mM ascorbic acid, 0.2 mM FeCl3, 0.2 mM EDTA, 1 mM H2O<sup>2</sup> and 20 mM phosphate buffer (pH 7.4) in a final volume of 500 µl. Additional tubes were incubated in the presence of 10% DMSO to evaluate the effect of an OH• scavenger on KYNA production. Samples were incubated for 15 min at room temperature. After incubation, KYNA production was quantified by high performance liquid chromatography (HPLC; see below).

Interactions between the two KYN enantiomers and ONOO<sup>−</sup> were determined using ONOO<sup>−</sup> synthetized in our laboratory (Lugo-Huitrón et al., 2011a). Briefly, the reaction mixture (in a final volume of 500 µl in HPLC grade water) consisted of L-KYN or D-KYN (20 µM each) and 25 µM ONOO−. In separate tubes, the ONOO<sup>−</sup> scavenger DL-penicillamine (300 µM) (Floriano-Sánchez et al., 2006); was added to evaluate its effect on KYNA formation. After 15 min of incubation at room temperature, KYNA levels were determined by HPLC.

#### In vitro Studies with Tissue

Cerebella were dissected out and immediately weighed and frozen on dry ice. Tissues were then homogenized (1:10, w/v) in Krebs buffer (118.5 mM NaCl, 4.75 mM KCl, 1.77 mM CaCl2, 1.18 mM MgSO4, 12.9 mM NaH2PO4, 3 mM Na2HPO<sup>4</sup> and 5 mM glucose; pH 7.4). In order to evaluate KYNA production by ONOO−, 80 µl of the tissue homogenate were incubated for 2 h at 37◦C in the presence of DAAO inhibitors (MPC, benzoic acid or kojic acid) or AOAA. L-KYN or D-KYN (100 µM) were added to the tissue homogenate, and each inhibitor (final concentration: 1 mM) was assessed in the presence or absence of ONOO<sup>−</sup> (25 µM) in a final volume of 200 µl. After incubation, samples were centrifuged for 10 min at 6,000 × g, and the supernatants were diluted 1:5 (v/v) for KYNA determination.

#### Microdialysis

Rats were anesthetized with a mixture of ketamine (80 mg/kg) and xylazine (100 mg/kg) (i.p.) and placed in a stereotaxic frame. A guide cannula was positioned and secured to the skull with stainless steel screws and acrylic dental cement at the following coordinates: AP: 11.0 mm posterior to bregma, L: ±2.0 mm from the midline, V: 4.0 mm below the dura. Three days later, a microdialysis probe (MD-220, membrane length: 2 mm; BASi, West Lafayette, IN, USA) was inserted through the guide cannula to protrude into the cerebellar cortex, and connected to a microperfusion pump set at a speed of 2 µl/min. Microdialysis samples were collected every 30 min. A stable baseline was first established by perfusing Ringer solution (pH 7.4) for 2 h. Production of KYNA from either L-KYN or D-KYN was then assessed by perfusing the bioprecursors, diluted in Ringer solution, for 2 h. The effect of ONOO<sup>−</sup> was examined by perfusing the compound for 30 min prior to the administration of either L-KYN or D-KYN. After the discontinuation of the experimental interventions, Ringer solution was perfused for an additional 4 h. Animals were then killed by decapitation, and the cerebellum was dissected to confirm the proper placement of the microdialysis probe. Microdialysate samples were diluted as needed and then analyzed directly by HPLC. Data were not corrected for recovery from the microdialysis probe.

#### KYNA Analysis

KYNA was measured by HPLC with fluorometric detection. Briefly, 20 µl of the sample (after either in vitro incubations or perfusions in vivo) were injected onto a 3-µm C<sup>18</sup> reverse phase column (80 × 4.6 mm; ESA, Chelmsford, MA, USA), and KYNA was isocratically eluted using a mobile phase containing 250 mM of zinc acetate, 50 mM sodium acetate and 3% of acetonitrile (pH adjusted to 6.2 with glacial acetic acid) at a flow rate of 1 ml/min. KYNA was detected fluorimetrically (excitation wavelength: 344 nm, emission wavelength: 398 nm, S200 fluorescence detector; Perkin-Elmer, Waltham, MA, USA). The retention time of KYNA was ∼7 min.

#### Data Analysis

One-way analysis of variance (ANOVA) followed by Bonferroni's post hoc test was used to analyze the effect of the different treatments used in vitro. In microdialysis experiments, the average of five samples collected immediately prior to the administration of test compounds was defined as the baseline value (100%). The effect of either L-KYN or D-KYN, alone or after in vivo pretreatment with ONOO−, was analyzed by two-way ANOVA with repeated measures followed by post hoc tests: a Student's t-test was used to compare two treatments at a specific time point, and a paired Student's t-test was used when comparing the effect of a treatment at two specific times. In all cases, a P-value <0.05 was considered significant.

# Results

# KYNA Production by Interaction of L-KYN or D-KYN with OH• and ONOO<sup>−</sup> in Synthetic Systems

Co-incubation of L-KYN or D-KYN (20 µM each) with OH• and ONOO−, respectively, resulted in the formation of KYNA in vitro. As illustrated in **Figure 1**, incubation for 15 min led to the de novo production of KYNA. KYNA was undetectable in the control solutions containing the enantiomers alone, and no KYNA was measurable when OH• was incubated on its own (not shown). Addition of the OH• scavenger DMSO (10%),

which did not contain measurable amounts of KYNA itself, reduced KYNA production from OH• + L-KYN or OH• + D-KYN by 65.8 ± 2.3% and 80.7 ± 1.0%, respectively (**Figure 1A**). KYNA production was substantially more pronounced following the co-incubation of either L-KYN or D-KYN with ONOO<sup>−</sup> (25 µM). No KYNA was detectable when ONOO<sup>−</sup> was incubated on its own (not shown). Addition of the ONOO<sup>−</sup> scavenger DL-penicillamine (300 µM), which did not contain KYNA on its own, decreased KYNA production from either of the two enantiomers by 78.5 ± 0.7% and 80.1 ± 0.8% for L-KYN and D-KYN, respectively (**Figure 1B**).

# KAT and DAAO Inhibitors Attenuate KYNA Production from L-KYN and D-KYN in Tissue Homogenate

Incubation of tissue homogenate with L-KYN (**Figure 2A**) and D-KYN (**Figure 2B**), respectively, resulted in the de novo production of KYNA. The KYNA levels recovered after 2 h incubation with 100 µM of the enantiomers were 18.1 ± 2.9 fold and 9.8 ± 0.7-fold higher, respectively, than basal levels. Incubation in the presence of the KAT inhibitor AOAA (1 mM) reduced KYNA formation from L-KYN and D-KYN by 85.1 ± 1.7% and 27.1 ± 4.5%, respectively, suggesting differences in the mechanisms by which the two enantiomers are converted to KYNA in the cerebellum. In line with this conclusion, the DAAO inhibitors kojic acid, benzoic acid and MPC (all at 1 mM) reduced the production of KYNA from D-KYN by ∼66% (**Figure 2B**) but did not inhibit KYNA formation from L-KYN (**Figure 2A**).

# ONOO<sup>−</sup> Potentiates KYNA Production from L-KYN and D-KYN in Tissue Homogenate: Attenuation by KAT and DAAO Inhibitors

The addition of ONOO<sup>−</sup> (25 µM) to tissue homogenate increased KYNA production from both L-KYN and D-KYN (each 100 µM) 2.6 ± 0.3 and 2.8 ± 0.3 times, respectively (**Figure 3**). Under these conditions, the presence of AOAA (1 mM) attenuated the ONOO−-induced potentiation of KYNA formation from L-KYN (**Figure 3A**) but not from D-KYN (**Figure 3B**). DAAO inhibitors (all at 1 mM) decreased KYNA production induced by the co-incubation of both L-KYN and D-KYN with ONOO−, though the effect of MCP vs. L-KYN + ONOO<sup>−</sup> did not reach statistical significance. Kojic acid and benzoic acid were particularly effective, reducing the total KYNA generated by the combination of D-KYN + ONOO<sup>−</sup> by 86.1 ± 2.0% and 75.6 ± 2.7%, respectively.

# Effect of L-KYN and D-KYN on Extracellular KYNA in vivo

We next designed microdialysis experiments to investigate the conversion of L-KYN or D-KYN to KYNA in the rat cerebellum in vivo. The enantiomers were infused for 2 h by reverse dialysis, and the content of KYNA was monitored in microdialysate samples for an additional 4 h. Perfusion with 100 µM L-KYN reversibly raised extracellular KYNA levels, reaching a maximum of 17.9 ± 3.7 times baseline levels 2 h after the beginning of the perfusion (**Figure 4**). Perfusion with 100 µM D-KYN produced a 10.7 ± 1.2-fold increase in extracellular KYNA levels after 2 h (**Figure 5**).

# ONOO<sup>−</sup> Enhances KYNA Production in vivo

A brief (30 min) perfusion with 50 µM ONOO<sup>−</sup> enhanced the concentration of extracellular KYNA, assessed by in vivo microdialysis in the cerebellum. This treatment raised KYNA levels, assessed in a single 30-min microdialysis fraction, from a basal value of 2.9 ± 0.3 nM to 11.4 ± 2.4 nM (n = 16; P < 0.01).

# ONOO<sup>−</sup> Enhances KYNA Production from L-KYN and D-KYN in vivo

The 30-min pre-perfusion with ONOO<sup>−</sup> substantially enhanced KYNA production from L-KYN or D-KYN (100 µM each) when the enantiomers were applied by reverse dialysis for 2

endogenous tissue levels of KYNA (control; 26.4 ± 2.2 pmoles

h immediately following the termination of perfusion with the pro-oxidant. In both cases, stimulation was greatest in the first 30 min and subsided gradually with time, probably indicating the waning influence of the discontinued perfusion with ONOO<sup>−</sup> (**Figures 4**, **5**). Peak potentiation, compared to control animals perfused without ONOO<sup>−</sup> pre-treatment, was 4.1 ± 1.1-fold for L-KYN (**Figure 4**) and 3.2 ± 0.6-fold for D-KYN (**Figure 5**).

# Discussion

The present study demonstrated that KYNA can be synthesized enzymatically from both L-KYN and D-KYN in the rat cerebellum and, furthermore, that KYNA production from either enantiomer is enhanced in the presence of ROS. These results, which were first obtained in vitro and then confirmed in vivo, suggest that KYNA levels in the cerebellum can normally be controlled by several biosynthetic mechanisms. Conceivably, the relative significance of these biosynthetic routes may differ under

KYNA/mg protein) and represent the mean ± SEM of 8 experiments per group. \*P < 0.05 vs. control, #P < 0.05 vs. L-KYN (A) or D-KYN (B) alone (one-way ANOVA followed by Tukey's post hoc test). AOAA: Aminooxyacetic acid, KA: kojic acid, BA: benzoic acid, MPC: 3-methylpyrazole-5-carboxylic acid.

D-KYN (B), &P < 0.05 vs. L-KYN + ONOO<sup>−</sup> (A) or D-KYN + ONOO<sup>−</sup> (B) (one-way ANOVA followed by Tukey's post hoc test). AOAA: Aminooxyacetic acid, KA: kojic acid, BA: benzoic acid, MPC: 3-methylpyrazole-5-carboxylic acid.

various physiological conditions as well as in various pathological situations involving the cerebellum.

Irreversible transamination of L-KYN by KATs is considered the main means of KYNA formation in the mammalian brain (Turski et al., 1989) and was verified in the present study using the non-specific KAT inhibitor AOAA as an experimental tool. However, D-KYN, too, can serve as a substrate of KATs in both peripheral tissues and the brain (Pérez-de la Cruz et al., 2012), and this comparatively minor synthesis route was confirmed here using cerebellar tissue homogenates. Moreover, in contrast to L-KYN, D-KYN is an excellent substrate of DAAO, which is highly concentrated in the cerebellum (Horiike et al., 1994; Moreno et al., 1999; Verrall et al., 2007). We were therefore not surprised to observe that the cerebellar production of KYNA from D-KYN was quantitatively similar to KYNA formation from L-KYN both in vitro and in vivo, and that the three DAAO inhibitors BA, KA and MPC all caused a substantial reduction in KYNA synthesis from D-KYN.

Our study also showed that the redox environment has a substantial influence on KYNA production in the cerebellum since the pro-oxidant agents OH• and ONOO<sup>−</sup> enhanced KYNA formation from either L-KYN or D-KYN in an artificial milieu in vitro. As the effect of ONOO<sup>−</sup> exceeded the effect of OH•, this pro-oxidant was then tested in cerebellar tissue homogenate where it greatly potentiated the ability of both KYN enantiomers to synthesize KYNA. Notably, subsequent in vivo experiments revealed a substantial increase in extracellular KYNA within 30 min after reverse dialysis of ONOO<sup>−</sup> alone, suggesting that a pro-oxidative environment also stimulates the conversion of endogenous tryptophan or KYN to KYNA (cf. Introduction). Tryptophan, through its metabolites indole-3-pyruvic acid and KYN, and the subsequent production of the anti-oxidant KYNA (Politi et al., 1991; Lugo-Huitrón et al., 2011b; Ugalde-Muniz et al., 2012), may therefore provide a defense mechanism against the detrimental effects of ROS in the brain (see below). Also of note in this context, L-KYN reduces chemiluminescence of luminol induced by H2O<sup>2</sup> or chloramine (Weiss et al., 2013) and is able to inhibit ROS production by neutrophils (Genestet et al., 2014).

ROS and reactive nitrogen species (RNS) are produced during physiological processes and, by interacting with proteins, fatty acids and DNA, perform numerous roles in the regulation of cellular function (Dröge, 2002; Koskenkorva-Frank et al., 2013). Increased production of ROS and RNS and/or insufficient endogenous defense mechanisms in neurons or astrocytes can lead to functional impairments and cause cellular injury (Dringen et al., 2000; Valko et al., 2007; Scherz-Shouval and Elazar, 2011). Specifically, ONOO<sup>−</sup> is a potent, short-lived

used to compare ONOO<sup>−</sup> + D-KYN vs. D-KYN alone at a specific timepoint).

oxidant species that is produced by the reaction of nitric oxide (NO•) and superoxide (O2• <sup>−</sup>). As NO• is a relatively stable and highly diffusible free radical (Szabó et al., 2007; Botti et al., 2010), the formation of ONOO<sup>−</sup> is spatially associated with the sources of O2• <sup>−</sup> (such as the mitochondrial respiratory complex). This allows ONOO<sup>−</sup> to inhibit antioxidant enzymes or neutralize antioxidants (Ischiropoulos et al., 1992; Quijano et al., 1997; MacMillan-Crow et al., 1998; Aykaç-Toker et al., 2001; Savvides et al., 2002) and, consequently, to cause apoptotic or necrotic cell death (Szabó et al., 2007; Franco et al., 2013). Additionally, ONOO<sup>−</sup> produces secondary reactive species such as nitrogen dioxide, hydroxyl and carbonate radicals, all of which interfere with a large number of cellular functions and increase cellular vulnerability (Radi et al., 1991; Bartesaghi et al., 2006). Interestingly, in the present study all three DAAO inhibitors attenuated KYNA production in the presence of ONOO<sup>−</sup> in vitro to various degree (**Figure 3**), indicating that these compounds also have antioxidant activity (see also Gomes et al., 2001).

In the brain, dysregulated redox processes have been proposed to constitute a critical factor in the pathophysiology of neurodegenerative disorders and in major psychiatric diseases including schizophrenia and depressive disorders (Okusaga, 2013; Cahill-Smith and Li, 2014; Salim, 2014; Black et al., 2015; Gu et al., 2015). Notably, oxidative stress in the brain increases with advancing age (Tian et al., 1998), so that redox phenomena may also play a causative role in age-related structural and cognitive deficits (Dröge and Schipper, 2007; Brawek et al., 2010). As abnormal cerebral disposition of KYNA, too, has been linked to various brain pathologies (for review, see Schwarcz et al., 2012), the present findings raise the possibility that the boosting of brain KYNA levels by ROS and/or RNS may be functionally

related to the pathological effects of the pro-oxidants. In other words, we speculate that the increased generation of KYNA in the presence of harmful free radicals may have evolved as a (neuro)protective mechanism to counter the effects of oxidative stress (Lugo-Huitrón et al., 2011a; Ugalde-Muniz et al., 2012). This increase in brain KYNA levels may also have detrimental consequences, however. Thus, even relatively modest elevations in brain KYNA cause a reduction in the extracellular concentrations of several classic neurotransmitters, including dopamine, glutamate and GABA (Carpenedo et al., 2001; Rassoulpour et al., 2005; Wu et al., 2010; Beggiato et al., 2014), and may therefore have adverse effects, especially on cognitive functions (Pocivavsek et al., 2011, 2012). This hypothesis, as well as the detailed cellular, subcellular and molecular mechanisms involved in the interactions between KYNA, ROS and RNS, is currently under investigation in our laboratories.

Our results also raise the question of a possible role of D-tryptophan or D-KYN in this context. Thus, whereas the biology of the essential amino acid L-tryptophan and its major catabolic product, L-KYN, in mammalian systems is reasonably well understood, information about possible roles of their respective D-enantiomers is still very sparse. In fact, D-KYN has so far not been identified as an endogenous constituent of mammalian tissues, though it is readily produced from D-tryptophan, which can originate from the microbial flora (Friedman, 1999; Rodríguez-Crespo, 2008), from the diet (Friedman, 2010) or, possibly, from enzymatic cleavage of D-tryptophan-containing polypeptides, which are found in several vertebrate species including humans (Jilek et al., 2005). Experimentally, D-KYN formation from D-tryptophan, and further degradation to KYNA, has been documented in several mammalian species and in several organs including the brain (Tashiro et al., 1961; Higuchi and Hayaishi, 1967; Loh and Berg, 1971; Ishii et al., 2010; Notarangelo et al., 2013). Importantly, in line with the high DAAO activity of the cerebellum (Horiike et al., 1994; Verrall et al., 2007), KYNA production from systemically applied D-tryptophan or D-KYN is especially pronounced in the cerebellum (Wang et al., 2012; Notarangelo et al., 2013). Oxidative processes in this brain region, which may play a substantive role in a considerable number of grave neurological and psychiatric diseases including cerebellar ataxia and autism (Chauhan and Chauhan, 2006; Kern and Jones, 2006; Wang et al., 2011; Goldani et al., 2014; Rossignol and Frye, 2014; Salim, 2014; Steullet et al., 2014), could therefore conceivably have etiological links to both L-KYN and D-KYN or their respective bioprecursors. These links could be especially relevant in situations involving dysfunctions of the immune system, since increased D-KYN formation from D-tryptophan, as well as enhanced L-KYN formation from L-tryptophan, is seen under inflammatory conditions due to a pronounced up-regulation of the nonstereospecific enzyme indoleamine-2, 3-dioxygenase (Johnson et al., 2009).

The fact that the cerebellum can produce KYNA by routes other than the canonical pathway has likely functional implications since both α7nAChR and NMDAR, which can serve as targets of KYNA (Kessler et al., 1989; Hilmas et al., 2001; Alkondon et al., 2011), are abundant in this area of the brain (Caruncho et al., 1997; Dumas, 2005; Llansola et al., 2005; Taslim and Saeed Dar, 2011). While early experiments, performed mostly using cultured cerebellar neurons, clearly documented inhibition of NMDAR function by high (millimolar) concentrations of KYNA (Gallo et al., 1987; Brockhaus and Deitmer, 2002), recent studies suggest that α7nAChRs may, in fact, be the preferential target of endogenous KYNA in the cerebellum. Thus, increases in cerebellar KYNA concentrations in the nanomolar range reduce extracellular glutamate levels locally in vivo and, as also seen in several regions of the forebrain (Albuquerque and Schwarcz, 2013), this effect can be duplicated by other α7nAChR—but not NMDAR—antagonists (Wu and Schwarcz, 2013 and unpublished data). Interestingly, and in line with the well-documented network connecting the cerebellum with the midbrain and the forebrain (Clower et al., 2005; Mittleman et al., 2008), even a moderate elevation of cerebellar KYNA levels controls the extracellular levels of glutamate

#### References


and dopamine in the distant prefrontal cortex (Wu and Schwarcz, 2013). Such insights provide a conceptual framework for studies designed to explore functional links between (fluctuations in) cerebellar KYNA and motor, cognitive and other forebrain functions that, are influenced by the cerebellum (Ichinohe et al., 2000; Hoshi et al., 2005; Akkal et al., 2007).

In summary, the present study demonstrates that mechanisms other than the classic enzymatic transamination of L-KYN, namely DAAO-catalyzed synthesis from D-KYN, and the interplay between L-KYN or D-KYN (and possibly L-tryptophan or D-tryptophan) with ROS, can contribute to the formation of KYNA in the rat cerebellum (**Figure 6**). These findings have ramifications for the role of KYNA in cerebellar physiology and pathophysiology and suggest novel strategies for normalizing impaired cerebral KYNA function in aging and major brain diseases.

#### Acknowledgments

This work was supported by CONACYT Grant 183867 and Becas Para las Mujeres en la Ciencia L'ORÉAL-CONACYT-UNESCO-AMC 2013.


kynurenic acid controls extracellular glutamate levels in the prefrontal cortex. J. Mol. Neurosci. 40, 204–210. doi: 10.1007/s12031-009-9235-2


and Chronobiology at the Institute of Zoology and Anthropology, University of Göttingen, Germany, ed. R. Hardeland (Göttingen: Cuvillier), 168–176.

Zsizsik, B. K., and Hardeland, R. (2002). Formation of kynurenic and xanthurenic acids from kynurenine and 3-hydroxykynurenine in the dinoflagellate Lingulodinium polyedrum: role of a novel, oxidative pathway. Comp. Biochem. Physiol. C Toxicol. Pharmacol. 133, 383–392. doi: 10.1016/s1532- 0456(02)00126-6

**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.

Copyright © 2015 Blanco Ayala, Lugo Huitrón, Carmona Aparicio, Ramírez Ortega, González Esquivel, Pedraza Chaverrí, Pérez de la Cruz, Ríos, Schwarcz and Pérez de la Cruz. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution and 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.

# Caloric restriction protects against electrical kindling of the amygdala by inhibiting the mTOR signaling pathway

Bryan V. Phillips-Farfán<sup>1</sup> \*, María del Carmen Rubio Osornio<sup>2</sup> , Verónica Custodio Ramírez <sup>2</sup> , Carlos Paz Tres <sup>2</sup> and Karla G. Carvajal Aguilera<sup>1</sup>

<sup>1</sup> Laboratorio de Nutrición Experimental, Instituto Nacional de Pediatría, México City, México, <sup>2</sup> Laboratorio de Neurofisiología, Instituto Nacional de Neurología y Neurocirugía, México City, México

#### Edited by:

Shawn Hayley, Carleton University, Canada

#### Reviewed by:

Rodrigo N. Romcy-Pereira, Universidade Federal do Rio Grande do Norte, Brazil Rubem C. A. Guedes, Universidade Federal de Pernambuco, Brazil Lydia Jimenez-Diaz, University of Castilla La Mancha, Spain

#### \*Correspondence:

Bryan V. Phillips-Farfán, Laboratorio de Nutrición Experimental, Instituto Nacional de Pediatría, Av., Insurgentes Sur 3700 Letra C, Del. Coyoacán, México City, C.P. 04530, México bvphillips@yahoo.com

> Received: 14 October 2014 Accepted: 26 February 2015 Published: 11 March 2015

#### Citation:

Phillips-Farfán BV, Rubio Osornio MC, Custodio Ramírez V, Paz Tres C and Carvajal Aguilera KG (2015) Caloric restriction protects against electrical kindling of the amygdala by inhibiting the mTOR signaling pathway. Front. Cell. Neurosci. 9:90. doi: 10.3389/fncel.2015.00090 Caloric restriction (CR) has been shown to possess antiepileptic properties; however its mechanism of action is poorly understood. CR might inhibit the activity of the mammalian or mechanistic target of rapamycin (mTOR) signaling cascade, which seems to participate crucially in the generation of epilepsy. Thus, we investigated the effect of CR on the mTOR pathway and whether CR modified epilepsy generation due to electrical amygdala kindling. The former was studied by analyzing the phosphorylation of adenosine monophosphate-activated protein kinase, protein kinase B and the ribosomal protein S6. The mTOR cascade is regulated by energy and by insulin levels, both of which may be changed by CR; thus we investigated if CR altered the levels of energy substrates in the blood or the level of insulin in plasma. Finally, we studied if CR modified the expression of genes that encode proteins participating in the mTOR pathway. CR increased the after-discharge threshold and tended to reduce the after-discharge duration, indicating an anti-convulsive action. CR diminished the phosphorylation of protein kinase B and ribosomal protein S6, suggesting an inhibition of the mTOR cascade. However, CR did not change glucose, β-hydroxybutyrate or insulin levels; thus the effects of CR were independent from them. Interestingly, CR also did not modify the expression of any investigated gene. The results suggest that the anti-epileptic effect of CR may be partly due to inhibition of the mTOR pathway.

Keywords: caloric restriction, epilepsy, mTOR signaling cascade, hippocampus, neocortex

# Introduction

At first glance neurodegenerative diseases and epilepsy may seem to have little in common. However, upon further research it is clear that they are surprisingly similar. As an example, abnormal neural activity -the hallmark of epileptic disorders- might cause the synaptic and cognitive deficits observed in neurodegenerative diseases (Sanchez et al., 2012). Another unexpected commonality between neurodegenerative diseases and

**Abbreviations:** CR, caloric restriction; mTOR, mechanistic target of rapamycin; β-HB, β-hydroxybutyrate; AMPK, adenosine monophosphate-activated protein kinase; PKB/Akt, protein kinase B; S6, ribosomal protein S6; S6K, ribosomal protein S6 kinase; TSC2, tuberous sclerosis 2; AL, ad libitum; AD, after-discharge; Cx, neocortex; Hp, hippocampus.

epilepsy may be a defect of autophagy (Wong, 2013; Lipton and Sahin, 2014) due to dysfunction of the mammalian or mechanistic target of rapamycin (mTOR) signaling pathway (Wong, 2013; Lipton and Sahin, 2014).

The ketogenic diet is a beneficial treatment for epilepsy that is frequently used in the clinic (Kossoff and Wang, 2013; Wang and Lin, 2013). Similarly, caloric restriction (CR) is a diet that has anti-epileptic and anti-epileptogenic effects in different animal models (Greene et al., 2001; Bough et al., 2003). Their mechanism of action is poorly understood; however it has been reported that the ketogenic diet might inhibit the mTOR cascade (McDaniel et al., 2011). Excessive activation of this pathway has been observed in diverse models of genetic or acquired epilepsy, suggesting that aberrant function of this cascade plays a crucial role in epilepsy generation (Wong, 2013; Lipton and Sahin, 2014). CR may also inhibit the mTOR pathway in yeast, flies and nematodes (Kapahi et al., 2004; Kaeberlein et al., 2005; Walker et al., 2005).

The mTOR cascade regulates protein synthesis and cell growth, among other things (**Figure 4**; Wong, 2013; Lipton and Sahin, 2014). Insulin and growth factors result in protein kinase B (PKB or Akt) phosphorylation, which inhibits tuberin (tuberous sclerosis 2 or TSC2). In contrast, TSC2 is activated by adenosine monophosphate-activated protein kinase (AMPK) in response to energy deficits. TSC2, together with hamartin (tuberous sclerosis 1 or TSC1), inhibits the activity a complex which includes the mTOR protein. This complex activates ribosomal protein S6 kinase (S6K), which phosphorylates ribosomal protein S6 (S6). These proteins only function when phosphorylated at specific residues. Therefore, the only way to study their activity with the western blot technique is to probe with antibodies that only can recognize the protein when phosphorylated at one or more of these residues.

CR may not hinder the mTOR cascade in all mammalian tissues (Sharma et al., 2012). We thus studied whether CR inhibits the mTOR pathway in the hippocampus and the adjacent temporal neocortex to clarify this issue. This was done by analyzing the phosphorylation of the α1/2 subunits of AMPK, PKB and S6 using the western blot technique. In addition, we also investigated if CR modified the expression of genes encoding for proteins participating in the mTOR cascade. This pathway is regulated by energy and insulin levels, which might be changed by CR; thus we also studied whether CR altered the blood levels of glucose or βhydroxybutyrate (β-HB) or the plasma concentration of insulin. Finally, we investigated the anti-epileptogenic effects of CR using electrical kindling of the amygdala, which allows investigation of epilepsy generation. The hippocampus and the adjacent temporal neocortex were chosen because the amygdala projects to the hippocampus and the adjacent temporal neocortex presumably includes hippocampal projection areas. In opposition, the amygdala was not analyzed to minimize any possible influence of electrode placement within it.

Most prior studies used weanling animals, since the ketogenic diet might be more effective the younger the animal (Bough et al., 1999). Thus, we used young rats due to the precedent set by previous reports. Additionally, the susceptibility to epilepsy is very high in childhood (Kotsopoulos et al., 2002). CR was accomplished by food restriction without vitamin or mineral supplementation; therefore the rats were subjected to mild CR (15%) to avoid any concerns relating to under-nutrition, especially given their young age. This low degree of CR was also selected given prior studies (Greene et al., 2001; Raffo et al., 2008; Linard et al., 2010).

# Materials and Methods

# Animal Handling

20 male Wistar rats were purchased from Harlan Laboratories at postnatal day 21 and then subjected to a 4 h fast to insure that all animals started at a similar metabolic set point. Afterwards, the rats were weight-matched and assigned to four groups: (1) 5 fed ad libitum (control), (2) 5 allowed to feed ad libitum and subjected to the kindling procedure (kindled control), (3) 5 calorically restricted (experimental) and (4) 5 subjected to CR plus kindling (kindled experimental). CR was achieved by feeding rats with a normal diet (2018S, Harlan Laboratories) to 85% of the daily allowance based on the body weights (recorded every day) of rats allowed to feed ad libitum (Rogers, 1979). The animals were kept 30 days on their respective diets before stereotaxic surgery; each group was housed separately under standard conditions. Rat management was performed according to the principles of the guide for the care and use of laboratory animals and NIH publication 85-23, 1985.

#### Stereotaxic Surgery

A group of rats fed ad libitum and a group of animals subjected to CR were operated, while the other 2 groups served as control groups. After deep ketamine (40 mg/kg) anesthesia was achieved, bipolar electrodes were implanted within the left basolateral amygdala and right sensory cortex (stereotaxic coordinates: anterior 2.8, 6.7 mm; lateral 5, 2.5 mm; ventral 8.5, 9 mm; from bregma and the interaural line, respectively). The electrodes consisted of two twisted strands of stainless steel (0.005′diameter) coated with Teflon, except the tips which were separated by 0.02′ . They were soldered to connectors and fastened to the skull with acrylic cement and screws, a screw served as an indifferent source of isoelectric reference. Electrode placement was verified by performing histological staining techniques after the animals were sacrificed.

# Electrical Kindling of the Amygdala

The after-discharge (AD) threshold, defined as the lowest electrical stimulus that elicited an amygdala AD lasting at least 5 s (from the end of the stimulus), was determined after the rats were allowed to recover from surgery for a period of at least 10 days. We employed an electronic circuit breaker device that switched to either the polygraph (Grass 78D, Grass Technologies) or the stimulator (Grass S88, Grass Technologies) connected to the electrode placed in the amygdala (1.8 cm long). This allowed us to stimulate and record through this electrode (Ferrer et al., 1978). The electrode located within the sensory cortex (1.5 cm long) served to verify propagation of epileptic activity (**supplementary figure**). The electrical signal was amplified 1000x and processed with a 16-bit sigma delta converter at a sampling frequency of

256 Hz. The recordings were monitored and saved for off-line analysis with Harmonie v.5.2 software (Stellate Systems).

After 2 min of polygraphic recording to establish the basal activity, the amygdala was stimulated for 1 s with rectangular pulses of 1 ms at 60 Hz (Paz et al., 1991). The first stimulation intensity was 5 V followed by subsequent increments of 1 V until elicitation of an AD (the intensity eliciting an AD lasting at least 5 s from the end of the stimulus was defined as the AD threshold). In the following daily trials, the threshold intensity was applied until at least 5 consecutive generalized convulsive (stage 5) seizures were obtained. Seizure severity was scored according to behavioral criteria; briefly: stage 1-mouth and facial movements, stage 2-head nodding, stage 3-forelimb clonus, stage 4-rearing, stage 5-rearing and falling (Racine, 1972). We measured the following kindling parameters: AD threshold and duration, latency to each Racine stage (number of stimulations/days required to reach each Racine stage), latency to reach criterion (number of stimulations/days needed to reach 5 consecutive generalized stage 5 convulsive seizures), number of stimulations/days where the rats showed each Racine stage and number of stimulations/days the animals had either focal convulsive seizures (stages 1–3) or generalized convulsive seizures (stages 4–5). The intracranial electroencephalographic activity of the amygdala and the sensory cortex was captured using a signal capture software (Galileo, Stellate Systems) and merged with digital video records (Diva, Stellate Systems). All the animals were sacrificed the day after kindled rats were last stimulated (for a total of 64 days subjected to CR, postnatal day 85).

#### Blood and Plasma Measurements

Glucose and β-HB levels were measured with a digital monitor system (MediSense Optium Xceed, Abbott Laboratories) after a 4 h fast, before beginning the experiment and once it ended, using blood samples collected from the tail vein. Blood was also obtained after a 4 h fast at sacrifice and processed to obtain plasma, which was kept at −70◦C. The plasma was used to determine the insulin concentration by an enzyme-linked immunosorbent assay (Alpco Diagnostics) according to standard protocols.

#### Hippocampus and Temporal Neocortex Samples

The brain was obtained rapidly and the hippocampus (Hp) plus adjacent temporal neocortex (Cx) were micro-dissected quickly, separated and cut into two portions. The samples were frozen with liquid nitrogen and stored at −70◦C. One piece was homogenized at 4◦C with a lysis buffer (pH 7) containing (in mM): 50 Tris, 150 NaCl, 1 EGTA, plus phosphatase and protease inhibitors. After a 30 min incubation at 4◦C, the homogenate was centrifuged at 16,000 g for 10 min at 4◦C and the supernatant was recovered. The concentration of total protein was determined using the Lowry method.

The other fragment of each brain area was homogenized at 4◦C with Trizol to obtain total RNA. Briefly, the protocol included chloroform addition, centrifugation at 12,000 g for 15 min at 4◦C, isopropanol addition, centrifugation at 12,000 g for 10 min at 4◦C and addition of ethanol followed by centrifugation at 7500 g for 10 min at 4◦C. The rest of the brain was fixed with 4% paraformaldehyde, cryo-preserved in a 30% sucrose solution, frozen, cut and processed by histological staining techniques to verify electrode placement.

#### Western Blotting

Fifty Microgram/Microliter of protein was separated in a 10% SDS-PAGE gel under denaturing conditions, transferred to PVDF membranes and probed with: phospho-AMPKα (Thr172), AMPKα, phospho-PKB (Thr308), PKB/Akt, phospho-S6 (Ser235/236) or S6 (all from Cell Signaling Technology). The membranes were revealed utilizing a chemiluminescence assay and the amount of protein was analyzed utilizing Quantity One software (Bio-Rad Laboratories). Actin, detected with an antibody (Sigma-Aldrich Corporation), was used as loading control.

# Quantitative Real-Time Polymerase Chain Reaction

Complementary DNA was obtained from total RNA by retrotranscription utilizing the M-MLV reverse transcriptase enzyme and random hexamers (Applied Biosystems) using the 2400 Geneamp PCR system (Perkin Elmer Inc.). The analysis of relative gene expression levels was done by quantitative real-time polymerase chain reaction using TaqMan gene expression assays. The procedure was performed following the instructions provided by the supplier (Applied Biosystems). The analyzed genes were: the α1 and α2 catalytic subunits of AMPK, mTOR, tuberous sclerosis 1 (TSC1), TSC2 and ribosomal protein S6 kinase (S6K, polypeptide 1). The level of these genes was normalized against the expression of the gene that encodes the eukaryotic 18S

rRNA (endogenous control) using the difference in cycle threshold method with the StepOne real-time PCR system (Applied Biosystems).

# Statistics

Unpaired student's t-tests were performed to evaluate differences between kindled animals and those not subjected to kindling

FIGURE 2 | Systemic concentration of energy substrates and the hormone insulin at the initial and final time points of the experiment in animals allowed food ad libitum (AL) or exposed to 15% caloric

restriction (CR). (A) Blood β-hydroxybutyrate (β-HB) levels (mmol/l). (B) Blood glucose concentrations (mg/dl). (C) Plasma insulin levels (ng/ml). ##p ≤ 0.01 and ### p ≤ 0.001, initial vs. final.

(control vs. kindled control, experimental vs. kindled experimental). All the variables behaved similarly between these groups (Supplementary Data). Thus, control and kindled control data were pooled together (hereafter AL, n = 10) and the results for experimental and kindled experimental rats were combined (henceforth CR, n = 10) for all subsequent analysis. Dissimilarities in body weight were evaluated utilizing two-way analysis of variance tests for each pair of days followed by Holm-Sidak post-hoc tests. The differences between groups (AL vs. CR) were evaluated using unpaired student's t-tests, whereas dissimilarities among groups (before vs. after) were analyzed with paired student's t-tests. A p ≤ 0.05 was considered to be significantly different. Figure data were expressed as mean ± standard error.

# Results

#### Bio-Indexes

Body weights increased in both groups (**Figure 1**), but mild CR significantly reduced the weight gain of young animals [i.e., AL vs. CR in days 6 vs. 7: F(1, 16) = 6.20, p < 0.05, t = 2.49; for all other comparisons see Supplementary Data]. On the other hand, fasting blood glucose and β-HB as well as plasma insulin (**Figure 2**) were similar between the groups [initial glucose: t(1, 18) = −0.14, p = 0.89; final glucose: t(1, 18) = 0.09, p = 0.93; initial β-HB: t(1, 18) = −0.15, p = 0.88; final β-HB: t(1, 18) = 0.50, p = 0.62; plasma insulin: t(1, 18) = 0.371, p = 0.72]. Nonetheless, β-HB decreased whereas glucose increased significantly with time in both groups [initial vs. final; AL glucose: t(1, 9) = −9.00, p ≤ 0.001; CR glucose: t(1, 9) = −13.33, p ≤ 0.001; AL β-HB: t(1, 9) = 4.03, p < 0.01 and CR β-HB: t(1, 9) = 4.96, p ≤ 0.001].

# Electrical Kindling of the Amygdala

CR significantly increased [t(1, 8) = −3.35, p = 0.01] the AD threshold (**Figure 3A**) and reduced the AD duration in days 14 [t(1, 8) = 2.73, p < 0.05], 15 [t(1, 8) = 4.63, p < 0.01] and 16 [t(1, 8) = 2.92, p < 0.05] of the kindling procedure (**Figure 3B**). However, CR did not significantly alter the AD duration in any other day nor did it affect any other kindling parameter (latency to each Racine stage: AL 4.0 ± 1.0 vs. CR 3.8 ± 0.2 to stage 2, AL 8.7 ± 2.3 vs. CR 9.2 ± 1.1 to stage 3, AL 11.7 ± 1.2 vs. CR 12.8 ± 1.7 to stage 4, AL 13.7 ± 0.9 vs. CR 14.6 ± 2.0 to sage 5; latency to criterion: 17.7 ± 0.9 vs. 18.6 ± 2.0; days in each Racine stage: AL 3.0 ± 1.0 vs. CR 2.8 ± 0.2 in stage 1, AL 4.7 ± 1.5 vs. CR 5.4 ± 1.0 in stage 2, AL 3.0 ± 1.2 vs. CR 3.6 ± 1.4 in stage 3, AL 2.0 ± 1.0 vs. CR 1.8 ± 0.4 in stage 4, AL 9.3 ± 0.7 vs. CR 9.4 ± 0.7 in stage 5; days with focal convulsive seizures: AL 10.7 ± 1.2 vs. CR 11.8 ± 1.7; days rats had generalized convulsive seizures: AL 11.3 ± 0.3 vs. CR 11.2 ± 0.5; statistical comparisons in Supplementary Data).

#### mTOR Signaling Pathway

CR tended to augment AMPK phosphorylation (**Figures 4**, **5**) in the Cx [t(1, 18) = −1.47, p = 0.16] and significantly increased AMPK phosphorylation in the Hp [t(1, 18) = −2.44, p < 0.05]. CR significantly reduced PKB/Akt and S6 phosphorylation in the Cx [PKB: t(1, 18) = 3.03, p < 0.01 and S6: t(1, 18) = 2.13, p < 0.05] and Hp [PKB: t(1, 18) = 3.13, p < 0.01 and S6: t(1, 18) = 2.39, p < 0.05], suggesting that CR inhibited the mTOR cascade. Finally, CR did not significantly change the expression level of the studied genes [**Figure 6**; AMPKα1: Cx: t(1, 18) = 0.40, p = 0.70, Hp: t(1, 18) = 0.62, p = 0.54; AMPKα2: Cx: t(1, 18) = 1.37, p = 0.19, Hp: t(1, 18) = 0.57, p = 0.58; mTOR: Cx: t(1, 18) = −1.09, p = 0.29, Hp: t(1, 18) = −1.71, p = 0.10; S6K: Cx: t(1, 18) = 1.78, p = 0.09, Hp: t(1, 18) = 0.08, p = 0.94; TSC1: Cx: t(1, 18) = 1.74, p = 0.10, Hp: t(1, 18) = −1.33, p = 0.20; TSC2: Cx: t(1, 18) = 0.27, p = 0.79, Hp: t(1, 18) = 0.58, p = 0.57].

# Discussion

#### Body Weight

Interestingly, mild CR significantly diminished weight gain of weanling rats; therefore our concerns with under-nutrition were well founded. This is not a universal finding given that some studies have shown a similar decrease in body weight (Greene et al., 2001; Cheng et al., 2003, 2004; Eagles et al., 2003), whereas other reports have not (Bough et al., 1999, 2003; Raffo et al., 2008). These contradictory findings have been mostly ignored, perhaps because little explanation can be offered as to the reasons for these discrepancies. The age at diet onset and length of time on it are the most likely factors that may account for the inconsistent results regarding the effect of CR on body weight. It is important to note that a reduction in body weight gain is consistent with an inhibition of the mTOR pathway. Even if a decrease in

FIGURE 4 | (A) Brief diagram of the mechanistic target of rapamycin (mTOR) pathway with analyzed proteins/genes (black arrows, activation; red lines, inhibition; AMPK, adenosine monophosphate-activated protein kinase; PKB, protein kinase B; TSC1, tuberous sclerosis 1; TSC2, tuberous sclerosis 2; S6K, ribosomal protein S6 kinase; S6, ribosomal protein S6). (B) Representative blots showing both phosphorylated (p) and total AMPK, PKB and S6 in the neocortex (Cx) and hippocampus (Hp) of rats fed ad libitum (AL) or subjected to 15% caloric restriction (CR). The molecular weight (MW) is shown in kilodaltons (kDa) to the right of each protein pair.

body weight was observed, CR showed anti-epileptic actions in all reports. Thus, it is unlikely the animals were malnourished given the adverse effects of malnutrition on epilepsy (Crepin et al., 2009).

# Biochemical Parameters

Blood glucose or β-HB and plasma insulin were similar between the groups, again showing that CR did not result in undernutrition. The normal developmental profile of increasing glucose and decreasing β-HB levels was observed (Edmond et al., 1985; Prins, 2012). CR has not been reported to affect blood β-HB levels in rats (Bough et al., 1999; Cheng et al., 2003, 2004; Eagles et al., 2003; Raffo et al., 2008; Linard et al., 2010), except for one report (Bough et al., 2003). On the other hand, Sprague-Dawley rats exposed to CR starting at weaning for 7 days had lower blood glucose compared to animals fed ad libitum (Cheng et al., 2003, 2004). In contrast, no significant alterations were seen in blood glucose levels of Wistar rats subjected to CR for 3 weeks starting at postnatal day 50 (Raffo et al., 2008; Linard et al., 2010). The most probable explanation is that the animals in the latter studies and ours adapted their metabolism to CR due to its duration, though other factors (species, age at diet onset) might account for the discordant findings. For example, CR significantly increased blood β-HB and decreased glucose levels in mice; suggesting a species difference (Greene et al., 2001; Mahoney et al., 2006). The finding that CR did not modify glucose, β-HB or insulin levels is very important because any effects of CR are thus independent from them.

# Electrical Kindling of the Amygdala

CR significantly increased the after-discharge threshold and diminished the after-discharge duration in a few days of the kindling procedure, but did not meaningfully alter any other kindling parameter. The number of stimulations to a generalized convulsive seizure (stage 5), known as the kindling rate, is a measure of the epileptogenic capability in this model. 15% CR did not alter the kindling rate in Wistar rats, thus CR was not antiepileptogenic in these outbred animals. The anti-epileptogenic action of CR has been validated in mice and rats inbred for epileptic susceptibility (Greene et al., 2001; Azarbar et al., 2010), generating doubts whether CR really can be anti-epileptogenic in humans. However, a higher degree of CR likely has antiepileptogenic effects in outbred animal models and humans. This is to be expected if a higher degree of CR has more prominent inhibitory effects on the mTOR signaling cascade. Alternatively, the anti-epileptogenic action of CR may require reductions in

glucose or insulin and/or an increase in β-HB (Greene et al., 2001; Yamada, 2008).

Increases in the after-discharge threshold can be predictive of the anti-convulsive capacity of drugs in pharmacological tests (Azarbar et al., 2010). Similarly, since the after-discharge is an electrographic seizure, a reduction in its duration is an anticonvulsive effect (Azarbar et al., 2010). CR augmented the afterdischarge threshold and reduced its duration, without affecting any other kindling variable (most importantly the kindling rate). This suggests an anti-convulsive profile rather than an antiepileptogenic action; which is totally independent from changes in glucose, insulin or β-HB.

#### mTOR Signaling Cascade

Mild CR significantly augmented AMPK phosphorylation in the hippocampus. Besides, it reduced PKB/Akt and S6 phosphorylation in the temporal neocortex and the hippocampus; suggesting that even mild CR inhibits the mTOR signaling pathway. Similarly, mice which were subjected to 30% CR starting at 3 months of age for 9 or 17 months showed reduced PKB/Akt and S6K phosphorylation levels in the hippocampus, suggesting that CR hindered this cascade (Yang et al., 2014). In contrast, 35–40% CR starting at 3–4 months of age for 6 months in Fischer 344 and Brown Norway F1 rats did not affect the phosphorylation level of several proteins that participate in this pathway in the cerebellum or neocortex (Sharma et al., 2012). These discordant findings are probably explained by the age at diet onset or its duration, although factors such as species, strain and brain region may have influenced the results.

The results suggest that CR inhibited the mTOR cascade in the hippocampus and adjacent temporal neocortex after the animals showed 5 consecutive generalized convulsive seizures (postnatal day 75). Thus, the status of the mTOR pathway in the amygdala when the after-discharge threshold (postnatal day 55) and duration (postnatal days 55–75) were determined is unknown. However, rats subjected to CR had been on this diet for at least 40 days before the kindling procedure started, thus CR probably likewise hindered the mTOR cascade in the amygdala. Furthermore, the effects of CR on the after-discharge threshold and duration were consistent with its known effects (Bough et al., 1999, 2003; Eagles et al., 2003; Azarbar et al., 2010) and consistent with regulation of neuron/network excitability by the mTOR cascade (Lasarge and Danzer, 2014; Lipton and Sahin, 2014).

Mild CR (15%) for 55 days starting on postnatal day 21 (this study) did not significantly change AMPKα1, AMPKα2, mTOR, S6K, TSC1, or TSC2 expression in the hippocampus or temporal neocortex. In agreement, 20% CR for 6 months starting at 4 months of age did not significantly affect gene expression of proteins involved in the mTOR cascade in the hippocampus of Sprague-Dawley rats. However, 40% CR augmented the expression of genes involved in this pathway (Martin et al., 2008). These results show that the degree of CR is a very important factor determining the results. Similarly, 30–40% CR starting at 3–4 months of age for ≥6 months (Sharma et al., 2012; Yang et al., 2014) did not alter the level of mTOR, S6K or TSC2 expression in the hippocampus, neocortex or cerebellum. But 30% CR for variable times (∼4, 11, 18, or 23 months) starting at weaning reduced the expression of genes that participate in this cascade (such as S6K) in the hypothalamus of inbred mice (Wu et al., 2009). Given the similarity between two of these reports (Wu et al., 2009; Yang et al., 2014), the most likely explanation is that CR has differential impact depending on the age at diet onset or the brain region, as previously suggested (Zeier et al., 2011). Although CR and the ketogenic diet both inhibit the mTOR pathway (McDaniel et al., 2011), this is unlikely to be the only anti-epileptic mechanism of action of these treatments since they differ in important respects (Bough et al., 2000, 2003; Hartman et al., 2010). This is emphasized by the fact that the ketogenic diet has anti-epileptogenic effects (Jiang et al., 2012), whereas mild CR displayed a profile more consistent with an anti-convulsive action.

Our findings suggest using CR as a therapeutic intervention to inhibit the mTOR cascade in the human brain. The advantage of CR is that it can be combined with other treatments; this is generally done when using the ketogenic diet (Kossoff and Wang, 2013; Wang and Lin, 2013). Moreover, CR may be used for seizureepilepsy prophylaxis; although more studies are needed to resolve current contradictory findings (Zeng et al., 2010; Macias et al., 2013). Unfortunately, even mild CR significantly reduced weight gain in weanling rats which may limit its use in children and adolescents due to concerns regarding their growth. In addition, despite the presumed inhibition of the mTOR pathway all the rats progressed to generalized convulsive seizures with the same kindling rate. This was probably due to the low degree of CR used; a higher amount of CR likely would result in anti-epileptogenic effects.

In summary, CR elevated the after-discharge threshold without changes in other kindling variables; suggesting an anticonvulsive action. CR increased AMPK phosphorylation in the hippocampus, while it decreased phosphorylation of PKB/Akt and S6 in the temporal neocortex and hippocampus; implying an inhibition of the mTOR signaling cascade. The effects of CR were likely independent from any changes in glucose, β-HB, insulin or gene expression (AMPKα1, AMPKα2, mTOR, S6K, TSC1, or TSC2) since none were modified by CR. Our main conclusion is that CR may inhibit the mTOR signaling cascade in the brain, which results in an anti-convulsive increase in convulsive seizure threshold that does not depend on changes in energy or insulin levels or gene expression of proteins that participate in this cascade. Thus, CR might be used as a therapeutic tool to inhibit the mTOR pathway to achieve anti-convulsive effects.

# Acknowledgments

The present work was possible thanks to grants from CONA-CYT (151136 to KGCA) and our institution (Fondos Federeales 48/2011 to KGCA and 55/2012 to BVPF).

# Supplementary Material

The Supplementary Material for this article can be found online at: http://www.frontiersin.org/journal/10.3389/fncel.2015. 00090/abstract

#### Supplementary Figure | Fragment of an electrographic record, the space between each line represents one second. The channels show the

electrographic activity of the sensory cortex (SC) and the amygdala (A) before and after stimulating the amygdala. SA, stimulus artifact.

# 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.

Copyright © 2015 Phillips-Farfán, Rubio Osornio, Custodio Ramírez, Paz Tres and Carvajal Aguilera. 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.

# Cerebellar transcriptional alterations with Purkinje cell dysfunction and loss in mice lacking PGC-1**α**

#### *Elizabeth K. Lucas 1,2, Courtney S. Reid1, Laura J. McMeekin1, Sarah E. Dougherty1,3, Candace L. Floyd4 and Rita M. Cowell <sup>1</sup> \**

*<sup>1</sup> Department of Psychiatry and Behavioral Neurobiology, University of Alabama at Birmingham, Birmingham, AL, USA*

*<sup>2</sup> Department of Neuroscience, Icahn School of Medicine at Mount Sinai, New York, NY, USA*

*<sup>3</sup> Department of Neuroscience, Johns Hopkins University School of Medicine, Baltimore, MD, USA*

*<sup>4</sup> Department of Physical Medicine and Rehabilitation, University of Alabama at Birmingham, Birmingham, AL, USA*

#### *Edited by:*

*Marco Antonio Meraz-Ríos, Centro De Investigación Y De Estudios Avanzados, Mexico*

#### *Reviewed by:*

*Laurens Bosman, Erasmus MC, Netherlands María Del Carmen Cárdenas-Aguayo, Center for Research and Advanced Studies of the National Polytechnic Institute, Mexico Jose Tapia-Ramirez, Center for Research and Advanced Studies of the National Polytechnic Institute, Mexico*

#### *\*Correspondence:*

*Rita M. Cowell, Department of Psychiatry and Behavioral Neurobiology, University of Alabama at Birmingham, 1720 7th Avenue S. SC 729, Birmingham, AL 35294-0017, USA e-mail: rcowell@uab.edu*

Alterations in the expression and activity of the transcriptional coactivator peroxisome proliferator-activated receptor γ coactivator-1α (*ppargc1a* or PGC-1α) have been reported in multiple movement disorders, yet it is unclear how a lack of PGC-1α impacts transcription and function of the cerebellum, a region with high PGC-1α expression. We show here that mice lacking PGC-1α exhibit ataxia in addition to the previously described deficits in motor coordination. Using q-RT-PCR in cerebellar homogenates from PGC-1α−/<sup>−</sup> mice, we measured expression of 37 microarray-identified transcripts upregulated by PGC-1α in SH-SY5Y neuroblastoma cells with neuroanatomical overlap with PGC-1α or parvalbumin (PV), a calcium buffer highly expressed by Purkinje cells. We found significant reductions in transcripts with synaptic (complexin1, Cplx1; Pacsin2), structural (neurofilament heavy chain, Nefh), and metabolic (isocitrate dehydrogenase 3a, Idh3a; neutral cholesterol ester hydrolase 1, Nceh1; pyruvate dehydrogenase alpha 1, Pdha1; phytanoyl-CoA hydroxylase, Phyh; ubiquinol-cytochrome c reductase, Rieske iron-sulfur polypeptide 1, Uqcrfs1) functions. Using conditional deletion of PGC-1α in PV-positive neurons, we determined that 50% of PGC-1α expression and a reduction in a subset of these transcripts could be explained by its concentration in PV-positive neuronal populations in the cerbellum. To determine whether there were functional consequences associated with these changes, we conducted stereological counts and spike rate analysis in Purkinje cells, a cell type rich in PV, from PGC-1α−/<sup>−</sup> mice. We observed a significant loss of Purkinje cells by 6 weeks of age, and the remaining Purkinje cells exhibited a 50% reduction in spike rate. Together, these data highlight the complexity of PGC-1α's actions in the central nervous system and suggest that dysfunction in multiple cell types contribute to motor deficits in the context of PGC-1α deficiency.

**Keywords: PPARGC1A, cerebellum, ataxia, Catwalk, stereology, Refsum disease, Friedreich Ataxia**

#### **INTRODUCTION**

Since the identification of peroxisome proliferator activated receptor γ coactivator-1α (PGC-1α) as a master regulator of mitochondrial biogenesis in brown adipose tissue (Puigserver et al., 1998), a number of studies have demonstrated an association between PGC-1α dysfunction and neurological disorders, including Huntington Disease (Cui et al., 2006; Weydt et al., 2006; Chaturvedi et al., 2010; Hathorn et al., 2011; Johri et al., 2012; Puddifoot et al., 2012; Soyal et al., 2012), Parkinson Disease (St-Pierre et al., 2006; Zheng et al., 2010; Clark et al., 2011; Shin et al., 2011; Thomas et al., 2012), and Alzheimer Disease (Qin et al., 2009; Sheng et al., 2012; Pedros et al., 2014). Germline (Lin et al., 2004; Leone et al., 2005; Lucas et al., 2012) and nervous system-specific (Lucas et al., 2012) deletion of PGC-1α causes pronounced neurodegeneration in brain regions involved in motor control and deficits in motor coordination, and PGC-1α overexpression can promote neuronal survival and function in some mouse models (Cui et al., 2006; Mudo et al., 2012; Tsunemi et al., 2012). However, little is known about the physiological impact of PGC-1α dysfunction on specific cell types and circuits in the brain.

