A vitamin D receptor/SMAD genomic circuit gates hepatic fibrotic response.
Journal: 2013/June - Cell
ISSN: 1097-4172
Abstract:
Liver fibrosis is a reversible wound-healing response involving TGFβ1/SMAD activation of hepatic stellate cells (HSCs). It results from excessive deposition of extracellular matrix components and can lead to impairment of liver function. Here, we show that vitamin D receptor (VDR) ligands inhibit HSC activation by TGFβ1 and abrogate liver fibrosis, whereas Vdr knockout mice spontaneously develop hepatic fibrosis. Mechanistically, we show that TGFβ1 signaling causes a redistribution of genome-wide VDR-binding sites (VDR cistrome) in HSCs and facilitates VDR binding at SMAD3 profibrotic target genes via TGFβ1-dependent chromatin remodeling. In the presence of VDR ligands, VDR binding to the coregulated genes reduces SMAD3 occupancy at these sites, inhibiting fibrosis. These results reveal an intersecting VDR/SMAD genomic circuit that regulates hepatic fibrogenesis and define a role for VDR as an endocrine checkpoint to modulate the wound-healing response in liver. Furthermore, the findings suggest VDR ligands as a potential therapy for liver fibrosis.
Relations:
Content
Citations
(148)
References
(67)
Grants
(204)
Conditions
(1)
Drugs
(2)
Chemicals
(3)
Genes
(1)
Organisms
(5)
Processes
(3)
Anatomy
(2)
Affiliates
(1)
Similar articles
Articles by the same authors
Discussion board
Cell 153(3): 601-613

A Vitamin D Receptor/SMAD Genomic Circuit Gates Hepatic Fibrotic Response

+8 authors

INTRODUCTION

Hepatic fibrosis, defined by excessive accumulation of extracellular matrix (ECM) and resultant loss of pliability and liver function, is the result of wound-healing responses triggered by either acute or chronic liver injury (Bataller and Brenner, 2005; Hernandez-Gea and Friedman, 2011; Lee and Friedman, 2011). The main causes of liver injury leading to fibrosis in industrialized countries include chronic hepatitis virus (HBV/HCV) infection, alcohol abuse, and increasingly, nonalcoholic steatohepatitis (NASH) (Friedman, 1999, 2003; Friedman and Bansal, 2006; Siegmund et al., 2005). With persistent injury, there is progressive deposition of fibrillar collagens, eventually leading to parenchymal nodules surrounded by collagen bands, the histological signature of hepatic cirrhosis (Bataller and Brenner, 2005; Friedman, 2003).

Chronic liver disease and cirrhosis represents a major global health concern (Bataller and Brenner, 2005). In Australia and the UK, chronic liver disease is the 5 most common cause of death, after heart disease, cancer, stroke and chest disease (Williams, 2006). In the US, they are ranked as the 8 most common cause of mortality (Kim et al., 2002). Currently, no anti-fibrotic therapies for chronic liver disease have been approved by the FDA (Cohen-Naftaly and Friedman, 2011), and where the underlying cause of the liver disease cannot be ameliorated, therapeutic options are limited to addressing the consequent complications, such as portal hypertension, hepatocellular carcinoma and liver failure. Therefore, a greater understanding of molecular mechanisms regulating the hepatic fibrogenic response in liver is needed for identification of novel targets for successful anti-fibrotic therapies.

The central players in liver fibrosis are non-parenchymal cells (NPCs) such as hepatic stellate cells (HSCs) (Bataller and Brenner, 2005; Bouwens et al., 1992), which are the main producers of ECM (Friedman, 2008; Friedman et al., 1985; Reynaert et al., 2002). In the healthy liver, HSCs are retinoid (Vitamin A) storage cells located in the space of Disse, between the sinusoidal endothelium and hepatocytes (Friedman, 2008). Following injury, paracrine stimuli cause HSCs to undergo dramatic phenotypic changes (in a process called activation), whereby they exhibit proliferation, contractility and loss of retinoid stores, accompanied by secretion of chemokines, cytokines and pathological extracellular matrix components (Friedman, 2008; Geerts, 2001). While the precise mechanisms regulating this process have yet to be elucidated, transforming growth factor β1 (TGFβ1) signaling is recognized as one of the most potent pro-fibrotic pathways responsible for ECM synthesis (Breitkopf et al., 2006; Inagaki and Okazaki, 2007).

TGFβ is a multifunctional cytokine with profound effects on cell division, differentiation, migration, adhesion, organization and death. There are three major isoforms of TGFβ (TGFβ1, TGFβ2 and TGFβ3) and TGFβ1 is the principal isoform implicated in liver fibrosis (Inagaki and Okazaki, 2007). Following liver injury, TGFβ1, derived from both paracrine and autocrine sources, binds to type I and type II serine/threonine receptor kinases on the cell surface of HSCs (Inagaki and Okazaki, 2007). Subsequently, its downstream effectors SMAD2 and SMAD3 are phosphorylated and released into the cytosol, where they form a complex with SMAD4. This SMAD complex can then translocate into the nucleus, recognize SMAD-binding elements (SBE) on the genome and directly regulate target genes (Feng and Derynck, 2005; Massague et al., 2005). Thus, deciphering the TGFβ-SMAD transcriptional network in HSCs and understanding how it can be controlled by extracellular and intracellular factors is key to development of effective anti-fibrotic strategies.

