# **GENOME-WIDE VIEW ON THE PHYSIOLOGY OF VITAMIN D**

**Topic Editor Carsten Carlberg**

#### *FRONTIERS COPYRIGHT STATEMENT*

© Copyright 2007-2014 Frontiers Media SA. All rights reserved.

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

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

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

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

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

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

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

**ISSN** 1664-8714 **ISBN** 978-2-88919-349-3 **DOI** 10.3389/978-2-88919-349-3

## *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

## **GENOME-WIDE VIEW ON THE PHYSIOLOGY OF VITAMIN D**

Topic Editor: **Carsten Carlberg,** University of Eastern Finland, Finland

Structural model of the full length RXR-VDR heterodimer. A surface representation of the RXR (blue) - VDR (green) complex on a DR3-type DNA binding site.

Cover by Ferdinand Molnar, modified from: Molnár F (2014) Structural considerations of vitamin D signaling. Front. Physiol. 5:191. doi: 10.3389/fphys.2014.00191

The main physiological actions of the biologically most active metabolite of vitamin D, 1α,25-dihydroxyvitamin D3 (1α,25(OH)2 D3 ), are calcium and phosphorus uptake and transport and thereby controlling bone formation. Other emergent areas of 1α,25(OH)2 D3 action are in the control of immune functions, cellular growth and differentiation. This fits both with the widespread expression of the VDR and the above described consequences of vitamin D deficiency. Transcriptome-wide analysis indicated that per cell type between 200 and 600 genes are primary targets of vitamin D. Since most of these genes respond to vitamin D in a cell-specific fashion, the total number of vitamin D targets in the human genome is far higher than 1,000. This is supported by the genome-wide view on VDR binding sites in human lymphocytes, monocytes, colon and hepatic cells. All genomic actions of 1α,25(OH)2 D3 are mediated by the transcription factor vitamin D receptor (VDR) that has been the subject of intense study since the 1980's. Thus, vitamin D signaling primarily implies the molecular actions of the VDR. In this research topic, we present in 15 chapters different

perspectives on the action of vitamin D and its receptor, such as the impact of the genomewide distribution of VDR binding loci, ii) the transcriptome- and proteome-wide effects of vitamin D, iii) the role of vitamin D in health, iv) tissue-specific functions of vitamin D and v) the involvement of vitamin D in different diseases, such as infections, autoimmune diseases, diabetes and different types of cancer.

# Table of Contents



Joerg Reichrath, Sandra Reichrath, Kristina Heyne, Thomas Vogt and Klaus Roemer


Rene F. Chun, Philip T. Liu, Robert L. Modlin, John S Adams and Martin Hewison


**EDITORIAL** published: 02 September 2014 doi: 10.3389/fphys.2014.00335

## The physiology of vitamin D—far more than calcium and bone

## *Carsten Carlberg\**

*School of Medicine, Institute of Biomedicine, University of Eastern Finland, Kuopio, Finland \*Correspondence: carsten.carlberg@uef.fi*

*Edited and reviewed by:*

*Geoffrey A. Head, BakerIDI Heart and Diabetes Institute, Australia*

**Keywords: vitamin D, vitamin D receptor, genomics, physiology, immune system**

Vitamin D is a molecule displaying an important physiological impact. Average human diet is neither rich in vitamin D2 (of plant origin) nor in vitamin D3 (of animal origin). Therefore, humans have to rely on the endogenous production of vitamin D3 in UVB exposed skin. This process was an import evolutionary driver for skin lightening after our ancestors decided some 100,000 years ago to move North out of Africa toward Asia and Europe (Juzeniene et al., 2009). Did this happen only to extract calcium efficiently from our diet and to keep our bones strong?

Vitamin D3 exerts most, if not all, of its physiological effects via its metabolite 1α,25-dihydroxyvitamin D3 (1,25(OH)2D3), which acts as a nuclear hormone, since it is the only high affinity ligand for the transcription factor vitamin D receptor (VDR). VDR is expressed in the majority of human tissues and cell types, i.e., at far more places than needed for calcium homeostasis and proper bone formation. The aim of this Research Topic is to explore the physiology of vitamin D from the perspective of the genomewide distribution of VDR-binding sites in cells as different as B lymphocytes, monocytes, macrophages, colon cancer cells, and hepatic stellate cells (Tuoresmäki et al., 2014). The choice of these cell types as experimental models is already a clear indication that the physiology of vitamin D involves also actions on the adaptive and innate immune system and on cancer cells.

This Research Topic starts with an overview on the recently explored genome-wide locations of VDR and their link to the accessibility of chromatin and its 3-dimensional organization (Carlberg, 2014). This genomic view is extended by a structural view on the interaction of the VDR with DNA, natural and synthetic ligands and co-regulatory proteins (Molnar, 2014). The VDR-mediated genome-wide actions of vitamin D result in a change of the transcriptome in all VDR expressing tissues and cell types. Taking all human tissues together this does not only affect thousands of protein-coding mRNAs but also a comparable number of non-coding RNAs (Campbell, 2014). The signal transduction of the lipophilic molecule 1,25(OH)2D3 is straightforward, since it reaches the VDR directly in the nucleus. Nevertheless, vitamin D signaling functionally interacts with a number of other signal transduction pathways, many of which start with receptors at the cell membrane (Larriba et al., 2014). The introductory section of this Research Topic is complemented by a view on the epigenome-wide effects of vitamin D, such as DNA methylation and histone modifications (Fetahu et al., 2014).

The general physiological function of vitamin D is to keep us healthy by promoting strong bones, properly functioning muscles and a potent immune system. When weather and season allows, we can keep our vitamin D levels up through endogenous production during carefully dosed exposure to sunlight (Reichrath et al., 2014). However, at winter above latitudes of 40◦ North or below 40◦ South insufficient amounts of UVB radiation pass the atmosphere. This implies that at least during winter we have to consider vitamin D as an essential micronutrient that we should supplement via fortified food compounds, such as milk and margarine, or appropriately dosed pills (Bendik et al., 2014). Both sun exposure in summer and supplementation during winter should keep our vitamin D status on an optimal level, which most likely is individual for each of us (Carlberg et al., 2013). Under these conditions cells of our innate and our adaptive immune system, such as monocytes and macrophage as well as B and T lymphocytes, can take maximal benefit from the gene regulatory potential of vitamin D (Chun et al., 2014).

In addition to the cells of the immune system VDR is expressed in most other tissues that origin from mesenchymal cells, such as bone (Van De Peppel and Van Leeuwen, 2014), myocytes (Polly and Tan, 2014), and adipose tissue (Mutt et al., 2014). This demonstrates that the well-known role of vitamin D in bone extrapolates to skeletal muscle and fat.

Most common diseases, such as type 2 diabetes, cancer and autoimmune diseases, are associated with chronic inflammation. Inflammation is mediated by tissue-associated macrophages, dendritic cells, and T lymphocytes, in which vitamin D has important gene regulatory functions (Wöbke et al., 2014). This may also be a key mechanism for the beneficial effects of vitamin D in cancers of breast (Narvaez et al., 2014) and prostate (Wang and Tenniswood, 2014). Furthermore, the pleiotropy of vitamin D suggests additional mechanisms for its anti-cancer effects, such as the modulation of intracellular metabolism. However, in case when supra-physiological concentrations of 1,25(OH)2D3 would be required, in order to obtain a therapeutic effect, the application of synthetic vitamin D analogs is suggested (Leyssens et al., 2014).

Taken together, the 15 chapters of this Research Topic present the wide physiological impact of vitamin D and link it to its molecular basis, which is the genome-wide action of the transcription factor VDR in most human tissues and cell types.

## **ACKNOWLEDGMENT**

The author acknowledges support by the Academy of Finland and the Sigrid Juselius Foundation.

## **REFERENCES**


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

*Received: 22 July 2014; accepted: 15 August 2014; published online: 02 September 2014. Citation: Carlberg C (2014) The physiology of vitamin D—far more than calcium and bone. Front. Physiol. 5:335. doi: 10.3389/fphys.2014.00335*

*This article was submitted to Integrative Physiology, a section of the journal Frontiers in Physiology.*

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

**REVIEW ARTICLE** published: 29 April 2014 doi: 10.3389/fphys.2014.00167

## *Carsten Carlberg\**

*School of Medicine, Institute of Biomedicine, University of Eastern Finland, Kuopio, Finland*

#### *Edited by:*

*Debra Diz, Wake Forest Universtiy School of Medicine, USA*

#### *Reviewed by:*

*Moray J. Campbell, Roswell Park Cancer Institute, USA Gregory A. Hawkins, Wake Forest University Health Sciences, USA*

#### *\*Correspondence:*

*Carsten Carlberg, School of Medicine, Institute of Biomedicine, University of Eastern Finland, Room 3179, Yliopistonranta 1 E, PO Box 1627, FI-70211 Kuopio, Finland e-mail: carsten.carlberg@uef.fi*

For a global understanding of the physiological impact of the nuclear hormone 1α,25-dihydroxyvitamin D3 (1,25(OH)2D3) the analysis of the genome-wide locations of its high affinity receptor, the transcription factor vitamin D receptor (VDR), is essential. Chromatin immunoprecipitation sequencing (ChIP-seq) in GM10855 and GM10861 lymphoblastoid cells, undifferentiated and lipopolysaccharide-differentiated THP-1 monocytes, LS180 colorectal cancer cells and LX2 hepatic stellate cells revealed between 1000 and 13,000 VDR-specific genomic binding sites. The harmonized analysis of these ChIP-seq datasets indicates that the mechanistic basis for the action of the VDR is independent of the cell type. Formaldehyde-assisted isolation of regulatory elements sequencing (FAIRE-seq) data highlight accessible chromatin regions, which are under control of 1,25(OH)2D3. In addition, public data, such as from the ENCODE project, allow to relate the genome-wide actions of VDR and 1,25(OH)2D3 to those of other proteins within the nucleus. For example, locations of the insulator protein CTCF suggest a segregation of the human genome into chromatin domains, of which more than 1000 contain at least one VDR binding site. The integration of all these genome-wide data facilitates the identification of the most important VDR binding sites and associated primary 1,25(OH)2D3 target genes. Expression changes of these key genes can serve as biomarkers for the actions of vitamin D3 and its metabolites in different tissues and cell types of human individuals. Analysis of primary tissues obtained from vitamin D3 intervention studies using such markers indicated a large inter-individual variation for the efficiency of vitamin D3 supplementation. In conclusion, a genome-wide (over)view on the genomic locations of VDR provides a broader basis for addressing vitamin D's role in health and disease.

**Keywords: vitamin D, vitamin D receptor, chromatin, gene regulation, epigenomics, genomics**

## **INTRODUCTION**

During evolution the secosteroid vitamin D3 became a pleiotropic signaling molecule (Jones et al., 1998). Initially, the molecule was used by early unicellular organisms to protect their DNA against UV-B irradiation (Holick, 2011). Far later, when the first fish with bones evolved, the endocrinology of vitamin D3 was established, and still is very conserved in all higher organisms, including humans (Bouillon and Suda, 2014). In this system, the energy of UV-B is used to convert 7-dehydrocholesterol into pre-vitamin D3, i.e., UV-B became essential for the synthesis of vitamin D3 (Holick, 2004) (more details in the article by Reichrath et al. in this issue). The central importance of this step is emphasized by the step-wise depigmentation of human skin, when modern humans started to move out of Africa some 100,000 years ago (Hochberg and Templeton, 2010). Two hydroxylation steps are necessary for the conversion of vitamin D3 via 25-hydroxyvitamin D3 (25(OH)D3) into the biologically active vitamin D3 metabolite, 1,25(OH)2D3 (Norman, 2008). The latter molecule participates in a large number of physiological processes, such as bone formation, immune function and cellular growth and differentiation (Deluca, 2004) (more details in the articles by van Leeuwen et al., Hewison et al. and Munoz et al. in this issue).

The transcription factor VDR is the only high-affinity target for 1,25(OH)2D3 within the cell nucleus (Haussler et al., 1997). VDR is one of approximately 1900 transcription factors, which are encoded by the human genome (Vaquerizas et al., 2009). In addition, VDR is a member of the superfamily of nuclear receptors, most of which are specifically activated by lipophilic molecules in the size of cholesterol (Carlberg and Molnár, 2012). Its lipophilic allows 1,25(OH)2D3 to pass through all biological membranes, i.e., gene regulation by vitamin D

**Abbreviations:** 1,25(OH)2D3, 1α,25-dihydroxyvitamin D3; 25(OH)D3, 25 hydroxyvitamin D3; ALOX5, arachidonate 5-lipoxygenase; CBS, cystathionine β-synthase; CCNC, cyclin C; CDKN1A, cyclin-dependent kinase inhibitor 1A; CHD7, chromodomain helicase DNA binding protein 7; ChIA-PET, chromatin interaction analysis by paired-end tag sequencing; ChIP, chromatin immunoprecipitation; ChIP-seq, ChIP coupled with massive parallel sequencing; CTCF, CCCTC-binding factor; CYP, cytochrome P450; DNase-seq, DNase I hypersensitivity sites sequencing; DR3, direct repeat spaced by 3 nucleotides; FAIRE-seq, formaldehyde-assisted isolation of regulatory elements sequencing; IGV, Integrative Genomics Viewer; LPS, lipopolysaccharide; LRP5, low density lipoprotein receptor-related protein 5; MYC, v-myc avian myelocytomatosis viral oncogene homolog; PBMC, peripheral blood mononuclear cell; RXR, retinoid X receptor; THBD, thrombomodulin; TNFSF11, tumor necrosis factor (ligand) superfamily, member 11; TRPV6, transient receptor potential cation channel, subfamily V, member 6; TSS, transcription start site; VDR, vitamin D receptor; ZMIZ1, zinc finger, MIZ-type containing 1.

does not involve additional signal transduction steps, as they are known for hydrophilic signaling molecules, such as peptide hormones, growth factors and cytokines. Moreover, VDR is rather ubiquitously expressed, i.e., most human tissues and cell types are responsive to 1,25(OH)2D3 (Wang et al., 2012).

VDR shares the main structural characteristics of nuclear receptors, which is a highly conserved DNA-binding domain and a structurally conserved ligand-binding domain (Mangelsdorf et al., 1995). VDR's DNA-binding domain specifically contacts the hexameric consensus sequence RGKTSA (R = A or G, K = G or T, S = C or G) within the major groove of genomic DNA (Shaffer and Gewirth, 2002). However, like most other transcription factors, VDR uses a partner DNA-binding protein, in order to bind efficiently to its target sites. More than 20 years ago, this heterodimeric partner turned out to be the nuclear receptor retinoid X receptor (RXR) (Sone et al., 1991; Carlberg et al., 1993). Steric constraints of the dimerizing DNA-binding domains of VDR and RXR determine the optimal binding site of the VDR-RXR complex as a direct repeat of two hexameric nuclear receptor binding motifs spaced by three nucleotides (DR3) (Umesono et al., 1991; Shaffer and Gewirth, 2004). Within VDR's ligandbinding domain, a network of some 40 mostly non-polar amino acids forms a ligand-binding pocket, in which 1,25(OH)2D3 and its synthetic analogs are specifically fixed with high affinity (Molnár et al., 2006). This ligand binding process induces a conformational change to the surface of VDR's ligand-binding domain, which results in a significant change of VDR's proteinprotein interaction profile: it transforms from a repressor to an activator (Moras and Gronemeyer, 1998; Carlberg and Campbell, 2013) (more details on VDR structure in the article by Molnar in this issue).

Taken together, vitamin D signaling primarily comprises the molecular actions of the VDR, i.e., the physiological effects of 1,25(OH)2D3 are largely identical to those of its receptor. This reduces vitamin D signaling to one central question: which are the most important genomic targets of VDR in a given tissue and which genes are controlled via these sites? Thus, this review focuses on the description of the genome-wide binding of VDR and its mechanistic implications. This analysis will be in the context of genome-wide information on chromatin accessibility and the presence of other nuclear proteins, such as provided by the ENCODE consortium.

## **GENOME-WIDE VDR BINDING**

The method chromatin immunoprecipitation (ChIP) was developed, in order to monitor the binding of transcription factors to their genomic targets (Orlando, 2000). The core of the method is (i) mild chemical cross-linking of living cells or tissues, e.g., with 1% formaldehyde, in order to fix nuclear proteins to genomic DNA, (ii) sonication of chromatin into small (200–400 bp) fragments, and (iii) immunoprecipitation with an antibody specific for the chosen nuclear protein (Maston et al., 2012). In this way, chromatin regions, which, at the moment of cross-linking, had been in contact with the protein of choice, are specifically enriched. A specific ChIP signal, in reference to a control (often unspecific IgGs), is a strong indication that the protein of choice had been in contact with the selected genomic region at the moment of cross-linking.

At earlier times, the isolated chromatin template was analyzed by site-specific quantitative PCR (ChIP-qPCR). This approach had been used to study, for example, the extended promoter regions of the primary VDR target genes *CYP24A1* (Väisänen et al., 2005), *CYP27B1* (Turunen et al., 2007), *CCNC* (Sinkkonen et al., 2005), and *CDKN1A* (Saramäki et al., 2006, 2009). Alternatively, the abundance of immunoprecipitated chromatin fragments had been detected by tiled microarrays (so-called "chips,") which covered a selection of promoter and enhancer regions or any other subset of the genome (ChIP-chip). The group of Pike et al. had extensively used ChIP-chip, in order to locate VDR binding sites within the regulatory regions of the mouse genes *Vdr* (Zella et al., 2006), *Trpv6* (Meyer et al., 2006), *Lrp5* (Fretz et al., 2007), *Tnfsf11* (also known as *Rankl)* (Kim et al., 2006), *Cyp24a1* (Meyer et al., 2010), and *Cbs* (Kriebitzsch et al., 2011). The latest development of the ChIP method is the unbiased analysis of the precipitated chromatin by massively parallel DNA sequencing (ChIP-seq), i.e., the detection of the binding sites of the transcription factor of choice in the complete genome. To date, ChIP-qPCR is primarily used for the confirmation of ChIP-seq results, while ChIP-chip got outdated shortly after its introduction. This leaves, at present, ChIP-seq as the method of choice for analyzing VDR's genomic binding loci.

At present, the readouts of massive parallel sequencing are small sequence tags (35–50 nucleotides), but in the future there will be in majority longer reads used, which will lead to improved significance of the results. These sequence tags are aligned to a reference genome (for human samples this is, at present, hg19) and specifically represent the enriched chromatin fragments. Then "peak calling" software is used to identify genomic regions, in which significantly more sequence tags are detected than in control reactions. Therefore, tags that accumulate as "peaks" at specific genomic loci mark the presence of the investigated nuclear protein (Park, 2009; Furey, 2012). At present, ChIP is still performed with millions of cells; in case of a prominent binding site, most of these cells contribute to the ChIP signal, i.e., it can be assumed that in the majority of cells the locus is occupied by VDR. However, when only in some cells a site is bound by VDR, the respective peak is far less prominent, i.e., most likely of less impact for the regulation of 1,25(OH)2D3 target genes.

To date, VDR ChIP-seq data are available from (i) the immortalized lymphoblastoid cell lines GM10855 and GM10861 (Ramagopalan et al., 2010), (ii) undifferentiated THP-1 monocyte-like cells (Heikkinen et al., 2011), (iii) lipopolysaccharide (LPS)-polarized THP-1 macrophage-like cells (Tuoresmäki et al., 2014), (iv) LS180 colorectal cancer cells (Meyer et al., 2012), and (v) LX2 hepatic stellate cells (Ding et al., 2013). The original publications reported between 1600 and 6200 VDR binding sites (in ligand-stimulated samples) within the human genome. However, these numbers are not directly comparable, since different peak calling software, alternative threshold settings and even an older version of the reference genome (hg18) were used. A harmonized re-analysis of all six VDR ChIP-seq datasets with identical peak calling settings (MACS, version 2) resulted for 1,25(OH)2D3-stimulated and unstimulated cells, respectively, in following number of binding sites: 6172 and 3144 (GM10855), 12,353 and 4072 (GM10861), 774 and 609 (undifferentiated THP-1), 953 and 529 (LPS-differentiated THP-1), 3777 and 165 (LS180) and 1532 and 1474 (LX2) (Tuoresmäki et al., 2014).

In total, the six VDR ChIP-seq datasets indicated 21,776 nonoverlapping VDR binding sites when allowing a distance of up to 250 bp between the peak summits (Tuoresmäki et al., 2014). However, the vast majority of these VDR loci (67%) are unique for one of the analyzed cellular models. In contrast, under the above mentioned conditions only 54 sites are common within all six datasets. In general, this indicates that VDR displays a very individual pattern of cell-specific genomic locations, which overlaps between multiple tissues only at key sites. The VDR binding site of the 1,25(OH)2D3 target gene *ZMIZ1*, which is located 15.3 kb downstream of the transcription start site (TSS), represents an example of such a locus (**Figure 1**). In general, the rates of overlaps between the cell types follow roughly their developmental and functional relatedness, i.e., the two lymphoblastoid cell lines, GM10855 and GM10861, or LPS-differentiated and undifferentiated THP-1 cells show more overlapping VDR binding sites than all other comparisons between the VDR ChIP-seq datasets. Moreover, the VDR binding profiles of ligand-stimulated cells matched better than those of unstimulated cells (Tuoresmäki et al., 2014).

Genome-wide studies on VDR binding have changed the view on vitamin D signaling. The few dozens rather well characterized VDR binding sites in less than 10 kb distance to the TSS of 1,25(OH)2D3 target genes (Haussler et al., 2013), which were known before, were complemented by thousands of additional VDR loci spread over the whole genome. However, the very most of the loci, which were highlighted by ChIP-seq, have not yet been validated by ChIP-qPCR or similar methods (and many will never be confirmed). Some previously known VDR binding sites, such as those controlling the genes *MYC* (Toropainen et al., 2010), *VDR* (Zella et al., 2010), *CCNC* (Sinkkonen et al., 2005), and *ALOX5* (Seuter et al., 2007), could be confirmed by the VDR ChIP-seq datasets. However, for many known 1,25(OH)2D3 target genes the ChIP-seq data suggest additional or alternative VDR binding sites, many of these being far more distant to the gene's TSS region than previously foreseen. In the past, many of these VDR binding sites had been overlooked due to a focus to only a few kb upstream of the TSS of 1,25(OH)2D3 target genes. However, in accordance with the results of the ENCODE project (ENCODE-Project-Consortium et al., 2012), VDR binding sites are found with equal probability upstream and downstream of the TSS region of 1,25(OH)2D3 target genes. In addition, VDR loci in distance of even more than 1 Mb from the gene's TSS are accepted as regulatory sites (more details below).

In summary, there seem to be 1000–10,000 genomic VDR binding sites per cell type. This is far more than the number of primary 1,25(OH)2D3 target genes, which is in the order of 100-500 per tissue. This even holds true for undifferentiated THP-1 cells, where 774 VDR loci in ligand-stimulated cells are facing 408 statistically significantly up-regulated early 1,25(OH)2D3 responding genes (Heikkinen et al., 2011). The indicates that some genes are controlled by more than one VDR binding site, i.e., they may have a higher potential to be regulated by 1,25(OH)2D3 than target genes with only one active VDR locus (more details on the transcriptome-wide response to 1,25(OH)2D3 in the article by Campbell et al. in this issue).

## **MECHANISTIC INSIGHT FROM VDR ChIP-SEQ STUDIES**

The close to 22,000 non-overlapping VDR peaks, which are indicated by the public ChIP-seq datasets (Tuoresmäki et al., 2014), show rather different characteristics. Despite the rather different total number of reported VDR peaks per cellular model, each of the six ChIP-seq datasets contains an in part overlapping subset of less than 200 sites, where a stimulation with 1,25(OH)2D3 resulted in a significant increase of VDR binding compared to unstimulated samples. These VDR loci are far more prominent than most of the other sites, for which ligand treatment was either repressive, had no effect or was only minor stimulatory.

Another important parameter for the characterization of a VDR binding site is the presence or absence of a high confidence DR3-type binding site below the summit (±100 bp) of the respective ChIP-seq peak. This can be investigated with the help of binding site screening algorithms, such as provided by HOMER (Heinz et al., 2010). Depending on the threshold settings the software detects binding sites that deviate more or less from the consensus sequence. For example, for a moderate setting of a HOMER score of 7, from the total of 21,776 non-overlapping VDR sites in all six ChIP-seq datasets only 3801 (17.5%) contain a DR3-type sequence. Interestingly, the percentage of DR3-type motifs differs significantly between the datasets and ranges from 38.2% (483 of 1264 sites) in LPS-polarized THP-1 cells via 36.4% (373 of 1023) in undifferentiated THP-1 cells, 28.6% (1062 of 3706) in LS180 cells, 27.8% (611 of 2194) in LX2 cells, 13.0% (909 of 6975) in GM10855 cells to 9.0% (1118 of 12,438) in GM10861 cells. This indicates that the total number of identified VDR binding sites in each cell line inversely correlates with the percentage of peak summits with DR3-type sites. However, when the analysis is restricted to the top 200 VDR sites (based on fold enrichment scoring), for all six ChIP-seq datasets a DR3-like sequence rate of more than 60% is observed, i.e., DR3 motifs are found preferentially at highly ligand responsive VDR loci. In this way, the different VDR ChIP-seq datasets show a very similar relationship between VDR occupancy and DR3 percentage. This suggests that the mechanistic basis for the action of the VDR is independent of the cell type and the total number of identified binding sites.

Transcription factor binding site screening software, such as HOMER, suggests that DR3-type binding sequences are the most abundant sites below the summits of VDR ChIP-seq peaks. However, a significant number of the genomic VDR loci (depending on the dataset 60-90% of all, see above) do not associate with a DR3-type site. This indicates that at these loci VDR uses a different mode of interaction with genomic DNA. This could be either the use of a different heterodimeric binding partner or an indirect binding "backpack" of a DNA-binding transcription factor (Carlberg and Campbell, 2013). In both scenarios the specific DNA binding site would be different to a DR3-type sequence. Interestingly, for the VDR ChIP-seq datasets originating from hematopoietic cells, HOMER indicated binding sites for the transcription factors PU.1 (also called SPI1), ESRRB (also called NR3B2) and GABPA as significantly enriched (Tuoresmäki

et al., 2014). PU.1 is well-known as a pioneer factor (Zaret and Carroll, 2011), i.e., as a transcription factor with (i) a high number of genomic binding sites, (ii) a greater binding promiscuity and (iii) higher diversity of interactions. Pioneer factors are the first that bind regulatory genomic regions, such as promoters and enhancers, and interact with chromatin modifying enzymes, in order make the chromatin more accessible for regular transcription factors, such as VDR. At present, a direct protein-protein interaction of VDR with PU.1, ESRRB or GABPA has not been demonstrated, but for the Ets family, to which PU.1 belongs, there were indications for an interaction (Tolon et al., 2000). However, for the pure function as a pioneer factor a direct protein-protein interaction with the "settler factor" is not needed. Moreover, there is older evidence from single gene studies that DNA binding of VDR is modulated by the transcription factors AP1 (Schüle et al., 1990) and RUNX2 (Sierra et al., 2003). In contrast, a genomewide study on the interaction of VDR with the transcription factor TCF7L2 did not provide any evidence that the latter acts as a pioneer factor for VDR (Meyer et al., 2012).

Below VDR peak summits no dominant non-DR3 binding sequence could be identified. Moreover, the six VDR ChIP-seq datasets differ in the ranking and identity of the non-DR3 sites found below the peaks (Tuoresmäki et al., 2014). This suggests that in total there must be a larger number of VDR partnering proteins. Most likely, these proteins have a cell-specific expression pattern and may explain in part the cell-specific actions of VDR and its natural ligand 1,25(OH)2D3. Moreover, ChIP-seq datasets have indicated that, in contrast to steroid receptors, VDR binds a number of its genomic targets already in the absence of ligand. These ligand-independent genomic VDR loci have a clearly lower rate of DR3-type sequences than ligand-dependent sites (Heikkinen et al., 2011). In contrast, they associate preferentially with proteins related to gene repression, such as demonstrated for the example of the *CYP27B1* gene (Turunen et al., 2007). This implies that the functional profile of VDR is larger than that of its ligand (Polly et al., 2000) as previously shown for other members of the nuclear receptor superfamily, such as thyroid hormone receptor or liver X receptor (Perissi et al., 2010).

Taken together, all VDR ChIP-seq studies confirm the preferential binding of VDR to DR3-type sequences. However, only one in six of some 22,000 presently known VDR loci within the human genome carry a DR3 site. Thus, there have to be additional mechanisms for the association of VDR with its genomic loci, which may include partnering with presently undefined partner proteins or the tethering to other DNA-binding transcription factors, such as pioneer factors. These should explain some of the cell-specific actions as well as repressive functions of 1,25(OH)2D3 and the VDR.

## **RESPONSES OF CHROMATIN TO 1,25(OH)2D<sup>3</sup>**

Histone proteins forming the nucleosome core are DNAbinding proteins but do not show any sequence specificity. Therefore, the complex of nucleosomes and genomic DNA, which is referred to as chromatin, has an intrinsic repressive potential: it prevents access of transcription factors to their genomic targets (Razin, 1998). This provides essential stability to the epigenetic landscape for long-lasting regulatory decisions, such as gene expression in terminally differentiated cells (Mohn and Schubeler, 2009). In contrast, some regions of the epigenome show highly dynamic changes in response to extra- and intracellular signals, such as the activation of VDR by 1,25(OH)2D3 binding (Talbert and Henikoff, 2006). These changes involve the methylation of genomic DNA and/or reversible post-translational modifications of histone proteins, such as acetylation or deacetylation at exposed lysine residues (Narlikar et al., 2002). Dynamic chromatin modifications change the access to regulatory genomic regions, such as promoter and enhancers, for the binding of transcription factors, i.e., they determine whether at these regions chromatin is open or closed. This can be monitored genome-wide by using the method DNase I hypersensitive sites sequencing (DNase-seq), which highlights genomic regions being most sensitive to cleavage by the enzyme DNase I (Crawford et al., 2006). A very similar technique is Formaldehyde-Assisted Isolation of Regulatory Elements sequencing (FAIRE-seq), which identifies genome-wide accessible DNA regions (Giresi et al., 2007) (more details on the relation of the epigenome and 1,25(OH)2D3 in the article by Kallay et al. in this issue).

At present, the only publically available dataset describing genome-wide effects of 1,25(OH)2D3 on the epigenome, is a detailed FAIRE-seq time course in THP-1 cells (Seuter et al., 2013). These data demonstrate that some 87% of the more than 1000 VDR binding sites in this cellular model co-localize with open chromatin. Interestingly, at 165 of these VDR loci a strong 1,25(OH)2D3-dependent increase of chromatin accessibility is found. Importantly, at 66% of these chromatin regions a DR3 type sequence is found, i.e., they overlap with loci, at which VDR binding is enhanced most by 1,25(OH)2D3 stimulation (Seuter et al., 2013). Moreover, the binding of VDR to its genomic loci is a dynamic process, which takes at least some 2 h to saturate the sites. One example is a site located 225 kb downstream of the TSS of the chromodomain helicase DNA binding protein 7 (*CHD7*) gene (**Figure 2**). It demonstrates that at the same locus, where a strong ligand-dependent increase of VDR binding is observed, the rate of open chromatin more than doubled already 40 min after incubation of THP-1 cells with 1,25(OH)2D3. At some 200 additional VDR binding loci the chromatin shows detectable but less prominent response to 1,25(OH)2D3 treatment, while at the remaining 500 sites the VDR ligand did not affect chromatin accessibility. Accordingly, only at less than 20% of the latter sites DR3-type sequences are found. At many of these sites, VDR binds already in the absence of ligand and may have a different mode of DNA recognition and action (see above).

In summary, at approximately a third of its genome-wide binding loci VDR dynamically controls the epigenetic state of chromatin. At these sites, VDR binding and chromatin opening are tightly interconnected and provide indications for primary 1,25(OH)2D3 target genes. This allows a better understanding of the 1,25(OH)2D3 signaling cascade.

## **THE USE OF ENCODE DATA FOR UNDERSTANDING 1,25(OH)2D<sup>3</sup> SIGNALING**

In addition to the above described 1,25(OH)2D3-triggered chromatin open, in the future there will be much more data available on the interaction of VDR and 1,25(OH)2D3 with the epigenome. This will include FAIRE-seq and DNase-seq studies in further 1,25(OH)2D3-responsive tissues as well as investigations on changes of histone modifications and/or co-factor binding in 1,25(OH)2D3-responsive genomic regions. However, already at present existing genome-wide data on the annotation of the genomes of human, mouse and other species can be used. The best example is the large range of data collected by the ENCODE consortium (ENCODE-Project-Consortium et al., 2012). The core of the ENCODE datasets are publically available ChIP-seq results on approximately 100 transcription factors and 20 histone modifications from more than 100 human cellular systems. From the latter, the human monocytic leukemia cell line K562 is represented with highest number of datasets, while the majority of the other cells has not been studied with the same intensity. At present, ENCODE data describe primarily the basal status of cells, i.e., only in a very few cases a stimulation with hormones, growth factor, cytokines or similar molecules had been performed. Neither data on 1,25(OH)2D3 stimulations nor VDR ChIP-seq data are comprised in the ENCODE dataset. Nevertheless, the examples shown below will illustrate, how already on this stage ENCODE data are useful for a more detailed understanding of 1,25(OH)2D3 signaling.

All six VDR ChIP-seq datasets agree with observations of the ENCODE project that (i) transcription factors bind equally likely both up- and downstream of their target gene TSSs and (ii) the likelihood of detecting functional transcription factor binding sites for a given gene decreases by distance from its TSS region (ENCODE-Project-Consortium et al., 2012). This means that, in relation to the TSS of primary 1,25(OH)2D3 target genes, the distribution of the VDR binding sites has a Gaussian shape. In turn, this suggest that on the same chromosome there would be no threshold distance for the interaction between a VDR binding locus and the TSS of a primary 1,25(OH)2D3 target gene. However, there are limitations provided by higher-order structures of chromatin.

Chromatin forms loops (Kadauke and Blobel, 2009), which contribute to many nuclear functions, such as the control of gene expression (Misteli, 2007). Chromatin loops segregate each chromosome into domains, which are separated by an insulator region (Van Bortle and Corces, 2013). Most insulator regions are associated with the highly conserved transcription factor CCCTCbinding factor (CTCF) (Schmidt et al., 2012). Therefore, ChIPseq data for CTCF binding from multiple human cell lines, such as provided by ENCODE (ENCODE-Project-Consortium et al., 2012), allow a first estimation of the chromatin domain borders (**Figure 3**). However, only 15–20% of all genomic CTCF binding sites are involved in insulator function. The method chromatin interaction analysis by paired-end tag sequencing (ChIA-PET)

(Fullwood et al., 2009) allows an assessment of the 3-dimensional structure of chromatin. When applied for CTCF in K562 cells it mapped more than 120,000 intra-chromosomal, CTCF-mediated chromatin interactions (ENCODE-Project-Consortium et al., 2012). The high conservation of CTCF binding sites allows a reliable extrapolation of the CTCF ChIA-PET data from K562 cells to THP-1 cells, for which VDR ChIP-seq data is available. The combination of both datasets suggests that in THP-1 cells there are some 1600 chromatin domains, which contain at least one VDR binding site (Seuter et al., 2014). When the TSS region of a gene is within one of these chromatin regions, it may be a primary 1,25(OH)2D3 target. In case of the *CD14* gene, CTCF ChIA-PET data from K562 cells defined a chromatin domain spanning from 1.5 kb upstream to 57 kb downstream of the gene's TSS (**Figure 3**). This domain spans over the whole *CD14* gene and comprises two VDR binding sites 24 and 26 kb downstream of the gene's TSS. This provides a straightforward gene regulatory scenario explaining (i) the primary response of *CD14* to 1,25(OH)2D3 and (ii) why the neighboring genes of *CD14* do not respond to VDR ligand treatment.

The chromatin domain of the *CD14* gene is with a size of less than 60 kb rather small (**Figure 3**). In contrast, one of the largest VDR-containing chromatin domains spans over 2.3 Mb of chromosome 8 and combines the *MYC* gene with four VDR binding sites, of which the most prominent is 1215 kb of the gene's TSS (Ryynänen et al., 2013). This suggests that under the condition of sufficiently large chromatin domains, gene regulation by VDR can be over a distance of more than 1 Mb.

At present, there is no VDR ChIA-PET data available but will come in the future. However, via the ENCODE experiment matrix (http://encodeproject*.*org/ENCODE/dataMatrix/ encodeDataMatrixHuman*.*html) there is access to ChIA-PET data for RNA polymerase II in HeLa human cervix carcinoma and MCF-7 human breast carcinoma and for estrogen receptor α in MCF-7 cells. The latter may be of special interests for the breast cancer field.

Taken together, public ENCODE data are important tools, which can be used in combination with genome-wide data on VDR for an extrapolation on the 3-dimensional organization of gene regulation by 1,25(OH)2D3.

## **1,25(OH)2D<sup>3</sup> TARGET GENES AS BIOMARKERS FOR THE VITAMIN D STATUS OF HUMAN INDIVIDUALS**

In contrast to a number of other nuclear receptor ligands, such as cortisol or estrogen, the endocrinology of 1,25(OH)2D3 does not imply any fast changes (Deluca, 2004; Norman, 2008). Under normal circumstances, either the production in UV-B exposed skin or the intake of from diet or supplements should provide sufficient amounts of vitamin D3, in order to achieve optimal serum 25(OH)D3 concentrations. The latter vitamin D metabolite is the widely accepted indicator of the vitamin D3 status of the human body (Hollis, 2005). The serum 25(OH)D3 concentrations change only in the order of weeks and months, such as the result of seasonal variations in sun exposure (Virtanen et al., 2011). This indicates that stimulation experiments with 1,25(OH)2D3 over a few hours, as performed in *in vitro* experiments, do not represent

the physiological reality. In contrast, the effects of more longlasting changes of serum 25(OH)D3 concentrations should be considered. On a genome-wide level, this was investigated first with primary T cells isolated from nine human individuals with variant serum 25(OH)D3 concentrations (Handel et al., 2013). The number of the observed VDR ChIP-seq peaks, which varied between 200 and more than 7000, correlated with the 25(OH)D3 levels of the individuals, i.e., the higher the circulating 25(OH)D3 concentrations, the more VDR loci were identified in T cells. Unfortunately, the raw data of this study is not available, i.e., a harmonized re-analysis in comparison with other published VDR ChIP-seq data cannot be performed. However, from the 14,044 unique VDR peaks reported for the sum of the nine individuals, only 442 (3.1%) associated with a DR3-type sequence (based on HOMER score 7 settings).

Serum 25(OH)D3 concentrations vary widely from person to person based on (i) varied diet and sun exposure, (ii) different age and/or level of adiposity and (iii) genetic and epigenetic variations (Engelman et al., 2008; Orton et al., 2008; Snellman et al., 2009). The Institute of Medicine recommends a serum 25(OH)D3 level of 50 nM (Institute-of-Medicine, 2011), but it is under debate, whether this is sufficient for every individual (Holick, 2007). In fact, a substantial proportion of the world's population could be considered as vitamin D deficient. This condition may accelerate age-related bone loss and morbidity from falls and fractures. In addition, vitamin D insufficiency is associated with a number of diseases, such as cancer, autoimmune disorders and all components of the metabolic syndrome (more details in the article by Bendik et al. in this issue).

This important medical problem guided to the question, whether an insight into the genome- and transcriptome-wide actions of VDR and 1,25(OH)2D3 can help in a more accurate evaluation of the human individual's responsiveness to, and needs for, vitamin D. A first approach in this direction was done by studying peripheral blood mononuclear cells (PBMCs) and adipose tissue biopsies from 71 elderly, pre-diabetic individuals, which participated in a 5-month vitamin D3 intervention trial (VitDmet) during Finnish winter (Carlberg et al., 2013). The changes in the mRNA expression of the primary 1,25(OH)2D3 target genes *CD14* and thrombomodulin (*THBD*), which had been identified in a recent comparative study as most reliable biomarkers (Standahl Olsen et al., 2013), in both PBMCs and fat samples were correlated with the alterations in the serum 25(OH)D3 levels of the 71 individuals. Interestingly, only for a subset of individuals significant correlations between the up-regulation of both genes and the intervention-induced raise in serum 25(OH)D3 concentrations were obtained. This suggests that, on a molecular level, not all study participants benefited from the vitamin D3 supplementation, because (i) they had already reached their individual optimal vitamin D status before the start of the intervention, (ii) they carry a genetic polymorphism making them less responsive to vitamin D3 or (iii) other undefined reasons (Carlberg et al., 2013). Interestingly, the categorization of the human individuals by their vitamin D responsiveness unmasked a negative correlation between changes in serum concentrations of 25(OH)D3 and the inflammation marker interleukin 6, i.e., the more responsive the study participants were to vitamin D3 supplementation, the lower was their inflammatory status. At present, a number of other primary 1,25(OH)2D3 target genes, which were highlighted in the comparison of VDR ChIP-seq data, are evaluated for their potential to serve as even better biomarkers for the vitamin D status of human individuals than *CD14* and *THBD*.

In summary, vitamin D deficiency may negatively contribute to a number of diseases. Genome-wide insight led to the use of mRNA expression changes of the genes *CD14* and *THBD* as biomarkers for a molecular evaluation of vitamin D3 supplementation studies. The results allow a classification of human individuals based on their responsiveness to vitamin D3.

#### **CONCLUSIONS**

Genome-wide data on (i) the location of transcription factor binding sites in living cells, (ii) histone modifications and (iii) accessible chromatin, such as provided by ChIP-seq, DNase-seq and FAIRE-seq studies, have significantly changed the view on, and the understanding of the regulation of the entirety of the genes of our genome. This applies also to the transcription factor VDR, for which at present ChIP-seq data from six human cell lines and the T cells of nine human individuals are available. The abovementioned modern genome-wide techniques allow a more unbiased identification of transcription factor binding sites compared to previous studies, which were mostly focused on regions a few kb upstream of a primary 1,25(OH)2D3 target gene. VDR binding loci have now been shown to be localized equally likely up- and downstream of TSS regions in distances of even more than 1 Mb.

Genome-wide studies have confirmed DR3-type sequences as the preferential binding sites for VDR (most likely as a heterodimer with RXR), but only one in seven of the close to 20,000 known VDR binding loci carry such a motif. This is unanimously observed in all investigated cellular models. Therefore, there must be other types of binding motifs and partnering proteins that attract VDR to its genomic targets. These presently poorly understood alternative binding modes may explain some of VDR's function in the (trans)repression of its target genes. Moreover, the VDR cistrome seems to be largely cell-specific with only some 50 loci overlapping in all investigated models. However, these conserved sites could be fundamental entry ports of VDR to the human genome, which may serve as the unified core of the various pleiotropic functions of 1,25(OH)2D3.

The 22,000 detected VDR binding loci within six cell lines as well as the 14,000 peaks found in primary T cells from nine human individuals may be far more than what is needed to control the physiological actions of 1,25(OH)2D3, i.e., many sites may represent rather "noise" than having a specific function. Therefore, different approaches to categorize VDR loci are useful. It turned out that VDR binding sites that (i) carry a DR3-type sequence, (ii) show ligand-stimulated VDR association, (iii) co-locate with ligand-induced chromatin opening and (iv) are conserved between several cellular systems may play a more important role in mediating the functions of 1,25(OH)2D3 than the vast majority of other VDR sites that lack most of these properties. Therefore, for a cellular system of interest, the combination of (i) a genome-wide assessment of open chromatin by DNase-seq or FAIRE-seq, (ii) the monitoring of genomic VDR loci by ChIPseq and (iii) a screening for DR3-type sequences below the peak summits is an efficient tool for the prediction and identification of primary 1,25(OH)2D3 target genes.

## **FUTURE PERSPECTIVES**

Although historically 1,25(OH)2D3 was understood to be a hormone controlling calcium homeostasis and bone formation, to date the genome-wide the actions of VDR are best understood and monitored in cells of the hematopoietic system. This emphasizes the impact of 1,25(OH)2D3 on the function of innate and adaptive immunity. There are first indications that the core actions of 1,25(OH)2D3 and its receptor VDR can be extrapolated from hematopoietic cells to other tissues and cell types of the human body (Carlberg et al., 2013). If this holds true, the vitamin D status and responsiveness of a human individual can be derived from the response of, for example, PBMCs. Technically, various types of leukocytes can be collected far easier from blood samples than any other tissue biopsy. Like a glucose tolerance test is used to monitor the functionality of the carbohydrate metabolism of an individual, there may be in the future a higher dose vitamin D3 challenge test, where the (epi)genomic and transcriptomic profiles of leukocytes before and after supplementation are measured. Routine measurements of healthy individuals may also in the future more likely be based on a few selected biomarkers, such as *CD14* and *THBD*, while more complex scenarios in disease settings, such as cancer or autoimmune diseases, will be assessed genome- or transcriptome-wide. In the same way, basic research on 1,25(OH)2D3 and VDR will shift more and more from cell culture models to primary tissues and cell types and will eventually reach the single cell level.

## **ACKNOWLEDGMENTS**

The author thanks Dr. Sabine Seuter for critical reading of the manuscript and the Academy of Finland and the Juselius Foundation for support.

#### **REFERENCES**


of 25-hydroxyvitamin D and 1,25-dihydroxyvitamin D levels in Hispanic and African Americans. *J. Clin. Endocrinol. Metab.* 93, 3381–3388. doi: 10.1210/jc.2007-2702


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

*Received: 14 February 2014; accepted: 10 April 2014; published online: 29 April 2014. Citation: Carlberg C (2014) Genome-wide (over)view on the actions of vitamin D. Front. Physiol. 5:167. doi: 10.3389/fphys.2014.00167*

*This article was submitted to Integrative Physiology, a section of the journal Frontiers in Physiology.*

*Copyright © 2014 Carlberg. 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.*

## Structural considerations of vitamin D signaling

## *Ferdinand Molnár\**

*Faculty of Health Sciences, School of Pharmacy, Institute of Biopharmacy, University of Eastern Finland, Kuopio, Finland*

#### *Edited by:*

*Carsten Carlberg, University of Eastern Finland, Finland*

#### *Reviewed by:*

*Dragomir N. Serban, Grigore T. Popa University of Medicine and Pharmacy, Romania Natacha Rochel, Institut Génétique Biologie Moléculaire Cellulaire, France*

#### *\*Correspondence:*

*Ferdinand Molnár, Faculty of Health Sciences, School of Pharmacy, Institute of Biopharmacy, University of Eastern Finland, Yliopistonranta 1C, Canthia 2031, 70210 Kuopio, Finland e-mail: ferdinand.molnar@uef.fi; ferdinand.molnar@gmail.com*

Crystal structures represent the static picture in the life of a molecule giving a sneak preview what it might be in reality. Hence, it is very hard to extrapolate from these photos toward dynamic processes such as transcriptional regulation. Mechanistically VDR may be considered as molecular machine able to perform ligand-, DNA- and protein recognition, and interaction in a multi-task manner. Taking this into account the functional net effect will be the combination of all these processes. The long awaited answer to explain the differences in physiological effects for various ligands was one of the biggest disappointment that crystal structures provided since no substantial distinction could be made for the conformation of the active VDR-ligand complexes. This may have come from the limitation on the complexity of the available ligand-VDR structures. The recent studies with full length VDR-RXRα showed somewhat more comprehensive perspective for the 3D organization and possible function of the VDR-RXRα-cofactor complex. In addition to *in vitro* approaches, also computational tools had been introduced with the aim to get understanding on the mechanic and dynamic properties of the VDR complexes with some success. Using these methods and based on measurable descriptors such as pocket size and positions of side chains it is possible to note subtle differences between the structures. The meaning of these differences has not been fully understood yet but the possibility of a "butterfly effect" may have more extreme consequences in terms of VDR signaling. In this review, the three functional aspects (ligand-, DNA- and protein recognition, and binding) will be discussed with respect to available data as well as possible implication and questions that may be important to address in the future.

**Keywords: VDR, crystal structure, molecular dynamics, molecular mechanism, cofactors, response elements, ligand-binding**

## **INTRODUCTION**

One way of understanding life at molecular level is to obtain the three-dimensional (3D) structures of the molecules. Such structural views represent a static picture in the life of a molecule giving a sneak preview what it might be in reality. For the understanding of the functional implication of vitamin D (VD) signaling it is also important to look at various structural complexes of the vitamin D receptor (VDR), which may outline its possible dynamics and mechanics. VDR is able to perform the ligand-, DNA- and protein recognition, and interaction in a multi-task manner thus can be viewed as molecular machine which will regulate gene expression with the combination/sum of all these particular functions.

Before the year 2000 people in the VD field were only guessing how may the 3D structure of VDR look like. Some implications were coming from already known crystal structures e.g., receptors for retinoids such as retinoid X receptor (RXR) (Bourguet et al., 1995) and retinoic acid receptor (RAR) (Renaud et al., 1995). Not until exactly 14 years ago the structure of VDR-1α,25 dihydroxyvitamin D3 (1,25D3) complex has been solved (Rochel et al., 2000) and a long journey started in understanding the binding of various VDR analogs and the structure-based analog design. Within the first 5 years more structures have been solved (Tocchini-Valentini et al., 2001, 2004; Eelen et al., 2005) but the long awaited answer to explain the differences in physiological effects for various ligands was one of the biggest disappointment that crystal structures provided since no substantial distinction could be made for the conformation of the active VDR-ligand complexes.

Compared to the beginning of the last decade a huge number of X-ray crystal structures are available for VDR. In detail, there are VDR LBDs from three different species *H. sapiens* (34), *R. norvegicus* (40) and *D. rerio* (13) and four DBD-DNA complexes from *H. sapiens*. The basic information about these complexes is summarized in **Table 1**. This data makes also possible to analyze orthologous molecules with reflection to functional and structural differences. However, to understand this aspect well it would be beneficial to have more data from numerous organisms. In this review evolutionary aspects and species-specific difference will be not discussed in depth. More space will be given to

**Abbreviations:** 1,25D3, 1α, 25-dihydroxyvitamin D3; 3D, three-dimensional; 3Klca, 3-keto lithocholic acid; 9-*cis* RA, 9-*cis* retinoic acid; aa, amino acids; cryo-EM, cryo-electron microscope; *Cyp27b1*, 25-hydroxyvitamin D3 1α-hydroxylase; *D.*rerio, Danio rerio; DBD, DNA-binding domain; DNA, deoxyribonucleic acid; DR, direct repeat; DRIP, vitamin-D-receptor interacting protein; h, human; *H. sapiens, Homo sapiens;* HDX, H/D exchange; LBD, ligand-binding domain; LBP, ligand-binding pocket; LCA, lithocholic acid; MD, molecular dynamics; OH, hydroxyl; *P. marinus*, Petromyzon marinus; PXR, pregnane X recptor; RAR, Retinoic acid receptor; r, rat; *R. norvegicus, Rattus norvegicus;* RE, response element; RID, receptor interaction domain; RXR, retinoid X receptor; SAXS, short angle Xray scattering; TR, thyroid hormone receptor; VD, vitamin D; VDR, vitamin D receptor; VDRE, VD response element; z, zebrafish.


**Table 1 | List of crystal structure data available for LBD and DBD of VDR (source: www.pdb.org).** *(Continued)*


Molnár Structural data for VDR


*(Continued)*


**Ortholog**

**www.frontiersin.org** June 2014 | Volume 5 | Article 191 |

**20**


*(Continued)*


*dSynthetic peptide* 

*corresponding*

 *to NR2 Box of SRC1.* general domain organization, binding mode of natural ligands and recognition of DNA by VDR. Some data coming from molecular dynamics (MD) simulations will be also discussed since this approach represents a compromise in obtaining 3D structural models and have been proven to be well aligned with the wet lab data. At last, recent data from small-angle X-ray scattering (SAXS), cryo-electron microscopy (cryo-EM) and H/D exchange (HDX) experiments will be discussed with some perspectives highlighted.

## **WHAT DO VDR STRUCTURES TELL US?**

From functional and structural organization point of view VDR is formed by DNA-binding domain (DBD:24–89 aa; domain C), ligand-binding domain (LBD:126–427 aa; domain E), a connective hinge (domain D) between them and a short A/B domain located at the N-terminus. Compare to RXR it completely lacks the F domain, which is the very last part of the LBD after helix 12. The LBD is formed by a three layer anti-parallel α-helical sandwich (**Figures 1A,B** highlighted in green, blue and red). Based of the particular structure it contains all together 11–13 α-helices (Li et al., 2003) (**Figure 2A**). The internal structure of the LBDs of the respective nuclear receptors (NRs) shows a high similarity with specialized diversity based on functional properties of the particular receptor. In VDR, the LBD is responsible for active ligand recognition and interaction with partnering proteins such as coregulators, and RXR to form the functionally active RXR-VDR heterodimer. In particular, helices 3, 4, and 10–12 are involve in the interaction with protein partners. Interestingly, to date all solved VDR crystal structures show very ubiquitous and conserved organization of the of overall structural fold not reflecting the divergent nature of the bound natural or synthetic ligands. What are in fact the differences in structures that reflect various physiological effects of the particular ligands? Allegedly there will not be a simple answer to this question since we may face the limitation on the complexity of the available ligand-VDR structures or have to allow the possibility that the subtle differences between the structures may cause a "butterfly effect" that have more extreme consequences in terms of VDR signaling than initially thought. By all means there are important differences in the metabolism of various synthetic ligands and a possible unique coactivator recruitment may also play its role. However, none of these possibilities can be fully explored using the available VDR crystal structures. Nevertheless, what we may agree on is that all VDR crystal structures show agonistic conformation, surprisingly even in case of antagonists, that is canonically represented by a closed conformation of the helix 12 providing a docking platform for the recruitment of coactivators. This may be due to the shifted equilibrium that drives VDR for closed helix 12 with minimal energy conformation. In addition, the VDR structures do tell us the binding mode, anchoring points and subtle changes

in the position of residues that may be effectively used for *de novo* design of superagonist such as AMCR277A (PDBID:3CS4) (Hourai et al., 2008). The frequently mentioned subtle changes that are characteristic for the ligand-binding pocket (LBP) may be further analyzed and can explain some of the binding differences between various ligands in correlation to their functional and biological properties.

## **THE INNER CIRCLE: LOOKING INTO THE VDR POCKETS**

The "lower part" of the LBDs of all ligand-activated NRs contains a LBP which volume size range between 400 and 1400 Å<sup>3</sup> (**Figures 2A,B** light-brown mesh) (Nagy and Schwabe, 2004). It is not quite different in case of VDR where LBP primarily serves for effective recognition of various natural ligands such as 1,25D3 and its metabolites or bile acids. In addition, this is one of the most important parts to modulate VDR's activity via various synthetic compounds. The VDR pocket can be placed in the middle range of the volume scale showing rather high dynamic plasticity toward various ligands.

The first VDR crystal structure confirmed the conserved contact or anchoring points for the interaction of VDR with 1,25D3 (Rochel et al., 2000) (**Figure 2B**). The residues involved in the positioning of the 1,25D3 in the LBP are Y143 (helix 1) and S278 (helix 5) that contact the 1,25D31α-hydroxyl (OH) group, S237 (helix 3) and R274 (helix 5) contact 3β-OH, and H305 (loop 6–7) and H397 (helix 11) interact with 25-OH. In most of the cases if these anchoring point are disrupted a decrease in the activation potential of the ligand can be observed. Besides these residues the interior surface of the VDR LBP is formed of about 40 mostly non-polar amino acids. An interesting approach is to look how the LBP is changed upon ligand-binding especially its volume or how much volume (%) is occupied by the ligand. A rather straightforward example is the comparison of the 1,25D3 and MC1288 LBPs. Although the volumes of these ligands are highly comparable 434 Å<sup>3</sup> and 427 Å<sup>3</sup> (probe radius 1.9 Å), respectively, the volumes of the LBP show slight differences 776 Å<sup>3</sup> and 643 Å<sup>3</sup> (same probe radius) leading to ratio of 56 and 66% of ligand occupancy of the LBP volume (Molnár et al., 2006). When a ligand occupancy is higher compared to 1,25D3 then it increases the stability of VDR. This factor holds true for ligands that show high structural similarity and binding mode to 1,25D3. It seems to some extend that the ratio of the ligand to LBP volume can be a good descriptor of the ligands' activation potential. In addition, the actual shape of the cavity also reflects differences in the binding of various ligands which is illustrated in **Figure 3**. There are slight differences in the shapes of the LBP depending on the bound ligand. The red arrow shows the part where the shape is conserved well. Topologically this is the place where the 1,25D3 A-ring is located e.g., helices 1, 3, and 5. The more plastic part is the one where the 1,25D3 side chain is found with helices 6, 7, and 11. This region shows the highest variation between the ligandbound structures and is indicated with green arrow **Figure 3**. In addition, small changes can be also detected for instance the part highlighted with green circle. Very interesting in the difference between the two Gemini structures 2HCD and 4IA1, where the largest difference for the two side chains is that "C17–20 threo 20S—Gemini" has the hydrogens in its methyl groups substituted with deuterium. Interestingly, in this case the pocket shows more compact shape in the region with the double side chains and the volume of the location highlighted with green circle is decreased, which may be due to modification of the A-ring. Another interesting question is the maximum volume to which the VDR LBP can be stretched. MD simulations showed that by docking a Gemini with fluorinated methyls groups, (CF3)2-Gemini, the LBP could

be expanded by 1/3 of the 1,25D3 LBP volume. In proportion to this, the compound's volume in the pocket is also increased about 30%.

An appealing view opens up when a comparison of the VDR structure with one of its closest relative pregnane X receptor (PXR) is made (Watkins et al., 2001). As for VDR there are plenty of PXR crystal structures available and it can extend its LBP to very large volume (∼1400 Å3) to accommodate various compounds. The published VDR structures lack the insertion domain (*-*166–216 aa), but even without it the LBP as discussed earlier is as large as 700–800 Å3 with MD simulations showing that it can expand beyond 1000 Å3, which is already comparable to PXR's LBP. This may suggest that VDR is able to accommodate variety of other compounds in addition to 1,25D3 such as LCA/3-keto-LCA (Makishima et al., 2002). The later have been crystallized with rVDR and will be discussed in this review. The indications for binding additional compounds besides 1,25D3 are coming also from *P. marinus* (sea lamprey) where despite of lack of the calcified skeleton and teeth it may serve as a xenobiotic activator for detoxification by regulating P450 enzymes (Whitfield et al., 2003; Krasowski et al., 2005). However, it is yet to be determined whether there are existing other ligands that bind to VDR. One part of the LBD, which may allow the binding of these compounds is the insertion domain. Although its clear functional role has not been identified and it seems that it is not directly required for the binding of the 1,25D3 (Rochel et al., 2001), it may play some other roles. A mutation C190W was reported in patients that results in loss of 1,25D3 binding (Malloy et al., 1999), though this may be due to the disruption of the VDR structure by introducing a large bulky tryptophan residue. Secondly, the homologous part of PXR (142–431 aa) especially the occurrence of the two βstrands and the associated coiled regions are responsible for the expandability of the PXR's LBP thus an analogous role cannot be out ruled in case of VDR as well. There has been also proposed that an alternative pocket, which can be considered as an enlargement of the original pocket, is formed in the VDR that extends toward the helix 2/β-sheet region of the LBD (Mizwicki et al., 2004). Especially this can be observed with covalently locked 1,25D3-derived compounds such as 1,25(OH)2-lumisterol, which has been showed by *in silico* docking studies.

Due to the space limitation this review cannot address and discuss all the VDR-ligand complexes, for some more details see text below or recent reviews (Carlberg and Molnár, 2012; Carlberg et al., 2012), but some of the structures with natural ligands will be discussed in more details. One of them is 3-epi-1α,25-dihydroxyvitamin D3 (3-epi-1,25D3) a 1,25D3 metabolite that has been shown to exhibit tissue specific activities comparable to 1,25D3 (Norman et al., 1993; Reddy et al., 2000). The structural analysis showed a binding mode very similar to that of 1,25D3 with interesting compensation for the lacking S278-3α-OH hydrogen bond for the epimer using water mediated contacts **Figure 4** (Molnár et al., 2011). Interestingly, the same water channel is present in the 1,25D3 complex and was observed with other complexes as well (Tocchini-Valentini et al., 2001; Hourai et al., 2006).

A widely accepted fact is that precursor 25-hydroxyvitamin D3 (25D3) and its metabolite 24R,25-hydroxyvitamin D3 (24,25D3) does not posses significant biological activities. However, it has been shown that in *Cyp27b1*−*/*<sup>−</sup> cells, that are unable to produce actively 1,25D3, the VD signaling may be primary mediated via 25D3 (Lou et al., 2010). It stays a matter of discussion that under physiological condition how much of the VD signaling is mediated via 1,25D3 vs. 25D3. 24,25D3 shows only weak potency of influencing VD signaling at concentration 500nM although an enhancement for human osteoblast differentiation has been shown at concentration 1μM (van Driel et al., 2006). Docking and subsequent MD simulations have been done to see the

binding mode of these compounds and it has been confirmed that the position of residues mediating the anchoring hydrogen bonds are conserved with the exception of R274 which is located further than 3.5 Å from 24,25D3 **Figure 5**. Another residue that is unable to make a binding contribution is S237, but its position in the pocket is maintained. The void created by the missing R274 increases the LBP and lowers the occupancy factor for this compound. The analysis of the simple binding mode confers the activity range from 1,25D3 *>* 25D3 *>* 24,25D3. This also shows in general the importance of the 1α-OH group for the potency of the VDR agonist.

Recently, the crystal structures for another group of natural ligands have been solved. From the identification of secondary bile acids as VDR agonist (Makishima et al., 2002) the interesting question remained how these compounds bind to VDR. From the structural data it is evident that the litocholic acid (LCA) and 3-keto litocholic acid (3kLCA) are located in the opposite orientation than 1,25D3 (**Figure 6**). The 24-carboxyl group faces the β-turns of VDR, the β-region of the steroid backbone the helix 6–7/11 region and the A-ring is in the direction of helix 12. The S274 (hVDR S278) and Y143 hydrogen bonds are conserved

dashed lines.

in all structures (**Figure 6**). The difference for this part of the ligand-binding comes from the water mediated contacts with both R270 (hVDR R274) and S233 (hVDR S237). These contacts seem to be weaker compare to 1,25D3 but not so weak as in case of 25D3 and 24,25D3, where in fact they are missing due to lack of 1α-OH group. The opposite part of the bile acids shows also weaker hydrogen bonding network than for the previously discussed VD metabolites including 1,25D3. The structural comparison between the two bile acids shows a less complex, more straightforward binding mode for 3kLCA with only one water molecule involved. Whereas, for LCA there are no direct contacts with H301 (hVDR H305) and H393 (hVDR H397). All these contacts are mediated thorough waters. This also may explain the lower VDR activation potential of 3kLCA compared to LCA.

I would seems that the evolutionary design of VDR reached by recognizing 1,25D3 its perfection. All the anchoring points show effective hydrogen bonding and by looking at the limited data for natural ligands any deviation from 1,25D3 will result in the use of alternative bridging contacts such water molecules. This shows a rather limited adaptability of VDR in effectively recognizing its natural ligands, yet larger than we envisioned a decade ago. One must keep in mind that the activation of VDR with one or another metabolite will largely depend on the local cellular concentration of this compounds leading for instance for favored activation via bile acids instead of classical 1,25D3 binding.

## **DNA-RECOGNITION AND BINDING**

VDR belongs to the class of zinc finger transcriptional factors with DBD that consists of a highly conserved 66 aa residue core (Khorasanizadeh and Rastinejad, 2001) and an adjacent Cterminal extension. The conserved core has two zinc fingers where one contains fours cystein residues per atom of zinc (**Figure 7**). This feature allows VDR to effectively recognize and bind hormone response elements (REs) - termed VDREs. VDREs are

typically made up of two hexameric half-sites whose consensus sequence is 5 -RGKTCA-3 , where R = A or G and K = G or T. The half-sites may be arranged in various orientation most commonly forming a direct repeat with three neutral base pairs separating the half-sites (DR3) (Umesono et al., 1991). The unliganded VDR can occupy its REs also as a homodimer (Carlberg et al., 1993). Upon binding of ligand, VDR interacts with RXR and forms a heterodimeric complex that binds to VDREs with 5 -prime bound RXR.

The data for structural view on VDR-DNA recognition is scarce. To date we have only four structures published where three represent VDR homodimers on direct repeat 3 (DR3) from mouse osteopontin (mSPP) (CAC*GGTTCA*CGA*GGTTCA*), rat osteocalcin promoter (rOC)(CAC*GGGTGA*ATG*AGGACA*) and a canonical DR3 element (cDR3) (CAC*AGGTCA*CGA*AGGTCA*) (**Table 1**) (Shaffer and Gewirth, 2002). The last structure represents the DBDs of the heterodimeric RXR-VDR on canonical DR3 element (cDR3) (TT*AGGTCA*CGA*AGGTCA*A) (**Table 1**) (Shaffer and Gewirth, 2004). The 66 to 70 aa of the DBD are structurally coordinated by two zinc atoms that create a structure (**Figure 7**), in which one short α-helix is interacting directly with the major groove of the DNA (Härd et al., 1990; Lee et al., 1993). The VDR homodimers show asymmetric head to tail arrangements. The experimentally determined range of affinities of DR3s used for crystallographic studies are mSPP *>* cDR3 *>* rOC with mSPP supporting both VDR homodimer and RXR-VDR heterodimer binding, cDR3 requiring 10x higher VDR homodimer protein levels and weak heterodimer binding, whereas rOC is unable to bind VDR homodimers and has a very weak heterodimer binding ability (Freedman and Towers, 1991; Nishikawa et al., 1993; Toell et al., 2000). The change at third position, a purine to pyrimidine, of the consensus half sites AG*G*TCA allows the additional water mediated hydrogen contact of E42 with the DNA, which increase the stability of the mSPP-VDR complex. In case of rOC the reason for diminished VDR homodimer binding is because there is a G at position five of the upstream half-site GGGT*G*A, where supposedly RXR is bound in case of heterodimer binding. In high affinity half-sites in this location there is a C in the first strand but a complementary G in the second DNA strand. The interaction, which involves hydrogen bonds is between the R50 of VDR and the G from the second complementary strand. In rOC upstream half-site instead of G there is a C in the second strand, which is not a hydrogen bond acceptor thus R50 cannot interact with it. There is also some agreement between the strength of the homodimer binding and a sum of all existing DNA-VDR contacts within 3.5 Å calculated with the *ncont* program of the CCP4 suite (Winn et al., 2011). The results show 88, 85, and 83 contacts for rOC, cDR3, and mSPP, respectively. Interesting is also the interacting surface ratio between the two VDR homodimer molecules in percentage (5 upstream: 3 downstream) 38.52:61.48% for rOC, 46.21: 53.8% for the cDR3 and a reversed ratio of 57.3:42.7% for mSPP. This would suggest that for strong VDR homodimer binding there is a more contribution from the upstream half-site, but a clear conclusion cannot be reached based on this limited data set. In the RXR-VDR crystal structure there are two asymmetric units where the full complex shows an orientation RXR to VDR for unit 2 and a reversed orientation VDR to RXR for unit 1 with RXR bound on the downstream part of the VDRE. This may be due to the stabilization contacts between the adjacent VDR molecules where the hinge of one VDR molecule is stabilized with the DBD of the second VDR molecule. The general organization of the unit 2 is depicted also on the **Figure 7** bottom. The two zinc atoms (light blue color) with the respective cysteins are visualized (the second heterodimer unit is missing from the representation). The overall number of contacts is only 65, which is much lower than in case of VDR homodimers discussed earlier. This indicates that the binding of the RXR-DBD heterodimer to cDR3s is not optimal and for an effective binding a certain point of VDRE degeneration is needed. Moreover, the contribution of the monomers to the binding is rather interesting. The ratio between interacting monomeric surfaces is 56.22:44.26% (RXR-VDR) suggesting a higher contribution of RXR to the binding **Figure 7** top (see also the H/D exchange experiments discussed below). This is most likely due to the reverse orientation of the RXR-VDR on the cDR3. The surfaces of the interacting atoms are visualized in different color DNA (red), RXR (blue) and VDR (green). It is to be noted that compared to DNA-protein interaction there are hardly any interactions between the protein monomers. None for RXR-VDR, for VDR homodimers there are two for mSPP and rOC, and five on cDR3. This is quite in agreement with full length receptor studies which suggest that most of the heterodimerisation is contributed from hinges and LBDs.

## **THE OUTSIDE SHELL: PARTNERING AND THE COMPLEX VIEW**

### **VDR INTERACTING PROTEIN PARTNERS**

For VDR to function effectively as a regulator of transcription it is inevitable to interact with various protein partners. They show high structural and functional diversities ranging from enzymes, co-integrators and cofactors to components of distinct transduction signaling pathways. A comprehensive list of these partners with the accompanying citations are listed in **Table 2**.

One of the first complex identified using co-immunoprecipitation from mammalian cells was the VDR interacting protein DRIP complex, which is recruited to VDR in a completely ligand-dependent manner (Rachez et al., 1999). Many of the its components were shared with the earlier identified thyroid hormone receptor (TR) interacting protein complex TRAP (Fondell et al., 1996). It has been no surprise that majority of the interacting proteins can be related directly to transcriptional regulation such as subunits of the mediator complex MED1, 4, 6, 7, 12, 16, 17, and 23 (Rachez et al., 1999) or cofactors such as coactivators NCOA1-3 (Hong et al., 1997; Castillo et al., 1999; Molnár et al., 2005), NCOA6 (Mahajan and Samuels, 2000), ARA54 (Ting et al., 2005), SKIP (Baudino et al., 1998), RBP2 (Chan and Hong, 2001), SRB7 (Ito et al., 1999) and corepressors Alien (Polly et al., 2000), NCOR1 and 2 (Tagami et al., 1998; Kim et al., 2009), SIN3A (Fujiki et al., 2005), LCOR (Fernandes et al., 2003). Others show more selective properties functioning as coactivators and corepressors depending on the particular conditions such as RIP140 (Albers et al., 2005) or TIF1α (Thénot et al., 1997). Some are implicated in cellular processes such as cell cycle regulation CDK7 (Nevado et al., 2004), RAP46 (Guzey et al., 2000), DNA repair

#### **Table 2 | List of VDR interacting proteins.**


*(Continued)*


*\*no direct physical interaction but has positive effect on VDR transactivation.*

TDG (Chen et al., 2003) or signaling cascade FRG (Ellison et al., 2005). Interesting is the interaction and thus possible crosstalk with other NRs such as SHP (Albers et al., 2005), which lacks DBD and has corepressor-like behavior, and NGFIB, which has been shown to have a role during liver regeneration (Marshall et al., 2012). Some of the newly identified protein partners that may implicate VDR's involvements in new processes are XIRP1 that protects actin filaments from depolymerization or MYL3 a regulatory light chain of myosin (Marshall et al., 2012).