PGC-1α can interact with a number of transcription factors and coactivators to drive expression of metabolic transcriptional programs in peripheral tissues (Lin, 2009). Throughout the brain, PGC-1α protein is highly concentrated in neurons expressing the enzyme glutamic acid decarboxylase 67 (GAD67; Cowell et al., 2007). We recently determined that in GAD67-positive interneurons of the cortex, PGC-1α is involved in regulating the calcium buffer parvalbumin (PV), the synaptic proteins synaptogamin 2 (Syt2) and complexin 1 (Cplx1), and the structural protein neurofilament heavy chain (Nefh; Lucas et al., 2010, 2014). These data indicate that PGC-1α can drive genes besides those involved in metabolic regulation and that PGC-1α's role in neurons may be more complex than previously appreciated.

Considering the high concentration of PGC-1α in the cerebellum and the role of the cerebellum in motor coordination, we hypothesized that mice lacking PGC-1α would show transcriptional deficits and evidence of cerebellar dysfunction. Here we report that PGC-1α−/<sup>−</sup> mice exhibit ataxia and reductions in genes involved in metabolism, synaptic function, and structural support. Alterations in protein expression are restricted to specific neuronal populations within the cerebellum, especially Purkinje cells, and Purkinje cell number and firing rate are decreased in PGC-1α−/<sup>−</sup> animals. These experiments suggest that PGC-1α is required for proper gene expression in the cerebellum and that cerebellar deficits contribute to motor abnormalities associated with PGC-1α deficiency.

#### **METHODS**

#### **ANIMALS**

All experimental protocols were approved by the Institutional Animal Care and Use Committee of the University of Alabama at Birmingham. PGC-1α−/<sup>−</sup> mice (Lin et al., 2004) were used for experiments, and PGC-1α+/+, <sup>+</sup>/−, and <sup>−</sup>/<sup>−</sup> mice were obtained from offspring of PGC-1α+/<sup>−</sup> breeding pairs. Conditional deletion of PGC-1α was produced by crossing mice with LoxP sites flanking the exon 3–5 region of the PGC-1α gene (Lin et al., 2004; gift of Bruce Spiegelman, Harvard University, Cambridge, USA) with mice expressing Cre recombinase driven by the PV promoter (Hippenmeyer et al., 2005; #008069 from the Jackson Laboratory, Bar Harbor, ME, USA). For conditional knockout experiments, littermates expressing Cre recombinase without loxP sites were used as controls. To determine specificity of the Cre-mediated recombination pattern, PV-Cre mice were crossbred to mutant tomato/mutant green reporter mice (Muzumdar et al., 2007; #007676 from Jackson Laboratory). All mice were maintained on a C57BL6/J genetic background and housed 2–5 in a cage at 26 ± 2◦C room temperature with food and water *ad libitum*. With the exception of behavioral analysis, experiments were conducted on male and female mice, and no significant differences were found between male and female mice in any measure.

#### **CATWALK GAIT ANALYSIS**

Gait analysis was performed on 6-month-old male PGC-1α−/<sup>−</sup> mice with the CatWalk system (Noldus Information Technology, Leesburg, VA) as previously described (Hamers et al., 2006). Briefly, mice traversed a glass walkway (109 × 15 × 0.6 cm) with dark plastic walls spaced 15 cm apart in a dark room. Light from an encased fluorescent bulb was internally reflected within the glass walkway and scattered when the plantar surface of the paw contacted the walkway floor, thereby producing paw prints. Paw prints were recorded by a high-speed CCD camera mounted below the walkway at 50 half-frames/s and were stored on a computer by the associated CatWalk 7.1 acquisition software. Trials in which the animal stopped or changed direction were excluded from subsequent analysis, and three uninterrupted trials were analyzed and averaged to obtain the final gait analysis values. Paw print designations were assigned and data analyzed using the CatWalk analysis software (v 7.1) by an experimenter who was blinded to the genotype of the animals. Mice were tested at 6 months of age because PGC-1α−/<sup>−</sup> animals weighed significantly less than PGC-1α+/<sup>+</sup> animals at earlier time points (data not shown), which could bias Catwalk detection and subsequent analyses of paw prints.

#### **GENE EXPRESSION ANALYSES**

Mice, aged 1 month for germline and 6 months for conditional PGC-1α deletion, were anesthetized with isoflurane before sacrifice by decapitation. Cerebella were collected in centrifuge tubes, flash frozen on dry ice, and stored at −80◦C. Before processing, samples were incubated in RNA*later*®-ICE (Ambion, Austin, TX, USA) according to manufacturer's instructions. Tissue was homogenized with Tissue-Tearor (Biospec, Bartlesville, OK, USA) in Trizol reagent, and RNA was isolated by the Trizol/choloformisopropanol method following the manufacturer's instructions (Invitrogen, Carlsbad, CA, USA). RNA concentrations and purity were quantified using a Thermo Scientific NanoDrop2000 (Fisher Scientific, Pittsburg, PA, USA). Equivalent amounts of RNA (1μg) were treated with DNase I (Promega, Madison, WI, USA) at 37◦C for 30 min, and DNase was inactivated at 65◦C for 15 min. RNA was reverse-transcribed using the High-Capacity cDNA Archive Kit (Applied Biosystems, Carlsbad, CA, USA). Taqman qPCR was performed with JumpStart Taq Readymix (Sigma, St. Louis, MO, USA) and the following mouse-specific Applied Biosystems primers listed in Supplementary Table 1.

Reaction protocols consisted of an initial ramp time (2 min, 50◦C; 10 min, 95◦C) and 40 subsequent cycles (15 s, 95◦C; 1 min, 60◦C). Relative concentrations of the genes of interest were calculated in comparison to a standard curve calculated from dilutions of cDNA (1:5, 1:10, 1:20) from a pooled sample of wildtype littermate controls. Values were normalized to β-actin (ABI# Mm00607939\_s1) or 18S rRNA (ABI# Hs99999901\_s1) for values for the same sample and then expressed as ratio to control samples ± SEM.

#### **WESTERN BLOT ANALYSIS**

Primary antibodies included rabbit anti-Cplx1,2 (Synaptic Systems, Goettingen, Germany), chicken anti-Nefh (Abcam, Cambridge, MA, USA), rabbit anti-pacsin2 (Sigma), and mouse anti-β-actin (Chemicon, Billerica, MA, USA). One-month-old PGC-1α+/<sup>+</sup> and <sup>−</sup>/<sup>−</sup> mice were anesthetized with isoflurane before sacrifice by decapitation. Cerebella were collected in centrifuge tubes, flash frozen on dry ice, and stored at −80◦C. Cerebella were homogenized in RIPA buffer (150 mM NaCl, 50 mM Tris, 1% Triton X-100, 1% sodium dodecyl sulfate, 0.5% deoxycholic acid; pH 8.0) containing a protease inhibitor tablet. Total protein concentration was determined with a bicinchonicic acid protein assay kit (Thermo Scientific, Waltham, MA, USA), and absorbance was measured at 540 nm. Protein was denatured in sample buffer (62.5 mM Tris-HCl, 20% glycerol, 2% sodium dodecyl sulfate, 5% β-mercaptoethanol; pH 6.8) at 95◦C. Equivalent amounts of protein were loaded into precast polyacrylamide NuPage gels (Invitrogen). One interblot control sample was loaded onto every gel to permit comparison among gels. Protein was transferred onto nitrocellulose membranes. Membranes were blocked with 5% milk in tris buffered saline (TBS; pH 7.6) with 1% Tween (TBS-T) and probed with primary antibodies in 5% IgG-free bovine serum albumin (BSA; Jackson ImmunoResearch, West Grove, PA, USA) TBS-T overnight at 4◦C and peroxidase-conjugated secondary antibodies (Jackson ImmunoResearch) in 5% milk TBS-T for 1 h at room temperature. Membranes were incubated in chemiluminescent substrate (Thermo Scientific) and exposed to film. The optical density of bands was calculated after background subtraction using UN-SCAN-IT gel analysis software (Silk Scientific Inc., Orem, UT, USA). All bands were normalized to the interblot control band, then to actin, and expressed as optical density mean ± SEM.

#### **IMMUNOFLUORESCENCE**

Animals were anesthetized with isoflurane and perfused intracardially with phosphate-buffered saline (PBS, pH 7.4) and 4% paraformaldehyde in PBS. Brains were removed, postfixed in 4% paraformaldehyde for 24–72 h, cryoprotected in graded sucrose (5–20%), embedded in a mixture of 20% sucrose and Tissue-Tek O.C.T. Compound (Sakura Finetek, Torrance, CA), and frozen at −80◦C. Tissue blocks were sectioned at 20μm, mounted onto charged slides (Fisher, Hampton, NH), and allowed to dry overnight before freezing at −80◦C.

The same primary antibodies were used for immunofluorescence as Western blotting (see above) with the addition of mouse anti-GAD67 (Chemicon), mouse anti-PV (Sigma), and rabbit anti-PV (Swant, Marly, Switzerland). Slides were thawed, washed in PBS, and blocked with 10% serum from the host of the secondary antibody in PBS for 1 h. Sections were then incubated in the primary antibody overnight with 3% BSA and 5% serum from the host of the secondary antibody in PBS with Triton-X100 (PBS-T; Sigma) at 4◦C. Slides were washed in PBS-T and PBS and incubated 2 h at room temperature with the corresponding fluorescence-conjugated secondary antibody (Jackson ImmunoResearch) with 5% serum for the host of the secondary antibody in PBS-T. For colabeling with GAD67, a mouse-onmouse kit (Vector Laboratories, Burlingame, CA) was used following the manufacturer's instructions to reduce background staining. Sections were coverslipped with Prolong Antifade Gold (Invitrogen) and stored at 4◦C. All images were captured with a Leica confocal microscope (Buffalo Grove, IL, USA). All confocal settings, including laser intensity, gain, offset, and zoom, were held constant across all groups for a given protein. Images were imported into Adobe Photoshop CS3 (Adobe, San Jose, CA) for adjustments to contrast and brightness.

#### **STEREOLOGY**

Animals were transcardially perfused, and brains were processed as described above for immunofluorescence. Brains were sectioned 30μm thick in the sagittal plane and stored at −80◦C until use. Hematoxylin and Eosin (H&E; Sigma) staining was conducted according to the manufacturer's instructions on serial cerebellum sections to visualize Purkinje cells. Stereological analyses were performed with a computerized stereology workstation, consisting of a modified Olympus BX51 light microscope (Center Valley, PA, USA) equipped with 4×–100× Plan Apo objectives, a motorized specimen stage (Ludl Electronic Products Ltd., Hawthorne, NY, USA), CCD color video camera (CX9000, Microfire Optronics, Goleta, CA, USA) and StereoInvestigator 10 software (MBF Biosciences, Williston, VT, USA). The optical fractionator principle (West et al., 1991) was used for estimation of Purkinje cells in the cerebellum. Systematic random sampling in every 20th section was performed in a total of 11 sections per animal. All cells whose nucleus top came into focus within the unbiased counting spaces throughout the delineated regions were counted. The systemic random sampling grid layout was 253 × 188μm with 100 desired sites per animal. A 60 × 60μm square dissector frame was used to sample the sections.

#### **LOOSE PATCH RECORDINGS**

Loose patch recordings of Purkinje cells were conducted as in Dougherty et al. (2012). Mice (P42 ± 5 days) were anesthetized with isoflourane and then killed by decapitation. Brains were places in ice-cold artificial cerebral spinal fluid (ACSF) containing the following (in mM): 125 NaCl, 2.5 KCl, 2 CaCl2, 1 MgCl2, 25 NaHCO3 1.25 Na2HPO4and 25 D-glucose with a pH of 7.4 and osmolality of 295 ± 5 mOsm. ACSF was bubbled with 95%O2/5%CO2. Sagittal cerebellar slices (300 μm thick) were cut using a Vibratome (7000 smz, Campden Instruments, Sarasota, FL, USA). The slices were allowed to rest for 60 min at room temperature (22–23◦C), and all subsequent procedures were performed at room temperature. Slices were superfused continuously with oxygenated recording ACSF at room temperature. Slices were viewed with an upright microscope (Zeiss Axio Examiner A1, Thornwood, NY, USA) using infrared-differential interference contrast optics. Loose patch recordings were acquired from visually identified Purkinje cells using Axio Vision 4.8 software (Zeiss). Cellular activity was recorded using internal solution containing the following (in mM): 140 K-gluconate, 1 EGTA, 10 HEPES and 5 KCL, pH 7.3. Pipette tip resistance was 2–5 M-. The extracellular recording pipettes containing the internal solution were placed under visual control directly adjacent to the soma of the cell of interest. Positive pressure was applied throughout this process followed by a brief release of pressure to form a seal averaging approximately 45 M-. Cell-attached, loose patch-clamp recordings were obtained using an Axon CNS Molecular Devices amplifier (Multiclamp 700B, Molecular Devices, Sunnydale, CA, USA), filtered at 10 kHz and digitized at 20 kHz (Digidata 1440A, Molecular Devices). A 3 min gap-free protocol on Clampex 10.2 software (Molecular Devices) was used for data collection. Detection and analysis of event frequency and inter-event interval were performed semi-automatically using the program Clampfit 10.2. The detection threshold was set for analysis based on the event amplitude from a given cell.

#### **STATISTICAL ANALYSES**

All statistical analyses were conducted with SPSS software (IBM, Armonk, NY, USA). For initial gene expression studies, Oneway ANOVA comparing <sup>+</sup>/+, <sup>+</sup>/−, and <sup>−</sup>/<sup>−</sup> mice followed by Fisher's LSD was implemented. After determination of putative gene targets in the cerebellum, confirmation of protein expression in <sup>−</sup>/<sup>−</sup> mice and gene expression in conditional knockouts was analyzed with one-tailed *t*-tests with the *a priori* hypothesis that gene and protein expression would decrease in mice lacking PGC-1α. All other differences between PGC-1α+/<sup>+</sup> and <sup>−</sup>/<sup>−</sup>

mice were detected by two-tailed *t*-tests. All data are presented as mean ± SEM.

#### **RESULTS**

#### **MICE LACKING PGC-1α EXHIBIT ATAXIA**

PGC-1α−/<sup>−</sup> mice were previously described to have severe motor abnormalities indicative of neurological dysfunction, such as impaired rotarod performance, decreased rearing behavior, hindlimb clasping, and tremor (Lucas et al., 2012). Given the high expression of PGC-1α in the cerebellum compared to other brain regions (Cowell et al., 2007), we hypothesized that motor impairment in PGC-1α−/<sup>−</sup> animals would include ataxia and conducted CatWalk gait analyses on male PGC-1α+/<sup>+</sup> and <sup>−</sup>/<sup>−</sup> littermates at 6 months of age. CatWalk gait analyses have been validated as indicators of motor abnormalities in rodents, and several CatWalk variables indicating paw print dimensions and the time and distance relationships between footfalls are affected in rodent models of neurological diseases in which PGC-1α has been implicated, such as Huntington (Vandeputte et al., 2010) and Parkinson (Vlamings et al., 2007; Chuang et al., 2010) Diseases. We assessed paw print area, which indicates the average area of each paw print over a 4 step cycle sequence, and found that PGC-1α−/<sup>−</sup> mice exhibited reduced hindpaw area compared to PGC-1α+/<sup>+</sup> mice [*t*(18) = 2.53, *p* = 0.02; **Figure 1A**]. We also evaluated the stand index, which is ratio of all maximum paw contact values (stance) over the stance duration (seconds), normalized for camera acquisition rate. PGC-1α−/<sup>−</sup> animals exhibited an increased hindpaw stand index [*t*(18) = 2.53, *p* = 0.02], indicating greater duration stance than PGC-1α+/<sup>+</sup> mice (**Figure 1B**). Similarly, the average speed in the swing phase of the step cycle, or swing speed, was found to be decreased for the forepaws [*t*(18) = 3.74, *p* = <sup>0</sup>.002] and hindpaws [*t*(18) <sup>=</sup> <sup>3</sup>.25, *<sup>p</sup>* <sup>=</sup> <sup>0</sup>.004] of PGC-1α−/<sup>−</sup> mice (**Figure 1C**). Missteps indicate an instance in the step cycle wherein a paw was not placed in a step sequence, and PGC-1α−/<sup>−</sup> mice exhibited an increased number of missteps compared to their wildtype littermates [*t*(15) = 2.58, *p* = 0.02; **Figure 1D**]. **Figure 1E** is comprised of representative traces showing colorcoded, digitized paw prints and corresponding step cycles for PGC-1α−/<sup>−</sup> and PGC-1α+/<sup>+</sup> mice. Taken together, these data demonstrate that the motor phenotype of PGC-1α−/<sup>−</sup> mice includes ataxia characterized by altered gait kinematics, including increased stance duration, increased paw placement, and stepping mistakes.

#### **NOVEL CEREBELLAR PGC-1α-DEPENDENT GENES**

Little is known about the downstream gene targets of PGC-1α in the cerebellum, despite the high concentration of PGC-1α in this brain region (Cowell et al., 2007). To identify novel PGC-1αdependent transcripts in the cerebellum, we used an approach our laboratory recently used to identify PGC-1α-dependent genes in the cortex in which we mined microarray data comparing human SH-SY5Y neuroblastoma cells overexpressing PGC-1α and GFP in tandem to cells expressing GFP alone (Lucas et al., 2014; GEO NCBI database registration in progress). Transcripts were selected to measure in cerebellar homogenates based on three criteria: (1) all of the top 10 transcripts significantly upregulated by PGC-1α overexpression with a murine homolog, (2) transcripts listed on the whole-brain PGC-1α Neuroblast feature from the Allen Brain Atlas, and (3) genes listed on the wholebrain parvalbumin (PV) Neuroblast feature from the Allen Brain Atlas. The Neuroblast feature of the Allen Brain Atlas identifies genes with similar 3D spatial expression profiles by conducting Pearson's correlation coefficients of the expression intensity of each gene with other genes coexpressed in 200μm cubes (Lein et al., 2007). Although PGC-1α is not required for the expression of PV in the cerebellum as it is in the forebrain (Lucas et al., 2010), the PV Neuroblast was implemented because of the significant overlap between the neuroanatomical localization of PV and PGC-1α in this brain region; transcript and protein expression of both PGC-1α and PV is enriched Purkinje cells, molecular layer interneurons, and neurons of the deep cerebellar nuclei (Celio, 1990; Cowell et al., 2007). Mined transcripts fitting our three criteria but with unknown functions or ubiquitous expression patterns when viewed on the Allen Brain Atlas were excluded from subsequent analysis (Supplementary Table 2). The final list consisted of 37 transcripts (Supplementary Table 3).

Expression levels of all 37 transcripts were measured, and eight transcripts were found to be significantly reduced in cerebellar homogenates from PGC-1α−/<sup>−</sup> animals compared to their littermate controls at 30 days of age (**Figure 2**). These included the synaptic transcripts complexin 1 [Cplx1; *F*(2, 14) = 8.84, *p* = 0.005] and pacsin2 [*F*(2, 31) = 4.66, *p* = 0.02], the structural transcript neurofilament heavy chain [Nefh; *F*(2, 15) = 11.87, *p* = 0.001], and the metabolism-related transcripts isocitrate dehydrogenase 3a [Idh3a; *F*(2, 15) = 8.75, *p* = 0.004], neutral cholesterol ester hydrolase 1 [Nceh1; *F*(2, 31) = 6.42, *p* = 0.005], pyruvate dehydrogenase alpha 1 [Pdha1; *F*(2, 31) = 6.77, *p* = 0.004], phytanoyl-CoA hydroxylase [Phyh; *F*(2, 31) = 11.46, *p* = 0.0002], and ubiquinol-cytochrome c reductase, Rieske ironsulfur polypeptide 1 [Uqcrfs1; *F*(2, 31) = 4.47, *p* = 0.02]. Phyh was the only gene to exhibit a dose-dependency with PGC-1α expression, with significant differences between all genotypes, while deletion of one allele was sufficient to cause decreases in Cplx1, Idh3a, Pdha1, and Uqcrfs1 (Fisher's LSD, *p* < 0.05).

#### **NEUROANATOMICAL OVERLAP OF PGC-1A WITH NOVEL PGC-1α-DEPENDENT GENES IN THE MOUSE CEREBELLUM**

If Cplx1, Pacsin2, Nefh, Idh3a, Nceh1, Pdha1, Phyh, and Uqcrfs1 are truly PGC-1α-dependent genes, we surmised that they would exhibit neuroanatomical overlap with PGC-1α in the cerebellum. With the exception of Cplx1, Pacsin2, and Nefh, antibodies against the protein products of PGC-1α-dependent transcripts are either not commercially available or do not work well for immunohistochemistry. Therefore, we collected *in situ* hybridization images published on the Allen Brain Atlas (http://www. brain-map.org) to determine the cellular localization of PGC-1α and its putative targets in the adult mouse cerebellum (**Figure 3**). At low magnification, PGC-1α mRNA exhibited high expression in the Purkinje cell layer (PCL; arrowheads) and deep cerebellar nuclei. Higher magnification of PGC-1α images in the cerebellar cortex revealed high expression in the PCL, moderate expression in the molecular layer (ML; arrows), and non-specific background expression in the granule cell layer (GCL), consistent with published data on PGC-1α protein expression in this brain region

**FIGURE 1 | Ataxia characterized by altered gait kinematics in mice lacking PGC-1α.** CatWalk gait analysis was performed on male PGC-1α+/<sup>+</sup> and <sup>−</sup>/<sup>−</sup> littermates at 6 months of age. PGC-1α−/<sup>−</sup> mice exhibited decreased hindpaw area **(A)**, increased hindpaw stand index **(B)**, decreased forepaw and hindpaw swing speed **(C)**, and an increased number of missteps **(D)**

**FIGURE 2 | Novel cerebellar PGC-1α-dependent genes.** q-RT-PCR of microarray-identified PGC-1α regulated transcripts was performed on cerebellar homogenates from PGC-1α+/+, <sup>+</sup>/−, and <sup>−</sup>/<sup>−</sup> littermates at 30 days of age. Expression of transcripts spanning synaptic (Cplx1, Pacsin2; **(A)**, structural [Nefh; **(B)**], and metabolic [Idh3a, Nceh1,

Pdha1, Phyh, Uqcrfs1; **(C)**] functions were significantly decreased in PGC-1α−/<sup>−</sup> compared to littermate control cerebella. One-Way ANOVA followed by Fisher's LSD. <sup>∗</sup>*p* < 0.05, ∗∗*p* < 0.005, ∗∗∗*p* < 0.0005. n/group indicated on last bar histogram. Data are presented as mean ± SEM.

**FIGURE 3 | Neuroanatomical mRNA localization of PGC-1α and its dependent genes in the cerebellum.** Sagittal *in situ* hybridization images of PGC-1α **(A)**, Cplx1 **(B)**, Pacsin2 **(C)**, Nefh **(D)**, Idh3a **(E)**, Nceh1 **(F)**, Pdha1 **(G)**, Phyh **(H)**, and Uqcrfs1 **(I)** transcript expression in the adult mouse cerebellum were collected from the Allen Brain Atlas (http://www.brain-map. org). All transcripts are highly expressed in the Purkinje cell layer (PCL;

arrowheads) and the fastigial (FN) and interposed (IN) deep cerebellar nuclei. Most transcripts also exhibited expression in the molecular layer (ML; arrows), and Cplx1, Nefh, Phyh mRNA exhibited expression in the granule cell layer (GCL; concave arrowheads). Boxed portion in A represents area of higher magnification images in **A –I** . Scale bar = 525μm for **A–I** and 100μm for **A –F** .

(Cowell et al., 2007). All eight PGC-1α-dependent transcripts exhibited high expression in the PCL and deep nuclei. Pascin2, Idh3a, Nceh1, Pdha1, and Uqcrfs1 most closely resembled the overall spatial pattern of PGC-1α, while Cplx1, Nefh, and Phyh exhibited higher expression in the GCL (concave arrowheads).

#### **REGION-SPECIFIC LOSS OF Cplx1 AND NEFH PROTEIN EXPRESSION IN MICE LACKING PGC-1α**

To investigate the likelihood of these changes in transcription influencing cell function, we determined whether protein, as well as transcript, expression of putative downstream targets of PGC-1α was decreased in the cerebellum of PGC-1α−/<sup>−</sup> animals. To initially quantify overall protein levels, we conducted Western blot analysis on cerebellar homogenates from +/+ and −/− animals at 1 month of age with commercially available antibodies validated in our laboratory (Lucas et al., 2014). Of the tested antibodies, only Cplx1, Nefh, and Pacsin2 produced specific bands at the correct molecular weight. Expression of Cplx1 [*t*(12) = 2.32, *p* = 0.02] and Nefh [*t*(12) = 2.16, *p* = 0.03], but not Pacsin2, was significantly decreased in PGC-1α−/<sup>−</sup> compared to <sup>+</sup>/<sup>+</sup> littermates (**Figure 4**).

To localize the changes in protein expression to specific cell types, we conducted immunofluorescence colabeling of Nefh and Cplx with GAD67 in PGC-1α+/<sup>+</sup> and <sup>−</sup>/<sup>−</sup> mice (**Figure 5**). In <sup>+</sup>/<sup>+</sup> animals, Nefh expression was highly concentrated in the PCL with clear labeling in Purkinje cell bodies and their dendritic trees extending into the ML. However, in <sup>−</sup>/<sup>−</sup> animals, Nefh labeling of Purkinje cells was greatly reduced, both in the soma and the dendrites (**Figure 5A**). Contradictory to transcript expression of Cplx1 (**Figure 2B**), labeling with an antibody that recognizes both Cplx1 and Cplx2 did not reveal appreciable staining of Cplx in the PCL. In both <sup>+</sup>/<sup>+</sup> and <sup>−</sup>/<sup>−</sup> animals, Cplx was colocalized to GAD67-positive puncta surrounding Purkinje cell bodies, and this expression did not appear to differ between <sup>+</sup>/<sup>+</sup> and <sup>−</sup>/<sup>−</sup> mice (**Figure 5B**). In the deep cerebellar nuclei, however, there was a high intensity of Cplx staining that was greatly reduced by deletion of PGC-1α (**Figure 5C**, interposed nucleus shown). Cplx-positive cell bodies were not GAD67-positive in this region, but Cplx exhibited a high degree of colocalization with GAD67-positive processes, presumably from Purkinje cell axon terminals. Cplx staining was greatly reduced in both the GAD67-positive processes and GAD67-negative cell bodies of deep cerebellar nuclei of <sup>−</sup>/<sup>−</sup> animals. Nefh expression, on the other hand, did not appear to differ between <sup>+</sup>/<sup>+</sup> and <sup>−</sup>/<sup>−</sup> mice in the deep cerebellar nuclei (data not shown).

#### **CONDITIONAL DELETION OF PGC-1α IN PV-POSITIVE CELLS DISRUPTS TRANSCRIPTION OF CEREBELLAR PGC-1α-DEPENDENT GENES**

Our immunofluorescence experiments suggest that the effects of PGC-1α ablation appear to be specific to different cellular populations, depending on the target transcript. However, the differential effects of loss of PGC-1α in distinct cellular populations are consistent with the previously published protein (Cowell et al., 2007) and mRNA (**Figure 3**) expression patterns demonstrating high levels of PGC-1α in Purkinje cells, interneurons in the ML, and cells of the deep cerebellar nuclei. Interestingly, all of these populations express high levels of the calcium binding protein parvalbumin (PV), so to test whether deletion of PGC-1α specifically within these neuronal populations could recapitulate the transcriptional deficiencies observed in the germline knockout animal, we crossed mice with LoxP sites flanking the 3–5 exon region of the PGC-1α gene (Lin et al., 2004) with mice expressing Cre recombinase driven by the PV promoter (Hippenmeyer et al., 2005). Cell-specific recombination was confirmed using reporter mice that express red fluorescent protein in the absence of recombination and green fluorescent protein (GFP) in the presence of recombination (Muzumdar et al., 2007). At 3 months of age, reporter mice exhibited GFP expression in the PCL and ML but not the GCL of the cerebellar cortex and in the deep cerebellar nuclei (**Figures 6A,B**), with the highest GFP expression in Purkinje cell bodies and axon terminals in the deep nuclei.

tomato/mutant green reporter mice to determine the specificity of Cre-mediated recombination at 3 months of age. PV expression is pseudocolored red from Cy5. **(A)** Green fluorescent protein (GFP) expression was high in the Purkinje cell layer (PCL; arrowheads), moderate in the molecular layer (ML), and low in the granule cell layer (GCL). Scale bar = 25μm. **(B)** GFP expression was greatest in the deep cerebellar nuclei (interposed nucleus shown) and exhibited a high degree of colocalization to PV-positive terminals and, to a lesser extent, PV-positive cell bodies (arrowheads). Note the absence of GFP expression outside the borders of the nucleus (arrows). Scale bar = 75μm. **(C)** Gene expression of PGC-1α and its putative targets was measured in cerebellar homogenates from 6-month-old PGC-1αWT and fl/fl mice expressing PV-Cre by q-RT-PCR. Expression of PGC-1α, Cplx1, Nefh, Idh3a, and Uqcrfs1 were significantly reduced in PGC-1αfl/fl:PV-Cre mice. One-tailed *t*-tests. ∗∗∗*p* < 0.0005, <sup>∗</sup>*p* < 0.05. n/group indicated on last bar histogram. Data presented as mean ± SEM.

Gene expression was measured by q-RT-PCR in cerebellar homogenates from PGC-1αWT:PV-Cre and PGC-1αfl/fl:PV-Cre littermates at 6 months of age (**Figure 6C**). Conditional deletion of PGC-1α by PV-Cre resulted in a 50% decrease in the expression of PGC-1<sup>α</sup> [*t*(23) <sup>=</sup> <sup>7</sup>.69, *<sup>p</sup>* <sup>=</sup> 4.21 <sup>×</sup> <sup>10</sup>−8] and also significantly reduced expression of Cplx1 [*t*(23) <sup>=</sup> <sup>8</sup>.57, *<sup>p</sup>* <sup>=</sup> 6.52 <sup>×</sup> <sup>10</sup>−9], Nefh [*t*(23) <sup>=</sup> <sup>5</sup>.29, *<sup>p</sup>* <sup>=</sup> <sup>1</sup>.<sup>34</sup> <sup>×</sup> <sup>10</sup>−5], Idh3a [*t*(23) <sup>=</sup> <sup>7</sup>.48, *<sup>p</sup>* <sup>=</sup> <sup>6</sup>.<sup>68</sup> <sup>×</sup> <sup>10</sup>−8; **Figure 6C**], and Uqcrfs1 [*t*(20) <sup>=</sup> <sup>1</sup>.83, *<sup>p</sup>* <sup>=</sup> <sup>0</sup>.04]. Interestingly, Pacsin2, **Nceh1,** Pdha1, and Phyh gene expression was unaffected by conditional deletion of PGC-1α, despite their apparent concentration in Purkinje cells (**Figure 2**). It is possible that Pacsin2, Nceh1, Pdha1, and Phyh are dependent on PGC-1α in PV-negative cell types or that cell-cell interactions in the complete knockout are required for decreases in their expression.

#### **PGC-1α ABLATION DECREASES PURKINJE CELL VIABILITY AND FUNCTION**

Purkinje cells exhibit a particularly high concentration of PGC-1α (Cowell et al., 2007), and our immunofluorescence and conditional knockout experiments suggest that Purkinje cells are dependent on PGC-1α for transcriptional regulation. Thus, we questioned whether the combined decrease of synaptic, structural, and metabolism-related genes caused by ablation of PGC-1α would lead to decreased neuronal viability and/or function in this cellular population. As a measure of Purkinje cell viability, we conducted stereological counts of H&E stained Purkinje cells in PGC-1α+/<sup>+</sup> and <sup>−</sup>/<sup>−</sup> mice at 6 weeks of age, a time point after the onset of the previously described neurodegeneration in the striatum of this mouse line (Lucas et al., 2012). H&E labeling of Purkinje cells in PGC-1α+/<sup>+</sup> animals exhibited a clear nuclear haematoxylin stain, while the haematoxylin stain of Purkinje cells in <sup>−</sup>/<sup>−</sup> mice was often very faint and cell bodies were at times undetectable between the ML and GCL (**Figure 7A**). Stereological estimation with the optical fractionator method demonstrated <sup>a</sup> <sup>∼</sup>30% reduction of the total number of Purkinje cells in <sup>−</sup>/<sup>−</sup> compared to <sup>+</sup>/<sup>+</sup> animals [*t*(8) <sup>=</sup> <sup>3</sup>.83, *<sup>p</sup>* <sup>=</sup> <sup>0</sup>.005; **Figure 7B**].

Purkinje cells are one of the few neuronal populations that are spontaneously active in the absence of synaptic input (Hausser and Clark, 1997), allowing their firing rates to be measured in the absence of stimulation or drug application. As a measure of Purkinje cell basal function, we performed loose patch electrophysiology on Purkinje cells *in vitro* in acute cerebellar slices from PGC-1α+/<sup>+</sup> and <sup>−</sup>/<sup>−</sup> animals at 6 weeks of age. Slices were perfused with oxygenated ACSF; once patched, Purkinje cells were allowed to acclimate to the pipette for 3 min to avoid events in response to mechanical stimulation. Spontaneous events were then recorded for 3 min with no stimulation (**Figure 7C**). We observed a 50% reduction in Purkinje cell spike rate in PGC-1α−/<sup>−</sup> compared to <sup>+</sup>/<sup>+</sup> animals [*t*(5) <sup>=</sup> <sup>3</sup>.09, *<sup>p</sup>* <sup>=</sup> <sup>0</sup>.03; **Figure 7D**].

# **DISCUSSION**

In this manuscript, we present data showing that mice lacking PGC-1α exhibit a behavioral phenotype indicative of cerebellar dysfunction and show reduced cerebellar expression of transcripts involved in synaptic (Cplx1, Pacsin2), structural (Nefh), and metabolic (Idh3a, Nceh1, Pdha1, Phyh, and Uqcrfs1) functions. Using mice expressing Cre recombinase under the control of the PV promoter, we postnatally ablated PGC-1α in neuronal types enriched in its expression and observed reduced transcript expression of PGC-1α and its dependent genes Cplx1, Nefh, Idh3a, and Uqcrsf1, suggesting that PGC-1α regulation of these transcripts is specific to Purkinje cells, interneurons of the cerebellar cortex, and neurons of the deep cerebellar nuclei (Supplementary Table 4). Recombination in our conditional

undetectable stain in the Purkinje cell layer (PCL) of <sup>−</sup>/<sup>−</sup> animals. GCL,

Data presented as mean ± SEM.

knockout animals was highest in Purkinje cells, and germline deletion of PGC-1α resulted in Purkinje cell loss and physiological dysfunction.

As the novel cerebellar PGC-1α-dependent transcripts are both upregulated by PGC-1α overexpression (*in vitro* data from Lucas et al., 2014) and reduced by its deletion (*in vivo* data from **Figure 2**), we hypothesize that Cplx1, Pacsin2, Nefh, Idh3a, Nceh1, Pdha1, Phyh, and Uqcrfs1 are all directly regulated by PGC-1α. To definitively determine whether these transcripts are direct targets of PGC-1α, chromatin immunoprecipitation assays are required; however, considering the cell type-specific differences in vulnerability to PGC-1α loss, these experiments are impractical. An alternative approach would be to identify the common intermediate transcription factors that PGC-1α interacts with to drive their expression. It is possible that genes within certain functional categories (neurotransmitter release, axonal structural support, and metabolism) are driven by PGC-1α's interaction with distinct sets of transcriptional regulators; for example, transcription of metabolic genes may be driven by the PGC-1α-interacting factor nuclear respiratory factor 1 (Wu et al., 1999). Further elucidation of the PGC-1α interactome may clarify the mechanisms by which PGC-1α can regulate parallel programs for functional, structural, and metabolic maturation.

Cplx1 and Nefh, but not Pascin2, were also decreased at the protein level in the cerebellum of PGC-1α−/<sup>−</sup> mice. When we investigated the cellular distribution of protein expression, we found that Nefh was decreased in Purkinje cell bodies and dendritic processes, while Cplx was decreased in the deep cerebellar nuclei. Contradictory to the transcript pattern of Cplx1, very little Cplx protein was observed in the PCL. It is possible that the reduction in Cplx protein expression that was observed in the deep cerebellar nuclei was due to a loss of this synaptic protein in the axonal projections of Purkinje cells, whose sole synaptic targets are the deep cerebellar nuclei. However, we did not observe changes in Nefh expression in the white matter tracts from the Purkinje cells to the deep cerebellar nuclei, suggesting that the Purkinje cell axons themselves are still present. At the synapse, Cplx1 mediates synchronous neurotransmitter release by clamping the vesicle to the presynaptic membrane in a primed but unfused manner, thus controlling the timing of vesicular exocytosis (Sudhof and Rothman, 2009). Future experiments should be conducted to determine whether neurotransmitter release from Purkinje cells to the deep cerebellar nuclei is asynchronous, as abnormalities in synchronous neurotransmitter release were observed in the motor cortex of PGC-1αfl/fl:PV-Cre mice (Lucas et al., 2014).

While it has been established that loss of PGC-1α leads to vacuolization in motor regions of the forebrain (Lin et al., 2004; Leone et al., 2005; Lucas et al., 2012), no investigations of PGC-1α loss on cell viability have been conducted in the cerebellum to date. We here show a ∼30% reduction in Purkinje cell number in PGC-1α−/<sup>−</sup> mice, indicating that Purkinje cells are vulnerable to cell death in the absence of PGC-1α. Of the novel PGC-1α-dependent genes identified in this manuscript, Phyh is the only gene that has been shown to influence Purkinje cell viability. Mutations in the Phyh gene result in an autosomal recessive lipid storage disorder known as Refsum disease (Jansen et al., 1997). Clinically, patients present with severe cerebellar ataxia (Wierzbicki, 2007), and loss of Purkinje cells accompanied by ataxia has been reported in a Phyh <sup>−</sup>/<sup>−</sup> mouse model of this disease (Ferdinandusse et al., 2008).

While decreased Phyh expression is the most direct link between loss of PGC-1α and reduced Purkinje cell viability, it is also possible that compromised metabolism contributes to the observed cell loss. We previously found that the mitochondrial fusion gene mitofusin 2 (Mfn2) and the antioxidant gene manganese superoxide dismutase (MnSOD), both wellestablished targets of PGC-1α, were significantly reduced in cerebellar homogenates from PGC-1α−/<sup>−</sup> mice (Lucas et al., 2010). Here, we reported that Uqcrfs1 is also a cerebellar PGC-1αdependent gene. Uqcrfs1 is a component of complex III of the mitochondrial electron transport chain, and while other members of complex III can associate to form a complex in the absence of uqcrfs1, the resultant complex has limited enzymatic activity, leading to the degradation of complexes I and IV (Diaz et al., 2012). Thus, the combined decreases in expression of the metabolism-related genes Phyh, Idh3a, Mfn2, MnSOD, and Uqcrfs1 may lead to abnormalities in energy homeostasis or the increased production of reactive oxygen species, contributing to decreased viability of Purkinje cells of PGC-1α−/<sup>−</sup> animals. Reduced energy production could also explain the decrease in spike rate, as none of the other PGC-1α-dependent genes identified to date, alone or in combination, have been reported to produce this physiological abnormality. Of note, cortical PVpositive interneurons also exhibit a decrease in firing rate in PGC-1α−/<sup>−</sup> mice (Dougherty et al., 2014).

We previously reported that, in contrast to the cerebrum, PV expression in the cerebellum is not dependent on the presence of PGC-1α (Lucas et al., 2010). Here, no changes in the expression of synaptotagmin 2 (Syt2) were observed in the cerebellum of PGC-1α−/<sup>−</sup> mice (Supplementary Table 3), while Syt2 is reduced by ∼80% in the cortex of these mice (Lucas et al., 2014). This suggests that the mechanisms of PGC-1α-mediated gene regulation differ by region and cell type in the brain. Since PGC-1α's gene targets are defined by the transcription factors with which it interacts, it is likely that a different complement of factors interact with PGC-1α in cortical neurons than in cerebellar neurons. Multiple PGC-1α-interacting factors such as CREB, myocyte enhancing factor 2, and thyroid hormone receptor are expressed highly in the cerebellum; identification of the predominant factors that interact with PGC-1α in individual cell types would assist in predicting the specific subset of transcripts that would be affected by PGC-1α deficiency.

Regarding our observation of ataxia in PGC-1α−/<sup>−</sup> mice, it is not currently clear which cell populations are contributing to this phenotype. A recent publication suggests that gait abnormalities in PGC-1α−/<sup>−</sup> mice are more consistent with Purkinje cell dysfunction than dysfunction of granule cells or molecular layer interneurons (Vinueza Veloz et al., 2014). However, we did not perform CatWalk analysis on PGC-1αfl/fl:PV-Cre mice because our previous study did not reveal deficits in locomotor activity or motor coordination in these animals (Lucas et al., 2014). It is possible that the ataxia phenotype arises from cells that do not express PV, such as granule cells. As granule cell loss has been reported in Cplx1 <sup>−</sup>/<sup>−</sup> mice (Kielar et al., 2012), which exhibit profound ataxia (Glynn et al., 2005), PGC-1α may play an important role in regulation of Cplx1 and other dependent genes in granule cells. Other cells that could contribute to the ataxia phenotype could be motor neurons; it has not been shown whether motor neurons express PGC-1α, although overexpression of PGC-1α is capable of slowing disease progression in a mouse model of amyotrophic lateral sclerosis (Liang et al., 2011). A preliminary analysis of PGC-1α mRNA localization in the spinal cord using Allen Brain Atlas shows that PGC-1α mRNA is present in a subset of large neurons in the ventral gray matter of the spinal cord; gene expression loss, especially of Nefh which is highly concentrated in those cells, could be contributing to some of the motor deficits. Muscle cells are also affected by a lack of PGC-1α, although mice with muscle-specific deletion show only subtle motor impairments (Handschin et al., 2007).

It is also important to consider age differences when interpreting our behavioral data (conducted at 6 months of age) in light of our molecular findings (conducted at 4–6 weeks of age). Due to significantly decreased weight in PGC-1α−/<sup>−</sup> mice at younger ages, gait analysis was conducted on mice at 6 months of age, when weight does not differ from PGC-1α+/<sup>+</sup> mice, as differences in weight between genotypes can bias paw print detection with the automated Catwalk program. While we previously observed that striatal vacuolization and impaired rotarod performance does not significantly improve or decline between 1 and 3 months of age (Lucas et al., 2012), it is possible, given the additional metabolic targets in the cerebellum compared to the forebrain, that gait abnormalities are more pronounced at 6 months than earlier ages. While we did not perform transcriptional analysis from cerebella of 6 month-old PGC-1a −/− mice, we would predict that the putative targets that are reduced at 30 days of age would still be reduced at later ages (as demonstrated for the striatum; Lucas et al., 2012), but that other "off-target" transcripts may also be affected, due to compensatory mechanisms. Future studies will evaluate balance and gait in a cohort of younger animals with semi-quantitative approaches (Guyenet et al., 2010).

Recently, PGC-1α was also found to be involved in the pathology of Friedreich Ataxia. Friedreich Ataxia is an autosomal recessive disorder and is the most common of the early-onset hereditary ataxias (Pandolfo, 2009). Friedreich Ataxia is caused by an expanded GAA triplet repeat in the first intron of the frataxin gene, resulting in a 65–95% reduction in frataxin expression in Friedreich Ataxia patients (Campuzano et al., 1996). Transcript expression of PGC-1α and its downstream target MnSOD was found to be decreased in fibroblasts and lymphoblasts of patients with Friedreich Ataxia (Coppola et al., 2009; Marmolino et al., 2010). In fact, transcript expression of frataxin is highly correlated with that of PGC-1α in Friedreich Ataxia and control cells, and knock down of frataxin in cell culture also decreases transcript expression of PGC-1α (Coppola et al., 2009).

Related to the potential roles for PGC-1α in cerebellar gene dysregulation in other movement disorders, we have recently observed that the firing rate of Purkinje cells is reduced in transgenic and knock-in mouse models of Huntington Disease (Dougherty et al., 2012, 2013). Due to the interconnectivity between the basal ganglia and cerebellum and their joint roles in motor coordination and gait control, it is possible that cerebellar deficits contribute to the motor symptoms in Parkinson Disease as well (Wu and Hallett, 2013). Of note, the kinematic ataxia phenotype observed in PGC-1α−/<sup>−</sup> mice (**Figure 1**) closely resembles the gait abnormalities reported in rodent models of both Huntington (Vandeputte et al., 2010) and Parkinson (Vlamings et al., 2007; Chuang et al., 2010) Diseases. Further experiments are required to determine whether dysregulation of PGC-1αdependent programs of gene expression underlies some of the cell- and circuit-specific vulnerability in these disorders.

#### **ACKNOWLEDGMENTS**

This work was funded by National Institutes of Health (NIH) Grants 5K01MH077955-05 and 5R01NS07009-05 (Rita M. Cowell), a Civitan Emerging Scholar Award (Elizabeth K. Lucas) and a Civitan McNulty Scientist Award (Rita M. Cowell). Many thanks are due to John Hablitz, Karen Gamble, and Andrew Bohannon for electrophysiology guidance and resources, David Standaert for use of his confocal microscope, and the UAB Histology and Microscopy Core (NIH Blueprint Core Grant P30 NS57466).

#### **SUPPLEMENTARY MATERIAL**

The Supplementary Material for this article can be found online at: http://www.frontiersin.org/journal/10.3389/fncel. 2014.00441/abstract

#### **REFERENCES**


intervention in Parkinson's disease. *Sci. Transl. Med.* 2:52ra73. doi: 10.1126/scitranslmed.3001059

**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 October 2014; paper pending published: 07 November 2014; accepted: 08 December 2014; published online: 06 January 2015.*

*Citation: Lucas EK, Reid CS, McMeekin LJ, Dougherty SE, Floyd CL and Cowell RM (2015) Cerebellar transcriptional alterations with Purkinje cell dysfunction and loss in mice lacking PGC-1*α*. Front. Cell. Neurosci. 8:441. doi: 10.3389/fncel.2014.00441 This article was submitted to the journal Frontiers in Cellular Neuroscience.*

*Copyright © 2015 Lucas, Reid, McMeekin, Dougherty, Floyd and Cowell. 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.*

# Differential expression of sirtuins in the aging rat brain

Nady Braidy <sup>1</sup> , Anne Poljak 2, 3, Ross Grant 2, 4, Tharusha Jayasena<sup>1</sup> , Hussein Mansour <sup>5</sup> , Tailoi Chan-Ling<sup>5</sup> , George Smythe2, 3, Perminder Sachdev 1, 6 and Gilles J. Guillemin<sup>7</sup> \*

<sup>1</sup> Centre for Healthy Brain Ageing, School of Psychiatry, University of New South Wales, Sydney, NSW, Australia, <sup>2</sup> Faculty of Medicine, School of Medical Sciences, University of New South Wales, Sydney, NSW, Australia, <sup>3</sup> Bioanalytical Mass Spectrometry Facility, University of New South Wales, Sydney, NSW, Australia, <sup>4</sup> Australasian Research Institute, Sydney Adventist Hospital, Sydney, NSW, Australia, <sup>5</sup> Retinal and Developmental Neurobiology Lab, Discipline of Anatomy and Histology, School of Medical Sciences, University of Sydney, NSW, Australia, <sup>6</sup> Neuropsychiatric Institute, Prince of Wales Hospital, Sydney, NSW, Australia, <sup>7</sup> Neuropharmacology Group, MND and Neurodegenerative Diseases Research Centre, Macquarie University North Ryde, NSW, Australia

Although there are seven mammalian sirtuins (SIRT1-7), little is known about their expression in the aging brain. To characterize the change(s) in mRNA and protein expression of SIRT1-7 and their associated proteins in the brain of "physiologically" aged Wistar rats. We tested mRNA and protein expression levels of rat SIRT1-7, and the levels of associated proteins in the brain using RT-PCR and western blotting. Our data shows that SIRT1 expression increases with age, concurrently with increased acetylated p53 levels in all brain regions investigated. SIRT2 and FOXO3a protein levels increased only in the occipital lobe. SIRT3-5 expression declined significantly in the hippocampus and frontal lobe, associated with increases in superoxide and fatty acid oxidation levels, and acetylated CPS-1 protein expression, and a reduction in MnSOD level. While SIRT6 expression declines significantly with age acetylated H3K9 protein expression is increased throughout the brain. SIRT7 and Pol I protein expression increased in the frontal lobe. This study identifies previously unknown roles for sirtuins in regulating cellular homeostasis and healthy aging.

Keywords: aging, sirtuins, p53, brain, longevity

# Introduction

Sirtuins or "silent information regulators" of gene transcription, are a family of enzymes which are expressed throughout all phyla of life. Accumulating evidence suggests that this unique class of histone deacetylases are key regulators of numerous physiological processes, particularly in aging in multiple organisms (Porcu and Chiarugi, 2005; Berdichevsky and Guarente, 2006; Longo and Kennedy, 2006; Smith and Denu, 2006; Pallas et al., 2008; Schwer and Verdin, 2008). Gene silencing by this family of enzymes has been correlated directly with longer lifespan in yeast and worms (Yang and Sauve, 2005). In yeast, Sir2 plays a critical role in transcriptional silencing and in genomic stability (Lamming et al., 2004; Denu, 2005). The key question is whether sirtuins regulate healthier aging in mammals.

Seven Sir2 homologs (SIRT1-7) have been identified in mammals (Porcu and Chiarugi, 2005; Pallas et al., 2008). Mammalian sirtuins have diverse locations and multiple targets, and affect a broad range of cellular processes (Supplementary Figure 1). SIRT1 is the human homolog of sir2 and appears to be involved in several physiological functions including the control of gene

#### Edited by:

Victoria Campos-Peña, Instituto Nacional de Neurologia y Neurocirugia, Mexico

#### Reviewed by:

Graziamaria Corbi, University of Molise, Italy Claus Jürgen Scholz, University of Würzburg, Germany

#### \*Correspondence:

Gilles J. Guillemin, Neuropharmacology Group, MND and Neurodegenerative Diseases Research Centre, Australian School of Advanced Medicine, Macquarie University, Balaclava Road, North Ryde, NSW 2109, Australia gilles.guillemin@mq.edu.au

> Received: 28 October 2014 Accepted: 15 April 2015 Published: 08 May 2015

#### Citation:

Braidy N, Poljak A, Grant R, Jayasena T, Mansour H, Chan-Ling T, Smythe G, Sachdev P and Guillemin GJ (2015) Differential expression of sirtuins in the aging rat brain. Front. Cell. Neurosci. 9:167. doi: 10.3389/fncel.2015.00167 expression, cell cycle regulation, apoptosis, DNA repair, metabolism, and aging (Anastasiou and Krek, 2006; Qin et al., 2006; Smith and Denu, 2006). SIRT1 is localized in the nucleus and can deacetylate numerous proteins such as tumor suppressor protein (p53), Ku70, NF-κB, and forkhead proteins which modulate genes that control cellular stress resistance (Smith, 2002). The deacetylase activity of specific sirtuin proteins is dependent on the intracellular NAD<sup>+</sup> content (Sauve et al., 2006). They catalyze a unique reaction that releases nicotinamide, acetyl ADP-ribose (AADPR), and the deacetylated substrate (Sauve et al., 2006). It has been shown that increased SIRT1 activity in human cells can delay apoptosis and rescue vulnerable cells with additional time to repair after repeated exposure to oxidative stress (Howitz et al., 2003).

Mammalian SIRT2 is predominantly a cytoplasmic protein (North et al., 2003). It can deacetylate several cytoskeletal proteins, including α-tubulin, histones, and forkhead proteins, although the physiological effect of deacetylation of these proteins by SIRT2 remains unknown (Choudhuri et al., 2003; North et al., 2003; Brunet et al., 2004; van der Horst et al., 2004). SIRT2 protein expression levels also appear to increase during the mitotic phase of the cell cycle, and overexpression can delay mitosis (Dryden et al., 2003). Consistent with the idea that SIRT2 can protect against neurodegenerative pathology in mouse models of Alzheimer's disease, overactivation of SIRT2 has been shown to protect against axonopathy and neurodegeneration in a mouse model of Wallerian degeneration (Arraki et al., 2004; Tang and Chua, 2008). Small-molecule inhibitors targeting SIRT2 have been shown to attenuate several models of neurodegeneration in vivo (Choudhuri et al., 2003; North et al., 2003; Brunet et al., 2004; van der Horst et al., 2004). Therefore, it is of significant interest to determine the anatomical and functional changes of SIRT2 in the brain during aging.

Mitochondria represent the primary site for the production of reactive oxygen species (ROS) through one-electron carriers in the respiratory chain (Beal, 1995). This dynamic organelle is highly vulnerable to the cytotoxic effect of oxidative stress, as evidenced by extensive lipid peroxidation, protein oxidation and mitochondrial DNA (mtDNA) mutations (Beal, 1995, 2007; Mawrin et al., 2003; Andersen, 2004). Experimental evidence of respiratory chain defects (Braidy et al., 2011) are in accordance with the mitochondrial theory of aging. The role of the mitochondrial SIRT3-5 is of great interest with regard to mammalian aging and age-related brain disorders. Do mammalian sirtuins regulate metabolism and the oxidative stress response in the brain during aging? Recently, Ozden et al. (2011) showed that SIRT3 responds to changes in mitochondrial redox status by altering the enzymatic activity of specific downstream targets, including manganese superoxide dismutase (MnSOD). MnSOD is the primary mitochondrial ROS scavenging enzyme which modulates ROS levels as well as metabolic homeostatic poise (Ozden et al., 2011). However, no study to our knowledge has examined SIRT3-5 mRNA and protein expression levels over a mammalian aging time course.

SIRT6, is another nuclear specific protein, is a histone H3 lysine 9 (H3K9) deacetylase, that is necessary to promote longevity (Liszt et al., 2005; Mostoslavsky et al., 2006; Koltai et al., 2010). SIRT6 knockout mice display premature aging symptoms, including excessive loss of subcutaneous fat and a significant reduction in bone density, and die within 4 weeks of birth (Mostoslavsky et al., 2006). These animals also demonstrate impaired DNA base excision repair (BER) and several metabolic phenotypes (Mostoslavsky et al., 2006). However, the mechanism by which SIRT6 regulates BER remains unknown. It is interesting to know whether metabolic changes during aging are associated with SIRT6.

The least characterized of the sirtuins, SIRT7, is localized in the nucleolus of mammalian cells. Ford et al. (2006) showed that SIRT7 protein expression levels correlated with tissue proliferation and its expression is reduced in non-proliferating tissue, such as the heart, brain and muscle (Ford et al., 2006). Recently, SIRT7 has been associated with cellular growth and metabolism (Ford et al., 2006). In particular, SIRT7 has been associated with rDNA and interacts with RNA polymerase I (Pol I), suggesting a role in NAD-dependent regulation (Ford et al., 2006). Interestingly, both SIRT7, an inducer of rRNA transcription, and SIRT1, an inhibitor of p53, share similar features that exhibit a pro-survival role in cells.

Owing to the importance of mammalian sirtuins in physiology and aging, we hypothesized that they may be differentially expressed in the rat brain and may regulate various targets involved in metabolic processes and neurodegenerative diseases. To date, very little is known about the anatomical distribution of these sirtuins throughout the brain or changes which may occur during aging. Using real-time RT-PCR and western blot analysis, we have quantified changes in SIRT1-7 mRNA and protein expression levels in the brains of female wistar rats aged from 3 to 24 months, spanning life stages from young adulthood to old age (Coleman, 1989). We also sought to examine the functional role of sirtuin expression/activity by assaying superoxide and fatty acid oxidation levels as well as known targets of sirtuin deacetylation activity, including p53, FOXO3a, MnSOD, CPS-1, histone H3K9, and polymerase-I. The implications of these findings for the aging process are discussed within the context of key sirtuin-related metabolic processes.