The vitamin D receptor (VDR) is a member of the nuclear hormone receptor (NHR) superfamily and is a key regulator of calcium homeostasis and skeletal health (Bouillon et al., 2008; Goltzman et al., 2004). The endogenous activators of this receptor are the biologically active form of vitamin D (1,25(OH)2D3) and bile acids such as lithocholic acid (LCA) and its derivatives (LCA-acetate, LCA-formate, 3-keto LCA) (Makishima et al., 2002; Nagpal et al., 2005). Interestingly, the closest structural and functional relatives of VDR within the NHR superfamily include farnesoid X receptor (FXR), constitutive androstane receptor (CAR) and pregnane X receptor (PXR), all of which are regulators of bile acid homeostasis and xenobiotic detoxification in the liver (Bookout et al., 2006; Bouillon et al., 2008). However, a physiological role for vitamin D in hepatic function has long been dismissed due to low levels of VDR expression in liver (Bookout et al., 2006; Han et al., 2010). Nonetheless, the finding of robust VDR expression in HSCs led us to consider it as a possible modulator of liver fibrosis (Gascon-Barre et al., 2003).

Here, we demonstrate that liver fibrosis in a standard mouse model of hepatic injury can be ameliorated by administration of the synthetic VDR agonist calcipotriol, which reduces both collagen deposition and fibrotic gene expression. We also show that Vdr knockout mice develop spontaneous liver fibrosis, proving a role for this receptor in normal liver homeostasis. Mechanistic studies revealed that activation of VDR signaling antagonizes a wide range of TGFβ/SMAD-dependent transcriptional responses on pro-fibrotic genes in HSCs. Mapping of genome-wide binding sites of VDR and SMAD3 revealed overlapping DNA occupancy of these transcription factors on cis-regulatory elements of pro-fibrotic genes. Interestingly, TGFβ-SMAD signaling enhanced the accessibility of liganded VDR with these genomic loci, which in turn antagonized recruitment of SMAD3. This dynamic VDR/SMAD genomic feedback circuit represents a previously unrecognized mechanism for regulating hepatic fibrogenesis.

RESULTS

VDR Prevents Liver Fibrosis

Consistent with previous results (Abramovitch et al., 2011; Gascon-Barre et al., 2003), we found that Vdr is expressed in HSCs but is not detectable in either whole liver or purified hepatocytes (Figure S1 A, B&C). Moreover, the HSC-expressed VDR is fully functional as determined by ligand induction of CYP24A1 expression by either 1,25(OH)2D3 or its low calcemic analogue, calcipotriol (Cal) (Nagpal et al., 2005) (Figure S2A), in both primary HSCs and LX-2 cells, a well-established TGFβ1 responsive human HSC cell line (Xu et al., 2005) (Figure S1D&E).

To address whether VDR signaling could suppress fibrotic gene expression and counteract hepatic fibrogenesis in vivo, liver fibrosis was induced by carbon tetrachloride (CCl4), a widely used hepatotoxic agent, at a dose of 0.5ml/kg administered by intraperitoneal (IP) injection 3 times per week in wild type C57BL/6J mice. By four weeks, CCl4-treated mice exhibited extensive liver bridging fibrosis with substantial collagen deposition, whereas CCl4/calcipotriol-co-treated mice had a significant reduction in fibrosis as demonstrated by quantitation of Sirius red staining, hepatic hydroxyproline content and histological fibrotic scoring (Fig 1A–D). The serum calcium concentration was not significantly altered by calcipotriol treatment (Figure S2B). Examination of key fibrotic marker genes such as Col1a1, Tgfb1 and Timp1 revealed between 50–70% down-regulation by calcipotriol (Figure 1E–G). Interestingly, when the mice were pre-treated with calcipotriol for 5 weeks prior to CCl4/calcipotriol-co-treatment, the fibrogenic response in liver was nearly completely abrogated (Figure S2C–F), suggesting that the VDR agonist possesses not only the ability to attenuate fibrosis but also potential to proactively prevent liver fibrosis in vivo.

An external file that holds a picture, illustration, etc.
Object name is nihms474458f1.jpg
Systemic Administration of Calcipotriol Attenuates Liver Fibrosis in CCl4-Treated Mice while Genetic Abrogation of Vdr Results in Spontaneous Liver Fibrosis

(A) Livers from 4 wk-treated C57BL/6J mice (vehicle (DMSO) (n=3), carbon tetrachloride (CCl4, 0.5ml/kg i.p., n=6), calcipotriol (Cal, 20 μg/kg oral gavage, n=3) and CCl4 plus calcipotriol (n=6)) stained with Sirius red (left) and H&amp;E (right). Scale bar, 200μm. Fibrosis quantified by (B) Sirius red staining, (C) hydroxyproline content and (D) H&amp;E staining (Ishak score). Asterisks denote statistically significant differences (Student’s unpaired t-test, **p < 0.01, ***p < 0.001). (E)–(G) RT-qPCR measurement of hepatic gene expression levels of Col1a1, Tgfβ1 and Timp1. Data represents the mean ± SEM. Asterisks denote statistically significant differences (Student’s unpaired t-test, **p < 0.01, ***p < 0.001). (H) Sirius red (top) and H&amp;E (bottom) stained liver sections from Vdr+/+ (n=3), Vdr+/- (n=4) and Vdr−/− (n=2 of 4) mice maintained on a calcium- and phosphate-supplemented rescue diet (2% Calcium, 1.25% Phosphorus, 20% Lactose) for 6 months prior to sacrifice. Arrows indicate peri-sinusoidal fibrosis (Vdr+/− mice) and inflammatory cell infiltrate (Vdr−/− mice), respectively. Scale bar, 50μm. (I) Fibrosis quantified by hydroxyproline content and (J) Col1a1 mRNA expression using the two of four livers from Vdr−/− mice exhibiting the least fibrosis on Sirius red staining (refer to results). Data represents the mean ± SEM. Asterisks denote statistically significant differences (Student’s unpaired t-test, *p < 0.05). See also Fig. S1 &amp; S2.