## **STRUCTURAL DATA OF SRC1 AND MED1/DRIP205 INTERACTION WITH VDR**

There is a big limitation in obtaining large transcriptional complexes, which is firstly due to the transient nature of the complex where VDR serves as a docking and acquiring platform bringing other proteins that either act as chromatin modifiers, parts of the mediator, of various cofactors and bridging factors to the close proximity of the functional VDREs. The complex may be assembled for a short moment to initiate and/or repress the transcription thereafter fulfilling this action it falls apart. The second reason might be that many of the interacting proteins such as cofactors show a high degree of disorder. The crystallization of unfolded proteins is very tricky and many times even impossible. The intrinsic disorder of a VDR interacting proteins is an expected structural property since e.g., members of the p160 general coactivator family have to adopt and interact with many various transcription factors. Thus structural data from only short interacting peptides derived from steroid receptor coactivator 1 (SRC1) with zVDR (**Figure 8A**) and mediator complex subunit 1 (MED1/DRIP205) with rVDR are available showing only the core interaction between VDR and the LXXLL motif of coactivators (**Figure 8B**). Both peptides interact in a very similar fashion. The α-helix of the peptide is oriented with its N-terminus toward helix 12. The two peptides interact through their LXXLL motifs, LHRLL in SRC1 and LMNLL in MED1, and most of the interaction is contributed from hydrophobic contacts of coactivator's leucine residues with the hydrophobic core from VDR helices 3, 4, and 12. The anchoring points of the short α-helix are based on the interaction with the "charge clamp" consisting of the conserved glutamate in helix 12 and lysine in helix 3, and the backbone amides of the coactivator peptide. The similar interaction of the two LXXLL motifs raises the question on how the

specificity is achieved in the interaction. The situation is complicated with the fact that some of the coactivators have more than one interaction motif such as SRC1 has five of them that are similar or related to LXXLL motifs, but so far only three of them were reported/studied in detail.

## **SAXS, CRYO-EM, AND H/D EXCHANGE STUDIES WITH FULL LENGTH RXR-VDR COMPLEXES**

The recent studies with VDR complexes in solution that employed the use of modern techniques such as SAXS, cryo-EM, HDX with full length VDR-RXRα showed somewhat more comprehensive perspective for the 3D organization and possible function of the VDR-RXRα-cofactor complex. The SAXS (Rochel et al., 2011) and cryo-EM (Orlov et al., 2012) derived model is shown in **Figure 9A**.

Upon binding to DNA from osteocalcin VDRE, the SAXS data derived RXR-VDR shows an elongated asymmetric open conformation with separate DBDs and LBDs and a well structured VDR hinge with VDR located downstream and RXR on upstream half sites (Rochel et al., 2011) (**Figure 9B**). On contrary the coiled structure of the RXR hinge allows its adaptability to different REs. The hinges play an important role in an open conformation. The hinges also underly one of the very important feature of the RXR-VDR complex, the dynamic character. The DBD binding to DR3 results in a rotation of the LBD dimers take around 90◦ with respect to the DNA (**Figure 9A**). The same study showed that coactivators MED1/DRIP205 and SRC1 have higher affinity to VDR compared to RXR. This points to the binding of only one molecule of coactivator through VDR, which is not supporting the "deck model" of binding (one LXXLL motif to VDR and the other to RXR) for these coactivators. Studies using mutants show the preferential binding of VDR to the second LXXLL motif of MED1/DRIP205 compared the weak recruitment of the first motif to RXR (Ren et al., 2000). However, both motifs are crucial for the effectivity of the NR activation complex *in vivo* (Malik et al., 2004). RXR may play a role in coactivator recruitment as well by associating to some other factors.

The cryo-EM studies of the RXR-VDR have a higher resolution than SAXS data and it is possible to obtain more precise electron densities for VDR A/B and hinge domains for both receptors (Orlov et al., 2012). The heterodimer takes an L-shape form on the DR3 with a proper orientation of RXR on the upstream and VDR on the downstream half site (similar to SAXS model). The complex supports also the asymmetric open architecture from SAXS data. Both LBDs are positioned above the 5 upstream half site as shown on **Figure 9A**. This result also emphasizes the importance of the hinges for the correct function of the complex. The main nature of the flexible hinge of RXR allows to contact the CTE helix with helix 1 and makes it possible to adopt differential spacing in DR REs. In addition, the coiled RXR hinge has to be long to reach DBD to its LBD which is located on the opposite side of the DR3 VDRE. The cryo-EM data points to one interesting feature, the potential in modulation of the DNA-binding using the 17 aa long A/B domain of VDR, which in fact interacts with the major groove of the DNA.

The H/D exchange (HDX) is a great tools to address the dynamic properties of the RXR-VDR-SRC1 complex (Zhang et al., 2011). The addition of RXR to VDR stabilizes region with helices 6–7, very similar to what is seen when some of agonists bind to VDR. As expected, upon 1,25D3 binding in the VDR LBD the helices 1, 3, 5–7, and 11 (the actual region that forms the LBP) have been stabilized, but binding efficiency of RXR to VDR is not enhanced. For RXR the helices 7 and 10 are stabilized (increase of the heterodimerisation) and an allosteric communication has been shown for the helix 3 of RXR. The 9-*cis* RA binding in general stabilized RXR, but in contrast to the crystal structure, it increased the fluctuation in helix 12. This observation may also indicate that the crystal structure take the minimized

energy conformation of the complex, but in reality it is more dynamic. The allosteric communication in VDR upon 9-*cis* RA binding was seen in the helices 3, 5, and 7 that are adjacent from the heterodimerisation interface pointing to modulation of the complex upon only 9-*cis* RA binding. Surprisingly, the concurrent binding of 1,25D3 and 9-*cis* RA has a destabilizing affect on the VDR DBD. A stronger increase of the VDR binding has been observed compared to RXR DBD in the presence of cDR3 element and in the absence of ligands, pointing to the higher contribution of the VDR for DNA interaction. In addition, interesting allosteric stabilization were observed for the VDR hinge and for helices 7–8 and 9–10. Moreover, unexpectedly the helix 12 of VDR showed increased fluctuation. However, it should be noted that in this experimental setup a coactivator has been missing from the protein complex (see below). The binding to the natural VDRE from *CYP24A1* gene showed similar result except it seemed that 5 -AGGTCA-3 half-site was occupied by VDR and helix 12 of RXR was more stable. Unexpectedly, the stability of the heterodimer on *CYP24A1* VDRE was reduced although the binding affinities of the two VDREs are in the same magnitude. The interaction of the coactivator SRC1, that contained three LXXLL motifs, with the heterodimer bound to both 1,25D3 and 9-*cis* RA increased the stability of VDR's helix 12 and helices 3 and 10–11 of RXR. Helices 3 and 4 of VDR cannot be further stabilized since they achieve maximal stabilization upon 1,25D3 binding. For RXR the loop between helices 10 and 11 is important in the formation of the hydrophobic groove facilitating coactivator binding. Besides the classical charge clamp RXR contains the so called "aromatic clamp" consisting of residues in helices 3 (F437, F277) and 12 (F450) that is important for coactivator binding. As expected in the absence of both ligand no coactivator interaction was observed and the separate addition of the 1,25D3 or 9-*cis* RA recruited the coactivator in a ligand-specific manner. Further HDX and cell based experiments showed that a simultaneous binding of the coactivator to both receptors is important and in the interaction with the RXR-VDR heterodimer only one SRC1 molecule is required (Zhang et al., 2011). This is in contrast to the SAXS derived model (Rochel et al., 2011). For the RXR-VDR-SRC1 complex the helix 12 of VDR has been stabilized upon addition of the DNA. In addition, HDX shows that the DNA-binding enhances the recruitment of SRC1 to RXR, thus the binding of the DNA stabilizes the recruitment of SRC1 to the whole heterodimer not just to the partner such as VDR.

Both, the SAXS and cryo-EM, studies highlighted the open architecture conformation in solution unlike it has been shown for the full length PPAR-RXR (Chandra et al., 2008). The recent crystal structure of the full length RXRα–LXRβ on DR4 RE provides also a support for the open conformation of the NR heterodimer complex (Lou et al., 2014). It will have to be seen whether the closed PPAR-RXR complex on DR1 is and exception, although more plausible is the open conformation giving a rather high dynamic freedom for the NR heterodimer in the large transcriptional complex.

### **CONCLUSIONS**

The main aim of this review was to collect and discuss structural data that is related to vitamin D signaling. The structural data for various isolated domains (LBD and DBD) show their organization on atomic level. This data is sufficient for understanding the particular ligand- and DNA-protein interactions but fails to provide spatial information on the mutual orientation of the domains of RXR-VDR on its natural promoters. It also fails to highlight inter-domain communication after DNA-, ligand- or cofactor-binding. However, they are irreplaceable tools for structure-based drug design and mechanistic view of the VDR action. The dissemination of the information derived from structural data and *in silico* models may help to understand how VDR works in its natural settings and provides a space for the intervention in various diseases. The recent SAXS, Cryo-EM and H/D exchange studies with full length RXR-VDR complexes show more complex views on VDR function and provide first tools for the integration of structural information with genomic, epi-genomic, transcriptional and functional data.

#### **FUTURE PERSPECTIVE**

There are numerous scientific questions connected to VDR that may be interesting to answer on structural level. The list is not complete but here are some of them:


#### **REFERENCES**


with conformationally restricted side chains: evaluation of biological activity and structural determination of VDR-bound conformations. *Arch. Biochem. Biophys.* 460, 161–165. doi: 10.1016/j.abb.2006.11.029


**Conflict of Interest Statement:** The author declares that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest. The Editor Carsten Carlberg declares that, despite being affiliated to the same institution as the author Ferdinand Molnár, the review process was handled objectively and no conflict of interest exists.

*Received: 13 March 2014; accepted: 30 April 2014; published online: 06 June 2014. Citation: Molnár F (2014) Structural considerations of vitamin D signaling. Front. Physiol. 5:191. doi: 10.3389/fphys.2014.00191*

*This article was submitted to Integrative Physiology, a section of the journal Frontiers in Physiology.*

*Copyright © 2014 Molnár. 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.*

## Vitamin D and the RNA transcriptome: more than mRNA regulation

## *Moray J. Campbell\**

*Department of Pharmacology and Therapeutics, Roswell Park Cancer Institute, Buffalo, NY, USA*

#### *Edited by:*

*Carsten Carlberg, University of Eastern Finland, Finland*

#### *Reviewed by:*

*Alberto Muñoz, Consejo Superior de Investigaciones Científicas, Spain Thomas William Dunlop, University of Eastern Finland, Finland*

#### *\*Correspondence:*

*Moray J. Campbell, Department of Pharmacology and Therapeutics, Roswell Park Cancer Institute, Buffalo, NY 14263, USA e-mail: Moray.Campbell@ RoswellPark.org*

The GRCh37.p13 primary assembly of the human genome contains 20805 protein coding mRNA, and 37147 non-protein coding genes and pseudogenes that as a result of RNA processing and editing generate 196501 gene transcripts. Given the size and diversity of the human transcriptome, it is timely to revisit what is known of VDR function in the regulation and targeting of transcription. Early transcriptomic studies using microarray approaches focused on the protein coding mRNA that were regulated by the VDR, usually following treatment with ligand. These studies quickly established the approximate size, and surprising diversity of the VDR transcriptome, revealing it to be highly heterogenous and cell type and time dependent. With the discovery of microRNA, investigators also considered VDR regulation of these non-protein coding RNA. Again, cell and time dependency has emerged. Attempts to integrate mRNA and miRNA regulation patterns are beginning to reveal patterns of co-regulation and interaction that allow for greater control of mRNA expression, and the capacity to govern more complex cellular events. As the awareness of the diversity of non-coding RNA increases, it is increasingly likely it will be revealed that VDR actions are mediated through these molecules also. Key knowledge gaps remain over the VDR transcriptome. The causes for the cell and type dependent transcriptional heterogenetiy remain enigmatic. ChIP-Seq approaches have confirmed that VDR binding choices differ very significantly by cell type, but as yet the underlying causes distilling VDR binding choices are unclear. Similarly, it is clear that many of the VDR binding sites are non-canonical in nature but again the mechanisms underlying these interactions are unclear. Finally, although alternative splicing is clearly a very significant process in cellular transcriptional control, the lack of RNA-Seq data centered on VDR function are currently limiting the global assessment of the VDR transcriptome. VDR focused research that complements publically available data (e.g., ENCODE Birney et al., 2007; Birney, 2012), TCGA (Strausberg et al., 2002), GTEx (Consortium, 2013) will enable these questions to be addressed through large-scale data integration efforts.

**Keywords: VDR, microRNA, transcriptome, epigenetic, microarray**

## **THE TRANSCRIPTIONAL LANDSCAPE OF THE HUMAN GENOME**

An appreciation of the diversity of transcription across the human genome has undergone a rapid expansion in recent years, in large part thanks to the efforts of integrative genomic approaches such as those of ENCODE consortium (Birney, 2012; Maher, 2012; Stamatoyannopoulos, 2012; Rosenbloom et al., 2013). From these studies it has become apparent that there is considerable variation and diversity in; the distribution of transcription factor binding across the human genome; the interplay between transcription factors and different co-regulating partners; the extent of the genome that is transcribed; the number and functionally different RNA-based molecules that are transcribed, the impact of mechanisms that process and edit RNA molecules that generate even greater diversity of gene expression.

In this context it is timely to review the functions of the vitamin D receptor (VDR/NR1I1) (Pike et al., 1980; Baker et al., 1988; Carlberg and Campbell, 2013), and consider how its actions contribute to this diversity of transcriptional and post-transcriptional events.

## **THE VDR ACTS IN MULTIMERIC PROTEIN COMPLEXES TO REGULATE TRANSCRIPTION**

The VDR, like many other members of the nuclear receptor superfamily are relatively well-understood transcription factors. Their actions have been dissected and modeled, and have generated the concept of cyclical gene regulation in which transcription factors oscillate between on and off states (Metivier et al., 2003; Reid et al., 2003; Kim et al., 2005; Vaisanen et al., 2005; Carroll et al., 2006; Meyer et al., 2006; Saramaki et al., 2006; Yang et al., 2006; Zella et al., 2006; Malinen et al., 2008; Seo et al., 2007).

A direct consequence of VDR genomic interactions and gene regulation is the control of the epigenetic states at receptor binding regions, and more broadly across target gene loci. Epigenetic events play a central role for transcriptional complexes and the various components in these multimeric complexes are able to initiate, sustain, and finally terminate transcription (Dobrzynski and Bruggeman, 2009). For example, different histone modifications can control the rate and magnitude of transcription (reviewed in Goldberg et al., 2007). These events are intertwined with levels of CpG methylation (Kangaspeska et al., 2008; Metivier et al., 2008; Le May et al., 2010). Thus the histone modifications regulated by VDR actions, and other epigenetic events including DNA methylation processes, combine during transcription to generate highly flexible chromatin states that are either transcriptionally receptive and resistant (Mohn and Schubeler, 2009). That is, the specific transcriptional potential of a gene is flexibly controlled by the combination of epigenetic events. These events are varied in space across the genomic loci, and in time through the course of the transcriptional cycle.

The diversity of histone modifications, and their association with different DNA functions formed the basis for the histone code hypothesis. This concept, first proposed in 1993, held that these modifications were governed in a coordinated manner and formed a code that mirrored the underlying DNA code to convey heritable information on transcription and expression (Turner, 1993). Given the rapid expansion of the understanding in the number of histone modifications, their genomic distribution and their combinatorial manner, it is only relatively recently that the true diversity of the range of histone states, and their functional outcomes, has become apparent (Goldberg et al., 2007). The strongest evidence that histone modifications at the level of metachromatin architecture form a stable and heritable "histone code," is perhaps seen with X chromosome inactivation (reviewed in Turner, 1998). The extent to which similar processes operate to govern the activity of micro-chromatin contexts at gene promoter regions, is an area of debate (Jenuwein and Allis, 2001; Turner, 2002).

The regulation of transcription and the patterns of mRNA expression have been related to the expression of these histone modifications through a wide range of correlative and functional studies. For example, histone H3 lysine 4 tri-methylation (H3K4me3) is found in the promoter regions of actively transcribed genes. This mark is mutually exclusive with H3K9me, which instead is associated with transcriptionally silent promoter regions. Acetylation of H3K9 is found along with methylation of H3K4 at active promoter regions. Similarly, H3K27 can be either acetylated or methylated, with acetylation associated with active gene transcription and methylation associated with gene silencing.

In many experimental cases it has been established that VDR activation by natural ligand 1α,25-dihydroxyvitamin D3 (1α,25(OH)2D3) or synthetic analogs, can lead to a highly dynamic exchange of co-factors by releasing co-repressors and inducing a receptor conformation that attracts binding of coactivator proteins (**Figure 1**). This exchange of associations induces a more relaxed, or open, chromatin status and the recruitment of linking factors and subsequently the basal transcriptional machinery. However, this is not an indefinite signal and the ligand, is rapidly metabolized. Also the VDR itself is limited in function by proteasome-mediated receptor degradation (Peleg and Nguyen, 2010). In the absence of ligand, some basal level of receptor remains in the nucleus associated with co-repressor complex and leads to silencing of transcription (Malinen et al., 2008; Saramaki et al., 2009; Thorne et al., 2011; Doig et al., 2013). The sequencing, or choreography of these actions, give rise to the periodicity of transcriptional activation and pulsatile mRNA and protein accumulation and reflect intrinsic control mechanisms required to tightly regulate the expression of important signaling molecules. At closer resolution, for example on shorter time-scales, the patterns of regulation show some degree of coordinated pulsatile regulation, and probably reflect aspects of specific VDR binding sites impacting gene regulation for example through emerged to support chromatin looping within the same VDR target gene loci (Vaisanen et al., 2005; Saramaki et al., 2006, 2009).

Whilst these scenarios are relatively well characterized for the positive regulation of gene expression, it is probably not a complete understanding as the distribution of VDR binding in the genome (Ramagopalan et al., 2010; Heikkinen et al., 2011; Meyer et al., 2012; Ding et al., 2013) and the patterns of associated gene regulation suggest that the VDR is actually associated broadly with gene activation and repression. The mechanisms that drive gene repression appear more diverse than gene activation and reflect differences in complex formation, and the choreography of binding.

The patterns of protein-protein interaction identified for the VDR allude to both positive and negative gene regulation (**Table 1**). The VDR commonly forms a heterodimer with RXRα (Quack and Carlberg, 2000). However, the identification of partners that interact in the same complex supports a broad role for the VDR complex to regulate other signal transduction events and RNA processing activities.

A number of proteins with transcriptional activator function have been identified in complex with VDR. For example, CBP/p300 is a transcriptional co-integrator with histone acetlyase activity (Wang et al., 2011a,b, 2013a) and is associated with the VDR. Other proteins such as SNW1/NCOA62 which has function as a transcriptional co-activator(Baudino et al., 1998), as well as other proteins that have more recently been characterized to have coativator function (CCND3) (Cenciarelli et al., 1999; Lazaro et al., 2002; Despouy et al., 2003; Sarruf et al., 2005). Similarly the corepressor HR is also identified by such protein-protein interactions (Hsieh et al., 2003).

Aside from these traditional roles to modulate transcriptional actions, there is evidence to support a wider range of actions for the VDR in the control of mRNA. The coactivator SNW1 also plays a role as a splicing factor. This latter function is also shared by another VDR interacting protein, namely SRPK1, which is a protein kinase that regulates the activity of various splicing factors (Hayes et al., 2006; Aubol et al., 2013). Other interactions allude to the cross-talk between the VDR and different signal transduction processes. For example, the VDR interacts with negative regulators of WNT signaling (NKD2) (Katoh and Katoh, 2007), substrates for PKC signaling (PRKCSH) (Gkika et al., 2004), p53 (Kommagani et al., 2006; Lambert et al., 2006; Maruyama et al., 2006; Saramaki et al., 2006; Ellison et al., 2008) and SMAD3 (Ding et al., 2013; Ito et al., 2013; Zerr et al., 2014). The VDR functionally interacts with a range of co-repressors such as NCOR1 (Saramaki et al., 2009; Doig et al., 2013), NCOR2/SMRT (Khanim

et al., 2004; Gynther et al., 2011), TRIP15/ALIEN (Polly et al., 2000; Cui et al., 2009) and DREAM (Scsucova et al., 2005) but, interestingly, agnostic protein-protein interaction tools such as INTACT (**Table 1**) have not identified direct VDR co-repressor interactions. It is unclear why these proteins are not identified in such protein-protein screens, but may reflect an experimental artifact as a result of investigators using ligand stimulated VDR to capture interacting proteins.

activated VDR also represses a number of miRNA for example including

## **THE DIVERSITY OF VDR-PROTEIN INTERACTIONS IS REFLECTED BY THE DISTRIBUTIONS OF GENOMIC BINDING SITES**

To identify VDR binding sites through the genome several groups have now applied ChIP-Seq approaches in different human cell types including immortalized lymphoblastoids (Ramagopalan et al., 2010), hepatic stellate cells (Ding et al., 2013) and cancer cell lines representing monocytic leukemia (Heikkinen et al., 2011) and colon cancer (Meyer et al., 2012). These studies also differed in the time of cell exposure to 1α,25 (OH)2D3 ranging from 40 min to 36 h identify binding sites, on the order of hundreds to several thousands of different binding sites depending on time of treatment with 1α25(OH)2D3 and cell background, with longer treatment time points tending to be associated with more binding sites. However, as yet there are not uniform standards for the analyses of NGS data, and therefore it is likely that different analytical approaches are influencing the number and significance of the VDR enriched peaks identified.

disease and how can this be monitored and exploited?

These differences in treatment and analytical approaches aside, these VDR binding data sets reveals that fewer than 20% of the VDR binding sites are in common between the different cell types. These finding perhaps offers strong support for the concept that VDR transcription is extremely tailored in different cell types, presumably through interactions with either equal or more dominant co-factors that combine to determine its binding. Another important finding from these studies is that the canonical binding site for the VDR, termed the direct repeat (DR) spaced by 3 nucleotides (DR-3), which was identified by traditional biochemical approaches in candidate gene studies, appears to be the minority genomic element that directly binds the receptor. Fewer that 30% of genomic VDR binding sites contain a DR-3, although this number is increased following ligand treatment

#### **Table 1 | Proteins known to interact with the VDR.**


*(Continued)*


*The INTACT database curated by EBI http://www.ebi.ac.uk/intact/ was interrogated for interactions with the VDR. Interactions of VDR with NCOR1, NCOR2, and COPS2 were curated from the literature.*

when there is increased enrichment for VDR binding to DR-3 elements (reviewed in Carlberg and Campbell, 2013). Nonetheless, a range of other genomic elements were enriched in VDR binding peaks suggesting that the VDR co-operates closely with other factors to associate with the genome, both in the absence and presence of ligand. Indeed, the study of Evans and co-workers in the hepatic stellate cells (Ding et al., 2013) and the work of Pike and co-workers in colon cancer cells (Meyer et al., 2012) both address this significant cross-talk of the VDR. In the case of the hepatic cells this is considered in the context of TGF® and in the case of colon cancer cells this is with TCF4, downstream of ®-catenin. Both of these studies therefore reflect the finding of VDR interactions with SMAD3 specifically, and more generally with regulators of WNT signaling (**Table 1**).

## **THE HUNT FOR PIONEER FACTORS TO EXPLAIN THE DIVERSITY OF VDR FUNCTION**

Together these ChIP-Seq studies suggest that the VDR combines with other proteins in a network of interactions, quite likely in a cell type specific manner, to participate in diverse gene regulatory networks. It remains to be established how targeted or stochastic this is. The variation observed in both the type and position of binding sites for the VDR, depending on cell phenotype and disease state, suggests it is directed, and at least will establish a paradigm for hypothesis testing concerning what directs the VDR to bind and participate in gene transcription. The specificity of VDR signaling may arise due to integration with other perhaps more dominant transcription factors. Again, for other nuclear receptors (e.g., AR and ER) the concept has emerged that receptor binding is guided by the actions of more dominant so-called pioneer factors including the Forkhead (FKH) family members (Lupien et al., 2008; Serandour et al., 2011; Sahu et al., 2013). Efforts to define the major pioneer factors for the VDR have proved to be less consistent between the different VDR ChIP-Seq studies and may reflect the biology of the VDR which, given that it exists in the nucleus both in the presence and absence of ligand, potentially is a more interactive protein such that a single dominant pioneer factor is not so deterministic.

Another approach to identify the interacting partners of the VDR has been to examine the gene networks it regulates and to cluster genes by known regulating transcription factors. Novershtern et al. (2011) measured the transcriptome profiles of a large number of hematopoietic stem cells, multiple progenitor states and terminally differentiated cell types. They found distinct regulatory circuits in both stem cells and differentiated cells, which implicated dozens of new regulators in hematopoiesis. They identified 80 distinct modules of tightly co-expressed genes in the hematopoietic system. One of these modules is expressed in granulocytes and monocytes and includes genes encoding enzymes and cytokine receptors that are essential for inflammatory responses. Major players in this module are VDR together with the factors CEBPα and SPI1/PU.1. This indicates that VDR works together with this small set of transcription factors, in order to regulate granulocyte and monocyte differentiation. It is reasonable to anticipate that such modules exist in multiple cell types but are guided by the tissue specific expression of such factors.

#### **VDR REGULATION OF THE PROTEIN-CODING TRANSCRIPTOME**

Anti-proliferative effects of 1α,25(OH)2D3 have been demonstrated in a wide variety of cancer cell lines, including those from prostate, breast, and colon (Colston et al., 1982, 1989; Peehl et al., 1994; Campbell et al., 1997; Koike et al., 1997; Elstner et al., 1999; Welsh et al., 2002; Pálmer et al., 2003). Following on from these, VDR transcriptional studies were initially undertaken at the candidate level to identify processes by which the VDR mediated its cellular effects. These approaches identified of the gene encoding the 1α,25(OH)2D3 metabolizing enzyme *CYP24A1* (Dwivedi et al., 2000; Anderson et al., 2003) and *CDKN1A* (encodes p21*(waf1/cip1)*) (Schwaller et al., 1995) as VDR targets. Subsequently, with the emergence of differential expression and membrane array technology, workers applied these wider screening approaches to identify multiple genes regulated by the VDR. For example, Freedman and colleagues applied differential expression approaches in the context of 1α,25(OH)2D3 induced myeloid differentiation and identified a number of cyclin dependent kinase inhibitors including *CDKN1A* and undertook functional confirmation studies to suggest the importance of the regulation of these targets to trigger myeloid differentiation (Liu et al., 1996). Others undertook so-called focused array technology whereby cDNA probes for selected genes involved in key biological processes or disease states were arranged on macro scale membrane arrays. Such arrays contained anywhere from several hundred to several thousand probes, and so were not genomewide in terms of coverage but rather were candidate arrays often focused around specific pathways or disease states such as cancer. Despite the limitations, these approaches yielded important information supporting the links between VDR action and the regulation of growth and signaling (Savli et al., 2002). Similarly, first generation arrays chips, for example from Affymetrix which contained 4500 probes (Akutsu et al., 2001), also enabled sufficient genomic coverage to begin to define specific regulated gene networks. This particular study from White (Akutsu et al., 2001) and co-workers identified 38 genes that were responsive to 1α,25(OH)2D3 exposure, which represented approximately 1% of the transcriptome studied, and included *GADD45A*. These earlier studies already suggested at the footprint of the VDR dependent transcriptome (reviewed in Rid et al., 2013). In many ways these studies highlighted the heterogeneity of VDR actions that was to be identified subsequently by ChIP-Seq studies. This heterogeneity may in part reflect experimental conditions with very different cell line differences, and genuine tissue-specific differences of cofactor expression that alter the amplitude and periodicity of VDR transcriptional actions.

Even within this diversity there is some consistency on a certain targets and the biological actions they relate to, including cell cycle regulation (Akutsu et al., 2001; Pálmer et al., 2003; Eelen et al., 2004; Wang et al., 2005). A common anti-proliferative VDR function is associated with arrest at G0/G1 of the cell cycle, coupled with up-regulation of a number of cell cycle inhibitors. Candidate promoter characterization studies have demonstrated a series of VDR binding sites in the promoter/enhancer region of *CDKN1A* (Liu et al., 1996; Saramaki et al., 2006). By contrast the regulation of the related CDKI p27*(kip1)* is mechanistically enigmatic, and included translational regulation and enhanced mRNA translation, and attenuating degrading mechanisms (Hengst and Reed, 1996; Wang et al., 1996; Huang et al., 2004; Li et al., 2004). The up-regulation of p21*(waf1/cip1)* and p27*(kip1)* principally mediate G1 cell cycle arrest, but 1α,25(OH)2D3 has been shown to mediate a G2/M cell cycle arrest in a number of cancer cell lines via direct induction of *GADD45A* (Akutsu et al., 2001; Jiang et al., 2003; Khanim et al., 2004).

In the transition to genome wide understanding, workers applied more comprehensive array approaches to define VDR mRNA transcriptomes. For example, investigations of squamous cells (Lin et al., 2002; Wang et al., 2005) identified networks of genes that trigger the response to wounding, protease inhibition, secondary metabolite biosynthesis, cellular migration, and amine biosynthetic processes. Another approach has been to examine vitamin D sensitive and responsive isogenic cell pairs and undertake analyses to identify key networks that are critical for mediating antiproliferative sensitivity toward 1α25(OH)2D3. In this manner a critical role for TGF® signaling was again revealed, to associate with VDR antiproliferative sensitivity toward 1α25(OH)2D3 in breast cancer cell models (Towsend et al., 2006). Exploiting leukemia cell models with differential responsiveness toward 1α25(OH)2D3 triggereddifferentiation (Tagliafico et al., 2006) identified that certain VDR transcriptional targets could distinguish the aggressiveness of the leukemia, again, focused around cell cycle and included *MS4A3* which can modulate the phosphorylation of CDK2 and therefore exert control over the cell cycle. This concept of VDR sensitive vs. resistant models was also exploited in prostate cancer to identify the critical VDR transcriptional targets that mediate antiproliferative sensitivity and again also identified cell cycle and signal transduction components including *GADD45A* and *MAPKAPK2* that were required to mediate the sensitivity of cells to 1α,25(OH)2D3. Furthermore, these studies examined the epigenetic basis for the transcriptionally inert state of these targets in resistant models (Rashid et al., 2001; Khanim et al., 2004).

Identification of the VDR-dependent transcriptome using microarray approaches is heavily dependent on a range of statistical and technical considerations (Do and Choi, 2006; Zhang et al., 2009) including hybridization variations and limitations, background effects, normalization procedures, the choice of statistical test to identify differentially expressed genes, which in turn relies on study design and the numbers of arrays and samples chosen for study. Many of these study components were only formally agreed upon with the establishment of the MIAME compliant protocols in 2001 (Brazma et al., 2001), and as these became accepted standards for journal publication these approaches became widespread through the biological community.

MIAME compliant array experiments are subsequently published in public archives, such as GEO (Barrett et al., 2005, 2013) and ArrayExpress at EMBL (Parkinson et al., 2007) and Stanford microarray database (Marinelli et al., 2008). These three repositories between them contain thousands of genome-wide microarray experiments, containing millions of individual microarrays. Mining these repositories reveals a range of experiments (not all published) where cells have been treated with 1α,25(OH)2D3 and other vitamin D compounds, and RNA effects studied between short time points (1–2 h) to several days (**Table 2**). Again, these studies have supported consistent themes in terms of the VDRmediated control of cell cycle and signal transduction processes, the suppression of WNT and NF-κB, and the regulation of IGF1 signaling (Kovalenko et al., 2011), and integrated actions with TGF® signaling. A final area to emerge from these agnostic studies of the VDR transcriptome is the impact that 1α25(OH)2D3 exposure exerts on a range of chromatin remodeling components. Interestingly, NCOR2/SMRT appears to be a target of VDR signaling (Dunlop et al., 2004), and adds to the concept that VDR signaling is cyclical and based on the functions of various negative feedback loops. Similarly, KDM6B/JMJD3 is a histone H3 lysine demethylase and expression is induced by the activated VDR. In this manner, VDR action can feed-forward its own transcriptional program by promoting H3K9 acetylation and gene action (Pereira et al., 2011).

Given the number of arrays available, it is now timely to consider meta-analyses across the arrays to reveal common themes; this forward compatibility is one the key benefits of MIAME compliance. Meta-integration of array data has been shown to be surprisingly revealing in a range of studies. For example at the larger scale various workers have integrated multiple microarray data to reveal underlying patterns in the context of disease classification (Shah et al., 2009; Engreitz et al., 2011) but can also be applied to consider that specific phenotypes (Martinez-Climent et al., 2010; Rantala et al., 2010; Lai et al., 2014). It is

#### **Table 2 | Publically available MIAME compliant microarray studies of VDR function.**


therefore timely for these data to mined, and integrated with related nuclear receptor actions or other transcription factors that appear to co-operate with the VDR, for example SMADs.

#### **VDR REGULATION OF NON-CODING RNA SPECIES**

The human genome project in many ways was a race to define the protein coding genes within the human genome. Bacterial artificial chromosomes (BAC) clones enabled relatively large pieces of DNA, upto 300 kb to be inserted for sequencing (Osoegawa et al., 2000). However, a key step in the initial alignment process was to leverage cDNA and EST libraries and therefore naturally steered workers to protein coding genes, and the significance and extent of non-protein coding RNA remained largely unexplored.

Although non-coding RNA forms were well described in terms of ribosomal function it was little understood beyond this. The interpretation of the human genome, and other large scale approaches to investigating chromosomal function (Consortium et al., 2007; Tress et al., 2007) all led to a growing awareness of the extent of non-coding RNA and at least suggested that their was an unknown. This uncertainty has been reflected in the debate within the biological community over the extent and roles of so-called Junk DNA (Kapranov and St Laurent, 2012; Doolittle, 2013). As a result researchers have considered roles for non-coding RNA in the regulation of cell function and have begun to examine the interplay between the at least 20 different types of different non-coding RNA (reviewed in Ling et al., 2013). Many of these RNA species are gene regulatory RNA and include microRNA (miRNA), long non-coding RNA (long ncRNA), whereas others are involved in the posttranscriptional modification of RNA for example small nucleolar RNA (snoRNA).

Workers have now principally examined miRNA regulation by the VDR and evidence has emerged to support a role for the VDR to control regulation. For example Studzinski and coworkers revisited the mechanistically enigmatic VDR-mediated control p27*(kip1)*. They elegantly demonstrated a role for VDR to down-regulate miR181a, which when left unchecked degrades p27*(kip1)*(Wang et al., 2009) (**Figure 1**). Thus, indirectly, VDR activation elevates expression of p27*(kip1)*, initiates cell cycle arrest and commits cells toward differentiation. These studies illuminated the earlier ones that suggested that p27*(kip1)* protein levels appeared to be regulated by a range of post-transcriptional mechanisms, such as enhanced mRNA translation, and attenuating degradative mechanisms (Hengst and Reed, 1996; Huang et al., 2004; Li et al., 2004). Similar integration of miRNA and mRNA was revealed to control the regulation of *CDKN1A*. Dynamic patterns of *CDKN1A* mRNA accumulation have been observed in various cell systems (Thorne et al., 2011). This is in part explained by the epigenetic state of different VDR binding sites on the *CDKN1A* promoter. However, VDR-dependent co-regulation of miR-106b also appears to modulate the precise timing of *CDKN1A* accumulation and expression of p21*(waf1/cip1)* in a feedforward loop and determine the final extent of the cell cycle arrest. 1α,25(OH)2D3 regulates the DNA helicase *MCM7* (Khanim et al., 2004) that encodes the miR-106b, in intron 13 of the *MCM7* gene, and together these co-regulation processes control p21*(waf1/cip1)* through the balance of MCM7 and *CDKN1A* (Saramaki et al., 2006; Ivanovska et al., 2008) (**Figure 1**).

MicroRNA (miRNA) contribute negative regulatory aspects to normal gene regulation, for example as part of feed-forward loop motifs (Mangan and Alon, 2003; Mangan et al., 2006). The co-regulation of mRNA and miRNA in motifs that included feed forward structures appears quite common (Song and Wang, 2008; Gatfield et al., 2009; Ribas et al., 2009; Sun et al., 2009; Wang et al., 2009). Other established miRNA targets of the VDR include miR-627 (Padi et al., 2013) that in turn targets *JMJD1A* (another histone H3 lysine demethylase) miR-98 (Ting et al., 2013) and let-7a-2 (Guan et al., 2013). However, one of the more explored relationships between VDR and miRNA is the relationship between VDR and miR-125b. MiR-125b inhibits VDR (Mohri et al., 2009; Zhang et al., 2011) and in turn VDR can down-regulate miR-125b (Iosue et al., 2013; Zhou et al., 2014), and the other targets of miR-125b include NCOR2/SMRT (Yang et al., 2012) (itself a VDR target gene) suggesting multiple levels of co-regulation and interdependent relationship between the VDR, and the mRNA and miRNA transcriptomes, and the epigenome. Finally, it is interesting to note that altered levels of miRNA are associated with cancer states and progression risks and indeed miR-125b is associated with aggressive prostate cancer (Shi et al., 2011; Amir et al., 2013; Singh et al., 2014).

Beyond these candidate studies, a number of investigators have undertaken miRNA microarray analyses (**Table 2**) and these approaches have identified various networks, including the control of lipid metabolism and PPAR© function (Wang et al., 2013b). It is likely that with the increased application of array approaches and next gen sequencing approaches will identify the key networks downstream of the VDR miRnome. This is unfortunately a more challenging research question owing to the many-to-many nature of miRNA; a given miRNA target many mRNA, and a given mRNA may have many miRNA targeting it. Thus, the computational challenges to resolve these relationships are not insignificant. Together these findings suggest that co-regulated miRNA may form an integral part of VDR signaling to control gene expression.

Of the other types of non-coding RNA, their regulation by VDR remains far less explored. Recently the group of Bickle have begun to dissect VDR regulation of lncRNA in keratinocytes and identified a number of target lncRNA and in doing so have raised the curtain on new avenues of exploration (Jiang and Bikle, 2014).

## **CONCLUSIONS AND FUTURE CHALLENGES FOR UNDERSTANDING THE VDR TRANSCRIPTOME WHAT ARE THE PROTEINS, OR PROCESSES THAT GUIDE WHERE THE VDR BINDS IN THE GENOME?**

It is unclear what key pioneer factors will be identified and if the VDR is in a strong relationship with a specific family of pioneer factors, in the same that the AR is profoundly influenced by the Forkhead family members. Indeed, the precise pioneer factor may even be tissue specific as also revealed for the AR (Pihlajamaa et al., 2014). This may reflect the fact the ligand activation of the VDR is more associated with re-distribution of the VDR through the genome, rather than triggering movement into the nucleus (as in the case of the classic steroid hormone receptors).

The specific epigenetic niche that characterizes the VDR binding may also be revealing of where and why the VDR binds to the genome. These analyses will require agnostic integration of multiple genomic data sets, for example histone modifications, transcription factor binding, chromatin conformation and transcriptomic data and application of machine learning approaches to reveal the significance of the underlying patterns, VDR binding and transcriptional activity. Whilst the VDR was not included in the ENCODE project, judicious choice of a cell line model for these studies, most likely a Tier 1 cell line from ENCODE, will enable leverage of a considerable volume of cistromic and epigenomic data to be combined with *de novo* VDR ChIP-Seq data. In this manner the question of how TGF® and/or WNT signaling interacts directly or indirectly with VDR binding can be addressed relatively easily.

Another major knowledge gap in VDR understanding concerns the spatial relationships between VDR binding and the control of transcription. It is clear that chromatin looping processes can transiently bring distal regulatory regions into physical proximity to the proximal regions of a gene and lead to dynamic gene expression (Saramaki et al., 2009). To transition from examination of looping of this process at a single locus using established binding sites to the genome wide investigation is technically and statistically very challenging. Again, ENCODE analyses may be useful here, as Chromatin Interaction Analysis by Paired-End Tag Sequencing (ChIA-PET) data are available for specific cell lines; for example using RNA-PolII in K562 cells and therefore VDR ChIP-Seq in these cells would again be able to leverage this data to begin to understand how the VDR distributes and loops across the genome.

## **WHAT IS THE ROLE OF GENETIC VARIATION IN DETERMINING HOW AND WHERE THE VDR BINDS?**

Genetic variation exists throughout the genome and by definition is predominately in non-RNA coding regions. This realization has been the catalyst for examining how genetic variation impacts transcription factor binding and activity. Perhaps the most comprehensive integration of these concepts has been the development of the RegulomeDB tool (Boyle et al., 2012) which considers the impact of genetic variation on the function of all transcription factors analyzed by ENCODE. To date this question has not been seriously considered in terms of the VDR. A major hurdle to addressing this question is very large potential for Type 1 error owing to the large-scale data sets that need to be integrated, namely; all SNPs and those in linkeage disequilibrium that are significantly associated with disease in replicated studies and all binding sites for a given transcription factor against the backdrop of the number of SNPs for a given trait, the platform used for identification and the total number of SNPs in the human genome. In the context of the VDR specifically, this challenge is compounded by the fact that the majority of VDR binding sites through the genome do not contain a canonical DR3 type binding element and therefore a critical question will remain around what protein is the VDR interacting with in the genomic context and how is this influenced by genetic variation. This challenge is clearly intertwined with developing a comprehensive knowledge of how the VDR binds to the genome.

## **WHAT IS THE COMPLETE VDR TRANSCRIPTOME AND HOW DOES IT DIFFER BY TISSUE AND BY DISEASE?**

Surprisingly, no RNA-Seq data are yet available for the VDR. Therefore to capture all RNA regulated by the VDR will require RNA-Seq approaches applied to libraries that capture short and long RNA species. The ENCODE consortium have undertaken over 400 RNA-Seq experiments focused on different RNA species in multiple cell models, and this makes a compelling case for exploiting these data, especially in Tier 1 or Tier 2 cell lines. For example undertaking VDR ChIP-Seq and a limited number of RNA-Seq approaches in K562 cells has the potential to leverage a remarkable volume of data. Predictions from such integrative analyses could then be tested in other ENCODE resources, in RNA-Seq from multiple tissues in normal healthy donors, through the GTEx consortium, or through the vast numbers of microarrays that are publically available. For example, in the case of K562, which is a CML cell line, there are many large-scale microarray analyses of patients with CML to examine how the VDR transcriptome relates to disease state and drug response. A parallel outcome of investigating VDR function will be to address the role it plays in regulating splice variation as suggested by the interactions with proteins such as SNW1.

## **HOW WILL THIS KNOWLEDGE BE EXPLOITED IN PERSONALIZED MEASURES OF VDR SYSTEM IN HEALTH AND DISEASE?**

Many aspects of the relationships identified above can be interpreted by serum borne measurements, which are highly attractive owing to their ease of measurement. Serum levels of the prohormone, 25(OH) vitamin D3, are strongly correlated with the generation of the active hormone 1α,25(OH)2D3 and VDR function. For example, reduced serum levels of 25(OH) vitamin D3 levels are associated with increased risk of either cancer initiation and/or progression (Drake et al., 2010; Shanafelt et al., 2011). Therefore the serum level of 25(OH) vitamin D3can yield the "potential" of the VDR system to signal (Brader et al., 2014). This potential is impacted by the various cellular mechanisms outlined above. Of these, genetic variation that impacts VDR binding can obviously be measured in any cell in the body. The total transcriptome can be challenging to measure but perhaps small non-coding RNA represent a highly attractive marker of activity. Remarkably, miRNA are readily secreted into serum where they remain stable (El-Hefnawy et al., 2004; Goyal et al., 2006; Taylor and Gercel-Taylor, 2008; Valadi et al., 2007) and can be reliably extracted and measured (Chen et al., 2008; Liu et al., 2008; Mitchell et al., 2008; Rabinowits et al., 2009). Using serumborne molecules as prognostic markers is highly attractive for several reasons. First, they can overcome the limitations of inaccurate sampling for example in the case of the presence of cancer within a tumor biopsy. Second, they can encapsulate the effects of heterotypic cell interactions, again, for example within the tumor microenvironment. Third, they form a non-invasive test procedure. From a biostatistical perspective, given there are fewer miRNA than protein coding mRNA, genome-wide coverage is more readily achieved and avoids the statistical penalties typically associated mRNA genome wide testing (Lussier et al., 2012).

This raises the very exciting possibility that generating integrated models of VDR binding, the impact of genetic variation, the tissue specific differences in the transcriptome and identifying miRNA contained within transcriptional circuits offers the opportunity of exploiting their serum expression levels of 25(OH) vitamin D3, genetic variation and miRNA expression will be able to be exploited to predict accurately the capacity of VDR function.

## **ACKNOWLEDGMENTS**

Moray J. Campbell acknowledges support in part from National Institute of Health grants [R01 CA095367-06 and 2R01-CA-095045-06] and the NCI Cancer Center Support Grant to the Roswell Park Cancer Institute [CA016056].

## **REFERENCES**


(MIAME)-toward standards for microarray data. *Nat. Genet.* 29, 365–371. doi: 10.1038/ng1201-365


MC3T3-E1 cells. *J. Steroid Biochem. Mol. Biol.* 89–90, 405–407. doi: 10.1016/j.jsbmb.2004.03.008


role for vitamin D in osteoblast mTOR signaling. *FASEB J.* 25, 937–947. doi: 10.1096/fj.10-172577


circuits control cell states in human hematopoiesis. *Cell* 144, 296–309. doi: 10.1016/j.cell.2011.01.004


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

*Received: 16 March 2014; accepted: 21 April 2014; published online: 14 May 2014. Citation: Campbell MJ (2014) Vitamin D and the RNA transcriptome: more than mRNA regulation. Front. Physiol. 5:181. doi: 10.3389/fphys.2014.00181*

*This article was submitted to Integrative Physiology, a section of the journal Frontiers in Physiology.*

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

**REVIEW ARTICLE** published: 18 February 2014 doi: 10.3389/fphys.2014.00060

## Interaction of vitamin D with membrane-based signaling pathways

#### *María Jesús Larriba1, José Manuel González-Sancho1, Félix Bonilla2 and Alberto Muñoz <sup>1</sup> \**

*<sup>1</sup> Instituto de Investigaciones Biomédicas "Alberto Sols," Consejo Superior de Investigaciones Científicas, Universidad Autónoma de Madrid, Madrid, Spain <sup>2</sup> Department of Medical Oncology, Hospital Universitario Puerta de Hierro Majadahonda, Majadahonda, Spain*

#### *Edited by:*

*Carsten Carlberg, University of Eastern Finland, Finland*

#### *Reviewed by:*

*Enikö Kallay, Medical University of Vienna, Austria Johannes Van Leeuwen, Erasmus MC, Netherlands*

#### *\*Correspondence:*

*Alberto Muñoz, Instituto de Investigaciones Biomédicas "Alberto Sols," Consejo Superior de Investigaciones Científicas, Universidad Autónoma de Madrid, Arturo Duperier 4, E-28029 Madrid, Spain e-mail: amunoz@iib.uam.es*

Many studies in different biological systems have revealed that 1**α**,25-dihydroxyvitamin D3 (1**α**,25(OH)2D3) modulates signaling pathways triggered at the plasma membrane by agents such as Wnt, transforming growth factor (TGF)-**β**, epidermal growth factor (EGF), and others. In addition, 1α,25(OH)2D3 may affect gene expression by paracrine mechanisms that involve the regulation of cytokine or growth factor secretion by neighboring cells. Moreover, post-transcriptional and post-translational effects of 1α,25(OH)2D3 add to or overlap with its classical modulation of gene transcription rate. Together, these findings show that vitamin D receptor (VDR) cannot be considered only as a nuclear-acting, ligand-modulated transcription factor that binds to and controls the transcription of target genes. Instead, available data support the view that much of the complex biological activity of 1α,25(OH)2D3 resides in its capacity to interact with membrane-based signaling pathways and to modulate the expression and secretion of paracrine factors. Therefore, we propose that future research in the vitamin D field should focus on the interplay between 1α,25(OH)2D3 and agents that act at the plasma membrane, and on the analysis of intercellular communication. Global analyses such as RNA-Seq, transcriptomic arrays, and genome-wide ChIP are expected to dissect the interactions at the gene and molecular levels.

**Keywords: 1α,25(OH)2D3, VDR, membrane-based signaling, Wnt, growth factors, cytokines, paracrine effects**

## **INTRODUCTION**

The active vitamin D metabolite 1α,25-dihydroxyvitamin D3 (1α,25(OH)2D3) is a key regulator of gene expression in higher organisms. It modulates the activity of the vitamin D receptor (VDR), a member of the superfamily of nuclear hormone receptors that regulate gene transcription. Genome-wide chromatin immunoprecipitation studies have shown that VDR binds to hundreds of genome sites even in the absence of 1α,25(OH)2D3 and that ligand binding increases and partially changes these binding sites, which depend on the cell type and the duration of treatment (Carlberg and Campbell, 2013). While a subset of VDR binding sites may be responsible for the control of gene expression (VDREs or vitamin D response elements), others might be temporary anchorage places for a population of unliganded "dormant" VDR. According to the classical view, VDR binds DNA as heterodimers with a retinoid X receptor (RXRα, β, or γ) and, upon ligand binding, changes the transcription rate of neighboring genes.

However, many genes whose expression is altered by 1α,25(OH)2D3 do not contain VDREs. Putative mechanisms of this action include post-transcriptional regulation via changes in the levels of microRNAs that modulate the half-life and/or translation of their messenger RNAs (Thorne et al., 2011; Wang et al., 2011; Alvarez-Díaz et al., 2012; Kasiappan et al., 2012; Guan et al., 2013). Also, 1α,25(OH)2D3 may regulate genes post-translationally via changes in the phosphorylation or other modifications of proteins which affect their stability (Lin et al., 2003; Li et al., 2004), or through changes in the level or activity of proteases that target them (Alvarez-Díaz et al., 2010).

Increasing importance has recently been accorded to another mechanism of 1α,25(OH)2D3 action: the modulation of signaling pathways triggered by other agents at the plasma membrane. Indeed, a number of studies have shown that 1α,25(OH)2D3 modulates the effects of growth factors and cytokines by altering either their cytosolic signaling pathways or the activity of target transcription factors in the nucleus, or even in a paracrine fashion by inhibiting their synthesis and secretion by neighboring cells.

Here we review the available data on these non-classical, alternative mechanisms by which 1α,25(OH)2D3 modulates gene expression. Notably, for specific genes such as c-*MYC*, both direct transcriptional and indirect modes of regulation by 1α,25(OH)2D3 have been described (Pan and Simpson, 1999; Pálmer et al., 2001; Toropainen et al., 2010; Salehi-Tabar et al., 2012).

## **INTERACTION OF 1α,25(OH)2D<sup>3</sup> WITH WNT, HEDGEHOG, AND NOTCH PATHWAYS**

Wnt, Hedgehog, and Notch signaling pathways, which have long been known to play crucial roles during development, are now considered critical for many tumorigenic processes in which they function abnormally due to mutation and/or changes in expression of components.

Larriba et al. Vitamin D and signaling pathways

Wnt factors activate several signaling pathways upon binding to different plasma membrane receptors: the canonical or Wnt/β-catenin and the non-canonical (planar polarity, Ca2<sup>+</sup> *...*) pathways (Clevers and Nusse, 2012). Activation of the Wnt/β-catenin pathway by mutation of *APC* or *AXIN* tumor suppressor genes or of *CTNNB1*/β-catenin oncogene together with changes in the expression of a number of regulatory genes (*SFRPs*, *DICKKOPF (DKK)s...*) is a hallmark of most colorectal cancers and of a variable proportion of several other malignancies (Clevers and Nusse, 2012). A series of studies report that 1α,25(OH)2D3 antagonizes Wnt/β-catenin signaling in colon cancer cells by several mechanisms: the reduction of transcriptionally active β-catenin/T-cell factor complexes, the induction of β-catenin relocation from the nucleus toward the *adherens junctions* structures at the plasma membrane, and the increase in the level of the Wnt inhibitor DKK-1 (Pálmer et al., 2001; Shah et al., 2006; Aguilera et al., 2007) (**Figure 1**). In this way, the pathway endpoint, i.e., the activation of β-catenin target genes, is attenuated by 1α,25(OH)2D3 (Pálmer et al., 2001). Emphasizing the importance of this action, an additional indirect mechanism of Wnt/β-catenin antagonism in colon cancer has been proposed involving IL-1β, which will be reviewed in section 1α,25(OH)2D3 and Cytokines. Although 1,25(OH)2D3 inhibits β-catenin/TCF transcriptional activity in colon and other cancer cells, the upregulation of the Wnt/β-catenin pathway by either ligand-activated or unliganded VDR has been described in osteoblasts and keratinocytes, where it promotes bone formation and hair follicle differentiation, respectively (Larriba et al., 2013). However, the results reported in keratinocytes are controversial: while VDR enhances Wnt signaling through direct binding to Lymphocyte Enhancer-binding Factor (LEF)-1 independently of ligand and βcatenin (Luderer et al., 2011), ligand-activated VDR is believed to inhibit Wnt/β-catenin signaling (Bikle, 2011; Jiang et al., 2012).

Inhibition of Hedgehog (Hh) signaling by vitamin D compounds has also been suggested. In a study combining experiments in zebrafish, the yeast *Pichia pastori* and mouse fibroblasts, secreted vitamin D3, or its precursor 7-dehydrocholesterol (7-DHC), was shown to mediate the paracrine inhibition of Smoothened (Smo) by Patched (Ptch)1 which leads to pathway inactivation (Bijlsma et al., 2006). In the model proposed, which includes the binding of vitamin D3 to Smo at high (micromolar) concentrations, Hh ligands activate the pathway by blocking the induction of the secretion of vitamin D3/7-DHC by Ptch1 (Bijlsma et al., 2006) (**Figure 1**). The Hh pathway is aberrantly activated in basal cell carcinoma, the most frequent human tumor type. Interestingly, 1α,25(OH)2D3 inhibits proliferation and induces differentiation of mouse basal cell carcinomas and embryonal rhabdomyosarcomas with an activated Hh pathway due to *Ptch1* deletion (Uhmann et al., 2011, 2012). As in the previous study, 1α,25(OH)2D3 acts at the level of Smo in a VDRindependent manner (**Figure 1**). Curiously, Tang et al. found that vitamin D3 inhibits Hh and cell proliferation more effectively than 7-DHC, 25(OH)D3, or 1α,25(OH)2D3 in murine basal cell carcinoma cells (Tang et al., 2011). Vitamin D3 also inhibits proliferation and Hh pathway through inactivation of Smo in cultured mouse pancreatic adenocarcinoma cells, but has no antitumor activity *in vivo* (Brüggemann et al., 2010). A common concern in all these studies is the high concentration of vitamin D3 required to observe the reported effects. Research in *Vdr*-deficient mice and in mouse skin explants has shown that lack of VDR increases the expression of several components of the Hh pathway such as *Shh*, *Smo*, *Gli1*, *Gli2*, and *Ptch1*, while 1α,25(OH)2D3 suppresses their expression (Bikle et al., 2013) (**Figure 1**). However, the interaction between 1α,25(OH)2D3 and Hh signaling in human skin remains to be elucidated.

Few studies link vitamin D with Notch signaling. Differentiation of human osteoblasts with vitamin D3 and dexamethasone distinctly affects the expression of Notch receptor family members (Schnabel et al., 2002). In rodent osteoblasts, the transcription factor Hes-1, which is an effector of the Notch pathway, enhances the induction of *SPP1*/osteopontin transcription by 1α,25(OH)2D3, indicating the collaboration of 1α,25(OH)2D3 and Notch pathways in bone remodeling (Shen and Christakos, 2005) (**Figure 1**). Transcriptomic analyses in human RWPE1 immortalized non-tumorigenic prostate cells showed the reduction of the RNA levels of the NOTCH ligands *JAGGED* (*JAG)1*, *JAG2*, and *Delta-like* (*DLL)1* by 1α,25(OH)2D3 (Kovalenko et al., 2010) (**Figure 1**). By contrast, no changes in the expression of *NOTCH-1* and *JAG1* were detected in cultured human keratinocytes upon 1α,25(OH)2D3 treatment (Reichrath and Reichrath, 2012). As *JAG1* transcription and, consequently, Notch signaling are upregulated by Wnt/β-catenin in colorectal cancer cells (Rodilla et al., 2009), the repressive effect of 1α,25(OH)2D3 on the Notch pathway in this system may be secondary to the antagonism of the Wnt/β-catenin pathway (**Figure 1**).

## **INTERPLAY OF 1α,25(OH)2D<sup>3</sup> WITH AGENTS THAT TRIGGER SIGNALING PATHWAYS VIA PLASMA MEMBRANE KINASE RECEPTORS**

There is mutual antagonism between 1α,25(OH)2D3 and epidermal growth factor (EGF), a potent mitogen, in primary colon epithelial cells and in established colon (Caco-2) and breast (T47D) tumor cell lines. This is based on the cross-inhibition of the expression of their respective receptors, VDR and EGFR (Tong et al., 1998, 1999). However, it is a cell type-dependent effect as EGF increases VDR in the rat small intestine and 1α,25(OH)2D3 increases EGFR in BT-20 breast cancer cells (Bruns et al., 1989; Desprez et al., 1991). In addition, 1α,25(OH)2D3 inhibits EGFR signaling by increasing the level of E-cadherin protein at the plasma membrane, which downregulates EGFR (Pálmer et al., 2001; Andl and Rustgi, 2005), and by decreasing that of SPROUTY-2, a cytosolic protein that reduces EGFR ubiquitination, internalization and degradation (Cabrita and Christofori, 2008; Barbáchano et al., 2010).

Transforming growth factor (TGF)-β has opposite roles in carcinogenesis: it inhibits proliferation of normal epithelial cells, but it later induces epithelial-mesenchymal transition, immunosuppression and metastasis (Pickup et al., 2013). 1α,25(OH)2D3 induces the expression of the type I TGF-β receptor and both agents, 1α,25(OH)2D3 and TGF-β, cooperate in Caco-2 cell growth inhibition (Chen et al., 2002; Pálmer et al., 2003). Moreover, Smad3, a mediator of TGF-β signaling, is a co-activator of VDR and contributes to gene regulation by 1α,25(OH)2D3

(Yanagisawa et al., 1999), an effect that is abrogated by Smad7 in transfected COS-7 cells (Yanagi et al., 1999). Reinforcing the interaction between both signaling pathways, 1α,25(OH)2D3 induces the expression of Smad anchor for receptor activation (SARA) (Pálmer et al., 2003), which maintains the epithelial phenotype by recruiting Smads 2/3 to the activated TGF-β receptors and regulates endocytic trafficking of EGFR and other proteins (Tang et al., 2011; Kostaras et al., 2013). Notably, a recent study of R. M. Evans' group has revealed a genome-wide overlap of VDR and Smad3 binding sites that is responsible for the abrogation by VDR ligands of the TGF-β1-mediated activation of hepatic stellate cells during liver fibrosis (Ding et al., 2013). These authors show that TGF-β1 signaling redistributes VDR-binding sites in the genome and facilitates VDR binding at Smad3 profibrotic target genes. Upon ligand activation, VDR binding at coregulated genes decreases Smad3 occupancy at these sites, causing inhibition of fibrosis (Ding et al., 2013). This is a regulatory feedback mechanism in which VDR ligands limit the fibrotic process and so ensure an appropriate non-pathological tissue response. Given the crucial roles of TGF-β in carcinogenesis, future studies should examine whether vitamin D compounds play similar roles in the maintenance of epithelial integrity opposing the onset of carcinomas.

1α,25(OH)2D3 and TGF-β interact also in bone. Curiously, in rat (UMR 106 and ROS 17/2.8) and human (MG-63) osteoblastic cells TGF-β increases VDR expression but inhibits the stimulation of osteocalcin and osteopontin transcription and RNA levels by 1α,25(OH)2D3 (Staal et al., 1994). TGF-β exerts this inhibitory effect by reducing the binding of VDR-RXR complexes to VDREs localized in the promoter of these genes without affecting the nuclear availability of VDR at least in ROS 17/2.8 cells (Staal et al., 1996). In contrast to the stimulation of osteocalcin synthesis in human and rat cells, 1α,25(OH)2D3 decreases osteocalcin production in mouse fetal long bone cultures and neonatal osteoblastic MC3T3 cells while stimulating bone resorption (Staal et al., 1998). This bone resorption action of 1α,25(OH)2D3 is dose-dependently inhibited by TGF-β (Staal et al., 1998).

A complex, cell type-, context- and sometimes age-dependent relation exists between 1α,25(OH)2D3 and insulin-like growth factors (IGF)-I and II. For instance, in C2C12 myoblasts 1α,25(OH)2D3 decreases IGF-I expression while it increases that of IGF-II (Garcia et al., 2011). In HT29 colon carcinoma cells several vitamin D compounds inhibit the secretion of IGF-II thus attenuating its cell proliferation activity (Oh et al., 2001). In addition, 1α,25(OH)2D3 blocks the mitogenic activity of insulin and IGF-I in MCF7 breast cancer cells, at least in part due to the inhibition of c-FOS upregulation (Vink-van Wijngaarden et al., 1996). 1α,25(OH)2D3 and IGF-I have also opposite effects on mouse long bones: IGF-I increases osteocalcin production, which is completely blocked by 1α,25(OH)2D3, and inhibits the enhancement of bone resorption caused by 1α,25(OH)2D3 (Staal et al., 1998). Furthermore, 1α,25(OH)2D3 variably regulates the expression of several IGF binding proteins (IGFBPs), a group of molecules with pleiotropic actions that transport IGFs and also modulate cell survival/apoptosis: 1α,25(OH)2D3 induces IGFBP3 expression in SW480-ADH colon carcinoma, SaOS-2 osteosarcoma, PC3 prostate cancer, MCF7 breast carcinoma and MCF-10A normal mammary cells (Pálmer et al., 2003; Matilainen et al., 2005; Malinen et al., 2011; Brosseau et al., 2013), IGFBP6 in SaOS-2, SW480-ADH and colon carcinoma HT29 cells (Oh et al., 2001; Pálmer et al., 2003; Matilainen et al., 2005), IGFBP1 and IGFBP5 in SaOS-2 and PC3 cells, and IGFBP4 in SaOS-2 cells (Matilainen et al., 2005). Conversely, 1α,25(OH)2D3 represses IGFBP4 in HT29 and SW480-ADH cells, and IGFBP2 in HT29 cells (Oh et al., 2001; Pálmer et al., 2003). In ovarian cells, 1α,25(OH)2D3 alone induces IGFBP1 production but, conversely, it enhances the inhibitory effect of insulin (Parikh et al., 2010). Curiously, recent studies show that IGFBP3 interacts with VDR (Li et al., 2013) and that IGFBP6 binds VDR and blocks the induction of osteoblast differentiation by 1α,25(OH)2D3 (Cui et al., 2011).

Cell type-dependent effects of 1α,25(OH)2D3 have also been described for hepatocyte growth factor (HGF) signaling. 1α,25(OH)2D3 activates the *HGF* gene promoter and induces HGF expression and secretion in rat NRK-49F renal interstitial fibroblasts (Li et al., 2005) and in human keloid fibroblasts (Zhang et al., 2011). Consistently with these results, vitamin D deficiency reduces HGF and HGF receptor/c-Met expression during liver regeneration in rats (Goupil et al., 1997). Conversely, 1α,25(OH)2D3 decreases the level of *HGF* RNA in human HL-60 promyelocitic leukemia cells (Inaba et al., 1993), smooth muscle cells (Shalhoub et al., 2010) and MG-63 osteosarcoma cells (Chattopadhyay et al., 2003). Moreover, the expression of c-Met is inhibited by 1α,25(OH)2D3 in human MHCC97 hepatocellular cell line (Wu et al., 2007). Curiously, 1α,25(OH)2D3 and HGF cooperate to increase osteogenic differentiation of human bone marrow stem cells and maturation of chondrocyte progenitor cells (Grumbles et al., 1996; D'Ippolito et al., 2002; Chen et al., 2011, 2012). Also, 1α,25(OH)2D3 and HGF additively inhibit proliferation of androgen-unresponsive prostate cancer cells (Qadan et al., 2000).

In concordance with its regulatory role in the organism, 1α,25(OH)2D3 favors physiological and homeostatic angiogenesis but inhibits angiogenesis in pathological conditions. Thus, 1α,25(OH)2D3 promotes myogenic differentiation of C2C12 cells by increasing the expression of two key angiogenic factors: vascular endothelial growth factor (VEGF) and fibroblast growth factor-1 (Garcia et al., 2013). In addition, 1α,25(OH)2D3 stimulates pro-angiogenic properties of endothelial progenitor cells by increasing VEGF levels (Grundmann et al., 2012). 1α,25(OH)2D3 also upregulates VEGF expression in osteoblast-like cells but not in breast cancer cells (Schlaeppi et al., 1997). Likewise, ED-71, a vitamin D analog, enhances VEGF expression and promotes angiogenesis in a murine bone marrow ablation model (Okuda et al., 2007). Indeed, increased production of VEGF in vascular smooth muscle cells results from the activation of a VDRE present in the *VEGF* gene promoter (Cardus et al., 2009). By contrast, 1α,25(OH)2D3 downregulates hypoxia-inducible factor (HIF)-1 and VEGF protein expression in several human colon, prostate and breast cancer cell lines (Ben-Shoshan et al., 2007), decreases VEGF production by human lumbar annulus cells (Gruber et al., 2008), and protects against diabetic retinopathy in rats by inhibiting VEGF expression in the retina (Ren et al., 2012).

1α,25(OH)2D3 also modulates the activity of signaling pathways mediated by other types of plasma membrane receptors such us G protein-coupled receptors. Shen et al. found that 1α,25(OH)2D3 suppresses the expression of parathyroid hormone-related protein (PTHrP) in prostate cancer cells via a negative VDRE localized within the non-coding region of the gene, thus antagonizing the induction of cell proliferation and of the expression of the pro-invasive integrin α6β<sup>4</sup> exerted by PTHrP signaling (Shen et al., 2007).

## **1α,25(OH)2D<sup>3</sup> AND CYTOKINES**

The anti-inflammatory and immunomodulatory actions, and thus some of the anticancer and antimicrobial effects of 1α,25(OH)2D3, are mediated by the regulation of cytokine production and/or through the control of their receptors or downstream signaling pathways. Globally, 1α,25(OH)2D3contributes to the autocrine and paracrine control of innate and adaptative immune responses (Adorini and Penna, 2008).

1α,25(OH)2D3 regulates the function of antigen-presenting cells and T-lymphocytes. It inhibits Th1 cells differentiation and, therefore, the secretion of Th1-type cytokines, enhances the development of Th2 cells, and induces tolerogenic monocytes and dendritic cells. IL-4 and IL-10 are among the commonly increased cytokines, while IL-1, IL-2, IL-6, IL-17, tumor necrosis factor (TNF)-α and interferon (IFN)-γ are decreased (Adorini and Penna, 2008).

Mechanistically, ligand-activated VDR directly downregulates the expression of IL-10, IL-2, and IL-12B in lipopolysaccharidetreated human monocytes (THP-1) through its binding to VDREs located in the genomic regions of these genes and the recruitment of the co-repressor NCOR/SMRT and histone deacetylases (Matilainen et al., 2010a,b; Gynther et al., 2011). Remarkably, IL-10 is downregulated by short 1α,25(OH)2D3 treatment (8 h) but upregulated at late time points (48 h) (Matilainen et al., 2010a). In addition, direct VDR binding to a single VDRE mediates the upregulation of *IL-8* gene by 1α,25(OH)2D3 in undifferentiated and differentiated THP-1 cells (Ryynänen and Carlberg, 2013).

1α,25(OH)2D3 also changes the expression of target genes in immune cells by repressing crucial transcription factors such as nuclear factor *kappa* B (NFkB) and signaling pathways such as Janus kinase-signal transducer and activator of transcription (JAK-STAT) (Yu et al., 1995; Muthian et al., 2006; Geldmeyer-Hilt et al., 2011). 1α,25(OH)2D3 also represses NFκB activity in fibroblasts and adipocytes (Harant et al., 1998; Mutt et al., 2012), and fibroblasts lacking VDR have increased NFκB activity (Sun et al., 2006). Direct (increase in IκBα expression and reduction of nuclear translocation of p65) and indirect (upregulation of IGFBP3 and clusterin) mechanisms contribute to the inhibition of NFκB activation (Krishnan and Feldman, 2010). D. Feldman's group has reported that, in addition to inhibiting NFκB, the antiinflammatory effects of 1α,25(OH)2D3 in prostate cancer cells include the reduction of pro-inflammatory prostaglandins (PG) production via suppression of ciclooxygenase-2, downregulation of PG receptors, and upregulation of 15-hydroxyprostaglandin dehydrogenase, which inactivates PGs (Krishnan and Feldman, 2010). Moreover, 1α,25(OH)2D3 decreases the synthesis of proinflammatory IL-6 through the inactivation of p38 kinase due to the upregulation of the mitogen kinase phosphatase (MKP)5 and the blockade of TNF-α (Krishnan and Feldman, 2010). In Jurkat cells, the repression of *IL-2* gene by 1α,25(OH)2D3 is at least partially due to the blockade of NFATp/AP-1 complex formation at a positive regulatory NFAT-1 site, which is bound by VDR-RXR heterodimers (Alroy et al., 1995).

1α,25(OH)2D3 reduces the secretion of interleukin (IL)1-β in THP macrophages by blocking the activation of STAT1 (Kaler et al., 2009). As IL1-β activates the Wnt/β-catenin pathway in colon carcinoma cells via inhibition of GSK3β activity and subsequent stabilization and nuclear translocation of β-catenin, this mechanism may contribute to the antagonism of Wnt signaling by 1α,25(OH)2D3 (Kaler et al., 2009) (**Figure 1**). Curiously, IL-1α is believed to be upregulated and to mediate the antiproliferative effects of 1α,25(OH)2D3 in prostate progenitor/stem cells (Maund et al., 2011).

In human osteoblasts, 1α,25(OH)2D3 completely overrules the inhibitory effect of IFN-β on mineralization. This dominant effect on osteoblast differentiation and bone formation is reflected in the downregulation of IFN-related and -regulated genes by 1α,25(OH)2D3 (Woeckel et al., 2012). Concomitantly, 1α,25(OH)2D3 also induces activin A, a strong inhibitor of mineralization, and represses follistatin, the natural antagonist of activin A, to ensure a fine-tuned regulation of the mineralization process (Woeckel et al., 2013b).

Recent findings have underscored the complexity of 1α,25(OH)2D3 action and its role in the antimicrobial response as part of innate and adaptative immunity. Thus, activation of macrophage Toll-like receptors (TLRs) by intracellular bacteria such as *Mycobacterium tuberculosis* upregulates *VDR* and *CYP27B1* genes that allow the induction of the antimicrobial peptide cathelicidin by 1α,25(OH)2D3 (Liu et al., 2006). In monocytes, TLR activation triggers induction of defensin β4 (*DEFB4*) gene requiring the cooperation between IL-1β and 1α,25(OH)2D3, which is explained by the presence of one VDRE and two IL-1β-activatable NFkB sites in the *DEFB4* promoter (Liu et al., 2009). In addition, 1α,25(OH)2D3 is required for the antimicrobial effect of IFN-γ in human macrophages (Fabri et al., 2011). Moreover, by inducing the expression of TLR2 and CD14 receptors and cathelicidin, 1α,25(OH)2D3 mediates the effect of TGF-β favoring the response to microbial infection and wound injury by keratinocytes (Schauber et al., 2007). These findings show also unexpected cooperation of 1α,25(OH)2D3 with agents (IL-1β, TGF-β, IFN-γ) that are antagonistic in other cell types.