# Materials and Methods

# Animals

Female wistar rats were used in the following age groups: 3 months (equivalent to a young human adult aged 20 years), 12 months (equivalent to a middle-aged human aged 40 years) and 24 months (equivalent to an aged human greater than 80 years) (Collier and Coleman, 1991).The animals were individually housed in an environmentally controlled room under a 12 h alternating light/dark cycle at 23◦C and were fed commercial rat chow and water ad libitum. All experiments were performed with the approval of the Animal Ethics Committee of the University of Sydney. Anesthesia was induced with a mixture of O2, NO2, and 5% halothane, followed by an intraperitoneal injection of sodium pentobarbitone (60 mg/kg), and perfused transcardially as previously described (Chan-Ling, 1997; Mansour et al., 2008). Whole brain was removed, washed with phosphate buffered saline solution (Invitrogen) and used immediately for a variety of biochemical and histochemical procedures.

# RT-PCR for SIRT1-7 mRNA Expression

For the gene expression studies, RNA was extracted from the frontal, temporal, and occipital lobes and hippocampus, using Qiagen RNAeasy mini kits (Hilden, Germany). Quantitative and qualitative analysis of RNA samples was performed using a 2100 Bioanalyzer (Agilent Technologies) to ensure they were of sufficient quality to generate reliable data (RIN 10). The cDNA was prepared using the SuperScript III First-Strand Synthesis System and random hexamers (Life Technologies, Carlsbad, CA). Gene expression was determined using real-time PCR as described previously (Lee et al., 2010). Briefly, for each reaction, 2µL of diluted cDNA, 10µL of SYBR green master mix, 0.15µL of 10µM forward and reverse primers and 7.7µL of nucleasefree water was used, making a total volume of 20µL. Q-PCR was carried out using the Stratagene Mx3500P Real-Time PCR system (Sydney, Australia). The relative expression levels of SIRT1- 7 were calculated using a mathematical model based on the individual Q-PCR primer efficiencies and the quantified values were normalized against the housekeeping gene Glyceraldehyde Phosphate Dehydrogenase (GAPDH) (Lee et al., 2010). The primer sequences are shown in Supplementary Table 1.

# Western Blots for Assaying Protein Expression of SIRT1-7, Total/Acetylated p53, MnSOD and FOXO3

The frontal, temporal, occipital lobes and hippocampus were carefully dissected from the whole brain and homogenized in RIPA lysis buffer containing 50 mM Tris-HCl (pH 7.4); Igepal 1% (w/v); 0.25% (w/v) Na-deoxycholate; 1 mM EDTA, 150 mM NaCl; 1µg/ml each of protease inhibitors aprotinin, leupeptin and pepstatin; 1 mM Na3VO4; and 1 mM NaF. After 1 h, the homogenate was centrifuged (14,000 g, 30 min, 40176◦C) and the supernatant transferred to a clean polypropylene tube. Equal amounts of protein extract (30µg) were dissolved in Laemmli sample buffer (Bio-Rad Laboratories, Hercules, CA) and heated to denature the protein (5 min, 95◦C), and loaded onto the gel in order to execute protein separation. Electrophoresis (160 volts, 50 min) was performed using Tris-glycine precast 8– 12% (v/v) gradient polyacrylamide SDS-PAGE gels (Bio-Rad Laboratories, Hercules, CA) under reducing conditions. Proteins were transferred onto polyvinylidene difluoride membrane (0.45µm) (Invitrogen, CA, USA) using Tris/glycine buffer, pH 8.3 (25 mM Tris base, 192 mM glycine, 0.1% w/v SDS, 20% v/v methanol) at 140 mA for 3 h. Membranes were blocked with 5% (w/v) non-fat milk powder dissolved in Tris-buffered saline (TBS) for 1 h and incubated with primary antibody overnight at 4 ◦C. The primary antibodies used are detailed in Supplementary Table 2. Membranes were then washed in TBS-Tween-20 and incubated with HRP conjugated secondary antibodies from Sigma (Castle Hill, Australia) for 1 h at ambient temperature. After further washing in TBS-Tween-20, the membranes were incubated with an ECL plus reagent (RPN2132, Amersham) and protein bands visualized on X-ray films. The bands were quantified by integration of pixel intensity using ImageJ software (U.S. National Institutes of Health, Bethesda, Maryland), and normalized to β-actin, which served as an internal control.

# Isolation and Extraction of Nuclei for SIRT1 Deacetylase Activity Assays

Aliquots of brain homogenate from the frontal, temporal, occipital lobes and hippocampus (extracted without protease inhibitors) were centrifuged through 4 ml of 30% sucrose solution containing 10 mM Tris HCl (pH 7.4); 10 mM NaCl; and 3 mM MgCl<sup>2</sup> (1300 g, 10 min, 4◦C). The remaining pellets were washed with cold 10 mM Tris-HCl (pH 7.4) and 10 mM NaCl. The nuclei were then suspended in 50–100µl extraction buffer containing 50 mM Hepes KOH (pH 7.4); 420 mM NaCl; 0.5 mM EDTA; 0.1 mM EGTA; and glycerol 10%, sonicated for 30 s, and allowed to incubate on ice (30 min) followed by centrifugation (10,000 g, 10 min). The nuclear extract was immediately stored at −80◦C for later application of SIRT1 deacetylase activity assay.

# SIRT1 Deacetylase Activity

SIRT1 deacetylase activity was evaluated in nuclear extracts from the frontal, temporal, occipital lobes and hippocampus of young, middle-aged, and aged rats, using the Cyclex SIRT1/Sir2 Deacetylase Flourometric Assay Kit (Nagano, Japan). The final reaction mixture (100µl) contained 50 mM Tris-HCl (pH 8.8), 4 mM MgCl2, 0.5 mM DTT, 0.25 mU/ml Lysyl endopeptidase, 1µM Trichostatin A, 200µM NAD+, and 5µl of nuclear sample. The samples were mixed well and incubated for 10 min at room temperature and the fluorescence intensity (ex. 340 nm, em. 460 nm) was measured at 30 s intervals for a total of 60 min immediately after the addition of fluorosubstrate peptide (20µM final concentration) using Fluostar Optima Fluorometer (NY, NY) and normalized to the total protein content. The results are reported as relative fluorescence/µg of total protein (AU).

#### Measurement of Superoxide

Assay of superoxide was performed using the lucigenin-enhanced chemiluminescence assay as previously described (Brown et al., 2006). Briefly, brain homogenates from the frontal, temporal, occipital lobes and hippocampus were placed in a polypropylene tube containing 0.5 ml PBS and lucigenin (5µmol/L). The tube was placed in a Promega GloMax luminometer (Madison, WI) to detect the relative light units emitted. Background counts were determined from tissue-free preparations, and the luminescence subtracted from the brain sample readings.

# Manganese Superoxide Dismutase (MnSOD) Activity

MnSOD activity was measured as previously described (Spitz and Oberley, 1989). The enzymatic activity of MnSOD was determined based on the competition between MnSOD and an indicator molecule for the generation of superoxide from xanthine and xanthine oxidase, in the presence of 5 mM NaCN, an inhibitor of CuZnSOD activity.

#### Fatty Acid Oxidation

Fatty acid oxidation was measured as previously described (Nasrin et al., 2010). Briefly, brain homogenates from the frontal, temporal, occipital lobes and hippocampus were incubated in serum-free media containing 0.5 mg/ml bovine serum albumin and [3H] palmitate (10µM cold palmitate and 8.93µCi/ml [ <sup>3</sup>H] palmitate) for 90 min. Afterwards, 100µl of homogenate mixture was transferred to a 96-well filter plate containing 100µl of phosphate-buffered activated charcoal slurry. The plate was centrifuged (2800 g, 45 min, 25◦C), the charcoal-containing plate was discarded and the filtrate was counted using a Beckman LS6500 scintillation counter (Beckman-Coultier, Brae, CA).

# Bradford Protein Assay for the Quantification of Total Protein

SIRT1 activity, superoxide levels, MnSOD activity, and fatty acid oxidation were adjusted for variations in total protein concentration using the Bradford protein assay (Bradford, 1976).

### Data Analysis

Results obtained are presented as the means ± the standard error of measurement (Koch et al., 2006) of at least eight animals per age group analyzed in duplicate. One-Way analysis of variance (ANOVA) and post-hoc Tukey's multiple comparison tests were used to determine statistical significance between treatment groups. Differences between treatment groups were considered significant if p < 0.05.

# Results

# Regional Changes in Sirtuin mRNA and Protein

# SIRT1 Deacetylase Activity but Not Protein Expression Declines with Aging

SIRT1, an NAD<sup>+</sup> dependent deacetylase is primarily localized in the nucleus (Sauve et al., 2006). Therefore, we assessed the activity of SIRT1 in nuclear extracts from selected regions of the brain in aging rats. By 12 months a significant decline in SIRT1 activity in all brain regions was observed with the greatest decline again occurring between middle (12 month) and older age (24 months)

Expression in the Aging Rat Brain We performed RT-PCR and Western blotting analysis to examine changes in sirtuin mRNA expression in various regions of the brain from young, middle-age and aged animals. We observed a significant increase in SIRT1 mRNA expression in the frontal, temporal, occipital lobes and hippocampus with age (**Table 1**, Supplementary Figure 2), consist with an increase in SIRT1 protein expression in these regions (**Table 2**). Interestingly, SIRT2 mRNA (**Table 1**, Supplementary Figure 1) and protein expression (**Table 2**) levels increased only in the occipital lobe, with no significant increase observed in the other brain regions. SIRT3-5 mRNA and protein expression declined significantly in the hippocampus and frontal lobe (**Tables 1, 2**, Supplementary Figure 2). In contrast to the increase in SIRT1 expression, our data also shows that SIRT6 mRNA levels declined significantly with age, consistent with reduced SIRT6 protein expression in the same brain regions (**Tables 1, 2**, Supplementary Figure 2). As well, SIRT7 mRNA (**Table 1**, Supplementary Figure 2) and protein (**Table 2**) expression was increased only in the frontal lobe with aging.


TABLE 1 | Anatomical

 changes in sirtuin mRNA expression

 in the aging rat brain using RT-PCR.


(**Figure 1**). In contrast a significant (p < 0.01) increase in the level of SIRT1 protein was observed in all brain regions examined using western blotting (**Table 2**, Supplementary Figure 3).

To investigate the potential downstream effects of altered SIRT1 activity on the acetylation status of the tumor suppressor protein p53, we measured both acetylated and total p53 protein levels using western blotting. As shown in **Figure 2**, we found a significant age-dependent increase in acetylated p53 expression in all brain regions tested. However, no change was observed in total p53 protein content between young and aged rats in any brain region.

#### SIRT2 Can Regulate FOXO3 Expression

Since SIRT2 can deacetylate FOXO3 (Brunet et al., 2004), we set forth to determine the effect of aging of SIRT2 on FOXO3 expression. Our data demonstrates that increased SIRT2 expression in occipital lobe (**Tables 1, 2**, Supplementary Figure 4) is consistent with reduced acetylated FOXO3 expression (**Figure 3**). No change was observed in FOXO3 expression in the frontal, temporal lobes and the hippocampus, in line with the SIRT2 data.

#### Altered Superoxide Levels, MnSOD Protein Expression and MnSOD Activity

Our data indicate a significant decline in MnSOD expression and enzymatic activity with age in the hippocampus and frontal lobe (**Figures 4A–E**), consistent with a reduction in SIRT3 expression (**Tables 1, 2**, Supplementary Figure 5). To further confirm the effects of aging on MnSOD, we measured superoxide levels as a function of MnSOD activity, since MnSOD can catabolize superoxide. Superoxide levels significantly increased with age in the hippocampus and frontal lobe (**Figure 4F**). In contrast, measurements of superoxide in the temporal and occipital lobes were not significantly different than in young rats (**Figure 4F**).

#### Changes in Fatty Acid Oxidation

Nasrin et al. (2010) showed that SIRT4 regulates fatty acid oxidation and mitochondrial gene expression. To determine

TABLE 2 | Anatomical

 changes in sirtuin mRNA expression

 in the aging rat brain using western blotting.

(D) hippocampus in the brain with aging using anti-acetylated p53 and

established by comparison with 3 month old rats.

whether altered SIRT4 expression can affect metabolic function in the aging rat brain, we measured fatty acid oxidation as a target of SIRT4 activity in various brain regions. Fatty acid oxidation significantly increased with age in the hippocampus and frontal lobe (**Figure 5**) consistent with a decrease in SIRT4 expression (**Tables 1, 2**, Supplementary Figure 6). The amount of fatty acid oxidation and SIRT4 expression did not change significantly in the temporal and occipital lobes.

# Increased Acetylated CPS1 Correlates with Reduced SIRT5 Expression in the Hippocampus and Frontal Lobe

Nakagawa et al. (2009) recently showed that carbamoyl phosphate synthetase 1 (CPS1) is a SIRT5-binding protein. Consistent with this finding, we observed increased levels of acetylated CPS1 in the hippocampus and frontal lobe of the aging rat brain (**Figure 6**) in conjunction with decreased SIRT5 expression in these brain regions (**Tables 1, 2**, Supplementary Figure 7). Acetylated CPS1 levels in the temporal and occipital lobes remained unchanged, in line with observed pattern for SIRT5 expression.

# SIRT6 May Regulate Histone Acetylation in the Aging Rat Brain

To explore a potential functional effect of increased SIRT6 expression in the aging brain, we measured H3K9 histone acetylation. We observed a significant increase in H3K9 acetylation in various brain regions (**Figure 7**) consistent with an age-dependent decline in SIRT6 expression (**Tables 1, 2**, Supplementary Figure 8), parallel to an increase in SIRT1 expression (**Tables 1, 2**, Supplementary Figure 2).

# SIRT7 Can Influence Protein Transcription via Regulation of RNA Polymerase-I

Since SIRT7 can interact with Pol I (Ford et al., 2006), we tested the age-related effect of SIRT7 on Pol I expression. Our data shows that Pol I expression increases with age only in the frontal lobe (**Figure 8**), consistent with the observed increase in SIRT7 expression (**Tables 1, 2**, Supplementary Figure 9).

# Discussion

Sirtuins are key NAD-dependent class III histone deacetylase enzymes that have been extensively investigated to determine

expression in the aging rat brain. Western blotting for MnSOD in (A) frontal lobe, (B) temporal lobe, (C) occipital lobe, and (D) hippocampus in the brain with aging using anti-MnSOD antibody. The blots shown are representative tracings of an experiment done eight times. Graphs are mean ± S.E brains from brains from eight different represents the corresponding band for each age group. Significance \*p < 0.01 compared to 3 month old rats. (E) MnSOD activity in aged rat brain tissue. Significance \*p < 0.01 compared to 3 month old rats. (F) Superoxide levels in aged rat brain tissue. Significance \*p < 0.01 compared to 3 month old rats.

their role in various disease conditions (Chen et al., 2005; Denu, 2005, 2007; Anastasiou and Krek, 2006; Anekonda and Reddy, 2006; Belenky et al., 2007; Chen and Guarente, 2007; Dali-Youcef et al., 2007). Numerous studies have highlighted the myriad of intrinsic and extrinsic biological effects, which

play important neuroprotective roles in neurodegenerative and cerebrovascular conditions, including stroke, ischaemic brain injury, Alzheimer's disease, Parkinson's disease, Huntington's disease and amyotrophic lateral sclerosis (Longo and Kennedy, 2006; Gan, 2007; Okawara et al., 2007; Milne and Denu, 2008; Pallas et al., 2008). However, to date, very little has been reported regarding mammalian sirtuin distribution and functional role in the central nervous system and to our knowledge this is the first study examining sirtuin expression and functional changes with aging.

SIRT1, the human ortholog of Sir2α, is localized predominantly in the nucleus of neurons, and regulates several important physiological processes such a chromatin remodeling, gene transcription, and the activity of several apoptotic mediators, particularly p53 (Pillai et al., 2005; Sauve et al., 2006). Indeed, increased SIRT1 activity by mediators of caloric restriction, such as resveratrol have been shown to prolong lifespan by a number of different processes, including reduced apoptosis and improved DNA repair (Arraki et al., 2004; Borra et al., 2005; Raval et al., 2006; Qin et al., 2006). SIRT1 expression has been previously shown to progressively increase with age in both young and older endothelial cells. Although a significant increase in SIRT1 expression was reported

lobe, (C) occipital lobe, and (D) hippocampus in the brain with aging. The blots shown are representative tracings of an experiment done eight times. corresponding band for each age group. Significance \*p < 0.01 compared to 3 month old rats.

in young cell, a significant decline in SIRT1 was observed in older cells (Conti et al., 2015). This suggests that the SIRT1 pathway is more effective in younger cells. Oxidative stress may lead to a reduction in SIRT1 and its regulatory control on target proteins, thus promoting cellular senescence. Moreover, SIRT1 has been proposed to play a major role in neuroprotection. While several studies have demonstrated the protective roles of sirtuin activators during aging, little is known regarding the distribution or activity of SIRT1 in the aging brain.

We have shown that aging is associated with increases in SIRT1 expression level, but decreases in the activity of SIRT1 in nuclear extracts from selected brain regions. The decreased activity of SIRT1 in the aging rat brain is consistent with the observed decrease in substrate (NAD+) level that is observed during aging (Braidy et al., 2011). Oxidative damage may also potentially inhibit SIRT1 activity, as it does to several other proteins (Radak et al., 2009). The current observations are consistent with a previous study which reported decreased SIRT1 activity but not expression in skeletal muscle of aged rats (Koltai et al., 2010). It is also likely that the age-associated drop in NAD<sup>+</sup> content due to increased demand by the DNA repair process may induce a compensatory increase in SIRT1 production to enhance its competitive advantage for the available NAD<sup>+</sup> (Koltai et al., 2010). Exercise training has been shown to slow down the aging process by increasing SIRT1 activity, modulating an antioxidant response and mediating cell cycle regulation in aged rats (Ferrara et al., 2008).

On the contrary to previous studies, Gong et al. (2014) recently showed that SIRT1 expression is reduced with age at the transcriptional and translational levels in the brain, liver, skeletal muscle, and white adipose tissue in senescence-accelerated mouse prone (SAM-P8) and a control counterpart strain, senescence-accelerated mouse resistant 1 (SAM-R1). Moreover SIRT1 expression levels were significantly lower in SAM-P8 compared to SAM-R1 mice (Gong et al., 2014). We postulate that SAM series may exhibit differential genetic backgrounds apart from the accelerated senescence-related ones which may alter both the expression and function of sirtuins (Ito, 2013).

The neuroprotective effects of SIRT1 are thought to be mediated in part by the deacetylation and hence inhibition of p53. We therefore measured the expression of both total and acetylated p53 in aging brain tissue. Our results show a significant

increase in acetylated p53 protein while no change was observed in total p53 protein content. This is consistent with work done by others showing that cells derived from SIRT1-deficient mice had elevated levels of acetylated p53 (Ford et al., 2005), and that sirtuin inhibition leads to hyperacetylated p53 (Yamakuchi et al., 2008). These results suggest that age-related changes in SIRT1 activity can regulate the post-translational acetylation of p53 (Pillai et al., 2005). Increased SIRT1 activity has been previously shown to represses p53 activity to prevent p53-dependent cellular senescence. It is therefore likely that our observation of concurrently decreased SIRT1 activity and increased p53 expression level are causally linked. Our present study provides supporting evidence for an age related change in the brain SIRT1 p53 axis that is consistent across at least four cortical regions. These age related changes to SIRT1 activity, and in turn p53 expression, may be driven by availability and changes in brain levels of NAD+, which are known to decrease during aging

(Braidy et al., 2011). Ramadori et al. (2008) recently showed that reduced energy availability can lead to lower levels of acetylated p53 only in the hypothalamus and hindbrain within the normal brain, and the effect is altered in leptin-deficient obese mice (Ramadori et al., 2008). Level of NAD<sup>+</sup> may therefore be central to regulation of a variety downstream effects on senescence regulating proteins; however, further work is needed to investigate these relationships.

SIRT2 has been shown to promote longevity in yeast, nematodes and fruitflies, although the life-span promoting effect has not been observed in humans (Lamming et al., 2004). In human cells, SIRT2 has been shown to mediate cell survival through mitotic control (Dryden et al., 2003). SIRT2 regulates microtubule dynamics by deacetylating several cytoskeletal proteins including tubulin, and regulates cell cycle progression (North et al., 2003). Overexpression of SIRT2 has been reported to lengthen mitosis, and reduced expression of SIRT2 has an antiapoptotic effect (Dryden et al., 2003; North et al., 2003). Compared to the other sirtuins, SIRT2 expression has been found to be greatest in the brain (Pandithage et al., 2008). Although SIRT2 is mainly found in oligodendrocytes, and myelin-forming glial cells (Pandithage et al., 2008), SIRT2 has been recently described in the cytoplasm of hippocampal neurons in the adult mouse brain (Li et al., 2007). There is argument for (Werner et al., 2007) and against the localization of SIRT2 in astrocytes (Li et al., 2007), the second major brain glial cell type. However, other studies have shown that SIRT2 is present in both neurons and astrocytes (Michan and Sinclair, 2007; Werner et al., 2007; Pandithage et al., 2008; Ramadori et al., 2008).

To investigate the effect of SIRT2 function during aging, we examined the levels of FOXO3 protein levels in the aging brain. Our study shows that both SIRT2 and FOXO3 undergo age related expression changes only in the occipital lobe, and that in this case their expression levels are inversely related. The FOXO transcription factors are regulated by post-translational modifications, and SIRT2-mediated deacetylation of FOXO3 can influence FOXO3 ubiquitination and degradation (Wang et al., 2011). Brunet et al. (2004) showed that SIRT2 deacetylation of FOXO protein can activate a myriad of genes that may regulate cell survival, thus shifting vulnerable cells from apoptosis toward growth arrest and DNA repair (Brunet et al., 2004). However, the significance of the upregulated SIRT2 in the occipital lobe with aging cannot be easily interpreted in the light of brain aging. The biological significance of its upregulation in the occipital lobe with aging, in all of its cellular locales, is not immediately apparent, but its cellular distribution supports currently known roles.

We and others have previously shown that increased ROS formation and reduced mitochondrial efficiency may contribute to impaired physiological function, increased incidence of disease, and a reduction in life span (Beal, 1995, 2003, 2007; Budd and Nicholls, 1996; La Piana et al., 1998; Budd et al., 2000; Gibson et al., 2000; Menzies et al., 2002a,b; Jacquard et al., 2006; Ahn et al., 2008; Braidy et al., 2011). Therefore, it is highly likely that acetylation of mitochondrial proteins may play a critical role in regulating mitochondrial ROS levels (Ozden et al., 2011). SIRT3 is the main mitochondrial deacetylase (Shi et al., 2005; Ahn et al., 2008), and SIRT3 knockout studies have shown an increase in ROS, including the levels of the highly reactive superoxide anion both in vitro and in vivo (Lombard et al., 2007). Therefore, SIRT3 appears to represent a regulatory molecule that maintains mitochondrial homeostasis by mediating the acetylation of metabolic target protein, including those that form part of the endogenous antioxidant defense system. This is the first study to show that brain SIRT3 expression levels decline with age and in parallel with lower MnSOD protein levels, MNSOD activity, and increased superoxide (O2. <sup>−</sup>) levels in the rat hippocampus and frontal lobe. This is relevant to Alzheimer's disease where the hallmarks of the disease (i.e., senile plaques and neurofibrillary tangles) are predominantly observed in the hippocampus and frontal lobe of the brain (Alafuzoff et al., 1987).

MnSOD is the primary mitochondrial antioxidant enzyme which neutralizes O2. <sup>−</sup> to the less reactive hydrogen peroxide (H2O2) followed by conversion to H2O by catalase in the mitochondrial matrix (Oberley and Oberley, 1988). Superoxide is a byproduct of normal oxidative phosphorylation and ATP production and can lead to extensive damage to lipids, proteins and DNA (Henderson et al., 2009). Since MnSOD enzymatically scavenges superoxide, whose levels are significantly increased in SIRT3-deficient cells (Spitz and Oberley, 1989), it seems likely that an age-related reduction in SIRT3 expression may lead to reduced MnSOD activity and thereby higher oxidative damage and altered redox signaling. Tao et al. (2010) recently showed that SIRT3-mediated deacetylation of Lysine 122 can regulate MnSOD activity in response to stress (Tao et al., 2010). While the current data does not indicate a definite relationship between SIRT3-mediated acetylation of MnSOD during pathological processes, it is highly likely that SIRT3 may play a protective role against ROS by regulating the enzymatic properties of MnSOD during aging and particularly in chronic age-induced oxidative stress.

SIRT4 is another mitochondrial sirtuin that may be altered in the brain during the aging process. Recently, SIRT4 has been shown to inactivate glutamate dehydrogenase, an enzyme which converts glutamate to α-ketoglutarate in the mitochondria in an NAD<sup>+</sup> dependent manner (Haigis et al., 2006). Although the function of SIRT4 remains unclear, we speculated that SIRT4 might be involved in mitochondrial oxidative metabolism. Nasrin et al. (2010) recently showed that SIRT4 regulates fatty acid oxidation and mitochondrial gene expression in liver and muscle cells. SIRT4 knockdown in hepatocytes also increased SIRT1 mRNA protein levels both in vitro and in vivo, suggesting that the effect of SIRT4 on fatty acid oxidation may be SIRT1 dependent (Nasrin et al., 2010). Here, we show a significant increase in fatty acid oxidation in the aging rat brain in the hippocampus and frontal lobe which is closely associated with a reduction in SIRT4 expression. How SIRT4 regulates fatty acid oxidation with aging is unknown, but could be related to changes in NAD+/NADH and/or AMP/ATP ratios. Another possibility is the role of SIRT4 in modulating AMPK-SIRT1 pathways (Nasrin et al., 2010). Activation of AMPK increases NAD<sup>+</sup> levels which can increase SIRT1-mediated deacetylation of LκB1 which increases acetyl-CoA carboxylase (ACC) phosphorylation, leading to increased fatty acid oxidation (Nasrin et al., 2010). Alternatively, it is possible that SIRT1 increases as a compensatory effect to replace SIRT4 function in SIRT4-knockout mice (Nasrin et al., 2010). Therefore, sirtuins may represent a synergistic network which regulates metabolic signals and mitochondrial function during aging.

While SIRT3 and SIRT4 appear to directly modulate the activity of mitochondrial enzymes associated with energy metabolism, little is known regarding the cellular role of SIRT5. Our data is the first to show an age related reduction in SIRT5 expression in the hippocampus and frontal lobe, consistent with the expression changes of the other two mitochondrial sirtuins, SIRT3 and SIRT4. Our data support the findings by Nakagawa et al. (2009) who showed that SIRT5 could deacetylate CPS1 in a NAD-dependent manner and this deacetylation increased CPS1 enzymatic activity (Nakagawa et al., 2009). Indeed, SIRT5 knockout mice have ∼30% reduction in CPS1 activity compared to wild type mice. During fasting conditions, calorie restriction or following consumption of a high protein diet, SIRT5 deficient mice failed to up-regulate CPS1 activity resulting in hyper ammonemia (Nakagawa et al., 2009). Taken together, these data indicate that SIRT5 has an emerging role in the metabolic changes that take place during aging.

Like SIRT1, SIRT6 is another chromatin-associated nuclear protein that has been shown to affect DNA repair, telomere maintenance, gene expression, and metabolism (Mostoslavsky et al., 2006). We have shown that SIRT6 expression declines with age in the frontal, temporal, occipital lobes and hippocampus in the aging rat brain. SIRT6 deficient mice have a significantly reduced lifespan and suffer from severe multisystemic phenotypes (Mostoslavsky et al., 2006). SIRT6 can deacetylate histone H3K9, a chromatin marker that is associated with longevity (Schwer et al., 2010). Histone acetylation is relevant to several neurodegenerative diseases including schizophrenia, depression, addiction, and various neurodevelopment disorders (Schwer et al., 2010). However, the mechanism by which SIRT6 can regulate histone acetylation in the brain during aging remains obscure. We analyzed H3K9 acetylation in several brain regions. Our data indicates a significant increase in H3K9 acetylation in various brain regions consistent with an age-dependent decline in SIRT6 expression, and occurs parallel to an increase in SIRT1 expression. Loss of SIRT6 has been shown to induce dramatic H3K9 hyperacetylation in the hypothalamus, cortex, hippocampus and cerebellum, and in purified brain nuclei. Similarly, increased acetylation of H3K9 has been reported in SIRT6 deficient mice, while the acetylation levels of other histones remained unaffected (Schwer et al., 2010). Together, these results suggest that SIRT6 is the main H3K9 deacetylase in the brain, suggesting a potential role in gene regulation with age.

We also investigated the role of SIRT7 in the aging rat brain. Our data shows that SIRT7 is upregulated in the frontal lobe with aging. Ford et al. (2006) showed that SIRT7 can interact with Polymerase-I (Pol I) (Ford et al., 2006). As transcription of rDNA by Pol I accounts for 65% of total transcription in mammalian cells, Pol I appears to be highly coordinated with cellular metabolism and cellular proliferation (Grummt, 2003). We found that Pol I expression increases with age only in the frontal lobe, consist with the observed increase in SIRT7 expression. In yeast, Sinclair and Guarente (1997) showed that Sir2 can mediate longevity, primarily through its silencing role at the rDNA (Sinclair and Guarente, 1997).Therefore, maintenance of cellular energy status may be coupled with levels of RNA synthesis, protein transcription, and therefore cell growth, which is dysregulated during the aging process.

Sirtuins represent a unique class of enzymes that not only regulate protein acetylation and metabolism, but also play prominent roles in promoting longevity, preventing disease and improving cell survival. Our present study describes the anatomical landscape of mammalian sirtuins and their downstream targets within the brain (Supplementary Figure 10). On the contrary to our finding, a recent study showed that no significant decreases in the expression of any sirtuin member were observed in any brain region between the 24 month old and 3 month old Wistar rats. As well, the mRNA expression patterns for specific sirtuins were never parallel to its corresponding translational expression pattern in that study (Sidorova-Darmos et al., 2014). However, the cellular biology of altered sirtuin expression in various brain regions remains unknown. It is unclear which cells in the brain contain sirtuins. Previous studies have suggested that only specific neuronal brain cells may express functional sirtuin protein (Hasahara et al., 2005; Sidorova-Darmos et al., 2014). Sidorova-Darmos et al. (2014) further examined the expression patterns of sirtuins in murine brain cells. While SIRT2 mRNA was largely expressed in both neurons and astrocytes, SIRT2 protein was expressed in astrocytes only. Similarly, SIRT5 was expressed at the translational level in neurons, although its mRNA was also identified in both astrocytes and neurons (Sidorova-Darmos et al., 2014). If this hypothesis is correct, then only certain cells may be responsible for the age-related effects in the brain. Additionally, the effects of calorie restriction on sirtuin activity in specific brain regions and certain brain cells have not been investigated. The fact that all mitochondrial sirtuins are expressed in brain neurons is important with respect to their protective roles against neurodegenerationTaken together, our results provide additional evidence for the role of sirtuins in regulating brain function at different stages of development. It also identifies the potential for pharmacologically targeting specific sirtuins to establish cell-specific effects within the brain.

# Author Contributions

Conceived and designed the experiments: NB, AP, PS, GS, TJ. Performed the experiments: NB, TJ, HM. Analyzed the data: NB, AP, PS. Contributed reagents/materials/analysis tools: GG, TC, PS. Wrote the paper: NB, AP, PS. All authors reviewed the manuscript.

# Acknowledgments

This work was supported by a Capacity Building Grant from the National Health and Medical Research Council of Australia, and a UNSW Faculty of Medicine Research Grant. NB is the recipient of an Alzheimer's Australia Viertel Foundation Postdoctoral Research Fellowship at the University of New South Wales. TJ is a recipient of the University of New South Wales Postgraduate Award (UPA).

# Supplementary Material

The Supplementary Material for this article can be found online at: http://journal.frontiersin.org/article/10.3389/fncel. 2015.00167/abstract

# References


Yang, T., and Sauve, A. A. (2005). NAD+ metabolism and sirtuins: metabolic regulation of protein deacetylation in stress and toxicity. AAPS J. 8, E632–E643. doi: 10.1208/aapsj080472

**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.

Copyright © 2015 Braidy, Poljak, Grant, Jayasena, Mansour, Chan-Ling, Smythe, Sachdev and Guillemin. 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.

# Dopamine induces the accumulation of insoluble prion protein and affects autophagic flux

#### *Marcio H. M. da Luz 1,2, Italo T. Peres 1, Tiago G. Santos 3, Vilma R. Martins 3, Marcelo Y. Icimoto4 and Kil S. Lee1 \**

*<sup>1</sup> Department of Biochemistry, Molecular and Cellular Biology, Universidade Federal de São Paulo, São Paulo, Brazil*

*<sup>2</sup> Biomedicina, Universidade Metodista de São Paulo, São Paulo, Brazil*

*<sup>3</sup> International Research Center, A C Camargo Cancer Center, Sao Paulo, Brazil*

*<sup>4</sup> Department of Biophysics, Universidade Federal de São Paulo, São Paulo, Brazil*

#### *Edited by:*

*Victoria Campos, Instituto Nacional de Neurologia y Neurocirugia, Mexico*

#### *Reviewed by:*

*Yong-Sun Kim, Hallym University, South Korea Daniel Rial, Center for Neuroscience and Cell Biology, Portugal Lifeng Yang, China Agricultural University, China*

#### *\*Correspondence:*

*Kil S. Lee, Department of Biochemistry, Universidade Federal de São Paulo, Rua Pedro de Toledo 669, 8*◦ *andar, Vila Clementino, São Paulo, SP 04039-032, Brazil e-mail: kil\_sun\_lee@yahoo.com.br; kslee@unifesp.br*

Accumulation of protein aggregates is a histopathological hallmark of several neurodegenerative diseases, but in most cases the aggregation occurs without defined mutations or clinical histories, suggesting that certain endogenous metabolites can promote aggregation of specific proteins. One example that supports this hypothesis is dopamine and its metabolites. Dopamine metabolism generates several oxidative metabolites that induce aggregation of α-synuclein, and represents the main etiology of Parkinson's diseases. Because dopamine and its metabolites are unstable and can be highly reactive, we investigated whether these molecules can also affect other proteins that are prone to aggregate, such as cellular prion protein (PrPC). In this study, we showed that dopamine treatment of neuronal cells reduced the number of viable cells and increased the production of reactive oxygen species (ROS) as demonstrated in previous studies. Overall PrPC expression level was not altered by dopamine treatment, but its unglycosylated form was consistently reduced at 100μM of dopamine. At the same concentration, the level of phosphorylated mTOR and 4EBP1 was also reduced. Moreover, dopamine treatment decreased the solubility of PrPC, and increased its accumulation in autophagosomal compartments with concomitant induction of LC3-II and p62/SQSTM1 levels. *In vitro* oxidation of dopamine promoted formation of high-order oligomers of recombinant prion protein. These results suggest that dopamine metabolites alter the conformation of PrPC, which in turn is sorted to degradation pathway, causing autophagosome overload and attenuation of protein synthesis. Accumulation of PrP<sup>C</sup> aggregates is an important feature of prion diseases. Thus, this study brings new insight into the dopamine metabolism as a source of endogenous metabolites capable of altering PrP<sup>C</sup> solubility and its subcellular localization.

**Keywords: dopamine, protein aggregation, prion, protein synthesis, autophagy, neurodegeneration**

#### **INTRODUCTION**

Aberrant protein aggregation is a common hallmark of many neurodegenerative diseases, while a specific protein predominantly aggregates in each type of diseases (Brundin et al., 2010). Mutations that alter amino acid sequence or modifications of side chains by reactive molecules can disrupt the native fold (Tyedmers et al., 2010; Petrov and Zagrovic, 2011). Considering that proteins perform numerous biological activities, the accumulation of their aggregates can be toxic and effective clearance of structurally altered proteins can be essential for cellular survival (Tyedmers et al., 2010).

Most cases of neurodegenerative diseases are idiopathic and protein aggregation can occur without defined mutations or clinical histories that clearly justify the manifestation of the diseases (Alkhuja, 2013; Musiek and Schindler, 2013). This observation raises the possibility that certain endogenous metabolites can induce protein misfolding, and protein aggregates may accumulate upon excessive formation of such metabolites and/or failure of degradation pathways (Morris, 2013). For instance, reactive oxygen species (ROS) are produced as byproducts of several endogenous metabolic pathways, such as mitochondrial respiration and oxidase catalyzed reactions, which include the metabolism of catecholamine (Andreyev et al., 2005; Dikalov, 2011; Kodama et al., 2013). Among catecholamines, dopamine is the one that has the greatest propensity for oxidation (LaVoie et al., 2005). During dopamine metabolism, ROS are generated not only by enzymatic reactions but also by auto-oxidation of dopamine, producing highly reactive dopamine quinone

**Abbreviations:** ROS, Reactive oxygen species; PrPC, Cellular prion protein; PrPSc, abnormal prion protein; SDS, Sodium dodecyl sulfate; mTOR, Mammalian target of rapamycin; 4EBP1, 4E binding protein 1; ER, Endoplasmic reticulum; eIF2-α, Eukariotic initiation factor 2-α; p62/SQSTM1, Sequestosome 1; LC3, Microtubule-associated protein 1 light chain 3; TSE, Transmissible spongiform encephalopathies.

(Asanuma et al., 2003). The accumulation of neuromelanin (which is synthesized from dopamine quinone) throughout the aging process is an indication of the production of dopamine quinone in physiological conditions (Herrero et al., 1993; Kim et al., 2006). Thus, enhanced dopamine metabolism can induce oxidative stress leading to mitocontrial dysfunction and disturbance in protein quality control system (Chen et al., 2008; Hastings, 2009). Although dopamine metabolism has been extensively linked to the aggregation of α-synuclein and Parkinson's disease (Cappai et al., 2005; Lee et al., 2011), dopamine oxidation can also induce the formation of adducts and aggregates of other proteins, such as cytoskeleton proteins and mitochondrial complexes (Van Laar et al., 2009). Moreover, dopamine can also oxidize prion protein *in vitro* in the presence of copper ions (Shiraishi and Nishikimi, 2002; Shiraishi et al., 2005).

Conformational alteration of cellular prion protein (PrPC) is an essential event for development of transmissible spongiform encephalopathies (TSE). However, in its normal conformation, PrPC plays important roles in cell growth, differentiation and survival (Linden et al., 2008). Previous studies have demonstrated that PrP<sup>C</sup> also possesses anti-oxidant properties, but the mechanisms related with this function are still under debate. The expression of PrP<sup>C</sup> appears to increase during the oxidative stress caused by ischemia and antioxidant defense mechanisms appear to require PrP<sup>C</sup> expression (McLennan et al., 2004; Beraldo et al., 2013). Direct reactions between PrP<sup>C</sup> and ROS have also been proposed (Bertuchi et al., 2012), suggesting that PrP<sup>C</sup> might play roles in scavenging ROS from the extracellular environment. PrP<sup>C</sup> is highly expressed in neurons and has flexible N-terminal domain enriched by amino acid residues that are more susceptible to oxidation (Linden et al., 2008; Abouelatta et al., 2009). Thus, it is plausible to hypothesize that PrP<sup>C</sup> can be a preferential target of ROS in the central nervous system with consequent inactivation of ROS activity. PrPC appears to participate in the regulation of dopamine metabolism (Lee et al., 1999; Adjou et al., 2008; Rial et al., 2014), but the role of PrPC in oxidative stress caused by dopamine oxidation has not yet been evaluated. In this study, we investigated the effects of dopamine toxicity on the expression levels, solubility and subcellular localization of PrPC, with concomitant analysis of cellular viability, protein synthesis and degradation pathways.

#### **MATERIALS AND METHODS**

#### **ANTIBODIES AND REAGENTS**

DMEM Media-Glutamax™-I and fetal bovine serum (FBS) were purchased from Life Technologies. Dopamine hydrochloride, thiazolyl blue tetrazolium bromide (MTT) and 2 , 7 dichlorofluorescein diacetate (DCFDA) were purchased from Sigma-Aldrich. Complete Protease Inhibitor cocktail tablets were purchased from Roche and Pierce™ phosphatase inhibitor mini tablets were purchased from Thermo Scientific. All primary antibodies were purchased from Cell Signaling, except antiprion SAF32, which was purchased from Cayman. Horseradish peroxidase conjugated secondary antibodies were purchased from Sigma-Aldrich and Alexa fluorophore conjugated antibodies and CellROX Green reagent were purchased from Life Technologies.

#### **N2a CELL CULTURE**

N2a cells were maintained at 37◦C in a humidified atmosphere of 5% CO2 in DMEM-Glutamax™-I (Dulbecco's Modified Eagle's Medium) supplemented with 10% FBS, Penicillin (100 U/mL) and Streptomycin (100μg/mL). For dopamine treatment, 4 × 104 cells/cm<sup>2</sup> were seeded. After 24 h, cells were treated with 50 or 100μM dopamine diluted in DMEM containing 0.5% of FBS for 24 h.

#### **CELL VIABILITY ASSAY**

Cells were treated with dopamine in 96-well plates for 24 h. After the treatment, cells were incubated with 100μl of 1.2 mM MTT diluted in Krebs solution (NaCl 126 mM, KCl 2.5 mM, NaHCO3 25 mM, NaH2PO4 1.2 mM, MgCl2 1.2 mM, CaCl2 2.5 mM, D-glucose 10 mM) for 2 h. The MTT formazan were solubilized in 100μl of DMSO. Insoluble materials were removed by centrifugation at 20,000 g for 2 min and absorbance of supernatant was measured at 540 nm.

#### **ROS MEASUREMENT**

Cells were treated with dopamine in 96-well plates for 24 h. After the treatment, cells were incubated with 10μM of DCFDA or 5μM CellRox Green in DMEM 0.5% FBS for 30 min. After washing twice with PBS or Krebs solution, the fluorescence was measured using plate reader (Bio-Tek). CellRox Green was also analyzed by Operetta high content screening system (Perkin Elmer).

#### **WESTERN BLOT**

Cells were lysed with lysis buffer (Tris 100 mM, pH 8.0, NaCl 130 mM, EDTA 10 mM, Triton X-100 1%, sodium deoxycholate 0.5%, complete protease inhibitor cocktail and Pierce phosphatase inhibitor mini tablet), and post-nuclear supernatants (PNS) were collected. The protein content in PNS was quantified using BCA protein assay kit (Thermo Scientific Pierce). Equal amounts of proteins were resolved by SDS-PAGE and transferred to PVDF membrane. For immunodetection, membranes were incubated with TBS-T (Tris 50 mM, NaCl 150 mM, Tween-20 0.1% pH 7.4) containing 5% BSA for 1 h. Then, the membranes were incubated with primary antibody for 1 h at room temperature or for 16 h at 4◦C, then washed three times with TBS-T. Secondary antibody conjugated with peroxidase and SuperSignal™ West Pico Chemiluminescent Substrate were used to detect antigens labeled with specific antibodies. Images were digitalized and quantified using Alliance mini 4 m (UVITEC Cambridge). For each protein, at least three independent experiments were performed. In every experiment, the band intensity of the proteins of interest (PI) was normalized by the intensity of the GAPDH bands. Western blot is semi-quantitative assay and direct comparison of PI/GAPDH ratio between independent experiments is not possible. Thus, in order to enable this comparison, the average of PI/GAPDH ratio of three experimental groups (cells treated with 0, 50, and 100μM of dopamine) was set as 100%, and the percentage of each group was calculated in each independent experiment. The mean percentage of all independent experiments was plotted with respective 95% confidence interval (CI95). An example of PrP<sup>C</sup> expression analysis was shown in supplementary data.

#### **ULTRACENTRIFUGATION**

Post-nuclear supernatants were incubated with 1% sarkosyl for 10 min on ice, then ultracentrifuged for 2 h at 100,000 × g and at 4◦C. Pellets were dissolved in sample buffer 4x (8% SDS, Tris HCl 250 mM, pH 6.8, 40% Glycerol, 0.08 mg/ml bromophenol blue and 1.4 M β-mercaptoetanol), and supernatants were incubated with 4 volumes of methanol for 2 h at −20◦C for protein precipitation. After centrifugation for 20 min at 25,000× g and 4◦C, methanol precipitated pellets were dissolved in Sample Buffer 4x. The level of PrP<sup>C</sup> in each fraction was assessed by western blot. To combine the results of multiple independent experiments, percentage of PrP<sup>C</sup> observed in each sample was calculated considering the average of six samples (Pellets and supernatants of 0, 50, and 100μM) as 100%. This normalization was necessary because the absolute value of the band intensity is not comparable from one blot to another. The mean percentage of all independent experiments was plotted with respective CI95.

#### **PURIFICATION OF BIOTINYLATED CELL SURFACE PROTEINS**

Cell surface proteins were biotinylated using Pierce Cell Surface Protein Isolation Kit (Thermo Scientific). The cells were harvested at time zero or after 1 h incubation in the presence or absence of 50μM of dopamine. Biotinylated proteins were purified from the cell lysates using NeutrAvidin-agarose, and PrPC was detected by western blot using SAF32 antibody.

#### **IMMUNOFLUORESCENCE**

Cells were fixed with 3.7% paraformaldehyde diluted in PBS (pH 7) for 5 min and then with alkalinized 3.7% paraformaldehyde (pH 10) for 15 min. After washing twice with PBS, excess of paraformaldehyde was inactivated with glycine 100 mM and cells were permeabilized with Triton X-100 0.1% diluted in PBS for 10 min. To avoid unspecific bindings, cells were incubated with 1% BSA diluted in PBS for 1 h and then with primary antibodies diluted in PBS containing Triton X-100 0.1% for 1 h. After washing three times with PBS, cells were incubated with secondary antibody conjugated with Alexa 488 or Alexa 594 fluorophore for 1 h and then washed five times with PBS. Average images of six frames were acquired for each visual field using confocal microscope (*Leica TCS SP8).* Images were analyzed using Image J 1.48r. In each image, total area stained by anti-PrP<sup>C</sup> was defined as 100% and percentage of the area co-stained by anti-LC3-II was assessed using Colocalization Threshold plugin with default parameters (Bolte and Cordelieres, 2006). To estimate the distribution of PrP<sup>C</sup> in intracellular vesicles with distinct sizes, we counted the number of PrPC-positive particles and then categorized them by their size. The percentage of each category was calculated in relation to total number of particles counted in each group.

#### **OLIGOMERIZATION OF RECOMBINANT PrP (rPrP)**

rPrP was prepared as described elsewhere (Zahn et al., 1997). The oligomerization was performed in 20 μl of Tris buffer (10 mM pH 7.4) containing 10μg rPrP and dopamine at designated concentrations. This mixture was incubated for 24 h at 37◦C protected from the light unless otherwise specified. Sodium metabisulfite (400μM) was added whenever necessary. The product of oligomerization was separated in 4–20% gradient SDS polyacrylamide gel and revealed by silver staining.

### **STATISTICAL ANALYSIS**

All experiments, except ROS measurement and biotinylation of cell surface proteins, were repeated 3–5 times. For ROS measurement, one experiment was performed in triplicate for each dye and detection methods. Mean of independent experiments or replicates was plotted with respective CI95. The groups were considered different when their CI95 did not overlap. In this criterion, *p-*values are estimated to be lower than 0.01 (Cumming et al., 2007).

# **RESULTS**

### **DOPAMINE ALTERS THE EXPRESSION LEVEL OF UNGLYCOSYLATED PrP<sup>C</sup>**

Treatment of neuronal cells with high concentration of dopamine for a prolonged period (typically 24 h) is known to evoke cytotoxic effects (Gomez-Santos et al., 2003; Yamakawa et al., 2010). To evaluate the effect of dopamine cytotoxicity on PrPC, we treated N2a cells with 50 and 100μM of dopamine for 24 h. This treatment did not induce morphological alterations, but cell viability was gradually reduced as dopamine concentration was increased (**Figures 1A,B**). Dopamine treatment also induced ROS production as assessed by CellROX Green reagent and 2 ,7 dichlorofluorescein diacetate (DCFDA), compounds that show enhanced fluorescence upon oxidation. Both reagents showed similar results in high-throughput plate reader (**Figure 1C**), but the increase of CellRox fluorescence induced by 50μM of dopamine was not statistically significant. This subtle difference between two dyes can be explained by their different reactivity. CellROX Green was also analyzed by high content cell analyzer. As shown in **Figure 1D**, brighter fluorescent spots were observed in cells treated with 50 and 100μM of dopamine. According to the manufacturer, this pattern is due to the primary location of CellROX in the nucleus and mitochondria upon oxidation.

The treatment with 50 and 100μM of dopamine did not alter the total amount of PrP<sup>C</sup> (**Figures 2A,B**). However, the fastmigrating band (#) was decreased with 100μM of dopamine (**Figures 2A,C**). This band probably represents newly synthesized unglycosylated form rather than N-terminally truncated PrP<sup>C</sup> because this band migrated close to 26 kDa marker and was recognized by SAF32, an antibody that binds to octa-repeat region located in N-terminal PrPC. The specific reduction of this immature form can indicate that dopamine toxicity might have affected protein synthesis. Thus, we verified the level of mTOR (a master regulator of growth and protein synthesis) together with 4EBP1 (a repressor of cap-dependent translation upon dephosphorylation). As shown in **Figures 3A–D**, phosphorylation of mTOR and 4EBP1 was significantly decreased by 100μM of dopamine. Also, faster migration of total 4EBP1 bands indicated lower degree of phosphorylation. These data indicate that protein synthesis might be compromised in this condition. In addition, the expression level of ER chaperone BiP was increased with 50μM of dopamine

treatment for 24 h **(A)**, but the viability assessed by the formation of MTT formazan was reduced after dopamine treatment **(B)**. Mean of four independent experiments was represented with respective CI95 **(B)**. ROS production was evaluated using CellROX Green reagent and DCFDA.

Fluorescence was measured by high-throughput plate reader and normalized by total protein contents/well or by cell viability. Mean of triplicate was plotted with respective CI95 **(C)**. CellROX Green was also analyzed by high content cell analyzer. Representative field images are shown **(D)**. ∗no overlap of CI95 between indicated group and control group (0μM). scale bar = 20μM.

(**Figures 3E,F**), suggesting augment of misfolded proteins in ER. However, at 100μM, BiP expression returned to control level (**Figures 3E,F**), presumably due to the attenuated protein synthesis. On the other hand, eIF2-α phosphorylation was not altered by dopamine treatment (**Figures 3E,F**).

All protein fold changes were calculated based on the loading control GAPDH. Thus, to ensure the constitutive expression of GAPDH in our experimental conditions, we compared the GAPDH expression to α-tubulin, another frequently used loading control. As shown in **Figure 4**, GAPDH expression was not significantly changed by dopamine treatment.

Altogether, these data indicate that 100μM of dopamine compromises cellular viability, redox balance and protein synthesis, which may affect the synthesis of PrPC.

#### **DOPAMINE REDUCED THE SOLUBILITY OF PrP<sup>C</sup> AND INDUCED ITS ACCUMULATION IN AUTOPHAGOSOMES**

An intriguing fact is that even with reduced protein synthesis, we did not observe a significant reduction of the mature form of PrPC. One possibility is that dopamine treatment may reduce the turnover of PrPC. To test this hypothesis, we biotinylated cell surface protein and observed a turnover of PrPC. After 1 h of incubation, a substantial amount of biotinylated PrP<sup>C</sup> was degraded in both dopamine-treated (50μM) and untreated cells compared to input (**Figure 5A**, comparing second and third lane with first lane). However, a higher amount of biotinylated PrP<sup>C</sup> remained in dopamine-treated cells compared to untreated cells, indicating that dopamine treatment reduced PrP<sup>C</sup> turnover (**Figure 5A**, comparing second lane with third lane). To verify

whether this reduced turnover is due to the altered biochemical characteristics, we evaluated the solubility of PrP<sup>C</sup> in 1% sarkosyl. As shown in **Figure 5B**, an increased amount of PrP<sup>C</sup> was recovered in insoluble pellet fraction after treatment with 50 and 100μM of dopamine (**Figure 5B**). To further evaluate the effects of dopamine on PrP<sup>C</sup> aggregation, *in vitro* experiments were conducted using recombinant mouse PrPC (rPrP). Dopamine induced the formation of SDS-resistant higher-order oligomers of rPrP in a time and concentration-dependent manner (**Figure 5C**). Oligomerization was prevented in the presence of antioxidant (sodium metabisulfite 400μM) (**Figure 5C**, last two lanes). These results indicate that oxidative metabolites of dopamine can induce the accumulation of insoluble PrPC. In the presence of metabisulfite, monomer rPrP migrated slowly. Even after reduction of proteins with DTT, sulfhydryl groups can be readily reoxidized and form disulfide bond (Wall, 1971), and reduced recombinant PrP migrates more slowly than oxidized form (Lee and Eisenberg, 2003). Likely, metabisulfite prevented the reoxidation of sulfydryl groups, resulting in slower migration (**Figure 5C**).

Autophagy is one of the pathways that degrade prion protein (Homma et al., 2014). Since dopamine treatment induced accumulation of insoluble PrPC, we evaluated whether these aggregates were sorted to autophagosomes, labeled by LC3-II (an autophagosome membrane associated protein). As shown in **Figure 6**, the degree of colocalization between PrP<sup>C</sup> and LC3-II was increased by dopamine treatment (**Figures 6**, **7A**). Of note, the intracellular organelles that contained PrP<sup>C</sup> were frequently enlarged in cells treated with dopamine (**Figures 6**, **7B**) and they were mostly positive for LC3-II staining (**Figure 6**).

In addition, treatment of the N2a cells with 100μM of dopamine increased the levels of p62/SQSTM1 and LC3-II (**Figures 8A,B**). The former transports ubiquitinated proteins to the autophagosome and remains associated to LC3-II, which is produced by lipidation of LC3-I (Ravikumar et al., 2010). The accumulation of both proteins occurred without the induction of Beclin-1, a component of the complex that initiates the formation of autophagosome (**Figures 8A,B**) (Ravikumar et al., 2010). Thus, these data suggest that the accumulation of autophagosomes was not due to the activation of autophagy, but rather the failure in further processing of autophagosomal cargo. The immunofluorescence data also showed the accumulation and enlargement of LC3-II positive-vesicles (**Figure 6**).

Overall, these results indicate that toxic concentration of dopamine altered the solubility of PrP<sup>C</sup> and promoted its accumulation in autophagosomes, affecting autophagic flux and attenuating protein synthesis.

#### **DISCUSSION**

Dopamine is an important neurotransmitter that controls several functions such as physical movement, emotional process and alertness (Kauer and Malenka, 2007). In spite of its importance in physiology, dopamine metabolism has been pointed out as the main etiology of Parkinson's disease due to the inherent instability and reactivity of dopamine and its metabolites, capable to induce oxidative damage in biomolecules (Hastings, 2009). In this study, we demonstrated that dopamine treatment altered the solubility of PrP<sup>C</sup> and promoted its accumulation in autophagosomes in neuronal cells. Auto-oxidation of dopamine also induced the formation of SDS-resistant oligomers of unglycosylated recombinant prion protein. Previous studies have demonstrated that inhibition of complex glycosylation during synthetic pathway or expression of unglycosylated PrP facilitated its conversion into PrPSc (Korth et al., 2000; Winklhofer et al., 2003). In our study, dopamine induced aggregation of both glycosylated neuronal PrPC and unglycosylated rPrP, but with different biochemical characteristics regarding to the resistance to SDS.