This led us to examine whether VDR deficiency could impact liver fibrogenesis. Indeed, 6 month-old Vdr−/− mice exhibited a spontaneous liver injury/fibrosis phenotype as demonstrated by increased collagen deposition with two of four mice developing frank cirrhosis (Figure 1H, right/top) associated with hepatocyte necrosis and foci of necroinflammation surrounding portal tracts (Figure 1H, right/bottom, arrow). As there was some variability in the degree of liver fibrosis observed using Sirius red staining of liver sections, liver hydroxyproline content was measured in the two Vdr−/− mice exhibiting the least fibrosis (non-cirrhotic mice) and still found to be significantly greater than that observed in either wild-type or Vdr+/− mice (Figure 1I). Moreover, Vdr+/− mice exhibited multiple foci of peri-sinusoidal fibrosis in the absence of an inflammatory response (Figure 1H, center/top, arrows), pathology not observed in control wild-type mice maintained on an identical calcium- and phosphate-supplemented diet (Figure 1H, left). Histological findings were confirmed by quantitation of hepatic hydroxyproline content as well as examination of key fibrotic marker gene, Col1a1 (Figure 1I&amp;J). These data suggest that both Vdr alleles are required for the maintenance of normal liver architecture and when completely abrogated, result in loss of control of the local inflammatory response in addition to dysregulation of fibrogenesis.

VDR Signaling Suppresses TGFβ-Induced Pro-Fibrotic Genes

Expression profiling was used to explore the potential impact of VDR signaling in TGFβ1 and TGFβ1+1,25(OH)2D3-treated primary rat HSCs. Notably, 1,25(OH)2D3 treatment attenuated the culture-induced activation of HSCs, such that the transcriptome of treated cells closely resembled that of freshly isolated quiescent cells (Figure 2A), and co-treatment of 1,25(OH)2D3 together with TGFβ1 resulted in considerable repression of a large set of TGFβ1 induced genes (Table S1). Among these we noted 39 genes central to hepatic fibrogenesis, including collagens (Bataller and Brenner, 2005; Tsukada et al., 2006), Tgf superfamily members (Inagaki and Okazaki, 2007), matrix metalloproteinase family members (Mmps) (Arthur, 2000; Han, 2006), tissue inhibitors of metalloproteinase (Timps) (Arthur, 2000; Yoshiji et al., 2002), integrins (Patsenker and Stickel, 2011) and lysyl oxidase family members (Barry-Hamilton et al., 2011; Kagan and Li, 2003; Vadasz et al., 2005) (Figure 2B).

An external file that holds a picture, illustration, etc.
Object name is nihms474458f2.jpg
VDR Signaling Suppresses TGFβ-induced Pro-Fibrotic Genes

(A) Heat map comparing 519 differentially expressed genes in freshly isolated rat HSCs (quiescent HSCs, Q-HSCs), activated HSCs (A-HSCs, 3 days culture on plastic) and cells cultures in the presence of 10nM 1,25(OH)2D3 (A-HSCs+1,25(OH)2D3). Euclidean clustering of both rows and columns using log2 transformed microarray expression data, n=2 per treatment group. (B) Heat map of fold change of genes involved in fibrosis in primary rat HSCs treated with TGFβ1 (1ng/ml) and TGFβ1 plus 1,25(OH)2D3 (100nM) for 24 hours, n=2 per treatment group. (C) Fibrotic gene expression in control (siCNTL) or VDR-specific (siVDR) siRNA transfected LX-2 cells treated with Vehicle (DMSO), calcipotriol (Cal, 100nM), TGFβ1 (1ng/ml), or TGFβ1+Cal for 16 hours. Data represents the mean ±SEM of at least three independent experiments performed in triplicate. Asterisks denote statistically significant differences (Student’s unpaired t-test, *p < 0.05, **p < 0.01).

Next, we confirmed that in both primary mouse HSCs and LX-2 cells, calcipotriol potently repressed fibrotic gene expression, suggesting that the anti-TGFβ properties of VDR agonists are likely conserved across mammalian species (data not shown). Finally, using RNAi in LX-2 cells we found that loss of VDR abolished calcipotriol-mediated repression of TGFβ1 induced gene expression (Figure 2C), collectively revealing that VDR regulates an anti-TGFβ/fibrotic network in vitro.

Defining VDR and SMAD3 Cistromes in HSCs

A major question raised by these observations was whether VDR was a direct or indirect regulator of the anti-fibrotic gene network. As SMAD2 and SMAD3 are required for TGFβ1-induced pro-fibrotic gene expression in HSCs (Figure S3A), and VDR activation did not significantly affect TGFβ1-induced phosphorylation and subsequent nuclear translocation of SMAD3 (Figure S3B), we proposed a direct regulatory role for VDR. To explore this possibility, we analysed the genome-wide binding sites of VDR and SMAD3 in LX-2 cells cultured with both calcipotriol and TGFβ1 using chromatin immunoprecipitation coupled with high-throughput deep sequencing (ChIP-Seq). The resulting cistromes identified 24,984 VDR and 23,581 SMAD3 high-confidence binding sites (FDR<0.0001) (Fig 3A&amp;E). Consistent with the reported global binding pattern for other transcription factors (Barish et al., 2010; Biddie et al., 2011; Heinz et al., 2010; Trompouki et al., 2011), the majority of VDR and SMAD3 binding sites localize to distant intergenic and intronic regions, whereas only 16–21% are found at gene promoters (Fig 3A&amp;E). From the list of VDR and SMAD3 binding sites, we confirmed a number of previously characterized functional vitamin D response elements (VDRE) for known vitamin D-inducible genes such as CYP24A1 (Fig 3B), SPP1, BGLAP (Fig S4A&amp;B), and SMAD-binding elements (SBE) for TGFβ signaling target genes including ID1 (Fig 3F), SMAD7 and TGFβI (Figure S4C&amp;D). Gene annotation analysis assigned peaks based on the proximity to the closest transcription start site and yielded 11,031 and 9,210 putative target genes within the individual VDR and SMAD3 cistromes, respectively. Gene ontology (GO) analysis of these annotated genes revealed that the most common classified functions for putative VDR and SMAD3 target genes were metabolism (47%) and cell signaling (34%) (Fig 3C&amp;G).