## **INTERPLAY OF 1α,25(OH)2D3/VDR WITH TRANSCRIPTION FACTORS**

Liganded or unliganded VDR interacts with or regulates the expression of a number of transcription factors that are downstream effectors of different signaling pathways (**Table 1**). An interesting example is the upregulation by 1α,25(OH)2D3 of *CDKN1B*/p27Kip1, a cell cycle regulator gene which lacks VDREs. 1α,25(OH)2D3 was first shown to induce *CDKN1B* transcription by stimulating the binding of Sp1 and NF-Y transcription factors to the *CDKN1B* promoter in the myelomonocytic U937 cell line (Inoue et al., 1999). Later, direct VDR-Sp1 interaction at the promoter Sp1 sites was described as responsible for this effect (Huang et al., 2004). In addition to the enhancement of transcription, 1α,25(OH)2D3 increases the stability of p27Kip1 protein by repressing p45Skp2, an F-box protein, through the induction of VDR-Sp1 complexes that together with histone deacetylase 1 are recruited to Sp1 sites at the p45Skp2 gene promoter (Lin et al., 2003; Li et al., 2004; Huang and Hung, 2006).

The granulocyte-macrophage colony-stimulating factor (*GM-CSF*) gene is another example of unusual regulation by 1α,25(OH)2D3. Ligand-activated VDR represses *GM-CSF* through a composite DNA element recognized by Jun-Fos heterodimers (AP-1) and nuclear factor of activated T-cells (NFAT)1 (Towers et al., 1999). In the absence of RXR, VDR binds to c-Jun and stabilizes AP-1 bound to DNA, which outcompetes NFAT1 and decreases *GM-CSF* transcription. In Caco-2 cells, 1α,25(OH)2D3 stimulates AP-1 via activation of protein kinase C-α, ERK and JNK leading to cell differentiation (Chen et al., 1999).

**Table 1 | Interplay between VDR and other transcription factors.**


In renal cells, 1α,25(OH)2D3 suppresses renin gene expression by blocking the cyclic AMP response element (CRE) through direct binding of VDR to CRE-binding protein (CREB) and so, inhibiting the binding of CREB to the CRE (Yuan et al., 2007). By a complex mechanism, 1α,25(OH)2D3 also regulates several Forkhead box (FOX) transcription factors. Ligandactivated VDR binds FOXO3a and FOXO4 together with their regulators, sirtuin 1 deacetylase and protein phosphatase 1, inducing deacetylation and dephosphorylation of FOXO proteins, thereby activating these (An et al., 2010). In the case of the p53 tumor suppressor protein a mutual regulation takes place: while mutated p53 interacts physically with VDR and changes VDR-target genes, converting 1α,25(OH)2D3 from a pro-apoptotic into an anti-apoptotic agent (Stambolsky et al., 2010), 1α,25(OH)2D3 activates the promoter of Mdm2 in a p53-dependent fashion promoting the expression of this negative regulator of p53 protein stability and function (Chen et al., 2013).

Multiple interplays between 1α,25(OH)2D3/VDR and other nuclear receptor ligands have been described. Among them, crosstalk between liganded VDR and peroxisome proliferatoractivated receptor (PPAR)-α/δ in melanoma cells (Sertznig et al., 2009a,b) that may involve the stimulation of PPAR-δ expression by 1α,25(OH)2D3 (Dunlop et al., 2005). A synergistic action of 1α,25(OH)2D3 and rosiglitazone, a PPAR-γ ligand, has been shown during osteoblast-mediated mineralization (Woeckel et al., 2013a), while in human T47D breast cancer cells PPARγ binds VDR and represses its transcriptional activity, possibly also by competing for RXR heterodimerization (Alimirah et al., 2012). Titration out of common co-activators, but not of RXR, may be the mechanism by which ligand-bound VDR represses retinoic acid receptor (RAR) transactivation in GH4C1 pituitary cells (Jiménez-Lara and Aranda, 1999). The relation between 1α,25(OH)2D3 and retinoic acid is however complex, as cooperative effects on target genes and cellular outcome (proliferation inhibition and differentiation) have been described in other systems (Tavera-Mendoza et al., 2006; Anand et al., 2008; Ng et al., 2010). As for estrogen receptor (ER), D. Feldman's group has shown that 1α,25(OH)2D3 exerts a multilevel protective effect against breast cancer that includes the inhibition of estrogen synthesis through the direct and indirect repression of aromatase (CYP19) and the downregulation of ER-α expression through two VDREs in its promoter region (Krishnan et al., 2010; Swami et al., 2013). Likewise, there is a complex and unresolved relationship between 1α,25(OH)2D3 and androgen receptor (AR) synthesis and signaling. 1α,25(OH)2D3 induces AR in LNCaP cells (Zhao et al., 1999) while AR reduces VDR transcriptional activity (Ting et al., 2005), perhaps in some cells by a mechanism mediated by prohibitin (Mooso et al., 2010). In addition, 1α,25(OH)2D3 inhibits glucuronidation and so, inactivation of androgen in prostate cancer cells through the repression of UDP-glucuronosyltransferases (UGT) 2B15 and 2B17, which is counterintuitive given the growth promoting action of androgen and the antiproliferative effect of 1α,25(OH)2D3 in prostate cancer cells (Kaeding et al., 2008). In human bladder, 1α,25(OH)2D3 and analogs inhibit cell proliferation promoted by androgen and keratinocyte growth factor and induce apoptosis at least in part by repressing Bcl-2 expression (Crescioli et al., 2005).

Pituitary transcription factor (Pit)-1 activates growth hormone and prolactin genes in the anterior pituitary and also in breast cancer cells (Seoane and Pérez-Fernández, 2006). In MCF7 cells, VDR homodimers bind the *PIT-1* promoter and inhibit its expression in the presence of 1α,25(OH)2D3 without involvement of RXR (Seoane and Pérez-Fernández, 2006).

### **CONCLUSIONS**

The available evidence shows that the classical view of VDR only as a nuclear-acting ligand-modulated transcription factor that regulates the rate of transcription of those genes to which it binds is outdated. Instead, VDR and its ligand constitute a multilevel main regulator of gene expression in higher cells acting directly or indirectly, and via a variety of different mechanisms, on many signaling pathways. Some of them are triggered from the plasma membrane by paracrine or endocrine agents, and 1α,25(OH)2D3 interacts at different levels: membrane receptors, cytosolic signaling molecules or effector nuclear transcription factors. In most cases 1α,25(OH)2D3 action is mediated by nuclear VDR but in a few others this is unclear and non-canonical VDR-independent or extranuclear effects have been proposed. Available studies show that 1α,25(OH)2D3 and these signaling pathways interact variably and with distinct outcomes in a cell/tissue-specific fashion and sometimes also differentially between normal and malignant cells.

#### **PERSPECTIVES**

The increasingly recognized importance of its non-cell autonomous actions has widened the scope of the study of 1α,25(OH)2D3. On the one hand, an in-depth study of the interplay between 1α,25(OH)2D3 and other agents, which seems to be cell-specific in terms of biological outcome, is necessary to elucidate the possibilities of combined therapies using vitamin D compounds and inhibitors or activators of a variety of signaling pathways. On the other hand, several of these interactions take place at the intercellular level. By using high-throughput techniques and genome-wide analyses, we expect to be able to identify secreted paracrine and intracellular mediators of the interaction between 1α,25(OH)2D3 and other signaling pathways responsible for the regulatory actions of 1α,25(OH)2D3 in the organism. Future research should aim to discern how vitamin D compounds modulate tissue and organ physiology and how they may be used to treat pathological processes such as infections, autoimmune disorders, or cancer.

#### **AUTHOR CONTRIBUTIONS**

María Jesús Larriba, José Manuel González-Sancho, Félix Bonilla, and Alberto Muñoz wrote the manuscript.

#### **ACKNOWLEDGMENTS**

We thank Robin Rycroft for his valuable assistance in the preparation of the English manuscript. The work in the authors' laboratory is supported by Ministerio de Economía y Competitividad of Spain (SAF2010-18302, BFU2010-19659), Fondo Europeo de Desarrollo Regional-Instituto de Salud Carlos III (RD12/0036/0021, RD12/0036/0041), and Comunidad de Madrid (S2010/BMD-2344, Colomics2).

## **REFERENCES**


A production to fine-tune osteoblast-induced mineralization. *J. Cell. Physiol.* 228, 2167–2174. doi: 10.1002/jcp.24388


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

*Received: 10 January 2014; paper pending published: 21 January 2014; accepted: 30 January 2014; published online: 18 February 2014.*

*Citation: Larriba MJ, González-Sancho JM, Bonilla F and Muñoz A (2014) Interaction of vitamin D with membrane-based signaling pathways. Front. Physiol. 5:60. doi: 10.3389/fphys.2014.00060*

*This article was submitted to Integrative Physiology, a section of the journal Frontiers in Physiology.*

*Copyright © 2014 Larriba, González-Sancho, Bonilla and Muñoz. 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.*

## Vitamin D and the epigenome

## *Irfete S. Fetahu , Julia Höbaus and Eniko Kállay ˝ \**

*Department of Pathophysiology and Allergy Research, Center of Pathophysiology, Infectiology and Immunology, Comprehensive Cancer Center, Medical University of Vienna, Vienna, Austria*

#### *Edited by:*

*Carsten Carlberg, University of Eastern Finland, Finland*

#### *Reviewed by:*

*Patsie Polly, The University of New South Wales, Australia Moray J. Campbell, Roswell Park Cancer Institute, USA*

#### *\*Correspondence:*

*Eniko Kállay, Department of ˝ Pathophysiology and Allergy Research, Medical University of Vienna, Währinger Gürtel 18-20, A-1090 Vienna, Austria e-mail: enikoe.kallay@ meduniwien.ac.at*

Epigenetic mechanisms play a crucial role in regulating gene expression. The main mechanisms involve methylation of DNA and covalent modifications of histones by methylation, acetylation, phosphorylation, or ubiquitination. The complex interplay of different epigenetic mechanisms is mediated by enzymes acting in the nucleus. Modifications in DNA methylation are performed mainly by DNA methyltransferases (DNMTs) and ten-eleven translocation (TET) proteins, while a plethora of enzymes, such as histone acetyltransferases (HATs), histone deacetylases (HDACs), histone methyltransferases (HMTs), and histone demethylases (HDMs) regulate covalent histone modifications. In many diseases, such as cancer, the epigenetic regulatory system is often disturbed. Vitamin D interacts with the epigenome on multiple levels. Firstly, critical genes in the vitamin D signaling system, such as those coding for vitamin D receptor (*VDR*) and the enzymes 25-hydroxylase (*CYP2R1*), 1α-hydroxylase (*CYP27B1*), and 24-hydroxylase (*CYP24A1*) have large CpG islands in their promoter regions and therefore can be silenced by DNA methylation. Secondly, VDR protein physically interacts with coactivator and corepressor proteins, which in turn are in contact with chromatin modifiers, such as HATs, HDACs, HMTs, and with chromatin remodelers. Thirdly, a number of genes encoding for chromatin modifiers and remodelers, such as HDMs of the Jumonji C (JmjC)-domain containing proteins and lysine-specific demethylase (LSD) families are primary targets of VDR and its ligands. Finally, there is evidence that certain VDR ligands have DNA demethylating effects. In this review we will discuss regulation of the vitamin D system by epigenetic modifications and how vitamin D contributes to the maintenance of the epigenome, and evaluate its impact in health and disease.

#### **Keywords: VDR, VDRE, 1,25-dihydroxyvitamin D3, CYP27B1, CYP24A1, DNA methylation, histone modifications, CpG island**

## **INTRODUCTION**

The role of vitamin D in regulating gene expression has become increasingly evident since the discovery of the transcription factor vitamin D receptor (VDR), a member of the steroid nuclear receptor superfamily. The effect of liganded VDR depends on the epigenetic landscape of the target gene. Genome wide analysis in the human leukemia cell line THP-1 showed that VDR binds mainly at loci of open chromatin. Upon treatment with the VDR ligand 1,25-dihydroxyvitamin D3 (1,25-D3), chromatin accessibility further increases in more than 30% of these loci (Seuter et al., 2013). The mechanism of action of the liganded VDR is dependent on binding and action of histone acetyltransferases (HATs) and histone methyltransferases (HMTs). It has been shown that co-treatment of cells with 1,25-D3, and histone deacetylase or DNA methyltransferase inhibitors often have synergistic effects (Pan et al., 2010).

Many common diseases have both genetic and epigenetic components, which communicate in an intricate and multilayered manner. Currently, it is not clear to what extent epigenetic alterations contribute to onset and progress of common diseases, such as cancer. Epigenetics refers to processes that alter gene activity without changing the DNA sequence. They play an important role in regulating key processes during development, including embryonic developmental events, gene imprinting, and inactivation of chromosome X in females (Bird, 2002; Meissner et al., 2008; Tsai and Baylin, 2011). Maintenance of normal functioning of these biological processes is dependent on the intricate interaction between several epigenetic mechanisms, such as DNA methylation, histone modifications, and non-coding RNAs (Jones and Baylin, 2007). Therefore, at a given promoter the marks arising from DNA methylation and histone modifications determine whether the chromatin is in an open (active) or a closed (repressed) state. Deregulation of the epigenetic mechanisms can lead to aberrant DNA methylation patterns and chromatin architecture, which is a common feature in cancer (Baylin and Jones, 2011; Tsai and Baylin, 2011; Helin and Dhanak, 2013).

## **EPIGENETIC CHANGES MEDIATED BY THE VITAMIN D RECEPTOR AND ITS LIGANDS**

The effect of nutrition on the methylation equilibrium of the genome is already accepted as one of the mechanisms preventing either promoter hyper- or global hypomethylation. Several nutrients are renowned for their impact on DNA methylation, such as folic acid, vitamin B, green tea, and alcohol (Arasaradnam et al., 2008). The effect of vitamin D is currently under debate.

Primary epigenetic effects of vitamin D are linked to histone modifications, mainly acetylation. The VDR/RXR dimer interacts with HATs to induce transcriptional activation (Karlic and Varga, 2011). Several studies have suggested that vitamin D may affect also DNA methylation. A recent study associated severe vitamin D deficiency with methylation changes in leukocyte DNA, although the observed differences were relatively small (Zhu et al., 2013). This study suggested that subjects with vitamin D deficiency were more likely to show reduced synthesis and increased catabolism of active vitamin D. Whether this was the cause of the vitamin D deficiency or the consequence thereof is not clear and needs further studies.

## **EFFECT OF VITAMIN D ON DNA METHYLATION**

DNA methylation is the most extensively studied epigenetic mark (Esteller, 2008). In humans, DNA methylation occurs on cytosines followed by guanine (CpG) (Bird, 1980; Gruenbaum et al., 1981). Regions of DNA enriched in CpG clusters form CpG islands (CGI) (Wang and Leung, 2004). DNA methylation is necessary for regulating and orchestrating key biological processes, including cell cycle, differentiation, as well as genomic imprinting (Feinberg et al., 2002; Reik and Lewis, 2005; Jones and Baylin, 2007). DNA hypermethylation is mainly found in intergenic regions and repetitive genomic sequences to maintain these in a transcriptionally inactive chromatin state (Herman and Baylin, 2003).

DNA methyltransferases (DNMTs) mediate DNA methylation (Robertson, 2005). DNMT1 encodes for a maintenance methyltransferase, whereas DNTM3A/3B encode for de novo methyltransferases, which are pivotal to maintain and establish genomic methylation (Okano et al., 1998, 1999; Jin and Robertson, 2013). However, *in vivo* studies suggest that all three DNMTs might exert both de novo and maintenance functions (Rhee et al., 2000, 2002; Kim et al., 2002; Esteller, 2007a). Recently, a new group of enzymes that induce active demethylation of the DNA was discovered, the ten-eleven translocation (TET) enzyme family, which plays an important role both in development and tumorigenesis (Kriaucionis and Heintz, 2009; Ficz et al., 2011; Williams et al., 2011; Yamaguchi et al., 2012; Hackett et al., 2013).

Alterations in the cancer epigenome are generally associated with loss of global DNA methylation and gain of methylation in specific gene promoters (Ting et al., 2006). Loss of global methylation may lead to chromosomal instability (Eden et al., 2003), loss of imprinting (Cui et al., 2003; Bjornsson et al., 2007), and activation of transposable elements, thereby leading to disturbances in the genome (Bestor, 2005; Esteller, 2008). Conversely, hypermethylation of promoter regions of tumor suppressor genes (Greger et al., 1989; Sakai et al., 1991; Esteller, 2008) leads to loss of expression of key genes affecting pathways involved in maintenance of cellular functions, including cell cycle, apoptosis, and DNA repair (Esteller, 2007b). Several bona fide tumor suppressor genes are silenced by promoter hypermethylation in tumors. For instance, hypermethylation of the promoter of the DNA repair gene *hMLH1* is associated with early stages of endometrial and colon cancer, and microsatellite instability phenotype (Esteller et al., 1999). Epigenetically mediated silencing of cyclin-dependent kinase inhibitor 2A, which is crucial for control of cell cycle has been reported in several cancers (Brock et al., 2008; Liau et al., 2014). Additionally, pathways regulated by microRNAs have been associated with DNA hypermethylation-dependent silencing (Saito et al., 2006).

Besides methylating cytosines, DNMTs may coordinate other chromatin-mediated aspects of gene expression at sites of gene promoters (Herman and Baylin, 2003). For example, hypermethylation of promoters of tumor suppressor genes is associated with recruitment of proteins belonging to the methyl CpG-binding domain (MBD) family, MeCP2, MBD1, MBD2, MBD3, and MBD4 (Ballestar and Esteller, 2005). It has been shown that MeCP2 represses transcription of methylated DNA by recruiting histone deacetylases (HDACs), providing the first evidence for interactions between DNA methylation and histone modifications (Jones et al., 1998; Nan et al., 1998).

There is evidence that 1,25-D3 is able to induce DNA demethylation, however, the mechanisms behind the effect of 1,25-D3 on DNA methylation are not clear. In most cases it is probably passive demethylation that happens over several cycles of DNA replication. However, in some cases demethylation occurs within 1–4 h, indicative of an active process (Doig et al., 2013). The fact that vitamin D can alter methylation of DNA in the promoter of certain genes is novel. Tapp and colleagues suggested that in healthy subjects global, age-related CGI methylation of human rectal mucosa was influenced not only by gender, folate availability, and selenium, but also by vitamin D status (Tapp et al., 2013). The authors show negative association between serum 25-D3 level and CGI methylation of the adenomatous polyposis coli (*APC*) promoter region, a tumor suppressor often inactive in colorectal cancer. Interestingly, they observed a weak positive correlation of vitamin D level with methylation of *LINE-1* (genomic long interspersed nuclear element-1), a mammalian autonomous retrotransposon, increasing stability of this region (Tapp et al., 2013). A recent study in colorectal cancer patients investigating two Canadian populations (from Newfoundland and Ontario) found that high dietary vitamin D intake was associated with lower methylation of the two WNT antagonists *dickkopf 1* (*DKK1*) and *WNT5A* (Rawson et al., 2012). This relationship became even more significant in females in the Newfoundland population, while in the Ontario population the association between vitamin D intake and lower methylation was observed only in early stage tumors, but not in late stage tumors (Rawson et al., 2012). These data confer further insights in the mechanisms regulating the transcriptional activating effect of vitamin D on *DKK1* expression described *in vitro* (Aguilera et al., 2007; Pendas-Franco et al., 2008).

Moreover, treatment of the triple negative breast cancer cell line MDA-MB-231 with 1,25-D3 reduced DNA methylation of the *e-cadherin* promoter (Lopes et al., 2012), while another study showed that 1,25-D3 induced demethylation of the *PDZ-LIM* domain-containing protein 2 promoter, leading to increased expression (Vanoirbeek et al., 2014). In non-malignant and malignant prostate epithelial cells, treatment with 1,25-D3 caused clear changes in site-specific methylation of the *p21* promoter, in a cell line-specific manner (Doig et al., 2013).

An interesting interaction between vitamin D and DNA methylation is induction of the expression of GADD45 (growth arrest and DNA damage) protein by 1,25-D3 in several tumor cells (Jiang et al., 2003; Zhang et al., 2006; Bremmer et al., 2012). GADD45A is one of the enzymes that promote epigenetic gene activation by repair mediated DNA demethylation in Xenopus laevis (Barreto et al., 2007).

In summary, alterations in DNA methylation lead to aberrant gene expression and disruptions of genomic integrity, which contribute to development and progression of diseases. Vitamin D can regulate these processes; the mechanisms behind need further investigations.

## **INTERACTIONS OF VITAMIN D WITH CHROMATIN MODULATORS AND REMODELERS**

Nuclear receptors, such as the VDR contain DNA-binding domains that mediate binding to the DNA, presuming the DNA is available and is not wound tightly around nucleosomes. The chromatin context determines nuclear receptor binding and determines which epigenetic modifications will occur thereafter. Upon binding to their genomic response elements, nuclear hormone receptors will then recruit different regulatory cofactor complexes (Lee et al., 2001). The unliganded VDR is able to bind also genomic DNA, where it usually forms complexes with corepressor proteins that either exert HDAC activity, e.g., ALIEN (Polly et al., 2000), or are associated with HDACs, such as NCOR1 and SMRT. The corepressors dissociate upon binding of 1,25-D3, and are replaced by coactivator complexes.

The chromatin environment dictates gene activity throughout the genome. Post-translational modifications of the N-terminal tails of histone proteins allow nucleosomes to shift, the chromatin to relax, and genes to become activated. Histone modifications change in response to environmental stimuli (Meyer et al., 2013). Histones are major protein components of chromatin that undergo post-translational modifications, including acetylation of lysines, methylation of lysines and arginines, and phosphorylation of serine and threonine residues (Esteller, 2008). In epigenetically silenced genes, hypermethylation of CGIs is often associated with loss of acetylation on histone 3 and 4 (H3 and H4), loss of methylation of lysine (K) 4 on H3 (H3K4), and gain of methylation of K9 and K27 on H3 (H3K9 and H3K27) (Esteller, 2008).

Histone acetylation generally correlates with transcriptional activation (Hebbes et al., 1988; Kouzarides, 2007) and is dependent on a dynamic interaction between histone acetyltransferases (HATs) and histone deacetylases (HDACs) (Marks et al., 2001). The balance between the actions of these enzymes is crucial in controlling gene expression, and governs several developmental processes and disease states (Haberland et al., 2009). Generally, HATs are defined as activators of transcription, whereas HDACs as transcription repressors (Parbin et al., 2014). In various cancer types, including prostate, gastric, and breast cancers, overexpression of HDAC1 is often associated with poor clinical outcome (Choi et al., 2001; Halkidou et al., 2004; Zhang et al., 2005). In colorectal cancer patients HDAC1, 2, and 3 are overexpressed, and high HDAC1 and 2 expression is linked with reduced patient survival (Zhu et al., 2004; Wilson et al., 2006). Overexpression of HDAC1 plays a crucial role in regulating proliferation by repressing the expression of the cyclin-dependent kinase inhibitor p21 (Lagger et al., 2003). Additionally, silencing of HDAC4 leads to re-expression of p21, which in turn induces cell growth arrest and tumor growth inhibition, both *in vitro* and *in vivo* in a human glioblastoma model (Mottet et al., 2009). In addition to classical HDACs, another group of enzymes, the sirtuins (silent information regulator 2 proteins) are involved in histone deacetylation (Schwer and Verdin, 2008). Sirtuins have been linked to metabolic disorders, cancer, aging, and also regulation of the circadian rhythm (Guarente, 2006; Longo and Kennedy, 2006; Jung-Hynes et al., 2010).

Many of the coactivators recruited by the VDR, including p160 steroid receptor coactivator proteins (SRC1, 2, and 3), p300, or CBP have lysine acetyltransferase activity (**Figure 1**). Indeed, treatment of THP-1 cells with 1,25-D3 increased H3K27ac at the promoter of several early VDR target genes (Seuter et al., 2013). In genetic hypercalciuric stone forming rats inhibition of bone morphogenetic protein 2 (BMP2) by 1,25-D3, seems to involve H3 deacetylation and H3K9 di-methylation (Fu et al., 2013).

In MDA-MB453 breast cancer cells 1,25-D3 treatment regulates expression of p21 through a mechanism involving both histone acetylation and methylation, probably by dynamic

chromatin looping from distal 1,25-D3 responsive elements to the TSS of *p21* (Saramäki et al., 2009).

Histone methylation can lead either to gene activation or repression, depending on the histone site that is methylated, the degree of methylation (e.g., mono-methylation, di-methylation, or tri-methylation), amino acid residues affected, and their position in the histone tail (Esteller, 2008). Methylation of histones depends on a dynamic process arising from the actions of methyltransferases (HMTs) and demethylases (HDMs) (Shi and Whetstine, 2007; Mosammaparast and Shi, 2010; Greer and Shi, 2012). So far, two protein families capable of demethylating lysines are known, the amine oxidases (Shi et al., 2004) and jumonji C (JmjC)-domain-containing proteins (Cloos et al., 2006; Tsukada et al., 2006). The first histone demethylase discovered was the lysine-specific demethylase 1 (LSD1/KDM1A), an amine oxidase, which demethylates H3K4me2/me1 (**Table 1**) (Shi et al., 2004). High expression of KDM1A in various cancers, including colorectal cancer, prostate cancer, and neuroblastomas is associated with increased cancer recurrence and poor clinical outcome (Kahl et al., 2006; Schulte et al., 2009; Ding et al., 2013). Additionally, it has been shown that LSD1 is essential for androgen and estrogen receptor-dependent gene activation via H3K9me2/me1 demethylation (Metzger et al., 2005; Garcia-Bassets et al., 2007; Perillo et al., 2008). There is a reciprocal regulatory effect between the activity of VDR and histone demethylases. In the colon cancer cell line SW480-ADH 1,25-D3 increased the expression of the lysine-specific demethylase 1 and 2 (Pereira et al., 2012).

1,25-D3 treatment affected also the expression of a series of different JmjC histone demethylases. The first identified member of the JmjC family was KDM2A/JHDM1A (Tsukada et al., 2006). Expression profiling data showed altered expression of KDM2A and KDM2B in several tumors, however, it seems that their proor antioncogenic functions are tissue-dependent (Frescas et al., 2007, 2008; Pfau et al., 2008). 1,25-D3 inhibited the expression of several histone demethylases (e.g., KDM4A/4C/4D/5A/2B, JMJD5/6, PLA2G4B), and induced the expression of others, JARID2 and KDM5B (Pereira et al., 2012). Members of the KDM4 family catalyze tri-demethylation of H3K9 and/or H3K36 (Cloos et al., 2006; Fodor et al., 2006; Klose et al., 2006; Whetstine et al., 2006; Lin et al., 2008). H3K9me3 is a mark for heterochromatin and demethylation of H3K9 is suggested to be linked with chromosomal instability (Cloos et al., 2006). Inhibition of expression of KDM4 family members by 1,25-D3 could thus contribute to genome stability. Members of KDM5 cluster catalyze demethylation of H3K4me3/me2, which is a mark for open chromatin (Christensen et al., 2007; Iwase et al., 2007; Klose et al., 2007; Tahiliani et al., 2007) and their upregulaton upon 1,25-D3 treatment might lead to gene repression (Pereira et al., 2011). Overexpression of KDM5B has been reported in breast and prostate cancers (Lu et al., 1999; Xiang et al., 2007). Deletion of *kdm5b* inhibits tumor growth in a syngeneic mouse mammary


**Table 1 | A simplified list of the members of the two classes of histone demethylases (mentioned in the manuscript).**

*Reviewed in Pedersen and Helin (2010), Greer and Shi (2012). The main histone demethylase families and submembers are indicated. Degree of methylation and site of lysine residue are given. References for individual enzymes can be found throughout the text. H, Histone; K, lysine; me1, mono-methylation; me2, di-methylation; me3, tri-methylation; KDM1A, lysine-specific demethylase 1A; JHDM1D, JmjC-domain-containing histone demethylation protein 1D; PHF, plant homeodomain finger protein.*

tumor (Yamane et al., 2007), suggestive of its potential role in tumor development. 1,25-D3 induced the expression of the histone demethylase KDM6B as well, which is the only other known enzyme, besides KDM6A that is able to demethylate H3K27me3, a histone mark that correlates with gene repression. Furthermore, the authors showed positive correlation between KDM6B and VDR in 96 colon tumor patients, and inverse correlation of KDM6B with SNAIL1, which is involved in epithelial to mesenchymal transition, indicating that probably the antiproliferative role of 1,25-D3 via KDM6B upregulation might take place *in vivo* (Pereira et al., 2011). Interestingly, treatment of SW480-ADH cells with 1,25-D3 had no effect on global H3K27me3 levels, in spite of KDM6B upregulation (Pereira et al., 2011, 2012). The effect of 1,25-D3 on the expression of histone demethylases may well be indirect and could be mediated by microRNAs (Padi et al., 2013). KDM2A is one of the direct targets of microRNA-627. 1,25-D3-dependent upregulation of the microRNA-627 expression both *in vitro*, in the HT-29 colorectal cancer cells and *in vivo*, in tumor xenografts, led to lower KDM2A levels (Padi et al., 2013).

In different pathologies, the expression pattern of the nuclear receptor cofactors is altered, compromising the effect of 1,25-D3 (Doig et al., 2013; Singh et al., 2013). The initial interactions between VDR and coactivators are the seed for the assembly of intricate multiprotein complexes that remodel the chromatin structure, recruit the core transcriptional machinery, and induce expression of 1,25-D3 target genes (**Figure 1**). Often, differences in responsiveness to 1,25-D3 depend on the expression pattern of the coregulators of VDR. In prostate cancer cells, the temporal distribution of the nuclear corepressor NCOR1 at VDR target genes is different in 1,25-D3 responsive cells compared with unresponsive cells (Doig et al., 2013; Singh et al., 2013).

The liganded VDR is able both to transactivate and transrepress target genes. The mechanisms of action are probably different between transactivation and transrepression, and also highly dependent on the motifs of the vitamin D response elements. A highly complex mechanism regulates the liganddependent repression of CYP27B1 (Kim et al., 2007a). CYP27B1 repression requires two epigenetic modifications: deacetylation of histones and methylation of the *CYP27B1* gene promoter and exon regions. This is dependent on the presence of the VDR interacting repressor (VDIR) and the chromatin remodeler Williams Syndrome transcription factor. In the absence of 1,25-D3, VDIR is bound directly to the E-box motifs in the negative VDRE and recruits histone acetyltransferases to induce *CYP27B1* gene transcription. In the presence of 1,25-D3, VDIR acts as a scaffold for the 1,25-D3-VDR complex to repress transcription of *CYP27B1* through recruitment of HDAC2, DNMT1, and DNMT3B (Kim et al., 2007a). It seems that VDIR and HDAC2 are involved also in the 1,25-D3-dependent transrepression of the human parathyroid hormone gene (Kim et al., 2007b). It is not clear, whether this mechanism of transrepression by liganded VDR also applies to other genes. In mesenchymal stem cells 1,25-D3 represses gene expression by binding to promoters with enhanced H3K9Ac and H3K9me2 (Tan et al., 2009). Whether H3K9 acetylation/methylation enabled VDR binding or VDR binding caused H3K9 acetylation, is not clear.

Proper orchestration of histone modifications in crosstalk with other chromatin regulators is crucial in maintaining the epigenetic landscape and governing gene expression. Any disturbances in these constellations may lead to aberrant gene expression. Whether 1,25-D3 affects regulation of other chromatin modulators as well, is not yet known.

## **REGULATION OF THE VITAMIN D SYSTEM**

The vitamin D system has pleiotropic functions and regulates approximately 3% of the human genome (Bouillon et al., 2008). To maintain balance, a strict regulation of the vitamin D system genes is of utmost importance. The main role of liganded VDR in tissues not involved in calcium homeostasis is to control expression of genes that regulate cell proliferation, differentiation, and apoptosis. One major limitation in the therapeutic exploitation of these effects is the resistance of cancer cells to 1,25-D3. Epigenetic corruption of VDR signaling is suggested to be one of the mechanisms that leads to reduced responsiveness to 1,25-D3 actions. This can be caused by promoter methylation of key vitamin D system genes or by skewed accumulation of VDR-associated corepressors, preferentially at promoters of anti-proliferative target genes (Abedin et al., 2006).

Expression of the vitamin D degrading and metabolizing enzymes is regulated through binding of 1,25-D3-liganded VDR to vitamin D responsive elements (VDREs). However, the major regulators of 1,25-D3 levels and signaling CYP2R1, CYP24A1, CYP27B1, and VDR, "the vitamin D tool" genes, are prone to epigenetic regulation. CpG islands span the promoters of *CYP2R1*, *CYP24A1*, and *VDR*, while a CpG island is located within the *CYP27B1* gene locus (**Figure 2**). Therefore, DNA methylation and histone modifications in these regions can change the chromatin state from an open to closed conformation and lead to transcriptional repression of these genes. Expression of vitamin D tool genes becomes deregulated in various types of cancer, and these changes may be partially attributed to epigenetic alterations (reviewed in Hobaus et al., 2013). As early as 1984, Yoneda et al. (1984) have shown that the histone acetyltransferase inhibitor butyrate augments 1,25-D3 actions. Several studies confirmed these findings (e.g., Rashid et al., 2001) and have suggested that the action of butyrate could be through upregulation of VDR or CYP27B1 expression (Gaschott and Stein, 2003). Whether this effect is mediated by direct acetylation of the *VDR* or *CYP27B1* promoters, has not been determined.

In this chapter we discuss evidence for epigenetic regulation through DNA methylation of these genes in health and disease.

### **EPIGENETIC REGULATION OF THE VITAMIN D RECEPTOR**

The VDR is a nuclear receptor mediating 1,25-D3 signaling. It is expressed by at least 38 cell types in the human body (Norman and Bouillon, 2010). In the absence of its ligand 1,25-D3, VDR is mainly found in the cytoplasm (Nagpal et al., 2005). Upon ligand binding, VDR heterodimerizes with the retinoid X receptor (RXR) and translocates to the nucleus, where it binds to vitamin D responsive elements (VDREs) to regulate transcription of 1,25-D3 target genes. This is achieved through recruitment of coactivators or corepressors to the VDR-RXR complex bound to DNA (Nagpal et al., 2005; Pike et al., 2012; Haussler et al., 2013;

Pike and Meyer, 2013). As VDR is rarely mutated during carcinogenesis (Miller et al., 1997), the disturbance of the vitamin D signaling and apparent 1,25-D3 insensitivity in cancer (Marik et al., 2010) must be attributed to other alterations, which may include epigenetic changes, such as promoter methylation.

The *VDR* gene is located on the long arm of chromosome 12 (12q13.11) and contains 4 potential promoter regions. Exons 1a, 1c, and 1d of the *VDR* are well conserved, while 1b, 1e, and 1f show low homology (Halsall et al., 2007). Exon 1a appears to contain a strong promoter, including several transcription factor binding sites (AP-2 and SP1). Transcription was reported to originate also in exons 1d, 1e, and 1f, while translation starts in exon 2 (Halsall et al., 2007). Marik and colleagues performed an *in silico* analysis of the *VDR* gene sequence and reported three CpG islands located in exon 1a spanning from −790 bp to 380 bp relative to the TSS in exon 1a (Marik et al., 2010). According to the UCSC Genome browser, however, only one large CpG island spanning 892 bp in length is found in the *VDR* promoter region (Gardiner-Garden and Frommer, 1987; Karolchik et al., 2014). This discrepancy is likely due to the different search parameters used for CpG island identification.

Epigenetic silencing of VDR was suggested to cause the slow normalization of VDR levels in the parathyroid glands of uremic rats after kidney transplantation (Lewin et al., 2002; Hofman-Bang et al., 2012). However, sequencing of the *VDR* promoter [−250 to 300 bp relative to exon 1 (43 CpGs)] in normal and uremic rats showed no difference between methylation patterns (Hofman-Bang et al., 2012). Further, the authors reported that methylation levels coincided with the negative control, thus showing that promoter methylation does not play a role in regulating VDR expression in the parathyroid glands.

In contrast, promoter methylation was reported to cause repression of *VDR* gene expression in HIV infected T cells. In normal T cells, activation or priming causes an upregulation of VDR expression (Von Essen et al., 2010). In comparison, infection of previously activated T cells with human immunodeficiency virus (HIV) led to upregulation of DNMT3B, increased promoter methylation of *VDR* (45–70%), and decreased *VDR* gene expression (Chandel et al., 2013). This downregulation of VDR could be reversed upon treatment with 5-azacytidine (AZA) suggesting that the decreased expression of VDR by HIV is, at least partially, caused by DNA methylation (Chandel et al., 2013). There is evidence for an inverse correlation between the vitamin D status and infections, however, many trials failed to show a protective effect of vitamin D (reviewed in Yamshchikov et al., 2009). Thus, reduced sensitivity to vitamin D metabolites due to, e.g., downregulation of VDR may account for inconclusive trials. This is supported by a study investigating methylation of the 3- end of *VDR* in two South African groups revealing differences with respect to ethnicity and tuberculosis status of the patients (Andraos et al., 2011).

In breast tumors, methylation of exon 1a of the *VDR* gene was significantly higher (65% of CpGs methylated) compared with normal breast tissue (15% of CpGs methylated) (Marik et al., 2010). *In vitro*, in breast cancer cell lines, three hypermethylated regions in exon 1a became demethylated after treatment with the DNMT1 inhibitor 5-aza-2- -deoxycytidine (DAC) and VDR mRNA expression increased. These regions were in proximity to the SP1 binding sites (approximately 790 bp from TSS), NFκB binding sites (approximately −480 from TSS), and the exon 1a TSS. Treatment with 1,25-D3 had no effect on methylation of these regions (Marik et al., 2010). In other types of cancer, e.g., the choriocarcinoma-derived trophoblast cell lines JEG-3 and JAR, the *VDR* promoter was densely methylated (Novakovic et al., 2009). In contrast, no methylation of the VDR promoter region was observed in colon cancer cell lines, and treatment with DAC did not increase gene expression (Habano et al., 2011; Höbaus et al., 2013a). In parathyroid tumors the expression of VDR is decreased (Gogusev et al., 1997; Carling et al., 2000), however, no differences in DNA methylation of *VDR* were observed between parathyroid tumors and healthy controls (Sulaiman et al., 2013). Similar results were seen in parathyroid adenoma samples, which showed decreased expression of VDR, but showed no promoter methylation (Varshney et al., 2013).

Additionally, it has been suggested that expression of 5 truncated variants of VDR is linked to methylation of the *VDR* promoter. These variants are predominantly found in breast cancer compared with the full length variants expressed in normal breast tissue. Treatment with DAC restored expression of the active transcript variant of VDR in breast cancer cell lines, indicating promoter methylation as cause of truncated protein expression (Marik et al., 2010). The significance of these potentially untranslated truncated variants remains to be investigated, however, as they are not found in normal breast tissue aberrant expression of truncated isoforms may further disrupt vitamin D signaling in tumor tissue.

A recent study suggested that in colorectal cancer metastases, VDR becomes the target of the polycomb group protein enhancer of zeste homolog 2 (EZH2) that mediates VDR downregulation by H3K27 trimethylation in the VDR promoter (Lin et al., 2013). The histone deacetylase HDAC3, one of the most frequently upregulated genes in cancer, seems to inhibit VDR expression. In two colorectal cancer cell lines, HCT116 and SW480 knock down of HDAC3 increased VDR expression and restored sensitivity of these cells to 1,25-D3 (Godman et al., 2008).

Taken together, there is evidence that in various diseases the decreased tissue sensitivity to 1,25-D3 could have been caused by the epigenetic silencing of the VDR.

## **EPIGENETIC REGULATION OF THE CYP2R1**

CYP2R1 is a microsomal P450 enzyme, which hydroxylates both vitamin D2 and D3 at position C-25 to form the circulating storage form of vitamin D 25-D3. The promoter region of *CYP2R1* is located within a CpG island, which can be subjected to epigenetic regulation. So far, only two studies investigated the promoter methylation status of this gene. Genome wide association studies found increased *CYP2R1* promoter methylation in leukocyte DNA from individuals with severe vitamin D deficiency compared with control group (Zhu et al., 2013). Further, methylation levels of *CYP2R1* promoter decreased within 12 months of vitamin D supplementation in DNA extracted from serum of non-Hispanic white American post-menopausal women aged ≥55 years (Zhu et al., 2013), indicating an effect of vitamin D supplementation on *CYP2R1* promoter methylation. These data indicate that under low vitamin D serum levels, the promoter of the major 25-hydroxylase *CYP2R1* may become methylated, and that event appears to be reversible upon exposure to increased vitamin D.

#### **EPIGENETIC REGULATION OF THE CYP27B1**

CYP27B1 is an inner mitochondrial membrane P450 enzyme that converts 25-D3 to its active form 1,25-D3. It is mainly expressed in the proximal tubule of the kidneys, but it is also expressed in many vitamin D target tissues, albeit at lower levels (Hendrix et al., 2004). The *CYP27B1* gene harbors a CpG island. However, recent sequence updates (Ensembl 74, November 2013) shift the CpG island from the *CYP27B1* promoter region into the gene coding sequence (Flicek et al., 2014). This explains the differences between the location of the CpG island depicted in **Figure 2** and the location of the CpG island described in literature. For simplicity, statements on nVDRES and CpG island location below refer to reports in the published articles and not to **Figure 2**.

The promoter region of *CYP27B1* contains a negative VDRE (nVDRE) located at around 500 bp, consisting of two E-box like motifs (Murayama et al., 2004). This region is responsible for 1,25-D3-dependent transrepression, which seems to be achieved through recruitment of both HDACs and DNMTs by VDR/RXR to the promoter region of *CYP27B1* (Takeyama and Kato, 2011). For further details see subsection Interactions of Vitamin D with Chromatin Modulators and Remodelers.

In cancer, expression of CYP27B1 is often downregulated. This may be explained by increased methylation of the CpG island located within *CYP27B1*. In the breast cancer cells MDA-MB-231, *CYP27B1* hypermethylation led to gene silencing, which could be reversed by treatment with deoxyC (Shi et al., 2002). In prostate cancer cell lines, combination of the DNMT1 inhibitor DAC and the HDAC inhibitor TSA resulted in increased activity of CYP27B1 (Khorchide et al., 2005). In the choriocarcinoma cell lines BeWo and JAR the promoter of *CYP27B1* was densely methylated (Novakovic et al., 2009). The *CYP27B1* promoter was hypermethylated (61%) in Non-Hodgkin's lymphoma, but not in benign follicular hyperplasia. Two out of four non-Hodgkin's lymphoma cell lines showed strong methylation of the *CYP27B1* promoter. Interestingly, all four responded to DAC-TSA treatments with upregulation of gene expression independent of the methylation status of their *CYP27B1* promoter, which may be explained by other regions prone to methylation not investigated in this study or by differences in silencing mechanisms (Shi et al., 2007). Further, the methylation level of *CYP27B1* was increased in primary lymphoma and leukemia cells also compared with normal peripheral blood lymphocytes (Lagger et al., 2003; Wjst et al., 2010).

Methylation of *CYP27B1* in diseases might cause reduced local activation of 25-D3 to 1,25-D3, thus reducing local 1,25-D3 levels and restricting its functions.

### **EPIGENETIC REGULATION OF CYP24A1**

The 1,25-dihydroxyvitamin D3 24-hydroxylase is an inner mitochondrial membrane P450 enzyme, which catalyzes both 25-D3 and 1,25-D3 (Kawashima et al., 1981; Pedersen et al., 1983; Sakaki et al., 2000). Its primary site of expression are the kidneys, where it plays a crucial role in regulating systemic vitamin D metabolite levels, however, expression is found in many other vitamin D target tissues.

The promoter of *CYP24A1* is spanned by a CpG island making it prone to regulation by DNA methylation. Several responsive elements are located within this area, including two VDREs, a vitamin stimulating element (VSE), and SP1 binding sites (see **Figure 2**).

In healthy kidney, skeletal muscle, whole blood, brain, skin fibroblasts, and sperm the *CYP24A1* promoter is not methylated (Novakovic et al., 2009), although the expression levels are highly variable. In peripheral blood lymphocytes methylation of *CYP24A1* was low (5%) (Wjst et al., 2010). Interestingly, in full term human placenta 56.5% of the *CYP24A1* promoter is methylated. *CYP24A1* methylation was also observed in the placenta of the marmoset and mouse, however, at a lower level.

In the choriocarcinoma cell lines JEG-3, BeWo, and JAR the promoter of *CYP24A1* was densely methylated and the methylation level correlated inversely with the low gene expression (Novakovic et al., 2009). Treatment of osteoblastic ROS cells with 1,25-D3 did not induce CYP24A1 expression. Considering the strong methylation of the *CYP24A1* promoter region, epigenetic silencing of CYP24A1 may account for the unresponsiveness of this gene to 1,25-D3 (Ohyama et al., 2002). In the human prostate cancer cell line PC3, methylation of the *CYP24A1* promoter reduced reporter gene expression in a methylationdependent manner (Luo et al., 2010). In prostate cancer cells, the methylation status of *CYP24A1* promoter inversely correlated with gene expression. Demethylating agents restored CYP24A1 expression only in cell lines where the *CYP24A1* promoter was methylated prior to treatment (Luo et al., 2010). Only DNA demethylation by DAC treatment permitted recruitment of VDR to the *CYP24A1* promoter (Luo et al., 2010). In patients, development from benign toward malignant prostate lesions was paralleled by increasing methylation levels of the *CYP24A1* promoter (Luo et al., 2010). Prostate tumor derived endothelial cells (TDEC) expressed less CYP24A1 compared with endothelial cells derived from normal cells or matrigel plugs, which may be attributed to increased *CYP24A1* promoter methylation in TDECs (Johnson et al., 2010). We have shown recently that in colon cancer cell lines DAC induced CYP24A1 expression in a cell line-specific manner, independent of the methylation level of the promoter. In these cells induction of CYP24A1 expression by DAC seems to be independent of *CYP24A1* promoter methylation (Höbaus et al., 2013a). Moreover, the methylation level of the *CYP24A1* promoter was comparably low both in colon adenocarcinomas and the adjacent mucosa, although the expression of CYP24A1 was significantly higher in the tumors (Höbaus et al., 2013b).

Taken together, the regulation of CYP24A1 by DNA methylation appears to be tissue-dependent, both in health and disease.

## **CONCLUSIONS**

There is a strong reciprocity between the vitamin D system and epigenetic mechanisms. The vitamin D system is, on the one hand regulated by epigenetic mechanisms and, on the other hand, is involved in regulating epigenetic events. Critical vitamin D tool genes can be silenced by DNA methylation. The VDR protein interacts, directly or indirectly, with chromatin modifiers and remodelers. Liganded VDR regulates expression of several of these chromatin modifiers and remodelers, and it might even regulate DNA methylation.

Epigenetic regulation of gene expression is a fine-tuned mechanism and its deregulation can lead to pathological conditions. The impact of vitamin D in the maintenance of the normal epigenetic landscape underlines the central role of this hormone in physiology.

## **PERSPECTIVES**

One of the most fundamental questions in the control of gene expression is the way how epigenetic marks are established, erased, and recognized. Regulating epigenetic events could be a further mechanism by which 1,25-D3 may prevent or delay tumorigenesis and onset of chronic diseases. Therefore, we need to understand better the impact of vitamin D on the epigenome, and plan thorough and comprehensive studies to examine this interplay.

### **ACKNOWLEDGMENTS**

We thank the Austrian Research Fund (FWF) Grant P22200-B11 and the European Union, FP7-PEOPLE-2010 Grant 264663 for support.

## **REFERENCES**


DNA/chromatin binding motifs is specifically up-regulated in breast cancer. *J. Biol. Chem.* 274, 15633–15645. doi: 10.1074/jbc.274.22.15633


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

*Received: 09 February 2014; paper pending published: 07 March 2014; accepted: 09 April 2014; published online: 29 April 2014.*

*Citation: Fetahu IS, Höbaus J and Kállay E (2014) Vitamin D and the epigenome. Front. Physiol. 5:164. doi: 10.3389/fphys.2014.00164*

*This article was submitted to Integrative Physiology, a section of the journal Frontiers in Physiology.*

*Copyright © 2014 Fetahu, Höbaus and Kállay. 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.*

## Tumor suppression in skin and other tissues via cross-talk between vitamin D- and p53-signaling

#### *Jörg Reichrath1 \*, Sandra Reichrath1, Kristina Heyne2, Thomas Vogt <sup>1</sup> and Klaus Roemer <sup>2</sup>*

*<sup>1</sup> Department of Dermatology, The Saarland University Hospital, Homburg (Saar), Germany*

*<sup>2</sup> José Carreras Centre and Internal Medicine I, University of Saarland Medical Centre, Homburg (Saar), Germany*

#### *Edited by:*

*Carsten Carlberg, University of Eastern Finland, Finland*

#### *Reviewed by:*

*Alberto Muñoz, Consejo Superior de Investigaciones Científicas, Spain Moray J. Campbell, Roswell Park Cancer Institute, USA*

#### *\*Correspondence:*

*Jörg Reichrath, Department of Dermatology, The Saarland University Hospital, Kirrberger Str., 66421 Homburg, Germany e-mail: joerg.reichrath@uks.eu*

P53 and its family members have been implicated in the direct regulation of the vitamin D receptor (VDR). Vitamin D- and p53-signaling pathways have a significant impact on spontaneous or carcinogen-induced malignant transformation of cells, with VDR and p53 representing important tumor suppressors. VDR and the p53/p63/p73 proteins all function typically as receptors or sensors that turn into transcriptional regulators upon stimulus, with the main difference being that the nuclear VDR is activated as a transcription factor after binding its naturally occurring ligand 1,25-dihydroxyvitamin D with high affinity while the p53 family of transcription factors, mostly in the nucleoplasm, responds to a large number of alterations in cell homeostasis commonly referred to as stress. An increasing body of evidence now convincingly demonstrates a cross-talk between vitamin D- and p53-signaling that occurs at different levels, has genome-wide implications and that should be of high importance for many malignancies, including non-melanoma skin cancer. One interaction involves the ability of p53 to increase skin pigmentation via POMC derivatives including alpha-MSH and ACTH. Pigmentation protects the skin against UV-induced DNA damage and skin carcinogenesis, yet on the other hand reduces cutaneous synthesis of vitamin D. A second level of interaction may be through the ability of 1,25-dihydroxyvitamin D to increase the survival of skin cells after UV irradiation. UV irradiation-surviving cells show significant reductions in thymine dimers in the presence of 1,25-dihydroxyvitamin D that are associated with increased nuclear p53 protein expression, and significantly reduced NO products. A third level of interaction is documented by the ability of vitamin D compounds to regulate the expression of the *murine double minute 2* (*MDM2*) gene in dependence of the presence of wild-type p53. MDM2 has a well-established role as a key negative regulator of p53 activity. Finally, p53 and family members have been implicated in the direct regulation of VDR. This overview summarizes some of the implications of the cross-talk between vitamin D- and p53-signaling for carcinogenesis in the skin and other tissues.

**Keywords: vitamin D, vitamin D receptor, p53, MDM2, cancer**

## **SKIN, VDR AND THE VITAMIN D ENDOCRINE SYSTEM/REGULATORY NETWORK: AN INTRODUCTION**

The skin is the largest organ of the human body, consisting of several compartments that are named epidermis, dermis and subcutis. The epidermis contains a basal layer (stratum basale), that is composed of self-renewing cells (keratinocytes) with limited proliferative capacity (transient amplifying cells), of stem cells with high proliferative capacity that need to be preserved, and of outwardly migrating layers (stratum spinosum, stratum granulosum and stratum corneum) of mostly resting keratinocytes at different stages of differentiation. The skin is one of the key tissues of the human body's vitamin D regulatory network (VDRN) (Lehmann et al., 2004; Holick, 2007; Reichrath and Reichrath, 2012; Mason and Reichrath, 2013). First, vitamin D is synthesized in the skin (**Figure 1**) by the action of solar or artificial ultraviolet B (UVB) radiation (under most living conditions, only a small amount of vitamin D is taken up by the diet) (Lehmann et al., 2004; Holick, 2007). Second, the skin represents an important target tissue for 1,25-dihydroxyvitamin D, the biologically active natural vitamin D metabolite, that is formed from vitamin D by consecutive hydroxylations at position 25 in the liver (mediated by CYP2R1 and by CYP27A1, resulting in 25-hydroxyvitamin D) and at position 1 in the kidneys and in many other tissues (mediated by CYP27B1) (Lehmann et al., 2004; Holick, 2007; Reichrath and Reichrath, 2012; Mason and Reichrath, 2013). 1,25-Dihydroxyvitamin D represents a potent seco-steroid hormone that regulates, via various independent mechanisms growth, many non-malignant and malignant cell types, including human keratinocytes (Lehmann et al., 2004; Holick, 2007; Haussler et al., 2012; Reichrath and Reichrath, 2012; Mason and Reichrath, 2013). It exerts its effects through the binding with high affinity to a corresponding receptor (VDR) that is located intranuclear in target tissues (Lehmann et al., 2004; Holick, 2007; Haussler et al., 2012; Reichrath and Reichrath,

2012; Mason and Reichrath, 2013). VDR is a member of a superfamily named trans-acting transcriptional regulatory factors, that also contains the retinoic acid receptors (RARs) and the retinoid-X receptors (RXRs), as well as the thyroid and steroid hormone receptors (Lehmann et al., 2004; Holick, 2007; Haussler et al., 2012; Reichrath and Reichrath, 2012; Mason and Reichrath, 2013). The farnesoid-X receptor (FXR) that controls bile acid metabolism and the pregnane-X receptor (PXR) which regulates xenobiotic detoxification are evolutionarily most closely related to the VDR (Haussler et al., 2012). Binding of its ligand 1,25 dihydroxyvitamin D induces conformational changes of the VDR that lead to heterodimerization with RXR and to zinc fingermediated binding to vitamin D response elements (VDREs) that are located in regulatory regions of target genes (Haussler et al., 2012). As a result, vitamin D activity in a particular cell largely depends upon sufficient expression of VDR and RXR proteins, the autocrine/paracrine production or the endocrine delivery of adequate amounts of the 1,25-dihydroxyvitamin D ligand, and of cell-specific programming of gene transcription to regulate expression of distinctive genes that encode proteins that finally exert the vitamin D effect (Haussler et al., 2012). cDNA microarray analyses of mRNAs and other investigations suggest that as many as 500–1000 coding genes may be regulated by the VDR, which may contact up to ∼8000 loci in the human genome (Haussler et al., 2012). 1,25-Dihydroxyvitamin D-mediated transcriptional regulation of many genes involved in cellular growth and differentiation has been demonstrated, including the genes for ß3-integrin, fibronectin, and cell cycle regulatory proteins such as p21/WAF-1 (CDKN1A) (Lehmann et al., 2004; Holick, 2007; Haussler et al., 2012). Like most other skin cells, keratinocytes express VDR (Lehmann et al., 2004; Holick, 2007); in these cells, 1,25-dihydroxyvitamin D, blocks proliferation and promotes differentiation *in vitro* (Lehmann et al., 2004; Holick, 2007; Haussler et al., 2012). Interestingly, it has been reported that the combination of 1,25-dihydroxyvitamin D and the retinoic acid metabolite isotretinoin is efficient in the therapy of precancerous skin lesions and of non-melanoma skin cancer (cutaneous squamous and basal cell carcinomas) (Tang et al., 2012a,b; Mason and Reichrath, 2013). Moreover, it has been demonstrated that VDR ablation promotes chemically induced skin carcinogenesis (Tang et al., 2012a,b; Mason and Reichrath, 2013).

VDR-signaling comprises much more than just ligand/ receptor triggering of gene expression. Distinct and fine-tuned responses indicate a complex regulation of this signaling pathway. Moreover, chemical and other modifications of the VDR signaling pathway govern such important parameters as intracellular trafficking, duration of interaction between the receptor and cofactors, the receptor and ligand, as well as turnover and stability of other relevant proteins (Haussler et al., 2012). Not least, regulation of VDR target genes is controlled by stability and turnover of relevant microRNAs and RNAs (Haussler et al., 2012).

Depending on cell type and context, both VDR- and p53-signaling regulate many cellular functions that are of relevance for cancer development, including proliferation, differentiation, apoptosis and cell survival (Murray-Zmijewski et al., 2006; Holick, 2007; McKeon and Melino, 2007; Vousden and Lane, 2007; Vousden and Prives, 2009; Haussler et al., 2012; Mason and Reichrath, 2013). Consequently, vitamin D- and p53-signaling pathways have a significant impact on spontaneous or carcinogen-induced malignant transformation of cells, with vitamin D receptor (VDR) and p53 representing important tumor suppressors (Murray-Zmijewski et al., 2006; Holick, 2007; McKeon and Melino, 2007; Vousden and Lane, 2007; Vousden and Prives, 2009; Haussler et al., 2012; Mason and Reichrath, 2013). Mutations in genes encoding for proteins of the p53 pathway represent a hallmark of many if not all types of cancer (Vousden and Lane, 2007; Vousden and Prives, 2009). Low serum 25(OH)D concentrations and distinct polymorphisms (SNPs) in the *VDR* gene and other vitamin D-related genes, on the other hand, are associated with an increased incidence and an unfavorable outcome of various malignancies (Mason and Reichrath, 2013). The VDR and the p53 family all function typically as activatable transcriptional regulators, with the main difference being that VDR is activated after binding its naturally occurring ligand 1,25-dihydroxyvitamin D (1,25(OH)2D or calcitriol) with high affinity (Haussler et al., 2012) while p53, mostly in the nucleoplasm, responds to a large and still growing number of alterations in cell homeostasis (Murray-Zmijewski et al., 2006; McKeon and Melino, 2007; Vousden and Lane, 2007; Vousden and Prives, 2009). In any event is the result of such activation manifested by conformational changes and heterodimerization with retinoid X receptor (RXR) of VDR and by chemical modifications and oligomerization of the p53 family—the direct contact with regulatory DNA. In both pathways the cell type- and contextdependent recruitment of nuclear co-regulators entails the stimulation or repression of a very large number, typically hundreds, of genes (Lin et al., 2005; Holick, 2007; Perez and Pietenpol, 2007; Sbisa et al., 2007; Riley et al., 2008; Haussler et al., 2012). Several of these code themselves for transcriptional regulators, adding a further level of complexity to the networks. It is obvious that transcription factor pathways may cross-talk, for instance, through the sharing of target genes or co-regulators, and through the engagement in interdependent regulatory loops. Indeed, all of these mechanisms, plus several others, seem to have been realized in the cross-talk of VDR and the p53 family (**Table 1**).

Intriguingly, both pathways are critically involved in cellular processes that are important for carcinogenesis such as cell differentiation/proliferation, in the regulation of stem cell maintenance, and in cell homeostasis. While VDR controls proliferation/differentiation of many cell types (Holick, 2007; Haussler et al., 2012), some members and isoforms of the p53 family, and in particular p53 itself, reduce the stem cell potential and stimulate differentiation (Lin et al., 2005). Interestingly, on the side of the p53 family, all three members (p53/p63/p73) can be expressed as truncated isoforms capable of counteracting their siblings' transactivating effects (Murray-Zmijewski et al., 2006). Not too surprising, VDR and p53 have been linked to many malignancies, including non-melanoma skin cancer (Mason and Reichrath, 2013). The present review aims at providing an overview on this interesting signaling network, with a focus on non-melanoma skin cancer. Future genome-wide analyses of the target genes will shed further light on the interaction of these pleiotropic regulators. Before the cross-talk is discussed, the p53 pathway shall be briefly outlined.

## **THE p53 FAMILY OF TRANSCRIPTIONAL REGULATORS**

p53, p63, and p73 (the p53 family hereafter) are homotetrameric transcriptional regulators that bind to very closely related DNA motifs, consisting of two consecutive 10-mers (half-sites),


#### **Table 1 | Overview of the cross talk between vitamin D- and p53 signaling.**

*1,25-D, 1,25-dihydroxyvitamin D; MDM2, murine double minute 2.*

preferentially spaced by no more than zero to 2 base pairs, with the consensus r,r,r,C,A/T,T,G,y,c,y (p53); r,r,r,C,G,T,G,y,y,y; t/a,a/t,a,C,A/T,T,G,t,t/a,t or r,r,r,C,A/G,T/A,G,y,y,y (p63), and a/c/g,g/a,g,C,A,T,G,c/t,c,c/t (p73; r = purines; y = pyrimidines) (Osada et al., 2005; Riley et al., 2008; Brandt et al., 2009; Roemer, 2012). They share a large number of target sequences, as expected given the high degree of homology within the DNA binding domains, among the consensus sequence motifs, and the degeneracy of the individual binding sites (Roemer, 2012). It is therefore perhaps no surprise that the regulation of a defined sequence by any of these transcription factors is controlled at several levels including posttranslational modifications and protein/protein interactions. Many of these are specific for each family paralog. Moreover, the binding of p53, p63, and p73 to DNA is affected by additional parameters such as the number of the half-sites, their orientation, their position relative to the target gene, and their overlap with binding sites for other transcription factors. Finally, differential recruitment of co-activator/co-repressor complexes to promoters has been documented. These may be coded, for example, by specific spacings between the 10-mers of the DNA binding motifs (Riley et al., 2008). Epigenetic CpG methylation does not seem to affect the binding to DNA significantly (Brandt et al., 2009; Roemer, 2012).

The p53 family proteins display a modular organization that is quite different from that of the VDR (see above). Typically, an N-terminal transactivation domain (TD), a central DNA binding domain (DBD) and C-terminal regulatory and protein/protein interaction domain is present. The DBDs are the most highly conserved regions among the paralogs, sharing ∼60% homology (Murray-Zmijewski et al., 2006; Roemer, 2012). In addition to the full-length variants, a large number of isoforms exists, owing to transcription initiation from internal promoters, alternative splicing and the use of alternative translation initiation sites; however, in most cases the DBD is maintained. More than 10 different isoforms of p53, more than six of p63 and at least 29 of p73 are currently known (Murray-Zmijewski et al., 2006; Hollstein and Hainaut, 2010; Roemer, 2012). In most cases, their biological functions are not fully understood. Furthermore, an arsenal of posttranslational modifications that are in part interdependent has evolved. These include phosphorylations, acetylations, ubiquitinations, sumoylations, neddylations, methylations, glycosylations, and oxidation/reduction, and they control the proteins' abundance, DNA binding, level of activity as transcription factor, cross-talk with other proteins and subcellular localization (Murray-Zmijewski et al., 2006; Toledo and Wahl, 2006; Kruse and Gu, 2009). All these levels of regulation are best studied in p53 and have revealed an enormous degree of complexity (Vousden and Prives, 2009; Roemer, 2012) which may be exemplified by the chemical modification "code" that seems to regulate p53 function in a tissue-specific manner through the sequential build-up of poly-phosphorylation patterns at different sites and that may even be accompanied by other chemical changes such as acetylations (Gu and Roeder, 1997; Ashcroft et al., 2000; Wang et al., 2004; Roemer, 2012). At the level of the cell, p53 is involved in the regulation of the cell cycle (Wang and El-Deiry, 2006), cell survival and autophagy, DNA repair, respiration, oxidative stress protection, glucose metabolism, cell adhesion/motility, the cytoskeleton and endo/exosome compartments, and of angiogenesis. At the organismal level, p53 is involved in tumor suppression and maintenance of genome stability, and the control of stem cell compartments, female fertility and ageing (Riley et al., 2008; Roemer, 2012).

The complexity of the regulation of the p53 family is further highlighted by the antagonistic partnership between p53 and its central negative regulators, the E3 ubiquitin ligases murine double minute 2 (MDM2) and MDM4 (Roemer, 2012). Activation of p53 is almost always involving inhibition of MDM2/4. For example, acetylation of p53 and MDM2 overcomes the inhibitory ubiquitination of p53 by MDM2 through the blocking of MDM2 enzymatic function, the dissociation of the p53/MDM2 complex and thereby, the stimulation of p53's interaction with DNA as well as the recruitment of co-activators (Gu and Roeder, 1997; Ashcroft et al., 2000; Wang et al., 2004; Roemer, 2012). MDM4 is not functioning as a ubiquitin ligase for p53 but can inhibit p53's transcriptional activity and modulate the p53/MDM2 interaction (Toledo and Wahl, 2006; Roemer, 2012). Since p53 can transactivate the *MDM2* gene, a negative feedback loop is formed (Toledo and Wahl, 2006; Kruse and Gu, 2009; Roemer, 2012). Such a loop is also established with p63 and p73; however, MDM2 inhibits these transcription factors at promoters yet in contrast to p53 cannot ubiquitin-mark them for degradation (Murray-Zmijewski et al., 2006; Roemer, 2012).

Cell context determines the respective function of individual p53 family members. In the absence of extra stress, i.e., under physiological background stress induced, for instance, by reactive oxygen species (ROS) as a by-product of respiration, p53, p63, and p73 primarily control cell fate, differentiation and development. Intriguingly, these functions seem to be predominantly mediated by the DNA binding competent yet transactivation impaired delta-N isoforms of the proteins (*-*Np63, *-*Np73). In cells or tissue that have been challenged by further stresses, as for example by overt ROS production, radiation, hypoxia, hypo/hyperthermia, metabolite shortages and imbalances, oncogene dysregulation, and virus/bacterial/parasite infections, the p53 family members, and in particular p53 itself, seem to mainly control repair, proliferative capacity and survival. Central to these functions are the transactivation-proficient isoforms (p53, TAp63, TAp73). Since many of the damaging stresses can support cell transformation, the p53 family, and here again, mostly p53 itself, thus act as tumor suppressors by inducing cell cycle arrest, temporary or permanent senescence, apoptosis, and differentiation (Vousden and Lane, 2007; Levine and Oren, 2009; Vousden and Prives, 2009; Roemer, 2012). Conversely, lack of proper function of p53 or p73, or overproduction of dominant-negative *-*Np63, support tumor formation in animals and humans. Along the same line, tumor-inducing viruses encode proteins that target p53, and perhaps there is no tumor in which the p53 pathway itself plus all ascending/descending pathways are fully intact (Gatza et al., 2007; Vousden and Lane, 2007; Feng et al., 2008; Hu et al., 2008; Roemer, 2012).

Like p53, p63, and p73 can act as tumor suppressors, although this does not seem to be their primary functions (Murray-Zmijewski et al., 2006; Roemer, 2012). For example, p63 and p73 are not as frequently mutated in human cancers as is p53. Rather, p63 is often overproduced in tumors (Park et al., 2000), which seems to contradict its function as a tumor suppressor, yet as mentioned above, this is often due to p63 isoforms that lack the transactivation domain but not their ability to bind to DNA and that thereby may act dominant-negatively (Candi et al., 2007; Roemer, 2012). In contrast and as expected from a tumor suppressor, transactivation competent p63 (TAp63) can sensitize cells to apoptosis in response to DNA damaging stress (Gressner et al., 2005; Roemer, 2012). Moreover, some p63± mice are tumor-prone, and the resulting tumors often display loss of the remaining wild-type allele (Flores, 2007; Roemer, 2012). Mice with a specific deficiency for TAp73 show genomic instability and a higher tumor incidence (Tomasini et al., 2008; Roemer, 2012). Furthermore, p63 and p73 seem to have p53-independent roles in DNA repair (Talos et al., 2007; Lin et al., 2009; Roemer, 2012).

p63 and p73, but not p53, are crucial for embryonic development in all organisms examined so far (Danilova et al., 2008; Roemer, 2012). Although p53-deficiency interferes with mesoderm/endoderm fate determination in the frog *Xenopus* (Wallingford et al., 1997; Roemer, 2012), this condition fails to generate significant early phenotypes in mice or humans (Choi and Donehower, 1999; Varley, 2003; Roemer, 2012). However, at a more subtle level, and since p53 can induce stem cell differentiation, lack of p53 function may cause unrestrained stem cell proliferation (Gil-Perotin et al., 2006; Dumble et al., 2007; Roemer, 2012). Other more subtle functions of p53 are in mitochondrial respiration and glucose metabolism (Matoba et al., 2006; Roemer, 2012). Overactivity of p53, by contrast, does indeed entail immediate and dramatic consequences in the development of the early mouse embryo—its apoptotic loss and one of the most striking functions of the p53 inhibitors MDM2 and MDM4 during embryonic development is the prevention of this consequence (Marine et al., 2006; Roemer, 2012). Later in embryonal development, for example during neurogenesis, the DNA binding-proficient yet transactivation-incompetent dominant-negative isoform of p73, *-*Np73, may serve as a p53 and p63 restraining factor to inhibit apoptosis (Jacobs et al., 2004; Roemer, 2012). p53 and possibly p63, but most importantly p73, help shape the nervous system during life, perhaps primarily by controlling apoptosis (Jacobs et al., 2005; Miller and Kaplan, 2007; Roemer, 2012).

p63 function during development is critical for epithelial stem cell maintenance (Yi et al., 2008; Roemer, 2012), squamous epithelial differentiation and skin renewal (Truong et al., 2006; Koster et al., 2007; Mikkola, 2007; Roemer, 2012). The *-*Np63 isoform acts mainly through controlling the expansion of epithelial layers while TAp63 seems to support differentiation, and it functions as the guardian of the female germ line by inducing apoptosis in damaged resting oocytes (Suh et al., 2006; Roemer, 2012). p73 deficiency in mice results in neuronal and olfactory dysfunctions as well as in chronic infection and inflammation (Murray-Zmijewski et al., 2006; Roemer, 2012). Collectively, the stem cell/differentiated cell decision seem to be regulated in part by the balance between the *-*Np63/TAp63 antagonists in the skin and - in an analogous manner - by the balance between the *-*Np73/TAp73 antagonists in the developing nervous and immune systems. Thus, p73 may be to neuronal development and homeostasis what p63 is to the development and homeostasis of the skin (De Laurenzi et al., 2000; Jacobs et al., 2004; Roemer, 2012). Since the pleiotropic VDR is important for tumor suppression as well as skin development and differentiation, it is perhaps no surprise that both transcriptional regulator pathways talk to each other.