Although the conversion of PrP<sup>C</sup> into abnormal prion protein (PrPSc) is an essential event for the development of TSE, molecular mechanisms of conformational transition between PrP<sup>C</sup> and PrPSc are poorly understood. Nevertheless, previous studies have demonstrated that PrPSc-like aggregates were found in normal

brains at low levels (Yuan et al., 2006). Glycosylation pattern, retrograde transport from ER to cytosol and inhibition of proteosomal activity might contribute to this spontaneous production of PrPSc-like aggregates (Korth et al., 2000; Yedidia et al., 2001; Ma and Lindquist, 2002; Winklhofer et al., 2003). The aggregation of PrP<sup>C</sup> in response to redox alteration has also been proposed as a cytoprotective mechanism (Das et al., 2010). However, if the aggregates are not effectively degraded by autophagy, other

**FIGURE 4 | Validation of GAPDH as loading control.** N2a cells were treated with dopamine at designated concentrations and cell lysates were used to evaluate the expression levels of GAPDH and **α**-tubulin in four

independent samples. The band intensity of GAPDH was normalized by the intensity of α-tubulin. Mean normalized intensity was plotted with respective CI95.

proteins might co-aggregate causing cytotoxicity (Das et al., 2010). In this study, we observed that dopamine treatment increased the amount of LC3-II and p62/SQSTM1, both involved in autophagosomal cargo recruitment. Concomitantly, PrP<sup>C</sup> was also accumulated in autophagosomes. These data indicate that the degradation of autophagosomal cargo by lysosome was not effective. Therefore, it is plausible to speculate that the reduced cell viability observed after dopamine treatment is at least partially

**FIGURE 6 | Dopamine induced PrPC sorting to autophagosomes.** After dopamine treatment, endogenous PrP<sup>C</sup> (green) and LC3-I/LC3-II (red) was detected using specific antibodies. Third column shows merged images. Right column shows colocalized pixels in gray scale. Scale bar = 10μm.

due to the accumulation of PrP<sup>C</sup> aggregates in autophagosomes. Similarly, PrP<sup>C</sup> aggregation might have negatively affected protein synthesis since a previous study has demonstrated that protein synthesis was corrupted in N2a cells infected with prions (Roffe et al., 2010).

Endocytosis of PrP<sup>C</sup> can be stimulated by several compounds such as Cu2+, nucleic acids and heme (Lee et al., 2001, 2007; Kocisko et al., 2006). An intriguing fact is that these compounds can alter the structure of PrP<sup>C</sup> and confer cytotoxicity at high concentration (Thakur et al., 2011; Macedo et al., 2012). These findings suggest that PrPC might function as a scavenging receptor for toxic molecules. Dopamine can induce toxicity by rapid auto-oxidation producing reactive dopamine-quinone (Asanuma et al., 2003; Chen et al., 2008). Binding of toxic molecules to PrPC, including dopamine metabolites, might induce structural alterations, which in turn, may trigger endocytosis for further degradation. Frequent binding of toxic molecules that cause conformational alterations of PrP<sup>C</sup> can eventually overload the degradation pathways, instigating the accumulation of protein aggregates.

Certain conditions, such as stress, drug addiction, iron deficiency, L-DOPA treatment or schizophrenia are known to increase dopamine metabolism (Nelson et al., 1997; de la Fuente-Fernandez et al., 2004; Elliott and Beveridge, 2005; Kim et al., 2005). Our findings raise a possibility that chronic exposure to these conditions may facilitate PrP<sup>C</sup> aggregation. Possibly, dopamine and its oxidative metabolites can have more general roles in other protein aggregation, but the specificity of the

**FIGURE 8 | Dopamine treatment induced accumulation of autophagosomes.** The level of Beclin-1 was reduced by 50μM of dopamine and restored to control level at 100 μM, while p62/SQSTM1 and LC3-II was increased by 100μM of dopamine **(A)**. Beclin-1 and LC3-II were analyzed in

the same gradient gel. Band intensity of Beclin-1, p62/SQSTM1, and LC3-II was normalized by the respective GAPDH band and mean of three independent experiments was plotted with respective CI95 **(B)**. ∗no overlap of CI95 between indicated group and control group (0μM).

target can be ruled by subcellular compartmentalization, brain region specificity and susceptibility of aggregation. Identification of endogenous metabolites that induce aggregation of a specific protein can contribute to better understanding of idiopathic neurodegenerative diseases and provide new molecular targets for treatment.

# **AUTHOR CONTRIBUTIONS**

Marcio H. M. da Luz conducted all experiments except rPrP oligomerization and analyzed the results. Italo T. Peres performed rPrP oligomerization assay. Tiago G. Santos acquired confocal images. Vilma R. Martins participated in study design and interpretation of data and revised the manuscript. Marcelo Y. Icimoto purified rPrP. Kil S. Lee participated in study design, data analysis and interpretation, and wrote the manuscript.

#### **ACKNOWLEDGMENTS**

This study was supported by grants from FAPESP (Fundação de Amparo à Pesquisa do Estado de São Paulo: 2008/06152- 9, 2009/14027-2, 2013/22413-5, 2012/18093-2), CAPES (Coordenação de Aperfeiçoamento de pessoal de Nível Superior), CNPq (Conselho Nacional de Desenvolvimento Científico e Tecnológico) and EMU (programa de equipamentos multiusuários).

# **SUPPLEMENTARY MATERIAL**

The Supplementary Material for this article can be found online at: http://www*.*frontiersin*.*org/journal/10*.*3389/fncel*.*2015*.* 00012/abstract

#### **REFERENCES**


brain of scrapie-infected hamsters. *Neurochem. Int.* 52, 1416–1421. doi: 10.1016/j.neuint.2008.03.002


Zahn, R., von Schroetter, C., and Wuthrich, K. (1997). Human prion proteins expressed in Escherichia coli and purified by high-affinity column refolding. *FEBS Lett.* 417, 400–404. doi: 10.1016/S0014-5793(97)01330-6

**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: 21 October 2014; accepted: 09 January 2015; published online: 02 February 2015.*

*Citation: da Luz MHM, Peres IT, Santos TG, Martins VR, Icimoto MY and Lee KS (2015) Dopamine induces the accumulation of insoluble prion protein and affects autophagic flux. Front. Cell. Neurosci. 9:12. doi: 10.3389/fncel.2015.00012 This article was submitted to the journal Frontiers in Cellular Neuroscience.*

*Copyright © 2015 da Luz, Peres, Santos, Martins, Icimoto and Lee. 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.*

# GABAergic alterations in neocortex of patients with pharmacoresistant temporal lobe epilepsy can explain the comorbidity of anxiety and depression: the potential impact of clinical factors

**Luisa Rocha<sup>1</sup>\*, Mario Alonso-Vanegas <sup>2</sup> , Iris E. Martínez-Juárez <sup>2</sup> , Sandra Orozco-Suárez <sup>3</sup> , David Escalante-Santiago<sup>3</sup> , Iris Angélica Feria-Romero<sup>3</sup> , Cecilia Zavala-Tecuapetla<sup>1</sup>† , José Miguel Cisneros-Franco<sup>2</sup> , Ricardo Masao Buentello-García<sup>2</sup> and Jesús Cienfuegos <sup>2</sup>**

<sup>1</sup> Department of Pharmacobiology, Center for Research and Advanced Studies of the National Polytechnic Institute (CINVESTAV), Mexico City, Mexico

<sup>2</sup> National Institute of Neurology and Neurosurgery "Manuel Velasco Suarez", Mexico City, Mexico

<sup>3</sup> Unit for Medical Research in Neurological Diseases, National Medical Center, Mexico City, Mexico

#### **Edited by:**

Rosalinda Guevara-Guzman, Universidad Nacional Autónoma de México, Mexico

#### **Reviewed by:**

Roberto Di Maio, University of Pittsburgh, USA Alberto Lazarowski, University of Buenos Aires, Argentina

#### **\*Correspondence:**

Luisa Rocha, Department of Pharmacobiology, Center of Research and Advanced Studies, Calz. Tenorios 235. Col. Granjas Coapa., Mexico City 14330, Mexico

e-mail: lrocha@cinvestav.mx

#### **†Present address:**

Cecilia Zavala-Tecuapetla, Nanotechnology Laboratory, National Institute of Neurology and Neurosurgery "Manuel Velasco Suárez", Mexico City, Mexico

Temporal lobe epilepsy (TLE) is a chronic neurodegenerative disease with a high prevalence of psychiatric disorders. Temporal neocortex contributes to either seizure propagation or generation in TLE, a situation that has been associated with alterations of the γ-aminobutyric acid (GABA) system. On the other hand, an impaired neurotransmission mediated by GABA in temporal neocortex has also been involved with the pathophysiology of psychiatric disorders. In spite of these situations, the role of the necortical GABA system in the comorbidity of TLE and mood disorders has not been investigated. The present study was designed to identify alterations in the GABA system such as binding to GABA<sup>A</sup> and GABA<sup>B</sup> receptors and benzodiazepine site, the tissue content of GABA and the expression of the mRNA encoding the α1–6, β1–3, and γ GABA<sup>A</sup> subunits, in the temporal neocortex of surgically treated patients with TLE with and without anxiety, and/or depression. Neocortex of patients with TLE and comorbid anxiety and/or depression showed increased expression of the mRNA encoding the γ2-subunit, reduced GABAB-induced G-protein activation in spite of elevated GABA<sup>B</sup> binding, and lower tissue content of GABA when compared to autopsy controls. Some of these changes significantly correlated with seizure frequency and duration of epilepsy. The results obtained suggest a dysfunction of the GABAergic neurotransmission in temporal neocortex of patients withTLE and comorbid anxiety and/or depression that could be also influenced by clinical factors such as seizure frequency and duration of illness.

**Keywords: GABA receptors, G-protein, temporal lobe epilepsy, temporal neocortex, anxiety, depression**

#### **INTRODUCTION**

It is known that a high percentage of patients with epilepsy have comorbid interictal psychiatric disorders (Briellmann et al., 2007; Kanner, 2011). This situation can be explained by pathophysiological mechanisms associated with specific neurotransmitters in the brain areas involved in both, epilepsy and psychiatric conditions (Rocha et al., 2014).

Temporal neocortex is a paralimbic structure that belongs to a visceromotor system associated with mood, emotions, and visceral reactions to emotional stimuli (Beauregard et al., 1988; Partiot et al., 1995;Ongür et al., 1998;Drevets et al., 2008). Neuroanatomical studies in non-human primates indicate that temporal cortex is involved in the sensory integration as well as codification of affective characteristics of stimuli (Ongür and Price, 2000; Saleem et al., 2007). Abnormalities in gray matter volume and glucose metabolism have been detected in the temporal neocortex of patients with mood disorders (Ongür et al., 2003). Temporal neocortex of patients with depression presents overactivation (Sheline et al., 2009) as well as abnormalities in cell communication and signal transduction systems identified by transcriptional profiling (Aston et al., 2005). Alterations of γ-amino-butyric acid (GABA) system in temporal neocortex have been proposed to participate in the pathophysiology of mood disorders (Nikolaus et al., 2010).

On the other hand, abnormal neurotransmission mediated by GABA system in the cortex has been suggested to play an important role in seizure generation and/or propagation (Chagnac-Amitai and Connors, 1989) as well as in the neuronal overactivation detected in this brain of patients with temporal lobe epilepsy (TLE) (Avoli et al., 1995; Koepp et al., 2000; Teichgräber et al., 2009). Nevertheless, studies aimed to identify alterations on GABA receptor binding in the temporal neocortex of patients with pharmacoresistant TLE have shown dissimilar results (la Fougère et al., 2009).

Although previous studies suggest that GABA disturbances in the temporal neocortex participate in the pathophysiology of TLE and comorbid mood disorders (Kondziella et al., 2007), this idea has not been investigated. The present study was focused to evaluate a possible association between alterations in the GABAergic system in the temporal neocortex of patients with pharmacoresistant TLE and comorbid anxiety and/or depression. Experiments were designed to analyze the binding to GABA<sup>A</sup> receptors involved in tonic ([3H]-Muscimol) and phasic ([3H]- Flunitrazepam) neurotransmission, as well as GABA<sup>B</sup> receptor binding. We also evaluated the G-protein activation mediated by GABA<sup>B</sup> receptors, the tissue content of GABA and the mRNA expression of some GABA<sup>A</sup> receptor subunits. Values obtained were correlated with clinical data to identify those clinical factors that could be involved in specific alterations of the GABAergic system.

#### **MATERIALS AND METHODS**

#### **PATIENTS CRITERIA AND SURGICAL SAMPLES**

Biopsy samples of temporal neocortex were obtained from 26 patients with diagnosis of pharmacoresistant TLE: 16 patients with mesial TLE, nine patients with TLE secondary to tumor or lesion, and one patient with dual pathology namely mesial TLE and tumor. All patients underwent epilepsy surgery after been submitted to an extensive pre-surgical evaluation according to the protocol of the Epilepsy Surgery Program of the National Institute of Neurology and Neurosurgery "Manuel Velasco Suarez," in Mexico (Table S1 in Supplementary Material).

The pre-surgical evaluation consisted of neurological evaluation, electroencephalogram (EEG) and video-EEG recordings, single photon emission computed tomography (SPECT) or positron emission tomography (PET), neuropsychological and neuropsychiatric evaluation, and magnetic resonance imaging (MRI). MRI findings were in concordance with those found in EEG recordings.

During the neurological evaluation, prevalence of depression and anxiety disorders was established using the Structured Clinical Interview for DSM-IV Axis I (SCID-I) (First et al., 1999) applied by a Psychiatrist, who was blinded to the epilepsy diagnosis. Spanish version of the Hospital Anxiety and Depression Scale (HADS) were applied to all patients to identify symptoms of anxiety and/or depression. HADS has been previously validated in a Spanish population (Herrero et al., 2003; Gómez-Arias et al., 2012). HADS scale considers symptoms over the previous week and is not affected by coexisting general medical conditions. Patients with other psychiatric or somatic disturbances interfering with mood disorders, such as addiction, were excluded from the present study.

The neurosurgeon (Mario Alonso-Vanegas) carried out all the surgeries. Patients with mesial TLE underwent epilepsy surgery using a T2 or T3 transtemporal approach and guided by electrocorticographic (ECoG) signals recorded from the brain surface (4 × 8-electrode grids, Ad-Tech, Racine, WI). The epilepsy surgery consisted of unilateral amygdalo-hippocampectomy that included removal of the uncus and parahippocampal gyrus, and the tailored resection of T2 and T3 (San-Juan et al., 2011). Patients with tumor or lesion, with and without TLE had a similar surgical procedure unaided by ECoG recordings, with standard temporal neocortical resection. In these patients, amygdalo-hippocampectomy was performed depending on the localization of the lesion or tumor and neuropsychological findings. After resection, T2 and T3 gyri were immediately frozen in milled dry ice and kept at −70°C until processing. When the tumor or lesion was restricted to the temporal neocortex, samples from the margins of the lesion were used for present study. The protocol did not include biopsies with tumor, cortical malformations or any cortical alteration identified by neuropathological evaluation.

The present study was approved by the scientific committees of the institutions involved in this research and informed authorization and consent were obtained from each patient.

#### **AUTORADIOGRAPHY EXPERIMENTS**

Previous studies specify that receptor binding and guanosine 5<sup>0</sup> - O-[γ-thio [35S]] triphosphate ([35S]GTPγS) binding stimulation by selective agonists to GABA<sup>B</sup> receptors are preserved for several hours after death (González-Maeso et al., 2000), while longer post-mortem delay has been associated with increased binding to benzodiazepine (BDZ) sites (Whitehouse et al., 1984). Considering this information, binding values acquired from autopsies of six men who died as consequence of diverse causes without clinical data of neurologic or psychiatric disorders with a postmortem interval of 2–14 h were compared to those obtained from the patients with pharmacoresistant TLE (Table S1 in Supplementary Material). T2 and T3 gyri were dissected at the time of the autopsy and quickly kept at −70°C. For each autoradiography assay, tissue samples of the different patients and autopsies were processed together in order to reduce the experimental variability.

#### **Preparation of tissue sections**

Frozen sections of 20µm were cut in a cryostat, thaw-mounted on gelatin-coated slides, and kept again at −70°C. Serial and parallel sections were obtained from each biopsy/autopsy for subsequent quantitative and functional autoradiography procedures.

#### **Quantitative autoradiography**

**Table 1** includes a summary of the different protocols for the quantitative autoradiography experiments. Brain sections were removed from the freezer, dried in a stream of cool air, and immediately washed to eliminate endogenous ligands. Then, sections were incubated in a solution with the specific ligand labeled with tritium ([*3*H]), in presence or absence of a non-labeled specific ligand. The specific binding values were established from the difference of values obtained from both experimental conditions. Incubation was concluded with two consecutive washes in buffer solution and a final rinsed with distilled water was carried out for 2 s at 4°C. Slices were quickly dried in a mild steam of cold air.

Slices from patients and autopsies as well as [*3*H] standards (Amersham) were arranged together in X-ray cassettes and all of them were exposed to [*3*H]-sensitive film (Kodak MR) at 22°C. After the appropriate exposure time (see **Table 1**), the film was developed at 18–20°C in Kodak D19 developer and fast fixer. Optical densities of cortical layers of each tissue sample were evaluated in three different sections using the JAVA Jandel image analysis software. Temporal neocortex was subdivided for autoradiographic analysis into an outer layer (cortical layers I and II), middle layer (cortical layers III and IV), and an inner layer (cortical layers V and VI). The distribution of receptor binding sites, as revealed by optical densities of the autoradiograms obtained


#### **Table 1 | Conditions for quantitative autoradiography experiments**.

SA, specific activity; RT, room temperature.

by outlining each layer, was matched to the cortical layers visualized on the sections counter-stained with 0.5% Cresyl Violet (**Figure 2**). Finally, the generation of a standard curve using the optical density values of the standards, the specific activity of each [ *<sup>3</sup>*H]-labeled ligand, and tissue thickness (20µm) was used to convert radioactivity values in fmol/mg of protein.

#### **Functional autoradiography**

Sections were washed in Tris buffer (50 mM Tris–HCl, 3 mM MgCl2, 0.2 mM EGTA, 100 mM NaCl, pH 7.4 at 25°C for 10 min), then incubated in the assay buffer containing 2 mM GDP (25°C for 15 min). GABAB-induced G-protein activation was evaluated in sections subsequently incubated in the same assay buffer with 2 mM GDP and 0.04 nM [35S]GTPγS (25°C for 2 h) in the presence of baclofen (100µM), a GABA<sup>B</sup> receptor agonist. In parallel sections, the effects of baclofen were evaluated in the presence of a GABA<sup>B</sup> antagonist (CGP55845A, 10 mM). Basal binding was determined in sections incubated in similar conditions, but lacking agonist and antagonist drugs. Thereafter, slices were washed twice for 2 min each in assay buffer (4°C, pH 7.4) and once in distilled water (4°C). Sections from autopsies and patients were dried overnight and exposed to film (Kodak-MR) for 5 days at 22°C in Xray cassettes containing [14C] microscales (American Radiolabeled Chemicals, Inc.). Optical density analysis of the different cortical layers was carried out as previously described for quantitative autoradiographic experiments. Results obtained from the different assays were expressed as nanocuries of [35S] per milligram of tissue. Net agonist-stimulated [35S]GTPγS binding was calculated in percentage by subtracting basal binding from agonist-stimulated binding.

#### **SEMIQUANTITATIVE RT-PCR ANALYSIS**

Expression levels of human GABA<sup>A</sup> receptor subunit mRNAs (α1 to α6-, β1- to β3-, and γ1- to γ3-subunits) were determined from resected human tissues of autopsies and patients with pharmacoresisant TLE by semiquantitative RT-polymerase chain reaction (PCR) procedure. For this purpose, a fragment was obtained from each brain sample of patients with epilepsy and from the autopsies. The fragmentation was done maintaining the tissue frozen, a situation that allowed to preserve the mRNA.

Total RNA was isolated using the TriPure isolation Reagent (ROCHE, USA) according to the manufacturer's instructions. Reverse transcriptase (RT) of 3µg of total RNA was synthesized to single-stranded cDNA using random primers (Promega, USA) and 200 U of M-MuLV RT (New England Biolabs, USA)

in a total reaction volume of 10µl. The RT reaction was performed for 10 min at 25°C, 50 min at 37°C and 15 min at 70°C according to the manufacturer's instructions. After RT, 20µl of ultrapure-grade water was added to the final reaction. Gene amplifications were performed using 1µl of diluted cDNA, 0.2–0.4µl of 60 mM MgCl2, 0.2µl of dNTP mix solution (Promega, USA), 0.2µl of each primer pair (10 mM) and 0.2µl of Taq DNA polymerase (Invitrogen, USA) in a total volume of 10µl. The final MgCl<sup>2</sup> concentration depended on each gene evaluated. The PCR was carried out in a DNA Thermal Cycler (Bio-Rad, USA) with a cycle program of 94°C for 3 min, 30–35 cycles at 94°C for 30 s, 54–58°C for 20–30 s, 72°C for 30–40 s, and one cycle of 72°C for 10 min. The cycle program was performed until the exponential phase was achieved; other PCR conditions were performed individually. The PCR products were separated by electrophoresis in a 2% agarose gel (Invitrogen) in TAE buffer at 75 V. The gels were captured and evaluated in an Alpha Innotech corporation IS-1000 digital imaging system, using EtBr (0.6 g/ml by gel) under UV light. Genomic DNA contamination was checked by carrying samples through a PCR procedure without adding reverse transcriptase.

The bands from images were analyzed using the NIH Image J system version 1.46 (http://imagej.nih.gov/ij/) and quantified as values of integrated density. The relative value of each gene was a ratio between the expression of each gene and β-actin as control.

#### **TISSUE CONTENT OF GABA**

Evidence exists indicating a significant degradation by proteolysis of the GABA-synthesizing enzyme (GAD) and reduced GABA tissue content in the brain tissue within hours after death (Lowe et al., 1988;Martin et al., 2003). Since this condition can represent a potential problem in the evaluation of the tissue content of GABA in autopsy samples, temporal neocortex of six patients (four men and two women) submitted to surgery with diagnosis of cerebral tumor without epilepsy was used as control tissue to be compared with values obtained from patients with pharmacoresistant TLE (Table S1 in Supplementary Material).

Gray matter (50–100 mg) of each brain sample was thawed and manually homogenized in perchloric acid (0.1 M, J. T. Baker). The homogenates were centrifugated at 13,200 rpm (15 min at 4°C) using a centrifuge (Eppendorf 5415R). Samples of the supernatant (100µl) previously filtered (Nalgene filters of 0.45µm) were suspended in 0.1 M perchloric acid in a 1:250 proportion. Subsequently, 20µl of the filtered supernatant were mixed with 6µl of o-phthalaldehyde (OPA) and agitated for 30 s. Two minutes later, the mixture was injected into the solvent stream of a high performance liquid chromatography (HPLC) system. For GABA quantification, the procedure required that OPA-amino acid were separated on a reversed-phase 3.9 × 150 mm column (Nova-Pack, 4µm, C18, Waters®) with solution A (sodium acetate dissolved in 90% miliQ water and 10% methanol; pH 5.75 with glacial acetic acid) as aqueous solvent and solution B as the other mobile phase (20% solution A and 80% methanol; pH 6.75 with glacial acetic acid) at a flow rate of 0.5 ml/min. Content of GABA was determined with fluorescent detection (Waters® model 474) by peak height measurements against standard solutions (Kendrick et al., 1988).

The pellets obtained from the centrifugation process were used to determine the amount of proteins (Lowry et al., 1951), a situation that allowed expressing in micromoles per milligram of proteins the values resulting from the fluorometric HPLC procedure.

#### **STATISTICAL ANALYSIS**

The results obtained were expressed as mean ± SE and analyzed employing ANOVA test and Bonferroni *post hoc* test. Pearson's correlation coefficients were estimated to establish the potential impact of clinical factors such as patient's age, age at seizure onset, duration of epilepsy, and seizure frequency on the GABAergic system.

#### **RESULTS**

#### **CLINICAL CHARACTERISTICS**

Anxiety and depression were detected in 10 patients with TLE. Their clinical data were as follows (mean ± SE): age of patients, 36.4 ± 3 years (ranged from 24 to 48 years); age at seizure onset, 14.2 ± 4.5 years; years of epilepsy duration,22.2 ± 3.3; and seizures per month, 16.4 ± 4.5. Regarding pharmacological therapy that these patients received during the epilepsy process and pre-surgical period, the following information was identified: (a) 90% (*n* = 9) received polytherapy [from 4 to 10 antiepileptic drugs (AEDs)]; (b) 80% (*n* = 8) were treated with AEDs that could induce psychiatric adverse effects, such as depression (levetiracetam, primidone, zonisamide, topiramate, and phenobarbital); (c) 90% (*n* = 9) received 2 or more AEDs with GABAergic properties (clobazam, clonazepam, diazepam, levetiracetam, lamotrigine, valproic acid, phenobarbital, and primidone); and (d) 80% (*n* = 8) were treated with AEDs that could induce positive effects on the mood (valproic acid, carbamazepine, and gabapentin). Four patients (40%) had been previously diagnosed with depression and treated with antidepressant drugs (amitriptyline, duloxetine, sertraline, or fluoxetine) during 9 months to 6 years before the epilepsy surgery. Six patients (60%) were diagnosed with anxiety and/or depression during the pre-surgical evaluation, but they did not receive pharmacotherapy for these disorders (Table S1 in Supplementary Material).

Preclinical evaluation did not reveal neuropsychiatric comorbidity in 16 patients with pharmacoresistant TLE. Their clinical data (age, 30.6 ± 2.2 years old, ranged from 17 to 60 years old; age at seizure onset, 12.6 ± 2.1 years; years of epilepsy duration, 18.9 ± 3.1; and seizures per month, 16.5 ± 5.7) were not significantly different when compared to those patients with TLE and comorbid anxiety and/or depression. The pharmacological

therapy that these patients received during the epilepsy process and pre-surgical period was similar to that administered to the patients with TLE and comorbid anxiety and depression: (a) 75% (*n* = 12) received polytherapy (from 3 to 7 AEDs); (b) 50% (*n* = 8) were treated with AEDs that may produce mood disorders; (c) 81% (*n* = 13) received 1 or more AEDs with GABAergic effects, and (d) 81% (*n* = 13) were treated with AEDs that induce a positive effect on the mood (Table S1 in Supplementary Material).

The mean age of patients with TLE with and without anxiety and/or depression was similar to the mean age of autopsies (39.5 ± 3.4 years, ranging from 29 to 51 years, *p* < 0.51) and patients with cerebral tumor without epilepsy and psychiatric disorders (39.6 ± 6.6 years, ranging from 25 to 63 years, *p* < 0.6) (Table S1 in Supplementary Material).

Nissl staining revealed a normal cytoarchitecture with no evident neuronal cell loss, cortical dysplasias, malformations, or tumor in the different tissues evaluated.

#### **CONTROL AND AUTOPSY SAMPLES**

Samples obtained from control patients with tumor without epilepsy presented 20.3 ± 1.3µM/mg of protein of GABA tissue

GABA<sup>A</sup> and GABA<sup>B</sup> receptors as well as benzodiazepine (BDZ) sites, labeled with [<sup>3</sup>H]-Muscimol, [<sup>3</sup>H]-CGP54626, and [<sup>3</sup>H]-Flunitrazepam, respectively. High binding appears as red and orange areas, low binding is indicated as yellow and green areas, whereas blue areas represent absence of binding.

levels (**Figure 1**). In autopsy samples, binding to [3H]-Muscimol (GABA<sup>A</sup> receptors) was widely distributed through the various cortical layers. Binding to [3H]-CGP54626 (GABA<sup>B</sup> receptors) demonstrated a gradient across the cortical layers, showing the highest in outer layer. Binding to [3H]-Flunitrazepam (BDZ sites) was elevated in outer and middle layers (**Figure 2** and **Table 2**). Functional autoradiography revealed [35S]GTPγS incorporation as consequence of the GABAB-induced G-protein activation (153% in layers I–II, 130% in layers III–IV, and 137% in layers V–VI) (**Table 2**).

In autopsy samples, the mRNA expression of the GABA<sup>A</sup> receptor subunits was unrelated to the age of the subjects and time required to obtain the tissue. High mRNA levels were observed for subunits β1, β2, β3, and γ1, whereas subunits α1–6, γ2, and γ3 were less prominent.

#### **PATIENTS WITH TLE**

Patients with TLE without psychiatric disorders presented a nonsignificant decrease of the GABA tissue content (22%, *p* > 0.05) when compared to the autopsies. In contrast, the temporal neocortex of subjects with TLE and comorbid anxiety and/or depression showed lower tissue content of this amino acid (42%, *p* < 0.05), a situation that correlated with a higher seizure frequency (*r* = −0.5822, *p* < 0.05, **Figure 1**).

Concerning binding evaluation, neocortex of patients with TLE with and without anxiety and/or depression did not show significant abnormalities in [3H]-Muscimol binding (**Figure 3**, **Table 2**). Patients without comorbid psychiatric disturbances showed higher [3H]-Flunitrazepam binding in all neocortex (layers I–II, 86%, *p* < 0.05; layers III–IV, 58%, *p* < 0.05; and layers V–VI, 103%, *p* < 0.001) compared to autopsies. In these patients,

**Table 2 | [3H]Ligand binding to GABA<sup>A</sup> and GABA<sup>B</sup> receptors as well as BDZ site, and GABAB-induced G-protein activation in specific cortical layers of autopsies and samples of patients with temporal lobe epilepsy with (A/D) and without (No A/D) anxiety and depression**.


Binding values are expressed as mean ± SME of fmol/mg of protein. The results of the GABAB-induced G-protein activation are expressed as mean ± SME of percentage of specific [<sup>35</sup>S]GTPγS binding with respect to basal value (100%). Statistical comparison was made using ANOVA and a post hoc Bonferroni test. \* p < 0.05; \*\*p < 0.01; \*\*\*p < 0.001, when compared with autopsy group.

**binding to GABA<sup>A</sup> receptors (- - - line) and benzodiazepine (BDZ) sites (——- line) in temporal neocortex (layers I–II, III–IV, and V–VI) of patients withTLE with (A/D, left side) and without (No A/D, right side) anxiety and depression, with respect to autopsies (100%)**. \*p < 0.05, \*\*\*p < 0.001.

the higher [3H]-Flunitrazepam binding correlated with a lower seizure frequency (layers I–II, *r* = 0.6582, *p* < 0.01; and layers III– IV, *r* = 0.6672, *p* < 0.01) and a shorter duration of epilepsy (layers III–IV, *r* = 0.5285, *p* < 0.05; layers V–VI, *r* = 0.4914, *p* < 0.05) (**Figure 4**). [3H]-Flunitrazepam binding in neocortex of patients

with TLE and anxiety and/or depression was not significantly different from autopsy group (**Figure 3**; **Table 2**).

RT-PCR experiments revealed high α4-subunit expression (114%, *p* < 0.05) in temporal neocortex of patients with TLE without mood disorders when compared with autopsies. The expression of the β-subunitis was not modified, whereas the γ2 subunit expression was increased in patients with (126%, *p* < 0.05) and without mood disorders (130%, *p* < 0.05).

[ <sup>3</sup>H]-CGP54626 binding was increased in both, patients with (layers III–IV, 102%, *p* < 0.01; layers V–VI, 88%, *p* < 0.01) and without anxiety and/or depression (layers III–IV, 74%, *p* < 0.01; layers V–VI, 90%, *p* < 0.01) (**Figure 5**; **Table 2**). In contrast, [ <sup>35</sup>S]GTPγS incorporation as consequence of activation of GABA<sup>B</sup> receptors was lower in all cortical layers of patients with TLE and comorbid anxiety and/or depression (layers I–II, 47%, *p* < 0.05; layers III–IV, 56%, *p* < 0.01; and layers V–VI, 46%, *p* < 0.01) (**Figure 5** and **Table 2**). In these patients, the lower GABABinduced [35S]GTPγS incorporation correlated with a higher seizure frequency (layers III–IV, *r* = 0.7380, *p* < 0.05; layers V– VI, *r* = 0.8859, *p* < 0.01). Patients without psychiatric disturbances showed a lower GABAB-induced [35S]GTPγS incorporation restricted to deep layers (V–VI, 33%, *p* < 0.05) (**Figure 5**; **Table 2**), a situation that correlated with a higher seizure frequency (*r* = 0.5317, *p* < 0.05) and a longer duration of epilepsy (*r* = 0.4975, *p* < 0.05).

#### **DISCUSSION**

Previous reports indicate important changes of the GABAergic neurotransmission in the neocortex of patients with TLE (Avoli et al., 1995; Koepp et al., 2000; Teichgräber et al., 2009). However, data concerning abnormalities on the GABA receptor binding in the lateral temporal neocortex of patients with pharmacoresistant TLE are still controversial (la Fougère et al., 2009). The results obtained in the present study could explain this disagreement suggesting that specific alterations in the neurotransmission mediated by GABA in the temporal neocortex of subjects with pharmacoresistant TLE are associated with the comorbidity of anxiety and/or depression. Our results also indicate that clinical factors such as seizure frequency and epilepsy duration may play an important role in the disturbances of the GABAergic system in the temporal neocortex of patients with pharmacoresistant TLE and comorbid anxiety and/or depression.

[ <sup>3</sup>H]-Muscimol binds to the same site on the GABA<sup>A</sup> receptor complex as GABA itself (Frølund et al., 2002), and it labels extrasynaptic GABA<sup>A</sup> receptors containing δ- and α4-subunits (Chandra et al., 2010). Extrasynaptic receptors are relevant sensors for ambient GABA and modulate the tonic inhibition (Nyitrai et al., 2006). We found no significant changes in [3H]-Muscimol binding in the neocortex of patients with TLE, regardless of the presence of anxiety and/or depression, suggesting that the binding to extrasynaptic GABA<sup>A</sup> receptors containing δ- and α4-subunits

is not altered and the GABAergic neurotransmission is preserved. Indeed, our RT-PCR experiments revealed enhanced expression of the mRNA encoding the α4-subunit in patients without anxiety and/or depression, a situation that could be related to enhanced tonic inhibition or may merely reflect compensatory changes.

[ <sup>3</sup>H]-Flunitrazepam is a BDZ that labels GABA<sup>A</sup> receptors containing γ2-subunit in combination with α1–3 subunits and are responsible for mediating phasic inhibition at synaptic sites (Lavoie and Twyman, 1996). BDZ can functionally potentiate the effects induced by sub-maximal concentrations of GABA by enhancing GABA affinity (Olsen, 1981). Previous studies in patients with TLE indicate an increased expression of γ2-subunit (Sperk et al., 2009), whereas the BDZ binding results are controversial (la Fougère et al., 2009). From our results, it seems that patients with TLE and mood disorders demonstrate absence of changes in BDZ binding in the temporal neocortex, but an upregulation of γ2-subunit. It remains to be elucidated whether, in these patients, the upregulation of γ2-subunit in temporal neocortex is primary or is secondary to the disease, and represents a compensatory response to the GABA deficit. Concerning patients with TLE without anxiety and depression, we found augmented BDZ binding and increased expression of γ2-subunits. This situation can be associated with fast synaptic GABA-induced inhibition and reduction in anxiety-like and depressive-like behaviors (Earnheart et al., 2007; Vithlani et al., 2013).

In agreement with the increased expression of the α4- and γsubunits, and the absence of changes in GABAA/BDZ binding found in the present study, results obtained from other authors using resected hippocampal tissue from patients with TLE support the idea that the inhibition mediated by GABAergic system is conserved or even augmented (Babb et al., 1989; Mathern et al., 1995). However, an important condition to be considered is that a

low ambient GABA represents an inadequate situation to activate GABA<sup>A</sup> receptors mediating tonic inhibition with a consequent facilitation of seizure activity, depression-like, and anxiety-like behaviors (Merali et al., 2004; Maguire and Mody, 2008; Hines et al., 2012). Our experiments revealed reduced GABA tissue levels in the neocortex of patients with TLE and comorbid anxiety and depression, a situation that can be associated with the low GABA release in pharmacoresistant epilepsy described by other authors (During and Spencer, 1993; Luna-Munguia et al., 2011).

Activation of GABA<sup>B</sup> receptors starts several signaling cascades at pre- and postsynaptic levels essential for the stability of the cortical network activity and modulation of gamma oscillations essential for cognitive processes (Kohl and Paulsen, 2010). In contrast, a deficiency in the neurotransmission mediated by GABA<sup>B</sup> receptors has been associated with changes in cortical gamma oscillations with consequent psychiatric disorders such as schizophrenia (Uhlhaas and Singer, 2010). We found that patients with TLE without anxiety and/or depression presented decreased GABAB-induced G-protein activation restricted to layers V–VI. In contrast, patients with TLE and comorbid anxiety and/or depression showed reduced GABA<sup>B</sup> receptor-induced G-protein activation in all cortical layers. These findings associated with low GABA tissue content suggest a deficit in the neurotransmission induced by GABA<sup>B</sup> receptors in temporal cortex of patients with TLE and comorbid anxiety and/or depression, a situation similar to that found in patients with mood disorders (Enna and Bowery, 2004; Croarkin et al., 2011). Indeed, the high GABA<sup>B</sup> binding found in cortical layers III–IV and V–VI (present study), as well as the up-regulation of GABA<sup>B</sup> in remaining neurons in hippocampus (Princivalle et al., 2002) can represent a compensatory mechanism resulting from the deficient neurotransmission mediated by these receptors in patients with TLE.

The comorbidity of mood disorders in patients with epilepsy has been associated with a higher perception of side effects of the therapy with AEDs (Gómez-Arias et al., 2012). AEDs therapy can induce uncoupling of GABA/BZD site interactions and alterations in GABA<sup>A</sup> and GABA<sup>B</sup> function (Mula and Sander, 2007); these changes may result from the repetitive administration of BDZs and/or AEDs that enhance GABA exposure such as tiagabine and vigabatrin (Suzuki et al., 1991; Gravielle et al., 2005; Perici´c et al., 2007). Carbamazepine and valproic acid increase the number of GABA<sup>B</sup> binding sites in the rat hippocampus when applied chronically, an effect that has been associated with the mood stabilizing effects induced by these drugs (Motohashi,1992). In contrast, long-term exposure to these AEDs may decrease the GABA<sup>B</sup> receptor function, an effect analogous to that produced by the subchronic administration of baclofen (Pacey et al., 2011). Our results do not support a significant association between the pharmacotherapy with AEDs inducing GABAergic effects in temporal neocortex and the comorbid anxiety and/or depression in patients with TLE. However, future studies including other brain areas and a larger number of patients should be carried out to support this hypothesis.

Affective disorders have been associated with disturbances in the second messenger signaling via G-protein function (Pacheco et al., 1996). Hyperfunction of G proteins leads to characteristics of a manic or depressive state caused by instability in the activities of protein kinases C (Avissar and Schreiber, 1992), a family of enzymes that are involved in the signal transduction mechanism of the GABA<sup>B</sup> receptors (Taniyama et al., 1992). This study indicates an imbalance of the G-protein activation induced by stimulation of GABA<sup>B</sup> receptors in the temporal neocortex of patients with TLE and comorbid anxiety and/or depression. This situation should be kept in mind when considering these receptors as promising targets for the therapy of psychiatric disorders associated with TLE.

The comorbidity of mood disorders in TLE has also been related to a higher seizure frequency (Grabowska-Grzyb et al., 2006; Mula and Sander, 2007; Peng et al., 2014) and longer duration of active epilepsy (Gonçalves and Cendes, 2011). Our results revealed that the lower values for BDZ binding, GABABinduced G-protein activation, and GABA tissue content correlated with a higher seizure frequency and the longer duration of epilepsy of patients with TLE and comorbid anxiety and/or depression. In contrast, the higher BDZ binding and elevated GABAB-induced G-protein activation correlated with a lower seizure frequency of patients with TLE without psychiatric disturbances. It is possible that the higher seizure frequency and longer duration of epilepsy augment the exposure to elevated extracellular GABA levels during the ictal period. This situation may result in GABA<sup>A</sup> desensitization, dysregulation of the neurotransmission mediated by GABA<sup>B</sup> receptors, and uncoupling of GABA/BZD site interactions (Gravielle et al., 2005), conditions that could facilitate the comorbid anxiety and/or depression of patients with TLE. Finally, further experiments should be carried out to identify if the receptor binding alterations detected in the present study are the consequence of changes in the number or affinity of the receptors evaluated. We also suggest future studies in patients with anxiety and depression disorders but without epilepsy to explain the findings obtained in the present study.

#### **AUTHOR CONTRIBUTIONS**

The author Luisa Rocha designed the study and prepared the manuscript. The author Mario Alonso-Vanegas performed the epilepsy surgery to all the patients. The author Iris E. Martínez-Juárez carried out the pre-surgical evaluation of the patients with pharmacoresistant epilepsy. The authors Sandra Orozco-Suárez, David Escalante-Santiago, and Iris Angélica Feria-Romero performed the semiquantitative RT-PCR analysis. The author Cecilia Zavala-Tecuapetla carried out the functional autoradiography experiments. Under the supervision of Mario Alonso-Vanegas, the authors José Miguel Cisneros-Franco, Ricardo Masao Buentello-García, and Jesús Cienfuegos analyzed the clinical data of the patients and their correlation with the results obtained. All authors contributed to manuscript revisions and have approved the final manuscript.

#### **ACKNOWLEDGMENTS**

We thank Francia Carmona and the staff members of the outpatient Epilepsy Surgery Clinic for their support. This study received funding from Consejo Nacional de Ciencia y Tecnología (Grant 98386).

#### **SUPPLEMENTARY MATERIAL**

The Supplementary Material for this article can be found online at http://www.frontiersin.org/Journal/10.3389/fncel.2014.00442/ abstract

### **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 October 2014; accepted: 08 December 2014; published online: 05 January 2015.*

*Citation: Rocha L, Alonso-Vanegas M, Martínez-Juárez IE, Orozco-Suárez S, Escalante-Santiago D, Feria-Romero IA, Zavala-Tecuapetla C, Cisneros-Franco JM, Buentello-García RM and Cienfuegos J (2015) GABAergic alterations in neocortex of patients with pharmacoresistant temporal lobe epilepsy can explain the comorbidity of anxiety and depression: the potential impact of clinical factors. Front. Cell. Neurosci. 8:442. doi: 10.3389/fncel.2014.00442*

*This article was submitted to the journal Frontiers in Cellular Neuroscience.*

*Copyright © 2015 Rocha, Alonso-Vanegas, Martínez-Juárez, Orozco-Suárez, Escalante-Santiago, Feria-Romero, Zavala-Tecuapetla, Cisneros-Franco, Buentello-García and Cienfuegos. 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.*

# HDAC4 as a potential therapeutic target in neurodegenerative diseases: a summary of recent achievements

#### *Michal Mielcarek1 \*, Daniel Zielonka2, Alisia Carnemolla1, Jerzy T. Marcinkowski <sup>2</sup> and Fabien Guidez <sup>3</sup>*

*<sup>1</sup> Department of Medical and Molecular Genetics, King's College London, London, UK*

*<sup>2</sup> Department of Social Medicine, Poznan University of Medical Sciences, Poznan, Poland*

*<sup>3</sup> INSERM UMRS 1131, Université Paris Diderot, Institut Universitaire d'hématologie (IUH), Hôpital Saint-Louis, Paris, France*

#### *Edited by:*

*Marco Antonio Meraz-Ríos, Centro De Investigación Y De Estudios Avanzados, Mexico*

#### *Reviewed by:*

*Vincenzo De Paola, Imperial College London, UK Henry Markram, Ecole Polytechnique Federale de Lausanne, Switzerland*

#### *\*Correspondence:*

*Michal Mielcarek, Department of Medical and Molecular Genetics, School of Medicine, King's College London, 8th Floor Tower Wing, Guy's Hospital Great Maze Pond, London, SE1 9RT, UK e-mail: mielcarekml@gmail.com*

For the past decade protein acetylation has been shown to be a crucial post-transcriptional modification involved in the regulation of protein functions. Histone acetyltransferases (HATs) mediate acetylation of histones which results in the nucleosomal relaxation associated with gene expression. The reverse reaction, histone deacetylation, is mediated by histone deacetylases (HDACs) leading to chromatin condensation followed by transcriptional repression. HDACs are divided into distinct classes: I, IIa, IIb, III, and IV, on the basis of size and sequence homology, as well as formation of distinct repressor complexes. Implications of HDACs in many diseases, such as cancer, heart failure, and neurodegeneration, have identified these molecules as unique and attractive therapeutic targets. The emergence of HDAC4 among the members of class IIa family as a major player in synaptic plasticity raises important questions about its functions in the brain. The characterization of HDAC4 specific substrates and molecular partners in the brain will not only provide a better understanding of HDAC4 biological functions but also might help to develop new therapeutic strategies to target numerous malignancies. In this review we highlight and summarize recent achievements in understanding the biological role of HDAC4 in neurodegenerative processes.

**Keywords: histone deacetylase, signaling, HDAC4, neurodegeneration, HDAC inhibitors, therapeutic potential**

#### **INTRODUCTION**

Transcription is a multistep process and its regulation involves a balanced coordination of several molecular factors. Epigenetic modifications of chromatin, including histone acetylation, represent priming events in the cascade leading to gene expression and are governed by the antagonistic activity of two families of enzymes: the histone acetyltransferases (HATs) and histone deacetylases (HDACs) (Fischer et al., 2010). The covalent modification of conserved lysine residues within histone proteins by acetyl groups leads to a nucleosomal relaxation and transcriptional activation; this reversible process provides a central mechanism to control gene expression and cellular signaling events. As such HDACs mediate epigenetic mechanisms that play a key role in homeostasis of histone functions and gene transcription. Mammalian HDACs are a family of 18 proteins divided into four groups based on structural and functional similarities: class I (HDACs: 1, 2, 3, 8), class IIa (HDACs: 4, 5, 7, 9), class IIb (HDACs: 6, 10), class III (sirtuins 1-7), and HDAC11 as the sole member of class IV (Saha and Pahan, 2006). It is well established that HDACs alter cell growth and differentiation by either governing chromatin structure or repressing the activity of specific transcription factors (Fischer et al., 2010). They are often deregulated in diseases and inhibition of their enzymatic activities remains of therapeutic interest.

Interestingly, the class IIa subgroup of HDACs shows a number of unique features in comparison to other HDACs (Verdin et al., 2003). Unlike the class I enzymes that are predominantly localized in the nucleus, class IIa HDACs shuttle between the nucleus and cytoplasm, a process that is controlled through the phosphorylation of specific serine residues (Chawla et al., 2003). The class IIa members are potent transcriptional repressors due to a high interaction propensity of N-terminal domains toward tissue specific transcription factors (Petrie et al., 2003; Saha and Pahan, 2006). Finally, the C-terminal catalytic domain of class IIa enzymes is characterized by a unique mutation of Tyr residue into His leading to the deacetylase domain inactivation (Lahm et al., 2007), emphasizing their distinctive properties compared to other HDACs (Verdin et al., 2003; Saha and Pahan, 2006).

HDAC4 shows approximately 60–70% sequence identity to HDAC5 and HDAC7 (de Ruijter et al., 2003). Among class IIa HDACs, HDAC4 has been described as a potent transcriptional repressor that is able to interact via its N-terminal domain with many different co-repressors, specifically in the brain. Hence, HDAC inhibitors have been recently used in the treatment of a wide-range of brain disorders characterized by pathological alterations in the transcriptome (Fischer et al., 2010) and they have displayed neuroprotective effects in animal models of such neurological disorders like Huntington's disease (HD), Alzheimer's disease (AD) and ischemic stroke.

#### **REGULATION OF HDAC4 EXPRESSION, CELLULAR LOCALIZATION AND FUNCTION**

HDAC4 is ubiquitously expressed throughout the body with enrichment in the brain, heart and skeletal muscle (Verdin et al., 2003; Broide et al., 2007). However, the transcriptional mechanisms involved in the regulation of HDAC4 expression are poorly understood. Recent studies have shown that HDAC4 expression might be tightly regulated by microRNAs. In podocytes, miR-29a reduced HDAC4 signaling and attenuated the glucocorticoidmediated β-catenin deacetylation and ubiquitination (Ko et al., 2013; Lin et al., 2014). Similarly, miR-365 reduced endogenous HDAC4 protein levels and led to inhibition of chondrocyte differentiation (Guan et al., 2011). Overexpression or inhibition of miR-206 in skeletal muscles has been associated with a decrease or increase of endogenous HDAC4 levels, respectively (Williams et al., 2009; Liu et al., 2012; Winbanks et al., 2013). HDAC4 expression has also been reported to be up-regulated under ER stress through its interaction with activating transcription factor 4 (ATF4). *In vitro*, HDAC4 overexpression caused ATF4 cytoplasmic retention and inhibition of ATF4 transcriptional activity, suggesting the presence of an autoregulatory loop. ER stress can ultimately promote cell apoptosis through up-regulation of ATF4 target genes such as CHOP and TRB3 and this effect was exacerbated by HDAC4 down-regulation (Zhang et al., 2014).

It is believed that HDAC4 undergoes a signal-dependent shuttling between the cytoplasm and nucleus, although in the brain it seems to be exclusively localized to the cytoplasm (Darcy et al., 2010; Mielcarek et al., 2013a,b). This nuclear-cytoplasmic shuttle is controlled by multiple mechanisms including activity of calcium/calmodulin-dependent kinase (CaMK) (Bolger and Yao, 2005) and salt-inducible kinases (Walkinshaw et al., 2013a), cAMP signaling (Walkinshaw et al., 2013b), and oxidative stress (Matsushima et al., 2013) (reviewed in Parra and Verdin, 2010, **Figure 1A**). Under normal conditions phosphorylated HDAC4 was retained in the cytoplasm through its association with 14-3-3 proteins (Grozinger and Schreiber, 2000; Verdin et al., 2003), (**Figure 1A**) and once dephosphorylated at Ser298 by the catalytic subunit of PP2A (Protein Phosphatase 2) it moved into the nucleus (Paroni et al., 2008).

It has been shown that HDAC4 can produce three specific nuclear pools including full length HDAC4 and two N-terminal fragments with different functions controlling cell death and differentiation *in vitro* (Paroni et al., 2004, 2007; Backs et al., 2011). Indeed, HDAC4 protein can be cleaved by caspases leading to a HDAC4-nuclear fragment generation (Paroni et al., 2004, 2007). Cleavage of HDAC4 occurred at Asp289 and resulted in the formation of a cytosolic carboxy-terminal fragment and an amino-terminal fragment that accumulated into the nucleus. This nuclear fragment exhibited a stronger cell death-promoting activity coupled with increased repressive effect on Runx2 (Runtrelated transcription factor 2) or SRF (Serum response factor) dependent transcription. Interestingly, this nuclear fragment was a less potent inhibitor of MEF2C (Myocyte enhance factor 2C) driven transcription, compared to full-length HDAC4 (Paroni et al., 2004), although such repressor activity has been described as independent from the acetylase domain. While caspase-2 and caspase-3 have been shown to cleave HDAC4 *in vitro,* caspase-3 was critical for HDAC4 cleavage *in vivo* during UV-induced apoptosis (Paroni et al., 2004). In the nucleus, a caspase-generated HDAC4 fragment was also reported to trigger cytochrome C release from mitochondria and cell death in a caspase-9 dependent manner (Liu and Schneider, 2013). In isolated skeletal muscle fibers expressing a HDAC4-green fluorescent protein, activation of PKA by the beta-receptor agonist isoproterenol or dibutyryl cAMP caused a steady HDAC4 nuclear influx. Thus, mutations of Ser265 and Ser266 (PKA targeted serines) enabled HDAC4 to respond to PKA activation (Liu and Schneider, 2013). Similarly, clenbuterol a potent β2-adrenoreceptor stimulator in skeletal muscles caused HDAC4 phosphorylation on Ser246 through activation of CaMKII (Ohnuki et al., 2014). In cardiomyocytes, PKA induced generation of the N-terminal HDAC4 cleavage product at Tyr202. This N-terminal fragment selectively inhibits activity of MEF2 but not SRF, thereby antagonizing a pro-hypertrophic potential of CaMKII signaling without affecting cardiomyocytes survival. Thus, HDAC4 may function as a molecular nexus for the antagonistic actions of the CaMKII and PKA pathways (Backs et al., 2011). In addition, sustained glycolysis induced by lipopolysaccharide (LPS) treatment activated caspase-3, which cleaved HDAC4 and triggered its degradation. Importantly, a caspase-3 resistant HDAC4 mutant escaped LPS-induced degradation and prolonged inflammatory cytokine production through the GSK3β (Glycogen Synthase Kinase-3 β isoform)–iNOS (inducible Nitric Oxide Synthase)–NO (Nitric Oxide) axis (Wang et al., 2014a). However, until now, there have been no data available suggesting a similar proteolytic pattern of HDAC4 in the healthy brain or in neurodegenerative disorders.

Interestingly, cleavage and phosphorylation sites are all located within the N-terminal region of HDAC4 protein highlighting this area as an important regulatory domain. While this N-terminal region seems to be critical for the repressive function of HDAC4, it also contains a transcription factor interacting domain that can bind MEF2 family members. HDAC4-MEF2 interaction was associated with the inhibition of MEF2 function resulting in neuronal cell death (Mao et al., 1999) and repression of MEF2-dependent genes in neuronal cells (Bolger and Yao, 2005) and skeletal muscles (Miska et al., 2001). In addition, the HDAC4 N-terminal region is characterized by a high glutamine content that is likely responsible for interactions with other glutamine-rich proteins leading to a spontaneous assembly of insoluble toxic amyloid-like structures (Fiumara et al., 2010). X-ray resolution of the human HDAC4 glutamine-rich domain showed that this domain is preferentially folding into a straight alpha-helix which assembles into a tetramer. In contrast to the coiled coil proteins, the HDAC4 tetramer lacked the regular arrangement of apolar residues and had an extended hydrophobic core that might lead to its rapid equilibrium with monomer and intermediate species (Guo et al., 2007). Overall, these studies provide a picture of a multifunctional protein and emphasize the presence of several mechanisms behind the tissue-specific regulation of HDAC4 expression and function.

# **REGULATION OF HDAC4 DEACETYLASE ACTIVITY**

Previous findings have suggested that class IIa HDACs are inactive on acetylated substrates, thus differing from class I and IIb enzymes (de Ruijter et al., 2003). HDAC4 catalytic domain purified from bacteria was 1000-fold less active than class I HDACs on standard substrates (Lahm et al., 2007). In contrast to the other HDACs, the C-terminal catalytic domain of the class IIa enzymes contains an amino acid substitution of a critical Tyr residue into His (Lahm et al., 2007). Mutation of this Tyr to His in class I HDACs severely reduced their activity, while a His-976- Tyr mutation in HDAC4 produced an enzyme with a 1000-fold higher catalytic efficiency (Lahm et al., 2007). Interestingly, mutations in the residues involved either in the coordination of the structural zinc binding domain of HDAC4 or the binding site of class IIa selective inhibitors prevented the association of HDAC4 with N-CoR/HDAC3 associated repressor complex (Bottomley et al., 2008). It has been proposed that HDAC4 binds directly to HDAC3 in order to activate its deacetylase domain (Mihaylova et al., 2011) and that the structural zinc-binding domain is crucial in the regulation of class IIa HDAC functions (Bottomley et al., 2008).

Finally, HDAC4 activity seems to be modulated by the ubiquitin-proteasome system. Serum starvation elicited the polyubiquitination and degradation of HDAC4 in nontransformed cells. Phosphorylation of Ser298 within the PEST1 sequence, a GSK3β consensus sequence, played an important role in the control of HDAC4 stability. Phosphorylation of HDAC4 by GSK3β has been described to occur *in vitro* upon phosphorylation of Ser302, which seems to play a role of a priming phosphate (Cernotta et al., 2011) and removal of growth factors fails to trigger HDAC4 degradation in cells deficient in this kinase (**Figure 1B**). One might conclude that HDAC4 is not a histone/protein deacetylase, however it can play a crucial role in many processes through its interaction with HDAC3 or with a general role of scaffolding protein.

# **HDAC4 BIOLOGICAL FUNCTION IN NON CNS ORGANS**

As mentioned, HDAC4 is ubiquitously expressed, however, its initial biological function was described in chondrocytes: direct inhibition of RUNX2 by HDAC4 led to chondrocyte hypertrophy (Vega et al., 2004). HDAC4-null mice displayed premature ossification of developing bones due to an ectopic and early onset chondrocyte hypertrophy, mimicking the phenotype associated with the constitutive *Runx2* expression in chondrocytes. On the other hand, overexpression of HDAC4 in proliferating chondrocytes *in vivo* inhibited their hypertrophy and differentiation, mimicking a *Runx2* loss-of-function phenotype (Vega et al., 2004).