An external file that holds a picture, illustration, etc.
Object name is nihms474458f3.jpg
VDR and SMAD3 Cistromes in Hepatic Stellate Cells

(A) and (E) Pie charts illustrating genomic locations of VDR and SMAD3 binding sites in treated LX-2 cells (calcipotriol (100nM) and TGFβ1 (1ng/ml) for 4 hours following 16 hours calcipotriol (100nM) pretreatment, FDR<0.0001). Promoter regions, < 2kb from TSS; intergenic regions, not promoter, intron or exon. (B) and (F) Representative ChIP-Seq reads for VDR and SMAD3 aligned to the CYP24A1 and ID1 genes, respectively. (C) and (G) Gene ontology (GO) classification of genes annotated with VDR and SMAD3 binding sites. (D) and (H) De novo motif analysis performed on sequences located within 100bp of VDR and SMAD3 peaks (FDR<0.0001). See also Fig. S3 &amp; S4.

Finally, we interrogated the most significantly enriched binding motifs for VDR and SMAD3. Among these sequence signatures, a direct hexamer repeat with a 3bp spacer (DR3) consensus sequence was the most enriched motif at VDR sites, explaining 74% of VDR binding peaks (Figure 3D, top), while the consensus SBE sequence, a GTCT motif, accounted for 83% of SMAD3 binding peaks (Figure 3H, top). Interestingly, our analysis revealed that the GTCT and DR3-type motifs are also co-enriched within nucleosomal distance at VDR and SMAD3 binding sites, respectively, suggesting VDR and SMAD3 communicate via intersecting cistromes (Figure 3D&amp;H, bottom).

Antagonism of TGFβ Signaling via VDR/SMAD3 Genomic Crosstalk

To address this possibility, we used bioinformatic analysis to quantify the extent of cistrome intersection by calculating the number of sites bound by both VDR and SMAD3. A total of 10,436 genomic sites were co-occupied (Figure 4A), and the co-occupancy pattern is genome-wide as visualized by a heatmap quantifying VDR sites surrounding SMAD3 binding peaks (Figure 4B). If this genomic intersection mediates VDR/SMAD3 crosstalk, VDR and SMAD3 could interact with their co-occupied sites simultaneously. Sequential ChIP (ChIP-re-ChIP) experiments confirmed that VDR and SMAD3 can, at least transiently, co-occupy the same genomic sites (Figure 4C).

An external file that holds a picture, illustration, etc.
Object name is nihms474458f4.jpg
Antagonism of TGFβ Signaling via VDR/SMAD3 Genomic Crosstalk

(A) Venn diagram depicting overlap of VDR and SMAD3 genomic binding sites in LX-2 cells treated as in Figure 3. (B) Intensity plots showing hierarchical clustering of ChIP-fragment densities as a function of distance from the center of statistically significant SMAD3 binding peaks (23,532 peaks, FDR=0.0001). Intensity around position 0 of VDR (blue) indicates overlapping VDR/SMAD3 sites with SMAD3 (red) acting as a positive control. (C) ChIP-re-ChIP of treated LX-2 cells analyzed by qPCR at VDR and SMAD3 co-bound sites. Occupancy is expressed relative to input chromatin. (D) Common human phenotypes enriched in genes co-occupied by VDR and SMAD3. (E) The number of TGFβ1/VDR-coregulated pro-fibrotic genes harboring genomic sites co-occupied by VDR and SMAD3. (F) The number of VDR/SMAD3 co-occupied sites observed in pro-fibrotic genes coregulated by TGFβ1 and VDR. LX-2 cells treated as in Figure 3. Data represents the mean ± SEM of at least three independent experiments performed in triplicate. Asterisks denote statistically significant differences (Student’s t-test, *p < 0.05, **p < 0.01). See also Suppl Table 2 &amp; Fig. S5.

Next, if anti-TGFβ signaling is mediated by a VDR/SMAD genomic intersection, then pro-fibrotic genes in HSCs should be overrepresented in jointly-bound regulatory elements. Indeed, GO analysis designating human phenotypes showed significant enrichment of “abnormal scarring” response (67%) for loci co-occupied by VDR and SMAD3 (Figure 4D) leading us to examine potential VDR/SMAD3 co-occupancy with the earlier identified 39 pro-fibrotic genes (Figure 2B). Within this subset, 34 were found to contain VDR/SMAD3 co-occupied sites (Figure 4E). Furthermore, many of these genes were found to contain multiple VDR/SMAD3 co-occupied sites (Figure 4F &amp; Table S2). Finally, we engineered luciferase reporter plasmids bearing VDR/SMAD3 co-bound sites on the COL1A1 gene and showed that these genomic elements could at least partially recapitulate the opposing actions of calcipotriol and TGFβ1, suggesting that these cis-elements function as enhancers of pro-fibrotic gene expression (Figure S5A).

VDR/SMAD Genomic Antagonism

Informatic analysis of the spatial relationships between VDR and SMAD3 in co-occupied genomic regions confirm that their respective response elements were co-localized within one nucleosomal window (≥200 base pairs) (Figure S5B), further supporting the possibility of genomic antagonism by proximal DNA binding (Barish et al., 2010; Hua et al., 2009).