## **CROSS-TALK BETWEEN THE VDR AND THE p53 FAMILY IN CANCER**

An increasing body of evidence points to a cross-talk between vitamin D- and p53-signaling occuring at different levels that might be of great importance for many malignancies, including non-melanoma skin cancer (**Table 1**). Both p53 and VDR act as tumor suppressors in several tissues, including skin. Much of the tumor suppressor function in the skin may be mediated through the interaction of the VDR and p53 pathways—either by mutual activation or inhibition. What is known about this interaction, in particular in non-melanoma skin cancer? DNA damage induced by solar or artificial ultraviolet (UV) radiation represents the most important environmental risk factor for carcinogenesis of cutaneous squamous cell carcinoma (SCC) (Reichrath and Reichrath, 2012; Mason and Reichrath, 2013). The predominant types of DNA damage which are directly induced by UV are promutagenic pyrimidine dimers (Wikonkal and Brash, 1999; Reichrath and Reichrath, 2012; Mason and Reichrath, 2013). Thymine-thymine dimers, which represent cys-syn cyclobutane pyrimidine dimers (CPDs), are the major form of pyrimidine dimers that are identified in human skin following UV-B exposure. In contrast, other types of DNA damage, including cytosine-cytosine, thyminecytosine bipyrimidines, and 6–4 photoproducts are less frequently detected (Douki et al., 2000; Cooke et al., 2003; Courdavault et al., 2004a; Mouret et al., 2006; Reichrath and Reichrath, 2012; Mason and Reichrath, 2013). CPDs are caused via disruption of the 5–6 double bonds in two adjacent pyrimidine bases, thereby inducing atypical covalent binding which connects the 2 bases by a stable ring configuration, resulting in a bipyrimidine (Ravanat et al., 2001; Pattison and Davies, 2006; Reichrath and Reichrath, 2012; Mason and Reichrath, 2013). In general, CPDs are induced by UV-B (290–320 nm) (Reichrath and Reichrath, 2012; Mason and Reichrath, 2013), although the production of thymine dimer by UV-A (320–400 nm) wavelengths below 330 nm has also been reported (Applegate et al., 1999; Jiang et al., 1999; Rochette et al., 2003; Courdavault et al., 2004b; Mouret et al., 2006; Reichrath and Reichrath, 2012; Mason and Reichrath, 2013). UV-radiation induces gene mutations which may result in photocarcinogenesis (Hart et al., 1977; Sutherland et al., 1985; Brash et al., 1991; Agar et al., 2004; Besaratinia et al., 2008; Reichrath and Reichrath, 2012; Mason and Reichrath, 2013). Moreover, it has been shown that DNA damage initiates and promotes cellular mechanisms which block the detection and elimination of transformed cells by immune surveillance (Applegate et al., 1989; Kripke et al., 1992; Reichrath and Reichrath, 2012; Mason and Reichrath, 2013). UV radiation induces different forms of DNA lesions which are generated either photochemically and directly or indirectly by UV activation of several photoreceptors which have the capacity to modulate the cellular redox equilibrium, thereby generating reactive oxygen species (ROS) (Reichrath and Reichrath, 2012; Mason and Reichrath, 2013). ROS induced cellular damage then leads both to oxidative DNA damage, and to lipid peroxidation (Reichrath and Reichrath, 2012; Mason and Reichrath, 2013). Additionally, UV-induced increased levels of nitric oxide synthase (Deliconstantinos et al., 1995; Bruch-Gerharz et al., 1998; Cals-Grierson and Ormerod, 2004; Reichrath and Reichrath, 2012; Mason and Reichrath, 2013) cause excess levels of nitric oxide (NO) (Paunel et al., 2005; Mowbray et al., 2009; Reichrath and Reichrath, 2012; Mason and Reichrath, 2013). It has been shown that these pathophysiologically increased concentrations of NO and ROS combine to generate genotoxic NO products, including peroxynitrite, which modify the bases and the sugar-phosphate scaffold of DNA via nitrosative and oxidative damage (Reichrath and Reichrath, 2012; Mason and Reichrath, 2013) Both UV- and ROS-induced damages activate p53.

UV-irradiation induces p53 to stimulate skin pigmentation via POMC derivatives including alpha-MSH and ACTH (Yamaguchi and Hearing, 2009), thereby protecting the skin against further UV-induced DNA damage and skin carcinogenesis. However, this reduces cutaneous synthesis of vitamin D. This may be important because, on a second level, 1,25-dihydroxyvitamin D can increase the survival of UV-irradiated keratinocytes and prevent further accumulation of DNA damage in these surviving skin cells (Gupta et al., 2007). Following UVR, the survival of 1,25 dihydroxyvitamin D-treated skin cells was significantly higher as compared to vehicle-treated cells (*P <* 0*.*01) (Gupta et al., 2007). In that study, UVR-surviving and 1,25-dihydroxyvitamin D-treated keratinocytes showed significantly reduced levels of thymine dimers (TDs) as compared to vehicle-treated cells (*P <* 0*.*001) (Gupta et al., 2007). Following UVR, nuclear p53 protein levels were elevated and, notably, became elevated to significantly higher levels in the presence of 1,25-dihydroxyvitamin D (*P <* 0*.*01). In contrast, NO derivatives were significantly decreased in 1,25-dihydroxyvitamin D-treated keratinocytes (*P <* 0*.*05) (Gupta et al., 2007). Both the elevated levels of nuclear p53 protein and the decreased production of nitric oxide products were suggested to be responsible at least in part for the decrease in TDs seen with 1,25-dihydroxyvitamin D-treatment after UVR (Gupta et al., 2007). In addition, a reduction in the number of TDs (*P <* 0*.*05) and in sunburn cells (*P <* 0*.*01) were demonstrated in skin sections from Skh:hr1 mice that had been treated with 1,25-dihydroxyvitamin D, at 24 h after UVR (Gupta et al., 2007). It was concluded that the vitamin D system in skin, in combination with p53, may represent an intrinsic mechanism that protects against UV damage (Gupta et al., 2007).

As a further molecular level of interaction it has been demonstrated that vitamin D compounds can regulate the expression of the MDM2 gene in dependence of p53 (Chen et al., 2013). As outlined above, *MDM2* represents a p53-inducible gene that encodes an E3 ubiquitin ligase mainly responsible for the degradation of p53 by the 26S proteasome (Roemer, 2012). A major function of MDM2 is its role as a key negative feed-back regulator of p53 activity (Bond et al., 2004), p53 activates *MDM2* expression via binding to corresponding p53 response elements (p53REs) in the P2 promoter of the *MDM2* gene. The increase of MDM2 protein then leads to its binding to p53 (primarily at the N-terminal 1–52 residues), which causes degradation of p53 or inhibition of p53 activity as a transcription factor (Chen et al., 1993). Vitamin D may thus prevent a lasting and overt p53 response in the face of damage and may thereby protect reparable cells from p53-induced apoptosis. However, MDM2 also exerts many p53-independent functions, and interacts with a broad variety of other proteins (including insulin like growth factor receptor, androgen receptor, estrogen receptor, Numb, RB, p300, etc.) that are of importance for various cellular functions including proliferation/differentiation, cell fate determination, and signaling (Ries et al., 2000; Ganguli and Wasylyk, 2003; Steinman et al., 2004; Zhang and Zhang, 2005; Lengner et al., 2006; Araki et al., 2010). VDR may activate the *MDM2* gene directly, through a VDR-response element in the promoter P2 of the *MDM2* gene (Barak et al., 1994; Zauberman et al., 1995; Roemer, 2012; Chen et al., 2013) However, even with this direct binding to *MDM2* sequences, p53 seems to be required for the induction of *MDM2* expression by VDR (Chen et al., 2013). Perhaps this is reflecting a cross-talk between the VDR and p53 bound to DNA since the p53 response element in the *MDM2* gene is also located in promoter P2.

VDR and p53 family members act first and foremost as transcription factors, and accordingly, much of the highly complex cross-regulation between them seems to happen at this level. For example, members of the p53 family including *-*Np63 can modulate VDR signaling through competitive binding to various VDR target genes including *p21Waf1/Cip (CDKN1A)*. Multiple VDREs have recently been identified in the promoter region of the *CDKN1A* gene, which is a transcriptional target of p53 and encodes a powerful blocker of the cell cycle in G1 and G2 phases (Saramaki et al., 2006). Notably, like with the *MDM2* gene, the VDR and p53 binding sites are in close proximity in the *CDKN1A* promoter (Saramaki et al., 2006). A much more detailed and unbiased (pathway-independent) genome-wide analysis of the VDR:p53 family interactions bound to DNA is in need. To this end, it shall be useful to employ chromatin immunoprecipitations (ChIPs) with either ChIPping with p53-antibodies first and re-ChIPping with VDR antibodies, or *vice versa*. Also, knock-in studies in which VDR response elements or p53 response elements in close proximity are deleted, should provide a deeper insight into the cooperativity or antagonism between these important tumor-suppressing transcription factors.

In the skin, p53/p63 play an important regulatory role in the maintenance of the stem cells as well as in the establishment of the differentiation gradient. In the undifferentiated proliferating basal layer of the skin, the dominant negatively acting, because DNA binding but transactivation impaired, *-*Np63 rules. Most effects exerted by the transactivation-competent p53 family members are inhibited by it (Yang et al., 1998; Lee and Kimelman, 2002; Roemer, 2012) In addition, *-*Np63 may inhibit differentiation by the blunting of VDR signaling through binding to various VDR target genes including *CDKN1A* (Pellegrini et al., 2001; Westfall et al., 2003; Roemer, 2012) TAp63 that is minor to *-*Np63 in this proliferating compartment of the skin, may become more dominant as *-*Np63 levels decrease in the course of differentiation (Nylander et al., 2002; Roemer, 2012).

Finally, p53 family members may regulate VDR directly (Maruyama et al., 2006; Kommagani et al., 2007). In a comparative genomics investigation in the human and mouse genome designed to locate conserved p53 binding sites, the VDR and 31 other genes were newly described as putative p53 targets. Reverse transcription-PCR and real-time PCR confirmed the responsiveness of these genes to p53 in human cancer cell lines (Maruyama et al., 2006). It was shown that VDR is upregulated by p53 and some other members of the p53 family. For example, an isoform of p63 (p63 gamma) can specifically upregulate VDR by directly associating with the VDR promoter *in vivo* (Kommagani et al., 2007). Moreover, ChIP analysis demonstrated that wild-type p53 protein binds to a conserved intronic site of the VDR gene (Maruyama et al., 2006). Conversely, transfection of VDR into cells resulted in upregulation of several p53 target genes and in growth suppression of colorectal cancer cells. In addition and as discussed above, p53 stimulated several VDR target genes in a 1,25-dihydroxyvitamin D-dependent manner, that is, in cooperation with VDR (Maruyama et al., 2006). Future, whole transcriptome-including studies will identify new transcripts that are initiated by VDR and p53 in concert.

An increasing body of evidence highlights the relevance of the cross-talk between VDR- and p53-signaling under various physiological and pathophysiological conditions. One study identified the VDRE as overrepresented in promoter sequences bound by mutated p53 (mutp53), and showed that mutp53 can interact functionally and physically with VDR (Stambolsky et al., 2010). In that investigation, mutp53 was recruited to VDR target genes and modulated their expression (increasing transactivation or relieving repression) (Stambolsky et al., 2010). Moreover, mutp53 promoted the nuclear accumulation of VDR and converted 1,25 dihydroxyvitamin D into an anti-apoptotic agent (Stambolsky et al., 2010).

Several investigations analyzed the cross-talk between VDRand p53-signaling in bone. It was demonstrated that hepatocyte growth factor (HGF) and 1,25-dihydroxyvitamin D act together to induce osteogenic differentiation of human bone marrow stem cells (hMSC) potentially through elevating p53 (Chen et al., 2012). The authors of this study hypothesized that the combination of HGF and 1,25-dihydroxyvitamin D can promote hMSC differentiation by up-regulation of 1,25-dihydroxyvitamin D and/or VDR expression to booster cell response(s) to 1,25 dihydroxyvitamin D. In line with this hypothesis, it was shown that HGF up-regulated gene expression of VDR and p63 and that p63 gene knockdown by siRNA eliminated the effects of HGF on VDR gene expression (Chen et al., 2012). Moreover, recent findings suggest that the cross-talk of VDR and p53 may directly target the human osteocalcin gene and positively affect osteocalcin gene expression. It was reported that osteocalcin promoter activity can be up-regulated both by exogenous and endogenous p53 and downregulated by p53-specific siRNA (Chen et al., 2011). It was shown that p53 binds to the human osteocalcin promoter *in vitro* and a p53 response element within the osteocalcin promoter region was identified (Chen et al., 2011). In this investigation, an additive effect of p53 and VDR on the regulation of osteocalcin promoter activity was observed. Another study demonstrated that p73 acts as an upstream regulator of 1,25-dihydroxyvitamin D-induced osteoblastic differentiation (Kommagani et al., 2010). In that investigation, silencing p73 significantly decreased 1,25-dihydroxyvitamin Dmediated osteoblastic differentiation; although p73 induced by DNA-damage increased 1,25-dihydroxyvitamin D-mediated differentiation of osteosarcoma cells (Kommagani et al., 2010).

### **CONCLUSIONS AND PERSPECTIVES**

VDR and the members of the p53 family are activatable transcriptional regulators that are at the hub of a common molecular network to control cell homeostasis, proliferation, differentiation and survival, and that way, act as classical tumor suppressors. Malfunction of either entails elevated susceptibility to transformation. A tissue archetypical of this interaction is the skin. Here, VDR as well as p53 and p63 control differentiation and the maintenance of the stem cell compartment. Accordingly, damage to skin cells such as induced by UV irradiation, or transformation of skin cells, typically come with characteristic responses of these proteins in the form of specific gene expression profiles to control differentiation, proliferation and survival. Since both classes of nuclear transcription factors act primarily through the regulation of genes, it is thus no great surprise to find functional interaction at several levels. Future, ChIP- and transcriptome analysis-based genome-wide studies of the DNA sequences that are targeted individually or by both factors together should provide us with new insights into this fascinating network.

## **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: 28 February 2014; accepted: 10 April 2014; published online: 03 June 2014. Citation: Reichrath J, Reichrath S, Heyne K, Vogt T and Roemer K (2014) Tumor suppression in skin and other tissues via cross-talk between vitamin D- and p53-signaling. Front. Physiol. 5:166. doi: 10.3389/fphys.2014.00166*

*This article was submitted to Integrative Physiology, a section of the journal Frontiers in Physiology.*

*Copyright © 2014 Reichrath, Reichrath, Heyne, Vogt and Roemer. 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.*

## Vitamin D: a critical and essential micronutrient for human health

## *Igor Bendik\*, Angelika Friedel , Franz F. Roos , Peter Weber and Manfred Eggersdorfer*

*Human Nutrition and Health (IB, AF, FFR), and Nutrition Science & Advocacy (PW, ME), DSM Nutritional Products Ltd., Basel, Switzerland*

#### *Edited by:*

*Carsten Carlberg, University of Eastern Finland, Finland*

#### *Reviewed by:*

*Katie Marie Dixon, The University of Sydney, Australia Miguel A. Maestro, University of A Coruña, Spain*

#### *\*Correspondence:*

*Igor Bendik, Department of Human Nutrition and Health, DSM Nutritional Products Ltd., Bldg. 201/219A, PO Box 2676, CH-4002 Kaiseraugst, Basel, Switzerland e-mail: igor.bendik@dsm.com*

Vitamin D is a micronutrient that is needed for optimal health throughout the whole life. Vitamin D3 (cholecalciferol) can be either synthesized in the human skin upon exposure to the UV light of the sun, or it is obtained from the diet. If the photoconversion in the skin due to reduced sun exposure (e.g., in wintertime) is insufficient, intake of adequate vitamin D from the diet is essential to health. Severe vitamin D deficiency can lead to a multitude of avoidable illnesses; among them are well-known bone diseases like osteoporosis, a number of autoimmune diseases, many different cancers, and some cardiovascular diseases like hypertension are being discussed. Vitamin D is found naturally in only very few foods. Foods containing vitamin D include some fatty fish, fish liver oils, and eggs from hens that have been fed vitamin D and some fortified foods in countries with respective regulations. Based on geographic location or food availability adequate vitamin D intake might not be sufficient on a global scale. The International Osteoporosis Foundation (IOF) has collected the 25-hydroxy-vitamin D plasma levels in populations of different countries using published data and developed a global vitamin D map. This map illustrates the parts of the world, where vitamin D did not reach adequate 25-hydroxyvitamin D plasma levels: 6.7% of the papers report 25-hydroxyvitamin D plasma levels below 25 nmol/L, which indicates vitamin D deficiency, 37.3% are below 50 nmol/Land only 11.9% found 25-hydroxyvitamin D plasma levels above 75 nmol/L target as suggested by vitamin D experts. The vitamin D map is adding further evidence to the vitamin D insufficiency pandemic debate, which is also an issue in the developed world. Besides malnutrition, a condition where the diet does not match to provide the adequate levels of nutrients including micronutrients for growth and maintenance, we obviously have a situation where enough nutrients were consumed, but lacked to reach sufficient vitamin D micronutrient levels. The latter situation is known as hidden hunger. The inadequate vitamin D status impacts on health care costs, which in turn could result in significant savings, if corrected. Since little is known about the effects on the molecular level that accompany the pandemic like epigenetic imprinting, the insufficiency-triggered gene regulations or the genetic background influence on the body to maintain metabolic resilience, future research will be needed. The nutrition community is highly interested in the molecular mechanism that underlies the vitamin D insufficiency caused effect. In recent years, novel large scale technologies have become available that allow the simultaneous acquisition of transcriptome, epigenome, proteome, or metabolome data in cells of organs. These important methods are now used for nutritional approaches summarized in emerging scientific fields of nutrigenomics, nutrigenetics, or nutriepigenetics. It is believed that with the help of these novel concepts further understanding can be generated to develop future sustainable nutrition solutions to safeguard nutrition security.

**Keywords: vitamin D, 25-hydroxyvitamin D, nutrition, micronutrients, hidden hunger, nutrition security, nutritional pathways, nutrigenomics**

## **INTRODUCTION**

Vitamin D is needed to maintain calcium concentrations within a narrow physiological range. This function is vital as the calcium ion is essential for a large variety of cellular and metabolic processes in the body (Berridge, 2012). To secure the calcium supplies besides intestinal absorption, calcium is stored in the skeleton and acts as a large calcium reservoir that is mainly controlled by PTH and vitamin D (Bouillon et al., 2014). Humans produce vitamin D by exposure to sunlight that includes ultraviolet B radiation; if ultraviolet B radiation is not available in sufficient amounts, vitamin D needs to be obtained from the diet or dietary supplements (Holick, 2007). The start of the vitamin D endocrine system is believed to have been initiated before the start of vertebrates and evolved over millions of years (Bouillon and Suda, 2014). Therefore, the vitamin D micronutrient either synthesized through the sun by the skin or through dietary uptake is well-adapted to the human body. The endogenously conjugated vitamin D metabolites have taken over many important roles in the maintenance of human health, of which many still await to be discovered.

In this paper, we summarize the knowledge on vitamin D as an essential micronutrient important for human health and discuss the new nutritional research on its way to gain further knowledge on the function of vitamin D for nutrition.

## **VITAMIN D PART OF NUTRITION AND CONTENT IN FOODS**

The history of vitamin D is linked to first scientific description of the classic bone disease rickets by Whistler in 1645 (Norman, 2012). Two centuries later it was Schütte who observed the usefulness of cod liver oil in the treatment of rickets and osteomalacia in 1824. The hunt for the anti-rachitic factor ended in early twentieth century, when Mellanby could demonstrate in a series of hallmark studies (1919–1924) that a nutritional component in the diet was the anti-rachitic factor to prevent rickets (Mellanby, 1919, 1976; Platt, 1956). Shortly after, vitamin D was inaugurated without the characterization of the chemical structure. In 1919, Hudschinsky showed in parallel that UV light was able to ameliorate rickets by increasing calcification in rachitic children (Huldschinsky, 1919, 1926). Both findings of the cod liver oil and the UV light preventing rickets remained independent observations until Hess and Weinstock elegantly could demonstrate that the anti-rachitic vitamin D was produce by UV irradiation in skin (Hess and Weinstock, 1925a,b). In 1936, Windaus and colleagues determined the chemical structure of the fat-soluble seco-steroid vitamin D (Windaus et al., 1936).

The vitamin D definition comprises a group of molecules called the calciferols. The main forms present in foods are cholecalciferol (vitamin D3) and ergocalciferol (vitamin D2), whereas the metabolite 25-hydroxycholecalciferol (25-hydroxyvitamin D3) is a natural part of the food chain by its occurrence in animal products. Vitamin D3 is unique by the fact that the same nutrient can be synthesized in the skin through the action of sunlight or being taken up by diet. This dual source of intake secures the body to maintain sufficient vitamin D levels in the body. The production in skin is usually the major vitamin D3 source for the body. However, in countries that receive insufficient sun exposure, people rely on dietary vitamin D as a major source. Exposure of the precursor 7-dehydrocholesterol in the basal and suprabasal layers of the epidermis to ultraviolet B (UVB) light with a wavelength of 290–315 nm is needed for the formation of the previtamin D3. The subsequent conversion is a non-enzymatic process that includes a thermal isomerization of the previtamin D3 to produce vitamin D3 (Collins and Norman, 2001; Holick, 2011). This vitamin D3 is rapidly converted to 25-hydroxyvitamin D3 in the liver. The vitamin D status is evaluated by measuring the circulating levels of serum 25-hydroxyvitamin D, which is the sum of cutaneous synthesis (vitamin D3) or dietary contribution (vitamin D3 and vitamin D2). The 25-hydroxyvitamin D3 needs to be further hydroxylated in the kidney (or locally in other organs Lehmann et al., 2001) to form 1,25-dihydroxyvitamin D3, the active endogenous hormone, which is responsible for most of the physiological actions of vitamin D through the binding to the vitamin D receptor (VDR). The plant-derived vitamin D2 is processed in the same way. For both vitamers, vitamin D2, and vitamin D3, the consecutive molecular action is believed to be identical, whereas only 1,25-dihydroxy vitamin D3 is the endogenous hormone, the activated vitamer 1,25-dihydroxyvitamin D2 is hormone mimetic. Therefore, it was not surprising that vitamin D3 has been reported to be superior to vitamin D2 in terms of bioavailability and maintaining the vitamin D status by the majority of studies (Trang et al., 1998; Armas et al., 2004; Romagnoli et al., 2008; Glendenning et al., 2009; Heaney et al., 2011; Lehmann et al., 2013). Only one study reported that the two vitamers were essentially equipotent (Holick et al., 2008).

The level of cutaneous vitamin D3 synthesis is mainly affected by the amount of solar UVB radiation reaching the human skin, which is a function that needs to take into account the wavelength, thickness of the ozone layer in the atmosphere and solar zenith angle. Furthermore, the geographic latitude, season of the year and time of day influence and restrict the skin-borne synthesis of vitamin D3 (Webb et al., 1988; Holick, 2011). It was described that vitamin D3 synthesis in the skin declines with age, which is due in part to a fall of 7-dehydrocholesterol and the morphological changes due to biological aging (MacLaughlin and Holick, 1985; Holick et al., 1989). Matsuoka et al. (1991) have shown that in Caucasians and Asian subjects having a lighter skin pigmentation UVB radiation produce significantly higher vitamin D3 serum levels than in African American and East Indian groups. It is not of a surprise that skin pigmentation reduces vitamin D3 formation. This skin tone dependent down regulation is easily overcome by increased sun exposures (Armas et al., 2007). Apart to darker pigmented skin, cutaneous vitamin D3 production can be reduced for many other reasons like severe air pollution in large cities, less outdoor activity as a consequence of an unhealthy lifestyle change, immobility of institutionalized elderly populations, topical application of sunscreens with a high sun protection factors or cultural dress codes (e.g., veiling). Therefore, dietary intake of vitamin D through foods or supplements plays a vital part to maintain healthy vitamin D levels.

Through nutrition, vitamin D intake is limited. There are few naturally-occurring food sources containing relevant levels of vitamin D. **Table 1** summarizes the vitamin D content in selected foods. Vegetarian diets are limited to the plant vitamin D2 that is only present in some mushrooms. Commercially dark cultivated white button mushrooms contain low amounts of vitamin D2, only wild mushrooms or sun-dried mushrooms contain elevated amounts of ergocalciferol (Mattila et al., 1994, 1999b, 2001; Teichmann et al., 2007). Some commercial producers include an UVB radiation step to increase the vitamin D2 content in their products (Mau et al., 1998; Roberts et al., 2008). Vitamin D2 is formed out from ergosterol in the mushrooms. Some plants that are used as foods however can contain ergosterol, but this provitamin form is not converted to vitamin D2. Vitamin D3 is not found in food-borne plants. In plants, the occurrence of vitamin D3-related compounds is scarce. Interestingly, species belonging to the botanical *Solanaceae* family, like *Solanum malacoxylon* (*Solanum glaucophyllum* and *Solanum glaucum*), contain a glycoside of the active 1,25-dihydroxyvitamin D3 hormone

### **Table 1 | Vitamin D content in raw products, processed foods, and fortified foods.**


(Boland, 1986; Boland et al., 2003; Japelt et al., 2013). This deciduous shrub (1.5–3.0 m stem length) is widely distributed in the provinces of Buenos Aires in Argentina and in Brazil and is responsible for the calcinotic disease in cattle and other grazing animal.

Animal food products are the main dietary source for naturally occurring vitamin D3 (Schmid and Walther, 2013). Since the discovery of vitamin D, vitamin D was associated with oily fish products. It was driven by the early observation that the amount of vitamin D in a teaspoon of cod liver oil was sufficient to prevent rickets in infants. It is still the fish liver oil that contains the highest amounts of vitamin D3. The highest reported concentration was found in skipjack liver oil 144,400μg/100 g (Takeuchi et al., 1984). The fish liver oils besides other nutritional ingredients might contain high levels of vitamin A. The vitamin A to vitamin D ratio in the fish liver oils is species and fishing area dependent. The ratio range starts with a factor of 0.5 for skipjack liver oil and can even reach an extreme ratio of 119 (pollack liver oil) (Takeuchi et al., 1984). This wide vitamin A to vitamin D ratio range is the reason why fish liver oils often need further processing. In fresh fish products we observe a huge variation in the vitamin D3 content per 100 g wet weight (Egaas and Lambertsen, 1979; Takeuchi et al., 1984, 1986; Kobayashi et al., 1995; Mattila et al., 1995a, 1997; Ostermeyer and Schmidt, 2006; Lu et al., 2007; Byrdwell et al., 2013) (**Table 1**). Large variations in vitamin D3 content were found within the same species, but also between the different fish species. Fish obtain their vitamin D3 requirements through their diet (Holick, 2003). Therefore, the vitamin D3 levels in the zooplankton, the primary food source of fish, or seasonal changes in the zooplankton reservoirs in the different habitats, might be the reasons for the observed fluctuation in the fish product. Interestingly, the weight, the sex, or the age of the fish could not be correlated to the vitamin D3 content. Furthermore, no significant correlation between the tissue fat content and vitamin D levels was detected (Mattila et al., 1995a, 1997). Significant differences in vitamin D3 content were found between muscle and skin tissues and even more pronounced between muscle and liver tissues (Takeuchi et al., 1986). The 25-hydroxyvitamin D3 compound was also detected, though at low concentrations (Takeuchi et al., 1986; Mattila et al., 1995a; Ovesen et al., 2003; Bilodeau et al., 2011).

Wild and sun-dried mushrooms can be a good dietary source of vitamin D2 (Mattila et al., 1994, 1999b, 2001; Kobayashi et al., 1995; Teichmann et al., 2007; Phillips et al., 2011). However, the commercially produced mushrooms, e.g., the white button mushroom, do not contain or contain only very low amounts of vitamin D2 (Mattila et al., 2001; Teichmann et al., 2007; Phillips et al., 2011). The vitamin D2 content in commercially produced mushrooms can be increased by UVB exposure during the culturing or the postharvest process (Mau et al., 1998; Roberts et al., 2008). The concentration of vitamin D in eggs can vary from 0.4 to 12.1μg (Parrish, 1979; Mattila et al., 1992, 1999a; Bilodeau et al., 2011; Exler et al., 2013), it is in a similar range like offal (Mattila et al., 1995b; Montgomery et al., 2000, 2002). Other animal products like pork, beef, and chicken muscle meat are low in vitamin D content (Mattila et al., 1995b; Montgomery et al., 2000, 2002; Bilodeau et al., 2011; Strobel et al., 2013). By adding vitamin D3 into the feed, the vitamin D3 content can be increased in muscle and liver of cattle, to 4.6μg per 100 g of tissue and 99.6μg per 100 g of tissue, respectively (Montgomery et al., 2004). Milk, unless fortified, has been shown to contain no or very little amounts of vitamin D, whereas in dairy products like butter and cheese the vitamin D content is higher, but in serving size amounts still very low (Kobayashi et al., 1995; Mattila et al., 1995b; Jakobsen and Saxholt, 2009; Trenerry et al., 2011). In general, household cooking seems to have some effect on vitamin D stability depending on the actual foodstuffs and the heating process used (Mattila et al., 1999b; Jakobsen and Knuthsen, 2014).

To meet the vitamin D needs in the countries some states fortify foods. Dairy products are ideal for vitamin D fortification. In Canada vitamin D fortification is mandatory for milk (1μg/100 ml) and margarine (13.3μg/100 g) (Health Canada, 2014). In other countries, like the United States, vitamin D fortification is optional for products like milk, breakfast cereals, and fruit juices (Calvo et al., 2004). In the U.S. Department of Agriculture (2013) of the US Department of Agriculture (USDA)'s Nutrient Databank System (Haytowitz et al., 2009), 5036 foods have been determined for their vitamin D content, of which only 259 food items had detectable vitamin D levels. The data showed that per serving only seven fish products had *>*15μg vitamin D. All 29 foods that contained between 2.5μg 15μg vitamin D per serving were either fortified foods (21) or fish produce (8). Two-thirds of all vitamin D containing foods were far below the 1.0μg level, whereas 20 percent had even negligible vitamin D content per serving (below 0.1μg).

Despite the fact that moderate sun exposure of arms and legs in summer for 5–30 min between the hours of 10 a.m. and 3 p.m. twice a week is enough to produce sufficient vitamin D3 in the body (Holick, 2007), it is astonishing that many populations that live at these privileged latitudes fail to achieve this goal (Holick and Chen, 2008; Lips, 2010; Wahl et al., 2012; Hilger et al., 2014). During winter time, when vitamin D3 production by the sun ceased, adequate vitamin D levels can only be achieved by UVB exposure from indoor tanning units, or by a daily diet of fortified foods or a few selected food items. This restricted list of options to achieve sufficient levels is one of the reasons, why the use of dietary vitamin D supplements has become so popular. It is currently the most applied and secure option to reach adequate vitamin D intake levels (Holick, 2007).

## **VITAMIN D MAP, MALNUTRITION, HIDDEN HUNGER, AND NUTRITION SECURITY**

An accepted biomarker for the vitamin D status in the general population is to measure the serum concentration of 25-hydroxyvitamin D levels, which is the major circulating form of vitamin D and reflects both dietary vitamin D intake and the endogenous vitamin D production (Lips, 2001, 2007). The serum concentration of 25-hydroxyvitamin D is linked to the serum level of the active hormone 1,25-dihydroxyvitamin D and also to the clinical relevant parathyroid hormone level. Lips has classified the 25-hydroxyvitamin D levels into four stages (Lips, 2001; Lips et al., 2013): severe deficiency (*<*12.5 nmol/L), deficiency (12.5–25 nmol/L), insufficiency (25–50 nmol/L), repletion (*>*50 nmol/L). The thresholds for severe deficiency and deficiency are undisputed; however, a controversy has arisen for defining the border between insufficiency and repletion. In 2011, the Institute of Medicine (IOM) suggested a serum level of 50 nmol/L as the value at which 97.5% of the vitamin D needs of the population would be covered (Institute of Medicine, 2011; Ross et al., 2011), whereas, the Endocrine Society (ES) defined it to be higher: 75 nmol/L (Holick et al., 2011). All deficiency levels including insufficiency, as so-called mild deficiency, must be prevented through focused supplementation.

In 2010, the Institute of Medicine (IOM) introduced new dietary reference intake (DRI) values for vitamin D after comprehensive reviewing of more than 1000 high quality research articles to renew thereby their first settings from 1997 (Institute of Medicine, 2011). The DRIs address an adequate nutritional intake of all sources. The IOM has set the dietary allowance (RDA) to 600 IU per day for the general population and at 800 IU per day for persons 70 years and older, whereas 1 IU is the biological equivalent of 0.025μg vitamin D3. The tolerable upper intake level or UL (Upper Level of Intake), which represents the safe upper limit, was set to 4000 IU per day for vitamin D intake (Ross et al., 2011). The new RDAs reflect the scientific outcome from large dietary studies that revealed vitamin D insufficiency (Looker et al., 2002; Zadshir et al., 2005). In 2012, Troesch et al. analyzed the vitamin intake from different dietary surveys that included the German Nutritional Intake Study (Nationale Verzehrstudie II) 2008 (Max Rubner-Institut, 2008), the US National Health and Nutrition Examination Survey (NHANES) from 2003 to 2008 (Centers for Disease Control and Prevention & National Center for Health Statistics, 2009), the UK (The British National Diet and Nutrition Survey, 2003) (Henderson et al., 2003) and the Netherlands (van Rossum et al., 2011), and could confirm that vitamin D is one of the critical vitamins, which intake is below the recommendation (Troesch et al., 2012).

A gap exists between the intake and the recommendation of vitamin D. The chronic insufficient intake of micronutrients like vitamin D without seeing immediate clinical signs is called Hidden Hunger. Hidden Hunger, in particular for vitamin D, is more prevalent in the populations of the developed countries as anticipated (Biesalski, 2013). Hidden Hunger is a threat for the nutrition security for a given country. Nutrition security mandates sufficient micronutrients in an adequate food supply and is required to safeguard an optimal nutritional status of a population.

Many groups have identified vitamin D deficiency or insufficiency to become a public health problem worldwide (Holick, 2007; Holick and Chen, 2008; Mithal et al., 2009; Lips and Van Schoor, 2011; Wahl et al., 2012; Hilger et al., 2014). Mithal et al. (2009) described in their global report that most populations do not achieve a desirable vitamin D status and particular people at risk and elderly people suffer from vitamin D deficiency. In two reports, the International Osteoporosis Foundation (IOF) and its partners published the global vitamin D status map (Wahl et al., 2012; Hilger et al., 2014). The vitamin D map was based on a systematic review of the worldwide vitamin D levels, using all available publications published between 1990 and February 2011 (Hilger et al., 2014). Eligible studies include 168,389 participants from the general populations throughout the world where the mean or median serum 25-hydroxyvitamin D levels were measured. Studies included had a cross-sectional design or were based on a population based cohorts. The analysis identified nearly 200 studies from 44 countries, whereas only half of the studies were included in the global vitamin D status map as 50.2% of the studies were not representative for the target populations. **Figure 1** shows the global vitamin D status map listed by countries and by continents. The largest numbers of studies were performed in Europe, followed by North America and Asia-Pacific. Available data from Latin America and even more from Africa are limited. Results of this review showed that 6.7% of the population were vitamin D deficient (mean 25-hydroxyvitamin D values *<*25 nmol/L), 37.3% were vitamin D insufficient according to IOM (mean values below 50 nmol/L) and 88.1% of the population showed an insufficient vitamin D status according to the ES (mean values below 75 nmol/L). No significant differences were found for gender or age, when looking at the worldwide data, but some regional differences could be identified (Hilger et al., 2014). The 25-hydroxyvitamin D serum levels were higher in Europe and the US, when compared to Middle East and Africa. This might be due to the vitamin D food fortification programs in North America (Prentice, 2008). Furthermore, the systematic analysis revealed that institutionalized elderly were more at risk to have low 25-hydroxyvitamin D levels in Europe and Asia/Pacific. The compared non-institutionalized elderly group showed higher levels, possibly due to spending more of time outdoors. The group of institutionalized elderly is therefore at high risk to become vitamin D deficient. Further research is needed to inform public health policy makers to reduce the risk for potential health consequences of low vitamin D status.

In the past few years the national recommendations for dietary vitamin D were adjusted in several countries; they are not harmonized across the European Union yet and vary from 200 to 800 IU. The higher recommendations for dietary vitamin D intake are increasingly being suggested in government documents, position statements and clinical practice guidelines for bone health. In 2008, the US Food and Drug Administration updated the health claim for the prevention of osteoporosis by including vitamin D to the consumption of calcium (Food and Drug Administration, 2008). In 2008, the American Academy of Pediatrics also reacted and issued an update of their guidelines for vitamin D intake and rickets prevention (Perrine et al., 2010). They doubled the recommended dose of vitamin D for children to 400 IU per day, beginning in the first few days of life and continuing throughout adolescence. In 2010, the Institute of Medicine (IOM) released the revised Dietary Reference Intakes (DRI's) for calcium and vitamin D and tripled the recommendations for vitamin D intakes to 600 IU per day for children and all adults up to age 69 years (Institute of Medicine, 2011). The IOM stated that there was insufficient evidence to make recommendations for non-skeletal benefits.

In 2012, the German, Austrian, and Swiss Nutrition Societies raised the recommended vitamin D intake to 800 IU per day, in case of absent UVB exposure, for all age groups starting from 1 year of age (German Nutrition Society, 2012). Furthermore, key opinion leaders are increasingly recommending higher daily intakes for vitamin D, between 800 and 1000 IU or even higher for people at risk or older adults. The recent statement by the IOF and the guidelines by the US ES suggest that higher vitamin D doses

25-hydroxyvitamin D3 serum thresholds starting from severe deficiency

consistent with severe vitamin D deficiency.

would be needed to achieve the desirable 25-hydroxyvitamin D serum level of 75 nmol/L for people at risk or older individuals.

Increasing the vitamin D levels in the population would also ameliorate health economics. Grant and colleagues calculated the benefit of increasing vitamin D levels to reduce the economic burden of diseases (Grant et al., 2009). A rise in the vitamin D serum level of all Europeans to 40 nmol/L would reduce the economic burden of different diseases and could save health care costs of up to 16.7%. Besides reducing the economic costs, vitamin D intake could in addition also reduce mortality rates and maintain a longer healthy life style.

## **NUTRITIONAL RESEARCH TO ADDRESS AND UNDERSTAND VITAMIN D INSUFFICIENCY**

Vitamin D deficiency is undoubtedly linked to severe consequences in the growing child by causing incomplete mineralization of the bone and in the adult accounting to wasteful osteomalacia. In the vitamin D insufficiency stage, this severity gets gradually less, but the outcome remains unchanged. Besides the established and accepted functional skeletal health relationship, more and more evidence is accumulating for falls (Pfeifer et al., 2000, 2009; Bischoff et al., 2003; Flicker et al., 2005; Broe et al., 2007; Prince et al., 2008; Bischoff-Ferrari et al., 2009) and physical performance (Bischoff-Ferrari et al., 2004; Houston et al., 2011; Ceglia et al., 2013; Redzic et al., 2013; Sohl et al., 2013; Tieland et al., 2013), which has been recognized by a health claim of the European Food and Safety Authority in 2011: "Vitamin D may reduce the risk of falling. Falling is a risk factor for bone fractures." This health claim is targeting men and women 60 years of age and older and the dose required is a daily consumption of 800 IU vitamin D, which can come from all sources. Further emerging vitamin D health relationships include physiological parameters like improved immune response (Baeke et al., 2010; Schwalfenberg, 2011; Hewison, 2012; White, 2012), improved respiratory health(Berry et al., 2011; Charan et al., 2012; Choi et al., 2013; Hirani, 2013) possibly also relate to reduced tuberculosis incidence (Nnoaham and Clarke, 2008; Martineau et al., 2011; Mitchell et al., 2011; Coussens et al., 2012; Salahuddin et al., 2013; Huaman et al., 2014); and reduced risk to develop autoimmune diseases like multiple sclerosis (Solomon and Whitham, 2010; Cantorna, 2012; Dobson et al., 2013) or type 1 diabetes (Hypponen et al., 2001; Holick, 2003; Ramos-Lopez et al., 2006; Baeke et al., 2010; De Boer et al., 2012; Dong et al., 2013; Van Belle et al., 2013). In chronic, non-communicable diseases, vitamin D deficiency is being discussed to possibly ameliorate the incidence of some neoplastic diseases like colorectal, lung, prostate, and breast cancers (Ng et al., 2008; Rosen et al., 2012; Welsh, 2012; Cheng et al., 2013); cardiovascular diseases (CVDs) including hypertension, myocardial infarction, stroke (Forman et al., 2007; Giovannucci et al., 2008; Gardner et al., 2011; Bischoff-Ferrari et al., 2012; Tamez and Thadhani, 2012; Karakas et al., 2013; Pilz et al., 2013a; Schroten et al., 2013); life-style diseases like obesity and type 2 diabetes (Pittas et al., 2007; González-Molero et al., 2012; Khan et al., 2013; Pilz et al., 2013b; Schottker et al., 2013; Tsur et al., 2013; Van Belle et al., 2013; Bouillon et al., 2014); diseases related to the decline in sight function including age-related macular degeneration (Parekh et al., 2007; Millen et al., 2011; Lee et al., 2012); and neurological disorders including Alzheimer and Parkinson disease (Buell and Dawson-Hughes, 2008; Annweiler et al., 2012; Eyles et al., 2013; Zhao et al., 2013). One may wonder about the width of possible implications being looked at, but considering the more than 1000 genes which vitamin D is regulating through the VDR (Carlberg and Campbell, 2013), this may actually not be a surprise. To determine the potential role of vitamin D supplementation in the prevention or treatment of chronic nonskeletal diseases notwithstanding, large-scale clinical trials are demanded. In this respect for the nutrition field, four new largescale ongoing long-term supplementation studies are expected to deliver results in near future (**Table 2**). The two very large studies, VITAL trial (*n* = 20*,* 000) and FIND study (*n* = 18*,* 000), are meant to deliver clinical evidence for the effect of vitamin D3 on cancer, CVD and diabetes outcomes. The two smaller trials, CAPS and DO-HEALTH, each having more than 2,000 participants are including cancer, infections, fractures, hypertension, cognitive function, and physical performance outcomes. In all four studies the placebo group will produce vitamin D3 in the skin and will possibly consume vitamin D through food, and therefore this will narrow the vitamin D serum level gap between the placebo and treatment groups. It remains to be seen whether the applied supplementation doses (2000 IU and 1600 IU, 3200 IU) will be sufficient to see a clear difference between the treatment and the control groups. An open likelihood will remain for the placebo group potentially obtaining sufficient vitamin D3 (600– 800 IU) levels that are considered to be sufficient for skeletal effects. In such a case only an incremental increase of an additional ∼1000 IU can be considered as the effective dose, for which no power calculation was available at the time before study begun. In light of such a situation, it will be of interest whether the micronutrient triage theory of Bruce Ames can be validated with vitamin D3 (Ames, 2006; McCann and Ames, 2009). The triage theory postulates, as a result of recurrent shortages of micronutrients during evolution, that the body has selected and developed a metabolic rebalancing response to shortage. These rebalancing favored micronutrient-needs for short term survival, while those only required for long-term health were starved. In the case of the micronutrient vitamin D3, calcium and bone metabolism can be considered to be secured with highest priority, therefore, it might be speculated that the 600–800 IU intake would satisfy this vitamin D3 serum level threshold. For the chronic non-skeletal diseases however, which have only secondary priority in an evolutionary perspective, higher serum vitamin D3 levels would be required. The ongoing four vitamin D3 studies that have chronic diseases as their main outcomes and use nutritionally relevant ∼2000 IU are therefore well-suited to address whether the triage theory holds also true for the micronutrient vitamin D3.

Vitamin D3 once in the blood immediately binds to the vitamin D-binding protein (DBP) and gets transported into the liver (Holick, 2007). The first hydroxylation at position 25 generates the major circulating metabolites 25-hydroxyvitamin D3. This metabolite circulates throughout all organs and undergoes hydroxylation at position 1, which occurs mainly in the kidney, but also in other organs, to form 1,25-dihydroxyvitamin D3, the active hormone. Besides the major circulating metabolite 25-hydroxyvitamin D3 and the hormonally active metabolite 1,25-dihydroxyvitamin D3, more than 35 additional vitamin D3 metabolites are formed by the body (Bouillon et al., 1995; Norman et al., 2001). It is speculated that they might be intermediates in the catabolism of 1,25-dihydroxyvitamin D3. The human body has evolved many CYP enzymes and invests energy to form these additional 35 vitamin D3 metabolites, whether this is for the purpose to catabolize 1,25-dihydroxyvitamin D3, remains still to be answered. More appealing is the theory that these metabolites are formed to fulfill yet unknown functions of vitamin D3. This perspective could potentially also account to the pleiotropic nonskeletal health benefits reported by the many vitamin D intake studies. For some of the vitamin D3 metabolites like the 24*R*,25 dihydroxyvitamin D3 potential function was explored *in vitro* (Norman et al., 2002).

The 24*R*,25-dihydroxyvitamin D3 has been shown to be an essential hormone in the process of bone fracture healing. The 24*R*,25-dihydroxyvitamin D3 most likely initiates its biological responses via binding to the ligand binding domain of a postulated cell membrane receptor VDRmem24*,*25, similar to the better studied, but still not cloned cell membrane receptor for 1,25 dihydroxyvitamin D3, VDRmem1*,*<sup>25</sup> (Norman et al., 2002). From the nutritional point of view, it will be of interest to investigate the function of the all vitamin D3 metabolites and relate the function to the level of vitamin D3 intake to secure the health benefit according to the triage theory.


**|ListofongoinglargenutritionalvitaminD3supplementationtrials(***>***2,000subjects)usingnutrition-relateddailyvitaminD3doses**

According to the current knowledge, the vitamin D endocrine system is funneled through the biologically most active metabolite 1,25-dihydroxyvitamin D3 that is mainly produced in the kidney, but also in other organs (Bouillon et al., 2013). Mechanistically 1,25-dihydroxyvitamin D3 binds the VDR directly on a DNA sequence, the 1,25-dihydroxyvitamin D3 response element (VDRE), in the regulatory region of primary 1,25-dihydroxyvitamin D3 target genes (Carlberg and Campbell, 2013). The VDR forms together with the retinoid X receptor or putative other transcription factors a heterodimer on the VDRE, recruiting tissue-specific transcriptional co-activators and regulates through a conformational change upon 1,25-dihydroxyvitamin D3 binding the downstream gene. The VDR is widespread in more than 30 tissues (Bouillon et al., 1995) and may trigger expression of more than 1000 genes through 1,25-dihydroxyvitamin (Carlberg et al., 2013; Hossein-Nezhad et al., 2013). The regulation of tissue-specific gene expression by 1,25-dihydroxyvitamin D3 is of high interest, as it guides us toward the better understanding of the mechanistic action of vitamin D3 in the different tissues. The gained knowledge from the mechanistic studies can help to design smaller and more focused nutritional intervention RCTs to answer whether vitamin D contributes to a specific health benefit of interest. In this respect the GeneChip-based transcriptomics methodology using high-density microarrays demonstrated the expression of genes in a variety of important functions of more than 100 different pathways that could be linked to vitamin D deficiency (Bossé et al., 2007; Tarroni et al., 2012; Hossein-Nezhad et al., 2013). The development of chromatin immunoprecipitation (ChiP) methodology linked to site-specific PCR amplification of the VDR bound genomic DNA fragment, and later the methods using tiled microarrays (ChiP-chip) applying the first unbiased genome-wide approach, which then was followed by the massive parallel NGS sequencing approach of the immunoprecipitated DNA segments, opened up new avenues to investigate 1,25-dihydroxyvitamin D3 target genes in selected tissues (Ramagopalan et al., 2010; Heikkinen et al., 2011; Carlberg et al., 2012, 2013; Pike et al., 2014). In an elegant study, Carlberg et al. identified in samples of 71 pre-diabetic individuals of the VitDmet study changes in serum 25-hydroxyvitamin D3 concentrations that were associated to primary vitamin D target genes (Carlberg et al., 2013). Based on their finding the authors proposed the genes CD14 and THBD as transcriptomics biomarkers, from which the effects of a successful vitamin D3 supplementation can be evaluated. These biomarkers are potentially suitable for displaying the transcriptomics response of human tissues to vitamin D3 supplementation.

Epigenetic alterations of the genome refer to heritable and modifiable changes in gene expression that are not affecting the DNA sequence. They may be inherited as Mendelian, non-Mendelian, or environmentally caused traits. One of the 1,25-dihydroxyvitamin D3 induced epigenetic modification was shown for the hypo-methylating effect on the osteocalcin promoter (Haslberger et al., 2006). 1,25-Dihydroxyvitamin D3 was associated with the demethylation of the osteocalcin promoter and induced the osteocalcin gene expression. The activity of VDR can be modulated by epigenetic histone acetylation. The VDR alone or in concert with other transcription factors can recruit histone-modifying enzymes like histone acetyl transferases (HATs) or histone deacetylases (HDACs) and epigenically direct transcriptional expression of downstream genes (Burrell et al., 2011; Karlic and Varga, 2011; Sundar and Rahman, 2011; Hossein-Nezhad et al., 2013). The trans-generational epigenetic inheritance of vitamin D3 triggered epigenome modification is not fully explored, however maternal vitamin D deficiency has been discussed with adverse pregnancy outcomes or potential susceptibility for diseases (Burrell et al., 2011; Hossein-Nezhad and Holick, 2012). For future nutritional research it would be of great value to identify and validate epigenetic biomarkers that could serve as risk assessment tool for vitamin D insufficiency related susceptibility to develop a disease later in life.

Variations in vitamin D status have been shown to be related to inheritance. The disparity of vitamin D levels according to ethnicity given skin pigmentation is well-established (Cashman, 2014; Ng et al., 2014). Dark skinned population individuals have compared to Caucasian descendants almost one-half the serum concentrations of 25-hydroxyvitamin D (Nesby-O'dell et al., 2002). From twin studies it has been estimated that the heritability of genetic regulation of vitamin D levels to be between 23 and 80% (Dastani et al., 2013). In addition, large-scale genetic association studies using linkage disequilibrium analysis have identified genetic loci correlating with serum vitamin D level within five candidate genes (Dastani et al., 2013). The identified SNPs are within the 1alpha-hydroxylase of 25-hydroxyvitamin D (CYP27B1) gene, the 25-hydroxylase of vitamin D (CYP2R1) gene, the vitamin D carrier protein (GC) gene, the VDR gene, and the cytochrome P-450 (CYP24A1) gene coding for an enzyme that inactivates 1,25-dihydroxyvitamin D. It is important to note that replication studies in separate populations have to follow to verify the validity of the identified SNPs. The SNP information will provide the additional guidance toward a personalized nutritional advice to reach a sufficient vitamin D status.

## **CONCLUSION AND FUTURE PERSPECTIVES**

In the recent years the knowledge about vitamin D and its implications have extended far beyond its classical role in bone health in either fields of basic research as well as in human trials. In particular, the evidence for the role of vitamin D in reducing the risk of fractures as well as decreasing the risk for falling is convincing and authorities have responded to it. Besides a health claim issued by the EFSA on the risk reduction for falling the dietary intake recommendations have been significantly increased in several countries such as the US and in Europe (Austria, Germany, Switzerland). A number of other countries around the globe are in the process of establishing new dietary intake recommendations as well. It turns out that on average a daily intake of 600–800 IU vitamin D appears to be required to meet fundamental needs of the human body, for specific applications higher daily intakes may be necessary, which will become clearer as the results of a number of ongoing clinical studies will become available.

The obvious question to answer is: do people obtain the recommended amounts of vitamin D? The diet is typically only a minor vitamin D source as only few food items contain relevant amounts of vitamin D, such as fatty sea fish. The primary vitamin D source for humans is the vitamin D synthesis in the skin from vitamin D precursors by the sunlight—provided the skin is sufficiently exposed to strong enough sun radiation. Several groups have reviewed the published results on 25-hydroxyvitamin D serum levels the established marker of the vitamin D status, showing that low 25-hydroxyvitamin D levels are found in many cohorts around the world. A recent systematic review of the global vitamin D status (Hilger et al., 2014) showed that 6.7% of the overall populations reported deficient 25-hydroxyvitamin D levels below 25 nmol/L, 37% had 25-hydroxyvitamin D levels below 50 nmol/L, and only 11% were above 75 nmol/L, which is considered an adequate status by the IOF and the ES. So a very important task ahead of us is to find efficient ways to improve the vitamin D status on the population level, be it by dietary means, food fortification, or dietary supplements.

In addition, it will be very important to gather sound and convincing evidence for the many additional implicated health benefits of vitamin D besides the ones that already reached a health claim status and to see which of them will actually hold up. This will require appropriate human studies on the one hand, and also involve the appropriate use of the novel experimental approaches like nutrigenomics, nutrigenetics, and nutriepigenetics on the other hand. In conclusion, the evidence we have for vitamin D in human health is exciting, however we have to make sure that appropriate measures are taken to improve the vitamin D status to the levels required to be beneficial for human health. In future, we will also need to further apply, exploit and invest in novel, innovative and break-through technologies in the vitamin D research to understand the underlying mechanisms by which vitamin D is exerting so many effects in the human body, which is knowledge needed to the purpose to obtain and secure optimal public health through nutrition.

## **REFERENCES**


M. Ruel, M. Van Ameringen, H. K. Biesalski, M. Bloem, et al. (Basel: Karger), 39–50.


Holick, M. F., Matsuoka, L. Y., and Wortsman, J. (1989). Age, vitamin D, and solar ultraviolet. *Lancet* 2, 1104–1105. doi: 10.1016/S0140-6736(89)91124-0


**Conflict of Interest Statement:** The authors are employees of DSM Nutritional Products Ltd. and declare to have no conflict of interest.

*Received: 24 March 2014; accepted: 14 June 2014; published online: 11 July 2014. Citation: Bendik I, Friedel A, Roos FF, Weber P and Eggersdorfer M (2014) Vitamin D: a critical and essential micronutrient for human health. Front. Physiol. 5:248. doi: 10.3389/fphys.2014.00248*

*This article was submitted to Integrative Physiology, a section of the journal Frontiers in Physiology.*

*Copyright © 2014 Bendik, Friedel, Roos, Weber and Eggersdorfer. 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.*

## Impact of vitamin D on immune function: lessons learned from genome-wide analysis

#### *Rene F. Chun1, Philip T. Liu1, Robert L. Modlin2, John S. Adams <sup>1</sup> and Martin Hewison1 \**

*<sup>1</sup> Department of Orthopaedic Surgery, Orthopedic Hospital Research Center, David Geffen School of Medicine, University of California at Los Angeles, Los Angeles, CA, USA*

*<sup>2</sup> Division of Dermatology, Department of Medicine, David Geffen School of Medicine, University of California at Los Angeles, Los Angeles, CA, USA*

#### *Edited by:*

*Carsten Carlberg, University of Eastern Finland, Finland*

#### *Reviewed by:*

*Dieter Steinhilber, Goethe Universitat, Germany Mieke Verstuyf, Katholieke Universiteit Leuven, Belgium*

#### *\*Correspondence:*

*Martin Hewison, Orthopedic Hospital Research Center, 615 Charles E. Young Drive South, Room 410D, Los Angeles, CA 90095-7358, USA*

*e-mail: mhewison@mednet.ucla.edu*

Immunomodulatory responses to the active form of vitamin D (1,25-dihydroxyvitamin D, 1,25D) have been recognized for many years, but it is only in the last 5 years that the potential role of this in normal human immune function has been recognized. Genome-wide analyses have played a pivotal role in redefining our perspective on vitamin D and immunity. The description of increased vitamin D receptor (VDR) and 1α-hydroxylase (CYP27B1) expression in macrophages following a pathogen challenge, has underlined the importance of intracrine vitamin D as key mediator of innate immune function. It is now clear that both macrophages and dendritic cells (DCs) are able to respond to 25-hydroxyvitamin D (25D), the major circulating vitamin D metabolite, thereby providing a link between the function of these cells and the variations in vitamin D status common to many humans. The identification of hundreds of primary 1,25D target genes in immune cells has also provided new insight into the role of vitamin D in the adaptive immune system, such as the modulation of antigen-presentation and T cells proliferation and phenotype, with the over-arching effects being to suppress inflammation and promote immune tolerance. In macrophages 1,25D promotes antimicrobial responses through the induction of antibacterial proteins, and stimulation of autophagy and autophagosome activity. In this way variations in 25D levels have the potential to influence both innate and adaptive immune responses. More recent genome-wide analyses have highlighted how cytokine signaling pathways can influence the intracrine vitamin D system and either enhance or abrogate responses to 25D. The current review will discuss the impact of intracrine vitamin D metabolism on both innate and adaptive immunity, whilst introducing the concept of disease-specific corruption of vitamin D metabolism and how this may alter the requirements for vitamin D in maintaining a healthy immune system in humans.

**Keywords: macrophage, dendritic cell, intracrine, antigen-presentation, antibacterial, CYP27B1, VDR**

## **INTRODUCTION**

Amongst the many reported extra-skeletal effects of vitamin D, its ability to regulate immunity through effects on both the innate and adaptive systems has received considerable attention. This stems in part from homage to studies carried out more than a century ago by a then relatively unknown scientist, Dr. Nils Finsen. In 1903 Dr. Finsen won the Nobel Prize for Medicine or Physiology for showing that he could cure the epidermal form of tuberculosis (TB), lupus vulgaris, using concentrated light irradiation (Moller et al., 2005). The subsequent discovery that exposure to ultra-violet light promotes epidermal synthesis of vitamin D led to further studies describing the successful use of oral vitamin D supplementation to treat lupus vulgaris, and other mycobacterial infections such as leprosy (Airey, 1946; Herrera, 1949). The advent of antibiotic therapies for infectious diseases appeared to have consigned these studies to the history books. However, in 2006 the work of Finsen returned to center stage as a consequence of a series of genome-wide analyses that revealed pathogeninduction of an intracrine vitamin D system in monocytes (Liu et al., 2006), and an associated mechanism for anti-mycobacterial actions of vitamin D (Wang et al., 2004), whilst also shedding light on how these responses may vary according to the vitamin D "status" of any given individual. With increasing awareness of vitamin D-deficiency across the globe (Holick, 2007), and ongoing discussions concerning the physiological and clinical relevance of this (Holick et al., 2011; Ross et al., 2011), these genome-wide analyses have played a pivotal role in defining our new perspective on non-classical vitamin D physiology. The current review will detail these developments and how they have helped to define a role for vitamin D in normal immune function.

#### **ANTIBACTERIAL RESPONSES TO VITAMIN D**

Despite its early use in the treatment of mycobacterial diseases such as TB and leprosy (Airey, 1946; Herrera, 1949), the immunomodulatory actions of vitamin D did not become clear until much later. Elucidation of this important non-classical action of vitamin D stemmed from two key observations. Firstly, most proliferating cells within the immune system express the nuclear receptor for active 1,25-dihydroxyvitamin D (1,25D) the vitamin D receptor (VDR). Initial studies focused on 1,25D binding capacity in cells from the adaptive immune system such as T and B lymphocytes (T and B cells) (Bhalla et al., 1983; Provvedini et al., 1983), with subsequent reports describing specific intracellular binding of 1,25D in cells from the innate immune system such as monocytes/macrophages (Kreutz et al., 1993), dendritic cells (DC) (Brennan et al., 1987), neutrophils (Takahashi et al., 2002), and monocytic cell lines (Mangelsdorf et al., 1984). The functional significance of these data was not immediately clear but, nevertheless, it was assumed that VDRexpressing immune cells were able to respond the circulating active 1,25D in a similar fashion to classical vitamin D target tissues such as the intestine, kidney, and bone. However, this assumption was challenged by the second major observation linking vitamin D and the immunity, namely the discovery of active vitamin D metabolism by cells from the immune system.

Elevated serum levels of 1,25D reported for some patients with the granulomatous disease sarcoidosis were shown to be due to conversion of pro-hormone 25D to 1,25D by tissue and systemic macrophages in these patients (Barbour et al., 1981; Adams et al., 1983). Similar observations for other inflammatory and granulomatous diseases (Kallas et al., 2010) suggested that immune activity of the enzyme that catalyzes metabolism of 25D to 1,25D, 25-hydroxyvitamin D-1α-hydroxylase (1α-hydroxylase) was a disease-related phenomenon. However, other studies, *in vitro*, highlighting the potential for macrophage 1α-hydroxylase activity in the absence of disease (Koeffler et al., 1985; Reichel et al., 1986) supported the exciting possibility that synthesis of 1,25D is part of normal immune function. Despite this, it was another 20 years before evidence to support this proposal was reported. Significantly, the major advances that provided this evidence involved genome-wide strategies that explored both the regulation and function of vitamin D by immune cells.

The first of the genome-wide studies to shed light on extraskeletal actions of vitamin D was published by John White and colleagues at McGill University in Montreal and utilized a combination of DNA array and *in silico* strategies. In this report, DNA array analysis of 1,25D-regulated genes in squamous cell carcinoma cells *in vitro* (Akutsu et al., 2001; Lin et al., 2002) was combined with *in silico* analysis of genomic VDR binding sites to provide a comprehensive overview of potential 1,25D-VDR target genes (Wang et al., 2005). Genome-wide analysis of DNA sequences that are able to bind liganded VDR revealed consensus vitamin D response elements (VDRE) within the gene promoters for two antibacterial proteins, cathelcidin (*CAMP*) and β-defensin 2 (*DEFB4*) (Wang et al., 2005). Interestingly, although both of these genes exhibited classical proximal promoter directrepeat 3 (DR3) consensus VDREs, only *CAMP* appeared to be transcriptionally induced by 1,25D in monocytes (Wang et al., 2005). The underlying mechanism for the differential regulation of monocyte *CAMP* and *DEFB4* by 1,25D was elucidated in subsequent studies, the first of which described increased expression of monocyte *DEFB4* following co-treatment with 1,25D and the inflammatory cytokine interleukin-1 (IL-1) (Liu et al., 2009). Based on these observations and promoter analysis for the *CAMP* and *DEFB4* genes, it was concluded that transcriptional induction of *DEFB4* requires cooperative occupancy of nuclear factor-κB (NF-κB) response elements as well as VDRE within the *DEFB4* gene promoter. By contrast, induction of *CAMP* appears to be primarily dependent on binding of VDR to promoter VDRE (Liu et al., 2009). The importance of NF-κB and VDR as co-inducers of *DEFB4* transcription was further emphasized by studies of the intracellular pathogen sensing protein NOD2 which is itself transcriptionally induced by 1,25D (Wang et al., 2010b). Cells cotreated with 1,25D and the ligand for NOD2, muramyl dipeptide (MDP), a cell wall product of Gram-positive and Gram-negative bacteria, showed potent NF-κB-dependent induction of *DEFB4* (Wang et al., 2010b). In these studies expression of *CAMP* was also enhanced by 1,25D-MDP co-treatment, suggesting that NFκB may cooperate with VDR in a variety of immunomodulatory functions (**Figure 1**).

Subsequent functional studies confirmed that dose-dependent vitamin D induction of *CAMP* transcription involves direct interaction between liganded VDR and VDRE in the *CAMP* gene promoter (Wang et al., 2004; Gombart et al., 2005). Intriguingly, the VDRE initially identified within the *CAMP* promoter appear to be specific for subhuman and human primates, as there are no similar motifs within equivalent genes for lower mammals (Gombart et al., 2005). Acquisition of a VDRE by the *CAMP* gene appears to have occurred following the introduction of an

vitamin D via nuclear factor-κB (NF-κB) (green arrows).

Alu short interspersed nuclear element (SINE) that placed *CAMP* under the control of 1,25D-VDR (Gombart et al., 2009b). This primate-specific adaptation has been conserved in humans and apes as well as Old World and New World primates, suggesting that a mechanism for transcriptional regulation of *CAMP* by vitamin D confers biological advantages. The assumption is that this mechanism will be potently activated by the relatively high circulating levels of 25D and 1,25D that are characteristic of non-human primates (Adams et al., 1985). A similar mechanism would also have been advantageous in early *Homo sapiens* whose existence is likely to have been defined by routine exposure to ultra-violet light and increased cutaneous vitamin D production, with associated high circulating levels of 25D. Conversely, in modern man where serum 25D status is more variable, this antibacterial mechanism may be less effective. Whilst this hypothetical mechanism broadly supports beneficial innate immune effects of vitamin D supplementation, the important question still remaining is how variations in serum levels of inactive 25D are able to influence immune responses driven by intracellular 1,25D and VDR. The answer to this question has been pivotal to our perspective on the non-classical actions of vitamin D and was, again, provided by genome-wide analyses.

For cells from both within and outside the classical immune system, recognition of and response to pathogens involves surveillance of pathogen-associated molecular patterns (PAMPs) by pattern recognition receptors (PRR). Prominent amongst these PRRs is the extended family of Toll-like receptor (TLR) noncatalytic transmembrane receptors which interact with specific PAMPs (Moresco et al., 2011). To clarify the role of the TLR2/1 heterodimer in mediating innate immune responses to the TB pathogen, *Mycobacterium tuberculosis* (*M. tb*), Liu and colleagues carried out DNA array analysis of gene expression in human macrophages and DCs following treatment with one of the putative PRRs for *M. tb* (Liu et al., 2006). Amongst the many macrophage genes shown to be differentially regulated by 19 kDa lipoprotein (a TLR2-interacting PAMP), expression of both *CYP27B1* and *VDR* was increased (Liu et al., 2006) (see **Figure 1**). This genome-wide approach provided the first unbiased evidence of a role for vitamin D metabolism and signaling in innate immune responses to a pathogen. Crucially, the TLR2/1-stimulus induced expression of both *CYP27B1* and *VDR*, suggesting that macrophage responses to *M. tb* involve an endogenous, intracrine, vitamin D system. Further studies showed that macrophages co-treated with the TLR2/1-ligand, 19 kDa lipoprotein were responsive to both active 1,25D and inactive 25D, confirming the functional efficacy of the intracrine model. Moreover, inhibition of either 1α-hydroxylase activity, or VDR function blocked the actions of 25D, underlining the importance of cell-specific metabolism as a determinant of vitamin D immunoregulation.

The specific functional read-outs used in the TLR2/1-*M.tb* study were induction of mRNA for the vitamin D catabolic enzyme 24-hydroxylase (*CYP24A1)* and the antibacterial protein *CAMP* (Liu et al., 2006). Parallel analysis of the other antibacterial target gene for 1,25D, *DEFB4,* did not reveal significant induction of this gene in the presence of 19 kDa lipoprotein and 25D. However, as outlined above, subsequent experiments demonstrated that co-treatment with either IL-1 (Liu et al., 2009), or the NOD2 ligand MDP (Wang et al., 2010b) cooperates with the TLR2/1 ligand and 25D to stimulate expression of *DEFB4*. Antibacterial proteins such as CAMP and DEFB4 play a crucial role in vitamin D-mediated killing of intracellular bacteria. Monocytes treated with increasing concentrations of CAMP peptide show a dose-dependent decrease in the viability of internalized *M. tb* (Liu et al., 2006); a similar inhibition of macrophage *M. tb* viability occurs in the presence of 25D, with this effect being abrogated by a VDR antagonist. In other studies knockdown of either *CAMP* or *DEFB4* decreased killing of *M. tb* in macrophages, suggesting that both antibacterial proteins are important in mediating vitamin D-induced responses to mycobacterial infection (Liu et al., 2009). Both CAMP and DEFB4 are detectable in the circulation where they are able to support innate immune responses to extra-cellular pathogens including anti-viral responses (Barlow et al., 2011; Tripathi et al., 2013). However, pathogens such as mycobacteria are internalized by phagocytosis, and bacterial killing can then take place following fusion of the resulting phagosome with a lysosome to form a phagolysosome. To evade this antibacterial process and maintain intracellular viability *M. tb* can subvert the transition of phagosomes to phagolysosomes (Vergne et al., 2004). In this situation, the host cell can restore pathogen encapsulation by evoking alternative mechanisms such as autophagy, in which encapsulation of organelles, cell proteins or intracellular pathogens in a double-membrane autophagosome occurs prior to fusion with lysosomes (Gutierrez et al., 2004). Although a well-recognized feature of eukaryotic cells cytosolic homeostasis (Klionsky and Emr, 2000), autophagy also appears to play a pivotal role in cellular response to infection (Gutierrez et al., 2004; Deretic and Levine, 2009). The ability of 1,25D and its synthetic analogs to promote autophagy is well established (Hoyer-Hansen et al., 2005; Wang et al., 2008), but recent data suggest that induction of autophagy may be particularly important for vitamin D-induced antibacterial responses to *M. tb* infection (Yuk et al., 2009; Shin et al., 2011) (**Figure 1**). The precise mechanism for this is not clear and may involve inhibition of the mammalian target of rapamycin (mTOR) intracellular signaling system (O'Kelly et al., 2006; Lisse et al., 2011) with mTOR acting to suppress the induction of autophagy (Sanjuan et al., 2009). It has also been suggested that vitamin D-induced autophagy occurs via an indirect mechanism, in that RNA-interference (RNAi) knockdown of antibacterial *CAMP* was sufficient to abrogate 1,25D-induced autophagy in monocytes (Yuk et al., 2009). In common with effects on expression of antibacterial proteins, it was noted that monocyte autophagy following activation of TLR2/1 involves enhanced expression of VDR and CYP27B1 (Shin et al., 2011), further highlighting the importance of intracrine 25D metabolism and action in normal human innate immunity.

Intracrine synthesis of 1,25D has also been shown to regulate expression of another antibacterial protein, hepcidin antibacterial protein (HAMP) (Bacchetta et al., 2013b). However, in contrast to CAMP and DEFB4, the direct microbiocidal properties of HAMP appear to be relatively weak. Instead, the major function of HAMP appears to be suppression of the cell membrane protein ferroportin, the only known exporter of intracellular iron (Ganz, 2011). This link between HAMP and ferroportin in cells such as enterocytes, hepatocytes and monocytes plays a key role in the so-called anemia of infection or chronic disease (Ganz, 2009). Because pathogens such as bacteria utilize iron to maintain growth, restriction of circulating iron concentrations provides an important host response to systemic infection (Drakesmith and Prentice, 2012). However, for pathogens such as *M. tb* that attempt to evade immune surveillance at the intracellular level, accumulation of iron within this environment may promote the growth of internalized pathogens such as *Salmonella typhimurium* (Nairz et al., 2007), *M. tb* (Schaible et al., 2002; Sow et al., 2007, 2009), and *Chlamydia psittaci* (Paradkar et al., 2008). Conversely, innate immune and viral stimuli are known to stimulate the expression of *HAMP* (Sow et al., 2009; Armitage et al., 2011). In this setting suppression, rather than induction, of HAMP by 25D and 1,25D may be beneficial by abrogating HAMP-induced suppression of ferroportin which, in turn, will favor iron export and lower intracellular concentrations of iron. In studies carried out by our group at UCLA, we have shown that 25D and 1,25D suppress transcription of *HAMP* in monocytes and hepatocytes, leading to increased membrane expression of ferroportin, and decreased expression of ferritin (a surrogate biomarker for intracellular iron concentrations) (Bacchetta et al., 2013b). Moreover, in contrast to CAMP and DEFB4, elevated serum 25D levels (but not 1,25D) following vitamin D supplementation of human subjects *in vivo* were associated with potent suppression of circulating concentrations of HAMP (Bacchetta et al., 2013b). It therefore appears that regulation of the HAMP-ferroportin axis is another key facet of vitamin D-mediated innate immune function, complimentary to its reported effects on antibacterial proteins (Liu et al., 2006; Adams and Hewison, 2008; Hewison, 2011), and autophagy (Yuk et al., 2009; Shin et al., 2011) (see **Figure 1**). However, it is important to recognize that the effect of vitamin D on serum levels of hepcidin may have additional consequences that are both positive (suppression of anemia) and negative (decreased hepcidin for systemic infection). This may be particularly important for patients with chronic kidney disease (CKD) who commonly present with impaired circulating levels of 25D and 1,25D, and who are at higher risk of infection. In CKD, low serum 25D has been shown to correlate inversely with anemia (Lac et al., 2010) and directly with blood hemoglobin levels (Kiss et al., 2011). These effects may be due to dysregulation of normal HAMP-ferroportin function under conditions of vitamin D-deficiency, further emphasizing the importance of vitamin D supplementation in these patients.