HDAC4 has been described as a critical factor that connects neural activity to the muscle remodeling program, nevertheless, its role in the physiology of this peripheral tissue is controversial and has not been entirely clarified. Inactivation of HDAC4 suppressed denervation-induced muscle atrophy while increased re-innervation (Williams et al., 2009; Winbanks et al., 2013). Although it has been observed that miR-206 could regulate HDAC4 expression in skeletal muscles, the postnatal expression of miR-206 is not a key regulator of a basal skeletal muscles mass or specific pathways of muscle growth and wasting (Winbanks et al., 2013). Previous studies established that muscle denervation strongly induced the expression of Gadd45a,

a small myonuclear protein that is required for denervationinduced muscle fiber atrophy. Interestingly, it was shown that HDAC4 mediated an induction of Gadd45a mRNA in denervated skeletal muscles (Bongers et al., 2013). Furthermore, HDAC4 has been described to induce AP1-dependent transcription by activating the HDAC4-MAPK-AP1 signaling axis essential for the neurogenic muscles atrophy (Choi et al., 2012). Interestingly, AP1 inactivation recapitulates HDAC4 deficiency and blunts the muscle's atrophy program. Surprisingly, HDAC4 stimulated AP1 activity by activating the HDAC4-MAPK-AP1 signaling axis essential for the neurogenic skeletal muscles atrophy (Choi et al., 2012). HDAC4 was also described as a member of the MEF2C repressor complex together with HDAC3 and Ser/Thr kinase homeodomain-interacting protein kinase 2 (HIPK2) in undifferentiated myoblasts (de la Vega et al., 2013). On the other hand, a recent study has shown that HDAC4 inactivation led to defective satellite cells proliferation, muscle regeneration and lipid accumulation (Choi et al., 2014).

Finally, it was shown that HDAC4 up-regulation was significantly greater in patients with rapidly progressive ALS (Amyotrophic lateral sclerosis) and its expression was negatively correlated with a degree of skeletal muscles re-innervation and functional outcome (Bruneteau et al., 2013). Similarly an increased level of HDAC4 has been found in SMA (Spinal muscular atrophy) model mice and in SMA patient muscles (Bricceno et al., 2012). Moreover, HDAC4 expression was increased in masseter muscles from a patient with a deepbite and was found to correlate negatively with slow myosin type I and positively with fast myosin type IIX (Huh et al., 2013). Overall, these studies provide evidence of an active role of HDAC4 in the neurogenic muscle's atrophy program, which becomes exacerbated in some neurodegenerative disorders, therefore, urging HDAC4 as a promising therapeutic target.

#### **HDAC4 FUNCTION IN THE BRAIN AND ITS IMPLICATION IN NEURODEGENERATION**

Compared to the other class IIa enzymes, HDAC4 is highly expressed in the mouse brain (Grozinger et al., 1999; Darcy et al., 2010) with a highest expression occurring during early postnatal life (Sando et al., 2012). Immunohistochemical analysis of brain sections revealed accumulation of HDAC4 in the cytoplasm of neurons, including neurons containing CRH, oxytocin, vasopressin, orexin, histamine, serotonin, and noradrenaline (Takase et al., 2013). HDAC4-immunoreactive puncta were uniform in size and were widely distributed in the neuropil of brain areas, including the PVN (Hypothalamic Paraventricular Nucleus), LHA (Lateral Hypothalamic Area), ARC (Hypothalamic Arcuate Nucleus), TMN (Tuberomammillary Nucleus), DR (Dorsal Raphe), and LC (Locus Coeruleus). Interestingly, these HDAC4 positive accumulations co-localized with PSD95-immunoreactive puncta (Mielcarek et al., 2013a; Takase et al., 2013 ), suggesting a role for HDAC4 in synaptic plasticity (Sando et al., 2012; Mielcarek et al., 2013a).

In neurons, dynamic changes in the subcellular localization of HDACs are thought to contribute to various signaling pathways. Treatment with the neuronal survival factor BDNF (Brain-derived neurotrophic factor) suppressed HDAC4 nuclear translocation, whereas a pro-apoptotic CaMK inhibitor stimulated HDAC4 nuclear accumulation. Moreover, as expected, an ectopic expression of the nuclear-localized HDAC4 led to neuronal apoptosis and repressed the transcriptional activities of survival factors in neurons like: MEF2 and cAMP response element-binding protein (CREB). In contrast, inactivation of HDAC4 by small interfering RNA or HDAC inhibitors suppressed neuronal cell death (Bolger and Yao, 2005). In cultured hippocampal neurons, localization of HDAC4 has been shown to be very dynamic and signal-regulated and spontaneous electrical activity was sufficient for HDAC4 nuclear export (Chawla et al., 2003). On the other hand, in various experimental models, it has been shown that loss of HDAC4 could lead to neurodegeneration of the retina (Chen and Cepko, 2009) and cerebellum (Majdzadeh et al., 2008). This might be explained by the chondrocyte hypertrophy that occurred in mice lacking HDAC4 causing developmental brain abnormalities due to a skull deformation (Vega et al., 2004) and might be further supported by major pathological changes in the *Hdac4* knock-out murine postnatal brain (Majdzadeh et al., 2008). In addition, a conditional knockout of *Hdac4* under the *CamkII* promoter in the mouse forebrain, showed impairments in the hippocampal-dependent learning and memory with a simultaneous increase in locomotor activity (Kim et al., 2012). However, it was recently shown that a selective deletion of *Hdac4* under the control of the *Thy1* or Nestin promoters resulted in a normal gross brain morphology and cytoarchitecture as well in a normal locomotor activity (Price et al., 2013). Moreover, the Affymetrix array data showed no effect of *Hdac4* knock-out on the transcriptional profile and global acetylation of the postnatal murine brain (Mielcarek et al., 2013b). Similarly, a hippocampal depletion of HDAC4 *in vivo* abolished long-lasting stress-inducible behavioral changes and improved stress related learning and memory impairments in mice (Sailaja et al., 2012). HDAC4 overexpression has been shown to accelerate the death of cerebellar granule neurons (Bolger and Yao, 2005; Li et al., 2012; Sando et al., 2012) by increasing their vulnerability to H202 insult due to an inhibition of PPARα activity (peroxisome proliferatorsactivated receptor α) (Yang et al., 2011). In addition, the HDAC4 viral-mediated overexpression in the rat hippocampus was sufficient to induce depression like behavior (Sarkar et al., 2014). Interestingly, overexpression of HDAC4 in the adult mushroom body, an important structure for memory formation, resulted in a specific impairment in long-term courtship memory but had no affect on short-term memory in *Drosophila* model (Fitzsimons et al., 2013). Similarly, HDAC4 and HDAC5 increased a cell viability through an inhibition of HMGB1, a central mediator of tissue damage following acute injury and it has been shown that NADPH oxidase-mediated HDAC4 and HDAC5 expression contributed to the cerebral ischemia injury through the HMGB1 signaling pathway that could be an effective therapeutic target to treat stroke (He et al., 2013).

# **HDAC INHIBITORS AND INHIBITION OF HDAC4 ACTIVITY**

Suberoylanilide hydroxamic acid, known also as SAHA or vorinostat, was the first HDAC inhibitor (HDACI) to be approved for the cancer therapy of advanced cutaneous T-cell lymphoma (Marks and Breslow, 2007). Initially, SAHA was identified as an inhibitor of class I and class II HDACs at nanomolar concentrations (Richon et al., 1998), but was further characterized as inhibitor of class I HDACs as well specifically HDAC6 within the class IIb enzyme (Parmigiani et al., 2008; Marks and Xu, 2009). More recently activity based probes have been used to demonstrate that SAHA can bind directly to both class I and IIa HDACs (Salisbury and Cravatt, 2007; Codd et al., 2009). Experiments performed on cancer cell lines revealed the ability of SAHA to induce the degradation via RANBP2-mediated proteasome of both HDAC4 and HDAC5 *in vitro* (Scognamiglio et al., 2008). Interestingly, other HDACIs, such as trichostatin A (TSA) and sodium butyrate, have also been reported to induce a reduction in HDAC4 levels when administered to embryoid bodies (Chen et al., 2011), suggesting that similarly to SAHA, these other HDACIs could induce HDAC4 degradation through a proteasome-dependent mechanism.

An increased expression of HDAC4 has been described in several *in* vitro and *in* vivo models of neuro-like disorders. Treatment of neuronal cell lines with SAHA led to a noticeable improvement of cell polarity and morphology, with longer processes in the rat H19-7 hippocampal cell line with folate deficiency. In this neuronal cell model, folate deficit led to a reduction in cell proliferation and decreased production of S-adenosylmethionine (the universal substrate for transmethylation reactions) concurrent with an increased expression of HDACs (including HDAC4,6,7) (Akchiche et al., 2012). Furthermore, cecal ligation and peroration (CLP) rats, used in the Sepsis-associated encephalopathy (SAE) study, were also characterized by an increased expression of HDAC4 (Fang et al., 2014). Administration of HDACIs (e.g., TSA or SAHA) restored Bcl-XL and Bax levels *in vivo* and decreased apoptotic cells *in vitro*. In addition, knock-down of HDAC4 by shRNA resulted in an enhanced histone acetylation like: H3 and H4 and reduced neuronal apoptosis. Consistently, CLP rats treated with TSA or SAHA exhibited significant spatial learning and memory deficits with no effect on their locomotive activity (Fang et al., 2014).

In preclinical settings, SAHA and other HDACIs have consistently improved the phenotype in HD mouse models (Ferrante et al., 2003; Hockly et al., 2003; Gardian et al., 2005) and are being developed as HD therapeutics. Recent findings have increasingly described a widespread peripheral organ pathology in HD, such as skeletal muscles atrophy (Zielonka et al., 2014a) and heart failure (Mielcarek et al., 2014a; Zielonka et al., 2014b), often associated with an increased HDAC4 expression (Mielcarek et al., 2014b). As such, class IIa HDACs inhibitors might be beneficial in delaying HD-related symptoms and, therefore, are under evaluation as HD therapeutics. It has been shown that the administration of SAHA to wild type and R6/2 mice decreased HDAC2 and HDAC4 at the protein but not RNA levels in different brain regions *in vivo* (Mielcarek et al., 2011), supporting previous observations from cancer cell lines (Scognamiglio et al., 2008). We have also shown that HDAC4 associates with huntingtin in a polyglutamine-length dependent manner and co-localizes with cytoplasmic inclusions. Consequently, reduction of HDAC4 levels delayed cytoplasmic aggregate formation in different brain regions of R6/2 mice and rescued cortico-striatal neuronal synaptic function in HD mouse models. This was accompanied by an improvement in motor co-ordination, neurological phenotypes and increased lifespan (Mielcarek et al., 2013a). Interestingly, SAHA treatment of R6/2 mice was accompanied by restoration of brain-derived neurotrophic factor (BDNF) cortical transcript levels (Mielcarek et al., 2011). An increased expression of BDNF has been associated with memory-enhancing and neuroprotective properties of HDACIs, as it has been shown that HDAC4 and HDAC5 might repress specific *Bdnf* transcripts in rats and primary hippocampal neuronal cultures and this effect was reversed by SAHA treatment (Koppel and Timmusk, 2013). However, the mechanism of BDNF induction by HDACIs is not yet fully understood. Surprisingly, HDAC4 reduction had no effect on global transcriptional dysfunction and did not modulate nuclear huntingtin aggregation in HD mousse models (Mielcarek et al., 2013a). Interestingly, elevated HDAC4 levels have been shown in *post mortem* HD (Yeh et al., 2013) and FTLD (Frontotemporal Lobar Degeneration) (Whitehouse et al., 2014) brains and HDAC4 has been described as a component of Lewy Bodies in Parkinson's disease brains (Takahashi-Fujigasaki and Fujigasaki, 2006) and of intranuclear inclusions in the neuronal intranuclear inclusion disease (Takahashi-Fujigasaki et al., 2006). Consistently, administration of SAHA has been shown to improve synaptic plasticity and learning behavior in an Alzheimer disease model (Kilgore et al., 2010). A causative role for HDAC4 has been also described in SCA-1 (Spinocerebellar Ataxia Type 1) as a modulator of Ataxin-1. It was shown that ataxin-1 bound specifically to HDAC4 and MEF2 and co-localized with them in the nuclear inclusion bodies. Significantly, these interactions were greatly reduced by the S776A mutation, which largely abrogates the cytotoxicity of ataxin-1 *in vitro* (Bolger et al., 2007). Moreover, HDAC4 has been found to be significantly overexpressed in specific cortical regions of autistic patients (Nardone et al., 2014).

Diabetes is one of major risk factors for dementia. However, the molecular mechanism underlying the risk of diabetes for dementia is largely unknown. Surprisingly, it has been shown that diabetes may cause epigenetic changes in the brain, which adversely affect synaptic function. These alterations were associated with an increased susceptibility to oligomeric Aβ-induced synaptic impairments in the hippocampal structure that eventually led to synaptic dysfunction. Use of pharmacological inhibitors against the HDAC IIa family restored synaptic function (Wang et al., 2014b). This therapeutic effect highlights the importance of HDACIIa members, including HDAC4, as a possible target in the brain.

# **CONCLUSION**

There is convincing evidence suggesting that HDAC4 plays a central role in the brain physiology and that it is deregulated in several neurodegenerative disorders, therefore representing a suitable therapeutic target, through which HDACs inhibition may occur. However, the use of currently available HDACIs is likely to involve adverse side effects due to the broad spectrum HDAC inhibition. Therefore, a search for selective HDAC inhibitors would likely be of benefit for targeted therapy. Likely, HDACs inhibition occurs through a dual mechanism, either by a direct inhibition of an active deacetylase domain (class I, IIb, III, and IV) or by a direct binding followed by proteosomal degradation (class IIa) (**Figure 2**). Crucially, relatively little is known regarding individual HDACs functions in the adult brain. Although, class I HDACs biological functions have been intensively studied in the brain, it appears that suitable HDAC targets could arise from HDAC IIa subfamily but their function and role are still poorly understood. Recent studies identified HDAC4 as a critical component of several neurological processes including neuronal survival and synaptic plasticity in healthy and diseased brains. However, little is known about HDAC4 cellular process like: mechanisms governing HDAC4 cellular localization, posttranslational modification and a proteolytic cleavage, especially in the diseased brains. In addition, HDAC4 transcriptional regulation has not been studied and therefore the description of the specific transcription factors and regulatory elements driving HDAC4 expression should be carefully undertaken. The presence of an inactive deacetylase domain within the class IIa HDACs might also suggest that newly designed small molecules should be rather directed to the HDAC4 known functions, including transcription binding domain, HDAC3 interaction or proteolytic cleavage sites than toward deacetylase domain. Therefore, much more research is needed to fully describe biological function of HDAC4 in the healthy and diseased brains to be able to shape future therapeutic strategies for a various disorders.

# **REFERENCES**


nuclear pools of histone deacetylase 4. *Mol. Cell. Biol.* 27, 6718–6732. doi: 10.1128/MCB.00853-07


interaction with ATF4. *Cell. Signal.* 26, 556–563. doi: 10.1016/j.cellsig.2013. 11.026


**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 October 2014; paper pending published: 19 November 2014; accepted: 28 January 2015; published online: 24 February 2015.*

*Citation: Mielcarek M, Zielonka D, Carnemolla A, Marcinkowski JT and Guidez F (2015) HDAC4 as a potential therapeutic target in neurodegenerative diseases: a summary of recent achievements. Front. Cell. Neurosci. 9:42. doi: 10.3389/fncel. 2015.00042*

*This article was submitted to the journal Frontiers in Cellular Neuroscience.*

*Copyright © 2015 Mielcarek, Zielonka, Carnemolla, Marcinkowski and Guidez. 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.*

# Hydrogels as scaffolds and delivery systems to enhance axonal regeneration after injuries

# **Oscar A. Carballo-Molina and Iván Velasco\***

Instituto de Fisiología Celular-Neurociencias, Universidad Nacional Autónoma de México, Mexico, D.F., Mexico

#### **Edited by:**

Victoria Campos, Instituto Nacional de Neurologia y Neurocirugia, Mexico

#### **Reviewed by:**

Ertugrul Kilic, Istanbul Medipol University, Turkey Carlo Di Cristo, University of Sannio, Italy

#### **\*Correspondence:**

Iván Velasco, Instituto de Fisiología Celular-Neurociencias, Universidad Nacional Autónoma de México, Circuito Exterior S/N, Ciudad Universitaria, México, D.F.-04510, México e-mail: ivelasco@ifc.unam.mx

Damage caused to neural tissue by disease or injury frequently produces a discontinuity in the nervous system (NS). Such damage generates diverse alterations that are commonly permanent, due to the limited regeneration capacity of the adult NS, particularly the Central Nervous System (CNS). The cellular reaction to noxious stimulus leads to several events such as the formation of glial and fibrous scars, which inhibit axonal regeneration in both the CNS and the Peripheral Nervous System (PNS). Although in the PNS there is some degree of nerve regeneration, it is common that the growing axons reinnervate incorrect areas, causing mismatches. Providing a permissive substrate for axonal regeneration in combination with delivery systems for the release of molecules, which enhances axonal growth, could increase regeneration and the recovery of functions in the CNS or the PNS. Currently, there are no effective vehicles to supply growth factors or cells to the damaged/diseased NS. Hydrogels are polymers that are biodegradable, biocompatible and have the capacity to deliver a large range of molecules in situ. The inclusion of cultured neural cells into hydrogels forming three-dimensional structures allows the formation of synapses and neuronal survival. There is also evidence showing that hydrogels constitute an amenable substrate for axonal growth of endogenous or grafted cells, overcoming the presence of axonal regeneration inhibitory molecules, in both the CNS and PNS. Recent experiments suggest that hydrogels can carry and deliver several proteins relevant for improving neuronal survival and axonal growth. Although the use of hydrogels is appealing, its effectiveness is still a matter of discussion, and more results are needed to achieve consistent recovery using different parameters. This review also discusses areas of opportunity where hydrogels can be applied, in order to promote axonal regeneration of the NS.

**Keywords: axotomy, growth factors, injury response, grafting, surgical intervention**

### **INTRODUCTION**

The nervous system (NS) is responsible for the interaction between organisms and their environment; it confers the ability to respond to external stimuli. However, when an injury occurs in this system, such ability is impaired. Understanding fundamental mechanisms involved in the response to damage might be used to design therapeutic interventions aimed to promote functional recovery. Axonal regeneration capacity is very limited in the Central Nervous System (CNS; Gurgo et al., 2002; Case and Tessier-Lavigne, 2005). Although the Peripheral Nervous System (PNS) is able to grow axons after a nerve injury, the lost function is not always restored, because the regenerated axons are unable to reinnervate areas previously connected by them (Johnson et al., 2005). This review describes first the elements that impede axonal regeneration following injury in CNS and PNS, and later discuss how hydrogels might attenuate the inhibitory elements for axonal regeneration in both systems.

#### **NERVOUS SYSTEM RESPONSE TO INJURY AND ITS ROLE IN AXONAL GROWTH INHIBITION**

An injury in the NS could imply a loss of tissue, interrupted communication caused by damage of synaptic contacts or disrupted information flow between cell soma and axons. Different events occur after NS damage, depending on several factors, such as the type of injury. Many of these events are responsible for the inhibitory environment during axonal regeneration. We next describe the differential responses to lesion of CNS and PNS.

#### **CNS**

#### **Glial and fibrous scar formation**

The disruption of the blood brain barrier (BBB) after damage allows the infiltration of blood proteins to the CNS, which triggers an inflammatory reaction (Kawano et al., 2012). White cells and macrophages enter through the lesion site and migrate to the surrounding neural tissue, releasing various cytokines and chemokines (Merrill and Benveniste, 1996; Donnelly and Popovich, 2008). These events lead to the activation of astrocytes, microglia and oligodendrocyte progenitor cells, to form the glial scar around the lesion site (Shearer and Fawcett, 2001; Kawano et al., 2012; **Figure 1A**). These activated cells release different molecules involved in inflammation, BBB restoration and neuroprotection (Yiu and He, 2006; Rolls et al., 2009). Glial scar isolates the damage area from adjacent tissue (**Figure 1A**); this contributes to maintaining homeostatic functions as ion and fluids balances, production of pro- and anti-inflammatory molecules, secretion of growth factors and free radicals elimination (Yiu and He, 2006; Rolls et al., 2009). In addition, a fibrotic scar (**Figure 1A**), which is produced by the intrusion of fibroblasts from the damaged meninges and that release extracellular matrix (ECM) proteins such as type IV Collagen, Fibronectin and Laminin, also forms around the site of lesion (Kawano et al., 2012). Moreover, fibroblasts and astrocytes cooperate to establish a continuous basal lamina around the glial scar (Mathewson and Berry, 1985; Shearer and Fawcett, 2001). The barrier formed by the glial and fibrous scars helps to contain the damage, preventing it from spreading and affecting surrounding tissue. The functions of these barriers are not fully understood yet, but their inhibitory effect on axonal growth has been extensively documented (Fawcett and Asher, 1999; Sandvig et al., 2004; Silver and Miller, 2004; Yiu and He, 2006; Fitch and Silver, 2008).

#### **Inhibitory elements for axonal growth**

Various elements have been described as being responsible of the adverse environment for axonal growth. The glial scar constitutes a physical barrier that impedes passage of axons across the lesion site (**Figure 1A**). In addition, the activated glial cells in the scar secrete ECM components, especially chondroitin sulphate proteoglycans (CSPGs) such as Neurocan, Brevican, Versican and NG2 (Fawcett and Asher, 1999; Tang et al., 2003), which exert an inhibitory influence on axonal growth (Shearer and Fawcett, 2001; Tang et al., 2003; Silver and Miller, 2004). Activation of Rho small GTPase proteins which are recognized by CSPGs blocks actin polymerization in growing neurites (Sandvig et al., 2004; Díaz-Martínez and Velasco, 2009). Some of the inhibitory molecules over-expressed in the site of lesion are Myelinassociated glycoprotein, Oligodendrocyte-myelin glycoprotein, Nogo protein and its receptor Nogo-66, Semaphorin 4D, Ephrin B3 (Kawano et al., 2012), Semaphorin 3D (Pasterkamp et al., 1999) and Ephrin B2 (Bundesen et al., 2003). The lesion site has been shown to secrete chemo-repulsive molecules like Tenascin (McKeon et al., 1991) and Semaphorin 3A (Pasterkamp et al., 1999). The fibrotic scar also produces inhibitory elements: it has been reported that fibroblasts express NG2 proteoglycan (Tang et al., 2003), Phosphocan (Tang et al., 2003), Tenascin-C (Tang et al., 2003), Semaphorin 3A (Pasterkamp et al., 1999) and Ephrin B2 (Bundesen et al., 2003). These data indicate that glial and fibrous scars contribute to the low rate of axonal regeneration. One option to bypass these inhibitory effects would be to prevent scar formation, but this could imply secondary effects, such as spreading of the damage. Alternatively, a modification of the lesion environment by introducing a device that is a permissive for axonal growth is feasible.

# **PNS**

The damage produce diverse signals that indicate the neuronal cell to either go into regeneration processes or to undergo programmed cell death (Maripuu et al., 2012). Ca2<sup>+</sup> entry (Maripuu et al., 2012) and the interruption of the retrograde transport system are the initial signals of neuronal damage (Dahlin, 2008). Damage to axons also release growth factors such as Ciliary Neurotrophic Factor, Leukemia Inhibitory Factor and Interleukin-6 in the site of injury (Hanz and Fainzilber, 2006; Raivich and Makwana, 2007). After a peripheral axon is cut or crushed, the distal part of the severed axon suffers a degenerative processes (**Figure 2A**) termed Wallerian degeneration (Mietto et al., 2008; Freeman, 2014), that consists in the degradation of axonal organelles and proteins, and disintegration of axonal structures (Mietto et al., 2008). In order to obtain successful axon regeneration, activation and proliferation of Schwann cells (SC) are needed (Dahlin, 2008). SC and macrophages phagocyte the disrupted myelin sheath and cell debris (**Figure 2A**) to clear the zone (Johnson et al., 2005). In addition, SC have been shown to provide a favorable substrate for regenerating axons (Maripuu et al., 2012). As a part of the regenerative processes, several axon growth cones emerge from the proximal stump (**Figure 2A**) and their growth is guided by Bugner bands, which are SC apposed around the basement membrane (Valls-Sole et al., 2011). Basement membrane is constituted by ECM proteins such as Laminin and contributes to the adhesion and guidance of cells during development (Silver and Miller, 2004). These emerging sprouts grow to reach the distal stump across the site of lesion. However, because the adverse environment, most of these developments disperse in various directions and become abortive (Johnson et al., 2005; Valls-Sole et al., 2011; **Figure 2A**). The few sprouts that successfully reach the distal stump grow in close apposition with SC (Johnson et al., 2005). To complete this processes, the regenerated axon is myelinated by SC to produce junctional nodes of Ranvier and internodal Schmidt-Lanterman incisures (Johnson et al., 2005). After the regeneration process, SC release Nerve Growth Factor (NGF) and Glial cell-Derived Neurotrophic Factor (GDNF) in the regenerated site (Maripuu et al., 2012).

#### **Elements involved in the axonal regeneration process in the clinic**

Although the PNS has some regeneration capacity, functional restoration is not always achieved (Hill et al., 2011; Valls-Sole et al., 2011). Factors involved in the recovery are age (Lundborg and Rosén, 2001; Ruijs et al., 2005), the period of time between injury and medical assistance (Ruijs et al., 2005), the type of nerve that is damaged and the magnitude of injury (Lundborg, 2004); for example, when the lesion consists in a moderate compression, the chances of regeneration are higher (Lundborg, 2004) than when the axon has been transected (Maripuu et al., 2012). Current treatments trying to reconnect the severed nerve offer a low success rate. One major problem in these surgical procedures is that axons do not innervate the correct area, since many axons

compete to connect and this lead to loss of selectivity (Valls-Sole et al., 2011). These mismatches could convey disturbances in signaling between CNS neurons and PNS neurons, which result in sensorimotor alterations (Valls-Sole et al., 2011). Although

PNS has more permissive environment in contrast to CNS, the majority of regeneration processes do not reach to innervate the pre-lesional area (**Figure 2A**), which leads to several adverse consequences.

#### **SCAFFOLDS THAT PROMOTE AND GUIDE AXONAL GROWTH BIOMATERIALS**

Biomaterials possess properties that make them biocompatible, meaning that they do not produce cytotoxic or immunologic reactions; their components should be susceptible to modifications, and the process of its production ideally should be easy and reliable (Holmes, 2002). It is important to mention that the specific requirements will dictate the origin of biomaterials. Natural biomaterials such as collagen matrix mimic largely the extracellular environment. However, natural materials present disadvantages such as lot variability (Zhang et al., 2005b), the generation of immune reactions and they may contain pathogenic agents (Holmes, 2002). On the other hand, synthetic materials can have a more consistent quality control but are not always compatible with host tissue (Metcalfe and Ferguson, 2007). Other advantages of synthetic materials are that they are free of pathogenic agents (Holmes, 2002) and could be readily modified in order to elicit a tissue response, for example, it is possible to add the isoleucine-lysine-valine-alanine-valine (IKVAV) motif which enhances cellular adhesion (Cheng et al., 2013).

#### **SCAFFOLDS**

The purpose of implanting a biomaterial in a tissue is to provide suitable physical support to cells in order to achieve regeneration. This substrate should mimic as much as possible the natural matrix, so it should constitute a three-dimensional (3D) structure with porous size that allows the exchange of nutrients and oxygen (Hollister, 2005; Zhang et al., 2005b). Furthermore, it needs to provide a substrate for cellular adhesion, and in selected cases, could promote guided growth, proliferation, differentiation or apoptosis by scaffold-cell or by cell-cell interactions (Hollister, 2005; Zhang et al., 2005a). Some scaffolds such as hydrogels, as we will see later, have the ability to deliver several components. Therefore, the scaffold is a dynamic element that might play an important role in the regenerative process.

#### **EARLY ATTEMPTS TO PROMOTE AXONAL REGENERATIONS USING SCAFFOLDS**

The first studies, aimed to promote axonal growth after injury, were performed in the PNS using rigid materials. These attempts mainly used tubular structures to protect and guide the growth of regenerating axons. However, the used materials were not biocompatible, such as silicone conduits (Cheng and Lin, 2004), mini guide channels of a polyacrylonitrile:polyvinylchloride copolymer (Bamber et al., 1999), polytetrafluoroethylene and collagen conduits (Vasconcelos and Gay-Escoda, 2000), guidance channels of polyvinylidene fluoride (Aebischer et al., 1987), cylinders made of poly (D,L-lactic-co-glycolic acid) (Gautier et al., 1998) and poly (D,L-lactic acid) macroporous guidance scaffolds (Patist et al., 2004). Although in some cases a partial axonal regeneration was achieved, most of them did not induce successful regeneration, even after addition of trophic factors. These materials did not possess the required properties to support cell attachment, axonal growth, and some of them even induced an immune reaction. Although many biomaterials, natural or synthetic, have been proved to posses some of these properties, currently there is not a consistent strategy to induce a complete axonal regeneration either in the CNS or in the PNS.

#### **SYNTHESIS AND DEGRADATION OF HYDROGELS POLYMERIZATION**

Hydrogels are hydrophilic polymers that can incorporate up to 90% of its dry weight of water in its structure (Aurand et al., 2012; Hoffman, 2012). Water incorporation occurs during the gelling process, in which a liquid polymer solution turns into a gel structure, by polymerization of monomers (Aurand et al., 2012). Gelling involves the formation of cross-links in response to different stimuli (Sawheny et al., 1993; Aurand et al., 2012). The density of a gel can be modified, and the change in stiffness or porosity impact on the interaction of gels with cells (Lee and Mooney, 2001; Aurand et al., 2012; Kirschner and Anseth, 2013).

Several stimuli trigger the polymerization of hydrogels: temperature (He et al., 2000; Jeong et al., 2000; Tate et al., 2001), pH (Srividya et al., 2001; Cheng et al., 2009; Chiu et al., 2009), UV light exposure (Sawheny et al., 1993; Mellott et al., 2001; Bryant and Anseth, 2002; Chatterjee et al., 2010), or ionic concentration (Ellis-Behnke et al., 2006; Nagai et al., 2006; Koutsopoulos and Zhang, 2012, 2013). The formation of gels *in situ* in living tissue limits the use of UV light, extreme pH or non-physiological temperatures. Therefore, many hydrogels have been devised to initiate the gelling process when in contact with physiological temperature (Jeong et al., 2000; Tate et al., 2001), ionic concentration (Ellis-Behnke et al., 2006; Nagai et al., 2006; Koutsopoulos and Zhang, 2012, 2013) or pH (Srividya et al., 2001; Chiu et al., 2009). The *in situ* gelling process of hydrogels is unique, because the resulting polymer can take the form of the receiving tissue. This is particularly important for some lesions of the NS, in which an irregular cavity is formed and this discontinuity impedes axonal regeneration as described previously. Hydrogels can fill completely the space, whereas preformed structures are not suitable for this application (Macaya and Spector, 2012).

#### **DEGRADATION**

Degradation of hydrogels occurs by breaking of covalent bonds (Aurand et al., 2012). Several factors influence the rate of degradation. Water access is one of them: it has been shown that when hydrogels are exposed to *in vitro* conditions, hydrolysis is the main reason of bond disruption due to high availability of water; in contrast, when hydrogels are in *in vivo* conditions, enzymatic activity, in particular metalloproteases, is the principal cause of degradation (Lutolf et al., 2003; Patterson and Hubbell, 2010). Properties of hydrogel also influence this process: in highbond density hydrogels the disruption begins from the surface, in contrast to low-bond density hydrogel, where it begins from the interior of the structure, due to the ability of water or enzymes to penetrate the hydrogel (von Burkersroda et al., 2002).

#### **HYDROGELS AS DELIVERY SYSTEMS**

Hydrogels have properties that could make them a good alternative as a drug release system. During the gelling process it is possible to incorporate different types of molecules or cells into the gel structure (Nagai et al., 2006; Kobsa and Saltzman, 2008; Censi et al., 2012; Koutsopoulos and Zhang, 2012). The incorporation of molecules into hydrogel is facilitated by the high quantity of water that permits the uptake and diffusion of soluble molecules (Nagai et al., 2006; Censi et al., 2012; Koutsopoulos and Zhang, 2012). The incorporation and release process is dictated by the characteristics of the hydrogel such as the size of porous and the molecular properties such as the monomer's molecular weight and its electrical charge (Nagai et al., 2006; Censi et al., 2012). In the case where the porous size is bigger than the molecule, the release occurs by diffusion (Amsden, 1998). On the other hand, when the molecule is larger than the porous, degradation, swelling and erosion of hydrogel permit the delivery of the molecule (Censi et al., 2012).

Administration of drugs is a necessity in the treatment of many injuries and diseases; however, commonly these drugs inside the body are metabolized and therefore have a limited time window to exert their actions. A local and controlled delivery of drugs could improve the treatments of many diseases or injuries, especially those that occur in the CNS. The delivery of drugs in the CNS implies a great challenge because the presence of BBB and the blood–spinal cord barrier, that impede the passage of many substances to the CNS (Pakulska et al., 2012). Some current delivery drugs methods into the CNS are bolus injection and catheter/minipump systems (Pakulska et al., 2012). Bolus injection into the intrathecal space is affected by the constant flow of cerebrospinal liquid, which disperse the drug, reducing its local effect (Pakulska et al., 2012). On the other hand, the use of a catheter/minipump system has high infection probabilities, due to the external minipump location. Furthermore, it is frequent that catheters suffer dislodgement, kinking, tearing and disconnection (Penn et al., 1995).

Because of the *in situ* gelling process, it is possible to use hydrogels as a local delivery system (Censi et al., 2012; Koutsopoulos and Zhang, 2012; Macaya and Spector, 2012; Pakulska et al., 2012), although it will be necessary to find out the best alternative to introduce the hydrogel into the brain or another site of the NS. It is feasible to put the hydrogel in a damaged area and release molecules there, allowing the local delivery of a drug, which could enhance the effectiveness of treatment. In addition, the use of biodegradable hydrogels is especially relevant for long-term treatments, since it will prevent repetitive invasive interventions. The data demonstrate that hydrogels are a very versatile release system because is possible to manipulate the rate of delivery and the rate of degradation. Hydrogels could be modified to release some medicament depending on the specific circumstances, like the half-life of the medicament, the dosage or the time that is required for treatment.

# **HYDROGELS PROMOTE AXONAL REGENERATION**

Hydrogels present characteristics that make them good candidates to fulfill the needs required to promote axonal regeneration after lesions of the CNS, such as filling up the cavity of a lesion with a suitable substrate (**Figure 1B**). Although the PNS environment is less restrictive for axonal growth, as described previously, there are many challenges to achieve a successful regeneration. It is essential to reduce the probability of mismatches by providing guidance cues for correct reinnervation, and hydrogels can help to attain this task (**Figure 2B**). However, a note of caution is appropriate because, in addition to act as scaffold and delivery tools, hydrogels might represent a physical barrier for both cellular and axonal reorganization. In this section we review the published evidence showing that hydrogels promote axonal regeneration both *in vitro* and *in vivo*.

#### **IN VITRO STUDIES**

Hydrogels properties such as its high water content, their porous constitution and the three dimensional (3D) networks formed during gelling, mimic to some extent the ECM found in tissues (Geckil et al., 2011; Aurand et al., 2012; Kirschner and Anseth, 2013), making possible to culture cells in 3D structures *in vitro*. These structures are closer to the *in vivo* environment than the classic two-dimensional cultures (Zhang et al., 2005b).

Early work (Holmes et al., 2000) aimed to obtain selfassembling peptides with motifs similar to the arginine-glycineaspartate (RGD) present in several ECM proteins. The authors substituted glycine with alanine (A) and repeated the RADA sequence several times. To assess the suitability of this selfassembling peptide hydrogel, a direct comparison with Matrigel (a commercial substrate containing ECM derived from carcinoma cells) was made after culturing rat hippocampal neurons. No differences in synaptic activity measured by the endocytosis marker FM1-43 were found, showing that this hydrogel can support neuronal maturation. Recently, RADA-containing peptides were used to form 3D structures to allocate neural stem progenitor cells (NSPC) in order to evaluate proliferation and neuronal differentiation (Koutsopoulos and Zhang, 2013). Both Matrigel and the self-assembled peptide-based hydrogels sustained these parameters. Matrigel was efficient during the first 2 weeks, but the hydrogel allowed neuronal survival for over 5 months. These data demonstrate that hydrogels support neuronal differentiation and long-term survival with signs of maturity.

In another study, using dopaminergic cells, Semaphorin 3A was coupled to a PEG hydrogel containing silica particles to assess the effects on axonal growth. Semaphorin 3A is a soluble protein implicated in the axonal growth of dopaminergic neurons during brain development (Hernández-Montiel et al., 2008). Application of recombinant Semaphorin 3A to these dopaminergic neurons obtained from the developing midbrain or from *in vitro* differentiated mouse embryonic stem cells caused increased growth of axons in a collagen gel system in culture. This effect was neutralized by anti-Neuropilin receptors (Tamariz et al., 2010). A significant increase in axonal length was observed with the PEG hydrogel containing either 2 or 5 µg/ml Semaphorin 3A compared to controls (Tamariz et al., 2011). Recently, a PEG hydrogel device was designed to evaluate the influence of distance in the application of a potential axonal growthpromoting molecule on murine embryonic stem cell-derived neurons. These authors conjugated Insulin-like Growth Factor 1 to poly-lactic-co-glycolic acid particles with several admixtures that resulted in different release kinetics: early, intermediate and late. The optimal conditions were: (i) a distance up to 2 mm between the poly-lactic-co-glycolic acid particles and neurons; (ii) a sequential array of early, intermediate and late release conjugates; (iii) the early release particles placed closer to the cells and those with late kinetics placed away (Lee et al., 2014).

An intermediate step between two-dimensional *in vitro* cultures and *in vivo* studies is the culture of organotypic slices, because they maintain the ECM and the 3D organization. In cultured spinal cord slices placed on different substrates such as membrane inserts, Collagen gel, soluble hyaluronic acid and hyaluronic acid-based hydrogel, different cell-type specific markers were analyzed. The hydrogel group preserved better than the other groups the characteristics of the slice: more neurons (NeuN+), a greater proportion of choline acetyltransferasepositive neurons, well-preserved astroglia and less number of activated microglial cells were reported (Schizas et al., 2014). The data of these works confirmed that hydrogels could be used as a delivery system to promote axonal growth, and preserve better the organotypic cultures. These characteristics might be useful to promote a successful axonal regeneration *in vivo*.

#### **IN VIVO STUDIES**

#### **Hydrogels as a strategy for promoting regeneration in the brain**

Several groups have assessed the biocompatibility of hydrogels in the absence of a lesion. PEG hydrogels with 13% and 20% macromer weight, and 20% PEG conjugated with GDNF were implanted into the cortex and striatum of nonhuman primates. Four months after implantation the astroglial and microglial reactions were present around the implant site of all groups, including sham. The 13% PEG hydrogel generated fewer reactions, probably due to its faster degradation (Bjugstad et al., 2008). This same group evaluated the biocompatibility of different weight percent of PEG hydrogel implanted as strands across the rat brain. The analysis in striatum revealed that both 13% and 20% hydrogels attenuate the acute response of reactive glia, compared to the sham group after 56 days (Bjugstad et al., 2010). However, when PEG was conjugated with silica particles and implanted into striatum, a higher amount of macrophages and glial cells were founded around the injection site after 30 days, compared to controls. The authors correlated this enhanced glial reaction with the presence of silica particles that were not degraded (Tamariz et al., 2011).

In some cases, one important limitation for axonal regeneration is the presence of a cavity in the damaged tissue (**Figure 1A**). Hydrogels have been proved to be able to fill such cavity and promote axonal growth (**Figure 1B**). After resection of a fraction of the cerebral cortex, the resulting cavity was filled with a hydrogel based on self-assembling peptides or with saline solution. After 6 weeks the hydrogel significantly reduced the lesion volume, and cellular ingrowth was detected. A significant decrease in astrocytic cells and macrophages was observed in the first 2 weeks, compared to the saline group (Guo et al., 2009). Hou et al. also caused a cortical damage in rats, but they used hyaluronic acid-based hydrogel, either alone or modified to incorporate Laminin in its structure, to fill the cavity. After six and twelve weeks hydrogels allow cell infiltration, angiogenesis and inhibition of the glial scar; however, only the hydrogel with Laminin was permissive for neurite growth (Hou et al., 2005).

One of the first attempts to evaluate the implantation of hydrogels together with living cells in the brain was made by Woerly et al. (1996). In this study SC, neonatal astrocytes or cells dissociated from embryonic cerebral hemispheres were entrapped in (N-(2-hydroxypropyl)-methacrylamide)-based hydrogel. Hydrogels containing SC were implanted into the rat neocortex, and promoted cellular and axonal ingrowth within the polymer. In another study, NSPC were encapsulated into self-assembling peptide hydrogels modified to include the IKVAV motif derived from Laminin (Cheng et al., 2013). Such cell-containing hydrogels were used to fill the cavity caused by a mechanical lesion in the neocortex of rats. Hydrogels made with the IKVAV motif and NSPC promoted better tissue regeneration and presented neurogenesis, compared to hydrogels without the motif, which promoted modest tissue regeneration and had prevalent glial differentiation. This study is in agreement with previous data (Hou et al., 2005) that demonstrate that the Laminin or Laminin-derived motif incorporated to hydrogel structure allow the recovery of tissue continuity. Another group tried to promote tissue recovery with a different strategy, which consisted in incorporating GDNF to gelatin-based hydrogels, with the objective to attract to the site of lesion the endogenous NSPC present in the adult subventricular zone. The hydrogel loaded with GDNF attenuated the astroglial reaction, promoted neurite growth into the site of lesion and induced the migration of neuroblasts towards the lesion site. However, cells did not reach the site of lesion, and the migration effect was observed only at 7 days post-lesion, disappearing after 21 days (Fon et al., 2014).

Probably the most important aspect in the CNS is to achieve re-connection of damaged areas, which in the long run might positively impact behavior. In a model where hamsters' optic tracts were severed and the resulting cavity was filled with a selfassembling peptide hydrogel or with saline solution, researchers observed that the hydrogel helped to reconnect the areas around the lesion after 6 weeks, in contrast to saline-treated animals. More importantly, vision was improved in the hydrogel-treated group (Ellis-Behnke et al., 2006). Although the degree of axonal re-growth varies depending on the strategy used, hydrogels have been demonstrated to be able to fill the lesion cavity with a suitable substrate for axonal growth, and the further addition of trophic factors or cells increases the possibilities of improvement after traumatic lesions in the brain (**Figure 1B**). However, in general terms, the evidence is still insufficient to say that hydrogels would substitute current treatments for brain lesions.

#### **Hydrogels promote regeneration in the spinal cord**

The studies showing that hydrogels promote axonal growth *in vitro* prompted investigators to test if implantation of hydrogels into the damaged spinal cord could promote recovery. A hyaluronic acid hydrogel was evaluated *in vitro* and *in vivo* to investigate if it was able to promote neurite growth. Although this hydrogel promoted neurite growth *in vitro*, it was not sufficient to achieve functional recovery when implanted in rats with complete transected thoracic spinal cord (Horn et al., 2007). Transection of the spinal cord in cats followed by filling of the cavity with NeuroGel hydrogel (N-(2-hydroxypropyl) methacrylamide) permitted the formation of neural tissue with myelinated axons across the damaged area, connecting both sides of the cavity, and allowing infiltration of glial cells and capillary vessels (Woerly et al., 2001). In additional work, the authors found that the hydrogel prevented scar formation and that the gliosis reaction was reduced in the interface between tissue and hydrogel (Woerly et al., 2004).

One possibility in the design of a suitable scaffold is the combination of two different types of hydrogels, to obtain a better substrate. A combined poly lactic acid (PLA) and poly (2-hydroxyethyl methacrylate, PHEMA) hydrogel was devised to obtain a degradable porous structure. This mixed hydrogel was implanted in the hemisected spinal cord of rats and demonstrated to promote axonal growth into the lesion area; moreover, animals improved in the widely used behavioral Basso, Beattie and Bresnahan scale (Pertici et al., 2014). Another strategy evaluated to bridge the two stumps after a spinal cord lesion is the use of tubular structures, which provide mechanical support for axonal regeneration. A tubular device made with poly (2-hydroxyethyl methacrylate-co-methyl methacrylate, PHEMA-MMA) hydrogel was used to join the transected spinal cord of rats and this strategy resulted in the re-establishment of tissue continuity, allowing axon regeneration with minimal scar tissue. Although the empty tubular structure promoted recovery by itself, it was proposed that filling the tubular structure with a suitable substrate could be a better option to promote recovery (Tsai et al., 2004). The same group evaluated this approach (Tsai et al., 2006), filling the tubular structure used previously with matrices of Collagen, Fibrin, Matrigel, methylcellulose or smaller PHEMA-MMA tubes. In addition, Fibrin and Collagen were supplemented with Fibroblast growth factor-1 (FGF-1) and NT-3. It was observed that almost all the matrices used promoted more axonal regeneration compared to unfilled structures. The addition of FGF-1 increased the axonal regeneration of vestibular neurons, and the addition of NT-3 decreased the total number of axons regenerating from brainstem neurons.

In addition to promote axonal regeneration, hydrogels can incorporate into its structure trophic factors and release them on site after a spinal cord lesion. PHEMA hydrogels soaked with BDNF and control PHEMA hydrogels were implanted into the hemisected spinal cords of rats. Only the BDNFcontaining hydrogel allowed axonal growth into the polymer structure (Bakshi et al., 2004). Similarly, BDNF was embedded into agarose hydrogel and implanted into hemisected spinal cords of rats. It was demonstrated that it promoted greater axonal growth in contrast to hydrogels without BDNF (Jain et al., 2006). Another trophic factor evaluated in spinal cord lesions is NT-3, which was combined to a hydrogel of acrylated PLAb-PEG-b-PLA to release it in hemisected cord of rats. Animals treated with hydrogel and NT-3 presented more axonal growth into the lesion site and improved in the behavioral parameters, in contrast to animals treated only with hydrogel (Piantino et al., 2006). Collectively, these experiments strongly suggest that

incorporation of molecules that promote axonal growth and/or cell survival increases the possibilities of recuperation after a spinal cord lesion.

As mentioned earlier, hydrogels can incorporate living cells, making grafting of cells together with hydrogel an additional strategy to promote recovery. SC and NSPC have been implanted with and self-assembling peptide-based hydrogel into transected spinal cord of rats. The ingrowth of tissue to the lesion was better in animals treated with hydrogel-embedded cells than those treated with hydrogel alone. NSPC and SC can survive, migrate, and differentiate into the site of lesion, with SC promoting greater axonal growth into damaged area (Guo et al., 2007). Another study demonstrated that when the same hydrogel was implanted with SC in a moderate spinal cord contusion model in rats, there was a reduction of astrogliosis reaction, a motor recovery and infiltration of endogenous SC to the lesion site was observed (Moradi et al., 2012).

#### **Hydrogels promote regeneration in the PNS**

Although the PNS allows some degree of regeneration, mismatches are frequent and limit its recovery. The best strategy to promote axonal regeneration in peripheral nerve injuries is the use of autografts, but this convey some problems such as donor tissue morbidity and loss of function in the tissue innervated by donor nerve (Schlosshauer et al., 2006). Some attempts have been made to substitute the autografts with variable results. Hydrogel porous tubes constructed with PHEMA-MMA were implanted into interrupted sciatic nerves. At early times the autografts were more effective as evaluated by histomorphology and electrophysiology. However, after 8 weeks the scaffold showed a bimodal recovery: 60% of animals surpassed the autografts but rest did not, probably due to tube collapse (Belkas et al., 2005). A more rigid hybrid conduit was designed with poly (3,4-ethylenedioxythiophene, PEDOT) and agarose hydrogel. This device was implanted in 10 mm peroneal nerve gaps and the regeneration was evaluated by muscle mass, contractile force measurements and nerve histomorphometry. The hybrid conduits promoted better regeneration compared to agarose-only hydrogel, but autografts presented much better results (Abidian et al., 2012).

Filling of conduits with a substrate, which allows axonal growth, could enhance the regenerative process (**Figure 2B**). An empty blood vessel filled with a self-assembling peptide hydrogel, implanted in a sciatic nerve gap of 10 mm, promoted higher numbers of growing and re-myelinated axons, more SC infiltration, less presence of lymphocytes and macrophages, greater gastrocnemius muscle recovery and better behavioral improvement, compared to empty conduits. However, the recovery was not comparable in retrograde labeling and electrophysiology, to unlesioned animals (Zhan et al., 2013). Another group developed a Keratin-based hydrogel to fill commercial tubular conduits, which improved histological characteristics such as number of blood vessels, axons per area, and axon size. Furthermore, electrophysiological features such as conduction delay and impulse amplitude were better than with the empty tubular structures, and comparable to autografts. These results were obtained in mice with a 4 mm gap in the tibial nerve (Sierpinski et al., 2008) and in rabbits with a 2–3 cm sciatic nerve break (Hill et al., 2011). The same group characterized the early cellular response after implantation of a commercial tubular structure filled with Keratin-based hydrogel, Matrigel or saline solution in rats presenting a 1 cm sciatic nerve injury. Significant differences present in the hydrogel group compared to others were: an earlier migration of dedifferentiated endogenous SC from the proximal end, faster SC dedifferentiation, higher myelin debris clearance, and decreased macrophage infiltration. However, others parameters, such as axon density, SC labeling or the amount of cells in the distal nerve did not present differences (Pace et al., 2013). It is worth mentioning that this is the only study in which the cellular response was characterized post-hydrogel implantation after peripheral nerve injury.

The release of molecules *in situ* after a lesion could enhance axonal growth through the lesion, contributing to reinnervation of correct areas (Macaya and Spector, 2012), increasing the possibilities of recovery (**Figure 2B**). Animals suffering from a 10 mm gap in the sciatic nerve were implanted with PHEMA-MMA hydrogel porous tubes, filled with Collagen matrices supplemented with NT-3, BDNF and FGF-1. The rats treated with growth factors presented better axonal regeneration compared to animals receiving empty tubes, or Collagen without factors. Tubes filled with collagen and 10 µg/ml FGF-1 presented similar number of fibers with diameters similar to animals that received autografts (Midha et al., 2003). Similarly, polysulfone tubes filled with agarose hydrogel containing Laminin-1 and NGF, implanted in the severed sciatic nerve caused equivalent recovery to animals that received autografts in parameters such as morphology of the regenerated nerve and the density of myelinated axons. However, although the functional recoveries of sciatic nerve were similar after hydrogel or autograft treatment, these values were significantly lower than the non-lesioned condition (Yu and Bellamkonda, 2003).

Ultrafiltration membrane conduits filled with a selfassembling peptide hydrogel containing SC were implanted into the damaged sciatic nerve. This device caused better axonal growth and linear alignment of nerve fibers with SC than conduits filled with: (a) self-assembling peptide-only hydrogel; (b) alginate/Fibronectin hydrogel or (c) alginate/Fibronectin with SC. (McGrath et al., 2010). These studies demonstrated that the combination of a tubular structure, which provides mechanical support, filled with hydrogels increases the possibilities of axonal regeneration after peripheral nerve injury. In addition, this system could be improved by the addition of trophic factors or cells (**Figure 2B**). The reported recoveries are to some extent similar to those resulting from autografts, the current gold standard to treat peripheral nerve damage. However, further studies that evaluate the recovery with additional parameters, such as electrophysiological studies and anterograde/retrograde labeling through regenerated axons across the damaged area are still needed.

#### **CONCLUSION**

Increasing the possibilities for axonal regeneration after neuronal damage is a complex challenge, because it is necessary to overcome several limitations, which might imply different strategies. Hydrogels have demonstrated to be useful to overcome some of these barriers, particularly by providing an adequate substrate for axonal growth. Their versatility allows modification of important parameters, which can positively impact on axonal regeneration, and this is a significant advantage compared to other biomaterials. Although different strategies such as implantation of hydrogel, alone or combined with trophic factors or with cells, have proved to promote axonal regeneration in different animal models, more research is needed to determine if hydrogels can be applied in the clinical setting in the future. Tissue regeneration seems to consistently occur after hydrogel application, but other parameters, particularly the electrophysiological and behavioral tests show more variable results, and these shortcomings will hopefully be resolved soon.

#### **ACKNOWLEDGMENTS**

The work in the laboratory of I.V. is supported by grants from Conacyt (131281) and Papiit-UNAM (IN208713). Oscar A. Carballo-Molina received a graduate fellowship from Conacyt. We thank Drs. Francisco Fernández de Miguel and Alfredo Varela-Echavarría for critical reading on early stages of this manuscript.

### **REFERENCES**


characteristics. *Proc. Natl. Acad. Sci. U S A* 100, 5413–5418. doi: 10.1073/pnas. 0737381100


**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 October 2014; accepted: 09 January 2015; published online: 17 February 2015*.

*Citation: Carballo-Molina OA and Velasco I (2015) Hydrogels as scaffolds and delivery systems to enhance axonal regeneration after injuries. Front. Cell. Neurosci. 9:13. doi: 10.3389/fncel.2015.00013*

*This article was submitted to the journal Frontiers in Cellular Neuroscience*.

*Copyright © 2015 Carballo-Molina and Velasco. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution and 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*.

# Identification of age- and disease-related alterations in circulating miRNAs in a mouse model of Alzheimer's disease

#### *Sylvia Garza-Manero1†, Clorinda Arias 1, Federico Bermúdez-Rattoni 2, Luis Vaca3 and Angélica Zepeda1 \**

*<sup>1</sup> Departamento de Medicina Genómica y Toxicología Ambiental, Instituto de Investigaciones Biomédicas, Universidad Nacional Autónoma de México, Coyoacán, México*

*<sup>2</sup> División de Neurociencias, Instituto de Fisiología Celular, Universidad Nacional Autónoma de México, Coyoacán, México*

*<sup>3</sup> Departamento de Biología Celular, Instituto de Fisiología Celular, Universidad Nacional Autónoma de México, Coyoacán, México*

#### *Edited by:*

*Rosalinda Guevara-Guzmán, Universidad Nacional Autónoma de México, Mexico*

#### *Reviewed by:*

*Hermona Soreq, The Hebrew University of Jerusalem, Israel Ertugrul Kilic, Istanbul Medipol University, Turkey*

#### *\*Correspondence:*

*Angélica Zepeda, Departamento de Medicina Genómica y Toxicología Ambiental, Instituto de Investigaciones Biomédicas, Universidad Nacional Autónoma de México, Circuito Exterior s/n, Coyoacán, CP 04510, México e-mail: azepeda@ biomedicas.unam.mx*

#### *†Present address:*

*Sylvia Garza-Manero, Epigenetics Unit, Institute of Cancer Sciences, University of Glasgow, Glasgow, UK* Alzheimer's disease (AD) is a neurodegenerative disorder characterized clinically by the progressive decline of memory and cognition. Histopathologically, two main hallmarks have been identified in AD: amyloid-β peptide extracellular neuritic plaques and neurofibrillary tangles formed by posttranslational modified tau protein. A definitive diagnosis can only be achieved after the *post mortem* verification of the histological mentioned alterations. Therefore, the development of biomarkers that allow an early diagnosis and/or predict disease progression is imperative. The prospect of a blood-based biomarker is possible with the finding of circulating microRNAs (miRNAs), a class of small non-coding RNAs of 22–25 nucleotides length that regulate mRNA translation rate. miRNAs travel through blood and recent studies performed in potential AD cases suggest the possibility of finding pathology-associated differences in circulating miRNA levels that may serve to assist in early diagnosis of the disease. However, these studies analyzed samples at a single time-point, limiting the use of miRNAs as biomarkers in AD progression. In this study we evaluated miRNA levels in plasma samples at different time-points of the evolution of an AD-like pathology in a transgenic mouse model of the disease (3xTg-AD). We performed multiplex qRT-PCR and compared the plasmatic levels of 84 miRNAs previously associated to central nervous system development and disease. No significant differences were detected between WT and transgenic young mice. However, age-related significant changes in miRNA abundance were observed for both WT and transgenic mice, and some of these were specific for the 3xTg-AD. In agreement, variations in the levels of particular miRNAs were identified between WT and transgenic old mice thus suggesting that the age-dependent evolution of the AD-like pathology, rather than the presence and expression of the transgenes, modifies the circulating miRNA levels in the 3xTg-AD mice.