The presence of VDR/SMAD genomic antagonism can be visualized by plotting the average ChIP-Seq signal intensity of VDR and SMAD3 to the center of their co-occupied sites. This demonstrated that, in the presence of calcipotriol, TGFβ-induced recruitment of SMAD3 was globally compromised by ~1.5 fold, whereas binding of VDR to these sites was globally enhanced by nearly 10 fold (Figure 5A&amp;B). In addition, the proposed genomic antagonism was illustrated by examining its impact along a pro-fibrotic gene harbouring VDR/SMAD co-occupied regulatory elements such as COL1A1. Visualization of sequencing tracks revealed that calcipotriol promoted VDR occupancy at all three major VDR/SMAD3 co-bound sites on the COL1A1 gene (Figure 5C, middle 2 tracks). In contrast, TGFβ-induced SMAD3 binding was typically diminished along the gene upon calcipotriol treatment (Figure 5C, top 2 tracks, and independently validated by ChIP-qPCR, Figure 5D&amp;F). Similar loss of SMAD3 coupled with VDR recruitment was also observed at the regulatory regions of other pro-fibrotic genes such as COL1A2, TGFB1, TGFB2, TIMP1, TIMP2 and LOXL2 (Figure S6A–F). Furthermore, RNAi-mediated depletion of VDR and SMAD2/3 reversed the calcipotriol-dependent loss of SMAD3 recruitment and TGFβ1-induced VDR binding to co-occupied regulatory elements respectively, demonstrating VDR and SMADs are required to mediate this genomic antagonism (Figure 5E&amp;G).

An external file that holds a picture, illustration, etc.
Object name is nihms474458f5.jpg
Genomic Antagonism between VDR and SMAD

(A) and (B) Plots of VDR and SMAD3 ChIP-Seq signal intensity relative to the center of VDR/SMAD3 co-occupied sites in LX-2 cells (TGFβ1 (1ng/ml) ± calcipotriol (100nM) for 4 hours). (C) Representative ChIP-Seq reads aligned to COL1A1 for VDR and SMAD3 in treated LX-2 cells (Vehicle (DMSO), Calcipotriol (Cal, 100nM), TGFβ1 (1ng/ml), or TGFβ1+calcipotriol). The three co-occupied sites are designated as 1, 2 and 3. (D) and (F) ChIP-qPCR at COL1A1 regulatory region #1 co-bound by VDR and SMAD3 in LX-2 cells treated as above. (E) and (G) ChIP-qPCR at COL1A1 regulatory region #1 of control (siCNTL), VDR-specific (siVDR), or SMAD3-specific (siSMAD3) siRNA transfected LX-2 cells treated as above. Occupancy is expressed relative to input chromatin. Data represents the mean ± SEM of at least three independent experiments performed in triplicate. Asterisks denote statistically significant differences (Student’s t-test, *p < 0.05, **p < 0.01). See also Fig. S6 &amp; S7.

Since recruitment of histone-modifying cofactors such as CBP and p300 and hyperacetylation of histone H3 have been established as landmark events of activation of TGFβ signaling (Massague et al., 2005), we asked whether VDR/SMAD genomic antagonism could restrain TGFβ signaling by interfering with this epigenetic pathway. We therefore examined the status of histone H3 acetylation as well as recruitment of CBP and p300 to VDR/SMAD co-occupied sites in cells treated with either calcipotriol or TGFβ1 or both. ChIP-qPCR demonstrated that TGFβ1 induced recruitment of p300 and CBP and histone H3 hyperacetylation at the VDR/SMAD co-occupied regulatory region of COL1A1. This effect was lost in cells co-treated with calcipotriol and TGFβ1 (Figure S7A), suggesting that VDR/SMAD genomic antagonism limits TGFβ activation by compromising coactivator recruitment and histone hyperacetylation.

Ligand-dependent corepressor recruitment or “transrepression” has been proposed as the major mechanism for nuclear receptors such as PPARγ and LXR to negatively regulate inflammatory gene expression (Glass and Saijo, 2010). To test whether transrepression contributes to the antagonism, we examined potential induced recruitment of co-repressors including NCoR, SMRT, HDAC3, CoREST, LSD1, and G9a to VDR/SMAD3 co-occupied regulatory regions of pro-fibrotic genes such as COL1A1 and COL1A2 in response to calcipotriol and TGFβ1. However, altered binding of these corepressors to these sites could not be detected (Figure S7B), suggesting that the loss of transcriptional activation complexes from these sites is not due to increased co- repressor recruitment.

TGFβ Unmasks a Signal Dependent VDR Cistrome

While establishing VDR/SMAD3 genomic antagonism, we noticed that TGFβ/SMAD signaling appears to enhance liganded VDR recruitment to the cis-regulatory regions of COL1A1 (Figure 5F&amp;G). To determine whether this effect is observed at other VDR binding sites of pro-fibrotic genes, we analyzed the VDR cistrome ± calcipotriol in the presence and absence of TGFβ1. Examination of binding data demonstrated that TGFβ1 promotes binding of liganded, but not unliganded VDR to cis-regulatory regions at all pro-fibrotic genes (Figure 5B &amp; S6A–F, lower 4 tracks).

Next, we compared calcipotriol-induced VDR global binding patterns in the presence or absence of TGFβ1. While 6,281 binding sites comprise the de novo VDR cistrome in the absence of TGFβ1, a new cistrome comprised of 24,984 sites was induced in the presence of TGFβ1 (Figure 6A). Interestingly, only 3,537 sites were shared by both cistromes and 85% (21,447 sites) of the TGFβ-induced liganded VDR binding sites were unique (Figure 6A), suggesting that TGFβ results in a dramatic shift of genome-wide binding locations of liganded VDR.