Vitamin D may also target other innate immunity mechanisms. For example, studies *in vitro* have shown that 1,25D promotes hyporesponsiveness to PAMPs by downregulating expression of TLR2 and TLR4 on monocytes (Sadeghi et al., 2006). In this way, vitamin D appears to promote feedback control pathways that limit antibacterial activity and other innate immune responses, thereby preventing potential inflammatory events that arise from an over-elaboration of immune responses, notably inflammatory T cell responses. Paradoxically, vitamin D can also promote responses that amplify innate immune function. Recent studies have described 1,25D-mediated induction of the triggering receptor on myeloid cells-1 (TREM-1) (Kim et al., 2013), a cell surface protein associated with cytokine and chemokine production (Bouchon et al., 2000) that can also act to amplify TLR signaling (Bouchon et al., 2001). The biological significance of this is still not clear and this mechanism may be more important for cells such as neutrophils which are the principal source of circulating CAMP. Neutrophils express VDR but, unlike monocytes/macrophages, they do not appear to express a functional 1α-hydroxylase and are therefore not subject to intracrine activation of innate immune function. In this setting, activation of proteins such as TREM-1 may help to promote neutrophil responses to circulating 1,25D rather than 25D through enhanced TLR-signaling. This, in turn, would stimulate expression of VDR and sensitivity to 1,25D.

A similar cooperative TLR response has also been described for epithelial keratinocytes, where basal expression of *CYP27B1* is insufficient to facilitate intracrine induction of antibacterial proteins by serum 25D. However, following skin wounding, locally generated transforming growth factor β (TGFβ) enhances expression of CYP27B1 (Schauber et al., 2007). The resulting TGFβdriven *CYP27B1* expression is then able to stimulate intracrine generation of antibacterial proteins such as CAMP to combat potential infections associated with epidermal injury (Schauber et al., 2007). Interestingly, the TGFβ-induced 1α-hydroxylase activity was also associated with increased keratinocyte expression of TLR2 which further enhances surveillance of infectious bacteria, but also suggests that the effects of vitamin D on TLR expression are likely to be cell-specific. TGFβ and 1,25D may also cooperate to promote expression of other pathways linked to enhanced innate immune responses to infection such as induction of the enzyme 5-lipoxygenase (5-LO) that catalyzes synthesis of leukotrienes. Expression of 5-LO in human monocytes is induced by both 1,25D and TGFβ (Harle et al., 1998), with 1,25D enhancing expression of 5-LO through novel promoterindependent VDRE within exons 10 and 12 and intron M of the 5-LO gene (Stoffers et al., 2010). Although commonly associated with bronchial dilation and asthma, leukotrienes are also known to participate in leukocyte accumulation at sites of infection and phagocytosis of bacteria (Peters-Golden et al., 2005). Leukotrienes have also been shown to trigger the processing of antibacterial CAMP by neutophils (Wan et al., 2007).

Vitamin D-mediated innate immune responses may also be species-specific. VDR-mediated induction of *CAMP* and *DEFB4*, as well as suppression of *HAMP*, appears to be primate-specific; other mammals may therefore utilize alternative innate immunity targets for intracrine 1,25D. For example reactive oxygen species (ROS) can be bacteriocidal; previous studies have shown that macrophages infected with *M. tb* in the presence of 1,25D produce high levels of the superoxide anion ROS via the NADPH oxidase system (Sly et al., 2001). More recent studies have shown that another ROS, nitric oxide (NO), is produced by mouse macrophages as part of innate immune responses to infection, with bacteriocidal consequences (Kohchi et al., 2009). The NO pathway appears to play a pivotal role in mouse responses to *M. tb* infection (Chan et al., 1992), but its importance to human *M. tb* infection is less clear. Moreover, one study using 1,25D and mouse macrophages has reported decreased expression of the enzyme inducible nitric oxide synthase (iNOS) and its NO product, suggesting that the link between vitamin D and NO in innate immune function is more complex than originally thought (Chang et al., 2004). Irrespective of the antibacterial mechanism that is utilized by animals such as mice, it is generally assumed that vitamin D-mediated induction of these responses will occur via the same intracrine monocyte mechanism that has been described for humans. Although expression of CYP27B1 and 1αhydroxylase activity has been described for murine macrophages *in vitro* (Esteban et al., 2004; Stoffels et al., 2007), the relative importance of this *in vivo* is still unclear. Indeed recent studies using the CYP27B1 KO mouse have suggested that CD8+ cytotoxic T cells are the predominant source of extra-renal 1,25D within the murine immune system (Ooi et al., 2014). Further studies are required to fully clarify the physiological importance of this observation.

Genome-wide analyses and associated *ex vivo* and *in vitro* experiments have clearly demonstrated the potential importance of vitamin D in maintaining optimal innate antibacterial responses in humans. However, these studies have also prompted three further crucially important questions: (1) how important is vitamin D for the adaptive immune system? (2) can vitamin D supplementation *in vivo* enhance these antibacterial responses? (3) what happens to the vitamin D system in human immune diseases? Each of these questions will be considered in the remaining sections of this review.

## **VITAMIN D AND ANTIGEN PRESENTATION**

In the seminal DNA array analysis of monocyte TLR2/1 responses by Liu et al that highlighted induction of *CYP27B1* and *VDR* by *M. tb*, it was notable that DCs did not produce the same response when challenged with 19 kDa lipoprotein, despite expressing TLR2/1 (Liu et al., 2006). Monocytes/macrophages belong to the same hematopoietic lineage as DCs, and both types of cells are able to act as antigen-presenting cells (APCs) to promote T cell and B cells responses. Furthermore like, monocytes, DCs express *VDR* and *CYP27B1*, and exhibit an active intracrine vitamin D system (Brennan et al., 1987; Fritsche et al., 2003; Hewison et al., 2003). However, in contrast to monocytes/macrophages, the primary function on intracrine vitamin D in DCs appears to be as a regulator of cell maturation, and ability of DCs to present antigen to T cells (Hewison et al., 2003). Differentiation of DCs toward a mature APC is associated with increased expression of *CYP27B1* but, paradoxically, a reciprocal decrease in *VDR* (Hewison et al., 2003). It therefore seems likely that DCs will utilize a paracrine vitamin D system, with immature DCs expressing VDR and responding to 1,25D produced by mature DCs with lower VDR expression. Such a mechanism may be biologically advantageous in that it allows some DCs to mature and promote T cell activation as part of normal adaptive immune responses, whilst preventing an over-elaboration of this response that could lead to inflammatory complications. A similar pattern of differential regulation of CYP27B1 and VDR has also been described for monocytes differentiating toward macrophages (Kreutz et al., 1993). The importance of 1,25D as a modulator of DC function is endorsed by studies of *VDR* and *CYP27B1* knockout mice, which present with lymphatic abnormalities consistent with increased numbers of mature DCs (Griffin et al., 2001; Panda et al., 2001) and dysregulated DC trafficking (Enioutina et al., 2009).

In a similar fashion to macrophages, DCs can be divided into distinct sub-types, specifically myeloid DCs (mDCs) and plasmacytoid DCs (pDCs). These cells exhibit different cytokine and chemokine profiles and exert complementary effects on T cells; mDCs are efficient promoters of naïve T cell function (Liu, 2005), whilst pDCs are more closely associated with attenuation of T cell function (Steinman et al., 2003). *In vitro*, 1,25D preferentially regulates mDCs, with associated suppression of naïve T cell activation (Penna et al., 2007). However mDC and pDC express similar levels of VDR, so tolerogenic pDC may also respond to 1,25D, possibly via local, intracrine mechanisms (Penna et al., 2007). Alternatively, 1,25D generated by pDCs may not act to regulate pDC maturation but may, instead, act in a paracrine fashion on VDR-expressing T-cells. The ability of vitamin D to influence the differentiation and function of DCs provides another layer of innate immune function that complements its antibacterial properties. However, this interaction between 1,25D and DC will also have downstream effects on cells that interact with APCs, namely cells from the adaptive immune system.

Consistent with the DNA array analyses that shed light on the antibacterial function of vitamin D in monocytes and macrophages (Liu et al., 2006), genome-wide analysis of DCs has revealed diverse responses to vitamin D in these cells. Proteomic analyses using matrix-assisted laser desorption/ionization (MALDI)-time of flight (TOF)/TOF strategies has defined the key proteins associated with tolerogenic responses to 1,25D (Ferreira et al., 2012). Intriguingly, the dominant effect of 1,25D treatment of monocyte-derived DCs described in this study was the alteration of proteins associated with the cytoskeleton and metabolic function. The induction of cytoskeletal proteins was shown to be consistent with DC responses to other tolerogenic steroid hormones, such as glucococorticoids (Ferreira et al., 2012). However, by contrast, the potent effects of 1,25D on metabolic pathways in DCs appear to be distinct from the effects of glucocorticoids. In particular, 1,25D induced significantly more proteins associated with carbohydrate metabolism, gluconeogenesis and the TCA cycle relative to the glucocorticoid dexamethasone, whilst 1,25D and dexamethasone shared induction of other groups of proteins such as those associated with glycolysis (Ferreira et al., 2012). This particular study also illustrates a key advantage of genome/proteome-wide analyses, which is the ability to group changes in gene/protein expression according to specific properties such as metabolism or cytoskeletal function. Moreover, more recent developments allow researchers to utilize tools such as Ingenuity Pathway Analysis (IPA) or DAVID to cluster altered genes and show protein or gene interaction networks (Hong et al., 2009). These "interactomes" provide a picture of the cooperativity of responses to a particular cell treatment. For example, recent proteomic analysis of responses to the synthetic 1,25D analog TX527 in immature and mature DCs showed that 65–75% of the proteins identified as TX527 responsive made up a statistically significant interactome, with some commonality between the two DC types (Ferreira et al., 2009). In this particular study the authors used multiple sets of data including the Biomolecular Interaction Network Database (BIND) (Bader et al., 2003), and the Molecular Interaction Database (MINT) (Zanzoni et al., 2002) to maximize potential interactions. The increasing availability of gene/protein expression databases for different cell types means that this type of strategy is likely to become a more prominent feature of genome-wide expression analyses in the future.

## **VITAMIN D AND ADAPTIVE IMMUNITY**

As outlined above, one of the initial observations linking vitamin D with the immune system was the presence of VDR in activated lymphocytes (Bhalla et al., 1983; Provvedini et al., 1983). The development of lymphocytes takes place in the thymus with VDR being expressed in medullary thymocytes but not in the less mature cortical thymocytes (Ravid et al., 1984). However, once cells leave the thymus and enter the circulation as T or B cells VDR expression is lost until these cells are activated to proliferate by mitogens (Bhalla et al., 1983; Provvedini et al., 1983). Indeed, 1,25D is a potent inhibitor of T-cell proliferation, blocking the transition from early G1 phase to late G1 phase (Bhalla et al., 1984; Nunn et al., 1986), but having no effect on transition from Go (resting) to early G1 or from late G1 to S phase (Rigby et al., 1985). Studies using T cells isolated from lymphatic tissue have shown that expression of VDR and responsiveness to 1,25D is proportional to the rate of cell proliferation (Karmali et al., 1991). Although these early studies have highlighted a role for 1,25D as a regulator of T and B cell proliferation, it has become increasingly clear that the predominant effects of vitamin D on adaptive immune function involve the modulation of T cell phenotype.

T cells consists of several sub-groups including cytotoxic CD8+ T cells, natural killer cells, γδ T cells, memory cells, CD4+ helper T cells (Th cells), and regulatory T cells (Treg). The best characterized vitamin D responses have been described for Th cells, with 1,25D regulating T cell proliferation and cytokine production (Lemire et al., 1985). Activation of naïve Th cells by antigen and APCs generates pluripotent Th0 cells which can then differentiate into further Th sub-groups based on distinct cytokine profiles. Two of these sub-groups, Th1 (IL-2, IFNγ, tumor necrosis factor alpha) and Th2 (IL-3, IL-4, IL-5, IL-10) T cells, respectively support cell-mediated and humoral immunity (Abbas et al., 1996; Romagnani, 2006). *In vitro* 1,25D inhibits expression of Th1 cytokines (Lemire et al., 1995), whilst promoting Th2 cytokines (Boonstra et al., 2001). More recently, other Th cell sub-groups have been identified, including interleukin-17 (IL-17)-secreting T-cells (Th17 cells) and these cells are also targets for vitamin D. In the autoimmune disease-susceptible nonobese diabetic (NOD) mouse treatment with 1,25D decreased expression of IL-17 (Penna et al., 2006). In a similar fashion, 1,25D suppression of murine retinal autoimmunity involves inhibition of Th17 activity (Tang et al., 2009).

In addition to its effects on Th cells, vitamin D may also act on CD8+ cytotoxic T cells which express relatively high levels of VDR (Rigby et al., 1987; Provvedini and Manolagas, 1989; Veldman et al., 2000). As outlined above, CD8+ cells in mice have also been reported to express the vitamin D-activating enzyme 1αhydroxylase (Ooi et al., 2014). CD8+ T cells are known to be involved in autoimmune disease such as multiple sclerosis (MS) (Babbe et al., 2000), but do not mediate the effects of 1,25D in suppressing the murine form of MS, experimental autoimmune encepholmyelitis (EAE) (Meehan and DeLuca, 2002). More recent studies have reported a link between vitamin D and a variant of CD8+ T cells, CD8αα cells. Unlike CD8+ T cells, CD8αα cells are not cytotoxic and may play a role in suppressing gastrointestinal inflammation (Cheroutre and Lambolez, 2008). VDR knockout mice exhibit decreased numbers of CD8αα cells (Yu et al., 2008), due to decreased T cell expression of the chemokine receptor CCR9 preventing T cell homing to the gastrointestinal tract. T cell homing defects provides a potential explanation for the increased colonic inflammation observed in VDR knockout mice when crossed with colitis disease-susceptible mice (Froicu et al., 2003). Vitamin D metabolites may also influence T cell homing in other tissues. In the skin, 1,25D stimulates expression of the chemokine receptor 10 (CCR10) which recognizes the chemokine CCL27 secreted by keratinocytes (Sigmundsdottir et al., 2007).

As well as acting as a modulator of Th cell phenotype and function, vitamin D can also influence adaptive immunity by promoting suppressor T cells known as regulatory T cells (Treg) (Barrat et al., 2002). The precise mechanism by which vitamin D regulates Tregs is still somewhat controversial. Initial studies *in vitro* suggested that the ability of 1,25D to promote CD4+ CD25+ Treg was due to indirect effects on antigen-presenting DCs, specifically suppression of DC maturation and increased expression of DC cytokines such as CCL22 (Penna et al., 2007). However, subsequent studies have also described direct effects of 1,25D on T cells to generate CTLA4-positive Treg (Jeffery et al., 2009). Significantly, these studies were focused on the use of active 1,25D as the immunomodulator, and it is only in more recent studies that the role of pro-hormone 25D in Treg development has been investigated (Jeffery et al., 2012). Data from this study demonstrated the ability of 25D to promote the generation of Treg through intracrine/paracrine effects on CYP27B1/VDRexpressing DCs. Notably, this report also highlighted the impact of vitamin D binding protein (DBP) on DC responses to 25D, and concluded that non-DBP-bound (free) 25D is the form of 25D that is biologically active for generation of Tregs (Jeffery et al., 2012). The importance of Treg as a facet of vitamin D immunomodulation is illustrated by various studies *in vivo*. In patients with CKD, systemic administration of 1,25D has been shown to increase numbers of circulating Treg (Ardalan et al., 2007). Conversely, in patients with MS, serum concentrations of 25D correlate with Treg activity (Royal et al., 2009; Smolders et al., 2009), underlining the importance of intracrine pathways in mediating effects of vitamin D on adaptive, as well as innate immunity. In mice, topical application of 1,25D (Gorman et al., 2007) or its synthetic analog calcipotriol (Ghoreishi et al., 2009) have been shown to increase numbers of Treg.

The effects of vitamin D on adaptive immunity have to date been very much focused on its ability to modulate T cell proliferation and phenotype. Nevertheless, early studies reported that 1,25D could also suppress the development of immunoglobulin (Ig)-secreting B cells following mitogenic stimulation (Shiozawa et al., 1985; Iho et al., 1986). Initial experiments suggested that the most likely mechanism for this was an indirect effect through inhibition of Th cells (Lemire et al., 1985), but more recent work has shown that 1,25D can suppress the differentiation of two types of B cell, plasma cells and class-switched memory cells, through apparent direct effects (Chen et al., 2007). Other reports have shown that 1,25D can regulate B cell IL-10 (Heine et al., 2008) and CCR10 (Shirakawa et al., 2008), suggesting that the effects of 1,25D on these cells is not restricted to their capacity to produce immunoglobulin.

Although genome-wide screening has played a pivotal role in identifying pivotal mechanisms for the interaction between vitamin D and innate immunity, the same cannot be said for vitamin D and adaptive immunity, where genome-wide analyses have complemented an already well-established field of research. Nevertheless, it is interesting to note reports where this strategy has been applied. In some cases these analyses have revealed a role for the vitamin D system, similar to the seminal studies of *M. tb* induction of CYP27B1 and VDR. For example, transcriptional profiling of γδT cells reported induction of VDR following activation of these cells with non-peptidic monoalkyl phosphate ligands (Chen et al., 2005). This small sub-set of T cells plays an important role in inflammatory diseases, and it was therefore speculated that 1,25D may act to suppress these cells as part of a more generalized anti-inflammatory response. Further array analyses have also identified VDR as one of a discrete number of genes involved in the formation of B cell germinal centers (Nakayama et al., 2006).

Array analyses have also been used to characterize the gene regulatory effects associated with immunomodulatory responses to 1,25D. These studies have focused primarily on the effects of 1,25D and its synthetic analogs on DCs, with results underlining the ability of 1,25D to promote decreased antigen presentation and a tolerogenic phenotype in these cells (Griffin et al., 2004; Shen and Zheng, 2004; Pedersen et al., 2009; Szeles et al., 2009). Notably, one of these array studies showed that the effects of 1,25D on DC gene expression were independent of DC differentiation status, suggesting a specific role for 1,25D as a regulator of DC function (Szeles et al., 2009). This particular study also reported that key changes in DC gene expression could be achieved using either 1,25D or 25D, further emphasizing the functional importance of the intracrine vitamin D system in these cells. DNA microarray analyses have also been used to assess 1,25D-mediated regulation of gene expression in CD4+ Th cells following activation of these cells by phorbol myristate acetate and a calcium ionophore to induce VDR (Mahon et al., 2003). The diverse array of gene targets regulated by 1,25D in this particular array analysis suggests that 1,25D can influence Th cells both directly, as well as via effects on antigen-presenting DCs.

## **VITAMIN D STATUS AND IMMUNE FUNCTION**

It is important to recognize that most of the genome-wide analyses that have explored the immunomodulatory effects of vitamin D *in vitro* have focused on treatments using active 1,25D or one of its synthetic analogs. However, as outlined above, pathogen-induction of an intracrine system in cells such as monocytes/macrophages strongly suggests that regulation of immunity *in vivo* is independent of endocrine, systemic 1,25D. Instead it is likely to be primarily driven by local activation of 25D, the major circulating form of vitamin D and determinant of vitamin D status in any given individual. Thus, it is not surprising that translational studies have focused on the relationship between serum 25D and human immune function, including effects on both innate and adaptive immunity.

Epidemiology has shown that vitamin D-insufficiency (serum 25D *<*30 ng/ml) is associated with increased risk of TB (Wilkinson et al., 2000; Ustianowski et al., 2005; Williams et al., 2008; Wejse et al., 2009). Several clinical trials of vitamin supplementation, as an adjunct to conventional antibiotic therapy, have also been reported with varying success (Nursyam et al., 2006; Martineau et al., 2007; Wejse et al., 2009). Supplementation using 4 × 100,000 IU vitamin D was successful in raising serum concentrations of 25D in TB patients, but this resulted in no overall improvement in sputum conversion time between vitamin D- and placebo-treated patients (Martineau et al., 2011). However, improved sputum conversion time was observed in a specific subset of TB patients with a *Taq1* single nucleotide polymorphism (SNP) within the VDR gene (Martineau et al., 2011), suggesting that genetic factors may influence immune responses to vitamin D supplementation. In a follow-up report to this TB vitamin D supplementation trial, it was shown that raised serum 25D was associated with improved resolution of TB disease (Coussens et al., 2012). Thus, in situations where infectious disease has already become established, it is possible that the role of 25D is primarily focused on anti-inflammatory adaptive immune responses. The link between vitamin D and infection is not restricted to TB. In patients with sepsis, circulating 25D levels have been shown to correlate directly with serum concentrations of CAMP, and inversely with critical illness in these patients (Jeng et al., 2009). Low serum 25D has also been linked to upper respiratory infections such as influenza (Cannell et al., 2006), and in patients with CKD low serum is associated with increased risk of infection and mortality (Gombart et al., 2009a).

To date, the application of genome-wide analyses to further elucidate the impact of serum vitamin D (25D) status on immune function has been limited. In a recent study by Holick and colleagues, array analysis of gene expression in peripheral blood mononuclear cells from vitamin D-sufficient (serum 25D *>*20 ng/ml, *n* = 4 subjects) and vitamin D-deficient (serum 25D *<*20 ng/ml, *n* = 4 subjects) revealed 66 differentially expressed genes (*>*1.5-fold change, *p <* 0*.*01) (Hossein-nezhad et al., 2013). However, after vitamin D supplementation (2000 IU vitamin D/day for 2 months), there was no significant change in the expression of these genes, even though serum 25D levels were increased in both sufficient and deficient subjects (Hosseinnezhad et al., 2013). Nevertheless, 291 additional genes were found to be differentially expressed in peripheral blood mononuclear cells following vitamin D supplementation (*>*1.5-fold, *p <* 0*.*01) (Hossein-nezhad et al., 2013). Similar array analyses carried out by our group using peripheral blood mononuclear cells from elderly vitamin D-deficient (18.8 ± 0.6 ng/ml serum 25D) and vitamin D–sufficient (58 ± 1.7 ng/ml serum 25D) patients, revealed 30 differentially regulated genes (**Figure 2**). These variations in gene expression occurred against a backdrop of no difference in serum 1,25D concentrations between vitamin D-sufficient and -deficient groups, underlining the importance of 25D, and the intracrine vitamin D system as regulators of immune cell

function. In both this study and the Holick report, array analyses were carried out using mixed populations of systemic immune cells including both innate immunity APCs and lymphocytes of the adaptive immune system. The array analyses will therefore encapsulate both intracrine and paracrine activity of 25D, but will also reflect inherent donor to donor variations in immune cell composition.

Genome-wide analyses of immune responses to altered vitamin D status in mice are also very limited. Data from our group using colon tissue from vitamin D-deficient (serum 25D = 2.5 ± 0.1 ng/ml) vs. vitamin D-sufficient (serum 25D = 24.4 ± 1.8 ng/ml) identified 31 genes that were differentially expressed *>*2-fold (*p <* 0*.*01) (Lagishetty et al., 2010). Amongst these, vitamin D-deficient mice showed decreased expression of angiogenin-4 (Ang4), an antimicrobial protein which acts to minimize tissue invasion by enteric bacteria (Hooper et al., 2003). Further studies showed that decreased Ang4 in vitamin D-deficient mice was associated with increased levels of bacteria in the colon epithelium, consistent with compromised innate immune surveillance. Given that dysregulation of innate immune responses to enteric bacteria has been linked to the initiation of tissue inflammation associated with some types of inflammatory bowel disease (Packey and Sartor, 2009), it is possible that vitamin D plays a role in protecting against this disease via the induction of antibacterial Ang4.

Another genome-wide strategy with implications for vitamin D and the immune system, arose from studies aimed at determining the genetic component of vitamin D-deficiency. A recent Genome-Wide Association Study of almost 34,000 individuals showed that SNPs within the *DBP* gene are a key inherited determinant of low vitamin D status (serum 25D *<*75 nM or 30 ng/ml). Gene variations in *DBP* appear to act by influencing the serum concentrations of DBP protein (Lauridsen et al., 2001) which are known to be linked to serum levels of total 25D and 1,25D (Lauridsen et al., 2005; Wang et al., 2010a). Studies of other *DBP* SNPs suggest that genetic variants of *DBP* are linked to different binding affinities of 25D for DBP protein (Arnaud and Constans, 1993). Both the concentration and binding affinity of DBP protein are important for the serum transport of vitamin D metabolites (notably 25D which binds to DBP with a higher affinity than 1,25D). However, DBP concentration and affinity also define the amount of 25D that is *not* bound to DBP. This "free" or "bioavailable" fraction of circulating 25D appears to be the form that accesses target cells such as monocytes (Chun et al., 2012), presumably via passive diffusion of lipid-soluble 25D through cell membranes—the so-called "free hormone hypothesis." Studies by our group have shown that antibacterial responses to 25D *in vitro* are more pronounced with low affinity forms of DBP that support higher levels of free 25D (Chun et al., 2010). Studies to date have been based on mathematical estimations of free 25D from total serum concentrations of 25D and DBP (Chun et al., 2012). However, future strategies using actual physical measurement of free 25D will greatly help to clarify the precise importance of total vs. free 25D in determining immune responses to vitamin D.

## **IMMUNE DISEASE AND THE DYSREGULATION OF VITAMIN D**

Genome-wide strategies have played a pivotal role in elucidating the core mechanisms that trigger the intracrine vitamin D system and associated immune responses in cells such as monocytes/macrophages and DCs. Whether these studies have been carried out using freshly isolated preparations of immune cells, or cultured immune cells the resulting data have reflected the potential vitamin D-mediated responses that may occur following a pathogen challenge. What is less clear is how these responses function under conditions of actual human immune disease. An illuminating example of this strategy is provided by the disease leprosy which, like TB, involves a mycobacterial infection—in this case *Mycobacterium leprae* (*M. lep*) or *Mycobacterium lepromatis. S*imilar to TB, vitamin D was at one time considered to be a putative therapy for leprosy (Herrera, 1949). However, unlike TB, leprosy can be divided into different disease sub-types, notably tuberculoid leprosy (T-lep) and lepromatous leprosy (L-lep). These two forms of leprosy have very different immune profiles and prognoses (Britton and Lockwood, 2004). DNA array analyses to define the gene expression profiles associated with the T-lep and L-lep forms of leprosy, highlighted elevated expression of CYP27B1, CYP24A1, and VDR in T-lep vs. L-lep lesions (Montoya et al., 2009). The over-arching conclusion from these studies is that the less aggressive form of leprosy, T-lep, is manifested by an intact vitamin D intracrine system that is able to support antibacterial responses to vitamin D. By contrast, L-lep, which is characterized by a high level of macrophage *M. lep* infection, and has a poor prognosis, exhibits an impaired vitamin D intracrine system. Thus, for patients with L-lep, successful elevation of serum 25D concentrations may be less effective in promoting intracrine-mediated regulation of antibacterial responses.

Several questions have arisen from the studies of vitamin D and leprosy. The first concerns the mechanism by which the vitamin D intracrine system is corrupted in L-lep patients. One possibility is that the T cell cytokine profiles that are characteristic of L-lep (e.g., increased IL-4, IL-10, and IFNα/β) exert a detrimental adjunct effect on the underlying TLR2/1-induced intracrine vitamin D system. At the same time, cytokine profiles associated with T-lep (e.g., increased IFNγ) may have more beneficial adjunct effects. Subsequent experiments *in vitro* support this hypothesis, with the Th1 cytokine IFNγ enhancing TLR2/1-induced vitamin D-activation and associated antibacterial activity (Edfeldt et al., 2010; Fabri et al., 2011). Conversely, the L-lep cytokines IL-4 (Edfeldt et al., 2010), IFNβ (Teles et al., 2013), and IL-10 (Teles et al., 2013) suppress antibacterial production (**Figure 3**). The effect of IFNβ appears to be mediated via IL-10 which acts to suppress expression of CYP27B1 (Teles et al., 2013), whereas IL-4 appears to act by stimulating activity of the vitamin D catabolic enzyme CYP24A1 (Edfeldt et al., 2010). The collective conclusion from these studies is that specific human diseases are characterized by T cell cytokines that act to either promote or corrupt the underlying pathogen-PRR-driven vitamin D intracrine system. Cytokine profiles such as this have also been described for active and inactive TB (Berry et al., 2010; Maertzdorf et al., 2012), providing an additional perspective on the varying success of vitamin D supplementation trials with this disease. Disease itself may therefore play a fundamental role in determining the efficacy of immunomodulatory vitamin D for any given patient; for example, it is possible that for diseases such as L-lep, higher levels of serum 25D will be required to achieve a specific antibacterial response. This is clearly an important topic for future research.

The differential regulation of the intracrine vitamin D pathway in T-lep and L-lep has also provided a platform for genomewide analyses aimed at identifying factors other than T cell cytokines that may be involved in corrupting monocyte vitamin D responses. In a follow-up to the previous DNA array studies for

**FIGURE 3 | Mechanisms for corruption of vitamin D-mediated antibacterial responses in monocytes.** Schematic representation of monocyte/macrophage responses to infection with a pathogen such as *Mycobacterium tuberculosis* (*M. tb*), and associated adjunct signals. Positive effects on the vitamin D intracrine system include: induction of CYP27B1 expression by cytokines such as interferon γ (IFNγ) and interleukin-15 (IL-15). Negative effects include suppression of CYP27B1 by IL-10 and IFNβ, and induction of 24-hydroxylase activity by IL-4. Fibroblast growth factor 23 (FGF23) also suppresses CYP27B1 and microRNA-21 (miR-21) associated with some pathogen infections (e.g., *M. leprae*) can suppress expression of CYP27B1 by degradation of RNA and/or translation. The serum vitamin D binding protein (DBP) may also attenuate intracrine vitamin D by restricting monocyte bioavailability of free 25D.

leprosy, further array analysis of T-lep and L-lep tissues revealed distinct patterns of microRNA (miRNA) expression in these tissues (Liu et al., 2012). Recent studies have shown that miRNAs play a key role in fine-tuning gene expression by interacting with RNA to silence gene expression either by degrading transcripts, or by blocking their translation (Ketting, 2011). In L-lep, 16 miRNAs were found to be differentially induced relative to T-lep tissue, with miRNA-21 (miR-21) being the most prominent of these (Liu et al., 2012). In the context of innate immunity and leprosy, miR-21 may target several important mechanisms, including suppression of IL-1 expression which may, in turn, attenuate intracrine induction of antibacterial DEFB4 (see **Figure 3**). However, importantly, miR-21 is also predicted to interact with *CYP27B1* mRNA and suppress activity of 1αhydroxylase and decrease localized synthesis of 1,25D. *In vitro*, siRNA knockdown of miR-21 in *M. lep* infected monocytes, restored CYP27B1 expression and 25D-mediated antibacterial responses (Liu et al., 2012). Despite these observations, relatively little is known about how miRNAs corrupt vitamin D signaling in disease situations. Based on genome-wide *in silico* analysis of miRNA target sequences, multiple miRNAs are predicted to influence the expression of proteins associated with vitamin D metabolism and signaling (reviewed in Lisse et al., 2013a). However, other than studies of miR-21, there have been few studies to validate the predicted effects of miRNAs on the vitamin D system. Analysis of ovarian granulosa and breast cancer cells has demonstrated increased expression of miR-125B in these tissue, and a concomitant dysregulation of two of its targets, *VDR* and *CYP24A1* mRNAs (Mohri et al., 2009). It seems likely that future studies will identify other miRNAs that modulate the vitamin D intracrine system in immune cells under disease conditions. Moreover, it is important to recognize that vitamin D itself is a potent regulator of miRNAs. To date, these studies have focused on cancer (Wang et al., 2009, 2011), and bone cells (Lisse et al., 2013b), but similar future studies of 1,25D-regulated non-coding RNAs in immune cells may provide an entirely new perspective on the immunomodulatory actions of vitamin D.

Some important questions about human disease and the immunomodulatory effects of vitamin D remain unanswered. For example, it is still not clear why there is aberrant synthesis of 1,25D by macrophages in granulomatous diseases (Kallas et al., 2010). It is also unclear what effect, if any, viral pathogens such as hepatitis C or HIV have on innate and adaptive immune actions of vitamin D, although HIV infection of some cells has been shown to suppress expression of *VDR* (Chandel et al., 2013). Future studies of vitamin D and the immune system may also explore non-traditional targets for immune regulation. For example, patients with end-stage kidney disease who routinely use dialysis are at high risk of infection and associated mortalities. These patients are also commonly vitamin D-deficient (Zehnder et al., 2007), and this may impair normal innate immune responses to infection. However, as with TB and leprosy, additional disease factors may act to further compromise the intracrine vitamin D system in these patients. Notably, circulating levels of fibroblast growth factor 23 (FGF23), which plays a key role in the endocrine regulation of phosphate homeostasis, are elevated very early in kidney disease (Danziger, 2008; Isakova et al., 2011). One of the important actions of FGF23 is to suppress renal production of 1,25D through the suppression of CYP27B1 expression (Shimada et al., 2004); in this way FGF23 acts as a counterpoint to parathyroid hormone which stimulates CYP27B1 and renal 1,25D production. Until recently, the effects of FGF23 were thought to be restricted to the mineral homeostasis endocrine system. However, work by our group has shown that FGF23 can also act on monocytes to suppress expression of CYP27B1 and the intracrine induction of antibacterial proteins (Bacchetta et al., 2013a). These data highlight a mechanism by which renal disease may compromise vitamin D-mediated immune function, similar to that observed for cytokines associated with infectious disease (see **Figure 3**). As well as providing an explanation for the increased risk of infection in kidney disease patients, these results also suggest a hitherto unrecognized link between the vitamin D endocrine system and its intracrine immune counterpart.

## **PERSPECTIVES**

For many years, the link between vitamin D and the immune system was considered to be a non-classical response with only a pathophysiological relevance. The advent of genome-wide analyses has enabled a complete change in this perspective by providing an unbiased picture of how the vitamin D system is induced by pathogens, and how the resulting intracrine cellular machinery can promote both innate and adaptive immune responses to the pathogen. A key challenge going forward will be to relate these mechanisms to patient vitamin D status, and this is likely to herald a new wave of genome-wide analyses linked to placebocontrolled vitamin D supplementation trials. Interpretation of these studies is likely to be complex. Recent genome-wide analysis of patient tissues has shown that some immune diseases are characterized by corruption of the vitamin D system, so that conventional notions of vitamin D-sufficiency and vitamin D-deficiency may be very different for patients with specific diseases. A key objective for future studies will be to determine whether vitamin D-mediated-immune function is also applicable to mouse models, where genome-wide screening will help to identify immune targets that are related to, or distinct from, human data. Future studies will also need to better characterize disease corruption of vitamin D responses. Screening for disease-specific microRNAs will be particularly important to identify non-coding RNAs that can target components of the intracrine vitamin D. The current shift away from DNA array technology to RNAseq strategies will help to achieve these new objectives in a single genome-wide screen, at increasingly affordable prices.

### **ACKNOWLEDGMENTS**

The authors acknowledge support from NIH grants DK0911672 and AR063910 (Martin Hewison), UL1TR000124 (Martin Hewison, Robert L. Modlin, John S. Adams), and AI073539 (Robert L. Modlin).

### **REFERENCES**

Abbas, A. K., Murphy, K. M., and Sher, A. (1996). Functional diversity of helper T lymphocytes. *Nature* 383, 787–793. doi: 10.1038/383787a0

Adams, J., Gacad, M. A., Baker, A. J., Gonzales, B., and Rude, R. K. (1985). Serum concentrations of 1,25-dihydroxyvitamin D3 in platyrrhini and catarrhini:a phylogenetic appraisal. *Am. J. Primatol.* 28, 263–270.


association studies. *Hum. Genet.* 126, 289–301. doi: 10.1007/s00439-009- 0676-z


to differentiate by 1,25-dihydroxyvitamin D3. *Cell Cycle* 8, 736–741. doi: 10.4161/cc.8.5.7870


Zehnder, D., Landray, M. J., Wheeler, D. C., Fraser, W., Blackwell, L., Nuttall, S., et al. (2007). Cross-sectional analysis of abnormalities of mineral homeostasis, vitamin D and parathyroid hormone in a cohort of pre-dialysis patients. The chronic renal impairment in Birmingham (CRIB) study. *Nephron Clin. Pract.* 107, c109–c116. doi: 10.1159/000108652

**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 January 2014; accepted: 30 March 2014; published online: 21 April 2014. Citation: Chun RF, Liu PT, Modlin RL, Adams JS and Hewison M (2014) Impact of vitamin D on immune function: lessons learned from genome-wide analysis. Front. Physiol. 5:151. doi: 10.3389/fphys.2014.00151*

*This article was submitted to Integrative Physiology, a section of the journal Frontiers in Physiology.*

*Copyright © 2014 Chun, Liu, Modlin, Adams and Hewison. 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.*

## Vitamin D and gene networks in human osteoblasts

## *Jeroen van de Peppel and Johannes P. T. M. van Leeuwen\**

*Department of Internal Medicine, Bone and Calcium Metabolism, Erasmus MC, Rotterdam, Netherlands*

#### *Edited by:*

*Carsten Carlberg, University of Eastern Finland, Finland*

#### *Reviewed by:*

*Martin Hewison, University of California Los Angeles, USA Lieve Verlinden, KU Leuven, Belgium*

#### *\*Correspondence:*

*Johannes P. T. M. van Leeuwen, Department of Internal Medicine, Bone and Calcium Metabolism, Erasmus MC, Room Ee585d, Dr. Molewaterplein 50, 3015 GE Rotterdam, Netherlands e-mail: j.vanleeuwen@erasmusmc.nl* Bone formation is indirectly influenced by 1,25-dihydroxyvitamin D3 (1,25D3) through the stimulation of calcium uptake in the intestine and re-absorption in the kidneys. Direct effects on osteoblasts and bone formation have also been established. The vitamin D receptor (VDR) is expressed in osteoblasts and 1,25D3 modifies gene expression of various osteoblast differentiation and mineralization-related genes, such as alkaline phosphatase (ALPL), osteocalcin (BGLAP), and osteopontin (SPP1). 1,25D3 is known to stimulate mineralization of human osteoblasts *in vitro*, and recently it was shown that 1,25D3 induces mineralization via effects in the period preceding mineralization during the pre-mineralization period. For a full understanding of the action of 1,25D3 in osteoblasts it is important to get an integrated network view of the 1,25D3-regulated genes during osteoblast differentiation and mineralization. The current data will be presented and discussed alluding to future studies to fully delineate the 1,25D3 action in osteoblast. Describing and understanding the vitamin D regulatory networks and identifying the dominant players in these networks may help develop novel (personalized) vitamin D-based treatments. The following topics will be discussed in this overview: (1) Bone metabolism and osteoblasts, (2) Vitamin D, bone metabolism and osteoblast function, (3) Vitamin D induced transcriptional networks in the context of osteoblast differentiation and bone formation.

**Keywords: vitamin D, osteoblast, differentiation, mineralization, autocrine/paracrine mechanisms, immune system**

## **BONE METABOLISM AND OSTEOBLASTS**

Bone is formed during fetal development by two processes: endochondral and intramembranous ossification (for review Bilezikian et al., 2002). Skull and flat bones are formed by intramembranous ossification where there is direct bone formation by condensation of the mesenchyme without a preformed cartilaginous scaffold. Long bones and most of the remaining bones are formed by endochondral ossification (Mackie et al., 2008, 2011; Nishimura et al., 2012). This type of bone formation is characterized by the transition of cartilage into mineralized bone tissue.

Two major processes occur in bone: bone modeling and bone remodeling. While bone modeling drives the growth of the skeleton, bone remodeling is responsible for the maintenance of healthy bone in the adulthood (Teti, 2011). Bone remodeling takes place throughout live and maintains the structural integrity and strength of the bone by removing old or damaged bone and replacing it by new, strong bone. Remodeling is a local process that can take place anywhere on the bone surface throughout the lifespan of a bone. Remodeling occurs in a temporary anatomic unit of osteoclasts and osteoblasts called a bone multicellular unit (BMU) (Martinello et al., 2012; Sims and Martin, 2014). The BMU is a sealed compartment in which bone resorption and subsequent formation are regulated. This coupled resorption and formation characterizes and differentiates bone remodeling from bone modeling, in which bone resorption and formation do not have to occur at the same time and site. For growth and for the maintenance of healthy bone, multiple cell types are of importance: mesenchymal stem cells (MSC), osteoblasts, osteocytes, and hematopoietic stem cells and osteoclasts. The osteoblasts play a pivotal role in bone metabolism by forming bone but also by controlling and regulating the formation and activity of the bone resorbing cell the osteoclast.

Osteoblasts originate from MSC. MSCs are located in the bone marrow but also in almost all other tissues undergoing continuous tissue homeostasis. MSCs can differentiate into osteoblasts, chondrocytes, fibroblasts, adipocytes or myocytes (Friedenstein et al., 1974; Minguell et al., 2001; Yin, 2006). During osteoblast differentiation several functional phases can be identified: proliferation, production and maturation of extracellular matrix (ECM) and ECM mineralization (Owen et al., 1991). Osteoblast differentiation can start by a trigger of certain growth factors (Wang, 1993) as well as hormones and other factors (Eijken et al., 2006). Mature osteoblasts produce and secrete ECM molecules (Owen et al., 1991). Osteoblasts synthesize the most abundant bone ECM protein collagen type I but also a broad range of non-collagenous ECM proteins. Mineralization of the ECM is likely induced by matrix vesicles which derive from osteoblasts (Anderson et al., 2005). When mature osteoblasts initiate mineralization of mature ECM, its fate may vary. Osteoblasts can further differentiate into osteocytes, become a bone lining cell or undergo apoptosis (Jilka et al., 1998; Weinstein et al., 1998). Osteoblasts become osteocytes by being entrapped in self-produced ECM, in which they may survive for decades. Osteocytes form a star-shaped network of cytoplasmic extensions. Osteocytes are thought to function as orchestrators of bone by sensing and communicating mechanical stress (i.e., bone damage) via these extensions (Bonewald, 2011). It has become evident by genetic approaches that osteocytes play a role in regulation of bone turn-over (resorption and formation) (Nakashima et al., 2011; Atkins and Findlay, 2012). Bone lining cells are less well understood. They are covering the bone surface and prevent it from being in direct contact with the bone marrow. It has been reported that these cells "clean" resorption pits after osteoclasts retreated (Everts et al., 2002). Bone lining cells are considered as inactive osteoblasts. It has been suggested that these cells can be activated to become osteoblasts (Dobnig and Turner, 1995; Chow et al., 1998) but they also may represent the osteoblastic part of the stem cell niche and interact with the hematopoietic stem cells.

The osteoblasts/osteocytes guarantee the close coupling between bone formation and resorption in healthy bone remodeling. Osteoblasts and osteocytes produce the soluble osteoclast stimulating factors RANKL and M-CSF which upon binding to its receptors (RANK and c-Fms, respectively) induce differentiation of osteoclast progenitors and fusion of mononuclear cells into multinucleated tartrate-resistant acid phosphatase positive osteoclasts (Boyle et al., 2003). Besides RANKL, osteoblasts also produce a soluble decoy-receptor: osteoprotegerin (OPG). OPG binds RANKL with high affinity leading to inhibition of osteoclast stimulation and thus leading to less bone resorption (Lacey et al., 1998; Kostenuik and Shalhoub, 2001).

## **VITAMIN D, BONE METABOLISM, AND OSTEOBLAST**

The biologically most active form of vitamin D, 1α,25 dihydroxyvitamin D3 [calcitriol or 1,25(OH)2D3 (1,25D3)] is formed by a stepwise process starting in the skin and involving the liver and kidney. Upon ultraviolet B exposure, 7-dehydrocholesterol (pro-vitamin D3) is transformed into (pre)vitamin D3 (cholecalciferol) in the skin. Subsequent hydroxylation at the C25 and 1α position in liver and kidney, respectively, produce 1,25D3 (Holick, 1995). In bone diseases, vitamin D is used as an anti-rickets agent (Kitanaka et al., 1998; McCollum et al., 2002; Tatsumi et al., 2007), which improves bone mineralization and is often prescribed in combination with other osteoporosis drugs to secure a positive calcium balance. However, a recent metaanalyes by Reid et al suggests that the supplementation of vitamin D for the prevention of osteoporosis is inappropriate without specific risk factors for vitamin D deficiency (Reid et al., 2014).

Up to today it is still in debate whether 1,25D3 effects on bone formation are indirect via intestinal and renal regulation of calcium levels or also via a direct effect on osteoblasts. It has been demonstrated that mice lacking the vitamin D receptor (VDR) gene display retarded growth, severe bone impairment, immune abnormalities, and premature death at only 15 weeks of age due to hypocalcemia (Li et al., 1997; Yoshizawa et al., 1997; Mathieu et al., 2001). A rescue diet restored all pathological effects suggesting that as long as calcium homeostasis is under control, bone itself does not seem to be affected by impaired VDR signaling. The importance of physiological 1,25D3 levels for bone is demonstrated by the mutation of the CYP27B1 gene. Subjects with a mutation in that gene develop vitamin-D-dependent rickets (Li et al., 1997). In a mouse model for rickets, greater extensibility and lower stiffness of fibrils resulted from a decreased grade of mineral deposition (Karunaratne et al., 2012). This further supports the importance of an optimal grade of mineralization for healthy bones (Kitanaka et al., 1998) and points to a role for 1,25D3 herein.

A direct positive effect an 1,25D3 analog on bone formation in ovariectomized rats with only slight changes in serum calcium points to the existence of a direct effect on bone formation (Shevde et al., 2002). This is supported by *in vitro* studies demonstrating direct effects on osteoblasts. The VDR is present in osteoblasts and its expression can be regulated by 1,25D3 itself and by other factors such as parathyroid hormone (PTH), glucocorticoids, transforming growth factor-β, and epidermal growth factor (Pols et al., 1988a,b; Reinhardt and Horst, 1990; van Leeuwen et al., 1991, 1992a,b; Godschalk et al., 1992). The expression of VDR allows 1,25D3 to directly affect osteoblast growth and differentiation. 1,25D3 has been shown to stimulate bone formation and mineralization in all studies using human osteoblasts and stimulate osteogenic differentiation from human mesenchymal stem/stromal cells (MSC) (Ueno et al., 1992; Prince et al., 2001; Jørgensen et al., 2004; Van Driel et al., 2006a,b; Zhou et al., 2006, 2012). 1,25D3 enhanced mineralization by effects on human osteoblasts prior to the onset of mineralization (Woeckel et al., 2010). Thus, 1,25D3 is not directly involved in the process of mineral deposition but more likely in a process preparing the environment/ECM for mineralization. 1,25D3 regulates the osteoblast differentiation marker ALPL and various bone ECM proteins such as COL1A1. Procollagen type I by human osteoblasts was stimulated (Franceschi et al., 1988; Hicok et al., 1998) as well as unaffected (Ingram et al., 1994; Hicok et al., 1998; Siggelkow et al., 1999) by vitamin D. However, gene expression profiling studies demonstrated that the 1,25D3 effect in the pre-mineralization phase is not likely primarily due to changes in expression of ECM proteins and thereby composition of the ECM (Woeckel et al., 2010). Production of alkaline phosphatase (ALPL) positive matrix vesicles was significantly induced by 1,25D3 in this period of osteoblast differentiation (Anderson, 1995) providing a means to enhance mineralization (Woeckel et al., 2010). In addition, previous studies have shown the importance of other factors like TGFβ, IGF-I, bone morphogenetic protein, interferon, PTH, hepatocyte growth factor, epidermal growth factor, and peroxisome proliferator-activated receptor ligands and Wnt signaling for the eventual effect of 1,25D3 on osteoblasts (Petkovich et al., 1987; Pols et al., 1988b; Scharla et al., 1991; Bonewald et al., 1992; Godschalk et al., 1992; van Leeuwen et al., 1992a,b; Ingram et al., 1994; Staal et al., 1994, 1996, 1998; Haussler et al., 1998; Yanagisawa et al., 1999; Sammons et al., 2004; Yarram et al., 2004; Fretz et al., 2007; Chen et al., 2012a, 2013; Woeckel et al., 2012; Yamaguchi and Weitzmann, 2012). These data stress the importance of studying and interpreting the effects of 1,25D3 on bone in a systems biological approach encompassing the different layers of regulation and interactions.

In contrast to human and rat studies, 1,25D3 inhibits differentiation and mineralization in cultures of murine osteoblasts (Shi et al., 2007; Chen et al., 2012a,b, 2013) and murine VDR deficient osteoblasts have increased osteogenic potential (Sooy et al., 2004). 1,25D3 increases in a VDR-dependent manner the expression of progressive ankylosis (ANK) and ectonucleotide pyrophosphatase phosphodiesterase (ENPP1) in murine osteoblasts. This leads to an increase in the level of pyrophosphate (PPi) that inhibits mineralization (Lieben et al., 2012). 1,25D3 also increases osteopontin shown to inhibit mineralization (Staal et al., 1996). However, transgenic murine models with osteoblast-specific VDR overexpression show increased bone formation and mineralization (Gardiner et al., 2000; Misof et al., 2003; Xue et al., 2006). An 1,25D3 analog had a positive effect on bone nodule formation and mineralization in murine calvarial osteoblast cultures of wild type but not VDR null mice (Shevde et al., 2002) while one study showed increased mineralization in MC3T3 cell cultures (Matsumoto et al., 1991). In a recent study, Yamamoto et al. (2013) illustrated that mice lacking VDR in osteoblasts had an increased bone mass, due to decreased bone resorption.

Overall the present data show variation in effects of 1,25D3 on differentiation and mineralization with overall stimulatory effects in human and rat osteoblasts while overall an inhibitory effect in murine osteoblasts (Van Driel et al., 2006a). Following this, 1,25D3 has been shown to increase RUNX2 expression in human osteoblasts (Prince et al., 2001; Viereck et al., 2002; Maehata et al., 2006) while 1,25D3 suppresses RUNX2 promoter and reduces RUNX2 expression in murine osteoblasts (Prince et al., 2001; Drissi et al., 2002). Osteocalcin (BGLAP) is an interesting gene considering differences in 1,25D3 effects in human and murine osteoblasts (Thomas, 2000). 1,25D3 stimulates BGLAP expression in human and rat osteoblasts while it inhibits BGLAP expression in murine osteoblasts (Lian et al., 1997; Zhang et al., 1997), supporting differences between human/rat osteoblasts and murine with respect to 1,25D3 responsiveness and mineralization.

A full explanation for this apparent discrepancy between human and murine osteoblasts is absent. Both the extracellular milieu (i.e., presence/absence of growth factors, cytokines and other signaling molecules) and the intracellular milieu (e.g., the insulin-like growth factor binding protein-6 that can bind to the VDR and inhibit 1,25D3 induction of ALPL activity) of the cell is important for the eventual effect of 1,25D3 (Cui et al., 2011). Also the extracellular phosphate concentration may affect the 1,25D3 action (Ito et al., 2013). These characteristics may contribute to the differences in 1,25D3 effects observed in human and murine osteoblasts.

Besides stimulation of bone formation /mineralization by osteoblasts 1,25D3 has certain protective control mechanisms in place to avoid pathological over-mineralization. For example, 1,25D3 induces BGLAP and SPP1, established inhibitors of mineralization (Noda et al., 1990; MacDonald et al., 1993) and a stimulator of mineralization, bone sialoprotein (IBSP), is inhibited by 1,25D3 (Li and Sodek, 1993). As mentioned above also the presence or absence of other growth factors, cytokines or signaling molecules may limit the 1,25D3 effect. Examples of this in relation to mineralization are Activin A and follistatin. Activin A inhibits osteoblast differentiation and mineralization (Eijken et al., 2007). Activin A expression in human osteoblasts is stimulated by 1,25D3 (Woeckel et al., 2013), implicating that 1,25D3 as stimulator of human osteoblast differentiation and mineralization also stimulates the production of a mineralization inhibitor. A function in the prevention of over-mineralization is supported by the data that the activin A blocker follistatin enhances 1,25D3 stimulated mineralization (Woeckel et al., 2013). The above mentioned induction of carboxylated osteocalcin by 1,25D3 may fit this hypothesis on preventing over-mineralization. Accumulation of osteocalcin in the ECM of human osteoblast cultures stimulated by 1,25D3 is inhibited by warfarin (antagonist of vitamin K) while vitamin K2 (cofactor of γ-carboxylase) enhanced the 1,25 D3 effect (Koshihara and Hoshi, 1997). 1,25D3 stimulated mineralization was significantly augmented by warfarin (Woeckel et al., 2013). These data on activin A, follistatin, warfarin, and vitamin K put forward a 1,25D3 induced regulatory mechanism to guarantee optimal mineralization (Woeckel et al., 2013). Differences in these regulatory loops may also be part of the differences in 1,25D3 effects in human and murine osteoblast studies.

The most well-known mechanism to limit the biological activity of 1,25D3 is its degradation via 24-hydroxylation. 1,25D3 potently induces CYP24A1, which encodes for the enzyme 24 hydroxylase, in osteoblasts. 24-Hydroxylation is the first step in the degradation cascade of active 1,25D3 (Ohyama et al., 1994). However, hydroxylation at the C-24 position doesn't directly lead to an inactive vitamin D molecule. Henry and Norman demonstrated the significance of 24,25-dihydroxyvitamin D3 (24,25D3) for normal chicken egg hatchability and calcium and phosphorus homeostasis (Henry and Norman, 1978; Norman et al., 1980). Already in 1980 it was shown that 24,25D3 directly stimulates calcification of bone in interaction with PTH and that the number and size of resorption sites in bone is decreased by 24,25D3 (Endo et al., 1980; Galus et al., 1980). Several other studies supported a positive effect of 24,25D3 on bone metabolism (Matsumoto et al., 1985; Tam et al., 1986; Kato et al., 1998) while one study showed no effect of 24,25D3 on histomorphometric parameters in ovariectomized rats (Erben et al., 1992). Administration of 24,25D3 in combination with 1,25D3 improved fracture healing in chickens (Seo et al., 1997) and interestingly, 24,25D3 serum levels correlated to fracture healing (Seo and Norman, 1997). Studies with the CYP24A knockout mouse supported a role for 24,25D3 in fracture repair (St-Arnaud, 2010). Albeit in a human study no positive association with femoral fracture was observed (Weisman et al., 1978). However, a study in pre-dialysis renal insufficiency patients supported a direct, i.e., PTH-independent, functional role of 24,25D3 in bone (Birkenhäger-Frenkel et al., 1995). These data suggest a direct effect on osteoblasts. *In vitro* studies with human osteoblasts have shown that indeed 24,25D3 has direct effects similar to that of 1,25D3 (Van Driel et al., 2006b). A recent comparative gene expression profiling study of 1,25D3, 24,25D3, and 25D3 in primary human and mouse fibroblasts suggested induction of metabolite specific sets of genes and pathways (Tuohimaa et al., 2013). It is important to note that the fact whether biological active levels of 24,25D3 or 1,24,25 trihydroxyvitamin D3 (1,24,25D3) can be reached fully depends on the velocity of the subsequent steps in the degradation pathway after the initial 24-hydroxylation step.

We have shown that osteoblasts besides degradation of active 1,25D3, are able to convert 25-hydroxyvitamin D3 (25D3) into the biologically most active form 1,25D3, suggesting a direct relationship between 1,25D3 synthesis and bone (Van Driel et al., 2006a). This study showed functionality of 1α-hydroxylation in human osteoblast differentiation. 25D3 induced expression of CYP24, osteocalcin and stimulated ALPL activity and mineralization, which was blocked by inhibition of 1α-hydroxylase by ketoconazole. Downregulation of CYP27B1 in human osteoblasts or perturbation of CYP27B1 supported the requirement of 1αhydroxylase for the effect on human MSC proliferation and osteogenic differentiation (Atkins et al., 2007; Geng et al., 2011a). CYP27B1 expression is reduced in MSC of older subjects and resistance to 25D3 induced osteoblast formation points to an aging effect (Geng et al., 2011b). The 1α-hydroxylase-dependent 25D3 stimulation of ALPL activity in human MSC was blocked by histone deacytylase inhibition (Zhou et al., 2013). Of interest, 25D3 has been shown to regulate gene expression in a gene expression profiling study with CYP27B1 deficient fibroblasts (Tuohimaa et al., 2013). This suggests that 25D3 may act independent of 1α-hydroxylation.

Up to now the data on 1,25D3 production by osteoblasts are derived from *in vitro* studies. *In vivo* significance of CYP27B1 and 1,25D3 formation in osteoblasts needs yet to be proven, for example by knocking out CYP27B1 specifically in osteoblasts. However, the observed discrepancies in effects on human-murine osteoblasts may hamper this approach. Although yet *in vivo* proof is lacking, the principal of local synthesis of 1,25D3 in bone may explain the observed associations of 25D3 and not of 1,25D3 with bone as well as other parameters (Hewison et al., 2004; Anderson et al., 2013). Besides CYP27B1, osteoblasts also express the receptors megalin and cubulin that are involved in cellular uptake of 25D3 via endocytosis of the vitamin D binding protein (DBP) (Van Driel et al., 2006a; Atkins et al., 2007). Linking back to the above discussed interaction between locally produced growth factors and 1,25D3 is the regulation of CYP27B1 in osteoblasts. Albeit 1,25D3 itself inhibits CYP27B1 expression in MSC as well as in the kidney (Zhou et al., 2010), the regulation appears to be different and more complex than in the kidney involving local regulators. Several locally in bone produced factors affects CYP27B1 expression: TGFβ suppresses 5- -flanking region of CYP27B1 (Turner et al., 2007) and interferon-β reduces while interleukin-1 and IGF-I increase CYP27B1 expression in mature human osteoblasts (Van Driel et al., 2006a; Zhou et al., 2010; Woeckel et al., 2012). The effect of interleukin-1 points to the involvement of NF-κB in stimulation of CYP27B1 expression in human osteoblasts. This is supported by the interferon-β inhibition of NF-κB in synoviocytes (Van Holten et al., 2004) and CYP27B1 regulation in human dendritic cells (Hewison et al., 2003).

1,25D3 plays an important role in maintaining bone health either via controlling calcium and phosphate homeostasis or via direct effects on osteoblasts. This latter is supported by the direct effects of 1,25D3 on osteoblast differentiation, expression and activity of bone formation related proteins and enzymes, and mineralization. The complete vitamin D endocrine system, from receptor to enzymes involved in 1,25D3 synthesis and breakdown, is present in the osteoblast, pointing to an autocrine/paracrine 1,25D3 function in bone. This is the more so interesting as over the past decade it has become clear that osteoblasts are not only involved in bone metabolism but that they also form the hematopoietic stem cell (HSC) niche controlling renewal of HSCs and differentiation of the immune cells (Calvi et al., 2003). Moreover, these HSC niches are also the sites of bone metastasis (Shiozawa et al., 2011). Considering the 1,25D3 effect on the immune system and tumor cell growth it is tempting to speculate that autocrine/paracrine action of 1,25D3 is also beyond bone metabolism and important for other regulatory functions of osteoblasts. It is therefore of critical importance to understand the full picture of 1,25D3 effects on osteoblasts. One of the approaches to obtain information on the effects of 1,25D3 on osteoblasts and MSC in an unbiased way is by omics approaches in combination with bioinformatics. In the next paragraph the current available 1,25D3 gene expression profiling studies of osteoblasts will be discussed.

## **VITAMIN D AND GENE TRANSCRIPTION IN THE CONTEXT OF OSTEOBLAST DIFFERENTIATION AND BONE FORMATION**

1,25D3 has been shown to regulate the expression of various genes related to osteoblast proliferation and differentiation. BMP-2 induced bone formation has been suggested to be enhanced by 1,25D3 induced c-MYC expression (Piek et al., 2010). Induction of Insulin-like growth factor-binding proteins (IGFBP)-2, -3, and -4 expression by 1,25D3 in human MSC may play a role in stimulation of osteogenic differentiation (Kveiborg et al., 2001). Recently, Li and coworkers (Li et al., 2013) demonstrated that IGFBP-3 interacts with the VDR and negatively regulates CYP24 and BGLAP expression. Overexpression of IGFBP-3 inhibited the 1,25D3 activation of ALP in MG-63 human osteosarcoma cells.

1,25D3 also regulated Forkhead Box O (FoxO) transcription factors in murine MC3T3 osteoblasts with FoxO3a being up-regulated while FoxO1 was down-regulated, and FoxO4 not affected. Knockdown of the FoxO's didn't change 1,25D3 inhibition of cell growth but led to increased accumulation of reactive oxygen species after 1,25D3 treatment (Eelen et al., 2013). This may be linked to cellular metabolism and the high energy demanding process of bone formation (Komarova et al., 2000; Chen et al., 2008; Bruedigam et al., 2010). Unfortunately, the effect of FoxO's knockdown on mineralization in these murine MC3T3 osteoblast cultures was not reported. 1,25D3 increased vascular endothelial growth factor (VEGF) expression in human and rat osteoblasts is interesting considering the relationship between bone formation and angiogenesis (Wang et al., 1996; Schlaeppi et al., 1997; Corrado et al., 2013). VEGF has been shown to be involved in the 1,25D3 bone anabolic effect (Wang et al., 1997).

Recent studies placed miRNAs in the 1,25D3 mechanism of action spectrum in osteoblasts. Five miRNAs were found to be differentially expressed in primary human osteoblast after 6 h of treatment with 1,25D3 (Lisse et al., 2013a,b). Interestingly, miR-637 and miR-1228 are two miRNAs located intergenic in DAPK3 and LRP1, respectively. miR-1228 was upregulated and coexpressed with its host gene LRP1 suggesting a conventional VDRE- mediated transactivation upon 1,25D3 treatment. Since LRP1 is known to mediate the canonical Wnt pathway in fibroblasts (Terrand et al., 2009), this suggests an indirect regulation of Wnt signaling by 1,25D3 adding to other data on 1,25D3 and Wnt signaling interaction (Fretz et al., 2007; Haussler et al., 2010). The target of miR-1228, BMP2K, was previously identified to be increased in mouse osteoblasts upon treatment with BMP2 (Kearns et al., 2001). Stable expression of BMP2K in mouse osteoprogenitor cells decreased ALPL activity and osteocalcin mRNA levels. This suggests that 1,25D3 induced expression of miR-1228 may affect osteoblast differentiation via down-regulation of BMP2K.

On the contrary, 1,25D3 upregulated miR-637 while its host gene was downregulated suggesting a different way of regulation of the two transcripts. miR-637 stimulated the degradation of COL4A mRNA levels that is expressed in the basement membrane and is downregulated during early differentiation of mouse MC3T3-E1 osteoblasts (Hong et al., 2010). It is becoming evident that miRNAs play an important role in osteoblast differentiation and bone formation (Lian et al., 2012) and in the near future more data on their role in 1,25D3 action in osteoblasts will come forward (Lisse et al., 2013a).

## **IDENTIFICATION OF 1,25D3 TARGET GENES IN OSTEOBLASTS**

In the past various studies have investigated the effects of 1,25D3 on target gene expression and VDR binding to DNA response elements. Only a few of these genome-wide studies have investigated the effects of 1,25D3 in the context of osteoblasts (**Table 1**). The studies that carried out are very heterogenic with regard to the differentiation stage of the cells (MSC vs. primary osteoblasts vs. Cell line), time points of treatment (2–6 h after treatment) and the 1,25D3 concentration that is used (1–100 nM). Together this makes it difficult to compare the different studies. Systematic analyses of both mRNA gene expression profiling and VDR binding experiments at early time points after induction with 1,25D3 will uncover direct target genes. Below we will address a few of these studies and the results obtained.

#### **ChIP ANALYSES IN OSTEOBLASTS**

Upon binding of 1,25D3 to the VDR, the VDR binds with its heterodimeric partner retinoid X receptor (RXR) on the vitamin D receptor response elements (VDRE). The VDRE consists of the hexameric sequence AGGTCAxxxAGGTCA (Ozono et al., 1990) but variants to the conserved sequence have been identified (Meyer, 2005). Due to the diversity of VDRE, bioinformatics approaches are limited in identifying whole genome VDR binding sites. To identify direct target genes of VDR, genome-wide approaches such as ChIP-chip or ChIP-seq approaches have to be performed. Systematic analyses of VDR binding upon activation by 1,25D3 combined with bioinformatics approaches identifies VDRE (VDR response elements) and subsequently direct targets of Vitamin D signaling. A few studies have started to identify 1,25D3 target genes is various cell types such as a human derived lymphoblastoid cell line (Ramagopalan et al., 2010) and monocytes (Heikkinen et al., 2011). Recently, the first VDR binding experiments in osteoblasts were published (Meyer et al., 2010). Meyer et al. analyzed the genomic locations that bind VDR, RXR, RNA polymerase II and acetylated H4 after 3 h treatment with 1,25D3 in mouse MCT3T-E1 osteoblasts. Interestingly, only 13% of the identified sites was located in classical promoter regions upstream vitamin D target genes. The majority of sites that were found to bind VDR, RXR and acetylated H4 were located distal (43%) and within intronic and exonic regions (44%). This demonstrates that distal transcriptional control contributes to the majority of vitamin D3-mediated transcription. Genome wide ChIP-seq analyses with human osteoblasts should illustrate whether binding of VDR at distal locations is conserved.

Pilot analysis of our gene expression profiles of osteogenic and adipogenic MSCs illustrated that many known 1,25D3 responsive genes (on basis of Ingenuity database; www*.*ingenuity*.*com) are dynamically expressed during adipogenic as well as osteogenic differentiation (data not published). This data does not directly show that these genes are regulated by 1,25D3 but it suggests that 1,25D3-responsive genes can have a role during the differentiation of mesenchymal precursors. Many of the two-fold regulated genes during osteogenic differentiation and those that were identified previously to be regulated by 1,25D3 are involved in Cell Cycle (41/162; GO:0007049), response to steroid hormone (21/162; GO:0048545), regulation of phosphate metabolic process (26/162, GO:0019220), regulation of apoptosis (31/162, GO:0042981), extracellular region part (36/162; GO:0044421). ChIP analyses using VDR and expression profiling of 1,25D3

#### **Table 1 | Genome-wide studies of vitamin D and osteoblasts.**


*Database searches were performed using Bone[Title/Abstract] OR osteoblast[Title/Abstract] AND vitamin D AND microarray in Pubmed (http://www.ncbi.nlm.nih.gov/pubmed) and GEO (http://www.ncbi.nlm.nih.gov/geo/).*

transcriptional activity against the backdrop of osteogenic MSC will be needed to demonstrate the importance of VDR—1,25D3 binding in osteoblast function either in bone formation, regulation of osteoclast formation and activity or in the stem cell niche.

## **GENE EXPRESSION PROFILING IN OSTEOBLASTS**

Besides binding of the VDR to gene regulatory elements, important information on the effect of 1,25D3 on osteoblasts comes from expression profiling studies upon 1,25D3 treatment. Several gene expression profiling studies have been performed to examine the effect of 1,25D3 on RNA expression in osteoblasts. Gene expression profiling in murine MC3T3 cells showed downregulation of DNA replication genes (Eelen et al., 2004) which fits the earlier observed inhibition of proliferation in these cells. Gene profiling of 1,25D3 treated human osteoblasts at multiple days during the differentiation phase before mineralization did not show regulation a specific set of DNA replication genes (Woeckel et al., 2010). Cell death, RNA splicing translation, and cell cycle genes were identified by Gene Ontology analyses as being most significantly overrepresented (Woeckel et al., 2010). Only 0.6 % (3 genes) of the genes changed in expression during the mineralizing period were also changed prior to mineralization (Woeckel et al., 2010). This study demonstrated that 1,25D3 has different effects on gene expression dependent on the differentiation stage of the cells and should be carefully addressed when investigating the effects of 1,25D3 on mesenchymal stem/stromal cells and differentiated osteoblasts.

Tarroni et al. found that upon 24 h treatment of human osteoblasts with 1,25D3 most genes were upregulated (136 up vs. 20 down) indicating the transcriptional activation of 1,25D3 (Tarroni et al., 2012). Pathway analyses identified various biological functions and/or diseases related to bone metabolism and cellular processes/molecular functions related to skeletal development. The link with skeletal development is supported by another study showing 1,25D3 induced expression in human and mouse osteoblasts of the odd-skipped related genes Osr1 and Osr2, known from expression in the developing limb (Verlinden et al., 2013).

Tarroni et al. also showed strong change in expression of genes linked to inflammation or immune and lymphatic system development (Tarroni et al., 2012). In line with this, is the observation of a gene profiling study showing interferon-related genes being overrepresented after 1,25D3 treatment of human osteoblasts. The interferon signaling related genes were down-regulated by 1,25D3 (Woeckel et al., 2012). The observations on processes related to the immune system are interesting from at least two points of view. Firstly, because of the link between the immune system and bone and the effect of immune cells-derived cytokines on bone metabolism, e.g., in conditions like rheumatoid arthritis. Secondly, considering the above mentioned role of the osteoblasts in the stem cell niche and control of hematopoietic stem cell renewal and differentiation. The expression profiling data and the identification of functions and processes related to the immune system may support a role of vitamin D in osteoblasts control the stem cell niche (Kawamori et al., 2010).

## **CONCLUSION**

Vitamin D can regulate bone metabolism in an indirect way via controlling calcium and phosphate homeostasis but also via direct effects on osteoblasts. In fact, the complete vitamin D endocrine system is present in osteoblasts. This enables osteoblasts to respond not only to vitamin D via the VDR but also to synthesize the biological most active vitamin D metabolite 1,25D3 and to act in an autocrine/paracrine manner. Vitamin D directly regulates gene expression and stimulates mineralization in *ex vivo* cultures of human and rat osteoblasts. The effect on mineralization may depend on species and/or environmental context that can alter the eventual vitamin D effect. Besides effects on bone metabolism, vitamin D effects on osteoblasts may be related to additional functions of osteoblasts such as the hematopoietic stem cell niche. Interesting in this respect is that gene expression profiling studies on vitamin D-treated osteoblasts revealed genes and processes related to the immune system. Further studies are needed to delineate these non-bone metabolism related effects of vitamin D in osteoblasts in greater detail at cellular and molecular level. A future challenge will be to construct networks representing the effects of vitamin D, either in bone metabolism- or in non-bone metabolism-related processes, against the backdrop of osteoblast differentiation by systems biological approaches.