**Keywords: early diagnosis, neurodegenerative diseases, blood-based biomarker, Alzheimer models, plasma, pathological aging, prognosis, 3x-Tg**

# **INTRODUCTION**

Alzheimer's disease (AD) is the most common cause of dementia clinically characterized by the progressive decline of memory. At the histopathological level two main hallmarks define it: the abnormal extracellular accumulation of amyloid-β peptide (Aβ) into neuritic plaques and the formation of intraneuronal neurofibrillary tangles (NFTs) composed by posttranslational modified tau protein. These pathological markers are mainly located in hippocampus and cortical regions (Selkoe, 2001; Ballard et al., 2011) and are accompanied by synaptic loss early in the disease, which constitutes the major morphological correlate of memory decline (Terry et al., 1991; Dekosky et al., 1996). AD is the major neurodegenerative disorder in the elderly; the prevalence of the disease increases with age, thus aging constitutes the main risk factor (World Alzheimer Report, ADI, 2009). Two recognized forms of the disease include the sporadic AD (SAD) and the familial AD (FAD). SAD accounts for more than 99% of total cases and is characterized by a late-onset of symptomatology, whereas FAD is an early-onset form of the disease associated with mutations in three genes encoding for the amyloid precursor protein (APP), presenilin 1 (PS1), and presenilin 2 (PS2) transmitted in an autosomal dominant pattern (Goate et al., 1991; Levy-Lahad et al., 1995; Rogaev et al., 1995; Sherrington et al., 1995).

Currently, amyloid-imaging positron emission tomography (PET) through the Pittsburgh Compound-B (PIB) tracer is the best tool for identifying amyloid deposits in patients (Klunk et al., 2004; Alkalay et al., 2013). However, a definitive diagnosis of AD can only be achieved after *post mortem* histopathological verification. Therefore, the development of sensitive and specific non-invasive methods to detect molecular markers or to best predict the disease progression is crucial. Among different prospects of a blood-based biomarker, the finding that circulating microRNAs (miRNAs) derived from different types of cells can be detected, make possible their use as non-invasive markers of brain dysfunction. miRNAs are small non-coding RNAs of 22–25 nucleotides length that posttrancriptionally regulate the level of many proteins via mRNA silencing or degradation (Bartel, 2009; Fabian and Sonenberg, 2012; Ameres and Zamore, 2013). They travel through blood within specialized vesicles known as exosomes, associated with protein complexes or into high density lipoproteins as well (Valadi et al., 2007; Arroyo et al., 2011; Vickers et al., 2011). Because endogenous miRNAs are protected in circulation, they are highly stable and their levels can be analyzed by conventional techniques of molecular biology through which changes in circulating miRNA levels have been associated with diverse pathologies (Chen et al., 2008; Lawrie et al., 2008; Mitchell et al., 2008).

Analysis of miRNA profiles in *post mortem* brain and cerebrospinal fluid (CSF) samples of AD patients have revealed changes in miRNA levels associated to the disease (Lukiw, 2007; Cogswell et al., 2008; Hébert et al., 2008; Wang et al., 2008b, 2011; Nunez-Iglesias et al., 2010; Alexandrov et al., 2012; Müller et al., 2014). Recent studies performed in plasma samples of potential AD cases suggest the possibility of finding differences between circulating miRNA levels of AD cases and control subjects (Geekiyanage et al., 2011; Kumar et al., 2013; Leidinger et al., 2013; Tan et al., 2013; Burgos et al., 2014; Kiko et al., 2014). In this regard, studies have aimed at analyzing miRNA levels at a single time-point, which may limit their use as biomarkers or as disease predictors. The possibility to evaluate circulating miRNA levels at different time-points in the progression of the disease could provide information regarding key physiopathological features of AD and could contribute in establishing early blood-based molecular biomarkers for the disease. Analyzing miRNA levels in blood samples from patients at an early time-point is hampered by the fact that symptoms do not develop until the disease is at an advanced stage. Thus, in this study we used the triple transgenic mouse model (3xTg-AD) which reproduces the age-dependent progress of amyloid pathology and develops the tauopathy similar to that found in AD (Oddo et al., 2003). In order to evaluate specific miRNA levels in plasma samples at different time-points of the evolution of the AD-like pathology, we performed multiplex qRT-PCR and compared the levels from 3xTg-AD and WT mice of 84 miRNAs previously associated to central nervous system (CNS) development (Giraldez et al., 2005; Yoo et al., 2009; Zhao et al., 2009; Edbauer et al., 2010) and diseases such as schizophrenia (Beveridge et al., 2010), Huntington disease (Johnson et al., 2008), Parkinson disease (Kim et al., 2007; Wang et al., 2008a), and AD.

No significant differences in circulating miRNA levels were detected between 3xTg-AD and WT young mice. However, agerelated significant changes in the abundance of certain plasmatic miRNAs were observed in both WT and transgenic mice. Interestingly, some of the miRNAs that changed were specific for the 3xTg-AD. In agreement, variations in the abundance of particular miRNAs were identified between WT and transgenic old mice thus suggesting that the age-dependent evolution of AD-like pathology, rather than the presence and expression of human transgenes, modifies the circulating miRNA profile in the 3xTg-AD mice.

#### **MATERIALS AND METHODS**

#### **ANIMALS**

Homozygous 3xTg-AD developed by Oddo et al. (2003) (*n* = 14) and wild-type (WT) (*n* = 13) male mice (strain 129/C57BL6) were used in the study. For developing the transgenic mice, human APP cDNA harboring the Swedish double mutation (KM670/671NL) and human tau cDNA harboring the P301L mutation were cloned into Thy1.2 expression cassette and comicroinjected into single-cells embryos harvested from homozygous PS1*M*146*<sup>V</sup>* knockin (KI) mice generated as a hybrid 129/C57BL6 background. For this study, all used 3xTg-AD mice were genotyped by identifying the three harbored human transgenes (Supplementary Figure 1). Brains from four 3xTg-AD male mice 2–3 (*n* = 2) and 14–15 (*n* = 2) months old were used for immunohistochemical analysis. Determination of circulating miRNA profiles included blood samples from six-seven animals per group and a total of four experimental groups were integrated: 2–3 months old-3xTg-AD mice (*n* = 6), 2–3 months old-WT mice (*n* = 6), 14–15 months old-3xTg-AD (*n* = 6), and 14–15 months old-WT mice (*n* = 7). The experiments were performed in accordance with local government rules and the Society for Neuroscience Guide for the Care and Use of Laboratory Animals with approval of the Animal Care Committee of the Instituto de Investigaciones Biomedicas, UNAM. Efforts were made to minimize animal suffering and to reduce the number of subjects used.

#### **GENOTYPING**

Genomic DNA was isolated from the 3xTg-AD mice tails (2–3 mm) by the *Hot-Shot* method. Tail samples were homogenized with alkaline lysis reagent (25 mM NaOH, 0.2 mM EDTA, pH 12) through mechanical action and incubated during 1 h at 95◦C. Tris-HCl buffer (40 mM, pH 5) was added to each sample for neutralization (pH 7.5). Samples were centrifuged during 2 min at 12 500 rpm, 4◦C, and the supernatant was used for PCR amplification. Human APPSwe and tauP301L transgenes were identified through their amplification products of 500 bp and 320 bp, respectively into a 2% agarose gel stained with ethidium bromide. The primers used for APP-tau PCR were 5tauRev (5 -TCCAAAGTTCACCTGATAGT-3 ), APPinternal (5 -GCTTG CACCAGTTCTGGATGG-3 ) and Thy12.4 (5 -GAGGTATTC AGTCATGTGCT-3 ). PS1M146V KI was detected by a PCR using the PS1-K13 (5 -CACACGCACACTCTGACATGCACAGGC-3 ) and PS1-K15 (5 -AGGCAGGAAGATCACGTGTTCCAAGTAC-3 ) primers, followed by an enzymatic digestion (1 h, 60◦C) with endonuclease BstEII (New England Biolabs). Products were separated into a 2% agarose gel, and identified as a 530 bp band from WT PS1, or 350 bp and 180 bp bands from mutation carriers PS1.

#### **IMMUNOHISTOCHEMISTRY**

3xTg-AD mice (*n* = 4) were anesthetized with sodium pentobarbital and transcardially perfused with 0.9% chilled saline followed by 4% chilled formaldehyde in 0.1 M phosphate buffer (pH 7.4). Brains were extracted and post-fixed during 24 h at 4◦C. Subsequently, brains were transferred to 15 and 30% sucrose solutions in 0.1 M phosphate buffer for 24 h in each concentration. 30μm-thick coronal sections were obtained in a cryostat (Microm HM550, Thermo Scientific). Free-floating sections were collected in 24-well plates (Corning) and were washed for 10, 15, and 30 min with 0.15 M PBS and 0.4% PBS-triton solutions. They were then incubated in blocking solution (0.4% PBS-triton, 3% NGS) during 2 h at room temperature, followed by an overnight incubation at 4◦C with 1:500 dilution of the primary antibody, either the MAB1560 6E10 clone (Chemicon) which recognizes aminoacids 1–17 from human Aβ peptide, or the MN1020 α-p-PHF-tau AT8 clone (Pierce, Thermo Scientific) which recognizes phosphorylated serine 202 and threonine 205 from human tau protein. Sections were washed as previously detailed and were then incubated in a 1:250 dilution of the secondary α-mouse antibody, either Alexa-488 or Alexa-555 for 2 h at room temperature. Finally, sections were washed for 10, 15, and 30 min in 0.15 M PBS, incubated for 3 min in 30 nM DAPI, and washed again. Immunofluorescence was detected using a Zeiss epifluorescence microscope under 10X and 60X.

#### **DETERMINATION OF CIRCULATING miRNA LEVELS** *Plasma collection*

Blood samples were collected from 3xTg-AD and WT mice at 2– 3 and 14–15 months of age. ∼500μL of blood were obtained from the facial vein of each mouse and collected into a new and sterile 1.5 mL centrifuge tube containing 10μL of 300 mM EDTA used as anticoagulant. Plasma was separated through two centrifugations, one at low (3000 rpm) and another at high (13,000 rpm) speed, both at 4◦C for 10 min. For determining circulating miRNA profiles, equal volumes of plasma samples from two mice within the same experimental group were pooled after plasma separation and a total of three different pools per experimental group were included in the analysis. Pooled samples were stored in ice and immediately treated following the RNA isolation protocol.

#### *RNA isolation*

RNA was isolated from each pool using the miRNeasy Serum/Plasma Kit (Qiagen) following the manufacturer instructions. Briefly, 200μL of pooled plasma were homogenized with 1000μL of Qiazol Lysis Reagent. After homogenization, 3.5μL of a 1.<sup>6</sup> <sup>×</sup> <sup>10</sup><sup>8</sup> copies/μL solution of the *C. elegans* miR-39 miRNA mimic used as a spike-in control were added to each sample. Then, a phenol-chloroform extraction was performed by incubating the sample in 200μL of chloroform for 5 min at room temperature and centrifuging for 15 min at 12,000 rpm at 4◦C. The aqueous phase was mixed with 1.5 volumes of 100% ethanol and passed through the RNeasy MinElute spin column. The RNA retained in the membrane was washed with buffer solutions, 80% ethanol and eluted in 14μL of RNase free water. RNA samples were stored in ice and immediately treated for qRT-PCR protocols.

#### *qRT-PCR*

miRNA levels from plasma of 3xTg-AD and WT mice 2–3 and 14–15 months old were determined using the miScript miRNA PCR Array System (Qiagen). The System consists in three different kits: the miScript II RT Kit which performs the cDNA synthesis, the miScript SYBR Green PCR Kit which prepares the mix for qPCR reactions, and the Pathway-Focused miScript miRNA PCR Array "Neurological Development and Disease" which detects 84 different miRNAs previously associated with CNS development and disease. For the cDNA synthesis, 1.5μL of RNA isolated from plasma were used as the starting material of a 20μL-reaction that consists of a universal synthesis driven by the addition of a poly-A tail and an oligo-dT that selectively retrotranscribes the small RNA molecules (<100 nucleotides) avoiding the retrotranscription of miRNA precursors. The reaction was incubated for 1 h at 37◦C followed by a 5 min-incubation at 95◦C to inactivate the RTase enzyme. cDNA from each sample was diluted by adding 200μL of RNase free water and stored at −20◦C overnight. After thawing the cDNA on ice, 100μL of sample were added to the qPCR reaction mix using SYBR Green as a detector. 20μL of qPCR reaction mix were added to each well of the array. The multiplex qPCR was performed using the Rotor Gene 6000 under the following conditions: pre-incubation for 15 min at 95◦C, 40 denaturation-alignment-elongation (15 s at 94◦C, 30 s at 55◦C, and 30 s at 70◦C) cycles and a dissociation melt protocol.

#### **DATA ANALYSIS**

Data were analyzed using the --Ct (2−-Ct) method of relative quantification with the support of the data analysis software for miScript miRNA PCR Arrays available for users at http://pcrdataanalysis.sabiosciences.com/mirna. The settings were manually fixed at 10 cycles of baseline and 0.02 of threshold line across all PCR runs. The software calculated the -Ct value between the threshold cycle value (Ct) of each miRNA and the Ct of the *C. elegans* miR-39 miRNA mimic spike-in control used for normalization. The software also calculated the (2−-Ct) value of each miRNA between arrays from different experimental groups. Each experimental group consisted of three samples (*n* = 3) and each sample contained the pooled plasma from two animals. The mean of the relative abundance ± SD of each miRNA contained in the samples (*n* = 3) was calculated. The statistical significance between miRNAs was determined by a *t*-test. A dissociation curve analysis was performed to guarantee the presence of only one PCR product.

### **RESULTS**

In order to detect possible variations in the circulating miRNA profiles related to the AD-like pathology present in the 3xTg-AD transgenic mice, we evaluated the levels of 84 miRNAs previously associated to CNS development and disease in plasma samples of 3xTg-AD and WT mice 2–3 and 14–15 months old. As previously shown (Oddo et al., 2003), these two time-points represent different stages of evolution of the pathological features in the 3xTg-AD mice: at 2–3 months old, histopathological markers are not evidently expressed in the hippocampus whereas at 14– 15 months old they become clear in cortical and hippocampal regions. The 2–3 month old-3xTg-AD mice express human transgenes; nevertheless at this time-point we did not find extracellular amyloid aggregates in the hippocampus and scarce phosphotau (p-tau) immunoreactivity for AT8 was observed (**Figure 1A**).

As previously reported (Oddo et al., 2003) we found that the 14–15 month old-3xTg-AD mice display extracellular amyloid aggregates. Also, neuronal somata and processes were strongly immunopositive for p-tau in AD relevant residues (**Figure 1B**).

100μm (top images) and 20μm (bottom images).

#### **ABUNDANCE OF CNS DEVELOPMENT- AND DISEASE-LINKED miRNAs IN MOUSE PLASMA**

Most of the 84 evaluated miRNAs were detected in mice plasma from all analyzed groups (Supplementary Tables 1–4). Highly abundant detected miRNAs belong to the let-7 family, miR-15 family, miR-30 family, miR-24-27 cluster, miR-29 cluster, miR-17-92 cluster and its paralogs miR-106a-363 and miR-106b-25 that are characterized by their presence in multiple cell lineages as well as by their role in fundamental cellular processes. We found a group of less or not detected miRNAs such as miR-135b, miR-302a/b, miR-488, and miR-9. The plasmatic abundance of detected miRNAs was compared between groups. No significant differences were observed in the circulating miRNA profile from 2–3 months-old 3xTg-AD mice when compared to the age-matched WT group (Supplementary Table 1).

When comparing the circulating miRNA profile between the young and the old mice for both WT and 3xTg-AD, we found that aging associated with changes in the plasmatic levels of a group of miRNAs. We detected significant differences in the levels of 33 miRNAs when comparing old vs. young WT mice and in 40 miRNAs when comparing old vs. young 3xTg-AD mice; 19 of these miRNAs were common to both comparisons (**Figure 2**, Supplementary Tables 2,3). These overlapping miRNAs include family members of let-7, miR-30, miR-17-92 cluster and its paralogs. When comparing young vs. old 3xTg-AD mice, we identified a particular group of miRNAs integrated by miR-132, miR-138, miR-146a, miR-146b, miR-22, miR-24, miR-29a, miR-29c, and miR-34a which show significant differences in plasma levels only in the transgenic group, raising the possibility of age-related changes that specifically occur in the 3xTg-AD mice (**Figure 3**, Supplementary Table 3).

#### **PATHOLOGY-RELATED CHANGES IN THE RELATIVE ABUNDANCE OF PLASMATIC miRNAs**

When comparing the plasma of 14–15 months old 3xTg-AD mice vs. age-matched WT mice, we detected a significant lower abundance of miR-132, miR-138, miR-139, miR-146a, miR-146b, miR-22, miR-24, miR-29a, and miR-29c as well as a higher abundance of miR-346 (**Figure 4**, Supplementary Table 4). This finding suggests that the relative abundance of these miRNAs was altered specifically in the 3xTg-AD mice only at age 14–15 months since they were not altered at 2–3 months of age. These results correlate with the appearance of the histopathological features (**Figure 1**).

# **DISCUSSION**

AD is a neurodegenerative disorder highly related to aging that exhibits progressive manifestations at both clinical and histopathological levels. The evolution of AD neuropathological features in the 3xTg-AD is also age-dependent (Oddo et al., 2003). In this work, we evaluated the levels of circulating miRNAs in four groups of animals: young WT; young 3xTg-AD; old WT and; old 3xTg-AD. For this purpose, we pooled plasma from two different animals from the same group and conformed three samples per group. Obtaining sufficient plasma for miRNA isolation can be problematic with young animals and getting sufficient animals with the triple transgene is difficult especially with old animals, since their health declines rapidly and mortality increases. Pooling the samples on the one hand, may pose a limitation given that each sample contains plasma from two different animals. But on the other hand, this design allowed us to avoid the bias from changes in circulating miRNAs associated to each individual. Even when the 3xTg-AD already bears the pathological transgenes, we did not find changes in the circulating miRNA profile between WT and 3xTg-AD young mice. This suggests that the presence of the transgenes is not sufficient to modify the circulating levels

**and 3x-Tg-AD analyzed by qRT-PCR. (A)** The Venn diagram depicts the number of miRNAs showing statistically different plasma levels between young and old mice in WT (33 miRNAs, in light gray) and in 3xTg-AD mice (40 miRNAs, in black) as shown by a *t*-test; *p* < 0.05. From these miRNAs, 19 are common to both comparisons (dark gray). **(B)** Relative abundance among

calculated considering the difference of the Ct of each miRNA and the Ct of the miR-39 of *C. elegans* mimic (-Ct) using the formula 2−-Ct. miR-39 from *C. elegans* was used as spike-in control in all qRT-PCR experiments. Only miRNAs that showed statistical differences when comparing young and old WT or Tg subjects are shown in the graph; *t*-test; *p* < 0.05.

**FIGURE 4 | Significant differences in the abundance of plasmatic miRNAs between aging groups analyzed by qRT-PCR.** The graph depicts the relative abundance (mean ± SD) of each one of the miRNAs that show statistically significant differences (*t*-test; *p* < 0.05) in plasma levels between 14–15 months-old WT and 3xTg-AD mice. Relative abundances were calculated with the difference of the Ct of each miRNA and the Ct of the miR-39 of *C. elegans* mimics (-Ct) using the formula 2−-Ct. Determination of circulating miRNA profiles included plasma samples from six-seven animals per group and a total of four experimental groups were integrated: 2–3 months old-3xTg-AD mice (*n* = 3), 2–3 months old-WT mice (*n* = 3), 14–15 months old-3xTg-AD (*n* = 3), and 14–15 months old-WT mice (*n* = 3) (*n* = samples).

of miRNAs at early stages of the pathology. Nevertheless, we found age-related changes in the plasmatic abundance of certain miRNAs in both WT and 3xTg-AD mice. Remarkably we found changes in specific circulating miRNAs in the aging AD model. In agreement with other studies (Grillari et al., 2010; Kato et al., 2011; Martinez et al., 2011) we found a decrease in abundance of members of the family of let-7, miR-30, miR-17-92 cluster and its paralogs miR-106a-363 and miR-106b-25 in WT and 3x-Tg-AD aged mice. Let-7 family is highly involved in developmental processes in multiple species; it is required for cell differentiation thus its levels increase in late developmental stages and remain high in the adult (Thomson et al., 2004). Accordingly, let-7 family members target important cell cycle molecules (Schultz et al., 2008). miR-17-92 cluster and its paralogs are other important cell cycle modulators; interestingly, they are up-regulated in cancer and suppression of these clusters' members induce cell growth arrest in cancer models, whereas they are down-regulated in aging and their overexpression induces cellular senescence (Grillari et al., 2010).

In addition, we found some miRNAs involved with an inflammatory response such as miR-146a and miR-146b that display altered levels in plasma of both groups of old mice. This is particularly relevant considering that it has been established that inflammation is a condition closely related to the aging process. miR-146a and 146b are NF-κB sensitive-miRNAs that suppress the immune response by inhibiting the pro-inflamatory immune cell signaling (Taganov et al., 2006). Interestingly, miR-146a displays a higher abundance in the 14–15-month old WT and a lower abundance in the 14–15-month 3xTg-AD as compared to their respective group of young mice. These bidirectional modulation agrees with recent data indicating that miR-146a is elevated in senescence models (Olivieri et al., 2013; Rippo et al., 2014), while it is decreased in AD (Kiko et al., 2014; Müller et al., 2014). A similar case is miR-34a, which is present with lower abundance in the 14–15-month 3xTg-AD mice as compared to the 2–3-month 3xTg-AD while showing a trend to increase with age in the WT. Higher levels of this miRNA are found in the brain, blood cells and plasma of old mice (Li et al., 2011) but the loss of miR-34a induces neurodegenerative phenotypes in *Drosophila*, and its overexpression induces longevity (Liu et al., 2012). Remarkably, lower levels of miR-34a have been reported in AD cases (Kiko et al., 2014; Müller et al., 2014).

Thus, it is evident that the age-dependent evolution of AD-like pathology produces specific changes in the circulating miRNA profile of the 3xTg-AD mice. As a consequence, we identified differences in the abundance of 10 miRNAs when comparing plasma samples of the 14–15- month 3xTg-AD with the 14–15-month WT mice. Most of the alterations observed in the aged 3xTg-AD mice consist on the reduction of plasmatic miRNAs levels as compared to the aged WT. This is in agreement with previous reports from brain, CSF and serum/plasma samples of AD patients where miR-15, miR-29, miR-101, miR-106, miR-107, and miR-181 have been shown to be diminished (Cogswell et al., 2008; Hébert et al., 2008; Wang et al., 2008b, 2011; Geekiyanage and Chan, 2011; Geekiyanage et al., 2011; Kumar et al., 2013; Tan et al., 2013; Burgos et al., 2014; Kiko et al., 2014).


**Table 1 | AD-like pathology-related miRNAs identified in this study and previously associated with AD.**

*The table contains data obtained from this study (left) and collected from others (right). It indicates the miRNA, the sample used for determination, the abundance respect to the control cases and the reference of the other studies. It also contains the mRNA-target for the miRNA: p250-GAP, a brain-enriched GTPase-activating protein for Rho Family GTPases involved in the N-Methyl-d-aspartate receptor (NMDAR) signaling; APT1, acyl protein thioesterase 1, an enzyme regulating the palmitoylation status of proteins that are known to function at the synapse; IRAK1, interleukin-1 receptor-associated kinase 1, a kinase that associates with the interleukin-1 receptor upon stimulation which is responsible for interleukin-1 transcription of NFkB; TRAF6, TNF receptor associated factor 6 involved in the regulation of inflammation response and apoptosis; BACE1,* β*-secretase, a protease that cleaves the APP in the* β *site to produce the A*β *peptide in the amyloidogenic APP processing.*

Some of the 10 AD-like pathology-related miRNAs identified in this study have been previously associated with AD in different studies (**Table 1**). For instance, the miR-29 cluster is decreased in brain samples of AD patients, which correlates with an increase of the levels of the β-sercretase protein. Furthermore, members of this cluster have been shown to regulate the levels of the β-secretase *in vitro* (Hébert et al., 2008). Levels of miR-146a and miR-146b have been reported diminished in brain, CFS and plasma of AD patients (Cogswell et al., 2008; Kiko et al., 2014; Müller et al., 2014). These miRNAs are related to the inflammatory response, and may constitute an interesting result since neuroinflammation is a prevalent and early pathological feature of AD (Lukiw et al., 2008). In agreement with our results, lower levels of miR-132, miR-138 y miR-139 have been reported in AD brain and CFS samples (Cogswell et al., 2008; Burgos et al., 2014). miR-132 has been reported to have an anti-inflammatory effect, blocking acetylcholinesterase, elevating acetylcholine levels and consequently blocking NF-kB (Shaked et al., 2009). In neurons, miR-132 and miR-138 are expressed in response to synaptic activity and have a role in the modulation of morphologic events of neuroplasticity taking place in memory and cognition processes (Wayman et al., 2008; Siegel et al., 2009; Edbauer et al., 2010; Impey et al., 2010; Hansen et al., 2012; Bicker et al., 2014). Synapses represent vulnerable structures in AD and several pathologic molecular events take place in these sites comprising their function (Selkoe, 2002). Since memory and cognition are highly impaired in AD, the lower abundance of miR-132 and miR-138 result very interesting for analyzing the evolution of AD at a molecular level. It would be therefore interesting to determine changes in plasmatic miRNAs with the status of patients already showing mild cognitive impairment in order to advance the early diagnosis of AD.

In this work we used a pre-designed miRNA microarray platform, which included a restricted number of miRNAs previously associated to different neuropathologies. Other miRNAs not included in this design may be associated also to AD. However, and since to our knowledge this is the first study of circulating miRNAs in this transgenic model, we decided to incorporate miR-NAs that have been already associated to several neuropathologies. It is clear that given the multifactorial nature of the disease, it is likely that not only a single biomarker will meet the needs for clinical diagnosis. Here we propose that a relative simple, non-invasive procedure may provide useful information about the AD pathophysiology, detection, and evolution. Thus, combining a panel of miRNA detection with additional biomarkers may increase the sensitivity and specificity for early AD diagnosis.

#### **CONCLUSION**

The aim of this study was to identify changes in the circulating miRNA profile of a transgenic mouse model of AD in two different stages of the evolution of AD-like pathology. Using these two time-windows it was not possible to detect early modifications in miRNA levels associated with prodromal AD. However, we distinguish variations in different miRNAs once the AD-like pathology is established which advances the molecular biomarkers field. Most of the miRNAs we found as potentially interesting biomarkers for AD have no identified targets that may be assoiciated to the disease and their role in the pathological mechanisms taking place in AD remains unknown. Our results however shed light on subtle molecular modifications associated to pathological aging and open new venues for studying the role of particular circulating miRNAs in the evolution of AD.

#### **ACKNOWLEDGMENTS**

We thank Perla Moreno-Castilla, Patricia Ferrera, and Gonzalo Acero for technical assistance. This work was supported by grants from CONACyT-176589 and PAPIIT-IA200312 to AZ, CONACyT-127822 to LV, CONACyT-155242 and DGAPA-IN209413 to FB-R. SG-M received an MSc degree fellowship from CONACyT. Additional support was provided by the Program "Posgrado en Ciencias Bioquímicas" at the Universidad Nacional Autónoma de México (UNAM).

#### **SUPPLEMENTARY MATERIAL**

The Supplementary Material for this article can be found online at: http://www.frontiersin.org/journal/10.3389/fncel.2015. 00053/abstract

# **REFERENCES**


FMRP-associated microRNAs miR-125b and miR-132. *Neuron* 65, 373–384. doi: 10.1016/j.neuron.2010.01.005


down-regulating p250GAP. *Proc. Natl. Acad. Sci. U.S.A.* 105, 9093–9098. doi: 10.1073/pnas.0803072105


**Conflict of Interest Statement:** The Guest Associate Editor Rosalinda Guevara-Guzmán declares that, despite being affiliated to the same institution as authors Sylvia Garza-Manero, Clorinda Arias, Federico Bermúdez-Rattoni, Luis Vaca and Angélica Zepeda, the review process was handled objectively and no conflict of interest exists. 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 October 2014; accepted: 03 February 2015; published online: 19 February 2015.*

*Citation: Garza-Manero S, Arias C, Bermúdez-Rattoni F, Vaca L and Zepeda A (2015) Identification of age- and disease-related alterations in circulating miRNAs in a mouse model of Alzheimer's disease. Front. Cell. Neurosci. 9:53. doi: 10.3389/fncel. 2015.00053*

*This article was submitted to the journal Frontiers in Cellular Neuroscience.*

*Copyright © 2015 Garza-Manero, Arias, Bermúdez-Rattoni, Vaca and Zepeda. 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.*

# Identification of the antiepileptic racetam binding site in the synaptic vesicle protein 2A by molecular dynamics and docking simulations

*José Correa-Basurto1, Roberto I. Cuevas-Hernández1, Bryan V. Phillips-Farfán2, Marlet Martínez-Archundia1, Antonio Romo-Mancillas3, Gema L. Ramírez-Salinas1, Óscar A. Pérez-González2, José Trujillo-Ferrara1 and Julieta G. Mendoza-Torreblanca2\**

*<sup>1</sup> Laboratorio de Modelado Molecular y Diseño de fármacos, Departamento de Bioquímica de la Escuela Superior de Medicina del Instituto Politécnico Nacional, México City, Mexico, <sup>2</sup> Laboratorio de Nutrición Experimental, Laboratorio de Oncología Experimental and Laboratorio de Neuroquímica, Instituto Nacional de Pediatría, México City, Mexico, <sup>3</sup> División de Estudios de Posgrado, Facultad de Química, Universidad Autónoma de Querétaro, Santiago de Querétaro, Mexico*

#### *Edited by:*

*Victoria Campos-Peña, Instituto Nacional de Neurologia y Neurocirugia, Mexico*

#### *Reviewed by:*

*Benjamín Florán, Centro de Investigación y de Estudios Avanzados del Instituto Politécnico Nacional, Mexico Alan Talevi, National University of La Plata, Argentina*

#### *\*Correspondence:*

*Julieta G. Mendoza-Torreblanca, Laboratorio de Neuroquímica, Instituto Nacional de Pediatría, Insurgentes Sur No. 3700-C, Colonia Insurgentes Cuicuilco, Delegación Coyoacán, México City 04530, Mexico julietamt14@hotmail.com*

> *Received: 06 November 2014 Accepted: 17 March 2015 Published: 10 April 2015*

#### *Citation:*

*Correa-Basurto J, Cuevas-Hernández RI, Phillips-Farfán BV, Martínez-Archundia M, Romo-Mancillas A, Ramírez-Salinas GL, Pérez-González OA, Trujillo-Ferrara J and Mendoza-Torreblanca JG (2015) Identification of the antiepileptic racetam binding site in the synaptic vesicle protein 2A by molecular dynamics and docking simulations. Front. Cell. Neurosci. 9:125. doi: 10.3389/fncel.2015.00125* Synaptic vesicle protein 2A (SV2A) is an integral membrane protein necessary for the proper function of the central nervous system and is associated to the physiopathology of epilepsy. SV2A is the molecular target of the anti-epileptic drug levetiracetam and its racetam analogs. The racetam binding site in SV2A and the non-covalent interactions between racetams and SV2A are currently unknown; therefore, an *in silico* study was performed to explore these issues. Since SV2A has not been structurally characterized with X-ray crystallography or nuclear magnetic resonance, a three-dimensional (3D) model was built. The model was refined by performing a molecular dynamics simulation (MDS) and the interactions of SV2A with the racetams were determined by docking studies. A reliable 3D model of SV2A was obtained; it reached structural equilibrium during the last 15 ns of the MDS (50 ns) with remaining structural motions in the N-terminus and long cytoplasmic loop. The docking studies revealed that hydrophobic interactions and hydrogen bonds participate importantly in ligand recognition within the binding site. Residues T456, S665, W666, D670 and L689 were important for racetam binding within the trans-membrane hydrophilic core of SV2A. Identifying the racetam binding site within SV2A should facilitate the synthesis of suitable radio-ligands to study treatment response and possibly epilepsy progression.

#### Keywords: SV2A, epilepsy, levetiracetam, brivaracetam, seletracetam

# Introduction

Epilepsy is the most common chronic brain disorder that affects people of all ages. More than 50 million people worldwide have epilepsy (WHO, 2014). It is characterized by recurrent seizures,

**Abbreviations:** AM1, Austin Model 1; BRIV, brivaracetam; CNS, central nervous system; H-bonds, hydrogen bonds; -G, free energy; LEV, levetiracetam; MFS, mayor facilitator family; MDS, molecular dynamics simulation; OPLS-AA, optimized potentials for liquid simulations all atoms; PET, positron emission tomography; POPC, 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine; Rg, radius of gyration; RMSD, root mean square deviation; RMSF, root mean square fluctuations; SAS, solvent accessible surface area; SD, standard deviation; SEL, seletracetam; SV2A, synaptic vesicle protein 2A; 3D, three-dimensional; TM, trans-membrane; Uniprot, universal protein resource; VMD, visual molecular dynamics.

which may vary from a brief lapse of attention or muscle jerks, to severe and prolonged convulsions (WHO, 2014). In most patients with epilepsy, seizures respond to available medications. However, a significant number of patients -especially in the setting of medical-intractable epilepsies- may experience different degrees of memory or cognitive impairment, behavioral abnormalities or psychiatric symptoms, which limit their daily functioning. As a result, in many patients, epilepsy may resemble a neurodegenerative disease (Ono and Galanopoulou, 2012). Epileptic seizures and/or epileptogenesis may functionally alter brain regions involved in cognitive processing, contributing to the progressive nature of epilepsy; additionally neurodegenerative cellular mechanisms may also participate. To cure epilepsy, both epileptogenesis and the associated neuro-degeneration have to be stopped and, if possible, reversed. This will require early detection through biomarkers that can reliably predict disease progression (Ono and Galanopoulou, 2012).

Synaptic vesicle protein 2A is an integral membrane protein present on all synaptic vesicles, being nearly ubiquitous in the CNS (Lynch et al., 2004). The existing evidence suggests that SV2A is critical for the proper function of the CNS. It regulates vesicle exocytosis by modulating the concentration of pre-synaptic calcium (Janz et al., 1999; Chang and Sudhof, 2009; Wan et al., 2010); maybe by binding to synaptotagmin 1 (Schivell et al., 1996, 2005; Yao et al., 2010), a calcium sensor protein implicated in synaptic vesicle exocytosis. In addition, it has been proposed that SV2A may control the release of neurotransmitters as a gel matrix or could transport ions, such as chloride or calcium (Bajjalieh et al., 1994; Janz et al., 1999; Reigada et al., 2003; Mendoza-Torreblanca et al., 2013). The SV2A protein has been associated to the physiopathology of epilepsy; newborn mice lacking SV2A experience severe seizures and die within 3 weeks, suggesting multiple neural alterations (Crowder et al., 1999; Janz et al., 1999). Moreover, SV2A protein expression is significantly reduced in brain tissue obtained from epileptic patients and in rats during epileptogenesis, suggesting that decreased SV2A contributes to the progression of epilepsy (Feng et al., 2009; Toering et al., 2009; van Vliet et al., 2009). SV2A is the molecular target of the second-generation antiepileptic drug LEV and its structural analogs BRIV, SEL and UCB-30889 (Gillard et al., 2003; Lynch et al., 2004; Frycia et al., 2010). Although LEV may acutely modulate ion channels and other targets; chronically, LEV may lead to decreased transmitter release predominantly by binding to SV2A (Surges et al., 2008; Crepeau and Treiman, 2010; Lyseng-Williamson, 2011; Deshpande and Delorenzo, 2014).

Knowing the *in vivo* expression pattern of SV2A in epilepsy patients may be essential for evaluating increased epileptogenicity and thus disease progression (van Vliet et al., 2009). Moreover, SV2A expression could also be useful as a biomarker for treatment response (de Groot et al., 2011). For this purpose, sensitive neuroimaging methods such as PET may enhance our ability to detect SV2A expression (Ono and Galanopoulou, 2012). Because PET requires the use of a radio-ligand to label the protein target, as a first approximation to synthesize a radio-ligand with a high affinity and specificity, we investigated the specific binding site(s) for racetams in SV2A and the residues involved in their interaction. Therefore, an *in silico* study was performed to identify the binding site for LEV and other racetams within SV2A. Since SV2A has not been structurally characterized with X-ray crystallography or nuclear magnetic resonance, it was necessary to generate and validate a 3D model of this protein. This model was refined by a MDS and docking studies were performed to decipher the interactions between the racetams and SV2A.

# Materials and Methods

## SV2A modeling

The SV2A protein sequence was retrieved from the National Center for Biotechnology Information (NP\_476558.2) and UniProt (Q02563). There was no difference between these databases; the protein sequence was thus used as an input for the I-Tasser server (Roy et al., 2010); which provided a cluster of five 3D models for this protein. The Rampage server was used to obtain the Ramachandran plot, which shows the ϕ and ψ torsion angles for all protein residues. Additionally, the CPHmodels-3.2 and 3D-JIGSAWv3.0 servers were used to confirm the reproducibility of the SV2A 3D model provided by I-Tasser.

### System Preparation for Molecular Dynamics Simulation

The model of the SV2A protein was inserted into a preequilibrated membrane consisting of POPC molecules. The resulting system was then placed in a hexagonal prism shaped box with its symmetry axis, which was perpendicular to the plane of the bilayer membrane, in the *z* direction. The system was then solvated using simple point charge model water molecules. The geometry and size of the simulation box were carefully selected to reduce to a minimum the number of water molecules and thus decrease the computation time. In order to remove solvent molecules accidentally introduced in the hydrophobic region of the lipid bilayer, 19 Na+ ions were added to neutralize the whole system.

The simulation box contained one molecule of SV2A, 342 POPC lipids, 47,671 water molecules and 19 ions, resulting in a total of 172,298 atoms. The OPLS-AA force field was used, it has been used to simulate and describe several TM proteins with good results (Tieleman et al., 2006). The simulation was done by applying the half-ε double-pair list method to ensure compatibility between the Berger united atom parameters, used for lipids (Lindahl and Edholm, 2000; Chakrabarti et al., 2010), with the OPLS-AA force field employed for protein and ions (Jorgensen et al., 1996; Kaminski et al., 2001).

#### Molecular Dynamics Simulation

The system in its entirety was minimized for 500 steps using the *steepest descent* optimization algorithm prior to generating the MDS trajectory. Solvent–protein and lipid–protein contacts were optimized by a simulation lasting 5 ns, with harmonic constraints in the heavy atoms of the protein (all, except for hydrogen). The force constants of these positional restraints were sequentially reduced. Afterward, an unrestrained trajectory lasting 50 ns was generated. The simulation was performed using a 2 fs time step. The pressure was set at 1 bar using a semi-isotropic Parrinello and Rahman (1981) barostat (Nosé and Klein, 1983) to allow independent modification of the box dimensions in the *XY* plane and in the *Z* dimension. The temperature was kept constant at 310 K by a Nosé (1984) and Hoover (1985) thermostat. The linear constraint solver algorithm was employed to remove bond vibrations (Hess et al., 1997). The particle mesh Ewald method (Darden et al., 1993; Essmann et al., 1995) coupled to periodic boundary conditions was used to simulate long-range electrostatic interactions using a direct space cutoff of 1.2 nm and a grid spacing of 0.15 nm. Because periodic boundary conditions can produce periodic artifacts, especially if combined with Ewald methods (Essmann et al., 1995), the minimum distance between protein molecules in adjacent boxes was calculated as a function of time. The simulation box size was adequate for the system, based on the cutoff used for electrostatic and short range contributions to the potential function. The van der Waals interactions were computed using periodic boundary conditions coupled to a spherical cutoff of 1.2 nm. The MDS was performed using GROMACS 4.6.1 (Hess et al., 2008). Additionally, a replicate MDS was carried out in order to confirm the results of the first MDS.

binding motif (arrow head) and a large loop between TM domains 6–7 with

another ATP binding motif (two arrows heads).

#### Docking Studies

Docking and MDS simulations were combined by screening different SV2A conformers. LEV, BRIV, SEL and UCB-30889 were the selected racetam ligands. The criteria to include these ligands were as follows: LEV specifically binds to SV2A (Lynch et al., 2004) and is the lead anti-convulsive racetam. BRIV, SEL and UCB-30889 also bind to SV2A, but with higher affinity than LEV (Gillard et al., 2003, 2011; Matagne et al., 2009). BRIV and SEL have been the two leading candidates to replace LEV, since they were more potent and effective in animals model studies (Kaminski et al., 2012). Finally, UCB-30889 has been extensively used to characterize the binding properties of SV2A (Gillard et al., 2003; Lambeng et al., 2005; Shi et al., 2011). Interactions between the racetam ligands and nine SV2A conformers corresponding to every 2 ns during the last 16 ns of the MDS trajectory (when the system reached equilibrium) were simulated. In order to obtain the minimum energy conformation for each ligand structure, they were preoptimized with AM1 (a semi-empirical method) and then optimized using the B3LYP/6-31G(d) basis set, both implemented in Gaussian 03 (Frisch et al., 1998). Dockings were performed using AutoDock 4.0.1 and AutoDock Tools 1.5.6 (Morris et al., 2008).

Before starting the docking simulations, hydrogen atoms were added to the polar atoms (considering a 7.4 pH value) and the Kollman charges were assigned for all atoms in the receptor. All rotating bonds, torsional degrees of freedom, atomic partial charges and non-polar hydrogens of the ligands were assigned. Then the ligands were docked onto the SV2A protein using a 70 Å × 70 Å × 70 Å grid box which covered all putative residues involved in racetam recognition according to the literature (Shi et al., 2011). A grid spacing of 0.375 Å was used under the hybrid Lamarckian Genetic Algorithm with an initial population of 100 randomly placed individuals and 1 <sup>×</sup> 107 as the maximum number of energy evaluations. All other docking parameters remained at their default settings. The resulting docked orientations within a root mean square deviation (RMSD) of 0.5 Å were clustered together. The lowest energy cluster for each compound, returned by AutoDock, was used for

FIGURE 3 | Root mean square deviation values for (A) the whole backbone and (B) the backbone of the TM region as a function of time. The gray bands correspond to the SD and the line within them represents the average value.

further analysis. The script and files used were prepared with AutoDock Tools 1.5.6 and this software was also used to visualize all the ligand–protein complexes together with VMD 1.9.1 (Humphrey et al., 1996). All the artwork was created using Gnuplot 4.6, VMD 1.8, Pymol 1.5 and Photoshop CS5 Extended 12.

# Results

# SV2A Modeling

After sending the SV2A protein sequence (FASTA format) to the I-Tasser server, it returned five possible 3D models with 12 α-helical TM segments, two long loops and a long cytoplasmic N-terminus. The *C*-score was between -2.84 to -3.02, the TM-score was 0.39 ± 0.13 and the RMSD was 15.4 ± 3.4 Å. The Ramachandran plot showed ϕ and ψ angles of 94.7–96.2 % in the favored-allowed regions of the SV2A models. In four of the models, the TM region was invaded importantly by the N-terminus. They were discarded for this reason and the model with an N-terminus separate from the TM region was selected (model 3; **Figure 1**). Moreover, the CPHmodels-3.2 and 3D-JIGSAWv3.0 servers were used to build 3D models of SV2A to confirm reproducibility. Both servers generated a structure quite similar to the I-Tasser model, predicting 12 TM regions.

#### Molecular Dynamics Simulations

The selected SV2A model was inserted into a POPC membrane and a MDS was performed to obtain a highly reliable 3D structure. **Figure 2** shows the simulation system after structural minimization (**Figure 2A**) and after 50 ns of MDS (**Figure 2B**). Notice the rearrangement of the SV2A protein, which indicates that conformational changes occur in the extra and intra-membrane domains, being greatest in loop structure components. The MDS allowed SV2A to re-accommodate within the membrane region, producing a more reliable system to perform docking studies of known ligands.

**Figure 3** shows the RMSD values for α-carbons of the whole backbone (**Figure 3A**) or α-carbons of the TM backbone helical domains (**Figure 3B**). RMSD values typically help to visualize global structural changes to determine protein stability. The lower RMSD values (0.21 ± 0.04 nm) of the TM helical domains (**Figure 3B**), compared to the whole backbone (**Figure 3A**), indicate strong conformational changes in extra-membrane regions. The RMSD slope became modest at ∼20 ns and tended to zero beyond ∼30–35 ns. The RMSD values suggest that the system reaches equilibrium, as confirmed by the Rg values (**Figure 4**).

The Rg of the whole protein backbone did not show a clear trend throughout the 50 ns trajectory (**Figure 4A**); it oscillated around an average value of 3.43 ± 0.02 nm. However, the variation was much lower when only the backbone atoms of the TM helical domain were considered (**Figure 4B**), although the Rg values for this region tended to increase slightly. This could be due to the fact that SV2A is an integral membrane protein with fairly stable TM domains and flexible extra-membrane portions.

The number of intra-molecular (**Figure 5A**), protein–water (**Figure 5B**) and protein–lipid (**Figure 5C**) H-bonds was determined as a function of time, based on geometrical criteria (donor–acceptor distance cutoff <3.5 Å and donor-H-acceptor angle <30◦). Changes in inter-helical hydrogen bonding are associated with the conformational dynamics of membrane proteins (Bondar and White, 2012). Most H-bonds corresponded to protein–solvent interactions, while only a few were formed between SV2A and the lipid heads. There was significant variation in the number of intra-molecular H-bonds that may correspond to a conformational change; however, this number reached an equilibrium value during the last part of the trajectory.

To obtain more structural details about the behavior of the protein, the RMSF of the α-carbon positions throughout the whole trajectory were calculated. These values indicated that the most flexible parts of SV2A were the loops located outside the membrane; i.e., the N-terminus, the large cytoplasmic loop between the sixth-seventh TM regions and the large intra-luminal loop between the seventh-eighth TM domains (residues 476–594; **Figure 6**).

On the other hand, protein solvent contacts were studied using SAS, which is a measure of the protein surface exposed to the solvent. The total SAS and the contributions of hydrophobic or hydrophilic groups to the SAS were determined as a function of time (**Figure 7A**). No significant changes were observed along the trajectory. The SAS of the protein exposed to water decreased slightly (**Figure 7B**), whereas the SAS of the protein exposed to lipids increased a little (**Figure 7C**).

Solvent accessible surface area by residues exposed to the membrane hydrophobic core helps to evidence the localization of several amino acids. **Figure 8** shows the total area exposed to the membrane by residues. Some residues in the loops, Nterminus and ATP binding motif interacted with the membrane. The TM helical domains can be determined, since the highest SAS values correspond to residues totally exposed to the membrane hydrophobic core (**Figure 8**).

In principal components analysis, the α-carbon matrix of SV2A shows a positive correlation between atomic motions of the large extra-vesicular loop and N-terminal segment (**Figure 9A**). It revealed that only 2–3 components are relevant for the trajectory. The projection of the trajectory onto the two principal eigenvectors shows a structure that reached equilibrium in the last portion of the trajectory; illustrated by denser yellow–red regions (**Figure 9B**).

The results of the replicate MDS were, in general, in agreement with the data obtained in the first MDS (data not shown). Although details varied, all MDS parameters behaved in a similar manner. In particular, the TM domains arrived at a conformational equilibrium; whereas the extra-membrane components remained mobile.

# Docking Studies

The values for the -G of binding between SV2A and the racetams varied between -5.26 and -8.82 Kcal/mol throughout the last 16 ns of the MDS trajectory (**Table 1**). The results showed the same affinity pattern in all evaluated conformations: UCB30889 > BRIV > SEL > LEV; which is in agreement with experimental SV2A binding data (**Table 2**; Noyer et al., 1995; Gillard et al., 2003, 2011; Bennett et al., 2007).

Ligand binding in dockings showed a clear preference for the following residues: T456, S665, W666, D670 and L689. These residues are within TM domains 7, 10 and 11. According to the SAS analysis these amino acids show values of zero or very close, which means that they are totally exposed to the membrane hydrophilic core (i.e., the pore or channel). Contacts between SV2A and the racetams were dominated by hydrophobic interactions and H-bonds. Every racetam showed a particular pattern of interactions when bound to SV2A; even if they shared some residues, the interaction was not the same in all cases. **Figure 10** shows schematic representations of the interactions between the last SV2A conformer (50 ns) of the MDS and LEV (**Figure 10A**), BRIV (**Figure 10B**), SEL (**Figure 10C**) or UCB30889 (**Figure 10D**).

# Discussion

Trans-membrane proteins are difficult to crystallize; thus computational techniques such as protein folding using homology modeling and MDS have been very useful tools to obtain information about their structural properties. The I-Tasser server was used to generate a full-length 3D model of SV2A with adequate structural predictions and acceptable quantitative assessments (Zhang and Skolnick, 2004; Zhang, 2008; Nugent and Jones, 2012). Modeling the extra-membrane loops was much more challenging than the TM regions, primarily due to sparse contacts or poor predictions of them; additionally, inherent loop flexibility imposed extra difficulties. Nevertheless, I-Tasser modeled the N-terminus and the two long loops quite similar to the secondary structure predicted by UniProt (2014). It is clear, however, that more sophisticated strategies will be required to overcome a number of current limitations in this area (Nugent and Jones, 2012).

FIGURE 6 | Graphic RMSF representations. (A) RMSF of the atomic positions and (B) representation of protein flexibility as a gradient from low to high fluctuations.

Although the obtained structure was acceptable, it was further refined by running a MDS. It showed the time dependent behavior of SV2A and provided detailed information on its fluctuations and conformational changes. The RMSD results showed that SV2A reached equilibrium in the last 15 ns of the MDS and this was in agreement with the Rg values. This was also true for the number of H-bonds, since they remained constant during the last part of the MDS. The results showed that the TM domains reached equilibrium and were likely to be correctly modeled; whereas the regions outside the membrane still remained mobile (also shown by the RMSF values). This is a general occurrence for TM proteins (Nugent and Jones, 2012) and it has been reported for other proteins, such as Bcl-2 (Ilizaliturri-Flores et al., 2013).

Extra-membrane domains, such as the long intra-luminal loops and N-terminus of SV2A, have post-translational modifications and interactions with other molecules, which might stabilize the protein conformation in natural conditions (Schivell et al., 1996, 2005; Nowack et al., 2010). The lack of post-translational changes and contacts with other molecules (which were not modeled due to their added complexity) may contribute to the extensive loop motion. Moreover, the extra-membrane regions of SV2A were evocative of intrinsically disordered domains within other proteins, such as Bcl-2 (Rautureau et al., 2010). All of the above, plus difficulties associated to proteins embedded in the lipid bilayer, may explain why SV2A has not been crystallized (Hipolito et al., 2014). The relevance of post-translational modifications on the conformational stability could be explored in future works with new, longer MD simulations. It is important to note that the racetam binding site is not part of the extramembrane regions and is within the pore, a region in equilibrium for at least the last 15 ns of the MDS; thus, our model proves to be of acceptable quality.

Synaptic vesicle protein 2A underwent amino acids rearrangements, mostly in the loop structure components. **Figure 2** and the SAS analysis showed an increase in the protein portion exposed to lipids; suggesting that amino acids initially exposed to the aqueous solvent were moving into the lipid bilayer. It is likely that they required a more hydrophobic environment or more favorable interactions with the lipid heads. In principal components analysis, the variance–covariance matrix of atomic motions showed which atoms moved together in a concerted way. It confirmed the conformational change that occurred in the large extra-vesicular loop and N-terminus, which were the most difficult regions to model and also the most flexible parts of the protein. The projection of the trajectory onto the two principal eigenvectors indicated that the protein experienced serious structural evolution during the first 15 ns of the MDS; however, its structure reached equilibrium in the last part of the trajectory.

We performed docking studies to explore the recognition pattern of racetams in SV2A. Docking studies can be performed on protein structures obtained from homology modeling, which maintains the backbone conformation (Tovchigrechko et al., 2002). However, the SV2A 3D structure obtained from I-Tasser was refined by performing a MDS; the advantage was that the side chain residues were energetically accommodated to their neighbors, resulting in several different conformers. Since the docking procedure lacks protein flexibility, we used the protein conformers from the MDS to obtain more reliable data, as has been published elsewhere (Sixto-Lopez et al., 2014). The MDS and docking simulations were thus combined, since using only



one conformation may omit several structural properties and result in non-biological dockings (Ramirez-Duran et al., 2013). Our docking studies suggested a consensus binding site for racetam ligands within SV2A constituted by five residues: T456, S665, W666, D670 and L689. Tryptophan 666 directly engages the racetams, mostly by electrostatic interactions. The SV2A protein with a mutated W666 did not restore synaptic depression in neurons lacking SV2A and SV2B; showing that this residue is vital for SV2A neurotransmitter function (Nowack et al., 2010). Additionally, binding of UCB30889 to SV2A is altered when this amino acid is mutated, indicating that it may participate in ligand recognition (Shi et al., 2011).

The SV2A protein has been proposed to be a transporter owing to its high degree of homology with other mayor facilitator super-family (MFS) carriers (Bajjalieh et al., 1992; Feany et al., 1992; Gingrich et al., 1992) and its compact funnel-shaped

structure with a visible indentation in the center indicative of a pore opening (Lynch et al., 2008). Tryptophan 666 is conserved in all SV2 isoforms and is homologous to a tryptophan in the 10th TM domain of MFS transporters which is vital for their activity (Nowack et al., 2010; Shi et al., 2011). Thus, it may provide the necessary hydrophobic milieu for transport of an endogenous substrate or LEV, in addition to its involvement in racetam binding. Residues Y461 and Y462 also frequently interacted with the racetams (**Table 1**); both are homologous to functional amino acids of MFS transporters (Shi et al., 2011). Additionally, Y461 is homologous to a residue important for chloride carrier activity (Bostick and Berkowitz, 2004).

It has been shown that 14 TM residues alter binding of UCB-30889 to SV2A when mutated (Shi et al., 2011). The present results coincide in the importance of W666 for ligand binding. The other four residues reported here were not mutated in the previous study. Y462 was also repeatedly found in this study;

TABLE 2 | Mean *-*G of binding during the MD trajectory and experimental SV2A binding data (see text for references).


however, the 12 other amino acids described in the prior work were not regularly detected. These discrepancies may be due to the following: (1) SV2A was transiently expressed in COS-7 cells which are far removed from neurons, (2) the true functionality of SV2A mutants was unclear since they were not likely to exit the endoplasmic reticulum or Golgi (Nowack et al., 2010), (3) the residues that altered binding to UCB-30889 were not necessarily part of the ligand pocket; they may affect other aspects, such as changing the native conformation or stability of the protein. This, in turn, may change important aspects, such racetam access, retention or expulsion. These phenomena are not modeled by docking studies, which only predict ligand affinity for a region of the protein. A MDS of SV2A together with the racetams would be useful to further refine the interactions of SV2A and its racetam ligands.

The 3D model generated by I-Tasser and the structure obtained after the MDS show that the TM portions of SV2A may be accessed by the racetams. As previously suggested, LEV and related racetams likely enter into recycling vesicles, gain access to SV2A from its luminal side and bind in its TM segments (Meehan et al., 2011, 2012; Mendoza-Torreblanca et al., 2013). Once bound to SV2A, racetams inhibit its function; reducing the ready releasable vesicular pool and synaptic transmission, preferentially during long high frequency activity (Xu and Bajjalieh, 2001).

Modeling of the complete SV2A protein was challenging due to its large size (742 amino acids), its long N-terminus and its two long loops. However, a valid 3D model was obtained and the 50 ns of MDS were enough to reach structural equilibrium of the protein. The racetam binding site in SV2A was identified by docking studies employing different SV2A snapshots to explore the protein motion involved in ligand recognition; it showed that five residues participate in ligand recognition. Given the biological and pharmacological importance of SV2A, knowing its conformation is essential to direct further experimental work in order to elucidate its function and attempt the discovery of new ligands. Additionally, knowing the racetam binding site within SV2A should facilitate synthesizing suitable radio-ligands for PET studies with a high affinity and specificity. This, in turn, should allow the *in vivo* evaluation of treatment response and possibly disease progression.