An external file that holds a picture, illustration, etc.
Object name is nihms474458f6.jpg
TGFβ Unmasks a Signal Dependent VDR Citrome

(A) Venn diagram displaying overlapping VDR cistromes in treated LX-2 cells (FDR<0.0001). (B) Plot of VDR ChIP-Seq peak locations depicted in (A) categorized as VDR Cal/TGFβ1+Cal (3,537 overlapping), VDR Cal only (2,744 calcipotriol-only), or VDR TGFβ1+Cal only (21,447 calcipotriol+TGFβ1-only) relative to the center of SMAD3 binding sites in LX-2 cells. (C) Plot of VDR ChIP-Seq signal intensity relative to the center of VDR/SMAD3 co-occupied sites in LX-2 cells treated as indicated. (D) Western blot for VDR in nuclear and whole cell extracts (NE, WCE) from LX-2 cells treated as above. TFIIH (p89) was used as a loading control. (E) The percentages of calcipotriol-only, calcipotriol+TGFβ1-only or calcipotriol/calcipotriol+TGFβ1-overlapping VDR ChIP-Seq peaks containing VDREs. (F) Plot of histone H3 ChIP-Seq signal intensity relative to the center of VDR/SMAD3 co-occupied sites in LX-2 cells treated as indicated.

Comparative studies of the two VDR cistromes revealed that TGFβ1+calcipotriol sites (but not calcipotriol-only sites) were highly enriched at SMAD3 binding sites (Figure 6B). Moreover, binding of VDR to these genomic sites was enhanced by TGFβ signaling (Figure 6C) and this effect was not likely due to a change of VDR expression (Figure 6D). We next analysed the DNA sequences of different subsets of VDR genomic loci and found more than 70% contain de novo VDR regulatory sites (Figure 6E), suggesting that VDR acts directly on the DNA, as opposed to SMAD-dependent tethering. Interestingly, we observed that TGFβ induced significant depletion of nucleosomes at VDR-SMAD3 co-bound sites (Figure 6F), indicating TGFβ-SMAD signaling may promote binding of VDR to its adjacent sites by potentiating local chromatin remodelling and resultant accessibility.

The Genomic Circuit between VDR and SMAD

Our findings suggest a dynamic relationship between VDR and TGFβ-SMAD signaling: perhaps, TGFβ induction of SMAD binding to chromatin creates a new genomic landscape that now becomes accessible to liganded VDR which could enable temporally delayed SMAD repression. To explore this spatio-temporal relationship, we determined the kinetics of SMAD3 and VDR recruitment to co-occupied cis-regulatory elements of fibrotic genes (such as COL1A1) in the presence of either calcipotriol or TGFβ1 or both. Specifically, ChIP-qPCR was employed to monitor binding of VDR and SMAD3 to the cis-regulatory region of COL1A1 at multiple time points (0, 1, 2, 4, 6, 16 hr). Notably, binding of both liganded VDR and SMAD3 to this site were maximally promoted by TGFβ1 after 4 hours of treatment, followed by a gradual decrease to basal levels after 16 hours (Figure 7A&amp;B), confirming the role of TGFβ1 in facilitating recruitment of VDR to chromatin. Interestingly, the binding curve of SMAD3 upon TGFβ1 stimulation was dramatically shifted by the presence of calcipotriol, with the maximum binding of SMAD3 observed just 1 hour post-TGFβ1 treatment. After 4 hours, SMAD3 recruitment was significantly reduced by 70% (Figure 7B). Furthermore, normalization of VDR and SMAD3 binding in the presence of both calcipotriol and TGFβ1 to their basal levels revealed that the occupancy of VDR and SMAD3 were inversely correlated (Figure 7C), suggesting TGFβ-induced chromatin accessibility produces a genomic architecture that facilitates VDR to reverse SMAD activation. Together, this VDR/SMAD genomic circuit provides a chromatin based mechanism for VDR to block fibrosis by antagonizing TGFβ signaling in HSCs.

An external file that holds a picture, illustration, etc.
Object name is nihms474458f7.jpg
VDR/SMAD Genomic Circuit

(A) and (B) Time course of VDR and SMAD3 binding at the COL1A1 regulatory region #1 in treated LX-2 cells (vehicle (DMSO), calcipotriol (100nM), TGFβ1 (1ng/ml), TGFβ1 (1ng/ml) + calcipotriol (100nM)) determined by ChIP-qPCR. LX-2 cells were pretreated with calcipotriol (100nM) for 16 hours prior to time course assay and occupancy is expressed relative to input chromatin. Data represents the mean ± SEM of at least three independent experiments performed in triplicate. Asterisks denote statistically significant differences compared to calcipotriol–induced VDR occupancy or TGFβ1-induced SMAD3 occupancy of corresponding time point (Student’s unpaired t-test, *p < 0.05, **p < 0.01). (C) Time course of TGFβ1+calcipotriol-induced VDR and SMAD3 binding, normalized to calcipotriol alone or TGFβ1 alone, respectively. Data represents the mean ± SEM of at least three independent experiments performed in triplicate. (D) Model depicting proposed VDR/SMAD genomic circuit controlling pro-fibrogenic responses in HSCs.

VDR Prevents Liver Fibrosis

Consistent with previous results (Abramovitch et al., 2011; Gascon-Barre et al., 2003), we found that Vdr is expressed in HSCs but is not detectable in either whole liver or purified hepatocytes (Figure S1 A, B&amp;C). Moreover, the HSC-expressed VDR is fully functional as determined by ligand induction of CYP24A1 expression by either 1,25(OH)2D3 or its low calcemic analogue, calcipotriol (Cal) (Nagpal et al., 2005) (Figure S2A), in both primary HSCs and LX-2 cells, a well-established TGFβ1 responsive human HSC cell line (Xu et al., 2005) (Figure S1D&amp;E).