## **REFERENCES**


(1,25D) on osteocyte like cells. *J. Steroid Biochem. Mol. Biol.* 136, 1–4. doi: 10.1016/j.jsbmb.2012.09.029


24,25-dihydroxyvitamin D3 levels are correlated with the healing process after chick tibial fracture. *J. Bone Miner. Res.* 12, 598–606. doi: 10.1359/jbmr.1997.12.4.598


17/2.8 osteoblast-like cells. *J. Steroid Biochem. Mol. Biol.* 103, 322–325. doi: 10.1016/j.jsbmb.2006.12.041


cross-talk in human osteoblasts with 1α,25-dihydroxyvitamin D3 being dominant over interferon-β in stimulating mineralization. *J. Cell. Physiol*. 227, 3258–3266. doi: 10.1002/jcp.24020


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

*Received: 05 February 2014; accepted: 20 March 2014; published online: 09 April 2014. Citation: van de Peppel J and van Leeuwen JPTM (2014) Vitamin D and gene networks in human osteoblasts. Front. Physiol. 5:137. doi: 10.3389/fphys.2014.00137 This article was submitted to Integrative Physiology, a section of the journal Frontiers in Physiology.*

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

## The role of vitamin D in skeletal and cardiac muscle function

## *Patsie Polly1,2\* and Timothy C. Tan1,3*

*<sup>1</sup> Inflammation and Infection Research Centre, School of Medical Sciences, Faculty of Medicine, UNSW Australia, Kensington, NSW, Australia*

*<sup>2</sup> Department of Pathology, School of Medical Sciences, Faculty of Medicine, UNSW Australia, Kensington, NSW, Australia*

*<sup>3</sup> Cardiac Ultrasound Laboratory, Department of Cardiology, Massachusetts General Hospital, Boston, MA, USA*

#### *Edited by:*

*Carsten Carlberg, University of Eastern Finland, Finland*

#### *Reviewed by:*

*Joerg Reichrath, Universitätsklinikum des Saarlandes, Germany Johannes Van Leeuwen, Erasmus University Medical Center, Netherlands*

#### *\*Correspondence:*

*Patsie Polly, Department of Pathology, School of Medical Sciences, Faculty of Medicine, Rm 420 Wallace Wurth East Building, UNSW Australia, Kensington, NSW 2052, Australia e-mail: patsie.polly@unsw.edu.au*

Myopathy is a feature of many inflammatory syndromes. Chronic inflammation has been linked to pathophysiological mechanisms which implicate 1,25 dihydroxyvitamin D3 (1,25-(OH)2D3)-mediated signaling pathways with emerging evidence supporting a role for the vitamin D receptor (VDR) in contractile and metabolic function of both skeletal and cardiac muscle. Altered VDR expression in skeletal and cardiac muscle has been reported to result in significant effects on metabolism, calcium signaling and fibrosis in these tissues. Elevated levels of serum inflammatory cytokines, such as IL-6, TNF-α and IFNγ, have been shown to impact myogenic and nuclear receptor signaling pathways in cancer-induced cachexia. The dysregulation of nuclear receptors, such as VDR and RXRα in muscle cells, has also been postulated to result in myopathy via their effects on muscle structural integrity and metabolism. Future research directions include generating transcriptome-wide information incorporating VDR and its gene targets and using systems biology approaches to identify altered molecular networks in human tissues such as muscle. These approaches will aid in the development of novel therapeutic targeting strategies for inflammation-induced myopathies.

#### **Keywords: cytokines, cancer cachexia, skeletal muscle, cardiac muscle and transcriptome**

## **VITAMIN D AND MUSCLE FUNCTION**

Vitamin D is necessary for the maintenance of structural integrity and function of the musculoskeletal system (Pfeifer et al., 2002). Severe deficiency results in impaired bone strength and deformation i.e., rickets in children and osteomalacia in adults. Vitamin D plays a significant role in calcium homeostasis and bone metabolism through its actions on target tissues (DeLuca, 2004). Serum vitamin D levels have been correlated to muscle cell contractility, muscle strength, and postural stability (Rodman and Baker, 1978; Marcinkowska, 2001; Grimaldi et al., 2013; Girgis et al., 2014). Low serum levels have been related to proximal muscle weakness, gait disturbance, paresthesia, and discomfort within the muscles (Skaria et al., 1975; Schott and Wills, 1976; Glerup and Eriksen, 1999; Glerup et al., 2000; Pfeifer et al., 2002; Ahmed et al., 2009).

The clinical features of myopathy associated with severe vitamin D deficiency are supported by findings of abnormal histological and electrophysiological changes in muscle. Histological analysis of skeletal muscle biopsied from adults with vitamin D deficiency reveal enlarged inter-fibrillar spaces, infiltration of fat, presence of glycogen granules, fibrosis, and type II muscle fiber (fast-twitch) atrophy (Sorensen et al., 1979; Yoshikawa et al., 1979; Boland, 1986; Sato et al., 2005). Recent interest has been in reversal of some of these pathological effects of this clinical syndrome. Meta-analysis of randomized controlled trials in the elderly with low serum levels of vitamin D, demonstrated a decrease in the risk of falls following supplementation with vitamin D (Rejnmark, 2011). This outcome has been attributed to the ability of vitamin D to impact muscle fiber composition hence skeletal muscle structure. Studies in vitamin D deficient patients revealed an increase in percentage of type II fibers, a significant increase in mean type II muscle fiber diameter and area particularly of type IIa muscle fibers following treatment with 1-α-hydroxyvitaminD3 and calcium. However, it is still unclear if vitamin D supplementation induced formation of new type II fibers or increased transition of existing type I fibers from to type II (Sorensen et al., 1979; Sato et al., 2005). Vitamin D has also been demonstrated to increase cell proliferation and inhibit apoptosis in injured rat soleus skeletal muscle, with positive functional outcomes such as faster recovery of contraction forces (Stratos et al., 2013). The therapeutic potential of vitamin D supplementation has also recently been tested on dysferlin gene regulation and dysferlinopathies (autosomal recessive neuromuscular disorder characterized by progressive muscle wasting due to dysferlin gene mutations and a deficiency of functional dysferlin protein). Vitamin D increased dysferlin gene expression in both HL60 monocytes and skeletal muscle cells via the activation of vitamin D receptor (VDR) which binds to the dysferlin promoter; and non-genomic MEK/ERK signaling and classical genomic effects. 1,25(OH)2D3 has also been reported to suppress myotube formation by decreasing Myf5 and myogenin gene expression resulting in increased myotube diameters but reduced myostatin expression potentially alleviating the myopathic effects of muscle weakness and reduced contractile function (Luna et al., 2012).

Experiments in C2C12 cells highlight some key molecular regulatory effects of 1,25(OH)2D3 including: (1) increased expression and nuclear translocation of the VDR, (2) decreased cell proliferation, (3) decreased IGF-I expression, and (4) increased IGF-II and follistatin expression and decreasing the expression of myostatin which appeared to promote myogenic differentiation and (5) altered differentiation and myotube size. Hence, vitamin D may also be considered for use in intervention studies for muscle conditions that involve these mechanisms (Garcia et al., 2011; Girgis et al., 2014).

## **THE ROLE OF VITAMIN D RECEPTOR IN MUSCLE FUNCTION**

The effects of vitamin D are modulated by its receptor, therefore the expression and distribution of VDR is of significant importance. Early studies demonstrated the presence of the VDR in cultured human myoblasts and myotubes which showed a response to physiological concentrations of 1,25-(OH)2D3. VDR is also present in human skeletal muscle cells within the nuclei and has been shown to play a role in skeletal muscle development, my fiber size and morphology (Simpson et al., 1985; Costa et al., 1986; Bischoff et al., 2001; Bischoff-Ferrari et al., 2006). Skeletal muscle development requires a co-ordinated series of transcription factor and growth factor events that enable progenitor cells to undergo myoblast determination (requiring Pax3, Pax7, MyoD, and Myf5) then myoblast to myotube determination (requiring p21Cip1, myogenin, MEF2C and Rb) then further myotube maturation requiring innervation, MRF4, MLP) (Ludolph and Konieczny, 1995; Perry and Rudnicki, 2000; Ryhänen et al., 2003; Miyazawa et al., 2005; Washington et al., 2011). VDR and myosin heavy chain isoform was shown to co-localize in skeletal muscle biopsies in older female subjects (Ceglia et al., 2010). VDR has also been shown to impact the expression of myogenic transcriptional regulators, in particular Myf5, myogenin, E2A, and early myosin heavy chain isoforms (Endo et al., 2003; Girgis et al., 2014). C2C12 myoblasts treated with 1,25-(OH)2D3 showed increased VDR and CYP24A1 expression above endogenous levels which resulted in inhibition in cell proliferation (Srikuea et al., 2012; Girgis et al., 2014). Furthermore, inhibition of myogenic differentiation of C2C12 and G8 cell lines was also achieved with suppression of VDR expression, suggesting that myoblasts require signals transmitted through VDR for differentiation into myocytes. Myogenic differentiation likely involves the orchestration of myogenic transcription factors in skeletal muscle (Girgis et al., 2013). Vitamin D signaling may modulate p21CIP1 and Rb as well as myogenin, which are important in myogenic differentiation of myoblasts to myotubes (Ludolph and Konieczny, 1995; Perry and Rudnicki, 2000). Autocrine vitamin D signaling has also been reported to regulate functional effects such as contraction and remodeling in smooth muscle cells although the autocrine effects in skeletal and cardiac muscles still require characterization (Weisman et al., 2005; Maghni et al., 2007; Eggersdorfer and Stöcklin, 2013).

Effects in VDR-null mutant mice further highlight the importance of the VDR in muscle biology. Apart from the observed growth retardation, osteomalacia and systemic metabolic changes such as secondary hyperparathyroidism and hypocalcemia, these mutant mice also had abnormal muscle structure and function (Burne et al., 2005). VDR-null mutant mice displayed a progressive decrease in their muscle fiber diameters compared to those of wild-type mice, which was evident early in the postnatal period (prior to weaning) and associated with an abnormally high expression of myogenic differentiation factors. These observations suggest alterations in muscle cell differentiation pathways and thus abnormal muscle fiber development and maturation (Endo et al., 2003). Interestingly, the muscle fiber abnormalities described were diffuse without any preference for type I or II fibers, which was different to myopathy due to vitamin D deficiency where there was a predominance of type II fiber loss. Additionally, the mutant mice had a total 33% body weight reduction compared to controls at maturity; implying a post-natal role for VDR in maintaining weight (Song et al., 2003). Increased VDR expression is also correlated with regeneration (Srikuea et al., 2012), but levels of VDR appear to decrease with increasing age, which has been proposed as a potential mechanism contributing to reduced muscle strength in the Bischoff-Ferrari et al. (2004).

In the context of muscle biology, VDR mediates both nongenomic and genomic effects of vitamin D (Buitrago et al., 2001; Capiati et al., 2002). VDR knock-down experiments demonstrated that 1,25-(OH)2D3-induced p38 MAPK activity occurs through Src phosphorylation, while also reducing ERK1/2 and Akt activity. These non-genomic effects include the stimulation of transmembrane second messenger systems involving adenylyl cyclase/cAMP/PKA and PLC/DAG+IP3/PKC to affect contractile function and myogenesis. Furthermore, 1,25−(OH)2D3has also been reported to mediate Ca2+release through voltage and store dependent calcium channels (SOC, CEE) in avian muscle cells (Santillan et al., 2004). Caveolae have also been shown to be involved in 1,25−(OH)2D3activation of in c-Src-MAPKs in C2C12 cells. Ca2<sup>+</sup> influx in caveolae is triggered by the interaction between VDR with TRCP3, an integral protein of capacitative Ca2<sup>+</sup> entry (CCE) (Buitrago and Boland, 2010; Buitrago et al., 2011).

## *New perspectives on vitamin D, chronic inflammation, and muscle physiology*

Valuable insights into the role of vitamin D and muscle function have arisen from the study of certain pathological settings such as chronic inflammatory conditions. These conditions share phenotypic characteristics to vitamin D deficiency states and VDR-null mice. Myopathy is a feature of a number of chronic inflammatory syndromes. Chronic inflammation has been linked to pathophysiological mechanisms which implicate non-genomic and genomic 1,25-(OH)2D3-mediated signaling pathways. Skeletal muscle may be considered as having a level of plasticity, allowing it to respond to environmental, physiological and pathophysiological stimuli that elicit alterations in size, fiber-type and metabolism. Molecular factors such as insulin-like growth factors, calcineurin, desmin, Myf5, Mrf4, MyoD and myogenin have been identified as positive regulators of muscle size, while tumor necrosis factor (TNF)-α, myostatin and components of the ubiquitin pathway, have been recognized as regulators of muscle wasting. Emerging evidence supports a role for VDR in the contractile and metabolic function of both skeletal and cardiac muscle in health and disease **(Figure 1)**. The expression of VDR and its interaction at

the molecular level with proteins that are involved in inflammation, signaling and ultimately contractile function of both skeletal and cardiac muscle is of importance. *In vitro* cell culture models, *in vivo* rodent models as well as clinical studies in humans are starting to clarify the mechanisms of vitamin D action mediated via the VDR in muscle in order to enhance our understanding of their role in inflammatory mediated myopathy and muscle weakness (Girgis et al., 2012, 2013, 2014).

Exercise-induced muscle damage has been shown to increase the expression of VDR while altering gene expression of inflammatory cytokines such as interleukin (IL)-6 and TNF-α and alterations in signaling molecules involved with vitamin D signaling pathways such as phosphorylation of AMPK, p38, ERK1/2, IKK, and IκB simultaneously (Choi et al., 2013). An inverse relationship is generally reported for vitamin D, cancer and muscle structure and function. Alterations in metabolic status and physical activity play a role, however paraneoplastic syndromes such as cancer cachexia integrate many metabolic and catabolic molecular mechanisms which result in pathophysiological skeletal and more recently cardiac muscle effects (Choi et al., 2013). Low serum vitamin D levels are highly prevalent in advanced cancer patients with cachexia or fatigue (Dev et al., 2011). Elevated levels of inflammatory circulating factors, include C-reactive protein (CRP), a currently utilized clinical marker. The VDR axis is reported to play a fundamental role with possible association between CRP and VDR gene polymorphisms, in cancer patients with cachexia. This suggests the notion of cachexia-prone genotypes or to cachexia-resistant genotypes (Punzi et al., 2012). It has been suggested that tumor associated effects such as these may in part be addressed by nutraceutical vitamin D supplemented diets to improve vitamin D status (Endo et al., 1998; Morley, 2009; Morley et al., 2009; Strohle et al., 2010).

Data arising from the study of muscle structure and function in cancer cachexia has revealed new insights into vitamin D. Cancer cachexia is a debilitating clinical syndrome which causes up to 30% of cancer related deaths by either immobility, respiratory and/or cardiac failure (Fearon, 2008) and is characterized by weight loss; up-regulation of inflammatory markers such as IL-6, IL-1, TNF-α and interferon gamma (IFN)γ; hypercalcemia; and insulin resistance (Argiles et al., 2003; Sato et al., 2003; Jackman and Kandarian, 2004; Evans et al., 2008; Tisdale, 2009; Asp et al., 2010). The interaction between host factors and tumor cells is proposed to cause an excess production of cytokines and improper stimulation of downstream signaling molecules which results in weakness and decreased physical activity; thus highlighting the detrimental effects of cachexia on quality of life (Dahele et al., 2007). Of these cytokines, IL-6 is thought to be a key mediator of skeletal and cardiac muscle wasting in the pathogenesis of CC (Argiles et al., 2003; Haddad et al., 2005; Baltgalvis et al., 2008; Tisdale, 2009; Carson and Baltgalvis, 2010). Current treatment strategies are limited and do little to improve survival (Michael and Tannock, 1998; Mantovani et al., 2008).

More recently, we have identified a link between IL-6, the myogenic transcriptional regulator MEF2C and muscle breakdown due to CC (Shum et al., 2012). Different underlying molecular effects may also underlie the pathological changes in skeletal vs. cardiac muscle due to cancer (Shum et al., 2012; Tan et al., 2013; Shum et al., 2013; Falconer et al., in press) "Exercise genes" have now been identified in humans, which now opens the gateway for analyses that focus on the genetic basis of performance. These include the genes encoding for: the angiotensin converting enzyme, alpha-actinin 3, bradykinin, ciliary neurotrophic factor, interleukin-15, insulin-like growth factor II, myostatin and the VDR which have been proposed to play a role in inter-subject variability in muscle strength or size. Current data is only available from healthy subjects, hence genetic variability that may account for these effects still requires further analysis particularly in the context of muscle disease (Stewart and Rittweger, 2006). Furthermore, conversion toward a fatigue prone, type II skeletal myofiber phenotype has been observed due to cancer, which potentially makes this condition treatable with vitamin D.

*Cardiac muscle effects due to cancer cachexia.* Cardiac weight loss is a relatively unreported feature in cancer cachexia although autopsy studies revealed that "cardiac atrophy" is a prominent feature in advanced cancer patients (Hellerstein and Santiago-Stevenson, 1950). Recent studies have demonstrated that the reversal of cardiac and skeletal muscle weight loss increased longevity in mouse models of cancer cachexia, implying that these effects on the heart may contribute to poor prognosis in cancer patients (Zhou et al., 2010). The molecular basis of this cardiomyopathy induced by cancer cachexia is unclear. We and others have established the IL-6 driven, colon 26 (C26) carcinoma cachexia mouse model to study cancer cachexia (Tanaka et al., 1990; Asp et al., 2010; Zhou et al., 2010; Shum et al., 2012). The C26 model demonstrates significant body wasting, has no metastases to the heart, thus effects seen are largely due to the tumor or the host-tumor response (Matsumoto et al., 1999; Schwarzkopf et al., 2006; Strassmann et al., 1992). Features of cardiac wasting in the end stages of cachexia (i.e., 20–25% body weight loss) observed in C26 and other cachectic animal models include: heart weight loss; marked fibrosis; oxidative modifications; reduced expression of contractile apparatus proteins; no increase of apoptosis; and lower ejection fraction (Fukuda et al., 2009; Springer et al., 2009; Marin-Corral et al., 2010; Tian et al., 2010; Shum et al., Unplublished Data). Genes that mediate muscle atrophy such as atrogin-1 and Murf-1, were unaltered in the heart unlike skeletal muscle; implying that cardiac wasting occurs via different molecular pathways (Zhou et al., 2010; Shum et al., 2013; Unplublished Data). Vitamin D and its gene effects in the context of functional consequences have been described in skeletal muscle cell culture models, cardiac muscle and smooth muscle (Meems et al., 2011; Girgis et al., 2014). However, the roles of VDR and 1,25-(OH)2D3 need further characterization in the context of muscle wasting due to cancer cachexia **(Figure 1)**.

*Vitamin D and cardiac pathology.* Vitamin D and its analogs may potentially have palliative effects in the cardiovascular system. Long term exposure to angiotensin II has been shown to induce hypertension, cardiac hypertrophy, activation of the hypertrophic fetal gene program atrial natriuretic peptide (ANP), B-type natriuretic peptide and alpha skeletal actin gene expression), increased expression of the pro-hypertrophic modulatory calcineurin inhibitor protein 1 (MCIP 1), and increased fibrosis with augmented procollagen 1 and 3 gene expression. Co-administration of paricalcitol (a vitamin D analog with agonist properties) in an animal model of non-renin-dependent cardiac hypertrophy partially reversed the reported AII-dependent effects. Interestingly, the effects of agonist-bound vitamin D receptor appeared to elicit potent anti-hypertrophic activity in this model of cardiac hypertrophy. The anti-hypertrophic activity appears to be at least partially intrinsic to the cardiac myocyte and may involve suppression of the MCIP 1 protein (Chen and Gardner, 2013). Though the cardiovascular system is not thought to represent a classical target for 1,25-(OH)2D3 and retinoic acid (RA), it is clear that both cardiomyocytes and vascular smooth muscle cells respond to these nuclear receptor hormones (NRHs) with changes in growth characteristics and gene expression (**Figure 1**). These NRHs suppress many of the phenotypic correlates of endothelin-induced hypertrophy in a cultured neonatal rat cardiac ventriculocyte model. Each of these NRHs reduced endothelin-stimulated ANP secretion in a dose-dependent manner and when the two were used in combination, they proved to be more effective than when either NRH was used alone. 1,25-(OH)2D3 abrogated the increase in cell size seen after endothelin treatment. These findings suggest that liganded vitamin D and retinoid receptors are capable of modulating the hypertrophic process *in vitro* and that agents acting through these or similar signaling pathways may be of value in probing the molecular mechanisms underlying hypertrophy (Wu et al., 1996) **(Figure 1)**.

### *Transcriptome-wide effects and muscle*

Recently, transcriptome-wide approaches have been applied to muscle in order to get a global view of changes that occur due to various stimuli, for example structural vs. metabolic. The transcriptional profile of VDR mRNA isoforms has been examined for differences in bone, cartilage and paravertebral muscles between tissues from curve concavity and convexity. VDR was differentially expressed in paravertebral muscles in patients with juvenile idiopathic scoliosis (JIS) and adult idiopathic scoliosis (AIS). The VDRl isoform appears to contribute to curve concavity in paravertebral muscles. Furthermore, muscular transcriptome differentiation was evident between curve concavity and convexity in JIS patients. Tob2 and MED13 gene expression in paravertebral muscles appear to differentiate the two types of idiopathic scoliosis (Nowak et al., 2012).

Gene expression has been examined in skeletal muscle tissue of obese insulin-resistant subjects before and after a euglycemichyperinsulinemic clamp to determine the pathogenesis of insulin resistance. Differential gene expression was demonstrated for enzymes, transcription, and translation regulators, transporters, G protein-coupled receptors, cytokines, and ligand-dependent nuclear receptors. Metabolic pathways that incorporated, inflammatory signaling and nuclear receptors were also significantly different. These included LXR/RXR activation, VDR/RXR activation, interleukin IL-8, acute phase response, IL-10, triggering receptor expressed on myeloid cells 1, peroxisome proliferatoractivated receptor, G-beta/gamma and hepatocyte growth factor and IL6 signaling (Rudkowska et al., 2013).

Comparisons between transcriptomes and proteomes in muscle tissues and activated CD4+ and CD8+ T lymphocytes (T-cells) analyzed using Affymetrix microarrays and mass spectrometry, from type 2 diabetes (T2DM) subjects and matched non-diabetic controls, demonstrated reduced gene expression for insulin receptor (INSR), VDR, insulin degrading enzyme, Akt, insulin receptor substrate-1 (IRS-1), IRS-2, glucose transporter 4 (GLUT4), and enzymes of the glycolytic pathway in the T2DM subjects compared controls. Increased gene expression was shown for plasma cell glycoprotein-1, TNFα, and gluconeogenic enzymes in T2DM subjects. Observed alterations in transcriptomes and proteomes between muscle and activated T-cells of T2DM were comparable suggesting a more global molecular basis for insulin resistance (Stentz and Kitabchi, 2007).

### *Conclusion and perspectives*

There is now clear evidence supporting a significant role for vitamin D in the biology and function of skeletal and cardiac muscle. Current evidence outlines a number of effects of vitamin D on these muscle types including intracellular calcium handling, differentiation and contractile protein composition. However further study using novel investigative strategies is still warranted to better delineate the role and functions of vitamin D in muscle. The molecular interplay between cytokine signaling, VDR expression, genetic variability in patients with myopathy due to chronic inflammatory conditions such as cancer cachexia may reveal the molecular basis for changes that have been observed in skeletal and cardiac muscle. Early transcriptomic studies on the effects of cytokines in muscle wasting due to cancer cachexia have provided clues regarding potential molecular mechanisms induced by cytokines that drive muscle wasting which may potentially also implicate vitamin D mediated transcriptional mechanisms although this still remains to be defined. A better characterization of the role of VDR in the context of inflammation-mediated muscle wasting and weakness may also potentially translate into significant clinical applications by informing nutraceutical approaches using vitamin D supplementation as a potential strategy for reversing muscle wasting.

## **REFERENCES**


p27, Cdk2, cyclin E, and the retinoblastoma protein. *Biochem. Pharmacol*. 66, 495–504.


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

*Received: 14 February 2014; accepted: 27 March 2014; published online: 16 April 2014. Citation: Polly P and Tan TC (2014) The role of vitamin D in skeletal and cardiac muscle function. Front. Physiol. 5:145. doi: 10.3389/fphys.2014.00145*

*This article was submitted to Integrative Physiology, a section of the journal Frontiers in Physiology.*

*Copyright © 2014 Polly and Tan. 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.*

## Vitamin D and adipose tissue—more than storage

## *Shivaprakash J. Mutt 1,2\*, Elina Hyppönen3,4,5, Juha Saarnio6, Marjo-Riitta Järvelin2,7, 8,9 and Karl-Heinz Herzig1,2,10\**


*<sup>10</sup> Medical Research Center Oulu and Oulu University Hospital, Oulu, Finland*

#### *Edited by:*

*Carsten Carlberg, University of Eastern Finland, Finland*

#### *Reviewed by:*

*Nina Eikelis, Baker IDI Heart and Diabetes Institute, Australia JoEllen Welsh, State University of New York at Albany Cancer Research Center, USA*

#### *\*Correspondence:*

*Shivaprakash J. Mutt and Karl-Heinz Herzig, Institute of Biomedicine and Biocenter of Oulu, University of Oulu, Aapistie 5, PO Box 5000, FIN-90014 Oulu, Finland e-mail: shivaprakash.jagalur@oulu.fi; karl-heinz.herzig@oulu.fi*

The pandemic increase in obesity is inversely associated with vitamin D levels. While a higher BMI was causally related to lower 25-hydroxyvitamin D (25(OH)D), no evidence was obtained for a BMI lowering effect by higher 25(OH)D. Some of the physiological functions of 1,25(OH)2D3 (1,25-dihydroxycholecalciferol or calcitriol) via its receptor within the adipose tissue have been investigated such as its effect on energy balance, adipogenesis, adipokine, and cytokine secretion. Adipose tissue inflammation has been recognized as the key component of metabolic disorders, e.g., in the metabolic syndrome. The adipose organ secretes more than 260 different proteins/peptides. However, the molecular basis of the interactions of 1,25(OH)2D3, vitamin D binding proteins (VDBPs) and nuclear vitamin D receptor (VDR) after sequestration in adipose tissue and their regulations are still unclear. 1,25(OH)2D3 and its inactive metabolites are known to inhibit the formation of adipocytes in mouse 3T3-L1 cell line. In humans, 1,25(OH)2D3 promotes preadipocyte differentiation under cell culture conditions. Further evidence of its important functions is given by VDR knock out (VDR−*/*−) and CYP27B1 knock out (CYP27B1 <sup>−</sup>*/*−) mouse models: Both VDR−*/*<sup>−</sup> and CYP27B1−*/*<sup>−</sup> models are highly resistant to the diet induced weight gain, while the specific overexpression of human VDR in adipose tissue leads to increased adipose tissue mass. The analysis of microarray datasets from human adipocytes treated with macrophage-secreted products up-regulated VDR and CYP27B1 genes indicating the capacity of adipocytes to even produce active 1,25(OH)2D3. Experimental studies demonstrate that 1,25(OH)2D3 has an active role in adipose tissue by modulating inflammation, adipogenesis and adipocyte secretion. Yet, further *in vivo* studies are needed to address the effects and the effective dosages of vitamin D in human adipose tissue and its relevance in the associated diseases.

**Keywords: 1,25-dihydroxycholecalciferol or calcitriol, vitamin D binding protein, gene regulation, adipose tissue, adipogenesis, secretion, adipokines**

**Abbreviations:** 25(OH)D, 25-hydroxyvitamin D; 1,25(OH)2D3, 1,25 dihydroxycholecalciferol; VDBPs, Vitamin D binding proteins; VDR, Vitamin D receptor; DKK1, dickkopf 1; SFRP2, Frizzled-related protein 2; BMSCs, bone marrow stromal cells; PPARγ, peroxisome proliferator-activated receptor gamma; RXRα, retinoid X receptor alpha; WNT10, wingless-type MMTV integration site family member 10; C/EBP (α, β, and γ), CCAAT/enhancer-binding proteins (α, β, and γ); ETO, C/EBPβ corepressor eight twenty-one; FABP4, fatty acid binding protein 4; LPL, lipoprotein lipase; FASN, fatty acid synthase; SCD1, stearoyl-coA desaturase 1; GLUT4, glucose transporter type 4; PEPCK, phosphoenolpyruvate carboxykinase; LPS, lipopolysaccharide; TLR, toll like receptor; IL-6R, IL-6 receptors; NFκB, nuclear factor kappa-B; P38MAPK, p38 mitogen-activated protein kinase; IL-6, interleukin 6; TNF-α, tumor necrosis factor alpha; IL-1β, interleukin 1 beta; IκBα, inhibitor kappa-B; UCPs, uncoupling proteins; VDR <sup>−</sup>*/*−, VDR knock out; CYP27B1−*/*−, CYP27B1 knock out; BMI, Body mass index; BMPs, bone morphogenetic proteins; FGFs, fibroblast growth factors; TGFβ, transforminggrowth factor β; IGF1, insulin like growth factor 1; JAK-STAT3, janus kinase-signal

## **INTRODUCTION**

Adipose tissue is no longer regarded as a simple storage organ since it has been convincingly shown that it secretes more than 260 different proteins/peptides (Lehr et al., 2012). Lean people have about 5 kg of adipose tissue, while in obese and severely obese individuals the adipose tissue/organ could amount to 50 kg

transducer and activator of transcription 3; S6K1, ribosomal protein S6 kinase 1; WNT, wingless family; Rb, protein of retinoblastoma family; Pref1, preadipocyte factor 1; Necdin, melanoma-associated antigen family of proteins member; SREBP1, sterol regulatory binding protein 1; MSCs, mesenchymal stem cells; AP2, adipocyte-binding protein 2; MCP1, monocyte chemoattractant protein 1; CYP27B1,(25(OH)D)-1α-hydroxylase; CPTII, carnitine palmitoyltransferase II; WAT, White adipose tissue; VDREs, vitamin D response elements; ChIP-seq, chromatin immunoprecipitation—sequencing; LCLs, lymphoblastoid cell lines.

or more (Frankenfield et al., 2001). Excess in adipose tissue has been attributed to a variety of diseases including cancer, diabetes, cardiovascular and neurodegenerative diseases and decrease in life expectancy (Adams et al., 2006; Despres and Lemieux, 2006; Kahn et al., 2006; Van Gaal et al., 2006). Adiposity is one of the most serious public health problems, associated with vitamin D insufficiency due to the decreased bioavailability of vitamin D3 (Wortsman et al., 2000). The Institute of Medicine (IOM) recommended 25-hydroxyvitamin D (25(OH)D) levels as reliable biomarker for assessment of Vitamin D status; currently values ≤50 nmol/l or ≤20 ng/ml are considered inadequate or not sufficient and values ≥50 nmol/l or ≥20 ng/ml as adequate or sufficient (Ross et al., 2011) (**Figure 1**). 25(OH)D levels have been determined by a variety of methods yielding different results. The National Institutes of Health's Office of Dietary Supplements together with National Institute of Standards and Technology (NIST) therefore developed a standard reference material-972 (SRM-972) for accuracy of laboratory vitamin D measurements (Phinney et al., 2012). A recent study by the D-CarDia consortium employed a Mendelian randomization (MR) approach to establish causality and direction of the association between vitamin D status and obesity measured by body mass index (BMI) using information from 21 adult cohorts (up to 42,024 participants) (Vimaleswaran et al., 2013). The consortium found that a higher BMI was causally related to lower 25(OH)D; no evidence was obtained for a BMI lowering effect of higher 25(OH)D.

However, the study did not provide insights into the cellular action of 1,25(OH)2D3 (1,25-dihydroxycholecalciferol or calcitriol). While the knowledge of the effects of 1,25(OH)2D3 as an essential hormone and transcription factor is further emerging, it is increasingly acknowledged that 1,25(OH)2D3 down regulates inflammatory responses in the adipose tissue. The anti-inflammatory effects of 1,25(OH)2D3 might have notable influences on population health and disease prevention, since inflammation is thought to be the underlying cause of a range of metabolic disorders (Hotamisligil, 2006; Huotari and Herzig, 2008; Vlasova et al., 2010).

## **VITAMIN D AND ADIPOGENESIS**

Adipose tissue expansion is a remarkable process characterized by the enlargement of adipocyte size known as hypertrophy and by the increase in the number of adipocytes known as hyperplasia, which is more strongly associated with severity of obesity (Arner and Spalding, 2010). Both processes emerge through sequential stages of differentiation to form mature adipocytes; this process is called adipogenesis. Mesodermal cells are influenced by various signals like bone morphogenetic proteins (BMPs), fibroblast growth factors (FGFs), transforming-growth factor β(TGFβ) and insulin like growth factor 1 (IGF1) to form preadipocytes (Lowe et al., 2011). Furthermore, preadipocytes undergo differentiation to mature adipocytes by several intracellular signaling molecules (**Figure 2**) including janus kinase-signal transducer and activator

Nutrition Board. Dietary Reference Intakes for Calcium and Vitamin D. Washington, DC: National Academy Press, 2011.

of transcription 3 (JAK-STAT3) (Zhang et al., 2011), glutathione (Vigilanza et al., 2011), SMAD proteins (Jin et al., 2006) and ribosomal protein S6 kinase 1 (S6K1) (Carnevalli et al., 2010) affecting adipogenic transcription factors. In preadipocytes, differentiation factors need to be released from their suppressive signaling molecules such as members of wingless (WNT) family (Ross et al., 2000), protein of the retinoblastoma (Rb) family (Scime et al., 2005), preadipocyte factor 1 (Pref1) (Smas and Sul, 1993) and Necdin, member of the melanoma-associated antigen family of proteins (Fujiwara et al., 2012) to undergo differentiation. The terminal differentiation to mature adipocytes is regulated by a number of transcriptional factors including early key regulator CAAT/enhancer binding proteins (C/EBPβ followed by C/EBPα, C/EBPδ), the master regulator PPARγ and sterol regulatory binding protein 1 (SREBP1) (Payne et al., 2009; White and Stephens, 2010). These transcriptional factors induce expression of various genes related to lipogenesis, lipolysis and insulin sensitivity including fatty acid binding protein (*FABP4*), lipoprotein lipase (*LPL*), glucose transporter (*GLUT4*) and fatty acid synthase (*FASN*) (Lefterova et al., 2008; Nielsen et al., 2008; Madsen et al., 2014).

Investigations of the molecular regulation of 1,25(OH)2D3 on adipogenesis have been conducted *in vitro*. In mouse 3T3-L1 preadipocytes, 1,25(OH)2D3 inhibits adipogenesis by acting on multiple targets suppressing C/EBPα and PPARγ expression, specifically antagonizing the transacting activity of PPARγ, and sequestering the nuclear receptor retinoic X receptor (RXR), a member nuclear receptor superfamily and down regulating both C/EBP β mRNA expression and C/EBP β nuclear protein levels (Kong and Li, 2006) (**Figure 2**). 1,25(OH)2D3 stimulates expression of the C/EBP β corepressor, eight twenty-one (ETO), and thus further inhibits the action of any remaining C/EBP β transcriptional effects required for adipogenesis (Blumberg et al., 2006).

Although early studies have established an inhibitory action of 1,25(OH)2D3 in 3T3-L1 preadipocytes differentiation, recently, a more specific effect of 1,25(OH)2D3 on WNT signaling emerged. WNT/β-catenin maintain preadipocytes in their undifferentiated state and thus preventing adipogenesis (Ross et al., 2000). The anti-adipogenic effect of 1,25(OH)2D3 is mediated by maintenance of WNT10B and nuclear β-catenin levels expression levels in 3T3-L1 preadipocytes, thereby suppressing transcription factor PPARγ (Lee et al., 2012). In addition, 1,25(OH)2D3 also inhibited mouse bone marrow stromal cells(BMSCs) differentiation into adipocytes by suppression of dickkopf1 (DKK1) and secreted frizzled-related protein 2 (SFRP2) expression levels via VDR mediated WNT signaling (Cianferotti and Demay, 2007).

In contrast, 1,25(OH)2D3 treatment of porcine mesenchymal stem cells (MSCs) stimulated both proliferation and differentiation in a dose dependent manner toward adipocytic phenotype by increasing PPARγ, LPL and adipocyte-binding protein 2 (AP2) mRNA levels (Mahajan and Stahl, 2009). In human tissue, 1,25(OH)2D3 promotes differentiation of already committed subcutaneous preadipocytes through increased expression of adipogenic markers *FABP4* and *LPL* (Nimitphong et al., 2012). Narvaez et al. (2013) demonstrated that mesenchymal cells differentiate in the presence of 1,25(OH)2D3 toward adipocytes with an enhanced lipid accumulation and increased expression of adipogenic marker genes (*FASN*, *FABP4*, and *PPAR*γ).

In conclusion, 1,25(OH)2D3 regulates adipogenesis at various levels of the entire differentiation process (**Figure 2**). However, there are significant differences summarized in **Table 1**; the reasons for these differences are not clear at the moment methodological differences as well as physiological roles of the adipose tissue in different species in their environments might affect these processes. Further studies are needed to address the effects of vitamin D in adipose tissue and its relevance in the associated diseases.

## **VITAMIN D AND ADIPOSE TISSUE INFLAMMATION**

In obesity, adipose tissue undergoes hypertrophic enlargement, which results in an imbalanced blood flow leading to hypoxia, inflammation and macrophage infiltration (Goossens, 2008; Trayhurn, 2013). The hypertrophied adipocytes are characterized by a reduced secretion of adiponectin and increased secretion of several proinflammatory cytokines such as interleukin IL-6, IL-8, TNF-α, resistin and MCP1 (Wellen and Hotamisligil, 2003; Maury and Brichard, 2010; Vlasova et al., 2010).

1,25(OH)2D3 acts at several levels to modulate the function of the immune system (Lemire, 2000). Several *in vitro* studies in the mouse 3T3-L1 cell line and human adipocytes have demonstrated that 1,25(OH)2D3 inhibits chronic inflammation in adipose tissue (**Table 2**). However, earlier studies performed in 3T3-L1 and human adipocytes demonstrate contradictory results favoring inflammatory cytokine expression (Sun and Zemel, 2008); the reasons for the contradictory findings are unclear. Recent evidence focuses on the involvement of 1,25(OH)2D3 in the regulation of adipose tissue inflammation by reducing the proinflammatory cytokines secreted from adipose tissue.

In differentiated adipocytes from human subcutaneous white adipose tissue 1,25(OH)2D3 attenuates TNF-α induced MCP-1 secretion, while it inhibited secretion of adiponectin without affecting its mRNA levels (Lorente-Cebrian et al., 2012). In human subcutaneous adipose tissue fragments 1,25(OH)2D3 reduced IL-1β induced expression of the inflammatory genes MCP-1, IL-6 and IL-8. However, results from the cell culture experiments have not been consistent with the *in vivo* findings. In a randomized controlled trial including fifty-five obese subjects, oral supplementation of vitamin D 7000 IU per day over 26

### **Table 1 | Effect of 1,25(OH)2D3 on adipogenesis in different species.**


## **Table 2 | 1,25(OH)2D3 and inflammation.**


Mutt et al. Vitamin D and adipose tissue

weeks did neither affect inflammation markers in the circulation nor in the adipose tissue (Wamberg et al., 2013). In mice on high fat diet, dietary supplementation of 1,25(OH)2D3 (0.05 mg/kg of diet) reduced their IL-6 protein content in epididymal adipose tissue and in the 3T3-L1 cell line stimulated by LPS (Lira et al., 2011).

Signal transduction of inflammatory pathways in adipose tissue involves activation of NF-κB and translocation of p65 to nucleus mediated by degradation of IκBα (Tourniaire et al., 2013). Mutt et al. (2012) have demonstrated that, 1,25(OH)2D3 suppressed LPS-stimulated IL-6 secretion in human isolated mature and MSC differentiated adipocytes. This was confirmed by Marcotorchino et al. (2012), who demonstrated that 1,25(OH)2D3 inhibits the inflammatory markers in both mouse and human adipocytes via the involvement of p38 MAP kinase and NF-κB classical inflammatory pathway and later by Gao et al. (2013) and Ding et al. (2013a).

In summary, the presence of 1,25(OH)2D3 inhibited chemokine and cytokine secretion in human adipocytes. 1,25(OH)2D3 strongly inhibited the activation of the NF-κB and MAPK signaling pathways, which prevent gene transcription of the proinflammatory factors (**Figure 3**). 1,25(OH)2D3 has been shown by different groups in different models to significantly reduce inflammation in the adipose tissue. However, further studies are needed to provide more evidence for the physiological relevance and the concentration levels of active 1,25(OH)2D3 in lean and obese subjects required to ameliorate the inflammation and associated complications.

## **VITAMIN D AND ADIPOSE TISSUE ENERGY HOMEOSTASIS**

The discovery of VDR expression in adipocytes was the cornerstone for the investigations of the effect of vitamin D on adipose tissue beyond its role in bone metabolism (Stumpf, 1995; Ding et al., 2012). Recent findings in genetically modified mouse models highlighted a new role for vitamin D and its receptor VDR in adipose tissue energy homeostasis. VDR knockout (VDR−*/*−) mice had reduced body weight and lower serum leptin concentrations, despite of an increased compensatory food intake compared to wild type mice of different genetic background C57BL6 and CD1. VDR−*/*<sup>−</sup> mice were highly resistant to high fat diet induced weight gain (Narvaez et al., 2009). In addition, these mice are characterized by a relatively short lifespan, alopecia, osteoporosis, ectopic calcification, progressive loss of hearing and balance (Keisala et al., 2009; Tuohimaa, 2009). Mice lacking CYP27B1 [the (25(OH)D)-1α-hydroxylase enzyme, converts 25(OH)D3 in to 1,25(OH)2D3], displayed features similar to VDR−*/*<sup>−</sup> with reduced body weight, hypoleptinemia and hyperphagia. Interestingly, uncoupling protein 1 (UCP-1) expression in white adipose tissue of the VDR−*/*<sup>−</sup> mice was increased 25-fold.

In addition to reduced body weight, VDR−*/*<sup>−</sup> mice had less body fat and lower levels of plasma triglycerides and cholesterol in comparison to the wild type counterparts even though mice were challenged with a high fat diet (Wong et al., 2009; Weber and Erben, 2013). The depletion of adipose tissue in younger VDR−*/*<sup>−</sup> mice progresses with aging and resulted in severe mammary adipose tissue atrophy, along with the increased respiration and energy expenditure (Welsh et al., 2011). The effect on plasma lipid profile and unaltered food intake in these mice was confirmed by an increased β-oxidation rate in isolated adipocytes mediated by the induction of carnitine palmitoyltransferase II (CPTII) (**Figure 3**). VDR−*/*<sup>−</sup> mice had an increased basal metabolism demonstrated by the total energy expenditure, oxygen consumption and CO2 production in comparison with the wild type mice (Wong et al., 2009). In addition, UCP1, UCP2, and UCP3 mRNAs were upregulated in brown adipose tissue of the VDR−*/*<sup>−</sup> mice fed high fat diet. In contrast to VDR knock out models with the

ablation of the receptor in the whole animal, adipose tissue specific overexpression of human VDR via the adipocyte fatty acid binding protein (aP2) promoter/enhancer element resulted in a decreased energy expenditure and oxygen consumption and thus the mice had an increased body weight and fat mass (Wong et al., 2011).

In conclusion, these transgenic animal models indicate a critical and complex role for 1,25(OH)2D3 and VDR signaling in energy homeostasis. However, notwithstanding the cell and mouse studies, further studies need to explore the role of vitamin D on human adipose tissue metabolism *in vivo*.

## **GENETIC VIEW ON THE ACTIONS OF VDR IN ADIPOCYTES: INTEGRATION WITH OTHER TISSUES**

The VDR genomic interactions in different types of cells and tissues have been mapped by *in vitro* experiments where target cells (primary or secondary) have been treated with 1,25(OH)2D3. Upon stimulation of VDR by its ligand, it forms a heterodimer with RXR and subsequently binds to the vitamin D response elements (VDREs) within the regulatory regions of target genes. The abundance of VDR binding sites and the regulation of changes in gene expressions are analyzed using array technology and the combination of chromatin immunoprecipitation (ChIP) with massive parallel sequencing (ChIP-seq). These advanced techniques have provided novel mechanistic insights of 1,25(OH)2D3 action via VDR in the regulation of cellular metabolism and disease states. However, studies on genome-wide actions of VDR in adipocytes are sparse.

Recent microarray studies of human adipocytes and preadipocytes incubated with macrophage-conditioned medium derived from U937 monocytes, confirmed the induction of genes associated with the metabolism and action of 1,25(OH)2D3, including CYP27B1 and VDR (Trayhurn et al., 2011). An earlier single microarray study in human subcutaneous adipose tissue derived preadipocytes differentiated to adipocytes demonstrated 237 1,25(OH)2D3 responsive genes (cell proliferation, angiogenesis, cell cycle, inflammation and response to oxidative stress) (Sun et al., 2008).

Most recent studies in the other cell types such as monocytes, primary CD4+ T-lymphocytes, adenocarcinoma, hepatic stellate and lymphoblastoid cell lines (LCLs) (Ramagopalan et al., 2010; Heikkinen et al., 2011; Meyer et al., 2012; Ding et al., 2013b; Handel et al., 2013; Tuoresmäki et al., 2014) contribute to a systemic understanding of 1,25(OH)2D3 induced gene regulation. Depending on the cell type, concentration and length of 1,25(OH)2D3 incubation approximately 2000 VDR genomic binding sites have been found in these studies. Yet, alterations in DNA accessibility in cell lines after short-term stimulation with 1,25(OH)2D3 may not reflect the physiological 1,25(OH)2D3 levels *in vivo* due to the different tissue environment and sympathetic influence. In primary CD4+ lymphocyte cells, isolated from nine healthy individuals with measured serum 25(OH)D levels, VDR binding sites ranged from 200 to 7118 across the genome and the corresponding 25(OH)D levels directly correlated with the number of VDR binding sites, suggesting far greater number of VDR binding sites in 1,25(OH)2D3 sufficient than the insufficient subjects (Handel et al., 2013).

Genome-wide VDR cistromes are not available in adipocytes, but recent VDR binding sites in other cell types has been mapped with ChIP-seq from both upstream and downstream of the transcription start site. Further genome wide view actions of VDR in adipocytes as well as integration of other tissue specific cell types are warranted.

## **CONCLUSION AND FUTURE DIRECTIONS**

Adipose tissue acts in addition to nutrient storage as an active endocrine organ. In the obese state, sub-clinical inflammation increases the risk of a variety of chronic diseases. Vitamin D deficiency is common in overweight and obese individuals, and it is possible that lower circulating concentrations may contribute to increases in metabolic risk. A genome-wide association study of 25(OH)D concentrations in 33996 individuals of European descent from 15 cohorts found variants near genes involved in cholesterol synthesis, hydroxylation, and vitamin D transport affect vitamin D status (Wang et al., 2010). Genetic variation at these loci identifies individuals who have a substantially increased risk of vitamin D insufficiency.

On the cellular level, 1,25(OH)2D3 has a significant role in adipogenesis and inflammation which might be species dependent. Holick et al. (1989) demonstrated that the peak circulating concentrations of 25(OH)D in the elderly are about 30% of that of the young. These findings suggest that there will be significant challenges in the translation of the finding from models and non-human primates to the targeted human populations (healthy, diseased, black, white, age, BMI, geographical latitude, race). More evidence accumulates that one dose does not fit all (Powe et al., 2014). Powe and colleagues evaluated vitamin D binding proteins (VDBP) and 25(OH)D levels in black and white Americans. Black adult Americans had low 25(OH)D levels and with the threshold of 20 or 30 ng/ml, 77–96% of them would be classified as vitamin D deficient. Surprisingly, the black study participants had higher bone mineral density, higher calcium levels and only slightly higher parathyroid levels than the white study participants due to VDBP gene polymorphisms (rs7041 and rs4588). The authors speculated that the low levels of VDBP might protect against the adverse effects of vitamin D deficiency. Sufficient levels of this essential hormone and the development of potent novel vitamin D receptor analogs (Peräkylä et al., 2005; Leyssens et al., 2014), which could be easily and cheaply substituted, are beneficial in the maintenance of health and prevention of a number of diseases associated with vitamin D deficiency. Recent systemic review and meta-analysis summary of observational studies and randomized interventions investigated the association between the circulating 25(OH)D concentrations and cause specific mortality in 900,000 subjects in 26 countries (Chowdhury et al., 2014). There was an inverse association of mortality risk and vitamin D levels, yet the observed association could be direct [suboptimal 25(OH)D concentrations] or indirect through higher BMI or disadvantageous social circumstances. Thus, prospective intervention studies are needed to establish potential causal associations between vitamin D levels and disease outcomes.

## **ACKNOWLEDGMENT**

This work was support in part by the grants from the Finnish Academy project 139900, Finnish Cultural Foundation, North Ostrobothnia Regional fund.

## **REFERENCES**


in white adipose tissue. *Endocrinology* 150, 651–661. doi: 10.1210/en.20 08-1118


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

*Received: 16 March 2014; accepted: 02 June 2014; published online: 24 June 2014. Citation: Mutt SJ, Hyppönen E, Saarnio J, Järvelin M-R and Herzig K-H (2014) Vitamin D and adipose tissue—more than storage. Front. Physiol. 5:228. doi: 10.3389/ fphys.2014.00228*

*This article was submitted to Integrative Physiology, a section of the journal Frontiers in Physiology.*

*Copyright © 2014 Mutt, Hyppönen, Saarnio, Järvelin and Herzig. 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.*

## Vitamin D in inflammatory diseases

## *Thea K. Wöbke , Bernd L. Sorg and Dieter Steinhilber\**

*Institute of Pharmaceutical Chemistry, Goethe University Frankfurt, Frankfurt, Germany*

#### *Edited by:*

*Carsten Carlberg, University of Eastern Finland, Finland*

#### *Reviewed by:*

*Martin Hewison, University of California, Los Angeles, USA Martin Tenniswood, University at Albany, USA*

#### *\*Correspondence:*

*Dieter Steinhilber, Institute of Pharmaceutical Chemistry, Goethe University Frankfurt, Max-von-Laue-Str. 9, Frankfurt, D-60438, Germany e-mail: steinhilber@em. uni-frankfurt.de*

Changes in vitamin D serum levels have been associated with inflammatory diseases, such as inflammatory bowel disease (IBD), rheumatoid arthritis, systemic lupus erythematosus, multiple sclerosis (MS), atherosclerosis, or asthma. Genome- and transcriptome-wide studies indicate that vitamin D signaling modulates many inflammatory responses on several levels. This includes (i) the regulation of the expression of genes which generate pro-inflammatory mediators, such as cyclooxygenases or 5-lipoxygenase, (ii) the interference with transcription factors, such as NF-κB, which regulate the expression of inflammatory genes and (iii) the activation of signaling cascades, such as MAP kinases which mediate inflammatory responses. Vitamin D targets various tissues and cell types, a number of which belong to the immune system, such as monocytes/macrophages, dendritic cells (DCs) as well as B- and T cells, leading to individual responses of each cell type. One hallmark of these specific vitamin D effects is the cell-type specific regulation of genes involved in the regulation of inflammatory processes and the interplay between vitamin D signaling and other signaling cascades involved in inflammation. An important task in the near future will be the elucidation of the regulatory mechanisms that are involved in the regulation of inflammatory responses by vitamin D on the molecular level by the use of techniques such as chromatin immunoprecipitation (ChIP), ChIP-seq, and FAIRE-seq.

**Keywords: 1α,25(OH)2D3, VDR, cyclooxygenase, NFκB, NFAT, MKP1, interleukins, innate immune system**

## **INTRODUCTION: 1α,25(OH)2D<sup>3</sup> AND INFLAMMATORY DISEASES**

It is now well established that the physiological importance of the vitamin D status extends far beyond the regulation of bone metabolism. According to its manifold functions in immune homeostasis, increasing evidence relates serum vitamin D levels as well as polymorphisms in enzymes involved in vitamin D metabolism to the incidence of chronic inflammatory diseases like asthma, atherosclerosis and autoimmune diseases (Stojanovic et al., 2011; Summerday et al., 2012; Szekely and Pataki, 2012). However, whether vitamin D exerts a salutatory or deteriorating role in such diseases is still under debate. This review will focus on the knowledge regarding the role of vitamin D in inflammatory diseases by the examples of asthma, atherosclerosis and autoimmune diseases.

## **1α,25(OH)2D<sup>3</sup> AND ASTHMA**

According to the World Health Organization (WHO), asthma is the most common chronic disease among children (http:// www*.*who*.*int/mediacentre/factsheets/fs307/en/index*.*html). In this context, several studies addressed the interrelationship of the maternal as well as infant vitamin D status and the prevalence and severity of asthma. Three studies by Brehm et al. analyzed the relationship between vitamin D levels and asthma severity in Costa Rican, North American and Puerto Rican children, respectively (Brehm et al., 2009, 2010, 2012). Collectively, they found high prevalences of vitamin D insufficiency in asthmatic children and vitamin D insufficiency was correlated with severe asthma exacerbations. However, the prevalence of vitamin D insufficiency was high in Puerto Rican children irrespective of the indisposition from asthma, with roughly comparable percentages between asthma patients and otherwise healthy children (Brehm et al., 2012). Although few studies showed no correlation between serum vitamin D levels and the presence of asthma (Menon et al., 2012; Gergen et al., 2013), many studies state a higher prevalence of vitamin D deficiency in asthmatic children (Freishtat et al., 2010; Chinellato et al., 2011a,b; Ehlayel et al., 2011; Hollams et al., 2011; Bener et al., 2012; Krobtrakulchai et al., 2013) and adults (Li et al., 2011b). Additionally, in many cases a relation between low vitamin D levels and reduced asthma control is found. Furthermore, metabolomic analysis of breath condensates revealed reduced levels of vitamin D metabolites in children with asthma (Carraro et al., 2013). Similarly, enhanced vitamin D binding protein levels were found in bronchoalveolar lavage fluid of asthmatic children (Gupta et al., 2012b). Interestingly, one study describes an age-dependent association between serum vitamin D level and asthma prevalence in children (Van Oeffelen et al., 2011).

A different relationship between the vitamin D status and asthma has been brought up by a northern Finland birth cohort study, which revealed an increased risk of asthma in adults who received high dose vitamin D supplementation in their childhood (Hypponen et al., 2004). In accordance with these findings, a prospective study by Tolppanen et al. revealed an increased risk of wheezing in association with higher vitamin D levels, but no correlation of lower vitamin D levels to respiratory sicknesses (Tolppanen et al., 2013). Another study reinforces the finding of increased susceptibility to asthma after vitamin D supplementation, yet only regarding supplementation of water soluble formulations and not in connection with vitamin D supplementation in peanut oil (Kull et al., 2006).

There is debate as to whether maternal vitamin D levels during the pregnancy influence the susceptibility to asthma of the progeny. Whereas some reports showed no correlation between maternal or cord blood vitamin D levels and an increased risk of childhood asthma (Camargo et al., 2011; Rothers et al., 2011; Morales et al., 2012; Pike et al., 2012; Magnus et al., 2013), another report indicates that high maternal vitamin D levels correlate with enhanced probability of asthma development in children (Gale et al., 2008). In contrast, some reports associate higher vitamin D intake during pregnancy with reduced risk of childhood wheezing and asthma (Camargo et al., 2007; Devereux et al., 2007; Erkkola et al., 2009).

Mechanistically, vitamin D induced protection against airway inflammation has been related to a modulated T cell response to allergens as well as induction of the immunoglobulin-like antiinflammatory cell surface protein CD200 on T cells, that acts on target immune cells which express the CD200 receptor (CD200R) (Dimeloe et al., 2012; Gorman et al., 2012; Urry et al., 2012). Many authors suggest that the beneficial effect of sufficient vitamin D levels on asthma development results from the immune enhancing effect of vitamin D and the simultaneous prevention of respiratory infections (Ginde et al., 2009; Camargo et al., 2011; Majak et al., 2011; Morales et al., 2012).

Furthermore, there is evidence that the serum vitamin D level has also an influence on asthma therapy, as vitamin D has been demonstrated to enhance glucocorticoid (GC) action and lower serum vitamin D levels are associated with higher corticosteroid requirement, at least in children, or even therapy-resistance (Searing et al., 2010; Goleva et al., 2012; Gupta et al., 2012a; Wu et al., 2012). Additionally, the therapeutic effect of specific allergen immunotherapy has been correlated to serum vitamin D levels (Majak et al., 2012).

Besides serum vitamin D levels also polymorphisms of genes of the vitamin D pathway such as the vitamin D receptor (VDR) have been associated with asthma (Poon et al., 2004; Raby et al., 2004; Saadi et al., 2009; Li et al., 2011a; Pillai et al., 2011; Maalmi et al., 2013), yet, not all studies revealed a correlation between vitamin D pathway polymorphisms and asthma prevalence (Vollmert et al., 2004; Fang et al., 2009).

## **1α,25(OH)2D<sup>3</sup> AND ATHEROSCLEROSIS**

Another chronic inflammatory disease that is more prevalent in the elderly population is atherosclerosis. Early studies on atherosclerosis development in several animal models revealed an accelerating effect of high doses of vitamin D. Vascular calcification was observed in some of these studies, but not all (Zemplenyi and Mrhova, 1965; Kudejko, 1968; Taura et al., 1979; Kunitomo et al., 1981; Toda et al., 1983, 1985). Moreover, 1α,25 dihydroxyvitamin D3, the active form of vitamin D, stimulated vascular calcification by *in vitro* by reducing the expression of parathyroid hormone-related peptide as well as stimulating alkaline phosphatase activity in bovine vascular smooth muscle cells (Jono et al., 1998). On the other hand, there is a large body of research from clinical studies in humans indicating that low levels of serum 25-hydroxy vitamin D are associated with atherosclerosis (Reis et al., 2009; Carrelli et al., 2011; Shanker et al., 2011; Cheraghi et al., 2012). In line with this, the incidence of osteoporosis, a disease known to be related to vitamin D inadequacy, correlates with the incidence of atherosclerosis (Stojanovic et al., 2011). Therefore, different mechanisms may account for the promotion of atherogenesis by high and low vitamin D levels, respectively, and calcification may be crucial in the case of hypervitaminosis. Moreover, differences between the animal and human system may account for the conflicting results.

With respect to atherogenesis, 1α,25-dihydroxyvitamin D3 has been demonstrated to reduce macrophage adhesion and migration as well as foam cell formation in monocytes isolated from type 2 diabetic patients (Oh et al., 2012; Riek et al., 2013a,b). Mechanistic investigations in the context of these studies attributed the beneficial effects of vitamin D to a reduction of endoplasmatic reticulum stress in macrophages. This has been investigated in two mouse models, where vitamin D deficiency facilitated atherosclerosis, which could be reversed in the course of macrophage endoplasmatic reticulum stress suppression (Weng et al., 2013). Further evidence on beneficial effects of calcitriol treatment on atherosclerosis development has been obtained from an investigation with apolipoprotein E knock-out mice. In this study, oral calcitriol treatment decreased the production of proinflammatory chemokines, led to a reduced amount of inflammatory effector cells in atherosclerotic plaques and simultaneously increased amounts of regulatory T cells (Takeda et al., 2010). A similar link between vitamin D, T cell modulation, and atherosclerosis has also been established in humans with chronic kidney disease (CKD) (Yadav et al., 2012).

The renin-angiotensin-system is known for its detrimental effects on the cardiovascular system and has been shown to play an important role in the development of atherosclerosis. Interestingly, numerous studies in mice document that vitamin D signaling suppresses the renin-angiotensin-system and that vitamin D deficiency is associated with an increased activity of the renin-angiotensin-system (Li et al., 2002; Zhou et al., 2008; Szeto et al., 2012; Weng et al., 2013). Moreover, the inverse associations which are described for vitamin D and the occurrence of inflammatory cytokines, C-reactive protein, and adhesion molecules suggest a inhibitory role for vitamin D in the genesis of atherosclerosis (Brewer et al., 2011). Additionally, there is experimental evidence that vitamin D reduces the expression of matrix metalloproteinases that are involved in vascular calcification (Nakagawa et al., 2005; Qin et al., 2006).

However, there are also studies that found no evidence for an association between low vitamin D and atherosclerosis in patients suffering from different autoimmune diseases (Mok et al., 2012; Sachs et al., 2013). Similarly, there was no evidence for an association of *Bsm*I polymorphism, an intronic single nucleotide variation of the VDR gene, with atherosclerosis (El-Shehaby et al., 2013). Yet, it has been shown that atherosclerosis in monkeys is associated with low levels of VDR expression in coronary arteries even in the presence of higher plasma vitamin D concentrations (Schnatz et al., 2012a,b). Moreover, the activation of vitamin D can occur locally in macrophages that infiltrate atherosclerotic lesions and local vitamin D response might thus not necessarily correlate with serum vitamin D levels (Richart et al., 2007).

## **1α,25(OH)2D<sup>3</sup> AND AUTOIMMUNE DISEASES**

It is well established that vitamin D plays an important role in the regulation of immune functions (Schwalfenberg, 2011; Zhang et al., 2013a). Accordingly, several inflammatory autoimmune diseases like rheumatic disorders and type 1 diabetes have been associated with vitamin D deficiency (Adorini and Penna, 2008; Shapira et al., 2010). Inflammatory processes in the central nervous system are a hallmark of the autoimmune disease multiple sclerosis (MS) (Deckx et al., 2013). Several studies indicate that MS patients have lower levels of vitamin D and that higher vitamin D levels as well as vitamin D supplementation have a protective effect against MS (Munger et al., 2004, 2006; Ozgocmen et al., 2005). Moreover, vitamin D levels have been shown to vary in concordance with MS exacerbations (Correale et al., 2009) and it is possible that low vitamin D levels are rather a consequence of the sun avoidance of MS patients and not a direct cause of the disease (Munger et al., 2006). Regarding the effectiveness of vitamin D supplementation in the course of MS treatment, there are studies in mice and humans that suggest a beneficial effect of treatment (Goldberg et al., 1986; Wingerchuk et al., 2005; Pedersen et al., 2007; Burton et al., 2010). Interestingly, a gender specific effect of vitamin D has been demonstrated in mice and humans, which points to greater effects of vitamin D in females (Spach and Hayes, 2005; Correale et al., 2010).

Overall, there have been only a few controlled trials documenting the outcome of vitamin D supplementation on disease activity in rheumatic conditions, and the role of vitamin D in rheumatoid arthritis is therefore controversially discussed (Gatenby et al., 2013). Yet, a metaanalysis of observational studies on the vitamin D intake and vitamin D serum levels suggests an inverse association with rheumatoid arthritis (Song et al., 2012). Additionally, *in vitro* experiments with macrophages from healthy donors and rheumatoid arthritis patients indicate an enhanced antiinflammatory potential of vitamin D in macrophages from the latter group (Neve et al., 2013).

It has been shown that the onset of autoimmunity in type 1 diabetes is preceded by a proinflammatory metabolic serum profile (Knip and Simell, 2012). Concurrently, a study in Italian children revealed reduced vitamin D serum levels in children at the onset of type 1 diabetes compared to children hospitalized for other reasons (Franchi et al., 2013). In conformity with these findings, metaanalyses suggest an association between vitamin D intake in early life and susceptibility for type 1 diabetes (Zipitis and Akobeng, 2008; Dong et al., 2013).

For inflammatory bowel disease (IBD), another autoimmune disorder, similar associations to that described above regarding vitamin D status and sunlight exposure have been reported (Garg et al., 2012; Ananthakrishnan, 2013). Animal studies in vitamin D deficient and VDR knockout (KO) mice reveal a dysregulation of T cells that might be of importance in the pathogenesis of IBD (Ooi et al., 2012).

In summary, there is considerable evidence for an association between vitamin D deficiency and inflammatory diseases. However, regarding the causality of this association and the benefit of vitamin D supplementation, only limited information is available and the existing data are still inconsistent.

## **INTERFERENCE OF 1α,25(OH)2D<sup>3</sup> WITH PRO-INFLAMMATORY TRANSCRIPTION FACTORS AND SIGNALING PATHWAYS**

Cell type specific up-regulation of proinflammatory genes and down-regulation of anti-inflammatory genes is a hallmark of the onset of an inflammatory reaction. Depending on the cell type, up-regulation of certain cytokines or enzymes which generate mediators of inflammation can occur at the transcriptional or posttranscriptional level. In addition, there is considerable crosstalk between various pathways which allows adaptation of the host defense reactions to the environment. According to their functions, the regulators of inflammatory reactions can be receptors such as toll like receptors, signal transducers as well as transcription factors which translate the activation of certain signal cascades into gene transcription. Additionally, regulation of gene expression during inflammatory processes can also occur on posttranslational level which is not focus of this review.

At the level of intracellular signal transduction, MAP kinases such as JNK or p38 have been identified as central signal transducers of inflammatory signals. Interestingly, it has been observed that there is a cross talk between VDR/RXR and MAP kinase signaling on many levels and the outcome, e.g., stimulation or inhibition, depends on the stimulus, cell type and the response (Miodovnik et al., 2012). Regarding inflammation, it has become obvious that vitamin D inhibits production of proinflammatory cytokines like IL-6 or TNFα in monocytes via the inhibition of p38 MAP kinase (Zhang et al., 2012). Inhibition of p38 in monocytes was found to be due to induction of MAPK phosphatase-1 (MKP1) which dephosphorylates p38 and thus reduces p38 activation (**Figure 1**). A similar mechanism was found in prostate cells where induction of MKP5 by 1α,25(OH)2D3 was responsible for down-regulation of IL-6 mRNA expression (Nonn et al., 2006). 1α,25(OH)2D3 increases MKP5 transcription by induction of VDR/RXR binding to a VDRE in the MKP5 promoter. Beside this indirect modulation of signaling cascades, 1α,25(OH)2D3 and its receptor complex VDR/RXR can interact with other transcription factors such as NF-κB, nuclear factor of activated T-cells (NFAT), or the glucocorticoid receptor (GCR) which leads to anti-inflammatory effects (**Figure 2**). Activation of VDR inhibits NF-κB activation and signaling. NFκB is a ubiquitously expressed transcription factor which represents a heterodimer. In the inactive state it interacts with IκB which keeps it in the cytosol (Karin and Lin, 2002). Upon cell activation by proinflammatory stimuli, IκB is phosphorylated and subsequently ubiquitinylated, which leads to proteasomal degradation of the IκB protein. Free NFκB translocates to the nucleus where it activates transcription of proinflammatory cytokines, antiapoptotic factors as well as of enzymes involved in the generation of proinflammatory mediators such as COX-2 (Karin and Lin, 2002; Tsatsanis et al., 2006). It has been shown that 1α,25(OH)2D3 down-regulates NF-κB levels in lymphocytes (Yu et al., 1995) and that the vitamin D analog TX 527 prevents NF-κB activation in monocytes (Stio et al., 2007). Inhibition of NFκB activation

by 1α,25(OH)2D3-mediated up-regulation of IκB expression was reported in human peritoneal macrophages (Cohen-Lahav et al., 2006) (**Figure 2**). Additionally, interference of vitamin D signaling with DNA binding of NFκB was found (Harant et al., 1998). It was shown that 1α,25(OH)2D3 inhibits NF-κB activity in human MRC-5 fibroblasts but not translocation of its subunits p50 and p65. The partial inhibition of NFκB DNA binding by 1α,25(OH)2D3 was dependent on de novo protein synthesis, suggesting that 1α,25(OH)2D3 may regulate expression of cellular factors which contribute to reduced DNA binding of NFκB (Harant et al., 1998). Thus, it seems that vitamin D is able to inhibit NFκB activation as well DNA binding (**Figure 2**).

Another interesting target for the anti-inflammatory signaling of vitamin D is NFAT (**Figure 2**). This transcription factor is activated by dephosphorylation by calcineurin which leads to translocation of this protein and transcriptional activation of proinflammatory genes such as interleukin 2 and cyclooxygenase-2 (Duque et al., 2005; Muller and Rao, 2010). In T-lymphocytes, it was shown for the interleukin 2 promoter that VDR-RXR heterodimers bind to an NFAT binding site and thus inhibit NFAT activity (Takeuchi et al., 1998). Similar data were obtained for interleukin 17 where 1α,25(OH)2D3 blocked NFAT activity which contributed to repression of interleukin17A expression in inflammatory CD4+ T cells by the hormone (Joshi et al., 2011).

Another interesting finding was that vitamin D enhances the anti-inflammatory activities of GCs (**Figure 1**). The strong anti-inflammatory activities of GCs are mediated by the GCR. It belongs to the nuclear receptor family. Upon ligand binding the receptor dimerizes and translocates into the nucleus where it binds to GC-responsive elements (GRE) and modulates gene expression (Barnes, 1998). In general, GCs downregulate expression of pro-inflammatory genes and up-regulate anti-inflammatory genes. It was found in asthmatics that dexamethasone-induced MKP-1 expression as a marker for GC responsiveness is significantly increased when serum vitamin D levels increase suggesting that vitamin D may enhance GC responsiveness (Sutherland et al., 2010). It is interesting to note that MKP-1 is also a vitamin D target gene as mentioned above (Zhang et al., 2012). Vitamin D enhancement of GC-induction of MKP1 was abolished both in purified CD14+ and CD14− cells and it was found that the synergism depends on vitamin D-induced GM-CSF release from CD14− cells and GM-CSFdependent MED14 induction in CD14+ cells (Zhang et al., 2013b). MED14 is part of the mediator complex involved in the regulation of transcriptional initiation and it was found to form a complex with VDR and mediate ligand-dependent enhancement of transcription by the VDR (Rachez et al., 1999) (**Figure 1**). Interestingly, MED14 also enhances gene activation by the GCR in a gene-specific manner (Chen et al., 2006). For MKP1 it was found in human monocytes that VDR and GCR bind to a corresponding VDRE and two GREs after ligand stimulation (**Figure 1**). After GM-CSF treatment, MED14 was recruited to the promoter after addition of 1α,25(OH)2D3 but not dexamethasone indicating that MED14 recruitment depends on the VDR (Zhang et al., 2013b). 1α,25(OH)2D3 enhanced the binding of the GCR to the GRE in close proximity to the VDRE in the presence of GM-CSF and ChIP analysis suggest a MED14-VDR-GCR complex at the MKP1 promoter with bridges the crosstalk between vitamin D and GCs (Zhang et al., 2013b). The data from single gene analyses such as MKP1 suggest that the VDR interacts with other signaling pathways.

At present there are genome-wide data available from immortalized lymphoblastoid cell lines (Ramagopalan et al., 2010), undifferentiated and LPS stimulated THP-1 cells (Heikkinen et al., 2011; Tuoresmäki et al., 2014), LS180 colorectal cancer cells (Meyer et al., 2012) and LX2 hepatic stellate cells (Ding et al., 2013). These six ChIP-seq data sets showed 21,776 nonoverlapping VDR binding sites whereas only 54 sites were common in all six data sets. The data suggest that, apart from a few sites, VDR binding is strongly cell and stimulus specific. In the non-overlapping binding sites, only 17.5% contain a DR3-type VDRE whereas the percentage of DR3-type response elements is enriched in highly ligand-responsive loci. All these data suggest that the VDR interacts with other transcription factors and that these interactions might only be in part ligand dependent. Regarding inflammation, the genome-wide effects of LPS on VDR location in THP-1 cells are of special interest (Tuoresmäki et al., 2014). From the 805 VDR binding sites, only 462 overlap in untreated and LPS-treated THP cells which were stimulated with 1α,25(OH)2D3. Thus, LPS treatment leads to a considerable change in VDR location. In THP-1 cells, bioinformatic searches for shared binding sites revealed motifs for CEBP1, PU.1 in stimulated THP-1 cells whereas NFYA, LHX3-like and NANOG were found for unstimulated cells but no transcription factor has been identified in conjunction with LPS treatment. Of note, binding

sites for JUN, a component of the AP1 transcription factor, were found to be enriched at VDR loci in LX2 hepatic stellate cells. This is of interest regarding inflammation as AP1 is known to be a transcription factor that regulates expression of many proinflammatory genes. At present, there are many data available on single gene levels but there is still a missing link between these data and the genome-wide observations. Since VDR signaling seems to be strongly cell type and stimulus-dependent, more genomewide data with different cell types and stimuli are required to understand the mechanisms how 1α,25(OH)2D3 modulates gene expression under inflammatory conditions.

## **REGULATION OF THE EXPRESSION OF PROINFLAMMATORY ENZYMES BY 1α,25(OH)2D<sup>3</sup>**

Arachidonic acid derived eicosanoids which comprise prostaglandins and leukotrienes play an important role in inflammatory processes (Harizi et al., 2008). Of the enzymes involved in prostaglandin synthesis, cyclooxygenase-2 (COX-2) and microsomal prostaglandin E synthase 1 (mPGES-1) have been shown to be induced in many inflammatory conditions (Tomasoni et al., 1998; Murakami et al., 2000; Cipollone and Fazia, 2006; Petrovic et al., 2006) and inhibition of both enzymes is a common approach in the treatment of inflammatory diseases (Fahmi, 2004; Ramalho et al., 2009; Dallaporta et al., 2010).