# Author Contributions

All authors contributed to the written manuscript. All authors have given approval to the final version of the manuscript.

# Acknowledgments

We are grateful to MD.USe Innovations for their collaboration with the SV2A MDS, Sara Navarrete Hernández for her administrative assistance and Julio C. López Dávila for his support in the implementation and use of GROMACS. The research was supported by the following grants: Consejo Nacional de Ciencia y

# References


Tecnología 180069 and 132353; Instituto Nacional de Pediatría 111/2012; Instituto de Ciencia y Tecnología del Distrito Federal PIRIVE09-9; Secretaria de Investigación y Posgrado 20150322; Estimulo al Desempeño de los Investigadores and Comisión de Operación y Fomento de Actividades Académicas del Instituto Politécnico Nacional; Programa Iberoamericano de Ciencia y Tecnología para el Desarrollo.

brain: relationship to anti-convulsant properties. *Eur. J. Pharmacol.* 664, 36–44. doi: 10.1016/j.ejphar.2011.04.064


the use of Protein Tomography. *Biochem. Biophys. Res. Commun.* 375, 491–495. doi: 10.1016/j.bbrc.2008.07.145


**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.

*Copyright © 2015 Correa-Basurto, Cuevas-Hernández, Phillips-Farfán, Martínez-Archundia, Romo-Mancillas, Ramírez-Salinas, Pérez-González, Trujillo-Ferrara and Mendoza-Torreblanca. 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.*

# Mitomycin-treated undifferentiated embryonic stem cells as a safe and effective therapeutic strategy in a mouse model of Parkinson's disease

Mariana Acquarone<sup>1</sup> , Thiago M. de Melo<sup>1</sup> , Fernanda Meireles <sup>2</sup> , Jordano Brito-Moreira<sup>3</sup> , Gabriel Oliveira<sup>4</sup> , Sergio T. Ferreira<sup>3</sup> , Newton G. Castro<sup>1</sup> , Fernanda Tovar-Moll 1,2 , Jean-Christophe Houzel <sup>1</sup> \* and Stevens K. Rehen1,2 \*

1 Institute of Biomedical Sciences, Federal University of Rio de Janeiro, Rio de Janeiro, Brazil, <sup>2</sup> D'Or Institute for Research and Education (IDOR), Rio de Janeiro, Brazil, <sup>3</sup> Institute of Medical Biochemistry Leopoldo de Meis, Federal University of Rio de Janeiro, Rio de Janeiro, Brazil, <sup>4</sup> Oswaldo Cruz Institute, Oswaldo Cruz Foundation, Rio de Janeiro, Brazil

#### Edited by:

Marco Antonio Meraz-Ríos, Centro de Investigación Y de Estudios Avanzados, Mexico

#### Reviewed by:

Afsaneh Gaillard, University of Poitiers, France Suzanne Earlene Peterson, The Scripps Research Institute, USA

#### \*Correspondence:

Jean-Christophe Houzel and Stevens K. Rehen, Instituto de Ciências Biomédicas, Cidade Universitária, Ilha do Fundão, Av. Carlos Chagas Filho, 373 – bloco F, Rio de Janeiro, RJ – CEP 21941-920, Brazil Tel: 55 21 39386390 jchouzel@icb.ufrj.br; srehen@lance-ufrj.org

> Received: 26 December 2014 Paper pending published: 04 February 2015 Accepted: 04 March 2015 Published: 08 April 2015

#### Citation:

Acquarone M, de Melo TM, Meireles F, Brito-Moreira J, Oliveira G, Ferreira ST, Castro NG, Tovar-Moll F, Houzel J-C and Rehen SK (2015) Mitomycin-treated undifferentiated embryonic stem cells as a safe and effective therapeutic strategy in a mouse model of Parkinson's disease. Front. Cell. Neurosci. 9:97. doi: 10.3389/fncel.2015.00097 Parkinson's disease (PD) is an incurable progressive neurodegenerative disorder. Clinical presentation of PD stems largely from the loss of dopaminergic neurons in the nigrostriatal dopaminergic pathway, motivating experimental strategies of replacement based on cell therapy. Transplantation of dopaminergic neurons derived from embryonic stem cells significantly improves motor functions in rodent and non-human primate models of PD. However, protocols to generate dopaminergic neurons from embryonic stem cells generally meet with low efficacy and high risk of teratoma formation upon transplantation. To address these issues, we have pre-treated undifferentiated mouse embryonic stem cells (mESCs) with the DNA alkylating agent mitomycin C (MMC) before transplantation. MMC treatment of cultures prevented tumorigenesis in a 12 week follow-up after mESCs were injected in nude mice. In 6-OH-dopamine-lesioned mice, intrastriatal injection of MMC-treated mESCs markedly improved motor function without tumor formation for as long as 15 months. Furthermore, we show that halting mitotic activity of undifferentiated mESCs induces a four-fold increase in dopamine release following in vitro differentiation. Our findings indicate that treating mESCs with MMC prior to intrastriatal transplant is an effective to strategy that could be further investigated as a novel alternative for treatment of PD.

Keywords: embryonic stem cell, stem cell therapy, Parkinson model, mitomycin C, contrast-enhanced MRI

# Introduction

Parkinson's disease (PD) is an incurable progressive neurodegenerative disorder with devastating consequences for patients and their families (Paulsen et al., 2013). Both genetic and environmental elements play a role in its etiology, with aging as the major risk factor. As the elderly population rises around the world, the number of affected individuals increases accordingly (Lees et al., 2009). Pharmacological treatments provide adequate symptomatic relief for many patients in the short term; however, severe adverse effects develop over time and drugs fail to prevent disease progression (Obeso et al., 2010).

Clinical presentation of PD stems largely from the loss of dopaminergic neurons in the nigrostriatal pathway. Many experimental strategies have therefore aimed to replace these cells. In humans, transplantation of dopamine (DA) neurons derived from fetal ventral mesencephalon has obtained clinical success (Piccini et al., 1999; Freed et al., 2001), but use of such neurons remains ethically controversial, besides the fact that resources are limited.

Mesenchymal stem cells have been suggested as a potential alternative source to fetal neurons (Khoo et al., 2011; White, 2011). However, they fail to readily differentiate into dopaminergic neurons, which has diminished the enthusiasm for using such cells for PD therapy. Embryonic stem cells retain the capacity to differentiate into DA neurons, both in vitro and in vivo. Importantly, transplantation of embryonic stem cell-derived DA neurons significantly improves motor functions in rodent and non-human primate models of PD (Deacon et al., 1998; Bjorklund et al., 2002; Muramatsu et al., 2009; Arenas, 2010; Kriks et al., 2011).

Several protocols have been described for differentiation of embryonic stem cells into dopaminergic neurons, albeit with low efficacy (Kirkeby et al., 2012; Sundberg et al., 2013). In addition, the resulting cell population often contains serotoninergic neurons, as well as residual undifferentiated embryonic cells that have a strong potential to give rise to teratomas, thereby limiting the clinical applicability of such cultures for transplantation into patients (Kriks and Studer, 2009; Chung et al., 2011; Lindvall, 2012). Moreover, published protocols are generally expensive and time consuming, and fail to yield large amounts of differentiated neurons, as required for clinical trials. Thus, there is an urgent need for alternative strategies that are safe and capable of generating purified embryonic stem cell-derived DA neurons on a large scale.

In order to minimize technical issues and eliminate the tumorigenic risk of transplanted cells, we tested for the first time the use of fully undifferentiated embryonic stem cells treated with the DNA alkylating and crosslinking agent mitomycin C (MMC) as an alternative source for cell transplant in a PD model. MMC is an FDA (2002)-approved chemotherapeutic agent for pancreatic and gastric adenocarcinoma that has also been tested in combination with other drugs in a wide variety of solid tumors (Akhlaghpoor et al., 2012; Werner et al., 2013). Our findings indicate that transplantation of undifferentiated embryonic stem cells pre-treated with MMC is a safe and effective strategy, and should be further investigated as a novel alternative for treatment of PD.

# Materials and Methods

#### Animals

All experiments were performed in accordance with International Guidelines for the Care and Use of Mammals in Neuroscience and Behavioral Research, and all procedures were approved by the University Committee on Ethics of Animal Use (protocol #DAHEICB060 for meningeal cell primary culture and for stem cell transplantation, and #DAHEICB027 for teratoma assay and 6-OHDA lesion). Mice were housed in self-ventilated cages (6 per cage) under standard laboratory conditions (12 h light/dark cycle, food and water ad libitum).

#### Embryonic Stem Cell Culture

USP1 mouse embryonic stem cells (mESCs) transfected with enhanced green fluorescent protein (eGFP) with retroviral vector (Paulsen et al., 2011) were maintained on a cell-free murine embryonic fibroblast-based substrate to avoid fibroblast contamination between cell passages (Stelling et al., 2013). Cells were maintained in D-MEM/F12 (Gibco) supplemented with 15% fetal bovine serum (JRH Biosciences), 2 mM glutamine (Gibco), 0.1 mM non-essential amino acids (Gibco), 0.1 mM β-mercaptoethanol (Gibco), 40 µg/mL gentamicin sulfate (Schering-Plough), and 0.2% of conditioned medium of CHO (Chinese hamster ovary) cells producing leukemia inhibitory factor (LIF).

#### Mitomycin C Treatment and Apoptosis Assay

For mitotic inactivation, undifferentiated mESC colonies were treated for 12 h with different concentrations of MMC (1, 2.5, 5, or 10 µg/mL). Immediately after MMC treatment, apoptosis was measured by the number of caspase 3 positive cells by flow cytometry. After incubation with MMC, mESCs were washed three times with D-MEM/F12, dissociated from the substrate into a single-cell suspension with 0.25% trypsin-EDTA (Gibco-Invitrogen) for 2 min, fixed with 4% paraformaldehyde in PBS for 20 min, and permeabilized with 0.3% Triton X-100 for 5 min. Cells were then blocked with 5% bovine serum albumin (BSA) solution for 30 min and incubated for 3 h with primary antimouse antibody against the active (cleaved) form of caspase 3 (Millipore), 1:400 in 1% BSA solution. Antibody-antigen reaction was revealed by a secondary antibody conjugated to Alexa Fluor 546, 1:1000 (Invitrogen). Acquisition was performed using a fluorescence-activated cell sorting (FACS) Calibur flow cytometer (Becton-Dickinson, USA) and data analysis was performed using WinMDI 2.8 software. A total of 100,000 events were acquired from three independent experiments for each tested MMC concentration.

# In Vitro Neural Differentiation Over Meningeal Cells

Dopaminergic differentiation was induced as described in Hayashi et al. (2008). Meningeal cell cultures were prepared from postnatal day 0 C57BL/6 mice. Briefly, the meninges were dissected from the calvaria and cultivated in α-minimum essential medium (MEMµ; Gibco-Invitrogen) containing 10% fetal bovine serum (JRH Biosciences), penicillin G (40 U/mL), and streptomycin (50 µg/mL). After the first passage, meningeal cells were allowed to reach confluence and were used as a feeder layer for mESCs. Mitomycin-treated and non-treated mESC colonies were dissociated with 0.25% trypsin-EDTA (Gibco-Invitrogen) for 2 min and plated on confluent meningeal layers (200 cells/cm<sup>2</sup> ). Co-cultures were maintained for 14 days in differentiation medium: G-MEM (Gibco-Invitrogen) supplemented with 5% knockout serum replacement (Gibco-Invitrogen), 2 mM glutamine, 0.1 mM non-essential amino acids, 1 mM pyruvate, and 0.1 mM β-mercaptoethanol.

# Electrophysiological Recordings

Voltage-clamp recordings were made from neuron-like cells 14 days after co-culturing control or mitomycin-treated mESCs eGFP positive cells on top of the meningeal cell layer. Whole-cell currents were recorded through borosilicate glass microelectrodes (WPI, USA) prepared on a P-97 horizontal puller (Sutter Instruments, USA). Current signals were acquired with an EPC-7 (HEKA, Germany) amplifier, low-pass filtered at 1 kHz, and digitized at 10 kHz with a LabMaster interface under the control of pClamp software (Axon Instruments, USA). Extracellular recording solution contained (in mM): 165 NaCl, 5 KCl, 2 CaCl2, 10 dextrose, 5 HEPES, 2 NaOH, pH 7.35. Cells were perfused at the rate of 1 mL/min at room temperature (23◦C) throughout the recordings. The microelectrode (intracellular) solution included (in mM): 80 CsCl, 80 CsF, 10 EGTA, 10 HEPES, and 26 CsOH, pH 7.30. Filled patch microelectrodes had resistances ranging from 2.7 to 7.4 MΩ when measured in the bath and a −7 mV liquid junction potential was added to the reported clamp potentials. Membrane potential was held at −70 mV and, approximately 2 min after achieving the wholecell configuration, voltage-sensitive currents were evoked by 200 ms depolarizing square pulses ascending in 10 mV steps from −60 to +60 mV, preceded by a 10 ms hyperpolarization to −90 mV. Leak-subtracted current traces were obtained by the fractional method (P/4) using four scaled hyperpolarizing subpulses. Data were analyzed using Clampfit 9 software (Axon Instruments, USA).

# Immunofluorescence for Dopaminergic Neurons

Following 2 weeks of neuronal differentiation, co-cultures were fixed with 4% paraformaldehyde in PBS for 20 min, permeabilized with 0.5% Triton X-100 for 5 min, blocked with 5% BSA for 60 min, and incubated overnight at 4◦C with the primary antibodies mouse β-tubulin III (1:400, Sigma), or rabbit anti-tyrosine hydroxylase (1:500, Millipore). Antibodyantigen reaction was visualized using anti-rabbit Alexa-594- or anti-mouse Alexa-594-coupled secondary antibodies (1:1000, Invitrogen). Cell cultures were analyzed and images were captured using a TCS SP5 laser confocal microscope (Leica Microsystems).

# Dopamine Release Assay

The amount of dopamine released spontaneously by mESC cells into the conditioned medium for 48 h, or after stimulation by elevated KCl solution, was measured by reverse phase chromatography coupled with electrochemical detection (0.5 V), as previously described (Arita et al., 2002). Cells were washed twice in a low KCl solution (in mM: HEPES-NaOH, 20, pH 7.4; NaCl, 140; KCl, 4.7; CaCl2, 2.5; MgSO4,1.2; KH2PO4, 1.2; glucose, 11) and incubated for 2 min in the same solution. Cells were then incubated for 15 min in a high KCl solution (in mM: HEPES-NaOH, 20, pH 7.4; NaCl, 85; KCl, 60; CaCl2, 2.5; MgSO4, 1.2; KH2PO4, 1.2; glucose, 11) to promote membrane depolarization. Briefly, 2 mL aliquots of culture medium and 0.8 mL of low or high KCl solutions, from control or mitomycin-treated differentiated mESCs, were first submitted to the following purification steps: 50 mg alumina (Al2O5) was weighed out in centrifuge tubes and the samples were added to 1 mL Tris-buffer, pH 8.0, 103 mM EDTA, plus 3 µL of 1 mM dihydroxybenzylamine (DHBA, internal standard). The suspension was mixed for 10 min at 4◦C, protected from light. Precipitated alumina was washed three times with 1 mL of ultrapure water and dopamine was eluted with 400 mL of 100 mM perchloric acid after 3 min of vortex agitation. After centrifugation for 3 min at 1,000 × g, 100 µL of medium and supernatant were analyzed. Isocratic separation was obtained using a C18 reverse phase column (Supelco 4.6 × 250 mm, Sigma Aldrich) eluted with the following mobile phase: 20 mM sodium dibasic phosphate, 20 mM citric acid, pH 2.64, containing 10% methanol, 0.12 mM Na2EDTA, and 566 mg/L heptanesulfonic acid. Total time for sample analysis was 30 min. Dopamine concentration was expressed as pg/g cell protein. Pierce BCA Protein Assay Kit was used for total protein quantification.

# Teratoma Assay

To confirm the pluripotency of mESCs and the effectiveness of the mitotic inactivation, mitomycin-treated or non-treated undifferentiated mESCs (500,000 cells) were injected into the thigh muscles of two-month-old nude immunodeficient mice (six animals). After mitomycin treatment, mESC colonies were washed three times with D-MEM/F12, dissociated from the substrate into a single-cell suspension with 0.25% trypsin-EDTA (Gibco-Invitrogen) for 2 min, and suspended in 400 µL of D-MEM/F-12. Using 27 gauge needles (Hamilton), the left upper thigh was injected with control mESCs (non-treated) and the right thigh with mitomycin-treated mESCs (500,000 cells in 2 µL). Animals were continuously monitored. Analgesic was added to the water bottle in order to minimize pain and discomfort (ibuprofen suspension, Abbott, 164 mg/L of drinking water). Twelve weeks after injection, they were anesthetized and submitted to magnetic resonance imaging (MRI, see below), then sacrificed to allow for muscle dissection. Muscles from both hindlimbs were fixed with 5% formalin, embedded in paraffin, and stained with hematoxylin and eosin (HE) for histopathological evaluation.

# Magnetic Resonance Imaging of Teratomas and Transplanted Animals

MRI was performed every 2 weeks after intrastriatal cell injections in all experimental groups. Mice were premedicated with atropine (0.2 mg/kg, s.c.) to reduce vagal tone and then anesthetized with an intraperitoneal injection of ketamine (90--120 mg/kg) and xylazine (10 mg/kg). Images were acquired in a 7-T magnetic resonance scanner (MRI System 7T/210 ASR Horizontal Bore Magnet, Agilent Technologies). Brain images were obtained using proton density (TR/TE: 10/2000 ms; matrix: 128 × 128; slice thickness: 1 mm; no gap, 12 averages), T1-weighted (TE/TR: 15/250 ms; matrix: 128 × 128; slice thickness: 1 mm; no gap, 16 averages), and T2-weighted (TE/TR: 15/2563 ms; matrix: 128 × 128; slice thickness: 1 mm; no gap, 14 averages) sequences in the axial, coronal, and sagittal planes, before and after gadolinium injection (0.2 mL/Kg i.p.). Hindlimb images were obtained using proton density (TR/TE: 10/2000 ms; matrix: 128 × 128; slice thickness: 1 mm; no gap, 16 averages), T1-weighted (TE/TR: 15/800 ms; matrix: 128 × 128; slice thickness: 1 mm; no gap, 5 averages), and T2-weighted (TE/TR: 15/2563 ms; matrix: 128 × 128; slice thickness: 1 mm; no gap, 14 averages sequences) in the axial, coronal, and sagittal planes, before and after injection of 0.2 mL/kg gadolinium (Mesentier-Louro et al., 2014).

Prior to image analysis, datasets were anonymized and randomized across groups. Brain and hindlimb morphology and tumor characteristics were evaluated in blinded fashion by two experienced researchers and compared across groups. For each dataset, all images were visually inspected for artifacts. Data processing was performed using Osirix Software.<sup>1</sup>

#### 6-OHDA-Induced Lesion and Transplantation of mESCs

Neurotoxic injections into the dorsal striatum were carried out according to a previously described protocol (da Conceição et al., 2010). Briefly, after premedicated with atropine (0.2 mg/kg, s.c.) and anesthetized with ketamine (90--120 mg/kg, i.p.) and xylazine (10 mg/kg, i.p.), male Swiss mice received one unilateral injection (2 µL) of 10 µg 6-OHDA (Sigma-Aldrich) dissolved in 0.9% sterile NaCl containing 0.1% ascorbic acid, at the rate of 0.5 µL/min (injection coordinates with respect to bregma were: AP = +0.5 mm; L = −2.0 mm; DV = −3.0 mm). Following the lesion, animals were allowed to recover for 4 weeks prior to an initial behavioral assessment. After additional 4 weeks, mice were re-anesthetized and 500,000 mitomycin-treated or non-treated undifferentiated mESCs, suspended in 2 µL of D-MEM/F12, were transplanted at the same coordinates. Lesioncontrol animals were injected with culture medium only.

#### Behavioral Analysis

Mice from all experimental groups: lesion-control (n = 8), untreated mESCs (mESC, n = 8), and mitomycin-treated mESCs (MMC-mESC, n = 12), were behaviorally assessed every second week, until 12 weeks after cell transplantation. The behavioral paradigms employed were open field, cylinder, balance beam, grip strength, and apomorphine-induced rotation tests, tested in that order.

Spontaneous activity in the open field was evaluated with the video-tracking EthoVision XT6 system (Noldus Information Technology). After a habituation period, the mice were placed in a 60 × 40 cm arena and locomotor activity, defined as the total covered distance (cm), was evaluated for 5 min. The video was recorded with a camera placed 1.0 m above the observation arena.

Motor coordination deficits were measured by the number of missteps and the latency to cross a 50 cm long, 2.5 cm in diameter round wooden beam. All mice were pre-trained on the beam for 5 min in order to encourage rapid crossing without reversals or stopping. Immediately after pre-training, animals were assessed on the test beam.

Forelimb grip strength was assessed with a grip strength meter with 1 g resolution (Insight EFF305, Brazil). Briefly, the mouse was allowed to grab a bar attached to a force transducer as it was gently pulled by the tail horizontally away from the bar. Force values from three consecutive trials with 5 s intervals were averaged to determine grip strength for each mouse.

The cylinder test was performed as a measure of spontaneous forelimb use (Schallert et al., 2000). Animals were placed individually in a glass cylinder (11 cm diameter, 20 cm height) and examined for spontaneous forepaw contact with the cylinder wall. After each trial, the cylinder was thoroughly cleaned with 70% ethanol. No habituation was allowed until a total of 20 contacts were recorded per animal and only weightbearing wall contacts made by each forelimb on the cylinder wall were scored. Data are presented as the percentage of lesion/grafted forepaw contacts relative to the total number of contacts, such that symmetric paw use would be scored as 50% (10/20 contacts).

To investigate dysfunction in dopaminergic signaling, analysis of apomorphine-induced rotation was performed as described previously (da Conceição et al., 2010). Apomorphine (0.5 mg/kg) was injected subcutaneously in the scruff and mice were placed in an opaque cylinder 30 cm in diameter. After a 5 min period, full body rotations were manually recorded over a period of 30 min, and data were expressed as net full body turns per minute. Experimental groups were unknown for the observer.

#### Histological Analysis

Four months after cell transplantation, mice were deeply anesthetized and transcardially perfused with PBS followed by 4% PFA. Brains were post-fixed in 4% PFA for 24 h before immersion in a standard sucrose gradient (10%, 20%, and 30%) until they sank. Brains were then embedded in OCT medium (Tissue-Tek). Serial coronal 60 µm thick sections were cut frozen on a cryostat (Leica) and collected in PBS. Sections through the substantia nigra and striatum were processed for immunohistochemistry. Sections were washed with PBS, permeabilized in 1% Triton X-100 for 30 min, incubated for 2 h with 3% H2O<sup>2</sup> diluted in methanol, blocked with 10% BSA for 3 h, and incubated overnight at 4◦C with rabbit anti-TH primary antibody, 1:1000 (Millipore). After PBS washing, sections were incubated for 2 h with biotinylated anti-rabbit secondary antibody (Vector Labs, 1:200) and staining was developed using ABC (Vectastain kit) and DAB (Sigma) reactions. Neurons displaying a TH-positive cytoplasm were counted as positive, and plotted throughout the substantia nigra and the striatum using a motorized Zeiss Axioplan microscope and the Neurolucida system (MBF Bioscience).

#### Statistical Analysis

Numerical values and error bars are expressed as mean ± SEM and mean ± SD for in vitro and in vivo measurements, respectively. Tests for differences between means were performed using a standard software package (GraphPad Prism version 5.00), using a 5% significance level. In the ANOVA procedures, Dunnet and Tukey tests were used for group comparisons (**Figures 1A**, **4B–F**, respectively). For

<sup>1</sup>www.osirix-viewer.com

the apomorphine-induced rotations (**Figure 4A**), treatments were compared at each time point by a Bonferroni test of inter-subject differences in a repeated measures ANOVA procedure.

currents from two neuron-shaped mESCs pre-treated with MMC and

# Results

## Treatment with Mitomycin C Increased Dopaminergic Differentiation in Vitro

We first determined the optimal concentration of MMC by testing different concentrations in vitro and comparing cell death rates to those observed in control cultures. We found that 1 µg/mL MMC was the maximal concentration of mitomycin that did not induce apoptotic cell death compared to control, untreated mESC cultures. Cultures treated with 1 µg/mL MMC for 12 h were morphologically indistinguishable from untreated controls (**Figures 1A--5**).

respectively; Mann-Whitney's test, n = 6).

We next examined the effect of MMC pre-treatment (1 µg/mL) on the ability of mESCs to differentiate into neurons in vitro. After 14 days of co-culture with meningeal cells in conditions promoting dopaminergic differentiation, many MMC-treated mESCs developed neuron-like morphology, and were immunoreactive for β-tubulin III (**Figure 1B**) and tyrosine hydroxylase (**Figure 1B**). We first sought to evaluate neuronal differentiation by determining whole-cell sodium current densities. However, only one out of 29 MMC-treated and none of 45 untreated neuron-like mESCs showed inward voltage-activated currents (**Figure 1C**). We then analyzed the dopaminergic phenotype by measuring neurotransmitter release. Differentiated cultures that had been pre-treated with MMC spontaneously released four times more dopamine than

untreated (left) mESCs. In all six animals tested, the right limbs were normal and the left limbs displayed large tumors. (B) In vivo MRI scans, coronal plane, proton density sequence, showing teratoma formation in the left limb, as opposed to the right limb (which received mitomycin-treated mESCs) of the

region. (D) High resolution photomontage of hematoxylin-eosin-stained section through the leg injected with untreated mESCs, showing teratoma containing a mixture of tissue types, including: (E) neural tube-like, (F) cartilaginous, and (G) glandular-like (asterisks in (D) indicate enlarged regions in (E--G)). Scale bar: 1000 µm for (D); 100 µm for (E--G).

untreated cells (untreated: 4.16 ± 0.47 and MMC-treated: 16.59 ± 2.13 pg/µg protein over 48 h. **Figure 1D**). DA release evoked by high KCl also increased about five times after MMC treatment (untreated: 0.69 ± 0.10 and MMCtreated: 3.55 ± 1.50 pg/µg protein after 15 min incubation in 60 mM KCl. **Figure 1D**). These results show that, although the differentiating mESCs were not electrically mature at 14 days, 1 µg/mL MMC pre-treatment promoted a functional dopaminergic phenotype, increasing both spontaneous and stimulated neurotransmitter release.

# Pre-treatment with 1 µg/mL Mitomycin C Prevents Teratoma Formation by mESCs Grafted into Nude Mice

Grafts of undifferentiated mESCs are prone to develop into tumors, particularly teratomas, as expected from their pluripotency (Fong et al., 2010). Standard protocols to assess tumorigenesis consist of injecting cells into the sub-renal capsule, the liver, the myocardium, the brain, or into subcutaneous or intramuscular sites. We chose to administer intramuscular injections into the hindlimb muscles because this is a straightforward strategy known to be efficient for assessing pluripotency capacity, besides being less harmful to the animal. Additionally, this strategy allowed us to use each animal as its own control by injecting control mESCs or MMC-treated cells in each limb.

Twelve weeks after injection of 500,000 cells, all left hind paws that received control cells developed prominent tumors, revealing the proliferative capacity of mESCs (**Figure 2A**). In vivo MRI confirmed formation of unilateral heterogeneous tumors (**Figure 2B**). Tumors were also evidenced in transverse images before (**Figure 2C**) and after gadolinium-contrast enhancement (**Figure 2C**). Histopathological analysis (**Figure 2D**) revealed that all tumors were mature teratomas containing ectodermal- (**Figure 2E**) as well as endodermal- (**Figure 2F**) and mesodermalderived (**Figure 2G**) tissue, further confirming the pluripotency of non-MMC-treated mESCs. Remarkably, the same number of

for (F,G).

MMC-treated mESCs did not induce any tumor in the right paws of the same animals (n = 6) (**Figures 2A--C**). These results indicate that in vitro pre-incubation with 1 µg/mL MMC for 12 h completely blocked further mitotic activity of grafted mESCs in vivo.

animal transplanted with 500,000 mitomycin-treated mESCs for as long as

# Striatal Graft with Mitomycin C-Treated mESCs Restores Motor Function in Hemi-Parkinsonian Mice Without Teratoma Induction

To test for potentially beneficial effects of mESC transplantation into a dopamine-deficient hemisphere and to investigate whether MMC could prevent tumor formation in the living brain, we injected MMC-treated pluripotent mESCs into the striatum of mice that had been previously lesioned by 6-OHDA. Every other week, transplanted animals were weighed, underwent behavioral testing for motor function, and were submitted to MRI scans to check for possible formation of brain tumors and to assess blood--brain barrier integrity. As expected, unilateral striatal lesion consistently resulted in severe motor dysfunction, but was not lethal per se. However, animals that received intrastriatal transplants of 500,000 untreated mESCs died between 3 and 7 weeks later (**Figure 3A**). In all animals from this group, intracerebral tumors were revealed by 7 Tesla MRI scans as early as 05 days after mESC transplant (**Figure 3B**). Importantly, contrast in intracerebral tumor images was significantly enhanced after gadolinium injection, further indicating a disruption of the integrity of the blood--brain barrier in these animals (**Figure 3B**, right). Post-mortem histological analysis with HE staining of all 8 transplanted hemispheres confirmed the presence of solid tumors after transplantation with untreated mESCs. In addition, immunohistochemistry for TH (**Figures 3E--G**) and computer-microscope reconstruction

(**Figure 3D**) revealed a population of dopaminergic cell bodies restricted to the injected striatum (dots indicate cell bodies positive for TH-immunostaining). In contrast, all animals receiving MMC-treated mESCs survived until the end of the observation period of 12 weeks post-transplant (n = 12 mice; **Figure 3A**), and no tumors could be detected in contrast-enhanced MRI scans (**Figure 3C**) or through histopathological investigation. Four additional MMC-mESC transplanted mice were monitored for as long as 60 weeks with no signs of pathology.

All animals lesioned by unilateral intrastriatal 6-OHDA injection developed motor signs of dopamine deficiency that were stable after four weeks, as previously described (Iancu et al., 2005). At this time point, the rate of apomorphineinduced contralateral full-body rotations ranged from 15 to 20 rpm amongst all 6-OHDA lesioned animals. For ease of comparison, rotation rates for each experimental group were normalized relative to the average rate measured on the week of transplantation (week 0). 6-OHDA-lesioned mice injected with culture medium (6-OHDA + medium, n = 8) maintained a steady behavioral response to apomorphine, displaying 0.98 ± 0.19 relative rpm at 20 weeks after the lesion (**Figure 4A**). Transplantation with untreated mESCs (6-OHDA + mESCs, n = 8) led to a marked reduction in the rate of apomorphineinduced rotations, before generating brain tumors that eventually killed the animals. Mice transplanted with MMC-treated mESCs (6-OHDA + MMC-mESCs, n = 12) displayed a slower but steady decrease in the number of apomorphine-induced rotations, reaching 0.13 ± 0.11 relative rpm 12 weeks after transplant (**Figure 4A**). Already at 2 weeks after transplant, the difference relative to 6-OHDA + medium control animals was significant (p < 0.001, Bonferroni test in RM ANOVA).

Long-term motor improvements resulting from MMC-treated mESC were further assessed 12 weeks after transplantation using open field, grip strength, beam walking, and cylinder tests. To assess the extent of recovery from the initial lesion, behavioral scores were compared across 6-OHDA + medium control, 6-OHDA + MMC-treated mESC, and healthy, non-lesioned, non-transplanted, agedmatched mice (normal group, n = 8). Locomotor activity in

the open field arena (measured as the distance covered by the animal over 5 min) was reduced by 39% on average in 6-OHDA + medium animals compared to healthy normal mice (normal: 2415 ± 373 cm; 6-OHDA + medium: 1468 ± 187 cm), but only by 17% in 6-OHDA + MMC-treated mESC animals (2003 ± 196 cm; **Figure 4D**). In the beam walking test for fine motor coordination and balance, 6-OHDA + medium mice slipped approximately three times as often (6.5 ± 1.8 slips, **Figure 4B**) and took twice the time to cross the 50 cm beam (11.3 ± 0.8 s, **Figure 4C**) as did healthy normal mice (2.0 ± 1.1 slips and 5.9 ± 1.5 s crossing time, respectively). Animals injected with 6-OHDA + MMC-treated mESCs displayed a significant improvement, so that both measures approached normal levels 12 weeks after transplant (2.6 ± 0.9 slips and 7.6 ± 1.9 s crossing time). None of the animals tested fell while crossing the beam. Forelimb grip strength was markedly reduced in 6-OHDA + medium mice (46.8 ± 13.1 g) as compared with healthy normal mice (123.0 ± 21.0 g), but 6-OHDA + MMCmESC transplanted animals completely recovered muscle strength (146.9 ± 36.7 g; **Figure 4E**). Finally, 6-OHDA + MMCtreated mESC transplanted hemi-parkinsonian mice behaved almost like healthy mice regarding symmetric forelimb use in the cylinder test (normal: 66.3 ± 7.4% and MMC-mESC: 71.1 ± 13.6% double contacts, respectively; **Figure 4F**), whereas 6-OHDA + medium mice made significantly fewer symmetric contacts with the cylinder wall (16.1 ± 5.4%), indicating marked motor asymmetry. Our data show that MMC-treated mESC transplant induced considerable recovery of motor function in a mouse model of Parkinsonism, without any signs of tumor formation for up to 15 months following transplant.

# Discussion

In the present study, we show that transplantation of mESCs, pre-treated with the FDA-approved chemotherapeutic agent MMC, in the 6-OHDA-lesioned mouse model of Parkinson's disease resulted in significant improvement in motor function and reduced akinesia without tumor formation. Furthermore, we show that halting mitotic activity of undifferentiated mESCs induced a four-fold increase in the release of dopamine after in vitro differentiation (**Figure 5**).

A central concern related to the therapeutic use of pluripotent stem cells in PD is the eradication of non-dopaminergic neurons and, especially, of residual stem cells with tumorigenic potential (Sykova and Forostyak, 2013). Additionally, most available studies use time-consuming protocols requiring multiple steps for generation of a limited number of DA neurons in vitro (Lindvall, 2012). Also, it is not fully understood whether in vitro-derived neurons are able to maintain their functional properties after in vivo transplantation (Park et al., 2006). All these caveats limit translation of this approach to the clinical practice. As shown in this study, the use of fully undifferentiated mESCs may represent a novel alternative.

Undifferentiated mESCs are easier to maintain than differentiated cultures; they are less expensive to cultivate and easier to obtain on a large scale. We have recently developed a method to cultivate human embryonic stem cells on suspended beads combined with the use of stirred microcarrier systems, with an optimized xeno-free culture medium (Marinho et al., 2013). Using this protocol, it is possible to produce over 160 million pluripotent cells per week, providing a step forward towards therapeutic use of human embryonic stem cells.

Much attention has been given to the benefits of using autologous induced pluripotent stem cells (iPSCs) instead of embryonic stem cells in preclinical and clinical studies. However, Araki et al. (2013) have shown that iPSCs and embryonic stem cells share similar limited immunogenicity. Importantly, undifferentiated embryonic stem cells are less immunogenic than their differentiated derivatives (Bonde and Zavazava, 2006).

Recently, several protocols for cell purification have been proposed as alternatives to eliminate potential tumorigenic cells (Kahan et al., 2011; Sundberg et al., 2013) or to obtain highly enriched populations of dopaminergic neurons (Pruszak et al., 2007; Bernstein and Hyun, 2012), without the need for genetic manipulation of cells. Thus, magnetic (MACS) or fluorescence-activated cell sorting (FACS), associated with the use of antibodies against specific cell surface marker antigens (generally, clusters of differentiation markers), have been incorporated as important tools for transplantation studies. However, cell sorting procedures are not fully efficient, besides the fact that such multi-step techniques may compromise the viability of fragile neuronal cell types (Emre et al., 2010).

MMC is a DNA-alkylating agent that irreversibly blocks cell division, inducing arrest mainly in the S phase of the cycle (Witte and Bradke, 2008). Acting as a DNA crosslinking agent, MMC could potentially be genotoxic to stem cells. Vinoth et al. (2008) investigated how genotoxic stress caused by MMC affects human embryonic stem cells as compared to a somatic human fetal lung fibroblast cell line, and found a 10% increase in the number of aberrant somatic cells vs. a 5% increase in human embryonic stem cells, indicating that the latter are more resistant to MMCinduced damage.

Due to its cytotoxic activity, we tested the impact of MMC pre-treatment on the viability and differentiation potential of mESCs. Most mESCs pre-treated with MMC and co-cultured onto a feeder layer of meningeal cells for neuronal induction exhibited typical neuronal morphology and expressed β-tubulin class III and TH. Interestingly, pre-treated cells released five times more dopamine, as compared to non-treated ones. These results indicate that MMC treatment promotes mESC differentiation towards a functional dopaminergic phenotype. Using a different cell type and experimental approach, Felfly et al. (2011) showed that MMC alone trigger the expression of several differentiation genes in murine neural stem cells.

To investigate the intrinsic tumorigenic capacity of pluripotent stem cells, we tested if MMC treatment blocks teratoma formation. Results with nude mice showed that pre-treatment of cells with 1 µg/mL MMC for 12 hours suppressed the tumorigenic capacity of transplanted mESCs after least 12 weeks. Untreated mESCs transplanted in the opposite hindlimbs of the same animals generated large teratomas.

As might be expected, mice injected in the striatum with nontreated mESCs developed tumors and exhibited blood–brain barrier disruption, as assessed by periodic MRI scans, which may explain why none of those animals survived more than seven weeks after transplantation. Nonetheless, despite the presence of intracerebral teratomas, all animals that received non-treated mESCs performed better on the apomorphineinduced rotation test. We suggest that this effect may be due to the differentiation of transplanted mESCs into dopaminergic neurons in vivo. In support of this notion, Felfly et al. (2011) described grafts able to generate DA neurons, which were surrounded by glial cells from the host, suggesting that neural differentiation of transplanted cells may be influenced by glial signaling in vivo. Moreover, mesencephalic neuroepithelial stem cells differentiated into tyrosine hydroxylase-positive neurons 3 times more efficient when transplanted into the brain of PD brains than in controls, which could also be explained by the presence of environmental cues (Mine et al., 2009). Finally, Takagi et al. (2005) demonstrated that FGF20, preferentially expressed by cells within the substantia nigra, increases the differentiation of DA neurons from ESC-derived neurospheres in monkeys.

Transplantation of MMC-treated mESCs in the brain of 6-OHDA lesioned animals significantly reduced the rate of apomorphine-induced rotations and nearly normalized motor performance at 12 weeks post-transplant. However, the beneficial effect of transplantation was somewhat slower in these animals than in mice transplanted with non-treated cells. This lies in apparent contradiction with our in vitro data showing higher spontaneous and stimulated dopamine release from differentiated MMC-treated mESCs. Nevertheless, untreated mESCs obviously proliferated faster, maybe generating more dopaminergic cells than MMC-treated mESCs. It will be of interest to assess whether survival, differentiation, and tissue integration are directly affected by prior MMC treatment such as proliferation.

In conclusion, results presented here show that MMC treatment allowed pluripotent stem cells to restore motor function without forming tumors for as long as 15 months in mice, suggesting that MMC-treated undifferentiated embryonic stem cells should be further investigated as a safe and effective strategy for the treatment of Parkinson's disease.

# Acknowledgments

This work is part of the PhD Thesis of MA. We thank Dr. Liisa Tremere for comments on an earlier version of the manuscript. Funding agencies: FAPERJ, CNPq, BNDES, CAPES.

# References


cells in hemiparkinsonian rodents. PLoS One 6:e19025. doi: 10.1371/journal. pone.0019025


**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.

Copyright © 2015 Acquarone, de Melo, Meireles, Brito-Moreira, Oliveira, Ferreira, Castro, Tovar-Moll, Houzel and Rehen. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution and 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.

# Pleiotrophin as a central nervous system neuromodulator, evidences from the hippocampus

#### **Celia González-Castillo<sup>1</sup>† , Daniel Ortuño-Sahagún<sup>2</sup>† , Carolina Guzmán-Brambila<sup>3</sup> , Mercè Pallàs <sup>4</sup> and Argelia Esperanza Rojas-Mayorquín<sup>5</sup>\***

<sup>1</sup> Doctorado en Ciencias en Biología Molecular en Medicina (DCBMM), CUCS, Universidad de Guadalajara, Guadalajara, Jalisco, México

2 Instituto de Investigación en Ciencias Biomédicas (IICB), CUCS, Universidad de Guadalajara, Guadalajara, Jalisco, México

<sup>3</sup> Tecnológico de Monterrey, División de Biotecnología y Salud, Escuela de Medicina, Campus Guadalajara, Guadalajara, Jalisco, México

<sup>4</sup> Department of Pharmacology and Medical Chemistry, Faculty of Pharmacy School of Pharmacy, Institute of Biomedicine (IBUB), Centros de Investigación

Biomédica en Red de Enfermedades Neurodegenerativas (CIBERNED), University of Barcelona, Barcelona, Spain

<sup>5</sup> Departamento de Ciencias Ambientales, Instituto de Neurociencias, CUCBA, Universidad de Guadalajara, Guadalajara, Jalisco, México

#### **Edited by:**

Victoria Campos, Instituto Nacional De Neurologia Y Neurocirugia, México

#### **Reviewed by:**

Corette J. Wierenga, Utrecht University, Netherlands Takumi Takizawa, Gunma University, Japan

#### **\*Correspondence:**

Argelia Esperanza Rojas-Mayorquín, Departamento de Ciencias Ambientales, Instituto de Neurociencias, CUCBA, Universidad de Guadalajara, Francisco de Quevedo 180, Col. Arcos Vallarta, Guadalajara, 44130 Jalisco, México e-mail: argelia.rojas@cucba.udg.mx †These authors have contributed

equally to this work.

# **INTRODUCTION**

Pleiotrophin (PTN) is a secreted cell signaling cytokine that acts as growth factor associated with the extracellular matrix, which has recently started to come to the fore as a significant neuromodulator with multiple neuronal functions. PTN is an 18-KDa protein and has 168 amino acids. It was discovered practically simultaneously by several laboratories nearly 25 years ago; thus, it initially received several names as follows: HBGF-8 (Heparinbinding growth factor; Milner et al., 1989); HB-GAM (Heparinbinding growth-associated molecule; Rauvala, 1989; Merenmies and Rauvala, 1990); HBNF (Heparin-binding neutrophil factor; Kovesdi et al., 1990); OSF-1 (Osteoblast-specific factor 1; Tezuka et al., 1990), and HARP (Heparin affinity regulatory peptide; Courty et al., 1991).

PTN shares high homology (>50%) with another peptide, denominated Midkine (MK); both are highly conserved throughout evolution and are found in species ranging from *Drosophila* to humans (Kadomatsu and Muramatsu, 2004). This means that although both have many functions in common and participate in similar functions, they also possess more particular, specific, and non-redundant functions. It is evident when both are simultaneously knocked out in mice, they display severe abnormality phenotypes. However, when independently knocked out, PTN−/− and MDK−/− mice are far from being

Pleiotrophin (PTN) is a secreted growth factor, and also a cytokine, associated with the extracellular matrix, which has recently starting to attract attention as a significant neuromodulator with multiple neuronal functions during development. PTN is expressed in several tissues, where its signals are generally related with cell proliferation, growth, and differentiation by acting through different receptors. In Central Nervous System (CNS), PTN exerts post-developmental neurotrophic and -protective effects, and additionally has been involved in neurodegenerative diseases and neural disorders. Studies in Drosophila shed light on some aspects of the different levels of regulatory control of PTN invertebrate homologs. Specifically in hippocampus, recent evidence from PTN Knock-out (KO) mice involves PTN functioning in learning and memory. In this paper, we summarize, discuss, and contrast the most recent advances and results that lead to proposing a PTN as a neuromodulatory molecule in the CNS, particularly in hippocampus.

**Keywords: pleiotrophin, neuromodulation, hippocampus, neuropeptide, miple**

completely normal and exhibit moderate but different abnormalities (Muramatsu et al., 2006; Zou et al., 2006; Gramage and Herradón, 2010; Himburg et al., 2012; Vicente-Rodríguez et al., 2013), which denotes that although both peptides could present overlapping or similar functions, they are also clearly involved in different roles.

## **PTN COULD SIGNAL THROUGH A MULTI-RECEPTOR COMPLEX**

PTN signals are generally related with cell proliferation, growth and differentiation, but PTN has also has been involved in other functions by acting through different receptors (**Figure 1**). Mainly, PTN can bind and signal via Receptor protein tyrosine phosphatase ζ (RPTPζ), EC = 3.1.3.48 (Maeda et al., 1996, 1999; Meng et al., 2000), which is a transmembrane chondroitin sulfate proteoglycan present in two isoforms (shorter and full-length), which in turn also binds with various cell adhesion molecules (NrCAM, L1/Ng-CAM, contactin, N-CAM, and TAG1), growth factors (PTN, MK, and fibroblast growth factor (FGF-2), and extracellular matrix molecules (amphoterin, tenascin-C, and tenascin-R) (reviewed in Maeda et al., 2010). Under certain circumstances, PTN can act via Anaplastic Lymphoma Kinase (ALK) receptor (Stoica et al., 2001, 2002; Powers et al., 2002), although some evidences suggest that the action of PTN on ALK

could occur through its previous interaction with RPTPζ (Perez-Pinera et al., 2007). Additionally, PTN; (1) promote neurite outgrowth via N-syndecan receptor (Raulo et al., 1994) or via Neuroglycan-C (NGC; Nakanishi et al., 2010), (2) interact with integrin αυβ3 (alpha nu beta 3) receptor, which is a mechanosensitive cell membrane receptor, for cell adhesion (Mikelis et al., 2009), and (3) interact with Low-density lipoprotein (LDL) Receptor-related protein (LRP; Kadomatsu and Muramatsu, 2004). Additionally, two different species of PTN, PTN15 and PTN18, have been described (Lu et al., 2005), but their differential interaction or their affinities to different receptors has not yet been established, which adds another level of complexity to their physiological functioning.

It has been recently proposed that PTN signaling may function through a multi-receptor complex (Xu et al., 2014), combining the previously mentioned receptors, and most probably other adaptor proteins, which interact under certain circumstances inside particular cell membrane microdomains, probably also associated with lipids in raft configuration, which could explain the variety of functions in different tissues, in terms of the combinatorial analysis of the elements present at each time and place. Then, PTN action over previously mentioned receptors could in turn signal through different signal pathways (**Figure 1**). Increasing our knowledge of the intricate molecular mechanisms involved would clarify the receptor complexes and signaling pathways implicated, as well as advance the discovery of other molecules involved, which in turn will lead us to fully explain its variety of functions.

# **DIFFERENTIAL EXPRESSION OF PTN RECEPTORS DURING DEVELOPMENT AND IN ADULT COULD INDICATE ITS DISSIMILAR PARTICIPATION IN DIFFERENT FUNCTIONS**

Although during early development PTN expression is widely distributed in Central Nervous System (CNS; Li et al., 1990), expression of PTN in adult brain appears to be constitutive and apparently limited to only a few cell types in brain cortex, hippocampus, cerebellum and olfactory bulb (Wanaka et al., 1993; Lauri et al., 1996; Basille-Dugay et al., 2013), as well as in some striatal interneurons (Taravini et al., 2005). At these locations, the differential expression of its receptors could exist, which might partially explain its diverse actions. RPTPζ is expressed in glial cells as well as neurons. In hippocampal cells, it is located at the postsynaptic membrane of pyramidal neurons in adult (Hayashi et al., 2005), and its expression is modulated by spatial learning (Robles et al., 2003); therefore, it is involved in learning and long-term memory. Additionally, it is highly expressed following injury in areas of axonal sprouting and glial scarring (Snyder et al., 1996), and its expression is induced in inner molecular-layer astrocytes of the dentate gyrus of the sclerotic hippocampus in patients with epilepsy (Perosa et al., 2002). Also, it is involved in regulating dendritogenesis and synaptogenesis of hippocampal neurons *in vitro* (Asai et al., 2009). Therefore, PTN signaling through RPTPζ could be involved in modulation of hippocampal plasticity during learning and also during recovery after a lesion or in neuropathological situations, by modulating dendritogenesis and synaptogenesis. RPTP-ζ/β might be implicated in plastic rearrangements of nigrostriatal connections, such as sprouting of dopaminergic terminals or postsynaptic changes triggered by L-DOPA treatment in a model of Parkinson disease (Ferrario et al., 2008).

Likewise, ALK receptor is expressed in adult mammalian hippocampus and has also been implicated in neurogenesis, memory, and learning (Weiss et al., 2012). In addition, it has been involved in basal hippocampal progenitor proliferation and its deficiency induces alterations in behavioral tests (Bilsland et al., 2008). Although MK has been postulated to be the ligand for ALK receptor, at least in controlling sympathetic neurogenesis (Reiff et al., 2011), PTN also appears to be able to interact with this receptor (Stoica et al., 2001), although this remains controversial (Mathivet et al., 2007), and it appears that PTN performs its action on ALK thought its previous interaction with RPTPζ (Perez-Pinera et al., 2007).

It is clear that overlapping of PTN and MK activities can occur in some cases, but certainly not under all circumstances, as mistakenly suggested by Xu et al. (see Figure 3 in Xu et al., 2014). Although both peptides exhibit similar actions under certain physiological conditions, at least in the CNS, each also exerts diverse effects and performs different actions, depending on the cerebral region, as mentioned later.

# **STUDIES IN DROSOPHILA ENLIGHTEN SOME ASPECTS OF THE DIFFERENT LEVELS OF REGULATORY CONTROL OF PTN EXPRESSION AND FUNCTION IN THEIR INVERTEBRATE HOMOLOGS**

*Drosophila* homologs to MK and PTN are Miple1 and Miple2, with 20 and 24% identical to human MK and human PTN, respectively (Englund et al., 2006). However, they cannot be assigned as respective homologs, but only as members of the same family. Respective genes are arranged in tandem, suggesting that they have arisen as a result of a gene duplication event at some point of evolution. However, these secreted proteins are expressed in restricted, non-overlapping patterns, with Miple1 mainly expressed in the developing embryonic nervous system, while miple2 is strongly expressed in the developing gut endoderm (Englund et al., 2006). Therefore, had they been generated by gene duplication, they were clearly submitted to different selective pressure expression regulation, and consequently diverge in their expression pattern, and most probably in functioning. Thus, it will be relevant to elucidate, in useful model such as *Drosophila*, the molecular interactions of these peptides during complex developmental processes.

The messenger RNA (mRNA) 3<sup>0</sup> -Untranslated region (UTR) binding protein HOW (Held out wing) is able to posttranslationally repress *miple*, downregulating its mRNA levels in mesoderm in order to enable proper mesoderm spreading during early embryogenesis in *Drosophila* (Toledano-Katchalski et al., 2007). This suggests that a similar mechanism could drive some regulatory action over PTN and MK expression in vertebrates.

Another point of regulation corresponds to the interaction of miple, as a signaling peptide, with other proteins. For example, by affecting the affinity of HTL ligands to the HTL receptor (Heartless, a *Drosophila* FGF receptor), thereby modulating the strength of HTL-dependent signaling (Toledano-Katchalski et al., 2007). Thus, it is feasible that PTN could interact with other peptides being a key modulator in the binding process to different complexes of receptors.

Interestingly, the combined expression pattern of Miple1 and Miple2 complements the expression pattern of the *Drosophila* ALK homolog (DAlk; Lorén et al., 2001, 2003). However, its ligand has been identified as a different peptide, namely Jelly belly (Jeb), which play roles in neuromuscular junction growth and function, early mesoderm development, and also in axon targeting of photoreceptors (Weiss et al., 2001; Englund et al., 2003; Lee et al., 2003; Bazigou et al., 2007; Rohrbough and Broadie, 2010). It is relevant to mention that *Drosophila* Jeb is not able to activate mouse ALK (Yang et al., 2007), and Jeb homologs in vertebrates have not yet been described. However, it is noteworthy that secreted Jeb contains a LDL receptor class A domain that contains 6 disulphide-bound cysteines (Bieri et al., 1995), and could constitute a binding site for LDL and calcium (Yamamoto et al., 1984). Given that LRP is a LDL receptor-related protein involved in PTN action in vertebrates (Kadomatsu and Muramatsu, 2004), it would be possible that Jeb signaling could be related with miple signaling and their vertebrate counterpart is unveiled to date.

Based on all previous cited evidences, and given the complexity of the molecular interactions in which PTN is clearly involved, it will be necessary to widely divulge approaches for disclosing its functioning. In this respect, one of the most useful approaches could be analysis by microarrays of the gene profile expression in PTN-defective Knock-out (KO) mice. Recently, in our laboratory, we performed these experiments and established the differential gene expression in the hippocampus of PTN KO mice (In preparation).

# **DIFFERENTIAL EFFECTS OF PTN VS. MDK INDICATE IT AS A NEUROMODULATORY PEPTIDE IN CNS, PARTICULARLY IN HIPPOCAMPUS**

PTN and MK have been shown to induce and stimulate neuronal differentiation (Jung et al., 2004; Ishikawa et al., 2009; Luo et al., 2012). More specifically, PTN has been involved in lineage-specific differentiation of glial progenitor cells, axonal outgrowth, synaptic plasticity, and angiogenesis (Mitsiadis et al., 1995; Kadomatsu and Muramatsu, 2004). PTN participates in axon regeneration after injury, being highly expressed by reactive astrocytes (Iseki et al., 2002) as a source of trophic support for neurons in brain (Dugas et al., 2008) and rescuing nigral dopaminergic neurons from degeneration (Hida et al., 2007; Moses et al., 2008). However, its precise molecular mechanisms remain unknown.

In addition to these widely recognized roles of PTN, functionally, PTN−/− mice exhibited a delayed response to nociceptive stimulus in the tail-flick test (Gramage and Herradón, 2010), and clonidine-induced analgesia was significantly enhanced (Vicente-Rodríguez et al., 2013) when compared with MK−/− and Wild-type (WT+/+) mice. These evidences strongly suggest that endogenous PTN modulates nociceptive transmission at the spinal level.

In addition, PTN has been involved in neurodegenerative disorders and in response to chronic drug consumption. PTN is upregulated in cortex and caudate-putamen after injection of a cannabinol (Mailleux et al., 1994), and in nucleus accumbens after acute administration of amphetamine (Le Grevès, 2005); in addition, it is also highly upregulated in substantia nigra of patients with Parkinson disease (Marchionini et al., 2007) and treatment with L-Dopa increases PTN levels in striatum (Ferrario et al., 2004). Thus, it has been involved, as is MK (Prediger et al., 2011), in regulation of the survival and function of dopaminergic neurons (Jung et al., 2004). Taken together, this evidence supports the hypothesis that PTN is upregulated in neurodegenerative and addictive disorders in order to induce trophic or neuroprotective effects on dopaminergic neurons (Herradón and Pérez-García, 2014).

After PTN expression was described in hippocampus (Bloch et al., 1992; Vanderwinden et al., 1992; Wanaka et al., 1993), it was suggested that it plays a role in injury-induced response (Takeda et al., 1995; Poulsen et al., 2000) and activity-dependent plasticity (Lauri et al., 1996; Rauvala and Peng, 1997) in rat hippocampus, by affecting early, synapse-specific stages of LTP production (Lauri et al., 1998). Later, it was demonstrated, in PTN-deficient mice, that hippocampal slices exhibit a lowered threshold for induction of LTP (Amet et al., 2001) and that LTP was attenuated in mice overexpressing PTN (Pavlov et al., 2002), possibly by enhancing GABAergic inhibition in CA1 (Pavlov et al., 2006) and affecting recognition memory (del Olmo et al., 2009). Together, these evidences indicate that PTN could act as inducible signal to inhibit LTP in the hippocampus. Therefore, taken collectively, these evidences add a new role to the previous functions referred for PTN, thus functioning as a neuromodulatory factor in the hippocampus (**Table 1**). However, molecular evidence continues to be incomplete regarding the complex signaling system involved in PTN modulation.