To address whether VDR signaling could suppress fibrotic gene expression and counteract hepatic fibrogenesis in vivo, liver fibrosis was induced by carbon tetrachloride (CCl4), a widely used hepatotoxic agent, at a dose of 0.5ml/kg administered by intraperitoneal (IP) injection 3 times per week in wild type C57BL/6J mice. By four weeks, CCl4-treated mice exhibited extensive liver bridging fibrosis with substantial collagen deposition, whereas CCl4/calcipotriol-co-treated mice had a significant reduction in fibrosis as demonstrated by quantitation of Sirius red staining, hepatic hydroxyproline content and histological fibrotic scoring (Fig 1A–D). The serum calcium concentration was not significantly altered by calcipotriol treatment (Figure S2B). Examination of key fibrotic marker genes such as Col1a1, Tgfb1 and Timp1 revealed between 50–70% down-regulation by calcipotriol (Figure 1E–G). Interestingly, when the mice were pre-treated with calcipotriol for 5 weeks prior to CCl4/calcipotriol-co-treatment, the fibrogenic response in liver was nearly completely abrogated (Figure S2C–F), suggesting that the VDR agonist possesses not only the ability to attenuate fibrosis but also potential to proactively prevent liver fibrosis in vivo.

An external file that holds a picture, illustration, etc.
Object name is nihms474458f1.jpg
Systemic Administration of Calcipotriol Attenuates Liver Fibrosis in CCl4-Treated Mice while Genetic Abrogation of Vdr Results in Spontaneous Liver Fibrosis

(A) Livers from 4 wk-treated C57BL/6J mice (vehicle (DMSO) (n=3), carbon tetrachloride (CCl4, 0.5ml/kg i.p., n=6), calcipotriol (Cal, 20 μg/kg oral gavage, n=3) and CCl4 plus calcipotriol (n=6)) stained with Sirius red (left) and H&amp;E (right). Scale bar, 200μm. Fibrosis quantified by (B) Sirius red staining, (C) hydroxyproline content and (D) H&amp;E staining (Ishak score). Asterisks denote statistically significant differences (Student’s unpaired t-test, **p < 0.01, ***p < 0.001). (E)–(G) RT-qPCR measurement of hepatic gene expression levels of Col1a1, Tgfβ1 and Timp1. Data represents the mean ± SEM. Asterisks denote statistically significant differences (Student’s unpaired t-test, **p < 0.01, ***p < 0.001). (H) Sirius red (top) and H&amp;E (bottom) stained liver sections from Vdr+/+ (n=3), Vdr+/- (n=4) and Vdr−/− (n=2 of 4) mice maintained on a calcium- and phosphate-supplemented rescue diet (2% Calcium, 1.25% Phosphorus, 20% Lactose) for 6 months prior to sacrifice. Arrows indicate peri-sinusoidal fibrosis (Vdr+/− mice) and inflammatory cell infiltrate (Vdr−/− mice), respectively. Scale bar, 50μm. (I) Fibrosis quantified by hydroxyproline content and (J) Col1a1 mRNA expression using the two of four livers from Vdr−/− mice exhibiting the least fibrosis on Sirius red staining (refer to results). Data represents the mean ± SEM. Asterisks denote statistically significant differences (Student’s unpaired t-test, *p < 0.05). See also Fig. S1 &amp; S2.

This led us to examine whether VDR deficiency could impact liver fibrogenesis. Indeed, 6 month-old Vdr−/− mice exhibited a spontaneous liver injury/fibrosis phenotype as demonstrated by increased collagen deposition with two of four mice developing frank cirrhosis (Figure 1H, right/top) associated with hepatocyte necrosis and foci of necroinflammation surrounding portal tracts (Figure 1H, right/bottom, arrow). As there was some variability in the degree of liver fibrosis observed using Sirius red staining of liver sections, liver hydroxyproline content was measured in the two Vdr−/− mice exhibiting the least fibrosis (non-cirrhotic mice) and still found to be significantly greater than that observed in either wild-type or Vdr+/− mice (Figure 1I). Moreover, Vdr+/− mice exhibited multiple foci of peri-sinusoidal fibrosis in the absence of an inflammatory response (Figure 1H, center/top, arrows), pathology not observed in control wild-type mice maintained on an identical calcium- and phosphate-supplemented diet (Figure 1H, left). Histological findings were confirmed by quantitation of hepatic hydroxyproline content as well as examination of key fibrotic marker gene, Col1a1 (Figure 1I&amp;J). These data suggest that both Vdr alleles are required for the maintenance of normal liver architecture and when completely abrogated, result in loss of control of the local inflammatory response in addition to dysregulation of fibrogenesis.

VDR Signaling Suppresses TGFβ-Induced Pro-Fibrotic Genes

Expression profiling was used to explore the potential impact of VDR signaling in TGFβ1 and TGFβ1+1,25(OH)2D3-treated primary rat HSCs. Notably, 1,25(OH)2D3 treatment attenuated the culture-induced activation of HSCs, such that the transcriptome of treated cells closely resembled that of freshly isolated quiescent cells (Figure 2A), and co-treatment of 1,25(OH)2D3 together with TGFβ1 resulted in considerable repression of a large set of TGFβ1 induced genes (Table S1). Among these we noted 39 genes central to hepatic fibrogenesis, including collagens (Bataller and Brenner, 2005; Tsukada et al., 2006), Tgf superfamily members (Inagaki and Okazaki, 2007), matrix metalloproteinase family members (Mmps) (Arthur, 2000; Han, 2006), tissue inhibitors of metalloproteinase (Timps) (Arthur, 2000; Yoshiji et al., 2002), integrins (Patsenker and Stickel, 2011) and lysyl oxidase family members (Barry-Hamilton et al., 2011; Kagan and Li, 2003; Vadasz et al., 2005) (Figure 2B).