In prostate cancer cells it has been demonstrated that 1α,25(OH)2D3 inhibits the expression of COX-2 on mRNA and protein level as well as the expression of prostaglandin receptors on mRNA level and simultaneously upregulates prostaglandin catabolism via 15-hydroxyprostaglandin dehydrogenase (Moreno et al., 2005). In addition, the combination of calcitriol with COX-inhibitors led to synergistic growth inhibition (Moreno et al., 2005). Similar results were obtained with the combination of 1α,25(OH)2D3 and COX-inhibitors in different leukemia cells (Jamshidi et al., 2008). In accordance with the previous findings, treatment with the vitamin D analog elocalcitol resulted in decreased COX-2 expression and diminished PGE2 synthesis in prostate cells (Penna et al., 2009). The COX-2/PGE2-pathway was also identified as the mediator of the growth inhibitory effect of calcitriol in breast cancer cells (Yuan et al., 2012). Furthermore, COX-2 upregulation in placental trophoblasts in response to oxidative stress and in myometrial cells in response to interleukin-1β was inhibited by 1α,25(OH)2D3 (Sun et al., 2013; Thota et al., 2013).

Thill et al. found correlations between VDR expression and expression of COX-2 as well as 15-hydroxy PG dehydrogenase in malignant breast cells and in cells from female reproductive tissues (Thill et al., 2009, 2010, 2012).

In human lung fibroblasts inhibition of PGE2-production by vitamin D was found which was not due to altered COXexpression. Yet, vitamin D inhibited IL-1β-induced mPGES-1 expression and simultaneously stimulated 15-hydroxy PG dehydrogenase (Liu et al., 2014).

5-lipoxygenase (5-LO) accounts for the first two steps in leukotriene biosynthesis. Leukotrienes exert potent proinflammatory actions and have been associated with several chronic inflammatory diseases (Haeggstrom and Funk, 2011).

In the myeloid cell line HL-60, treatment with 1α,25(OH)2D3 triggers differentiation into monocytic cells. Simultaneously, 1α,25(OH)2D3 has been shown to induce 5-LO expression on mRNA and protein level as well as to increase 5-LO enzyme activity (Bennett et al., 1993; Brungs et al., 1994). A similar effect was also observed in the monocytic cell line Mono Mac 6. Additionally, this effect was strongly enhanced by the combination of 1α,25(OH)2D3 with transforming growth factor β (TGF-β) (Brungs et al., 1995; Harle et al., 1998). Mechanistically, the effect of 1α,25(OH)2D3 on 5-LO expression was related to VDR binding sites in the 5-LO promoter and distal parts of the 5-LO gene (Sorg et al., 2006; Stoffers et al., 2010) and is due to stimulation of 5-LO transcript elongation (Stoffers et al., 2010).

Previous results suggest a modulatory role of vitamin D in the inflammatory response of cells of the monocyte/macrophage lineage, which is again modulated by TGF-β. In this context, it is interesting that macrophages contain 1α-hydroxylase and therefore are capable of autocrine or paracrine activation of vitamin D (Lagishetty et al., 2011). Moreover, in keratinocytes autocrine TGF-β production is induced by vitamin D (Kim et al., 1992). Crucial participation of monocytes/macrophages in diverse inflammatory processes has been demonstrated (Cutolo, 1999; Yoon and Jun, 1999; Moore et al., 2013). Besides induction of 5-lipoxygenase, the combination of TGF-β and 1α,25(OH)2D3 has been shown to induce the differentiation antigen CD69 in monocytic cells (Wobke et al., 2013). Overexpression of CD69 again, has been shown in the context of local dermal inflammation, systemic lupus erythematosus, hyperthyroid Graves' disease and autoimmune thyroiditis (Fernandez-Herrera et al., 1996; Portales-Perez et al., 1997; Crispin et al., 1998; Gessl and Waldhausl, 1998).

## **1α,25(OH)2D<sup>3</sup> AS REGULATOR OF CYTOKINE GENE EXPRESSION, PROTEIN PRODUCTION/RELEASE AND SIGNALING**

## **TGF-β AND Smad SIGNALING IN INFLAMMATION AND THE INFLUENCE OF 1α, 25(OH)2D<sup>3</sup>**

TGF-β is a pleiotropic cytokine with a broad range of biologic effects, which is involved in the regulation of inflammatory processes on several levels. A main mechanism in this respect is the maintenance of T cell tolerance to self or innocuous antigens (Li and Flavell, 2008). In cancer-associated inflammation, TGF-β suppresses the anti-tumor activity of diverse immune cells, including T-cells, natural killer (NK) cells, neutrophils, monocytes and macrophages (Bierie and Moses, 2010). A great number of studies focused on the role of TGF-β in fibrosis and associated inflammation. In these diseases, TGF-β regulates influx and activation of immune cells, as well as the actual fibrotic process, and thus the delicate balance between an appropriate inflammatory response and the development of pathologic fibrosis (Flanders, 2004; Sheppard, 2006; Lan, 2011). Several mechanistic links between inflammation and fibrosis are known, but the complete picture remains to be established (Lee and Kalluri, 2010). TGF-β signaling in these processes has been attributed both to canonical TGF-β signaling via the Smad proteins (signal-dependent transcription factors) as well as non-Smad signaling pathways (e.g., via MAPK pathways) (**Figure 2**).

Independent of inflammatory model systems, 1α,25(OH)2D3 and TGF-β/Smad signaling pathways have been found to be interrelated through three mechanisms: (i) the existence of a common regulator protein, the oncoprotein Ski, which can repress both pathways (Ueki and Hayman, 2003), (ii) the possibility of joint gene regulation via VDR and Smad recognition elements that are located in close proximity to a target promoter (Subramaniam et al., 2001) (**Figure 2**) or (ii) direct interaction of Smad3 and vitamin D signaling, whereby Smad acts as a coregulator specific for ligand-induced VDR transactivation (Yanagisawa et al., 1999).

The influence of vitamin D on inflammation-related signaling via TGF-β and Smad has mainly been investigated in models of fibrosis, and distinct mechanisms have been elucidated. Activation of 1α,25(OH)2D3 signaling by the natural ligand itself or its synthetic analogs reduces TGF-β expression (Kim et al., 2013) and interferes with the downstream signaling. The latter occurs via several mechanisms: downregulation of phosphorylated activatory Smads (Smad2/3 and 4) accompanied by upregulation of inhibitory Smad6 (Kim et al., 2013) (**Figure 2**); an inhibitory interaction between 1α,25(OH)2D3-bound VDR and Smad3 (Ito et al., 2013) or inhibition of Smad2 phosphorylation and nuclear translocation of Smad2/3, coincident with inhibited protein expression from TGF-β target genes (Halder et al., 2011). Similar findings have been made in studies with nephropathy models where suppression of TGF-β and p-Smad2/3 expression (Xiao et al., 2009) or a decrease in Smad2 and an increase in inhibitory Smad7 (Hullett et al., 2005) have been detected. In a large-scale study using hepatic stellate cells, TGF-β has been shown to cause chromatin remodeling events that led to a redistribution of genome-wide VDR binding sites (the VDR cistrome) with a shift toward VDR binding to Smad3-dependent, profibrotic target genes. In this study, VDR ligands led to a reduced Smad3 occupancy at these genes and thus antifibrotic effects (Ding et al., 2013). Although hepatic stellate cells do not belong to the immune system, and the interplay between VDR and TGF-β/Smad signaling may be dependent on the cell type, key aspects of this elaborate study deserve mention. More than 10<sup>4</sup> genomic sites were found to be co-occupied by both VDR and SMAD3 in these cells, and an analysis of the spatial relationships between the two transcription factors revealed that the respective response elements were located within a range of 200 base pairs (one nucleosomal window). Mechanistically, TGF-β signaling seems to deplete nucleosomes from the co-occupied sites and thus allow access of VDR to these sites. Vitamin D signaling on the other hand seems to limit TGF-β activation by inhibited coactivator recruitment. Spatiotemporal analysis revealed that 1α,25(OH)2D3/TGF-β-induced VDR and SMAD3 binding to the co-occupied sites were inversely correlated. The maximum of SMAD3 binding occurred 1 h after treatment and was reduced by 70% after 4 h, when VDR binding was maximal. Therefore, TGFβ signaling seems to change the chromatin architecture in a way in which liganded VDR can reverse Smad activation.

## **THE INFLUENCE OF 1α,25(OH)2D<sup>3</sup> ON INTERLEUKIN (IL) GENE EXPRESSION AND SIGNALING**

The finding that 1α,25(OH)2D3 interacts with the production of interleukins (Tsoukas et al., 1984) is of certain interest in the history of vitamin D research, as a crucial finding that expanded the view to roles beyond calcium homeostasis and crucially contributed to establish an immunoregulatory function of vitamin D (Tsoukas et al., 1984).

The interleukins are a large group of cytokines of central importance for the intercellular communication between the different cells generally involved in inflammatory responses. These cells mainly encompass the leukocytes in their various stages of differentiation (distinct T-cells subsets, monocytes, macrophages, dendritic cells (DCs), granulocytes and Blymphocytes) and cells of the connective tissue and vasculature (fibroblasts, endothelial cells). Furthermore, in specific organrelated diseases with inflammatory components (psoriasis, CKD, placental infection/inflammation, obesity, and others), further cell types are involved, e.g., keratinocytes, endothelial cells, trophoblasts, and adipocytes. All of them are capable of synthesizing interleukins, and the influence of 1α,25(OH)2D3 on IL gene expression has been investigated. The influence of 1α,25(OH)2D3 on IL gene expression and signaling in the different cell types will be outlined in the following.

## *Leukocytes*

Several studies, especially the early ones, included *ex vivo* experiments with cellular samples from healthy individuals, mainly with PBMC (Rigby et al., 1984; Tsoukas et al., 1984; Saggese et al., 1989; Muller and Bendtzen, 1992), (partly) isolated T-cells (Bhalla et al., 1986), (partly) isolated monocytes (Bhalla et al., 1986; Muller et al., 1992; Zarrabeitia et al., 1992; Lemire et al., 1995; Lyakh et al., 2005), or cocultures of T-cells and monocytes (Tsoukas et al., 1989).

## *PBMC and T-cells*

In stimulated PBMC, as a preparation that includes different cell types, 1α,25(OH)2D3 caused suppression of IL-2 production (Rigby et al., 1984; Tsoukas et al., 1984; Saggese et al., 1989) and reduced release of IL-1β, IL-6, and IL-10 (Joshi et al., 2011). Furthermore, the vitamin D analog paracalcitol led to reduced IL-8 production in stimulated PBMC (Eleftheriadis et al., 2010).

In more cell specific experiments with (partly) isolated Tcells, 1α,25(OH)2D3-mediated inhibition of IL-2 mRNA synthesis induced by lectin/phorbol ester (Matsui et al., 1986) or protein production induced by lectin (Bhalla et al., 1986), was confirmed. This was also observed for the two subsets of CD4+ and CD8+ T-cells (Jordan et al., 1989), which however displayed stimulus-dependency for IL-2 protein production. In a more detailed analysis, IL production by CD4+ and CD8+ cells was studied by flow cytometry on single cell level. In both populations, a decrease in IL-2 production was found. Conversely, regarding other IL class members analyzed in the same study, 1α,25(OH)2D3 increased the low percentage of IL-13-producing cells in both subsets and IL-6 producing CD4+ and CD8+ Tcells could only be detected after incubation with 1α,25(OH)2D3 (Willheim et al., 1999) (**Figure 3**).

The finding that IL-2 gene expression is reduced by 1α,25(OH)2D3 in T-cells has moreover been confirmed in two studies using the human T-cell line Jurkat, and the mechanisms have been studied. It has been found that the VDR seems to cause direct transcriptional repression of IL-2 gene expression via blockage of a positive regulatory element recognized by the transcription factor NFAT within the IL-2 promoter (Alroy et al., 1995). In a later study, the repression has been kinetically classified as a primary response to 1α,25(OH)2D3*,* and liganddependent VDR binding at the IL-2 gene locus was detected using ChIP assays (Matilainen et al., 2010b) (**Figure 2**). It has to be mentioned, however, that long term pretreatment of Jurkat cells with 1α,25(OH)2D3 before stimulation with mitogen and phorbol ester seems to enhance IL-2 mRNA expression (Prehn and Jordan, 1989). Studies using T-cells from other species confirmed the inhibitory effect of 1α,25(OH)2D3 on IL-2 production (Hodler et al., 1985).

Similar findings as for IL-2 have been made regarding the inhibition of IL-17 production by 1α,25(OH)2D3 from T-cells in a more recent report. It has been found that (i) the VDR competes for binding with NFAT and recruits histone deacetylase (HDAC) to the human IL-17 promoter, thus inhibiting its activation, (ii) binding of the activatory transcription factor Runx1 to the mouse IL-17A promoter was inhibited through sequestration of Runx1 by the VDR in the presence of 1α,25(OH)2D3 and (iii) 1α,25(OH)2D3 induced the IL-17 inhibiting transcription factor Foxp3 (Joshi et al., 2011). Other studies suggest a post-transcriptional mechanism of IL-17 inhibition by VDR via induction of the translation inhibitor C/EBP homologous protein (CHOP) (Chang et al., 2010).

Apart from studies with PBMC or T-cells from healthy individuals or experiments with cell lines, a few studies exist with cell samples from patients suffering from inflammatory diseases. In contrast to the findings with cells from healthy individuals after 1α,25(OH)2D3 treatment, PBMC isolated from hemodialysis patients responded to treatment with 1α(OH)D3 by enhanced IL-2 protein production, however, starting from a significantly lower level of IL-2 production compared to healthy controls (Tabata et al., 1988). The capacity of PBMC from Crohn's disease patients to produce IL-6 has been elevated by 1α,25(OH)2D3 treatment of the patients (Bendix-Struve et al., 2010). IL-6 and IL-8 production and mRNA expression have been found to be decreased by 1α,25(OH)2D3 in stimulated PBMC of psoriatic patients (Inoue et al., 1998). In PBMC from treatment-naive patients with early rheumatoid arthritis (RA), reduced IL-17A and increased IL-4 levels have been observed in the presence of 1α,25(OH)2D3. In the FACS-separated subpopulation of memory T-cells (CD45RO+), 1α,25(OH)2D3 suppressed IL-17A, IL-17F and IL-22 (Colin et al., 2010) (**Figure 3**).

## *Monocytes*

In an early report, IL-1 production by human monocytes/macrophages enriched from PBMC has been found to be elevated by single 1α,25(OH)2D3 treatment (Bhalla et al., 1986). In subsequent studies with stimulated, monocyte-enriched cultures from PBMC, either no 1α,25(OH)2D3 effect has been detected (Zarrabeitia et al., 1992) or a reduction of IL-1 (and IL-6) production has been found, which seemed to be based on post-transcriptional events (Muller et al., 1992). The decrease in IL-1 production has been confirmed for co-cultures of T-cells and monocyte-enriched PBMC (Tsoukas et al., 1989). However, it has to be pointed out that different stimuli to elicit IL-1 production had been used in these studies. In human monocytic cell lines, (U937, HL-60 or THP-1), no induction (THP-1), or upregulation of IL-1β mRNA (U937, HL-60) by 1α,25(OH)2D3 has been detected, which varied with the presence or absence and the type of the co-stimulus that was used (phorbol ester, lipopolysaccharide) (Bhalla et al., 1991; Blifeld et al., 1991; Fagan et al., 1991). Further it is noteworthy that conflicting data exist for studies with U937 cells regarding the actual secretion of IL-1β protein (Blifeld et al., 1991; Fagan et al., 1991; Taimi et al., 1993). In THP-1 cells stimulated with agonists for Toll-like receptor 8, IL-1β mRNA was induced and could be suppressed by 1α,25(OH)2D3 (Li et al., 2013).

In a more recent study, expression of IL-1 and IL-6 mRNA in freshly isolated monocytes and macrophages cultured for 7 days has been investigated. Interestingly, IL-1 and IL-6 gene expression has been regulated differently in these two distinct stages of monocyte/macrophage maturation. In the monocytes, basal IL-1 and IL-6 mRNA expression has been found to be slightly upregulated by 1α,25(OH)2D3 treatment compared to untreated controls. For 1α,25(OH)2D3 treated monocytes that were additionally stimulated with LPS or TNFα, no or only marginal differences have been found compared to LPS or TNFα treatment without 1α,25(OH)2D3 preincubation. In contrast, 1α,25(OH)2D3 treatment reduced basal IL-1 and IL-6 levels in macrophages. In 1α,25(OH)2D3 treated macrophages that were additionally stimulated with LPS or TNFα, only TNFα-stimulated IL-6 mRNA expression was influenced, whereas no significant changes were observed for IL-1 and IL-6 after 1α,25(OH)2D3/LPS-treatment. These findings show that in monocytes/macrophages, the influence of 1α,25(OH)2D3 on IL expression depends on the type of IL under consideration, the degree of maturation, and the stimulus that is employed (Di Rosa et al., 2012). In a second recent investigation, significant inhibition of IL-6 mRNA expression and protein secretion was observed in PBMC, and subsequent FACS-based analysis revealed a concomitant decrease in CD14+ IL-6-producing monocytes (Zhang et al., 2012) (**Figure 3**).

Apart from the two prominent monokines IL-1 and IL-6, the synthesis of IL-3 has been found to be influenced by 1α,25(OH)2D3 in the murine monocytic cell line WEHI-3. However, whereas one report describes dose-dependent inhibition of IL-3 production in this cell line (Abe et al., 1986), the second finds concentration-dependent stimulation or inhibition of IL-3 production (Hodler et al., 1985). Furthermore, the interleukin family members IL-8, IL10, and IL-12 have been studied more intensely on mechanistic level.

IL-10 and IL-12-production by stimulated primary human monocytes has been found to be negatively regulated by 1α,25(OH)2D3 (Lemire et al., 1995; Lyakh et al., 2005). These two genes have been identified as primary 1α,25(OH)2D3 target genes as judged by rapid VDR recruitment detected via ChIP assays in the monocytic cell line THP-1 (Matilainen et al., 2010b). Further studies with this cell line include extensive mechanistic analyses regarding the influence of 1α,25(OH)2D3 on the expression of IL-8, IL-10, and IL-12B. The IL-8 gene has been shown to be an up-regulated, primary target gene, located within an insulated cluster of CXC motif ligand (CXCL) genes. IL-8 and its neighboring genes CXCL1 and CXCL6 seem to be under the control of a consensus VDR binding motif located 22 kb downstream of the IL-8 transcription start site, which mediates 1α,25(OH)2D3–dependent chromatin opening (Ryynanen and Carlberg, 2013). As discussed in this report, this finding is seemingly in contradiction with other studies (e.g., Di Rosa et al., 2012). These studies used different cells and foremost, cells were stimulated with agents like LPS that activate transcription factors, e.g., NF-κB, that are themselves regulated by 1α,25(OH)2D3. As described above, NF-κB activity is inhibited by 1α,25(OH)2D3 (Harant et al., 1998) (**Figure 2**). It has been put forward that 1α,25(OH)2D3 may have a dual effect: primary up-regulation of genes like IL-8, which supports the inflammatory response in the early phase of inflammation, e.g., by IL-8 production, and secondary effects which would help to shut down the inflammatory process, e.g., by inhibition of NF-κB-mediated pro-inflammatory responses (Ryynanen and Carlberg, 2013). This could explain that in another study in which THP-1 cells were used, no significant effect of 1α,25(OH)2D3 on IL-8 expression was found on protein level. In this study, the cells have been stimulated with LPS after only 2 h of 1α,25(OH)2D3 treatment before IL-8 protein was analyzed after 24 and 48 h (Kuo et al., 2010). Similarly, U937 cells exposed to high glucose (a condition which leads to different stress responses like NF-κB or MAPK activation) (Stan et al., 2011; Yang et al., 2013) showed lower IL-8 secretion after pretreatment with 1α,25(OH)2D3 (Jain and Micinski, 2013). Therefore, the interference of 1α,25(OH)2D3 with cell signaling pathways of inflammatory or cell stress responses, like NF-κB or MAPK activation, and differences in treatment schedules may explain the different findings. In contrast to IL-8 as an upregulated gene, the primary effect of 1α,25(OH)2D3 on IL-10 expression is down-regulation, followed by up-regulation at a later stage (**Figure 3**). Cyclic binding of VDR to a distal promoter region with conserved VDREs, that loops 1α,25(OH)2D3 dependently to the transcription start site and induces epigenetic changes and chromatin remodeling, was detected (Matilainen et al., 2010a,b). IL-12B has been identified as a 1α,25(OH)2D3 dependently down-regulated gene in LPS-treated THP-1 cells. The gene harbors two VDR binding sites within ∼6 kb upstream of the transcription start site to which the VDR and its partner retinoid receptor (RXR) recruit co-repressors and consequently induce epigenetic changes associated with gene repression (Matilainen et al., 2010b; Gynther et al., 2011). An earlier report attributed the down-regulation of IL-12 via interference of 1α,25(OH)2D3/VDR with NF-κB binding to proximal IL-12 promoter regions (D'Ambrosio et al., 1998). It has been suggested in the more recent report that this suppression of proximal sites is due to epigenetic changes at that location via the distal VDRE binding sites identified in the more recent study (Gynther et al., 2011) (**Figure 3**).

In addition to data from experiments with monocytes, macrophages, and DCs as differentiated members of the monocytic lineage have been investigated.

In macrophages from vitamin D-deficient mice, IL-1, and IL-6 production (evaluated as biological activity) was significantly reduced relative to control mice. Notably, this was paralleled by a decrease in macrophage cytotoxicity. Furthermore, the vitamin D deficient mice had reduced serum levels of IL-1 and IL-6 after challenge with LPS (Kankova et al., 1991). In human monocyte-derived macrophages and PMA-differentiated U937 cells, which were stimulated with LPS or PMA, IL-1β production was strongly stimulated by 1α,25(OH)2D3. This effect was ascribed to increased IL-1β transcription, but not by RNA stabilization, and seemed to be mediated by Erk1/2. Moreover, 1α,25(OH)2D3 induced the expression and phosphorylation of CCAAT enhancer-binding protein β as a known IL-1 β-regulating transcription factor (Lee et al., 2011). The upregulation of IL-1β by 1α,25(OH)2D3 is also relevant for infection-induced inflammation, as in THP-1 cells or primary human macrophages infected with *Mycobacterium tuberculosis* (as well as in noninfected controls), 1α,25(OH)2D3 increased the expression of IL-1β mRNA. IL-1β is a critical factor for host defense in this disease. Notably, mature intracellular IL-1β protein was only detected in infected, 1α,25(OH)2D3 treated THP-1 cells, which represents a further level of gene expression control exerted by 1α,25(OH)2D3. Secretion of IL-1β was only seen in infected cells, and significantly enhanced by 1α,25(OH)2D3. With respect to the mechanism, the study revealed 1α,25(OH)2D3–dependent binding of VDR to a promoter-proximal consensus VDRE, which was paralleled by upregulated VDR-expression, and recruitment of RNA polymerase II to the transcription start site (Verway et al., 2013).

In a further study with mouse macrophages, 1α,25(OH)2D3 led to reduced mRNA expression of the IL-12 subunit p40 in response to LPS/interferon gamma (IFNγ) stimulation (Korf et al., 2012), which is in line with the effects seen in monocytes, as described above (**Figure 3**). Stimulation of the macrophages with 1α,25(OH)2D3 was accompanied by upregulation of VDR and the 1α,25(OH)2D3–catabolic enzyme CYP24. Further changes concerned the potential to stimulate T-cells, as assessed by co-culture experiments including FACS analysis of surface markers. These effects could not be observed with IL-10 deficient macrophages. Notably, the effects on IL-12 p40 expression and T-cell stimulation also occurred in monocytes/macrophages from non-obese diabetic (NOD) mice, which have a background of inflammatory features seen in type 1 diabetes (Korf et al., 2012).

Analogous studies have been conducted for DCs from NOD mice or non- obese diabetes-resistant (NOR) control mice. In both cases, 1α,25(OH)2D3 altered the phenotype of DCs and inhibited the LPS/IFNγ–induced mRNA expression and protein secretion of IL-10 and IL-12 (Van Etten et al., 2004). In general, it has been shown that 1α,25(OH)2D3 prevents *in vitro* differentiation of human monocytes into immature DCs, associated with decreased capacity to activate T-cells. Furthermore, 1α,25(OH)2D3 inhibits maturation of DCs. In maturating DCs, 1α,25(OH)2D3 reduces IL-12p70 and enhances IL-10 secretion upon stimulation of the DCs by CD40-crosslinking (Penna and Adorini, 2000). This has been independently confirmed for IL-12p70 production upon LPS stimulation (Sochorova et al., 2009). Additionally, these findings are in line with a study on the generation of regulatory DCs for therapeutic use from human monocytes, which were differentiated in the presence of 1α,25(OH)2D3. Apart from reduced LPS-induced IL-12 and enhanced IL-10 secretion of the maturating cells, a major characteristic of these 1α,25(OH)2D3–treated DCs is their low level of IL-23 secretion, which was apparent with or without stimulation with LPS (Pedersen et al., 2009) (**Figure 3**). A further recent investigation used monocyte-derived DCs from Crohn's disease patients. When the cells were cultured in the presence of 25(OH)D3 or 1α,25(OH)2D3 and matured with LPS, they exhibited significantly increased IL-6 production, and non-significant reductions in and IL-10 and IL-12p70. IL-1β and Il-8 levels were not affected in this study (Bartels et al., 2013).

## *B-cells and neutrophils*

B-cells and neutrophils have been less intensively studied, but the available data show that IL gene expression in these cells is also targeted by 1α,25(OH)2D3. In isolated human peripheral Bcells, IL-10 secretion can be induced by stimulation (cross-linking of B-cell receptor/CD40 antibody/IL-4). This production can be enhanced by 1α,25(OH)2D3. Besides the influence on IL gene expression, 1α,25(OH)2D3 induces the expression of VDR and Cyp24 mRNA in the stimulated B-cells. These activated cells also express Cyp27b1 mRNA and are able to produce 1α,25(OH)2D3 from 25(OH)D3. Binding of VDR to a VDRE in the proximal IL-10 promoter has been shown by ChIP assay, and binding of RNA-polymerase II could only be detected in IL-10 secreting B-cells (Heine et al., 2008).

Neutrophils respond to 1α,25(OH)2D3 by a slight reduction of IL-1β mRNA expression. Notably, the abundance of VDR mRNA in neutrophils has been found to be comparable with monocytes (Takahashi et al., 2002).

## *Fibroblasts, keratinocytes, endothelial cells*

In a first study where these cell types were used, IL-1-stimulated normal human dermal fibroblasts, normal human keratinocytes and normal human endothelial cells were investigated regarding changes of IL-8 mRNA and protein expression in dependence of 1α,25(OH)2D3 treatment. IL-8 expression was reduced by 1α,25(OH)2D3 on both levels of gene expression for fibroblasts and keratinocytes, but not for endothelial cells, where no significant changes have been found (Larsen et al., 1991).

For IL-8, and also for IL-6 protein production, this result has been confirmed in studies using phorbol ester stimulated human fibroblast cell lines (Srviastava et al., 1994), and in experiments employing TNF-α-stimulated human dermal fibroblasts (Fukuoka et al., 1998). Similar results have been obtained with fibroblast cultures obtained from surgery of patients suffering from nasal polyposis, which is defined as a chronic inflammatory process. However, rather high concentrations (10–100μM) of 1α,25(OH)2D3 were necessary to significantly reduce IL-6 and IL-8 production in these cells (Rostkowska-Nadolska et al., 2010).

In cultured normal human keratinocytes, only minor effects were observed for IL-1α and IL-8 production, when the influence of 1α,25(OH)2D3 was investigated for otherwise untreated cells. However, TNF-α-stimulation led to slightly enhanced IL-1α and markedly increased IL-8 secretion, which could be reduced by 1α,25(OH)2D3 (Zhang et al., 1994). This was confirmed for IL-8 (Koizumi et al., 1997). On the other hand, stimulation with phorbol ester plus LPS caused a rise in IL-8 production, but a decrease in IL-1α. 1α,25(OH)2D3 inhibited IL-8 secretion and restored IL-1α production (Zhang et al., 1994). Stimulation of normal human keratinocytes with IL-17A resulted in a pronounced increase in IL-6 mRNA and IL-8 protein secretion, which could be effectively blocked by 1α,25(OH)2D3 treatment (Peric et al., 2008). In a mechanistically insightful study, the effect of 1α,25(OH)2D3 on the expression of IL-1α, the intracellular IL-1 receptor antagonist (icIL-1Ra) and IL-18 was studied in mouse primary keratinocytes. Treatment with 1α,25(OH)2D3 induced IL-1α and icIL-1Ra mRNA and protein, however, the ratio of icIL-1Ra to IL-1, which determines the effect on IL-1 activity, was markedly increased, and indeed reduced IL-1 activity could be detected. The use of keratinocytes from VDR−*/*<sup>−</sup> mice confirmed that the effect was mediated by VDR. Regarding the mechanism of gene regulation, increased IL-1α mRNA stability was observed and enhanced icIL-1Ra gene transcription via a secondary mechanism have been suggested to account for the effects on these gene. 1α,25(OH)2D3 markedly suppressed IL-18 mRNA expression, and the effect was dependent on VDR, as no effect of 1α,25(OH)2D3 was seen in VDR−*/*<sup>−</sup> mice. These mice exhibit markedly elevated basal levels of IL-18 mRNA and protein, and expression of human VDR in these mice could restore basal levels (Kong et al., 2006).

A further cell type involved in inflammatory responses, especially in infection-mediated inflammation, are epithelial cells. Treatment of human microvessel endothelial cells with 1α,25(OH)2D3 suppresses LPS-induced IL-6 and IL-8 release, whereas 1α,25(OH)2D3 alone does not affect IL production. As assessed by reporter gene assay, this seems to be based on inhibition of LPS-induced NF-κB activation. This activation usually occurs via the MyD88-dependent branch of TLR4-signaling. In contrast, 1α,25(OH)2D3 did not influence the activity of interferon-β-promoter constructs, which has been determined as a measure of MyD88-independent LPS/TLR4 signaling (Equils et al., 2006). Reduced IL-6 and IL-8 production was also seen in 1α,25(OH)2D3-treated cystic fibrosis respiratory epithelial cell lines challenged with LPS. With respect to NF-κBsignaling, reduced IκBα phosphorylation and increased total cellular IκBα upon 1α,25(OH)2D3 treatment have been found in this study (McNally et al., 2011) (**Figure 2**). Similar findings have been made for human umbilical vein cord endothelial cells (HUVEC) incubated cultured in a CKD-like environment (hypocalcemia, advanced glycation end products, parathyroid hormone) and 1α,25(OH)2D3. This environment provoked enhanced IL-6 expression and secretion, increased DNA-binding of NF-κB-p65 and decreased IκBα expression. These changes were counteracted by 1α,25(OH)2D3 (Talmor-Barkan et al., 2011). In TNFα-stimulated human coronary arterial cells, a slight, but significant reduction of IL-8 production has been observed for 1α,25(OH)2D3 treatment in certain concentrations, but IL-6 production could not be influenced (Kudo et al., 2012). An interesting novel mechanism for interference of 1α,25(OH)2D3 and LPS-stimulated IL-8 production from epithelial cells has been proposed in a recent study, where a vitamin D3 derivative have been found to increase the release of the soluble form of CD14 (sCD14) via ERK1/2 activation. Neutralization of LPS by sCD14 could account for the effect of the vitamin D analog (Hidaka et al., 2013).

### *Trophoblasts, endometrial cells, myometrial cells*

Placental inflammation including release of interleukins is associated with preeclampsia, preterm labor, and abortion. Therefore, cell types involved in this inflammatory condition have been investigated regarding the influence of 1α,25(OH)2D3 on IL secretion. In cultured human trophoblasts, 1α,25(OH)2D3 reduced TNFα-induced IL-6 mRNA expression and protein secretion (Diaz et al., 2009). Mechanistic evidence regarding the influence of vitamin D signaling on IL gene expression in placental tissue was presented in a study with for Cyp27b1−*/*<sup>−</sup> (vitamin D-activating 1α-hydroxylase) mice and VDR−*/*<sup>−</sup> mice. In these mice, basal expression of IL-10 mRNA was decreased relative to wildtype placentas, and LPS stimulation resulted in higher levels of IL-6 mRNA in the <sup>−</sup>*/*<sup>−</sup> placentas compared to wildtype. PCR array analysis of LPS-stimulated placental tissue from Cyp27b1−*/*<sup>−</sup> mice revealed enhanced expression of IL-4, IL-15, and IL-18 mRNA relative to WT and the same experiments with VDR−*/*<sup>−</sup> mice yielded higher IL-1α and IL-6 mRNA levels. Further experiments with LPS-stimulated placentas from WT mice showed that treatment with 25(OH)D3 as the substrate of CYP27B1 reduces IL-6 mRNA expression. Moreover, LPS challenge of pregnant WT mice led to enhanced expression of Cyp27b1 and VDR. Apart from the mechanistic conclusion that VDR signaling is a factor that controls IL gene expression, these results show that pro-inflammatory stimuli are able to enhance the expression of crucial vitamin D signaling components which are able to mediate anti-inflammatory responses (Liu et al., 2011).

In line with these findings, experiments using human endometrial cells from women with unexplained recurrent spontaneous abortion (URSA) or in controls, significant down-regulation of IL-6 by 1,25(OH)2D3 was observed in two cell types (whole endometrial cells and endometrial stromal cells), but for IL-8, opposed effects were observed for the two cell types in URSA samples, which highlights the complexity of these responses given the fact that several cell types are involved in inflammatory processes (Tavakoli et al., 2011).

## *Adipocytes*

Obesity is a disease condition which is strongly associated with low-grade inflammation, therefore adipocytes have been used as a further model system regarding the interplay of vitamin D signaling and IL gene expression/production. In a recent report, human adipocytes from biopsies and from differentiated human mesenchymal stromal cells were studied with respect to IL-6 gene expression/release depending on the presence of 1α,25(OH)2D3. LPS-induced IL-6 mRNA and protein were reduced in both systems by cotreatment with 1α,25(OH)2D3. Regarding the underlying signal transduction events, it was shown that 1α,25(OH)2D3 inhibited IκB phosphorylation and thus NF-κB translocation into the nucleus (**Figure 2**). DNA binding of NF-κB complexes upon LPS stimulation was significantly reduced in 1α,25(OH)2D3 pretreated cells compared to controls (Mutt et al., 2012). A further recent investigation addressed the influence of *in vitro* and *in vivo* administered 1α,25(OH)2D3 on IL-6 and IL-8 gene expression from IL-1β-stimulated human adipose tissue. The adipose tissue samples have been either (i) treated *in vitro* with 1α,25(OH)2D3 or have been (ii) obtained from obese subjects with low plasma levels of 25(OH)D3 after *in vivo* (oral) treatment with high-dose 1α,25(OH)2D3 or placebo. In the *in vitro* study, reduced mRNA levels of IL-6 and IL-8 and reduced IL-6 and IL-8 protein (significance only shown for IL-8) have been found. However, although the *in vivo* treatment led to a small decrease of IL-6 and IL-8 mRNA expression in the adipose tissue, there were no significant differences between the 1α,25(OH)2D3–treated and the control group. Oral treatment with 1α,25(OH)2D3 did also not significantly change circulating levels of IL protein in the subjects preand post-treatment(Wamberg et al., 2013). These findings urge caution about the extrapolation of *in vitro* findings to the *in vivo* situation.

Apart from studies with primary cells, cultures of adipocytelike murine 3T3-L1 cells have been used, but contradictory results have been reported e.g., regarding IL-6 gene expression (Sun and Zemel, 2008; Marcotorchino et al., 2012).

## **VDR GENE VARIANTS, VDR GENE SILENCING, AND IL GENE EXPRESSION/PRODUCTION**

A further aspect that underscores the importance of vitamin D signal transduction on IL biosynthesis is the effect of the VDR receptor gene variants on IL gene expression. The singlenucleotide polymorphism *Fok*I, which comprises a shorter VDR protein of 424 aa or the long isoform with 427 aa, influences IL-12 expression. In human monocytes and DCs, presence of the short VDR isoform leads to a higher expression of IL-12 compared to the long isoform, a result which was reflected by results from reporter gene assays with IL-12 promoter fragments (Van Etten et al., 2007). Moreover, VDR gene promoter variants have an impact on the expression of IL-10 in blood mononuclear cells (Selvaraj et al., 2008).

Changes in IL production can be observed in VDR KO mice. VDR KO considerably facilitates development of IL-17 secreting T-cells (Th17 cells) in response to respective *in vitro* stimuli. Further, enhanced IL-17 production was observed in these Th17 cells compared to wildtype. Conversely, a reduction in regulatory T-cells and tolerogenic DCs was observed. Moreover, IBD can be induced experimentally in these mice by transfer of naive T-cells that develop into specific, IBD-inducing subsets. The severity of IBD was strongly enhanced in VDR KO mice compared to control animals, which was ascribed to the increased propensity for development into Th17 cells (Bruce et al., 2011).

## **INFLUENCE OF 1α,25(OH)2D<sup>3</sup> ON IL RECEPTOR EXPRESSION**

Apart from induction of IL gene expression/protein release, 1α,25(OH)2D3 may also modulate IL signaling via regulation of IL receptor expression. In early reports, moderate downregulation (Matsui et al., 1986) or no changes (Jordan et al., 1989) were found regarding IL-2 receptor expression in 1α,25(OH)2D3 treated, mitogen-stimulated PBMC, or mitogen/phorbol esterstimulated T-cells, respectively. However, IL-2 mediated expression of IL-2 receptor units was superinduced by 1α,25(OH)2D3 in mitogen-stimulated PBMC (Rigby et al., 1990). The vitamin D3 upregulated protein 1 (VDUP1), which is expressed in a 1α,25(OH)2D3-dependent manner, has been found to inhibit the activity of the IL-3 receptor promoter (Han et al., 2003). On the other hand, IL-1 and IL-4 receptor densities seem to be upregulated by 1α,25(OH)2D3 on a murine T-cell line and a murine osteoblast cell line, respectively (Lacey et al., 1993a,b). Furthermore, downregulated IL-22 mRNA and protein levels have been detected in cultured epidermis tissue treated with calcipotriol, a vitamin D analog (Moniaga et al., 2013).

## **THE INFLUENCE OF 1α,25(OH)2D<sup>3</sup> ON TNFα mRNA AND PROTEIN EXPRESSION AND RELEASE**

The impact of 1α,25(OH)2D3 on TNFα gene expression was primarily studied in PBMC, primary monocytes/macrophages or in monocytic cell lines. Expression was investigated both on mRNA level and/or on the level of protein production, and was sometimes reported in terms of protein release as a secondary readout.

In general, *in vitro* or *in vivo* treatment with 1α,25(OH)2D3 of PBMC caused a decrease in TNFα gene expression and/or TNFα production. This was the case for PBMC from healthy donors that were stimulated with different agents (LPS, Muller et al., 1992; Panichi et al., 1998; Rausch-Fan et al., 2002); live *Mycobacterium tuberculosis*(Prabhu Anand et al., 2009), as well as for PBMC from patients suffering from diseases with inflammatory features [renal disease (Riancho et al., 1993; Panichi et al., 1998); pulmonary tuberculosis (Prabhu Anand et al., 2009)]. Analogous findings were obtained with monocyte-enriched PBMC after stimulation with LPS (Muller et al., 1992), IFNγ or phorbol ester (Zarrabeitia et al., 1992) (however, not with LPS in this particular report). In one of the latter studies, nuclear run-off analysis did not indicate that TNFα gene transcription was affected by 1α,25(OH)2D3 (Muller et al., 1992).

In contrast to the findings with PBMC or monocyte-enriched PBMC, studies that used human primary monocytes or macrophages often found increased TNFα expression/secretion after 1α,25(OH)2D3 exposure, either regarding basal levels (human monocyte-derived macrophages; Bermudez et al., 1990) or with respect to stimulus-induced mRNA or protein levels (murine alveolar macrophages/LPS- or PMA-stimulation, Higashimoto et al., 1995; peritoneal macrophages from continuous peritoneal dialysis patients/LPS-stimulation, Cohen et al., 2001). In line with this, murine bone-marrow derived macrophages (BMMs) responded to 1α,25(OH)2D3 with an increase in TNFα mRNA abundance, which was synergistically enhanced by LPS stimulation. This study also addressed molecular mechanisms. Treatment with 1α,25(OH)2D3 and stimulation with LPS did not influence TNFα mRNA stability, but the data suggested that 1α,25(OH)2D3 regulates the TNFα gene on transcriptional level, as a VDR-binding sequence could be identified in the TNFα promoter region using electrophoretic mobility shift assays (Hakim and Bar-Shavit, 2003).

When human monocytic cell lines were studied, heterogeneous results were obtained, and the outcome seems to depend on the differentiation status of the cells (e.g., Bhalla et al., 1991). For the three cell lines that were mainly employed, the order of their stage of maturation is known. HL-60 cells are myelomonocytic stem-cells and thus are the least mature cell line; U937 are characterized as monoblasts, and represent an intermediate stage; and THP-1 cells are regarded as promonocytic cells and are therefore the most mature cell line (Frankenberger et al., 1994).

In HL-60 cells, 1α,25(OH)2D3 had no influence on PMAinduced TNFα mRNA expression, but enhanced it in U937 cells (Bhalla et al., 1991). In a second study, 1α,25(OH)2D3 preincubation of U937 cells accelerated LPS-induced TNFα mRNA expression and led to higher steady-state mRNA levels which were associated with enhanced TNFα protein production. Mechanistic analysis pointed to a secondary effect since 1α,25(OH)2D3 pretreatment was needed for more than 6 h in order to achieve enhanced TNFα protein synthesis. The requirement of 1α,25(OH)2D3-driven expression of the LPS co-receptor CD14, was suggested to be the mechanistic basis of his secondary effect (Prehn et al., 1992). In a further investigation, differentiation by 1α,25(OH)2D3 enhanced LPS-induced TNFα secretion in U937 and THP-1 cells. Concomitant increase in TNFα mRNA was confirmed for U937 cells (Taimi et al., 1993). In contrast, 1α,25(OH)2D3 was reported to significantly suppress TNFα release in LPS-stimulated THP-1 cells and human primary monocytes (Kuo et al., 2010), and a further study reported reduced TNFα production and secretion from 1α,25(OH)2D3-treated, IFNγ-activated THP-1 cells (Villaggio et al., 2012).

In one report, TNFα mRNA levels of 1α,25(OH)2D3–treated human PBMC, U937 and THP-1 cells, that were stimulated either with LPS or with phytohemagglutinin (PHA), were compared. Differences occurred between the two sample types and the two stimuli. In PBMC, LPS had no influence on TNFα expression in the presence of 1α,25(OH)2D3, whereas upon PHA-stimulation, reduced TNFα mRNA levels were observed. In contrast, U937 cells (but not THP-1 cells) responded by an increase in TNFα mRNA expression (Blifeld et al., 1991).

Taken together, several studies report an increase in TNFα mRNA and protein expression in 1α,25(OH)2D3-treated, subsequently stimulated U937 cells, but equivocal effects were found with the more mature THP-1 cells. In monocyte-derived DCs from patients that suffer from Crohn's disease, TNFα production was decreased when the cells were differentiated with LPS in the presence of 1α,25(OH)2D3 (Bartels et al., 2013).

T-cells have not been intensively studied, but regulation of TNFα-expression by 1α,25(OH)2D3 has been analyzed in T-cell subsets obtained from normal healthy subjects and pulmonary tuberculosis patients. Here, 1α,25(OH)2D3 reduced the percentage of TNFα-expressing T-cell subsets (CD3+, CD3+CD4+, CD3+CD8+) (Prabhu Anand et al., 2009) (**Figure 3**).

Other cell types that were analyzed are prostate cancer lines, where 1α,25(OH)2D3 reduced basal TNFα mRNA expression (Golovko et al., 2005), or 1α,25(OH)2D3/IL-1β-stimulated synoviocytes, where TNFα mRNA was decreased (Feng et al., 2013).

In summary, 1α,25(OH)2D3-mediated downregulation of TNFα gene expression has been found in cell preparations which contain a high percentage of T-cells (PBMC or monocyteenriched PBMC). In monocytic cells, upregulation has been reported for cell lines that represent an intermediate monocytic differentiation state, whereas for more mature cells, heterogeneous results have been found. Regarding the mechanism, it has been suggested that primary effects may play a role for 1α,25(OH)2D3 regulation of TNFα gene expression, since a VDR binding element has been found in the TNFα promoter region (Hakim and Bar-Shavit, 2003). On the other hand, kinetic analysis pointed to a secondary effect, where the expression of CD14 could play a role, at least for LPS-induced TNFα expression (Prehn et al., 1992). It has to be noted, however, that cell-type specific mechanisms have been found for T-cell specific expression of the TNFα gene. Cell type-specific DNA-protein-interactions have been identified for the TNFα gene when T-cells and monocytic cells were compared. A highly conserved region in intron 3 seems to be responsible for cell specificity, as this sequence induces specific activity of a TNFα-reporter plasmid in Jurkat T-cells, but not THP-1 cells (Barthel and Goldfeld, 2003). Possibly, cell specific protein complexes within this region interact with 1α,25(OH)2D3 signaling components in T-cells.

## **THE INFLUENCE OF 1α,25(OH)2D<sup>3</sup> ON INTERFERON γ GENE EXPRESSION**

IFNγ is a well-established effector in anti-infectious host reactions, autoimmune diseases and inflammation. IFNγ is mainly produced by NK and T-cells. Inhibition of IFNγ mRNA and protein secretion has been described for 1α,25(OH)2D3-treated human PBMC, peripheral blood lymphocytes or T-cells that were stimulated with phytohemagglutinin and phorbol ester (Matsui et al., 1986; Reichel et al., 1987; Rigby et al., 1987; Inoue et al., 1998) (**Figure 3**). Mechanistic insights exist from experiments using transient transfection of IFNγ promoter constructs in Jurkat T-cells. Here, it could be concluded that two VDR binding regions, one around −200 bp from the transcription start site and the second directly around the transcription start site, are involved in the regulation of IFNγ gene expression by 1α,25(OH)2D3 (Cippitelli and Santoni, 1998).

## **CONCLUSIONS**

It is well established that 1α,25(OH)2D3 influences cytokine gene expression and signaling in several different cell types. Firstly, this is the case for the pleiotropic mediator TGF-β, for which it has been shown that either the expression of the cytokine itself or expression of associated signaling components is downregulated by 1α,25(OH)2D3. In hepatocytes, 1α,25(OH)2D3 has been found to influence TGF-β signaling in a genome wide scale by directing binding of Smad proteins to target genes. These actions of 1α,25(OH)2D3 on TGF-β expression or signaling were able to inhibit fibrosis and associated inflammation. Second, the interleukins are a vast group of inflammatory cytokines that are clearly regulated by 1α,25(OH)2D3 in a cell-specific manner. However, for several members of this family (e.g., IL-1, IL-6, and IL-8), both positive or negative regulation by 1α,25(OH)2D3 has been observed. A closer look at the parameters that determine the outcome of 1α,25(OH)2D3 action on the expression of these genes is warranted. This applies in particular to the time-scale of changes in gene expression, as different responses may occur during separate stages of 1α,25(OH)2D3 action. Regarding the mechanisms, recruitment of VDR to the respective genomic regions, as well as interaction of 1α,25(OH)2D3 signaling with other transcription factors involved in IL expression (NFAT, NF-κB, Runx1), seem to occur. Concerning the p38 MAP kinase phosphatase MKP1, it was found that GCR and VDR/RXR act in a synergistic manner to induce MKP1 expression in monocytes. This results in reduced p38 activation and reduced formation of proinflammatory cytokines. As a further cytokine, the proinflammatory mediator TNFα has been identified as a 1α,25(OH)2D3 target gene. Also in this case, the vitamin D effects are cell-specific: With cell samples that mainly contain T-cells, downregulation of TNFα has been observed, whereas for monocytic cells, either positive or negative regulation occurred depending on the differentiation state. Finally, gene expression of the proinflammatory mediator IFNγ has been described to be suppressed by 1α,25(OH)2D3 in T-cells. Altogether, the influence of 1α,25(OH)2D3 on the expression of interleukins, TNFα, and IFNγ by different cell types, and the consequences for the cellular interplay that are to be anticipated, amounts to a complex picture. In **Figure 3**, the influence of 1α,25(OH)2D3 on the expression of these cytokines is summarized for the major immune cells (monocytes, DCs, and different T-cell subsets). The resulting pattern supports a shift of T-cell responses from a Th1 type toward Th2 reactions and a suppression of Th17 responses. The effect of 1α,25(OH)2D3 on cytokine expression in antigen presenting cells (monocytes, DCs) remains unclear and seems to depend on the time of stimulation, the differentiation state and other factors.

## **PERSPECTIVES**

Modulation of GCR, NFκB, NFAT as well as SMAD signaling plays a central role in the immunomodulatory activities of 1α,25(OH)2D3. Mechanistic studies on individual genes gave some mechanistic insights into the mechanisms involved in the interaction between VDR/RXR and the above mentioned transcription factors. These mechanisms include competitive binding as well as a crosstalk between the signaling pathways on multiple levels including the promoter level. However, by using ChIP seq and other techniques which allow a genome-wide view, we are just starting to understand the signaling network which is responsible for cell-type-specific and locus-dependent gene activation by ligand-regulated transcription factors such as VDR/RXR. For example, intersecting VDR/SMAD regulatory circuits have just been unraveled and it was shown that TGFβ signaling facilitates VDR binding to certain gene loci. More such data are required to increase our understanding of the complex gene regulatory network that is affected by 1α,25(OH)2D3. Especially, genome-wide data on VDR loci in conjunction with analyses of other, inflammation-related key transcription factors in different cell types and various stimuli are necessary to understand the complex regulation of gene transcription during inflammation.

## **ACKNOWLEDGMENT**

Studies in our laboratory were supported by the Dr. Hans Kröner Graduiertenkolleg, the LOEWE Initiative (Fraunhofer-TMP) the cluster of excellence Macromolecular Complexes (EXC 115) and DFG (SFB 1039).

## **REFERENCES**


inhibition of interleukin-2 and augmentation of interleukin-1 production. *Cell. Immunol.* 98, 311–322. doi: 10.1016/0008-8749(86)90291-1


bronchoconstriction in children with asthma. *Eur. Respir. J.* 37, 1366–1370. doi: 10.1183/09031936.00044710


of protein kinase C and Ca2+/calmodulin-dependent kinase. *Respiration* 62, 89–94. doi: 10.1159/000196398


benign prostatic hyperplasia stromal cell proliferation and inflammatory response by targeting the RhoA/Rho kinase and NF-kappaB pathways. *Prostate* 69, 480–493. doi: 10.1002/pros.20896


by 1,25-dihydroxyvitamin D3. *J. Clin. Endocrinol. Metab.* 69, 127–133. doi: 10.1210/jcem-69-1-127


**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: 22 April 2014; accepted: 12 June 2014; published online: 02 July 2014. Citation: Wöbke TK, Sorg BL and Steinhilber D (2014) Vitamin D in inflammatory diseases. Front. Physiol. 5:244. doi: 10.3389/fphys.2014.00244*

*This article was submitted to Integrative Physiology, a section of the journal Frontiers in Physiology.*

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

## The impact of vitamin D in breast cancer: genomics, pathways, metabolism

## *Carmen J. Narvaez 1, Donald Matthews 1,2, Erika LaPorta1,2, Katrina M. Simmons 1,2, Sarah Beaudin1,2 and JoEllen Welsh1,3\**

*<sup>1</sup> Cancer Research Center, University at Albany, Rensselaer, NY, USA*

*<sup>2</sup> Department of Biomedical Sciences, University at Albany, Rensselaer, NY, USA*

*<sup>3</sup> Department of Environmental Health Sciences, University at Albany, Rensselaer, NY, USA*

#### *Edited by:*

*Carsten Carlberg, University of Eastern Finland, Finland*

#### *Reviewed by:*

*Mieke Verstuyf, KU Leuven, Belgium Moray J. Campbell, Roswell Park Cancer Institute, USA*

#### *\*Correspondence:*

*JoEllen Welsh, Cancer Research Center, University at Albany, 1 Discovery Drive, Rensselaer, NY 12144, USA e-mail: jwelsh@albany.edu*

Nuclear receptors exert profound effects on mammary gland physiology and have complex roles in the etiology of breast cancer. In addition to receptors for classic steroid hormones such as estrogen and progesterone, the nuclear vitamin D receptor (VDR) interacts with its ligand 1α,25(OH)2D3 to modulate the normal mammary epithelial cell genome and subsequent phenotype. Observational studies suggest that vitamin D deficiency is common in breast cancer patients and that low vitamin D status enhances the risk for disease development or progression. Genomic profiling has characterized many 1α,25(OH)2D3 responsive targets in normal mammary cells and in breast cancers, providing insight into the molecular actions of 1α,25(OH)2D3 and the VDR in regulation of cell cycle, apoptosis, and differentiation. New areas of emphasis include regulation of tumor metabolism and innate immune responses. However, the role of VDR in individual cell types (i.e., epithelial, adipose, fibroblast, endothelial, immune) of normal and tumor tissues remains to be clarified. Furthermore, the mechanisms by which VDR integrates signaling between diverse cell types and controls soluble signals and paracrine pathways in the tissue/tumor microenvironment remain to be defined. Model systems of carcinogenesis have provided evidence that both VDR expression and 1α,25(OH)2D3 actions change with transformation but clinical data regarding vitamin D responsiveness of established tumors is limited and inconclusive. Because breast cancer is heterogeneous, analysis of VDR actions in specific molecular subtypes of the disease may help to clarify the conflicting data. The expanded use of genomic, proteomic and metabolomic approaches on a diverse array of *in vitro* and *in vivo* model systems is clearly warranted to comprehensively understand the network of vitamin D regulated pathways in the context of breast cancer.

**Keywords: 1α,25(OH)2D3, vitamin D, calcitriol, VDR, breast cancer, genomics**

## **INTRODUCTION TO BREAST CANCER**

In the United States in 2013, breast cancer was estimated to account for 29% of new cancer cases and 14% of cancer deaths in women, making it the most common cancer diagnosed and the second most common cause of cancer mortality in women. While standard tumor pathology focuses on the presence or absence of hormone (estrogen, progesterone) and growth factor (HER2) receptors, it is now clear that breast cancer is an extremely heterogeneous disease. Genomic profiling has identified several molecularly defined sub-types of breast cancer including Luminal A, Luminal B, Basal, HER2, and Claudin-low (Cancer Genome Atlas Network, 2012). The most frequently diagnosed sub-type is Luminal A (51%), followed by Luminal B (29%), Basal (17%), HER2 (12.5%), and Claudin-low (2%). The importance of these molecular subtypes cannot be underestimated as they allow for prediction of therapeutic targets and they display distinct clinical courses (Caan et al., 2014). Women whose tumors fit the Luminal A profile have the best prognosis whereas those whose tumors have Luminal B or Basal profiles have poor prognosis. Although many trials have assessed the impact of nutrients, including vitamin D, on breast cancer risk and progression, few have been designed to stratify results by molecular sub-type. This review will highlight the cumulative data on vitamin D actions in breast cancer while emphasizing the gaps in knowledge regarding its effects on specific molecular subtypes.

## **OBSERVATIONAL AND INTERVENTION STUDIES ON VITAMIN D AND BREAST CANCER**

Population studies on vitamin D in relation to chronic diseases such as breast cancer are complicated by difficulties in accurately assessing dietary sources (confounders include natural foods vs. fortified foods, supplement use, intake of D2 vs. D3, and calcium status) and in estimating the amount of vitamin D3 generated through sunlight exposure (confounders include lifestyle, latitude, pollution, sunscreen, skin pigmentation, and age). Thus, it is not surprising that studies designed to address the impact of vitamin D status on breast cancer have yielded mixed results. While much data is supportive that high vitamin D status as measured by serum 25-hydroxyvitamin D (25(OH)D3) is associated with decreased risk of breast cancer (Bauer et al., 2013; Bilinski and Boyages, 2013; Wang et al., 2013; Kim et al., 2014), longer disease free survival and reduced mortality (Rose et al., 2013; Maalmi et al., 2014; Mohr et al., 2014), some large studies have failed to support these associations (Kuhn et al., 2013). With respect to endogenous synthesis of vitamin D3, small scale studies supported the concept that sunlight exposure is associated with reduced risk of breast cancer, however, the associations were dependent on region of residence and skin pigmentation (John et al., 1999, 2007). Larger international studies have consistently demonstrated significant inverse correlations between incident solar radiation and breast cancer rates (Edvardsen et al., 2011; Engel et al., 2011, 2014; Grant, 2013; van der Rhee et al., 2013).

Data from randomized controlled trials (RCTs) of vitamin D supplementation in relation to breast cancer development have been inconclusive, with only slight benefits of supplementation sometimes observed (Sperati et al., 2013; Redaniel et al., 2014). The large (*>*30,000 women) Women's Health Initiative (WHI) trial assessed the impact of supplementation with both calcium and vitamin D on multiple health outcomes including cancer. However, the data from this trial was confounded by the low dose of vitamin D (400 IU/day), coadministration of calcium supplements, poor compliance, extensive pre-trial supplement use in the study population and the freedom for trial participants to take additional personal supplements of up to 1000 IU vitamin D per day. Thus it was not surprising that initial data from the WHI trial indicated no significant effects of vitamin D plus calcium supplementation on breast cancer. Subgroup and follow-up analyses of trial participants have yielded mixed results. One report indicated that higher intake of vitamin D was moderately associated with a lower risk of pre- but not post- menopausal breast cancer (Lin et al., 2007). In another sub-group analysis (including only women who were not taking personal calcium or vitamin D supplements at randomization), risk of invasive breast cancers was decreased in women supplemented with calcium and vitamin D (Bolland et al., 2011). The most recent analysis of all WHI diet study participants (assessed 5 years after intervention ended) indicated a reduced incidence of *in situ* breast cancers in the calcium and vitamin D cohort but also suggested that women with the highest vitamin D intakes may have an increased risk of invasive breast cancer (Cauley et al., 2013). Further sub-group analysis based on menopausal status or molecular subtype may shed light on these discrepancies. Regardless, these inconsistent data emphasize the continued need for rigorous, well-designed RCTs to specifically assess the impact of vitamin D supplementation on breast cancer development. Of note, the ongoing VITAL trial of 20,000 men and women taking daily supplements of 2000 IU vitamin D3 (http://www*.*vitalstudy*.*org/) is monitoring breast cancer as one outcome.

As discussed above, genomic studies have demonstrated that breast cancers are heterogeneous and that different subtypes exhibit distinct patterns of disease progression. It is likely that VDR expression or function and thus sensitivity to changes in vitamin D status may be subtype specific, yet this has not rigorously been examined. The limited epidemiologic data that has been stratified by subtype is mixed - one study reported that the relationship between serum 25(OH)D and reduced risk of breast cancer was strongest for high grade, ER negative or triple negative cancers (Yao and Ambrosone, 2013) whereas another found that low serum 25(OH)D was associated with poor prognosis only in women with the luminal subtype of breast cancer (Kim et al., 2011). It should be noted that while vitamin D deficiency is common in all breast cancer patient populations, it is particularly prevalent in those with triple negative/basal-like tumors, the most aggressive form of the disease (Rainville et al., 2009; Peppone et al., 2012; Yao and Ambrosone, 2013). Even without rigorous "proof" of a beneficial effect of supplemental vitamin D on breast cancer, correction of vitamin D deficiency in women at risk for, or living with, breast cancer should be standard practice.

## **VITAMIN D PATHWAY EXPRESSION IN NORMAL MAMMARY CELLS AND BREAST CANCER PREVENTION**

The VDR is present in rat, mouse and human mammary gland and its expression is highest during puberty, pregnancy and lactation (Berger et al., 1987; Zinser et al., 2002). In actively growing murine mammary ducts, VDR expression is high in the ductal epithelium and low in the proliferating terminal end buds. Consistent with this murine data, high content immunohistochemistry of normal human breast epithelium demonstrated that VDR positive cells are exclusively found in the luminal (differentiated) cell layer and do not co-localize with proliferating Ki67 positive cells (Santagata et al., 2014). VDR has also been identified in the stromal compartment including the mammary fibroblasts and adipocytes (Ching et al., 2011; Campos et al., 2013; Knower et al., 2013), highlighting the complexity of 1,25(OH)2D3 signaling in the breast tissue environment.

In addition to VDR, the 25(OH)D activating enzyme CYP27B1 is expressed in murine and human mammary tissue (Zinser and Welsh, 2004a; Townsend et al., 2005; Kemmis et al., 2006; Peng et al., 2009), suggesting that systemic 25(OH)D delivered to the mammary gland can be converted to the biologically active VDR ligand 1,25(OH)2D3. *In vitro* studies confirm that incubation of mammary epithelial cells with physiological (nanomolar) concentrations of 25(OH)D leads to temporal increases in 1,25(OH)2D3 detected in tissue culture media (Kemmis et al., 2006). Although there is still uncertainty regarding how 25(OH)D, which circulates tightly bound to the vitamin D binding protein (DBP), is internalized by non-renal cells, the presence of megalin and cubilin (Rowling et al., 2006) indicates that these accessory proteins could mediate uptake of 25(OH)D-DBP complexes in mammary gland as has been demonstrated for renal cells(Willnow and Nykjaer, 2002). Indeed, *in vitro* studies demonstrate that normal breast epithelial cells and some breast cancer cells internalize 25(OH)D via megalin-mediated endocytosis (Rowling et al., 2006), however, the function of this uptake pathway in intact mammary gland has yet to be confirmed. CYP27B1 is also expressed in mammary adipocytes, which too are capable of converting 25(OH)D to 1,25(OH)2D3 in organoid culture (Ching et al., 2011). Collectively, these studies provide a biological link between vitamin D status [i.e., serum 25(OH)D] and breast cancer risk that is observed at the population level.

The functions of CYP27B1 and VDR in prevention of breast cancer are supported by data from animal models. *In vivo*, both high dietary vitamin D (Jacobson et al., 1989) and treatment with synthetic VDR agonists (Hussain et al., 2003) inhibit the development of carcinogen induced mammary tumors. Furthermore, VDR ablation enhances the development of hyperplasias and hormone independent mammary tumors after DMBA administration, and VDR haploinsufficiency sensitizes the mammary gland to tumorigenesis driven by the neu oncogene (Zinser et al., 2002; Zinser and Welsh, 2004b). In organ culture models, VDR agonists such as 25(OH)D3 and 1α,25(OH)2D3 reduce the incidence of DMBA induced pre-neoplastic lesions (Mehta et al., 1997; Peng et al., 2009) suggesting direct anti-cancer effects of these metabolites on mammary tissue. Collectively, these data support the concept that systemic 25(OH)D delivered to mammary gland is converted to 1,25(OH)2D3 which activates VDR to protect against carcinogenesis.

Despite these consistent data, the precise mechanisms by which vitamin D inhibits cancer development have yet to be discerned. Both 25(OH)D3 and 1α,25(OH)2D3 exert profound effects on non-tumorigenic VDR positive mammary epithelial cells including inhibition of hormone stimulated growth and branching morphogenesis and induction of differentiation biomarkers such as E-cadherin. VDR and CYP27B1 expression in mammary adipocytes also contribute to the anti-cancer effects in the whole tissue, since in response to 25(OH)D adipocytes secrete diffusible signals that inhibit morphogenesis of the adjacent ductal epithelium (Ching et al., 2011). Other potential mechanisms for breast cancer prevention by vitamin D include reduction in DNA damage (possibly via up-regulation of p53 signaling), suppression of oxidative stress and inhibition of angiogenesis, many of which have been demonstrated in tissues or cell types other than mammary gland (Kallay et al., 2002; Peng et al., 2010; Bruce et al., 2011; Hopkins et al., 2011; Krishnan and Feldman, 2011; Bikle, 2012; Dogan et al., 2012; Gordon-Thomson et al., 2012; Ting et al., 2012; Alvarez et al., 2014; Nakai et al., 2014; Sun et al., 2014; Uberti et al., 2014; Zhong et al., 2014). In addition, recent data from our group and others implicate alteration of cellular energy metabolism and innate immune responses in the anti-cancer effects of vitamin D signaling on non-tumorigenic mammary epithelial cells as described below.

Although initially recognized by Otto Warburg more than 50 years ago (Warburg, 1956), attention has recently been re-focused on the role of cellular energy metabolism as a critical component of tumor initiation and progression. It is now recognized that many cancer cells preferentially rely on glycolysis to generate energy and macromolecules that are essential for rapid proliferation, even in the presence of normoxia. The metabolic switch (the "Warburg effect") from oxidative phosphorylation to aerobic glycolysis is triggered in normal cells by activation of oncogenes and/or loss of tumor suppressors which, in part, target glucose and glutamine metabolism. The best characterized oncogenic driver of the metabolic switch, myc, induces a cohort of genes that promote glycolysis and alter glutamine flux. Conversely, tumor suppressors such as p53 normally suppress glycolysis and enhance flux through the tricarboxylic acid (TCA) cycle and oxidative phosphorylation. p53 also increases intracellular glutamate which is shunted toward glutathione synthesis, enhancing antioxidant defense against reactive oxygen species. Recent studies support the concept that vitamin D might target energy utilization pathways in non-tumorigenic breast cells to prevent carcinogenesis. The regulation of cellular glucose metabolism by 1,25(OH)2D3 has been studied (Zheng et al., 2013) in pre-malignant mammary cells (MCF10A cells) compared to their ras transformed malignant derivatives (MCF10A-ras cells). In this model, 4-day 1,25(OH)2D3 treatment reduced flux of glucose through glycolysis in both MCF10A and MCF10A-ras cells, with a more pronounced effect in the transformed cells. 1,25(OH)2D3 also reduced the flux of glucose to acetyl-coA and oxaloacetate in both cell lines, suggesting a reduction in TCA cycle activity. The predicted consequences of these 1,25(OH)2D3 inducedchanges would be limitation in the availability of TCA-derived precursors for macromolecule synthesis, coincident with reduction in proliferation.

In microarray profiling of non-tumorigenic mammary epithelial cells (hTERT-HME1) cells, we identified SLC1A1, which encodes a plasma membrane glutamate transporter, and GLUL, which encodes glutamine synthetase (GS) as novel 1,25(OH)2D3 targets. We validated that 1,25(OH)2D3 increases SLC1A1 mRNA (*>*10-fold) and decreases GLUL mRNA (*>*4-fold) in these cells, and also demonstrated decreased expression of the cognate proteins GS and SLC1A1 by western blotting. These changes in metabolic gene/protein expression correlated with accumulation of glutathione and changes in respiratory capacity in 1,25(OH)2D3 treated cells. Furthermore, 1,25(OH)2D3 pretreatment hindered growth of hTERT-HME1 cells in glutamine-starved media and exogenous glutamine partially rescued 1,25D-mediated growth arrest. These findings are intriguing because: (a) glutamate uptake and glutamate transporters are enhanced during differentiation and deregulated in cancer cells; (b) SLC1A1 null mice exhibit GSH deficiency and high oxidative stress; (c) GS enzymatic activity is necessary for adaption of mammary cells to glutamine depletion; and (d) data compiled from The Human Protein Atlas indicates that SLC1A1 is reduced and GLUL is increased in human breast cancers relative to normal tissue. Thus, we propose that 1,25(OH)2D3 regulation of SLC1A1 and GLUL synergizes with p53 to alter metabolic flux, prevent the myc-driven metabolic switch and induce quiescence in normal mammary epithelial cells. Collectively, these emerging data demonstrating regulation of amino acid and glucose metabolism (Zheng et al., 2013) in mammary epithelial cells by 1,25(OH)2D3 provides another mechanism by which vitamin D may act to prevent carcinogenesis in normal breast tissue. These changes may be complemented by alterations in lipid and energy metabolism in adjacent stromal adipocytes by 1,25(OH)2D3 (Welsh et al., 2011; Narvaez et al., 2013).

Another emerging area in cancer prevention by vitamin D involves suppression of inflammation. We identified CD14, a component of the innate immune system, as the second most highly upregulated gene (second only to CYP24A1) in 1,25(OH)2D3 treated hTERT-HME1 cells. CD14 is a known VDR target in macrophages and other immune cells, but its regulation by 1,25(OH)2D3 in mammary cells has not been well studied. In contrast to macrophages where 1,25(OH)2D3 induces membrane bound CD14, both 1,25(OH)2D3 and 25(OH)D3 induce the secretion of large quantities of soluble CD14 (sCD14) from mammary epithelial cells. As a pattern recognition receptor, sCD14 binds lipopolysaccharide and contributes to protection against mastitis in mammary tissue (Lee et al., 2003; Zheng et al., 2006; Wall et al., 2009). The soluble form of CD14 is also secreted into human milk where it contributes to protection of the neonatal gut from infections (Vidal and Donnet-Hughes, 2008). However, even in the absence of infection or lactation, CD14 and other genes involved in innate immunity are highly induced during regression of the mammary gland after weaning (Stein et al., 2004). The role of CD14 during this period of glandular remodeling may be the recognition and disposal of apoptotic cells (Heidenreich, 1999; Devitt et al., 2004; Tennant et al., 2013). We speculate that vitamin D induction of soluble CD14 in mammary tissue inhibits activation of tissue resident macrophages, suppressing inflammation which is known to drive cancer development and progression(McMahon et al., 2011; Simpson and Brown, 2013). However, it remains to be determined whether vitamin D status regulates any of these proposed functions of CD14 in mammary tissue *in vivo*. One caveat to future experimentation on vitamin D regulation of CD14 is the apparent discordance between the human and murine genomes. 1,25(OH)2D3 does not induce CD14 in the murine mammary epithelial cell line HC11 and CD14 gene expression is not altered in the mammary gland of VDRKO mice (Welsh, unpublished).

## **VITAMIN D PATHWAY EXPRESSION IN ESTABLISHED BREAST CANCER**

Over 30 years ago, the recognition that VDR expression was retained in breast cancers prompted extensive studies to determine whether targeting VDR in tumors would provide therapeutic benefit. VDR expression is retained in the majority of rodent breast tumors, human breast cancers and established breast cancer cell lines (Colston et al., 1986; Buras et al., 1994; Zinser and Welsh, 2004b). In a study of 136 patients with primary breast cancer, it was found that women with VDR negative tumors relapsed significantly earlier than women with VDR positive tumors (Berger et al., 1991). Of note, some data suggests that receptor protein expression declines in highly aggressive tumors (Lopes et al., 2010). We reviewed the frequency of genomic VDR changes in human breast cancers using datasets publically available on The Cancer Genome Atlas (https://tcga-data*.*nci*.*nih*.*gov/ tcga/) which annotates mutations, amplifications, deletions and mRNA expression profiles in human tumors (**Table 1**). Analysis of the TCGA invasive breast cancer dataset (Cancer Genome Atlas Network, 2012) of over 450 breast tumors suggests that alterations in the VDR gene are rare in human breast cancer. As shown in **Table 1**, only 5% of human breast tumors exhibited any alteration in VDR sequence or expression. However, when the VDR gene was altered, the most common change was a reduction in mRNA expression (deletions and mutations did not occur). With respect to VDR expression in specific molecular sub-types, the Luminal B subtype had the highest frequency of VDR alterations with 10.5% of tumors displaying reduced VDR mRNA expression compared to 0–3% for Luminal A, Basal, HER2, or Claudin-Low subtypes. These results showing retention of VDR in the majority of human breast tumors are consistent with the data of Santagata et al. (2014) who used a multiplex immunohistochemical approach to map receptor proteins at the single cell level and confirmed that the majority of human breast tumors are VDR positive. Interestingly, this study also demonstrated that breast tumors with the highest expression of VDR, ER, and Androgen Receptor (AR) had the best prognosis. The retention of VDR in tumors may indicate that its function has been somehow abrogated, either by altered function of the VDR despite mutation (i.e., alteration of transcriptional co-regulators), reduced ligand availability (i.e., loss of CYP27B1 and/or gain of CYP24A1), or mutation/deregulation of critical anti-cancer VDR target genes.