To complete a whole view and to fully understand the modulatory role of PTN in CNS, and particularly in hippocampus, it is necessary first to establish which elements of the molecular machinery are present, and second, which are the ways in which they interact with each other. In this respect, immunohistochemical analyses reveal that RPTPζ and its substrate, GIT1/Cat-1, are

#### **Table 1 | PTN functions**.


PTN, Pleiotrophin; CNS, Central nervous system.

co-localized in the processes of pyramidal cells in hippocampus and neocortex in rat brain, and PTN increases tyrosine phosphorylation of GIT1/Cat-1 in neuroblastoma B103 cells (Kawachi et al., 2001). Also, PSD-95/SAP90 family proteins, along with RPTPζ, are distributed in the dendrites of pyramidal neurons of hippocampus and neocortex (Kawachi et al., 1999). Additionally, it has been demonstrated that P190 RhoGAP activity, regulated by PTN/RPTPζ pathway, is involved in hippocampus-dependent memory formation through the downstream Rho/Rock pathway, which plays an important role in cell migration, axonal growth, and synaptic plasticity (Tamura et al., 2006). Another receptor involved in PTN signaling is N-syndecan receptor (Raulo et al., 1994), which due to deficiency in hippocampus exhibits enhanced LTP and altered hippocampus-dependent memory (Kaksonen et al., 2002). Moreover, this KO mouse is not responsive to PTN.

On the other hand, PTN regulates neurite extension and plasticity in pig hippocampal neurons *in vitro*, signaling through chondroitin sulfate/dermatan sulfate hybrid chains (Bao et al., 2005; Raulo et al., 2005); this action could involve chondroitin sulfate E as a binding partner, co-receptor, or genuine receptor for PTN (Deepa et al., 2002), but it is also reasonable to speculate that this could involve NGC, a brain-specific chondroitin sulfate proteoglycan involved in neuritogenesis (Nakanishi et al., 2006) and which interacts with PTN (Nakanishi et al., 2010).

#### **CONCLUDING REMARKS**

Therefore, the different actions of PTN as a neuromodulatory peptide (**Table 1**) could vary during development depending on the signaling pathways that it mainly activates. During early brain development, PTN implication in regulating neurogenesis and neural migration and differentiation, regulating axonal outgrowth, dendritogenesis, and synaptogenesis, could principally involve signaling through PTN/RPTPζ, and also through integrin αυ β3, possibly acting coordinately. Later, in adult, the participation of PTN in learning and long-term memory, by modulating LTP by activity-dependent plasticity memory process in hippocampus, can again be principally mediated by its signaling through PTN/RPTPζ, possibly in combination with its signaling through N-syndecan pathway. Finally, its neuroprotective effects constitute a relevant role, suggesting that PTN signaling pathways are involved in neurodegenerative disorders, as well as in response to injuries and chronic drug consumption. Those signaling pathways may be functioning through a multi-molecular complex of receptors, combining previously mentioned receptors and other adaptor proteins, which interact inside membrane microdomains in raft configuration, which could explain each of these functions.

We are sure that a lot of molecules involved in PTN signaling pathways remain unknown to date. It is necessary to perform more integral studies, such as the use of proteomics and genomics approaches, as well as studies *in vivo* (employing PTN-KO) and *in vitro* (by mean of experiments with small interfering RNA [siRNA]), which will undoubtedly elucidate the complete molecular mechanisms involved.

#### **AUTHOR STATEMENT**

Daniel Ortuño-Sahagún, Argelia E. Rojas-Mayorquín, Mercè Pallàs Conceived the work, drafted and revised critically.

Celia González-Castillo, Carolina Guzmán-Brambila, Acquire and compilates information.

All authors, Write and revised the manuscript.

#### **ACKNOWLEDGMENTS**

This work was partially supported by Universidad de Guadalajara (UdeG) grant 222769 PRO-SNI 2014 to Daniel Ortuño-Sahagún, and CONACyT-México grants 2012-180268 and PROMEP/103.5/12/8143 to Argelia E. Rojas-Mayorquín. Authors wish to thank Dra. Consuelo Morgado and to Frontiers reviewers for their critical review of the manuscript and constructive comments. Our apologies to authors whose works have not been reviewed and to those whose papers have not received the emphasis that they merit. We also apologize to authors whose work has not been appropriately cited due to limitations of space and/or to limitations of our knowledge.

#### **REFERENCES**


neurogenesis of hippocampal neurons. *Neuroscience* 164, 1020–1030. doi: 10. 1016/j.neuroscience.2009.09.012


**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 October 2014; accepted: 10 December 2014; published online: 08 January 2015*.

*Citation: González-Castillo C, Ortuño-Sahagún D, Guzmán-Brambila C, Pallàs M and Rojas-Mayorquín AE (2015) Pleiotrophin as a central nervous system neuromodulator, evidences from the hippocampus. Front. Cell. Neurosci. 8:443. doi: 10.3389/fncel.2014.00443*

*This article was submitted to the journal Frontiers in Cellular Neuroscience*.

*Copyright © 2015 González-Castillo, Ortuño-Sahagún, Guzmán-Brambila, Pallàs and Rojas-Mayorquín. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution and 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*.

# Protection against neurodegeneration with low-dose methylene blue and near-infrared light

F. Gonzalez-Lima\* and Allison Auchter

*Department of Psychology and Institute for Neuroscience, University of Texas at Austin, Austin, TX, USA*

Keywords: mitochondrial respiration, near-infrared light, methylene blue, neuroprotection, neurotherapeutic potential

Neurons are metabolically protected against degeneration using low-level methylene blue and nearinfrared light interventions. Both of these novel interventions act by a cellular mechanism involving enhancement of the electron transport chain in mitochondria, which promotes energy metabolism and neuronal survival (Gonzalez-Lima et al., 2014). Methylene blue preferentially enters neuronal mitochondria after systemic administration, and at low-doses forms an electron cycling redox complex that donates electrons to the mitochondrial electron transport chain. Low-level nearinfrared light applied transcranially delivers photons to cortical neurons that are accepted by cytochrome oxidase, which causes increased cell respiration and cerebral blood flow. Breakthrough in vivo studies with these interventions suggest that targeting mitochondrial respiration may be beneficial for protection against different types of neurodegenerative disorders.

#### Edited by:

*Karla Guadalupe Carvajal, National Institute of Peadiatrics, Mexico*

#### Reviewed by:

*Michal Novak, Slovak Academy of Sciences, Slovakia Karla Guadalupe Carvajal, National Institute of Peadiatrics, Mexico*

#### \*Correspondence:

*F. Gonzalez-Lima, gonzalezlima@utexas.edu*

Received: *21 October 2014* Accepted: *24 April 2015* Published: *12 May 2015*

#### Citation:

*Gonzalez-Lima F and Auchter A (2015) Protection against neurodegeneration with low-dose methylene blue and near-infrared light. Front. Cell. Neurosci. 9:179. doi: 10.3389/fncel.2015.00179*

The purpose of this paper is to provide an update on the cellular mechanisms mediating the neuroprotective effects of low doses of methylene blue and near-infrared light, and to argue that the neurotherapeutic benefits of these two different interventions share the same cellular mechanism of action based on stimulation of mitochondrial respiration. Presented first is the explanation of the biochemical redox action of low-dose methylene as an electron cycler on the mitochondrial electron transport. Presented second is the explanation of the biophysical action of near-infrared light as a photon donor to cytochrome oxidase that also serves to stimulate mitochondrial electron transport. We finish with a comparison of these two interventions and how they share a common cellular mechanism with similar properties such as energy transfer, low-dose hormetic dose-responses, and enhanced capacity for oxidative metabolic energy production, which serve to protect nervous tissue from degeneration.

# Methylene Blue as Electron Donor

Low-dose methylene blue stimulates mitochondrial respiration by donating electrons to the electron transport chain. This is possible by a unique auto oxidizing redox chemical property. Methylene blue is unique among chemicals for several important reasons. Foremost is the auto oxidizing property that allows methylene blue at low concentrations to form a redox equilibrium by cycling electrons (i.e., serving as both an electron donor and acceptor). This property permits the cycling of electrons from chemicals inside the mitochondrial matrix to electron transport proteins in mitochondria. These transport proteins act as acceptors for electrons donated by methylene blue in mitochondria. The final acceptor of electrons in the respiratory chain is oxygen, which is obtained from oxyhemoglobin transported in the circulation. Molecular oxygen becomes reduced to water in a reaction catalyzed by the mitochondrial enzyme cytochrome oxidase (Complex IV, cytochrome c oxidase). The electron transport chain is coupled with the biochemical process of oxidative phosphorylation, which leads to increased oxygen consumption and the formation of ATP from ADP (**Figure 1**). Under normal physiological conditions the electrons that enter the electron transport chain come from electron donor molecules such as NADH and FADH2. These molecules derive from the Krebs cycle conversion of the food we eat. Methylene blue at low concentrations serves as another source of electrons for the electron transport chain that is part of mitochondrial respiration, leading to increased cytochrome oxidase activity and oxygen consumption (Riha et al., 2005; Wen et al., 2011; Rojas et al., 2012a; Rodriguez et al., 2014).

In addition, recent functional magnetic resonance imaging (fMRI) studies show that systemically administered lowdose methylene blue can directly reduce oxygen to water and support cerebral metabolic rate of oxygen consumption and metabolic energy production in normoxic and hypoxic conditions in vivo (Huang et al., 2013). Since rapid activation of oxygen consumption leads to a local transient hypoxia, cytochrome oxidase changes from reducing oxygen to catalyzing the formation of nitric oxide (Poyton and Ball, 2011). This in turns leads to a hemodymamic vasodilation response that increases cerebral blood flow and brain glucose uptake in vivo (Lin et al., 2012). This metabolic cascade promotes bigenomic regulation of mitochondrial and nuclear genes whose expression up-regulate cytochrome oxidase protein subunits and holoenzyme levels in nervous tissue (Rojas et al., 2012a).

FIGURE 1 | Two neuroprotective interventions for enhancing mitochondrial respiration. Low-dose methylene blue (MB) acts as an exogenous electron (e-) cycler, boosting oxygen consumption and cell respiration (molecular O2 reduced to H2O). Low-level red-to-near-infrared light directly energizes cytochrome oxidase (Complex IV) via photon absorption, facilitating its catalytic activity and leading to up-regulation of cytochrome oxidase levels. These interventions result in long-term increases in the amount of cytochrome oxidase in the electron transport chain by a process of enzymatic induction, which promotes oxidative energy metabolism and neuronal survival. Abbreviations: I–IV, refer to the four electron transport enzymatic complexes in the inner membrane of mitochondria; MB, is oxidized methylene blue (blue color); MBH2 , is reduced methylene blue (colorless); H+, stands for the protons pumped by Complexes I, III, and IV that enter the mitochondrial matrix via ATP synthase, which results in ATP production.

# Near-infrared Light as Photon Donor

Near-infrared light from low-power lasers and light-emitting diodes (LEDs) stimulates mitochondrial respiration by donating photons that are absorbed by cytochrome oxidase, a bioenergetics process called photoneuromodulation in nervous tissue (Rojas and Gonzalez-Lima, 2013). Photons are packets of luminous energy from electromagnetic waves. The amount of energy delivered by a photon depends on the light wave frequency. Photons in the red-to-near-infrared frequency range of approximately 620–1150 nm penetrate to the brain and intersect with the absorption spectrum of cytochrome oxidase (Rojas and Gonzalez-Lima, 2011). The absorption of luminous energy by the enzyme results in increased brain cytochrome oxidase enzymatic activity and oxygen consumption (Rojas et al., 2012b). Since the enzymatic reaction catalyzed by cytochrome oxidase is the reduction of oxygen to water, acceleration of cytochrome oxidase catalytic activity directly causes an increase in cellular oxygen consumption. Because increased oxygen consumption by nerve cells is coupled to oxidative phosphorylation, ATP production increases as a consequence of the metabolic action of near-infrared light. This type of luminous energy can enter brain mitochondria transcranially, and—independently of the electrons derived from food substrates—it can directly photostimulate cytochrome oxidase activity.

Red-to-near-infrared light not only stimulates mitochondrial respiration and brain oxygen consumption when light is on. Light also has a longer lasting effect on metabolic capacity because its acceleration of cytochrome oxidase activity causes enzymatic induction, a process dependent on gene expression and protein synthesis that up-regulates the levels of cytochrome oxidase (Hayworth et al., 2010). For example, 1 day after a single session of photoneuromodulation, there are significantly higher levels of cytochrome oxidase enzyme in the rat brain (Rojas et al., 2012b). Indeed, behavioral effects in humans can still be observed at 2 and 4 weeks after a single transcranial near-infrared light treatment (Barrett and Gonzalez-Lima, 2013). This enzymatic induction provides a long-tern mechanism for increasing the oxidative metabolic capacity of neurons, which is manifested in vivo by increases in cerebral rates of oxygen consumption and blood flow to the brain (Uozumi et al., 2010; Rojas et al., 2012b).

# Cellular Mechanisms of Neuroprotection

While low-dose methylene blue and low-level near-infrared light may produce different pleiotropic cellular effects, both interventions cause a similar up-regulation of mitochondrial respiration with similar benefits to protect nerve cells against degeneration. First, both interventions increase the expression of brain cytochrome oxidase in vivo (Gonzalez-Lima et al., 2014). Methylene blue accomplishes this by supporting the electron transport chain, while near-infrared light does it by directly energizing cytochrome oxidase via photon absorption (**Figure 1**). Still, their primary cellular mechanism of action is the same: enhancement of mitochondrial respiration.

Second, both interventions share high bioavailability. Methylene blue readily crosses the blood-brain barrier, builds up inside neurons, and resides inside respiring mitochondria. Indeed, methylene blue injected to live animals was first used as a supravital stain of nervous tissue by Paul Ehrlich and Ramon y Cajal in the 1890s. In the modern case of near-infrared lasers used transcranially, the degree of penetration increases with longer wavelengths and pulse durations within the effective near-infrared spectrum range. For example, approximately 8–10% of luminous energy reaches the rat cerebral cortex and 1.7–2% reaches the human cerebral cortex (Gonzalez-Lima and Barrett, 2014).

Third, similar conditions affect their neural effects, such as the redox and activational status of the target tissue and the dose-response. Hormesis has been documented for the doseresponses of both methylene blue (Bruchey and Gonzalez-Lima, 2008) and near-infrared light (Huang et al., 2011). This phenomenon means that low doses produce opposite effects than high doses, and that intermediate doses may be ineffective. For example, while low doses of methylene blue are used to treat methemoglobinimia, high doses cause methemoglobinimia (Bruchey and Gonzalez-Lima, 2008). Hence it does not make sense to refer to methylene blue without specifying the dose level, as different doses produce opposite effects. For example, while high doses may inhibit tau aggregation and nitric oxide formation in vitro, they are toxic in vivo (Riha et al., 2005; O'Leary et al., 2010). But systemic low-doses (0.5–4 mg/kg) of methylene blue that stimulate mitochondrial respiration in vivo are safe and effective in both animals and humans (Rojas et al., 2012a). Similarly, only low-level near-infrared light is beneficial because higher doses become ineffective or produce opposite effects. The dosing for transcranial near-infrared light depends on multiple parameters besides wavelength, such as transmission, fluency, irradiance, number of fractions, pulsing, etc. (Huang et al., 2011). For example, forehead transcranial stimulation of the human cerebral cortex has been done effectively with a continuous wave 1064 nm laser at 60 J/cm<sup>2</sup> (fluence), 250 mW/cm<sup>2</sup> (irradiance) for 4 min, which corresponds to about 1.2 J/cm<sup>2</sup> energy density reaching the cortical surface with a 2% transmission (Barrett and Gonzalez-Lima, 2013).

An effective mechanism of stimulation of mitochondrial respiration protects against neurodegeneration by increasing the oxidative metabolic energy capacity of neurons and reducing oxidative damage (Wen et al., 2011). With increases in the capacity to produce ATP by up-regulation of cytochrome oxidase, multiple secondary benefits accrue such as enhancement of neuronal metabolic energy and bigenomic responses, antiapoptotic signaling, DNA repair, mitogenic signaling, axonal sprouting, synaptogenesis and brain-derived neurotrophic factor (Martijn and Wiklund, 2010; Gomes et al., 2012; Poteet et al., 2012; Rojas and Gonzalez-Lima, 2013; Xuan et al., 2014). Lowdoses of methylene blue and near-infrared light that up-regulate mitochondrial respiration in vivo have similar neuroprotective effects in multiple model systems featuring neurodegeneration. These include models of neurotoxicity (Zhang et al., 2006; Rojas et al., 2008, 2009a), ischemia (Yip et al., 2011; Watts et al., 2013; Auchter et al., 2014; Rodriguez et al., 2014), neurotrauma (Bittner et al., 2012; Oron et al., 2012; Quirk et al., 2012; Shen et al., 2013; Xuan et al., 2013; Zhang et al., 2014), neurocognitive and emotional impairment (Callaway et al., 2004; Gonzalez-Lima and Bruchey, 2004; Wrubel et al., 2007a,b; Riha et al., 2011), Alzheimer's disease (Callaway et al., 2002; Auchter et al., 2014), and Parkinson's disease (Rojas et al., 2009b; Wen et al., 2011).

Although most of the mechanistic studies have been undertaken in animal models, the in vivo neuroprotective benefits of low-dose methylene blue and near-infrared light have been documented in humans as well (Rojas et al., 2012a; Rojas and Gonzalez-Lima, 2013). For example, lowdose methylene blue in humans leads to neuroprotection and reversal of ifosfamide-induced encephalopathy (Turner et al., 2003), and it also improves treatment of bipolar and unipolar depressive disorders (Naylor et al., 1986, 1987). In addition, a recent controlled, randomized, double-blind study in phobic patients provided the first peer-reviewed report of improvement in human memory functions by low-dose methylene blue, specifically fear extinction memory and contextual memory (Telch et al., 2014). These human studies suggest that low-dose methylene blue may have potential therapeutic applications in neurology as a neuroprotective agent, and in psychiatry and clinical psychology to facilitate psychotherapeutic interventions. Similarly, low-level near-infrared light improved human neurological outcome after ischemic stroke (Lampl et al., 2007), and enhanced emotional and neurocognitive functions such as sustained attention and working memory in humans (Gonzalez-Lima and Barrett, 2014).

# Conclusion

We have described two very different chemical-physical interventions resulting in a similar cellular mechanism targeting mitochondrial respiration (Gonzalez-Lima et al., 2014). One entails fostering electron cycling by an auto-oxidizing redox chemical, and the other using photon absorption from nearinfrared light. New in vivo evidence from animal models and human studies suggest that low-dose methylene blue and low-level near-infrared light share a common mechanism of enhancement of mitochondrial respiration that protects against neuronal degeneration in a broad range of animal models and human neurobehavioral disorders. We hope that this fascinating neuroprotective approach targeting mitochondrial respiration will stimulate more human studies of potential therapeutic applications in neurology and psychiatry.

# Acknowledgments

Professor FG gratefully acknowledges his support from a College Research Fellowship award from the University of Texas at Austin.

# References


requires neuroprotection and reduced soluble tau burden. Mol. Neurodegener. 5:45. doi: 10.1186/1750-1326-5 45


experimentally induced transient cerebral ischemia. Neuroscience 190, 301–306. doi: 10.1016/j.neuroscience.2011.06.022


**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.

Copyright © 2015 Gonzalez-Lima and Auchter. 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.

# Retinal aging in the diurnal Chilean rodent (Octodon degus): histological, ultrastructural and neurochemical alterations of the vertical information processing pathway

#### Edited by:

Victoria Campos-Peña, Instituto Nacional de Neurologia y Neurocirugia, Mexico

#### Reviewed by:

Rafael Linden, Federal University of Rio de Janeiro, Brazil Benjamín Florán, Centro de Investigación y de Estudios Avanzados del IPN, Mexico

#### \*Correspondence:

Maria Trinidad Herrero, Clinical and Experimental Neuroscience (NiCE), CIBERNED and Institute of Bio-Health Research of Murcia (IMIB), School of Medicine, Campus Mare Nostrum, University of Murcia, Campus Espinardo, 30100 Murcia, Spain Tel: +34 868 88 84 84, Fax: +34 868 88 41 50 mtherrer@um.es

† It is with great sorrow that authors notify that Krisztina Szabadfi passed away tragically during the submission of this manuscript.

> Received: 21 October 2014 Accepted: 17 March 2015 Published: 21 April 2015

#### Citation:

Szabadfi K, Estrada C, Fernandez-Villalba E, Tarragon E, Setalo Jr. G, Izura V, Reglodi D, Tamas A, Gabriel R and Herrero MT (2015) Retinal aging in the diurnal Chilean rodent (Octodon degus): histological, ultrastructural and neurochemical alterations of the vertical information processing pathway. Front. Cell. Neurosci. 9:126.

doi: 10.3389/fncel.2015.00126

Krisztina Szabadfi1,2† , Cristina Estrada<sup>3</sup> , Emiliano Fernandez-Villalba<sup>3</sup> , Ernesto Tarragon<sup>3</sup> , Gyorgy Setalo Jr. <sup>4</sup> , Virginia Izura<sup>3</sup> , Dora Reglodi <sup>5</sup> , Andrea Tamas <sup>5</sup> , Robert Gabriel 1,2 and Maria Trinidad Herrero<sup>3</sup> \*

<sup>1</sup> Department of Experimental Zoology and Neurobiology, University of Pecs, Pecs, Hungary, <sup>2</sup> Janos Szentagothai Research Center, Pecs, Hungary, <sup>3</sup> Clinical and Experimental Neuroscience (NiCE), CIBERNED and Institute of Bio-Health Research of Murcia (IMIB), School of Medicine, Campus Mare Nostrum, University of Murcia, Murcia, Spain, <sup>4</sup> Department of Medical Biology, University of Pecs, Pecs, Hungary, <sup>5</sup> Department of Anatomy, MTA-PTE "Lendulet" PACAP Research Team, University of Pecs, Pecs, Hungary

The retina is sensitive to age-dependent degeneration. To find suitable animal models to understand and map this process has particular importance. The degu (Octodon degus) is a diurnal rodent with dichromatic color vision. Its retinal structure is similar to that in humans in many respects, therefore, it is well suited to study retinal aging. Histological, cell type-specific and ultrastructural alterations were examined in 6-, 12- and 36-months old degus. The characteristic layers of the retina were present at all ages, but slightly loosened tissue structure could be observed in 36-month-old animals both at light and electron microscopic levels. Elevated Glial fibrillary acidic protein (GFAP) expression was observed in Müller glial cells in aging retinas. The number of rod bipolar cells and the ganglion cells was reduced in the aging specimens, while that of cone bipolar cells remained unchanged. Other age-related differences were detected at ultrastructural level: alteration of the retinal pigment epithelium and degenerated photoreceptor cells were evident. Ribbon synapses were sparse and often differed in morphology from those in the young animals. These results support our hypothesis that (i) the rod pathway seems to be more sensitive than the cone pathway to age-related cell loss; (ii) structural changes in the basement membrane of pigment epithelial cells can be one of the early signs of degenerative processes; (iii) the loss of synaptic proteins especially from those of the ribbon synapses are characteristic; and (iv) the degu retina may be a suitable model for studying retinal aging.

Keywords: Octodon degus, aging, retina, vertical pathway, ultrastructure, rod bipolar cells, synaptic proteins

# Introduction

The vertebrate retina, like other parts of the central nervous system, is subjected to degenerative changes caused by aging. The retina is also the site of diseases for which age is a major risk factor, including macular degeneration and glaucoma (Jackson and Owsley, 2003). The retina is arguably the best understood part of the vertebrate central nervous system with regard to its cellular patterning, circuitry, and function. It is composed of five major neuron types: photoreceptors, interneurons (horizontal, bipolar, and amacrine cells), and retinal ganglion cells (RGCs) that integrate visual information and send it to the brain (Sanes and Zipursky, 2010). Retinal neurons can be further subdivided into approximately 70 distinct functional subtypes (Masland, 2001) for many of which markers are available to identify the aging-specific alterations.

Age-related complications have been demonstrated in several mammalian species, including monkeys, cats, sheep, rats, mice and Octodon degus (degu). This latter species presents several advantages for studying different pathological conditions. The degu is a diurnal, highly visual South American hystricomorph rodent native to Chile, which in old age expresses cognitive deficits, anxiety (Popovi´c et al., 2009) and unstable circadian rhythms of low amplitude (Vivanco et al., 2007). Particularly notable is that the animals develop spontaneous Alzheimer-like pathology and show signs of significant white matter disruption, diabetes and cancer in aging (Inestrosa et al., 2005; Ardiles et al., 2012, 2013), resembling several aspects of pathological human aging (van Groen et al., 2011).

The visual system of degus is also comparable to the human visual system. It shows robust responses to both photic and nonphotic circadian Zeitgebers (Goel et al., 1999; Jacobs et al., 2003). Degus have the potential for dichromatic color vision on the basis of green-sensitive M cones and UV-sensitive (in the near UV) S cones, the most common type of mammalian color vision (Jacobs, 1993; Chávez et al., 2003; Palacios-Muñoz et al., 2014). In degus, the retinal projection is primarily contralateral, with a small ipsilateral component (Fite and Janusonis, 2001). Degu is also well suited model for studying eye pathology, because they have an increased susceptibility to cataract development and aging (Worgul and Rothstein, 1975; Brown and Donnelly, 2001; Peichl et al., 2005). However, no data are available on retinal aging in degu. In other rodents (rats and mice), retinas show some aging alterations. For example, the total retinal area expands while RGC dendritic arbors shrink with age, thus, each RGC covers a decreased fraction of the visual field in old animals. Amacrine and bipolar cells also exhibit age-related structural changes, some of which may contribute to reduced visual function (Samuel et al., 2011). Neuronal loss with age is characteristic for some, but not all species, leading to thinning of the cellular and synaptic layers (Miller et al., 1984; Limaye and Mahmood, 1987; Morrison et al., 1990; Gao and Hollyfield, 1992; Kim et al., 1996; Samuel et al., 2011). Ultrastructural changes have also been revealed in the neural, vascular and epithelial components. Even more prominent changes can be observed in the retinal pigment epithelial (RPE) layer than in the neuroretina during the early phases of aging. Signs include increased number of basal infoldings, phagolysosomes and lipofuscin deposits. In aged rat retina, organelle atrophy and whirling extensions of the basal membrane into the cytoplasm are characteristic in the RPE cells (DiLoreto et al., 2006). However, it is not known at present if these changes are also characteristic to degus. In spite of the similarities between human and degu retinas (Cuenca et al., 2010), surprisingly little is known about the degu retina and its retinal aging.

Therefore, the aim of the present study was to perform a complex retinal characterization of degu at histological, ultrastructural and immunohistochemical levels during aging focused on the elements of the vertical pathway (photoreceptors to bipolar to ganglion cells). Since degu retina is more similar to the human retina than to the retina of other rodents, this description will provide a strong foundation for future studies where experimental manipulations and/or neuroprotective agents can be studied.

# Materials and Methods

# Animals

A total of 28 female degus (body weight 180–270 g) of 6 (n = 8), 12 (n = 8), and 36-months of age (n = 12) were used. This latter group is considered as aging (but not old) group. Degus were housed individually in opaque glass cages (40 × 25 × 25 cm) at the animal facilities of the University of Murcia. Throughout the study, the experimental room was maintained under controlled temperature (21 ± 1 ◦C) and 12 h light/dark cycle (lights on at 7:00 a.m. and off at 19:00 p.m.). The floors of the cages were covered with wood shavings that were changed once a week. Food and water were provided ad libitum by placing 120 g food pellets (Harlan Tekland Global Diet®, Harlan Laboratories, USA) per day and water bottles on a grid located on the top of the tank. The water in the tank was changed daily. All experiments were performed in accordance with relevant regulatory standards, experimental guidelines and procedures complied with the European Community Council Directive (2010/63/UE) and the ethical committee of the University of Murcia.

#### Histological and Electron Microscopic Analysis

Animals were anesthetized with Isofluorane (Isoba® vet, USA), administered with a continuous flow vaporizer (MSS3, Medical Supplies and Services, England, UK), and then sacrificed by decapitation. Both eyes were immediately removed and distinctly disposed according to histological or electron microscopic procedure to be performed.

For histology eyes were fixed in 4% paraformaldehyde (PFA; Merck, Hungary) dissolved in 0.1M phosphate buffer (PB; Spektrum3D, Hungary). The eyecups were dissected and embedded in epoxy resin (Durcupan ACM resin; Sigma-Aldrich, Hungary) as we previously described (Szabadfi et al., 2012). Sections were cut at 2 µm, stained with toluidine blue (Sigma-Aldrich, Hungary), and examined in a Nikon Eclipse 80i microscope. Measurements were taken with the SPOT Basic program. Central retinal areas within 1 and 2 mm from the optic disc were used for measurements (n = 2–5 measurements from one tissue block). The following parameters were measured: (i) cross-section of the retina from the outer limiting membrane (OLM) to the inner limiting



Abbreviations: CtBP2—C-terminal Binding Protein 2; GFAP—glial fibrillary acidic protein; PKCα—protein kinase Cα; PNA—peanut agglutinin-conjugated with FITC; VGLUT1—vesicular glutamate transporter 1.

membrane (ILM); (ii) the width of individual retinal layers. Statistical comparisons were made using one-way ANOVA test followed by Tukey-B posthoc analysis. Data were presented as mean ± SEM (GraphPadPrism5.0).

Electron microscopy was performed on eyes fixed with 4% PFA supplemented with 1% glutaraldehyde dissolved in 0.1M PB. After washing in PB, tissue samples were treated with 1% OsO<sup>4</sup> in PB, dehydrated through ascending ethanol series and embedded in Durcupan ACM resin (Sigma-Aldrich, Hungary). Sections were cut at 70 nm in Reichert Ultracut S and counterstained with Reynold's lead citrate. Samples were examined and photographed in a JEOL 1200EX electron microscope.

#### Immunohistochemistry

Eyes were dissected immediately after sacrifice in ice-cold phosphate buffer with saline (PBS) and fixed in 4% PFA at room temperature. Tissues were then washed in PBS and cryoprotected in 20% sucrose at 4◦C. For cryostat sectioning, retinas were embedded in tissue-freezing medium (Shandon Cryomatrix, USA), cut in a cryostat (Leica, Germany) at 10 µm radially. Sections were mounted on subbed slides. Primary antibodies and peanut agglutinin-conjugated with FITC (PNA) were used overnight at room temperature (**Table 1**). Next day the sections were incubated for 2 h at room temperature with the corresponding secondary fluorescent antibodies in the dark, then coverslipped using Fluoromount-G (Southern Biotech, USA). For the colocalization study, we used 10 µm cryostat sections simultaneously with antibodies to Chx10 and protein kinase Cα (PKCα); C-terminal Binding Protein 2 (CtBP2) and Bassoon; PKCα and Bassoon; PKCα and postsynaptic density 95 protein (PSD95/SAP90); and PKCα and vesicular glutamate transporter 1 (VGLUT1), respectively. These were detected with corresponding secondary antibodies (**Table 1**); nuclei were counterstained with DAPI (4<sup>0</sup> , 6-diamidino-2-phenylindole; 1:10000), then coverslipped using Fluoromount-G (Southern Biotech, USA). For control experiments, primary antibodies were omitted, and cross-reactivity of the non-corresponding secondary antibodies with the primaries was also checked. Photographs were taken with Nikon Eclipse 80i Microscope (Nikon, Japan) and Fluoview FV-1000 Laser Confocal Scanning Microscope (Olympus, Japan) and further processed with Adobe Photoshop 7.0 program. Images were adjusted for contrast only, aligned, arranged, and labeled using the functions of the above program.

The number of RGCs (Brn3a-positive cells; Xiang et al., 1995; Nadal-Nicolás et al., 2009) ± SEM was measured in 100 µm ganglion cell layer (GCL) length. The cells expressing both Chx10 and PKCα were scored as rod bipolar cells, and cells expressing Chx10 but not PKCα were scored as cone bipolar cells as followed the protocol of Morrow et al. (2008). The number of all bipolar cells and rod bipolar and cone bipolar cells were counted in 100 µm<sup>2</sup> area of INL. Statistical comparisons were made using one-way ANOVA test followed by Tukey-B posthoc analysis. Data were presented as mean ± SEM (GraphPadPrism5.0).

# Results

The baseline characterization of 6-, 12- and 36-month-old degu retinas was initially done by routine histology.

### Descriptive Morphological and Morphometric Analysis

The characteristic layers of the mammalian retina were well distinguishable in degu: photoreceptor layer (PL), outer nuclear layer (ONL), outer plexiform layer (OPL), inner nuclear layer (INL), inner plexiform layer (IPL) and GCL (GCL; **Figure 1**). The typical cells of the mammalian retina (photoreceptor cell bodies and outer segments (OS) of cones and rods, bipolar cells, different types of amacrine cells, horizontal cells, displaced amacrine cells, ganglion cells and Müller glial cells) were also well visible at all ages.

There were only minor differences between the three groups (**Figures 1A–C**). A loose retinal structure could be observed in aging degu retinas (**Figure 1C**). This observation was manifested in a significantly increased OLM–ILM distance and IPL thickness in 36-month-old degu retinas (**Figure 1D**). However, the thickness of other layers did not change significantly in the aging retinas (**Figure 1D**; ##p < 0.001 vs. 6-month-old, ###p < 0.001 vs. 12-month-old degu retinas). The number of RGCs (Brn3apositive cells)/100 µm retina length was significantly decreased in the 36-month-old degu retina (∗∗p < 0.001 vs. 6- and 12 month-old degu retinas; **Figure 1E**) compared to the 6- and 12 month-old retinas.

#### Glial Cells and The Structure of Outer Retina

Glial fibrillary acidic protein (GFAP)-positivity was selectively localized to endfeet of Müller cells (**Figures 2A–D**). Müller glial cells respond rapidly to any alterations of the retinal microenvironment by elevated expression of GFAP as a specific metabolic stress signal in the mammalian retina (Cuenca et al., 2014). Increased GFAP immunoreactivity was observed in the entire width of the 36-month-old degu retina in few, but not all, Müller glial cells (**Figures 2C,D**), compared to 6- (**Figure 2A**) and 12-month-old (**Figure 2B**) degu retinas.

Basement membrane proliferation of the RPE in the areas of age-related retinal peripheral degeneration was observed in the 36-month-old degus retinas. The thickening of Bruch's membrane and fibrosis of the choriocapillary were evident. The somas of the RPE cells were pressed toward this thinner basement membrane in the 36-month-old retinas, suggesting an initial agerelated alteration of the RPE. Early peripheral changes of the RPE included increased basal infoldings, phagolysosomes and lipofuscin deposits, as well as atrophy and whirling extensions of the basement membrane into the cytoplasm (**Figure 3C**). The relation between RPE and the OS of the photoreceptors seemed intact in all groups (**Figures 3A–C**).

We observed that the cones composed a nest between the rods in all age groups, in accordance with the cone-dominant retinal structure of the degu (Cuenca et al., 2010). Most of these cone-nests were located near the OLM (**Figures 3D–F**). In the 36-month-old group, we detected altered photoreceptor ratio for the favor of cones suggesting that rods are more sensitive to

aging. We also observed degenerating rods in the ONL of the 36-month-old group (**Figure 3F**). PNA is used to label the OS and the terminals of cone photoreceptors in the OPL. No differences could be observed in the OSs between the three different age groups (**Figures 3G–I**) and the number of cone terminals (data not shown), which further supports the selective loss of rods at 36-months.

6- and 12-month-old degu retinas (A,B). In the 36-month-old degu retinas the GFAP immunoreactivity is very strong within their endfeet at the vitreoretinal

We found ribbon synapses in the OPL and observed differences between the rod and the cone terminals. The rod terminals (RT) were more electron dense and cup-shaped spherules, while cone terminals were bowl-shaped, less dense pedicles (**Figures 4A,B**). The only difference between the three groups was the reduced cytoplasmic density of the RT with older age. As a consequence we could not easily distinguish the rod and the cone terminals in the 36-month-old groups (**Figure 4C**).

Photoreceptors transmit their signals at ribbon synapses in the OPL, the first synaptic region in the retina, whereas bipolar cells make their ribbon synaptic contacts in the IPL. A large number of regularly aligned synaptic vesicles were tethered to the ribbon. The rod photoreceptor ribbon synapses had horseshoeshaped structure in the OPL, however, the cone photoreceptor (OPL) and bipolar cell ribbons (IPL) had a dot-like appearance at light microscopic level. We analyzed the structure of the OPL with retina-specific ribbon synapse markers presynaptic CtBP2 and Bassoon labeling. The rod ribbon synaptic profiles were marked with horseshoe-shaped ribbons by CtBP2 and continuous distribution of punctate staining by Bassoon protein (**Figures 4D,E**). Near the horseshoe-shaped ribbons, some degenerated synaptic structures (fragmented ribbons) could be observed in the OPL of 36-month-old degu retinas (**Figure 4F**). The localization of Bassoon was similar in the three groups (**Figures 4D–F**).

# Bipolar Cells and Ultrastructure of the Synaptic Layers

The pan-bipolar cell marker Chx10 reveals the organization of bipolar cells in the INL of the retina (Elshatory et al., 2007). We found that anti-Chx10 antibodies stained all bipolar cells in all groups (**Figures 5A,E,I**). The presence of PKCα was detected in the rod bipolar cell population. The labeled structures were the cell bodies of the INL, the dendrites in the OPL and cell processes extending into the IPL, close to the GCL. There were no alterations detected in the arborization and pattern of the cone and rod bipolar cells between the 6- and 12-month-old groups (**Figures 5A–L**). In the 36-month-old retinas empty cell bodylike shapes could be observed in the bipolar cell area of INL (**Figures 5I,J,L**).

acidic protein (GFAP); OLM—outer limiting membrane; ILM—inner limiting

membrane. Scale bar: 20 µm.

Differences could be observed in the total bipolar and rod bipolar cell numbers (∗p < 0.05 vs. 6- and #p < 0.05 vs. 12 month-old groups and ∗∗∗p < 0.0001 vs. 6- and ###p < 0.0001 vs. 12-month-old groups, respectively). The most prominent difference was at the age of 36 months with a significant reduction in these parameters (**Figures 6A,B**). However, the number of cone bipolar cells did not differ between the three groups (**Figure 6C**). These alterations resulted in an increased cone/rod bipolar cell ratio in the 36-month-old degu retina.

The dendritic field of rod bipolar cells was also altered. The progression of the degeneration was manifested by the retraction of rod bipolar cell dendrites. These dendrites became flatter and the loss of dendritic branches above the cell bodies was evident. In the 36-month-old group the PKCα positive dendrites were sparsely distributed (**Figures 7E,F**) compared to the 6- and 12-month-old groups. The pattern of PKCα positive dendrites was dense and each cell had a huge arbor in these latter two groups (**Figures 7A–D**). In normal retina the fine processes of rod bipolar cells penetrate the ONL. The dendritic trees appear brushy and candelabrum-like as we showed in the case of 6-month-old degus (**Figures 7A,B**). In 36 month-old degu retinas the rod bipolar cell dendrites were no longer erect and brushy, but appeared flattened (**Figures 7C,F**). The Bassoon-staining seemed to be unchanged during aging (**Figures 7A,C,E**). Glutamatergic photoreceptors affect the physiological properties of bipolar cells in the mammalian retina. To visualize the spatial pattern of glutamatergic input to the

bipolar cells in the OPL of degu retina, PSD95 was used to label these synapses. PSD95 puncta were regularly spaced along the membranes of the bipolar cell dendrites. The spatial distribution of PSD95 labeling of the bipolar cells were similar in the 6 and 12-month-old groups (**Figures 7B,D**). However, in the 36-month-old group the pattern of PSD95 was altered, only a few puncta were detected on the somas of rod bipolar cells and labeling was not shown along the non-brushy dendritic arbor (**Figure 7F**).

VGLUT1 was detected in the OPL and throughout the laminae of the IPL, a distribution consistent with the expected synaptic localization of the protein (Brandstätter et al., 1999). In degus, the coarse structure of the retina was not significantly affected, however, the axon terminals of rods and rod bipolar cells showed dramatic alterations in VGLUT1 expression in the 36-month-old group (**Figures 8C,F**). These degu retinas showed loss of most of the rod outputs to their bipolars (**Figures 8C,F**) compared to 6- (**Figures 8A,D**) and 12-monthold (**Figures 8B,E**) degu retinas. The axon terminals of the rod bipolar cells showed some alterations in the 36-month-old group, such as decreased staining both for PKCα and VGLUT1 in the innermost IPL. The loss of PKCα immunoreactivity in the axons and axon terminals of rod bipolar cells was also evident and their terminals were further reduced in size and density (**Figure 8F**) compared to 6- (**Figure 8D**) and 12-month-old (**Figure 8E**) retinas.

The IPL structure was well retained (**Figures 9A–F**) even at 36 months of age (**Figures 9D–F**). Both ribbon and conventional synapses were visible, synaptic vesicles were regularly distributed. Sometimes a few swollen neural profiles (lacking synaptic vesicles) were seen along with space-filling glial protrusions which can be clearly identified in the electron microscope. The structural elements of the GCL, nerve fiber layer (NFL) and ILM in general did not show major signs of degeneration in any of the groups (**Figures 9G–J**). The NFL fibers were embedded into the large endfeet of Müller glial cells (**Figures 9G,I,J**).

# Discussion

Progressive and irreversible functional decay during aging is characterized by region-specific neuron loss (Lossi et al., 2005). The retina is a potentially sensitive target of agedependent degeneration. In this paper, we report a comparison of the major retinal neuronal types of the vertical pathway in young adult (6-months-old), adult (12-months-old) and aging (36-months-old) degus. There are several remarkable differences along with numerous unchanged features in the different age groups.

In our study we focused on the elements of the vertical information processing pathway with the addition of the RPE and the Müller glial cells. In the retinas of 36-month-old animals we observed a slightly loosened tissue structure both at light and electron microscopic levels, elevated GFAP expression in Müller glial cells and reduced number of rod bipolar cells and RGCs. Other age-related differences were detected at ultrastructural level: alteration of the retinal RPE and degenerated photoreceptor cells, especially rods, was evident. Ribbon synapses in the OPL were sparse and often fragmented (**Figure 10**).

We revealed well-defined alterations in degus that are also characteristic in human, mouse and rat retinal aging. The neuroretina together with RPE cells form a functional unit of the visual system. The RPE usually bears very long sheet-like apical microvilli that project into a complex matrix. During aging the RPE undergoes a number of well characterized changes, including increase in the number of residual bodies and accumulation of basal deposits (Garron, 1963; Guymer et al., 1999). We observed degeneration of the RPE/Bruch's/choriocapillary complex in 36-month-old degu retinas, possibly leading to altered tissue oxygen levels and contributing to photoreceptor cell loss (DiLoreto et al., 2006). The degeneration process is complex and involves the accumulation of deposits, RPE cell loss leads to formation of hypopigmented areas, and the development of hyperpigmented areas. This stage can progress to a proliferative neovascular

(wet or exudative) form of degeneration characterized by the growth of choroidal vessels (choroidal neovascularization) or to a geographic form of atrophy characterized by damage of the RPE and of the neural retina (Limaye and Mahmood, 1987). Some signs of this process could already be seen at 36 months of age and we predict that these changes will become dominant at later ages. The observed elevated cone/rod bipolar cell ratio in the 36-month-old degu group indicates that the elements of cone pathway were more resistant to the agerelated degeneration than that of rods. This parallels well with the observation that in the human retinas there is a decrease of approximately 54% of the total rod photoreceptor density between the fourth and ninth decades of life (Gao and Hollyfield, 1992; Aggarwal et al., 2007) whereas cone density remains essentially unchanged (Curcio et al., 1993). The decrease in rod density ultimately triggered the associated decline of neurons connected to them (Gao and Hollyfield, 1992; Aggarwal et al., 2007).

Photoreceptors transfer the visual signals to the postreceptorial retinal network; malfunctioning of this process due to degeneration of rods, rod bipolar cells, or ribbon synapses will lead to impaired vision. During human rod degeneration, surviving rods, horizontal and amacrine cells similarly extend anomalous neurits throughout the retina (Li et al., 1995; Fariss et al., 2000). Photoreceptor degenerationdependent modifications in the synaptic machinery connecting photoreceptors with second-order neurons are evident: altered connectivity of rods and rod bipolar cells as well as horizontal cells affects retinal circuity. The normal pairing of presynaptic and postsynaptic markers are lost. The synaptic markers associated with photoreceptors and processes of bipolar and horizontal cells show abnormalities prior to significant photoreceptor loss (Cuenca et al., 2005). We report here agedependent structural changes at the ribbon synapses in the synaptic terminals of rod photoreceptor and rod bipolar cells, which conforms well with these observations.

In contrast to the observed initial decline of ribbon synapses and rod bipolar cells density, the loss of synaptic sites was not complete in the aging degu retina, since Bassoon staining persisted marking the functional integrity of the arciform density. Currently we do not know whether the rods or the rod bipolars are lost first. In other pathological conditions, for example in a rat model of hyperoxia, the loss of bipolar dendrites takes place before photoreceptor death (Dorfman et al., 2011).

Putative structural modifications of the inner retina can be a consequence of aging. Liets et al. (2006) have indeed shown aberrant processes in rod bipolar neurons as a consequence of aging. Aberrant processes establish normally structured synapses ectopically (Terzibasi et al., 2009). As the degeneration of rod bipolar cells progresses they display early retraction and loss of dendrites (Cuenca et al., 2005). In 36-month-old degu retinas the rod bipolar cell dendrites were no longer erect and brushy but appeared flattened in contrast to the normal retina, where the fine

###p < 0.0001 vs. 12-month-old degu retinas). No differences could be observed in the cone bipolar cell number between the three groups (C).

processes of rod bipolar cells penetrate the ONL, like those in the 6-month-old degus.

The PSD95 is a part of the dense structure attached to the postsynaptic membrane opposed to the presynaptic active zone to ensure normal synaptic transmission. The structural alterations in aged learning-impaired rats correlate with altered content of PSD proteins that are critically involved in normal synaptic function. The alterations in synaptic protein content resulted in reduced synaptic function (Nyffeler et al., 2007; Takada et al., 2008). Immunofluorescence for PSD95 was most prominent in the retina, the dendrites in the OPL opposed to the rod spherules and cone pedicles were strongly labeled (Koulen et al., 1998). The spatial distribution of PSD95 labeling on the bipolar cell dendrites was altered in the 36-month-old group: only a few puncta were detected, furthermore, labeling was not shown nearby and along the non-brushy dendritic arbors of rod bipolar cells, indicative of rod degeneration. At the same time, however, cones remained unaltered as it was proven by PNAlabeling. The specialization of rod and cone bipolar cells involves the differential expression of proteins involved in glutamatergic signaling. Hanna and Calkins (2007) described that 26 rod bipolar cells expressed at least one AMPA glutamate receptor subunit gene in monkey retina. This infers the presence of those scaffolding proteins (including PSD95) that are related to the ionotropic glutamate receptors (Hanna and Calkins, 2007). It is known that ischemia induces severe progressive inner retinal degeneration and down-regulation of synaptic proteins, such as PSD95 and synaptophysin (Guo et al., 2014). Although PSD95 is a very important protein in the retinal synaptic transmission in both the OPL and IPL, there is no information on its alteration in retinal aging. The decreased PSD95 expression in aging degu retinas suggested that the synthesis of PSD95 could be altered.

VGLUT1 was localized to photoreceptor and bipolar cell terminals, which is consistent with the function of photoreceptors and bipolar cells in vertical excitatory transmission with glutamate release in mammalian retina (Gong et al., 2006). As a result of photoreceptor degeneration VGLUT1 immunostaining was decreased in the OPL. Remodeling of bipolar cells during aging also affects their axon terminals. These reduced axon terminals of the rod bipolar cells showed reduced VGLUT1-staining and the shape and structure of terminals was also altered in 36-month-old degu retinas. Response of VGLUTs to diverse stimuli is altered with aging, for example after transient global cerebral ischemia in the rat brain (Llorente et al., 2013) or ischemia and excitotoxicity in the retina (Atlasz et al., 2008, 2010). Similarly to our findings, decreased VGLUT1 immunostaining was observed in aging ventral cochlear nuclei, possibly associated with age-related hearing loss (Alvarado et al., 2014).

These above observations suggest that both the input and the output synapses of the rod bipolar cells were affected in the aging degu retina. In contrast, the cone pathway appeared mostly unchanged in the aging degu retinas. It is possible that this only reflects the different time course of the degeneration of rods and cones in the aging process. Rod bipolar cells and rods disappear first, therefore, the secondary degeneration within the rod pathway is expected to occur earlier. This observed alteration in aging has serious consequences in the light of the fact that one rod bipolar cell makes synapses with multiple RT via their dendritic arbor (Wässle and Boycott, 1991). With advancing age, the reported significant decrease of the rods coincides with considerable reduction in the density of rod bipolar cells in humans (Gao and Hollyfield, 1992; Aggarwal et al., 2007), similarly to degus. Not all retinal cells are equally vulnerable to the effects of advancing age (Roufail and Rees, 1997). Marked differences in the 36-month-old degus were partly cell type specific, such as elevated GFAP expression in the Müller

FIGURE 7 | Spatial relationship of the dendritic arbor of rod bipolar cells and photoreceptor terminals. Double immunostaining with PKCα (green) and Bassoon (red) antibodies (A,C,E). The dendrites of rod bipolar cells were longer and more brushy in the 6- and 12-month-old degu retinas (A,C), than in the 36-month-old retinas (E). Double immunolabeling with PKCα (green) and PSD95 (red) antibodies shows the related alterations in the rod synaptic spherules and in the dendritic field of rod bipolar cells. PSD95

immunofluorescence is prominent in the OPL, where the axon terminals of rods, the so-called rod spherules, are labeled intensively in the 6- and 12-month-old degus retinas (B,D). Decreased PSD95 labeling in the reduced dendritic arbors of rod bipolar cells was shown in 36-month-old aging degu retina (F). Abbreviations: ONL—outer nuclear layer; OPL—outer plexiform layer; INL—inner nuclear layer; PKCα—protein kinase Cα; PSD95—postsynaptic density 95 protein. Scale bar: 10 µm.

VGLUT1-immunopositive structures were found in photoreceptor terminals in close apposition to PKCα-positive rod bipolar cell dendrites in the OPL of 6 and 12-month-old degu retinas (A,B). Altogether with the decreased dendritic arbor of the rod bipolar cells the VGLUT1-expression in photoreceptors also

glial cells, photoreceptors (especially rods) and rod bipolar cell loss with the alteration in their synaptic profile (ribbon synapse decreased in 36-month-old degu retinas (C). VGLUT1-positive large puncta were co-localized with PKCα in the inner part of the IPL indicating VGLUT1 expression in axon terminals of rod bipolar cell in 6- and 12-month-old degu retinas (D,E). In the 36-month-old degu retinas both the arbor of axon terminals of rod bipolar cells and their VGLUT1 expression decreased (F). Abbreviations: OPL—outer plexiform layer; INL—inner nuclear layer; IPL—inner plexiform layer; GCL—ganglion cell layer; PKCα—protein kinase Cα; VGLUT1—vesicular glutamate transporter 1. Scale bar: 10 µm.

and altered dendritic trees: no longer erect and brushy). In contrast the bipolar cells of mice show arbor-specific alteration;

FIGURE 9 | Ultrastructure of IPL, GCL and nerve fiber layer (NFL) in the 6-, 12- and 36-month-old degu retinas. In the IPL conventional (arrows) and ribbon synapses (arrowheads) were also present in all groups (6-month-old: (A,B); 12-month-old: (C); 36-month-old: (D–F)). The NFL is intact in all age

groups (6-month-old: (G,H); 12-month-old: (I); 36-month-old: (J); asterisks: nerve fibers). Abbreviations: IPL—inner plexiform layer; M—Müller glial cell; NFL—nerve fiber layer (NFL); RGC—retinal ganglion cell. Scale bars are indicated in the pictures.

their dendrites are sprouted but remain stable (Samuel et al., 2011).

Furthermore, the number of RGCs, the output neurons of the retina, was altered in degus. The prominent loss of RGC axons in the optic nerve as described across mammalian species must translate ipso facto to the corresponding decline in RGC bodies in the retina, whether RGC bodies in the retina are also susceptible to age-related loss (Calkins, 2013). The number of RGC bodies in the rat and mouse retina does not change with age though the retina itself enlarges and RGC shrink with a concomitant decrease in the density of IPL synapses (Harman et al., 2003; Samuel et al., 2011). In contrast, only the human retina appears to progress to actual RGC body loss with age; perhaps progression in normal aging depends on the actual extent of the lifetime (Calkins, 2013). Degu retina in this respect behaves like human retina, making it a suitable model for examining ganglion cell loss mechanisms. Altogether, these data reveal selective agerelated alterations in the neural circuitry in the degu retina. These changes in the degu retina seem more closely related to those observed in human retinal aging than alterations observed in other rodents, such as rats and mice. Animal models of retinal aging have usually employed nocturnal species (e.g., rats and mice), however, to better approximate the human retinal changes during aging, the diurnal rodent, Octodon degus is more useful, since the (i) rod/cone ratio is similar to the human ratio; (ii) the diurnal behavior is charactheristic to this species and its activity pattern resembles that of the human; (iii) its lifespan is considerably longer than those of the other experimental models, therefore the age-related changes can be better monitored and compared to those described in the human retina; and (iv) other, age-related diseases (e.g., cataract, diabetes, Alzheimer's disease, cancer) often appear spontaneously in this species (Jacobs, 1993; Brown and Donnelly, 2001; Chávez et al., 2003; Peichl et al., 2005; Ardiles et al., 2013; Palacios-Muñoz et al., 2014).

The proper functioning of the nervous system (including that of the retina) depends on the underlying structure of neural networks. Any loss of the pre- and/or postsynaptic profiles of the retinal neurons causes changes in their morphology and function. As a consequence, these retained neurons may be capable to establish new synaptic contacts (ectopic synapses, unusual synaptic arrangements), so the retina may undergo marked remodeling. In the aging degu retina, the synaptic rearrangements/alterations in the OPL and potentially

#### References


concomitant synaptic alterations in the IPL efficacy would reduce transmission, while the loss of function of the RPE cells would alter the homeostasis of the retinas. The neuronal elements of the vertical pathway such as rod bipolar and ganglion cells were seen affected already at 36-month of age. We think that these structural changes become more obvious, will reach other cells and will more seriously affect the synaptic complexes with advancing age.

The proven similarities between the degus and the human retina in their structure and aging processes offer a possibility to develop potential treatments and therapies for retinal age-related alterations and diseases.

# Authors Contribution

KS researched data and wrote, edited and reviewed the manuscript; CE, EF-V, ET, VI and GS Jr. researched data; RG and MTH researched data, reviewed and edited the manuscript; DR and AT reviewed the manuscript.

# Acknowledgments

The authors thank to Ms. D. Lopez and to Ms. C. M. Ros for their kind assistance. Supported by KTIA\_NAP\_13-1-2013- 0001 and MTA-PTE Momentum Program, AOK-KA-2013/14 to Mr. GS Jr. This work was also financially supported by Spanish Ministry of Science and Innovation (FIS PI13/01293), UJI (13I004) and partially the PharmaCog consortium by the European Community's Seventh Framework Programme for the Innovative Medicine Initiative under Grant Agreement no. 115009.

bassoon among ribbon synapses in the mammalian retina. Eur. J. Neurosci. 11, 3683–3693. doi: 10.1046/j.1460-9568.1999.00793.x


membrane and the choriocapillaris in the aged Fischer 344 rat. Curr. Eye Res. 31, 749–763. doi: 10.1080/02713680600864782


dual-phasing rodent, Octodon degus. J. Pineal Res. 42, 280–290. doi: 10.1111/j. 1600-079x.2007.00418.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.

Copyright © 2015 Szabadfi, Estrada, Fernandez-Villalba, Tarragon, Setalo, Izura, Reglodi, Tamas, Gabriel and Herrero. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution and 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.