Despite these data, *in vitro* studies have demonstrated that specific oncogenes can deregulate VDR expression. For instance, comparison of VDR expression in a series of isogenic,



*The data was calculated with the publically available breast invasive carcinoma dataset (Cancer Genome Atlas Network, 2012) at https://tcga-data.nci.nih.gov/tcga/as. Numbers indicate the percentage of the indicated genomic alterations observed in the total number of tumors analyzed within each group (shown in parentheses). LumA, Luminal A; LumB, Luminal B; CLN Lo, Claudin Low.*

progressively transformed human mammary epithelial (HME) cell lines indicated that VDR expression and function was reduced more than 70% in HME cells expressing SV40 and/or RAS compared to the non-transformed HME cells from which they were derived (Kemmis and Welsh, 2008). Likewise, SV40 and RAS have been shown to reduce VDR activity in other breast cancer cell model systems (Agadir et al., 1999; Escaleira and Brentani, 1999). Transcriptional repressors linked to the epithelial mesenchymal transition such as SNAIL and SLUG have also been shown to down-regulate VDR (Mittal et al., 2008; Larriba et al., 2010). Thus, it is clear that there are distinct mechanisms targeting both the VDR gene itself and its protein product. Data on receptor expression derived from whole tumors may be somewhat misleading since cancer progression is driven by genetic instability and outgrowth of cells with advantageous mutations, such as activation of oncogenes. These studies indicate that abrogated expression and/or function of VDR may be limited to certain subsets of cells within individual tumors that have sustained specific molecular genetic alterations.

In addition to genetic alterations and effects of oncogenes, VDR abundance is affected by many physiological agents, including 1,25(OH)2D3 itself, estrogens, retinoids and growth factors. Thus, cell sensitivity to 1,25(OH)2D3 may also reflect the activity of other hormone signaling pathways through their impact on VDR expression. In breast cancer, the regulation of VDR expression and activity by estrogens is likely to be clinically significant. ER positive cells tend to express higher levels of VDR than ER negative cells (Buras et al., 1994) and *in vitro* studies have demonstrated that estrogen up-regulates whereas anti-estrogens down regulate VDR in ER positive breast cancer cells (Nolan et al., 1998; Byrne et al., 2000). Further studies are therefore warranted to determine the degree to which estrogen status influences VDR abundance in different 1,25(OH)2D3 target tissues (i.e., breast, bone, uterus), and whether common therapeutic synthetic or natural estrogens act as estrogen agonists or antagonists in regulation of VDR expression. Consistent with this concept, some data suggests that phytoestrogens such as resveratrol and genestein can alter VDR expression and 1,25(OH)2D3 sensitivity in cancer cells *in vitro* (Wietzke and Welsh, 2003; Gilad et al., 2006).

Other tumor-associated changes that can lead to 1,25(OH)2D3 resistance in VDR positive tumors include disruption of VDR transcriptional activation and enhanced catabolism of its ligand. Data from breast, bladder and prostate cancer suggests that alterations in transcriptional co-regulators can abrogate signaling by the 1,25(OH)2D3-VDR complex (Malinen et al., 2008; Abedin et al., 2009). Enhanced mammary cell catabolism of 1,25(OH)2D3 would also be predicted to limit the formation of active VDR complexes. Indeed, amplification of the CYP24A1 gene was reported in human breast tumors (Albertson et al., 2000) and analysis of the datasets from The Cancer Genome Atlas confirms that a subset of human breast cancers (10–13%) exhibit alterations in the CYP24 gene, with the most frequent changes being amplifications and upregulation at the mRNA level (**Table 1**). There is no obvious subtype specificity to CYP24A1 changes, although amplifications were somewhat more frequent in Luminal B and HER2 subtypes whereas increased mRNA levels were more common in Basal and Claudin-low tumors. These data are consistent with analysis of tumor samples which demonstrated higher CYP24A1 protein expression in breast tumors compared to adjacent normal tissue (Townsend et al., 2005; Lopes et al., 2010). Furthermore, splicing variants of CYP24A1 have been reported in breast cancer cell lines (Scheible et al., 2014), suggesting that distinct forms of the enzyme with altered properties may be expressed in tumors. The significance of the CYP24A1 deregulation with respect to overall catabolism of vitamin D metabolites *in situ* has yet to be ascertained.

Given that normal mammary cells utilize 25(OH)D3 as substrate for local tissue generation of 1,25(OH)2D3, imbalanced expression of either CYP24A1 or CYP27B1 favoring catabolism could theoretically contribute to escape of developing tumor cells from anti-cancer VDR signaling. In the HME cell model, oncogenic transformation was associated with reduced CYP27B1 expression and activity [as measured by 1,25(OH)2D3 synthesis]. The reductions in CYP27B1 in the oncogene-transformed HME cells were of sufficient magnitude to reduce cellular sensitivity to growth inhibition by 25(OH)D3 approximately 100-fold (Kemmis and Welsh, 2008). However, clinical data on CYP27B1 expression in breast cancer is inconsistent (Segersten et al., 2005; Townsend et al., 2005; de Lyra et al., 2006; McCarthy et al., 2009; Lopes et al., 2010) and less than 2% of breast cancers annotated in The Cancer Genome Atlas datasets exhibit genomic alterations in CYP27B1. However, altered splice variants of CYP27B1 have been detected in breast cancer cells (Cordes et al., 2007; Fischer et al., 2007) suggesting the possibility that forms of the CYP27B1 enzyme with altered function could be expressed in breast tumors.

## **ACTIONS OF VDR AGONISTS ON BREAST CANCER CELLS AND TUMORS**

Numerous studies have profiled the cellular and molecular effects of 1,25(OH)2D3 on VDR positive breast cancer cells. Furthermore, a large number of structural analogs of vitamin D developed by pharmaceutical companies and academic researchers have been used to probe the mechanisms of vitamin D mediated growth inhibition. In general, the effects of VDR agonists on breast cancer cells include modulation of key cell cycle regulators to induce G0/G1, induction of differentiation markers, and/or activation of cell death (via apoptosis or autophagy). Notably, studies with cells derived from VDRKO mice has definitely established that the nuclear VDR is required for the anti-proliferative and pro-apoptotic effects of 1,25(OH)2D3 in transformed mammary cells *in vitro* (Zinser et al., 2003; Valrance et al., 2007). In addition to regulation of cell growth and survival, studies in ER negative breast cancer cells, representative of late stage disease, have provided evidence that 1,25(OH)2D3 alters genomic stability, inhibits angiogenesis and reduces invasion and metastasis. For instance, 1,25(OH)2D3 interacts with the 53BP1 protein to eliminate invasive breast cancer cells lacking BRCA1 (Grotsky et al., 2013). 1,25(OH)2D3 and analogs inhibit invasion as measured by the Boyden chamber assay (Flanagan et al., 2003) likely through suppression of extracellular proteases (MMP-9, urokinase-type plasminogen activator, tissue type plasminogen activator), protease inhibitors and adhesion molecules. Comparative studies of breast tumors and normal adjacent breast tissue in an explant system confirmed that malignant tissue is responsive to 1,25(OH)2D3 but that the magnitude of the response is highly disparate between individual patients (Suetani et al., 2012; Milani et al., 2013). Furthermore, tumor tissue was far less sensitive to 25(OH)D3 than 1,25(OH)2D3 (Suetani et al., 2012).

Animal models of established breast cancer have demonstrated that VDR agonists can reduce tumor growth (and in some cases trigger tumor regression) with minimal effects on calcemia [the most common and dangerous side effect of 1,25(OH)2D3 therapy]. In experimental metastasis paradigms, the vitamin D analog EB1089 inhibited secondary tumors, blocked skeletal metastases and improved survival (El Abdaimi et al., 2000; Flanagan et al., 2003). More recently, dietary modifications have been shown to alter breast tumor growth and progression. Increasing dietary vitamin D3 from 1000 IU/kg diet (rodent standard) to 5000 IU/kg diet significantly reduced growth of established MCF-7 xenografts, with equivalent potency to 1,25(OH)2D3 (Swami et al., 2012). In another study, vitamin D deficiency sufficient to enhance bone turnover promoted the skeletal growth of breast cancer metastases (Ooi et al., 2010). For further details on these and other *in vivo* studies, several recent comprehensive reviews on vitamin D and breast cancer are available (Krishnan and Feldman, 2011; Lopes et al., 2012; Welsh, 2012).

## **GENOMIC PROFILING OF VDR AGONISTS IN BREAST CANCER MODEL SYSTEMS**

Screening for molecular changes induced by 1,25(OH)2D3 or vitamin D analogs in various breast cancer cells has identified scores of VDR regulated genes and proteins, indicating a broad range of downstream targets. Here we will focus on the few comprehensive genomic studies that have allowed for identification of vitamin D responsive pathways in breast cancer (Lee et al., 2006; Campos et al., 2013; Milani et al., 2013; Laporta and Welsh, 2013). In the first study of vitamin D mediated genomic changes in breast cancer, Feldman's groups used early generation arrays to compare gene expression in ER positive MCF-7 cells and ER negative MDA-MB-231 cells treated with 100 nM 1,25(OH)2D3 (Swami et al., 2003). Due to the limited nature of these arrays, which comprised 2000 cancer related genes, direct comparisons with whole genome profiling arrays isn't appropriate. However, comparisons between the two cell lines is of interest. Using a 2-fold cutoff, 62 genes (47 up/15 down) in MCF-7 cells and 20 genes in MDA-MB-231 cells (10 up/10 down) were significantly altered by 24 h treatment with 100 nM 1,25(OH)2D3, with only seven genes commonly altered in both cell lines. The larger number of regulated genes in MCF-7 cells was not surprising as CYP24A1 induction was 10-fold higher in MCF-7 cells (52-fold) than in MDA-MB-231 cells (5.5-fold). Other highly regulated genes in MCF-7 cells treated with 1,25(OH)2D3 for 24 h included RBL2, CTNNA1, RAD23B, NCOA4, BMP5 and IFNG (up) and CEACAM1, CDH6, IL13, IL1R2 and ESR (down). In MDA-MB-231 cells, highly modulated genes at 24 h included CASP4, NF1B, ITGAV, TXNRD1 and TGFB2 (up) and ANGPT1, four MMPs (12, 10, 7, 1) and PRKD1 (down). Thus, in MCF-7 cells, many of the 1,25(OH)2D3 regulated genes were involved in growth factor signaling, cell cycle, apoptosis and immune responses, whereas in MDA-MB-231 cells genes related to disease progression (i.e., invasion and angiogenesis) were altered.

Since the availability of whole genome arrays, four studies (Lee et al., 2006; Campos et al., 2013; Milani et al., 2013; Laporta and Welsh, 2013) on the effects of VDR agonists [three with 1,25(OH)2D3, one with a synthetic VDR agonist] in breast cancer model systems have reported although only one of these datasets (Milani et al., 2013, accession #GSE27220) is publically available on the Gene Expression Omnibus website (http://www*.* ncbi*.*nlm*.*nih*.*gov/geo/). Lee et al. (2006) compared the effects of 1 nM RO3582, a Gemini 1,25(OH)2D3 analog, on pre-malignant (MCF10AT1) and malignant (MCF10CA1a) breast cancer cells and identified distinct gene expression profiles for each cell line. Similar to the comparison of MCF-7 and MDA-MB-231 cells (Swami et al., 2003), more significant changes in gene expression were observed in the less malignant MCF10AT1 cells than in the invasive MCF10CA1a cells (391 vs. 156, respectively, 12 h treatment; 2-fold cutoff). Despite the reduced sensitivity in the more aggressive cells, the overlap in target genes was considerable (about 55% of the genes altered in MCF10AT1 cells were similarly altered in the MCF10CA1a cells); the complete gene lists are available as supplemental data.

Using an approach designed to more accurately represent the tumor microenvironment *in situ*, tumor slices from postmenopausal breast cancer patients with stage I, II, or III breast cancer were cultured with 0.5 or 100 nM 1,25(OH)2D3 for 24 h (Milani et al., 2013). This study identified nine genes that were significantly altered within 24 h of exposure to 0.5 nM 1,25(OH)2D3, a concentration that is physiologically achievable in patients. Of these, CYP24A1 was induced over 7-fold and was validated in another set of tumor samples, clearly indicating activation of VDR signaling. Gene set enrichment analysis (GSEA) indicated a trend toward the enrichment of genes sharing DR3 binding sites, a consensus motif for VDR. Other genes identified in response to 0.5 nM 1,25(OH)2D3 included DPP4, CYP26B1, SPIN3, KCKN3, EFTUD1, TKTL1, and CA2 (up-regulated) and FCGR2C and SAMSN1 (down-regulated). At 100 nM 1,25(OH)2D3, 30 genes (28 up/2 down) were significantly regulated by 1,25(OH)2D3. In addition to those listed above, genes up-regulated by 3-fold or more included IL1RL1, CILP, PI15, TMEM37, and SHE. The two top down-regulated genes (2-fold or more) were P2RY1 and BCOR. Interestingly, CD14 and SLC1A1, two 1,25(OH)2D3 regulated genes we identified in the genomic profiles of normal mammary epithelial cells discussed above, were also induced by 1,25(OH)2D3 in the breast cancer slice model. The significance of this study is that it demonstrated for the first time that 1,25(OH)2D3 could induce genomic changes in intact breast tumor tissue, indicating the functionality of the VDR. Although patient-to-patient variability was considerable, a core set of 1,25(OH)2D3 modulated genes was identified that may represent biomarkers of vitamin D action for future studies.

A fourth genomic study was recently conducted in a mouse mammary tumor model of triple negative breast cancer (Laporta and Welsh, 2013). Cells derived from DMBA induced tumors generated in wildtype (WT) and VDRKO mice were studied after 24 h treatment with 100 nM 1,25(OH)2D3. A unique feature of this study was the inclusion of VDRKO cells in which the growth inhibitory effects of 1,25(OH)2D3 were restored via stable expression of human VDR (KOhVDR cells). Genomic profiling demonstrated that 1,25(OH)2D3 failed to alter gene expression in VDRKO cells whereas major changes were observed in cells derived from WT mice (WT145 cells) and in KOhVDR cells. With a 2-fold cutoff, 117 transcripts in WT145 cells and 197 transcripts in the KOhVDR clones were significantly altered by 1,25(OH)2D3 with 35 genes found to be commonly regulated in all VDRpositive cell lines (the complete list of genes is included in the manuscript). In addition to *Cyp24a1,* seven genes were validated as 1,25(OH)2D3 responsive and VDR dependent in this system: *Cib2*, *Prelp*, *Enpp1*, *Plau, Hbegf, Postn, and Has2*. The last four of these, whose expression was markedly down regulated by 1,25(OH)2D3, are known to drive breast cancer invasion and metastasis. These data support a model whereby 1,25(OH)2D3 coordinately suppresses multiple proteins that are required for survival of triple-negative/basal-like breast cancer cells.

In summary, while CY24A1 is commonly identified in microarray studies as the most highly upregulated gene in 1,25(OH)D treated cells, other target genes vary greatly depending on the model system. Integration of the existing genomic datasets generated in various mammary cell models with other normal and transformed array profiles and ChIP-Seq studies will assist in identifying common and tissue/cell specific genesets regulated by the 1,25(OH)2D3-VDR complex. In addition, the ENCODE project (http://genome*.*ucsc*.*edu/ENCODE/cellTypes*.* html) includes several breast cancer cell lines (MCF-7, MDA-MB-231, T47-D) which may provide relevant genomic information on nuclear receptor signaling The continued use of complex models such as tumor explants for vitamin D studies is desirable given the expression of VDR in most cell types and the critical interactions between tumor cells and their stromal microenvironment.

## **CONCLUSIONS**

Although meta-analyses of population studies demonstrate an inverse relationship between vitamin D status and breast cancer risk, questions remain regarding mechanisms, tissue specificity, and optimal intakes of vitamin D3 for potential benefits on cancer. In 2010, the Institute of Medicine recommended an increase in the adult intake of vitamin D3 (from 200 to 600 IU per day) based on its role in bone health, but concluded that current data is insufficient to support recommendations with respect to cancer prevention. Comprehensive genomic, metabolomic and proteomic profiling approaches combined with mechanistic studies remain highly valuable for identification of relevant biomarkers of tissue vitamin D action that are needed for translational investigations (i.e., supplementation trials).

#### **FUTURE VIEW**

Despite the extensive effort to understand the relationship between vitamin D and breast cancer, many issues remain unresolved. Much of the work conducted in cell systems or animal models is consistent, but epidemiological data is somewhat mixed and clinical studies are limited. As discussed above, population studies do support the concept that high serum levels of 1α,25(OH)2D3 and/or its precursor 25(OH)D3 are associated with lower risk of initial disease development and may retard progression. However, tissue uptake and cellular metabolism of these metabolites *in vivo* is likely to be highly relevant to cancer biology, yet few studies have successfully measured these parameters. In addition, there is little data on how systemic vitamin D status might interact with other known breast cancer risk factors including genetic (BRCA1, BRCA2, ATM), endocrine (estrogen, progesterone) and environmental (radiation, carcinogens) modulators of breast cancer development. Genomic profiling has characterized many 1α,25(OH)2D3 responsive targets in normal mammary cells and in breast cancers providing valuable insight into the molecular actions of 1α,25(OH)2D3 and the VDR in regulation of cell cycle, apoptosis and differentiation. New areas of emphasis suggested by recent studies include regulation of tumor metabolism and activation of innate immune responses. However, the role of VDR in individual cell types (ie epithelial, adipose, fibroblast, endothelial, immune) of normal and tumor tissues remains to be clarified. Furthermore, there has been limited attention directed at understanding how VDR integrates signaling between these diverse cell types and what soluble signals and paracrine pathways may be regulated by 1α,25(OH)2D3 in the tissue and tumor microenvironment. Finally, the possible interactions of VDR with other nuclear receptors and their ligands (particularly RXR family) in control of mammary cell fate/carcinogenesis will require additional study.

#### **REFERENCES**


the European prospective investigation into cancer and nutrition: a nested case-control study. *Int. J. Cancer* 133, 1689–1700. doi: 10.1002/ijc.28172


MCF-7 breast cancer cells. *Mol. Cell. Biochem.* 188, 13–20. doi: 10.1007/978-1- 4615-5763-0\_2


molecular patterns revealed by crossreactivity of anti-LPS antibodies. *Cell Death Differ.* 20, 698–708. doi: 10.1038/cdd.2012.165


**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: 08 April 2014; accepted: 17 May 2014; published online: 13 June 2014. Citation: Narvaez CJ, Matthews D, LaPorta E, Simmons KM, Beaudin S and Welsh J (2014) The impact of vitamin D in breast cancer: genomics, pathways, metabolism. Front. Physiol. 5:213. doi: 10.3389/fphys.2014.00213*

*This article was submitted to Integrative Physiology, a section of the journal Frontiers in Physiology.*

*Copyright © 2014 Narvaez, Matthews, LaPorta, Simmons, Beaudin and Welsh. 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.*

## Vitamin D, intermediary metabolism and prostate cancer tumor progression

## *Wei-Lin W. Wang and Martin Tenniswood\**

*Department of Biomedical Sciences, University at Albany, State University of New York, Albany, NY, USA*

#### *Edited by:*

*Carsten Carlberg, University of Eastern Finland, Finland*

#### *Reviewed by:*

*Stephen Assinder, University of Sydney, Australia Moray J. Campbell, Roswell Park Cancer Institute, USA*

#### *\*Correspondence:*

*Martin Tenniswood, Department of Biomedical Sciences, Cancer Research Center, University at Albany, State University of New York, One Discovery Drive, Rensselaer, NY 12144, USA e-mail: mtenniswood@albany.edu*

Epidemiological data have demonstrated an inverse association between serum vitamin D3 levels, cancer incidence and related mortality. However, the effects of vitamin D on prostate cancer biology and its utility for prevention of prostate cancer progression are not as well-defined. The data are often conflicting: some reports suggest that vitamin D3 induces apoptosis in androgen dependent prostate cancer cell lines, while others suggest that vitamin D3 only induces cell cycle arrest. Recent molecular studies have identified an extensive synergistic crosstalk between the vitamin D- and androgen-mediated mRNA and miRNA expression, adding an additional layer of post-transcriptional regulation to the known VDR- and AR-regulated gene activation. The Warburg effect, the inefficient metabolic pathway that converts glucose to lactate for rapid energy generation, is a phenomenon common to many different types of cancer. This process supports cell proliferation and promotes cancer progression via alteration of glucose, glutamine and lipid metabolism. Prostate cancer is a notable exception to this general process since the metabolic switch that occurs early during malignancy is the reverse of the Warburg effect. This "anti-Warburg effect" is due to the unique biology of normal prostate cells that harbor a truncated TCA cycle that is required to produce and secret citrate. In prostate cancer cells, the TCA cycle activity is restored and citrate oxidation is used to produce energy for cancer cell proliferation. 1,25(OH)2D3 and androgen together modulates the TCA cycle via transcriptional regulation of zinc transporters, suggesting that 1,25(OH)2D3 and androgen maintain normal prostate metabolism by blocking citrate oxidation. These data demonstrate the importance of androgens in the anti-proliferative effect of vitamin D in prostate cancer and highlight the importance of understanding the crosstalk between these two signaling pathways.

**Keywords: vitamin D, androgen, prostate, warburg, miRNA, mRNA**

## **OVERVIEW ON PROSTATE CANCER BIOLOGY**

Prostate cancer is the most commonly diagnosed non-cutaneous malignancy in males in North America (Altekruse et al., 2010). This disease is usually considered to be an androgen dependent cancer, since the normal prostate is clearly dependent on androgens for its structure and function. Paradoxically, the agedependent incidence and associated mortality of prostate cancer between 50 and 60 years of age increase after serum testosterone levels start to decline significantly, particularly after age of 65 (**Figure 1A**) (Siegel et al., 2014). Prostate adenocarcinomas are slow growing tumors that are characterized by low mitotic index and a long natural history (McNeal, 1968). The progression from normal prostate to prostatic intraepithelial neoplasia (PIN), and eventually to localized adenocarcinoma takes place over several decades (**Figure 1B**). Autopsy studies have shown that prostatic adenocarcinoma and the pre-malignant PIN are evident in men in their early and mid-thirties. The development of advanced, locally invasive prostate cancer and metastatic disease is a relatively late process for which there are limited treatments, and hormone ablation therapy used at this late stage applies selective stress that probably is responsible for the development of castration-resistant prostate cancer (CRPC).

## **VITAMIN D AND PROSTATE**

There are many epidemiological studies that suggest high serum vitamin D levels, usually measured as serum 25(OH)-vitamin D3 (25(OH)D3) may be important in preventing various cancers, including breast, ovarian and colon cancer (Thorne and Campbell, 2008; Giovannucci, 2009). The risk of developing and dying of these cancers appears to be inversely correlated with sun exposure, and/or vitamin D status, suggesting that vitamin D has chemopreventive properties (Garland et al., 2009). Some studies have also suggested that vitamin D may play a role in prostate cancer prevention (Tseng et al., 2004; Schwartz and Skinner, 2007), but the data are less conclusive than in other cancers and several recent meta-analyses have found weak or no associations between 25(OH)D3 levels and prostate tumor incidence or progression (Yin et al., 2009; van der et al., 2009; Barnett et al., 2010; Park et al., 2010; Holt et al., 2013). However, a recent study of men diagnosed with prostate cancer showed that 72% of men with recurrent disease and 68% with clinically localized disease were insufficient or deficient in serum 25(OH)D3 levels, less than 20 ng/mL (desirable levels *>*40 ng/mL) (Trump et al., 2010). These data suggest that the majority of men with prostate cancer have low circulating androgen and low 25(OH)D3 levels

at the time of diagnosis. Based on many *in vitro* studies (Miller, 1998; Blutt et al., 2000; Peehl et al., 2003), preclinical and clinical studies (Deeb et al., 2007), it has been suggested that vitamin D can be used either as chemopreventative or as therapeutic agent for prostate cancer. Despite extensive research, the importance of vitamin D as a chemopreventive agent for prostate cancer is still a matter of considerable controversy (van der et al., 2009; Park et al., 2010), and the results from therapeutic intervention using 1,25-dihydroxyvitamin D3 (1,25(OH)2D3), the active metabolite of vitamin D3 or its less calcemic analogs, have been generally disappointing (Vijayakumar et al., 2005; Beer and Myrthue, 2006; Wagner et al., 2013). In low-risk prostate cancer patients who enrolled in active surveillance, high dose of vitamin D3 supplementation decreases Gleason score or the number of positive cores in more than 50% of patient population (Marshall et al., 2012), whereas 1,25(OH)2D3 supplementation at adjuvant settings have provided mixed results in CRPC or recurrent diseases (Flaig et al., 2006; Chan et al., 2008; Srinivas and Feldman, 2009; Chadha et al., 2010; Scher et al., 2011; Shamseddine et al., 2013).

Various reports suggest that the action of vitamin D in prostate cancer cells is androgen dependent (Esquenet et al., 1995; Murthy et al., 2005; Weigel, 2007; Mordan-McCombs et al., 2010). In Sprague–Dawley rats, 1,25(OH)2D3 administration decreases prostatic size in intact males, but not castrated groups (Leman et al., 2003). Longitudinal studies have demonstrated a positive correlation between 25(OH)D3 levels and the production of androgen (Wehr et al., 2010; Pilz et al., 2011; Nimptsch et al., 2012), which has been further validated *in vitro* (Lundqvist et al., 2011). However, vitamin D also induces *CYP3A4* and *CYP3A5* expression, enzymes that metabolize and inactivate testosterone and androstanediol in prostate cells, suggesting that vitamin D signaling may play a role in limiting androgen levels in the prostate (Maguire et al., 2012). Previous *in vitro* studies have shown that 1,25(OH)2D3 also induces moderate increases in *AR*, *PSA*, and *TMPRSS2* transcript levels (Hsieh et al., 1996; Zhao et al., 1999; Krishnan et al., 2004; Washington and Weigel, 2010), however this finding does not translate into clinical setting where 1,25(OH)2D3 appears to decrease the PSA velocity (Krishnan et al., 2003). Based on these findings, serum vitamin D levels appear to have a significant impact on androgen-mediated signaling and the crosstalk between androgen and vitamin D probably plays an important role in prostate cancer biology. While there have been many studies examining the effects of androgens or 1,25(OH)2D3 individually on gene expression in prostate cancer cells, there have been very few studies that explored the crosstalk between the two signaling pathways and the biological consequences of this crosstalk.

## **GENOMIC OVERLAY OF VDR AND AR SIGNALING**

The crosstalk between VDR- and AR-mediated gene expression was first demonstrated in LNCaP cells (Qiao and Tuohimaa, 2004). Induction of *FACL3* (long-chain fatty-acid CoA ligase 3) is dependent on both vitamin D and androgen levels, and treatment with bicalutamide inhibits 1,25(OH)2D3-induced *FACL3* expression. This coordinated effect on gene expression has recently been validated by a comprehensive microarray study using the same *in vitro* model (Wang et al., 2011). 1,25(OH)2D3 and androgen share many common targets and coordinately modulate these transcript levels in the same direction (**Figures 2A,B**). More importantly, the combination of the two hormones regulates additional genes, including both mRNAs and miRNAs that have not been previously identified. The significance of this additional layer of transcriptional control is best illustrated by bioinformatic analysis which demonstrates the coordinated effect of 1,25(OH)2D3 and androgen on cellular processes, including cell homeostasis, proliferation, differentiation and metabolism, all of which have significant impact on prostate tumorigenesis (**Figure 2C**). Most of these processes are more significantly regulated by 1,25(OH)2D3 and androgen together than by either hormone alone, demonstrating the interaction between the two signaling pathways. Several genes identified from the expression microarray analysis are validated VDR and AR targets, contains functional VDRE (within 10kb upstream and 5 kb downstream of the structural gene) and ARE sites, and some genes exhibit additive induction after testosterone and 1,25(OH)2D3 stimulation. Both androgen and 1,25(OH)2D3 induces *PSA* mRNA levels while addition of testosterone blunts the early vitamin D dependent induction of *Cyp24A1*, the main enzyme involved in the catabolism of 1,25(OH)2D3. This suggests that the half-life of 1,25(OH)2D3 is extended in the presence of exogenous androgen.

More than 50% of the responsive genes found from microarray data lack functional response elements in their promoters when comparing to existing genome-wide screens for VDREs and androgen responsive database (Wang et al., 2005; Jiang et al., 2009), raising issues regarding the regulation of these genes, particularly genes that are only expressed if both hormones are present.

The anti-neoplastic effect of vitamin D has been linked to its regulation of miRNA levels. This include the repression of *miR-181ab* expression (Wang et al., 2009) and the induction of *miR-100*, *miR-125b*, and *miR-22* levels by 1,25(OH)2D3 (Alvarez-Diaz et al., 2012; Giangreco et al., 2013). Dysregulated miR-106b expression, which is required for the 1,25(OH)2D3-induced feed-forward loop regulating *p21* expression in non-malignant RWPE-1 cells, has also been implicated in prostate cancer biology (Poliseno et al., 2010; Thorne et al., 2011). Microarray analysis that interrogates the differential miRNA expression in LNCaP cells after treatment with 1,25(OH)2D3 and testosterone, either alone or in combination suggests that VDR plays a critical role in miRNA regulation (Wang et al., 2011) and further highlights the important interactions between VDR- and AR-mediated miRNA expression. These include the additive induction of *miR-22*, *miR-29ab*, *miR-134*, *miR-371-5p*, *miR-663*, and *miR-1207-5p* and the synergistic down-regulation of the oncogenic *miR-17/92* cluster by testosterone and 1,25(OH)2D3. Both *miR-22* and members of the *miR-29* family are candidate tumor suppressors (Alvarez-Diaz et al., 2012; Szczyrba et al., 2013; Wu et al., 2013) and their induction is consistent with the anti-proliferative effect of vitamin D in prostate cancer. In comparison, elevated *miR-371-5p* and *miR-663* expression have been correlated with cancer progression and *miR-663* expression positively associates with the Gleason score used to stage prostate cancer (Zhou et al., 2012; Liu et al., 2013; Jiao et al., 2014). In contrast, the *miR-17/92* cluster is known to play an oncogenic role and its expression has been linked to more advanced prostate cancer (He et al., 2005; Volinia et al., 2006; Sylvestre et al., 2007; Yu et al., 2008; Diosdado et al., 2009; Trompeter et al., 2011; Yang et al., 2013). In addition, this cluster is a well-validated target for c-Myc, which itself is a direct target of VDR (Simpson et al., 1987; O'Donnell et al., 2005), and a recent report has proposed a regulatory role for the *miR-17/92* cluster on PPARα levels, linking *miR-17/92* to energy metabolism in prostate cancer cells (Wang et al., 2013). This concurrent analysis of VDR- and AR-mediated mRNA and miRNA expression reveals an extensive and complex transcription network that interconnects c-Myc, PPARα and other transcription factor-mediated signaling, which is only active when both androgen and vitamin D are present. A recent comprehensive analysis of 24 nuclear receptors and 14 transcription factors (TFs) in the MCF-7 breast cancer cell line has demonstrated a similar finding and has identified genomic regions enriched with nuclear receptors and TFs binding sites, which generates extensive regulatory networks that may modulate target gene expression (Kittler et al., 2013). Such functional interactions between nuclear receptors and TFs, including the antagonistic interaction between RARs and AR and PPARδ (Rivera-Gonzalez et al., 2012; Kittler et al., 2013), and the agonistic interaction between VDR and AR (Wang et al., 2011) provide valuable information that can be used to improve cancer prevention and therapy. The functional interactions between AR and VDR, as well as other nuclear receptors and TFs may also be important for disease management, especially now that nutritional intervention has become more widely accepted as an effective approach to prevent cancer progression. These experimental data suggest that 1,25(OH)2D3, and androgens as well as other hormones and growth factors trigger at least three mechanisms to modulate global gene expression. These include AR- and VDR-mediated gene transactivation; miRNA-mediated mRNA degradation and translational repression; and transcription factor-mediated feed-forward signaling. These mechanisms do not appear to be mutually exclusive and act together to regulate many vitamin D- and androgen-mediated cellular processes that have significant implication in prostate carcinogenesis.

## **INTERMEDIATE METABOLISM: THE WARBURG EFFECT**

A number of studies have suggested that vitamin D has a novel role in regulating energy metabolism. The vitamin D receptor knockout (VDRKO) and the *Cyp27b1* knockout (Cyp27b1KO) mice exhibit elevated energy expenditure with subsequent loss of body fat over time (Narvaez et al., 2009; Wong et al., 2011). In human adipocytes, 1,25(OH)2D3 inhibits uncoupling protein-1 expression and alters Ca2<sup>+</sup> homeostasis, suggesting a regulatory role of vitamin D in thermogenesis and provides

expression microarray data [Green: 1,25(OH)2D3-moduated, Orange: androgen-modulated, Blue: synergistically modulated genes by androgen and 1,25(OH)2D3]. **(B)** Venn diagram analysis of the miRNA microarray data

enrichment analysis of representative gene sets identified as significantly enriched after 1,25(OH)2D3 treatment in the presence or absence of androgen in LNCaP cells. False discovery rate *<*5%.

rationale for the observed lean phenotype in VDRKO and Cyp27b1KO mice (Xue et al., 1998; Shi et al., 2001, 2002). Similarly, both 25(OH)D3 and 1,25(OH)2D3 promotes lipogenesis in primary human preadipocytes, adipocyte and adiposederived mesenchymal progenitor cells, which is associated with increased expression of differentiation markers *C/EBPα* and *PPARγ* (Nimitphong et al., 2012; Narvaez et al., 2013). However, this effect may be cell type and lineage specific since 1,25(OH)2D3 inhibits lipid accumulation in mouse 3T3-L1 preadipocytes and prevents high fat diet-induced fatty liver syndrome in Sprague–Dawley male rats (Rayalam et al., 2008; Yin et al., 2012).

In T47D breast cancer cells, 1,25(OH)2D3 induces lipid synthesis, which has been associated with its effect on cell differentiation and reduced cell growth (Lazzaro et al., 2000). This lipogenic effect of 1,25(OH)2D3 is recapitulated in LNCaP cells and is enhanced in the presence of androgen (Esquenet et al., 1997; Wang et al., 2013), highlighting the coordinated effect of AR and VDR signaling. Increase in PPARα expression and its associated lipogenic gene signature, including the elevation of fatty acid synthase (*FASN*) expression, accounts for vitamin D- and androgen-induced lipid production. However, this occurs without significant changes in nuclear sterol regulatory element-binding protein (SREBP-1) levels. Nuclear activation of SREBP-1 has been implicated in *de novo* lipogenesis in more aggressive cancers, including prostate cancer (Menendez and Lupu, 2007; Huang et al., 2012). A recent comprehensive parallel analysis of various genomic studies using prostate cancer cell lines has uncovered a critical regulatory role of AR in the energy metabolic network, with lipid synthesis being the predominate AR-regulated process. These data suggest that altered AR signaling and its effects on the downstream targets of calcium/calmodulin-dependent protein kinase kinase 2, beta (*CAMKK2*), which regulates the activity of a key energy sensor AMP-activated protein kinase (AMPK), promotes the metabolic switch that provides the energy for prostate cancer growth and progression (Massie et al., 2011). These data suggest a divergent role of lipid production in prostate tumors: SREBP-1 dependent up-regulation of fatty acids production for phospholipid and membrane synthesis and signaling molecules that are essential for tumor progression (Currie et al., 2013; Soga, 2013); or SREBP-1 independent elevation of neutral and inactive lipid accumulation which restricts energy expenditure and limits tumor growth.

In addition to the modulation of lipid metabolism by vitamin D and androgen, qPCR analysis has suggested a regulatory role of these two hormones on the TCA cycle in prostate cancer cells. In most normal cells, the TCA cycle is utilized to generate energy for normal cellular functions. This process is relatively slow and ATP production does not meet the demand for highly proliferative cancer cells. As a result, cancer cells often disengage mitochondrial oxidative phosphorylation from glycolysis for rapid ATP production by employing the fermentation process, a process referred as the Warburg effect (Warburg et al., 1927; Warburg, 1956; Soga, 2013). Prostate cancer cells are a notable exception, and the metabolic switch that occurs is more appropriately regarded as an "anti-Warburg" effect. The prostate gland normally secrets high levels of citrate into the seminal fluid, a function that is supported by a truncated TCA cycle activity. The prostate has the highest levels of intracellular zinc of any tissue in the body. This high level of zinc inactivates maconitase 2 activity, the enzyme that converts citrate to isocitrate in the mitochondria. In prostate cancer cells, zinc transporters are down-regulated, which leads to lower intracellular zinc levels. This restores m-aconitase 2 function and the conversion of citrate to isocitrate for ATP production via the TCA cycle (Costello and Franklin, 1991a,b). This is supported by both clinical and *in vitro* data, demonstrating a minimal reliance of prostate cancer cells on glycolysis for proliferation especially during the early phases of tumor progression. This precludes the usage of fluorine-18-labeled 2-deoxy-2-fluoro-*D*-glucose (FDG-PET) for prostate cancer detection and diagnosis (Hofer et al., 1999; Jadvar, 2011). In comparison, androgen stimulates glucose usage to facilitate citrate accumulation in normal prostate epithelial cells (Harkonen, 1981; Harkonen et al., 1982) and this androgenic effect is maintained in androgen responsive prostate cancer cells, although in these cells, elevated citrate is funneled for the production of lipid (Moon et al., 2011).

In LNCaP cells, 1,25(OH)2D3 and androgen together downregulate mitochondrial thiamine pyrophosphate (TPP) carrier (*SLC25A19*) and up-regulates two zinc transporters, (*SLC39A1* and *SLC39A11*) (supplemental data to Wang et al., 2011). Low expression of *SLC39A1* in adenocarcinomatous glands and PIN foci has been documented and linked to depleted zinc levels (Franklin et al., 2005). In comparison, *SLC39A11* is less well-characterized, but studies have shown that it is abundantly expressed in murine testes and digestive system, and is associated with zinc import (Yu et al., 2013). This suggests that vitamin D and androgen cooperate to reset zinc levels, inhibiting m-aconitase activity in prostate cancer cells. In comparison, down-regulation of the TPP carrier, *SLC25A19* (Lindhurst et al., 2006; Kang and Samuels, 2008) affects mitochondrial coenzyme TPP levels, leading to decreased activities of pyruvate dehydrogenase (PDH) and alpha-ketoglutarate dehydrogenase (OGDH) activities. In comparison, *in vivo* studies have shown that administration of testosterone up-regulates the expression and activities of PDH and mitochondrial aspartate aminotransferase to increase the substrate pools for citrate synthesis, acetyl-CoA and oxaloacetate (Costello and Franklin, 1993; Qian et al., 1993). This suggests that vitamin D and androgen supplementation facilitate the reversion of the metabolic switch that occurs during prostate carcinogenesis by preventing citrate oxidation, partially restoring the normal prostatic function and shunting citrate into the cytoplasm for secretion and lipid synthesis (**Figure 3**). This is supported by the observation that LNCaP cells retain the sensitivity to androgen-induced citrate production and accumulation (Franklin et al., 1995). This suggests that vitamin D facilitates and maintains this differentiated phenotype, rendering prostate cancer cells less aggressive. This also suggests that maintaining or restoring adequate levels of androgen, accompanied by vitamin D supplement will significantly delay prostate cancer progression in aging men.

To further highlight the impact of vitamin D and androgen on resetting cancer cell metabolism, 1,25(OH)2D3 and androgen also down-regulate c-Myc levels, whose many functions include metabolic reprogramming to drive tumor progression, including the induction of glycolysis and glutaminolysis (Shim et al., 1998; Wise et al., 2008; Soga, 2013; Zirath et al., 2013). While there is good evidence suggesting a positive correlation between serum glutamate levels and more aggressive prostate cancer (Koochekpour et al., 2012), the dependence of prostate cancer on glutaminolysis for energy generation and progression is not well-studied. Nevertheless, it is reasonable to suggest that in response to vitamin D and androgen stimulation, prostate cancer cells reverse or block the metabolic switch that occurs early in the course of the disease and further blocks c-Myc-mediated metabolic reprogramming, which may occur independently of the initial metabolic switch.

## **CONCLUSION**

Recent studies have shown a complex relationship between vitamin D3- and androgen-mediated signaling in the normal prostate and prostate cancer through their coordinated effect on mRNA and miRNA transcription, cell proliferation and cancer metabolism. These data suggest that the effect of vitamin D3 on global gene expression is dependent on the activity of androgen and their combined effect on miRNA transcription and other TFs. Phenotypically, the two hormones maintain normal prostatic metabolism to prevent de-differentiation of prostate cancer

cells into more aggressive phenotype. These newly emerging data provide a explanation for the discrepancies observed from epidemiological and experimental studies of vitamin D3 in prostate cancer since these studies do not take the synergistic interactions between the two pathways into account. These data also suggest that maintenance of adequate levels of vitamin D3 and androgen will slow or halt prostate cancer progression especially for patients diagnosed with early stage, locally confined disease. Case-control clinical studies will be needed to fully evaluate the risk and benefit of combining these two hormones in prostate cancer patients.

## **ACKNOWLEDGMENTS**

Wei-Lin W. Wang would like to acknowledge the DAMD for Predoctoral support (PC102080).

## **REFERENCES**


induces apoptosis in prostate cancer cells. *Clin. Cancer Res.* 18, 5888–5901. doi: 10.1158/1078-0432.CCR-12-1308


**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: 11 March 2014; accepted: 22 April 2014; published online: 15 May 2014. Citation: Wang W-LW and Tenniswood M (2014) Vitamin D, intermediary metabolism and prostate cancer tumor progression. Front. Physiol. 5:183. doi: 10.3389/ fphys.2014.00183*

*This article was submitted to Integrative Physiology, a section of the journal Frontiers in Physiology.*

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

## The future of vitamin D analogs

## *Carlien Leyssens , Lieve Verlinden and Annemieke Verstuyf\**

*Clinical and Experimental Endocrinology, Department of Clinical and Experimental Medicine, KU Leuven, Leuven, Belgium*

#### *Edited by:*

*Carsten Carlberg, University of Eastern Finland, Finland*

#### *Reviewed by:*

*Ferdinand Molnár, University of Eastern Finland, Finland Antonio Mourino, Universidad de Santiago, Spain*

#### *\*Correspondence:*

*Annemieke Verstuyf, Clinical and Experimental Endocrinology, Department of Clinical and Experimental Medicine, KU Leuven, Herestraat 49, bus 902, 3000 Leuven, Belgium e-mail: mieke.verstuyf@ med.kuleuven.be*

The active form of vitamin D3, 1,25-dihydroxyvitamin D3, is a major regulator of bone and calcium homeostasis. In addition, this hormone also inhibits the proliferation and stimulates the differentiation of normal as well as malignant cells. Supraphysiological doses of 1,25-dihydroxyvitamin D3 are required to reduce cancer cell proliferation. However, these doses will lead *in vivo* to calcemic side effects such as hypercalcemia and hypercalciuria. During the last 25 years, many structural analogs of 1,25-dihydroxyvitamin D3 have been synthesized by the introduction of chemical modifications in the A-ring, central CD-ring region or side chain of 1,25-dihydroxyvitamin D3 in the hope to find molecules with a clear dissociation between the beneficial antiproliferative effects and adverse calcemic side effects. One example of such an analog with a good dissociation ratio is calcipotriol (Daivonex®), which is clinically used to treat the hyperproliferative skin disease psoriasis. Other vitamin D analogs were clinically approved for the treatment of osteoporosis or secondary hyperparathyroidism. No vitamin D analog is currently used in the clinic for the treatment of cancer although several analogs have been shown to be potent drugs in animal models of cancer. Transcriptomics studies as well as *in vitro* cell biological experiments unraveled basic mechanisms involved in the antineoplastic effects of vitamin D and its analogs. 1,25-dihydroxyvitamin D3 and analogs act in a cell type- and tissue-specific manner. Moreover, a blockade in the transition of the G0/1 toward S phase of the cell cycle, induction of apoptosis, inhibition of migration and invasion of tumor cells together with effects on angiogenesis and inflammation have been implicated in the pleiotropic effects of 1,25-dihydroxyvitamin D3 and its analogs. In this review we will give an overview of the action of vitamin D analogs in tumor cells and look forward how these compounds could be introduced in the clinical practice.

#### **Keywords: vitamin D, analogs, pleiotropic effects, cancer**

## **INTRODUCTION**

The active form of vitamin D3, 1,25-dihydroxyvitamin D3 **[**1α,25(OH)2D3; **1]** (**Table 1**), is mostly known for its effects on bone, calcium, and phosphate homeostasis. Next to these classical effects, 1,25(OH)2D3 also exerts so-called non-classical effects on various tissues which express the vitamin D receptor (VDR) as well as the enzymes that are responsible for activating the hydroxylations of vitamin D3, which is essential for the formation of 1,25(OH)2D3. Thus, most tissues have the ability to convert vitamin D3 into its active form, 1,25(OH)2D3, which in turn will bind the VDR in order to positively or negatively influence target genes via binding of the 1,25(OH)2D3/VDR complex to vitamin D receptor elements (VDRE). Non-classical properties of 1,25(OH)2D3 include prodifferentiating and antiproliferative effects on normal and cancer cells (Colston et al., 1981; Jensen et al., 2001) as well as immunomodulatory effects. However, in order to obtain these non-classical effects, 1,25(OH)2D3 doses of the nanomolar range are necessary, while physiological 1,25(OH)2D3 serum concentrations are in the picomolar range. Since supraphysiological doses of 1,25(OH)2D3 result in hypercalcemia, 1,25(OH)2D3 analogs were developed to minimize the calcemic side effects while preserving or augmenting the beneficial effects of 1,25(OH)2D3. Both industry and academic institutions have synthesized a vast amount of vitamin D analogs. Some of these analogs have tissue-specific effects with low calcemic side effects and can be given at higher doses compared to the mother compound.

## **CLINICALLY APPROVED VITAMIN D ANALOGS**

Given the huge amount of vitamin D analogs that have been synthesized during the years, it is nearly impossible to give an overview of them all. In the first part we will discuss vitamin D analogs that are clinically approved (**Table 2**). For several conditions such as secondary hyperparathyroidism, psoriasis and osteoporosis, vitamin D analogs are frequently used as a treatment option. Paricalcitol **[2]** and doxercalciferol **[3]** are vitamin D2 analogs approved for therapeutic use of secondary hyperparathyroidism. In Japan falecalcitriol **[4]** and maxacalcitol **[5]** are also used to treat this disease. Secondary hyperparathyroidism is characterized by elevated parathyroid hormone (PTH) levels in response to hypocalcemia and is often caused by chronic kidney disease. Above-mentioned vitamin D analogs suppress PTH, as does 1,25(OH)2D3, but without inducing severe hypercalcemia. Clinical studies with chronic kidney disease patients show that different analogs induce a stronger PTH suppression compared to placebo treatment (Hamdy et al., 1995; Coburn et al., 2004; Coyne et al., 2006). Also, end-stage renal disease patients treated with these analogs often have a better survival (Teng et al., 2003;


#### **Table 1 | Overview of vitamin D analogs.**

## **Table 1 | Continued**

**Table 1 | Continued**


Tentori et al., 2006; Shinaberger et al., 2008). However, few studies with chronic kidney disease and end-stage renal disease patients directly compare the effects of 1,25(OH)2D3 with its analogs.

Psoriasis, a hyperproliferative condition of the skin, is also treated with vitamin D analogs. Tacalcitol **[6]**, calcipotriol **[7]** and the recently approved maxacalcitol **[5]** are used either as monotherapy or in combination with topical steroids such as betamethasone dipropionate to treat psoriasis. The analogs exert prodifferentiating and antiproliferative effects on keratinocytes and also possess important anti-inflammatory properties. Furthermore, alfacalcidol (**[**1α(OH)D3; **8]**, actually a pre-metabolite of 1,25(OH)2D3) and eldecalcitol (ED-71) **[9]** are used in Japan in the treatment of osteoporosis. The recently approved eldecalcitol **[9]** is more effective than 1,25(OH)2D3 and alfacalcidol **[8]** in increasing bone mineral density and mechanical strength in ovariectomized rats (Uchiyama et al., 2002). Various studies in mouse models as well as in patients show that treatment with eldecalcitol **[9]** leads to higher lumbar and hip bone mineral density and a lower incidence of new vertebral fractures (Ito et al., 2011; Matsumoto et al., 2011; Harada et al., 2012; Hagino et al., 2013), making eldecalcitol **[9]** a very promising new analog for the treatment of osteoporosis.

## **GENOME- AND TRANSCRIPTOME-WIDE EFFECTS OF VITAMIN D ANALOGS**

The exact mechanism of action of vitamin D analogs still has to be deciphered. The reason why specific analogs have superagonistic actions on specific tissues remains unknown, however several studies have tried to elucidate the mechanisms behind these tissue-specific effects. The catabolism of vitamin D analogs is one of the mechanisms that have an effect on their potency. Modifications of the side chain of 1,25(OH)2D3 are known to slow down its catabolism by CYP24A1 (Jones, 1997). Seocalcitol (EB1089) **[10]** and 20-epi-1,25(OH)2D3 **[11]** are degraded slower compared to 1,25(OH)2D3 leading to a longer exposure of these analogs to the tissues (Hansen and Maenpaa, 1997; Kissmeyer et al., 1997; Shankar et al., 1997; Zella et al., 2009). The metabolites formed after catabolism of lexicalcitol **[12]** are more active than the ones formed after 1,25(OH)2D3 is catabolized (Dilworth et al., 1997) and this analog is also more effective in slowing down the degradation rate of the VDR (van den Bemd et al., 1996). Moreover, since some cell types prefer specific catabolism pathways and enzymes above others, the degradation process may also contribute to the tissue-specific activity of vitamin D analogs. The affinity for the vitamin D binding protein (DBP) also plays a role in the activity of vitamin D analogs. Maxacalcitol **[5]** for example has a 500 times lower affinity for DBP and is thus cleared faster from the circulation than 1,25(OH)2D3 (Okano et al., 1989a). This analog has a short effect on bone and intestine, tissues responsible for calcium homeostasis, and a longer effect on PTH levels, making this analog ideal for the treatment of secondary hyperparathyroidism. However, it is still unknown why the duration of the effects is different in these tissues. Eldecalcitol **[9]** on the other hand has a higher DBP affinity compared to the mother compound, leading to longer sustained plasma levels and is thus more suitable for the treatment of osteoporosis (Okano et al., 1989b).

Another mechanism that contributes to the superagonistic effects of vitamin D analogs is their interaction with the VDR, co-activators and VDREs. 20-epi-1,25(OH)2D3, a C-20 epimer of 1,25(OH)2D3 **[11]**, promotes heterodimerization between VDR and retinoid X receptor (RXR) (Liu et al., 2001). 20-epi-1,25(OH)2D3 **[11]** and other analogs like maxacalcitol **[5]**, CD578 **[13]**, inecalcitol **[14],** and TX527 **[15]** require lower concentrations to recruit specific coactivators to the VDR/RXR/VDRE complex (Liu et al., 2000; Eelen et al., 2005, 2008; Schwinn and DeLuca, 2007). Approximately 10- fold lower doses of inecalcitol **[14]** and TX527 **[15]** are needed, compared to 1,25(OH)2D3, to acquire the same amount of co-activator interactions (Eelen et al.,

#### **Table 2 | Overview of clinically approved vitamin D analogs.**


2005). Vitamin D analogs might also be able to induce tissuespecific effects by favoring binding to specific *VDRE* motifs in target gene promoters. Analogs with a 20-methyl group as well as seocalcitol **[10]** bound to a VDR/RXR complex preferably interact with the IP9 type of VDRE (Danielsson et al., 1996; Quack and Carlberg, 1999).

On the genome level, studies using chromatin immunoprecipitation (ChIP) and micro-array techniques have investigated 1,25(OH)2D3-regulated genes in different cell lines. One ChIP study compared the binding sites of the VDR in intestinal tissue after 1,25(OH)2D3 or 20-epi-1,25(OH)2D3 **[11]** treatment. This study shows that both 1,25(OH)2D3 and 20-epi-1,25(OH)2D3 **[11]** induce VDR binding to *CYP24A1* and *TRPV6* loci in the intestine, but the analog elicits a prolonged VDR binding to these genes leading to its superagonistic characteristics such as hypercalcemia *in vivo* (Zella et al., 2009). Other ChIP studies have tried to investigate the molecular mechanisms of some analogs in different tissues. In osteoblast cell models 2MD **[16]** bound to the VDR is able to bind VDREs at lower concentrations compared to 1,25(OH)2D3 (Yamamoto et al., 2003). Seocalcitol **[10]**, on the other hand, mediates the dissociation of Williams syndrome transcription factor of the aromatase promoter leading to inhibition of aromatase expression and activity in breast cancer cells which is one of the main therapeutic strategies in breast cancer patients (Lundqvist et al., 2013). In a recent paper binding sites of VDR and mothers against decapentaplegic homolog 3 (SMAD3) were investigated in hepatic stellar cells. These transforming growth factor β1 (TGFβ1)-activated cells play an important role in liver fibrosis. In this study it is shown that VDR and SMAD3 can at least transiently co-occupy the same genomic sites and function as enhancers of pro-fibrotic gene expression. However, when calcipotriol **[7]** is added, the TGFβ1-induced recruitment of SMAD3 is compromised and binding of VDR to these genomic sites is enhanced 10-fold meaning that liganded VDR antagonizes SMAD residency on chromatin and thereby suppresses pro-fibrotic gene expression (Ding et al., 2013). This genomic feedback circuit is a previously unknown mechanism of calcipotriol **[7]**.

Micro-array studies in various cancer cell lines such as leukemia, prostate, breast, colorectal, and ovarian cancer show that a variety of gene clusters are influenced by 1,25(OH)2D3 and its analogs (reviewed in Kriebitzsch et al., 2009). Cell growth, apoptosis, cellular adhesion and extracellular matrix composition, oxidative stress, immune function, intra- and intercellular signaling and steroid/lipid metabolism are frequently modulated processes in cells by 1,25(OH)2D3 and its analogs. However, when different cell lines are compared, few 1,25(OH)2D3/analogregulated genes overlap, which suggests that 1,25(OH)2D3 and its analogs behave in a cell type- and tissue-specific way. Also studies using human T-cells (Baeke et al., 2011), rat ventricular heart tissue (Bae et al., 2011), and bone marrow-derived mouse dendritic cells (Griffin et al., 2004) have researched the impact of 1,25(OH)2D3 analogs on gene expression. In these studies genes important for cell growth, cell death and cell signaling are regulated, but also a large set of genes implicated in the migration of T-cells and dendritic cells are influenced. TX527 **[15]** imprints human T-cells with a migratory signature and targets them to sites of inflammation (Baeke et al., 2011). Paricalcitol **[2]** treatment of rats with cardiac hypertrophy prevents the progression of cardiac hypertrophy and the development into chronic heart failure. The genomic changes associated with cardiac hypertrophy in the ventricular heart tissue of these rats are, in part, reversed by paricalcitol **[2]** administration (Bae et al., 2011). Furthermore, other studies investigated if 1,25(OH)2D3 analogs are able to bind and regulate different genes compared to 1,25(OH)2D3. All conducted studies conclude that 1,25(OH)2D3 and its analogs induce or repress the same set of genes. Seocalcitol **[10]** induces a less malignant phenotype in SCC25 squamous cell carcinoma cells and modulates expression of genes important in cell cycle progression, cell adhesion, extracellular matrix composition, intraand intercellular signaling, G-protein coupled function, redox balance, and steroid metabolism. In these cells, seocalcitol **[10]** regulates the same genes compared to 1,25(OH)2D3, however gene regulation by 1,25(OH)2D3 is more transient (Lin et al., 2002). Also WY1112, a seco-9,11-*bis*nor-17-methyl analog lacking the C-ring and with a 21-epi side chain which is fluorinated on C26 and C27 **[17]**, was investigated in MCF-7 breast cancer cells. Despite the 400-fold stronger antiproliferative capacity of WY1112 **[17]**, the same genes are upregulated after 1,25(OH)2D3 or WY1112 **[17]** treatment. However, the induction ability is much higher for the analog (Vanoirbeek et al., 2009). When treating human coronary artery smooth muscle cells with equal amounts of 1,25(OH)2D3 or paricalcitol **[2]** the same genes are regulated (Wu-Wong et al., 2007; Shalhoub et al., 2010). In conclusion, differences in action and capacity of vitamin D analogs are more due to their specific sensitivities to metabolism and their specific interaction with the VDR, co-activators and VDREs than from different gene regulations. However, to our knowledge no studies have yet looked into the potential differences elicited by analogs compared to 1,25(OH)2D3 in the fields of proteomics and epigenetics, which could help to understand the molecular mechanism of 1,25(OH)2D3 and its analogs on different cell and tissue types.

## **EFFECTS OF VITAMIN D ANALOGS IN CANCER**

The use of 1,25(OH)2D3 for the treatment of cancer gained interest since many tissues express vitamin D metabolizing enzymes as well as the VDR and because 1,25(OH)2D3 has potent antiproliferative and prodifferentiating effects on normal and malignant cell lines. Several analogs evaluated *in vitro* show stronger antiproliferative and prodifferentiating effects compared to the mother compound in different cancer cell lines. These compounds are further evaluated in rodent models to assess their anti-cancer activity and safety *in vivo*. All *in vivo* studies using rodent cancer models that were published between 2007 and 2013 are summarized in **Table 3**. For studies preceding 2007, the reader is referred to other reviews (Eelen et al., 2007). In most studies the growth of the tumor is inhibited without inducing severe hypercalcemia when appropriate doses of vitamin D analogs are used (Abe et al., 1991; Kawa et al., 1996, 2005; Akhter et al., 1997, Blutt et al., 2000; Prudencio et al., 2001; Grostern et al., 2002; Flanagan et al., 2003; Albert et al., 2004a; Wietrzyk et al., 2004; Zhang et al., 2005; Fichera et al., 2007; van Ginkel et al., 2007; Ghous et al., 2008; Lee et al., 2008; Schwartz et al., 2008; Gonzalez-Pardo et al., 2010; Seubwai et al., 2010; Berkovich et al., 2013;


#### **Table 3 |** *In vivo* **studies in rodent cancer models treated with vitamin D analogs (intraperitoneal i.p.; subcutaneous s.c.) published between 2007 and 2013.**

*(Continued)*


*For studies preceding 2007 the reader is referred to other reviews (Eelen et al., 2007).*

Chiang et al., 2013). However, in some models the analog dose that is effective in inhibiting tumor growth also causes hypercalcemia and lower survival of the treated animals (Albert et al., 2004b). Not only tumor proliferation is modulated by vitamin D analogs, also apoptosis, angiogenesis, migration of tumor cells, etc. are affected by some analogs. In xenograft studies where apoptosis in the tumor was investigated after vitamin D analog treatment, apoptosis or the necrotic field in the tumor is augmented (James et al., 1998; VanWeelden et al., 1998; Hara et al., 2001; Vegesna et al., 2003; Lambert et al., 2010; Park et al., 2012). Inecalcitol (Hybrigenics, France) **[14]** treatment of mice with squamous cell carcinoma xenografts increases apoptosis in the tumors and this increase is higher for the analog compared to 1,25(OH)2D3, while the capacity of the analog to inhibit proliferation is equal compared to the mother compound (Ma et al., 2013). Most studies agree that vitamin D analogs also have an effect on tumor metastasis. Seocalcitol **[10]** reverses the growth-stimulatory effects of parathyroid hormone-related protein (PTHrP), which plays a major role in prostate cancer progression and metastasis, in a xenograft mouse model of prostate cancer. The same study shows that seocalcitol **[10]** also inhibits migration and invasion of these prostate cancer cells *in vitro* (Bhatia et al., 2009). This analog also reduces the number and surface area of bone metastasis originating from intracardially injected breast cancer cells (El Abdaimi et al., 2000). Vitamin D analogs are thus able to reduce the number and growth of metastasis originating from various types of cancer cells (Sato et al., 1984; Lokeshwar et al., 1999; Nakagawa et al., 2005; Park et al., 2012). However, in a study using mice with chemically induced breast cancer, the invasion capacity of the tumor after seocalcitol **[10]** treatment is not affected (Liska et al., 2012). The effect of vitamin D analogs on angiogenesis has also been studied *in vivo*, but here the results are more conflicting. Some studies show no effect of vitamin D analogs on angiogenesis (Oades et al., 2002; Valrance et al., 2007), while others find decreased angiogenesis of xenograft tumors. Intraperitoneal injections of inecalcitol **[14]** decrease the vascularity of xenografted prostate cancer cells (Okamoto et al., 2012) and oral treatment of colorectal tumors in rats with alfacalcidol **[8]** also inhibits tumor angiogenesis (Iseki et al., 1999). All *in vivo* studies conclude that vitamin D analogs inhibit tumor growth but vitamin D and its analogs often do not influence tumor number. Seocalcitol **[10]** was given as chemoprevention in a transgenic mouse model for androgen-independent prostate cancer. Tumor growth is inhibited, however, there is no prevention in the development of tumors (Perez-Stable et al., 2002).

Since vitamin D and its analogs do not possess cytostatic properties, many *in vivo* studies have focused on vitamin D analog cancer treatment combined with radiotherapy and/or chemotherapy. When seocalcitol **[10]** is combined with radiotherapy in a xenograft model for breast cancer, the anti-cancer effects are more effective compared to monotherapy (Sundaram et al., 2003). Another analog, tacalcitol **[6]**, has been investigated in colorectal cancer xenograft in combination with different standard chemotherapies. Different concentrations as well as administration routes of tacalcitol **[6]** or PRI-2205 (an analog of calcipotriol) **[18]** were used in combination with different concentrations of 5 fluorouracil (5-FU). Using specific analog doses and chemotherapy schedules, a synergistic effect on the prolongation of the life span of the mice is achieved (Milczarek et al., 2013a). Also the combination with irinotecan or oxaliplatin was investigated. In this study the mice also show a longer life span and a stronger tumor growth inhibition compared to monotherapy when certain doses of tacalcitol **[6]** and chemotherapy were used. However, some combinations were more toxic than the monotherapies (Milczarek et al., 2013b). Other studies report better effects when combining calcipotriol **[7]** and diclofenac in a non-melanoma skin cancer model (Pommergaard et al., 2013). However, the combination of vitamin D analogs with chemotherapy does not always result in additive or synergistic effects. Combining maxacalcitol **[5]** and 5-FU did not enhance anti-tumor effects in a chemically induced breast cancer model (Iino et al., 1992). Another study investigated calcipotriol **[7]** and its derivatives in breast and lung cancer *in vivo* models and concluded that these analogs and low dose cytostatics are not effective in the used models (Wietrzyk et al., 2007). Also tacalcitol **[6]** in combination with cyclophosphamide does not lead to a significant difference in tumor growth inhibition compared to the vehicle treatment (Wietrzyk et al., 2008).

In view of the promising results that certain vitamin D analogs show against cancer *in vitro* and *in vivo* animal models, some analogs have been tested in cancer patients (**Table 4**). Seocalcitol **[10]** is an analog that has been extensively studied *in vitro* and *in vivo* in different cancer models, however in clinical trials the results are rather disappointing. Patients with advanced breast or colon cancer were treated with different doses of seocalcitol **[10]** (most patients tolerate 7μg/d) but none of them showed a complete or partial response (Gulliford et al., 1998). Also oral seocalcitol **[10]** treatment in patients with inoperable pancreatic cancer exhibited no objective anti-tumor activity (Evans et al., 2002). Two out of 33 patients with inoperable hepatocellular carcinoma showed a complete response after oral seocalcitol **[10]** treatment, however the majority of the patients presented stable or progressive disease (Dalhoff et al., 2003). Inecalcitol **[14]** is in an early stage II of its clinical trial in chronic lymphocytic leukemia. Fifteen patients received 2 mg/d orally and one patient had a 90% decrease in blood lymphocyte count after 10 months of treatment, in 8 other patients blood lymphocyte count stopped growing when the treatment started (Hybrigenics, 2014). Intravenous administration of paricalcitol **[2]**, an analog that is approved for secondary hyperparathyroidism, also displayed no objective responses in patients with androgen-independent prostate cancer. However, elevated serum PTH levels, which are common for advanced prostate cancers, are reduced by the analog (Schwartz et al., 2005). Doxercalciferol **[3]**, also used in the treatment against secondary hyperparathyroidism, was investigated in androgen-independent prostate cancer patients. A phase I study administered oral doxercalciferol **[3]** between 5 and 15μg/d, which was well tolerated by the patients (Liu et al., 2002). In the following phase II study, patients were treated with 12.5μg/d for a minimum of 8 weeks and 30% of these patients experienced stable disease for over 6 months (Liu et al., 2003). Oral treatment of non-Hodgkin's lymphoma patients with 1μg/d alfacalcidol **[8]**, a pre-metabolite of 1,25(OH)2D3, resulted in a low overall response. Out of 34 treated patients, only 4 had a complete response and 4 others showed a partial response to the treatment (Raina et al., 1991). Calcipotriol **[7]** is often used to treat skin psoriasis and has thus been investigated in patients with locally advanced or cutaneous metastases from breast cancer. In both studies the analog was applied topically at a dose of 100μg/d. One study reported no response after 3 months of treatment (O'Brien et al., 1993), while in the other study 3 patients showed a 50% reduction in the diameter of treated lesions after 6 weeks (Bower et al., 1991). A more recently developed analog, ILX23-7553 **[19]**, was investigated in 16 patients with advanced solid tumors but no objective response was seen (Jain et al., 2011). Similar to the *in vivo* studies, clinical trials have also combined vitamin D analogs with standard radiotherapy or chemotherapy. Metastatic breast cancer patients were given oral paricalcitol **[2]** doses between 2 and 7μg/d in combination with taxane-based chemotherapy and this regimen was well tolerated by the patients (Lawrence et al., 2013). Oral inecalcitol **[14]** was given to patients with hormone-refractory prostate cancer in combination with docetaxel for maximum 18 weeks. This study had a response rate of 85% based on a PSA decline of at least 30% within 3 months of treatment (Hybrigenics, 2014). In a small study with acute non-lymphoid leukemia patients the combination of alfacalcidol **[8]** and chemotherapy resulted in 17% of the patients with a complete response and 45% with a partial response (Petrini et al., 1991). The same analog was combined with standard treatment of surgery, radiotherapy, and/or chemotherapy in glioblastoma and anaplastic astrocytomas. Here, 0.04μg/kg/d alfacalcidol **[8]** was administered resulting in 27% of the patients with progressive regression of the lesion and complete clinical remission (Trouillas et al., 2001). In metastatic renal cell carcinoma patients, oral treatment of 1μg alfacalcidol/d **[8]**

#### **Table 4 | Clinical trials with vitamin D analog supplementation.**



*(Continued)*


was combined with a 3 weekly administration of interferon-α for minimal 3 months. In these patients 25% had a partial response to the combination therapy (Obara et al., 2008). Randomized, placebo-controlled studies have been conducted with oral doxercalciferol **[3]** or alfacalcidol **[8]**. One study administered 10μg/d doxercalciferol **[3]** or placebo during 4 weeks to patients with localized prostate cancer or high-grade prostatic intraepithelial neoplasia. However, no beneficial effects in serum or tissue markers were seen (Gee et al., 2013). Another study used the same dose in metastatic androgen-independent prostate cancer patients but combined the treatment with docetaxel. Also here, no enhanced PSA response rate or survival rate was seen after 4 weeks of treatment (Attia et al., 2008). Oral alfacalcidol **[8]** or placebo was given to patients with myelodysplastic syndromes. In the patients treated with the analog, a prolongation of leukemic transformation-free survival was seen compared to the placebo group (Motomura et al., 1991). Despite the promising *in vitro* and *in vivo* results of vitamin D analogs in cancer models, clinical trials have failed to proof the effects of vitamin D analogs in cancer patients. Vitamin D and its analogs lack cytotoxic activity, so using these analogs in combination with standard therapies such as radio- and chemotherapy is probably more effective than using the analogs as monotherapy. Next to the combination of analogs with standard cancer therapies, it is also possible that these analogs need to be given for a longer period of time or that treatment with analogs has to be started earlier, for example in early stages of disease or even as chemoprevention.

## **CONCLUSIONS AND PERSPECTIVES**

Vitamin D and its analogs exhibit strong antiproliferative and prodifferentiating effects on different normal and malignant cell types. Several vitamin D analogs have been approved for treating psoriasis, osteoporosis, and secondary hyperparathyroidism and are often used as first or second-line treatment option. Despite promising *in vitro* as well as *in vivo* results in various cancer models, vitamin D analog treatment in clinical trials with cancer patients failed to prove efficacy in most trials. Different combinations of analogs and standard cancer therapies should be further explored as well as the correct duration and timing of administration. To unravel the exact working mechanisms of vitamin D analogs more research studies should compare the effects of vitamin D analogs in different cell types to the mother compound. Furthermore, differences between 1,25(OH)2D3 and its analogs are probably more due to their differences in metabolism and coactivator recruitment than from different genetic regulations. However, some fields such as epigenetics and proteomics remain largely unexplored in comparing the potentially distinctive effects of 1,25(OH)2D3 and its analogs. Since all current genomic and transcriptomic studies focus on established human cell lines, micro-array, and ChIP techniques comparing the effects of 1,25(OH)2D3 and its analogs on human primary tumor tissues should be investigated in the future.

## **ACKNOWLEDGMENTS**

This work was supported by grants from the Fund for Scientific Research (G.0587.09; G.0A17.14N) and the KU Leuven (GOA/14/010).

## **REFERENCES**


*J. Steroid Biochem. Mol. Biol.* 89–90, 443–448. doi: 10.1016/j.jsbmb.2004. 03.039


activities. *Biochem. Pharmacol.* 53, 1087–1097. doi: 10.1016/S0006-2952(96) 00816-7


with the vitamin D3 analog, EB1089. *Endocrinology* 139, 2102–2110. doi: 10.1210/en.139.4.2102


**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: 06 February 2014; paper pending published: 19 February 2014; accepted: 13 March 2014; published online: 03 April 2014.*

*Citation: Leyssens C, Verlinden L and Verstuyf A (2014) The future of vitamin D analogs. Front. Physiol. 5:122. doi: 10.3389/fphys.2014.00122*

*This article was submitted to Integrative Physiology, a section of the journal Frontiers in Physiology.*

*Copyright © 2014 Leyssens, Verlinden and Verstuyf. 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.*