An external file that holds a picture, illustration, etc.
Object name is nihms474458f2.jpg
VDR Signaling Suppresses TGFβ-induced Pro-Fibrotic Genes

(A) Heat map comparing 519 differentially expressed genes in freshly isolated rat HSCs (quiescent HSCs, Q-HSCs), activated HSCs (A-HSCs, 3 days culture on plastic) and cells cultures in the presence of 10nM 1,25(OH)2D3 (A-HSCs+1,25(OH)2D3). Euclidean clustering of both rows and columns using log2 transformed microarray expression data, n=2 per treatment group. (B) Heat map of fold change of genes involved in fibrosis in primary rat HSCs treated with TGFβ1 (1ng/ml) and TGFβ1 plus 1,25(OH)2D3 (100nM) for 24 hours, n=2 per treatment group. (C) Fibrotic gene expression in control (siCNTL) or VDR-specific (siVDR) siRNA transfected LX-2 cells treated with Vehicle (DMSO), calcipotriol (Cal, 100nM), TGFβ1 (1ng/ml), or TGFβ1+Cal for 16 hours. Data represents the mean ±SEM of at least three independent experiments performed in triplicate. Asterisks denote statistically significant differences (Student’s unpaired t-test, *p < 0.05, **p < 0.01).

Next, we confirmed that in both primary mouse HSCs and LX-2 cells, calcipotriol potently repressed fibrotic gene expression, suggesting that the anti-TGFβ properties of VDR agonists are likely conserved across mammalian species (data not shown). Finally, using RNAi in LX-2 cells we found that loss of VDR abolished calcipotriol-mediated repression of TGFβ1 induced gene expression (Figure 2C), collectively revealing that VDR regulates an anti-TGFβ/fibrotic network in vitro.

Defining VDR and SMAD3 Cistromes in HSCs

A major question raised by these observations was whether VDR was a direct or indirect regulator of the anti-fibrotic gene network. As SMAD2 and SMAD3 are required for TGFβ1-induced pro-fibrotic gene expression in HSCs (Figure S3A), and VDR activation did not significantly affect TGFβ1-induced phosphorylation and subsequent nuclear translocation of SMAD3 (Figure S3B), we proposed a direct regulatory role for VDR. To explore this possibility, we analysed the genome-wide binding sites of VDR and SMAD3 in LX-2 cells cultured with both calcipotriol and TGFβ1 using chromatin immunoprecipitation coupled with high-throughput deep sequencing (ChIP-Seq). The resulting cistromes identified 24,984 VDR and 23,581 SMAD3 high-confidence binding sites (FDR<0.0001) (Fig 3A&amp;E). Consistent with the reported global binding pattern for other transcription factors (Barish et al., 2010; Biddie et al., 2011; Heinz et al., 2010; Trompouki et al., 2011), the majority of VDR and SMAD3 binding sites localize to distant intergenic and intronic regions, whereas only 16–21% are found at gene promoters (Fig 3A&amp;E). From the list of VDR and SMAD3 binding sites, we confirmed a number of previously characterized functional vitamin D response elements (VDRE) for known vitamin D-inducible genes such as CYP24A1 (Fig 3B), SPP1, BGLAP (Fig S4A&amp;B), and SMAD-binding elements (SBE) for TGFβ signaling target genes including ID1 (Fig 3F), SMAD7 and TGFβI (Figure S4C&amp;D). Gene annotation analysis assigned peaks based on the proximity to the closest transcription start site and yielded 11,031 and 9,210 putative target genes within the individual VDR and SMAD3 cistromes, respectively. Gene ontology (GO) analysis of these annotated genes revealed that the most common classified functions for putative VDR and SMAD3 target genes were metabolism (47%) and cell signaling (34%) (Fig 3C&amp;G).

An external file that holds a picture, illustration, etc.
Object name is nihms474458f3.jpg
VDR and SMAD3 Cistromes in Hepatic Stellate Cells

(A) and (E) Pie charts illustrating genomic locations of VDR and SMAD3 binding sites in treated LX-2 cells (calcipotriol (100nM) and TGFβ1 (1ng/ml) for 4 hours following 16 hours calcipotriol (100nM) pretreatment, FDR<0.0001). Promoter regions, < 2kb from TSS; intergenic regions, not promoter, intron or exon. (B) and (F) Representative ChIP-Seq reads for VDR and SMAD3 aligned to the CYP24A1 and ID1 genes, respectively. (C) and (G) Gene ontology (GO) classification of genes annotated with VDR and SMAD3 binding sites. (D) and (H) De novo motif analysis performed on sequences located within 100bp of VDR and SMAD3 peaks (FDR<0.0001). See also Fig. S3 &amp; S4.

Finally, we interrogated the most significantly enriched binding motifs for VDR and SMAD3. Among these sequence signatures, a direct hexamer repeat with a 3bp spacer (DR3) consensus sequence was the most enriched motif at VDR sites, explaining 74% of VDR binding peaks (Figure 3D, top), while the consensus SBE sequence, a GTCT motif, accounted for 83% of SMAD3 binding peaks (Figure 3H, top). Interestingly, our analysis revealed that the GTCT and DR3-type motifs are also co-enriched within nucleosomal distance at VDR and SMAD3 binding sites, respectively, suggesting VDR and SMAD3 communicate via intersecting cistromes (Figure 3D&amp;H, bottom).

Antagonism of TGFβ Signaling via VDR/SMAD3 Genomic Crosstalk

To address this possibility, we used bioinformatic analysis to quantify the extent of cistrome intersection by calculating the number of sites bound by both VDR and SMAD3. A total of 10,436 genomic sites were co-occupied (Figure 4A), and the co-occupancy pattern is genome-wide as visualized by a heatmap quantifying VDR sites surrounding SMAD3 binding peaks (Figure 4B). If this genomic intersection mediates VDR/SMAD3 crosstalk, VDR and SMAD3 could interact with their co-occupied sites simultaneously. Sequential ChIP (ChIP-re-ChIP) experiments confirmed that VDR and SMAD3 can, at least transiently, co-occupy the same genomic sites (Figure 4C).

An external file that holds a picture, illustration, etc.
Object name is nihms474458f4.jpg
Antagonism of TGFβ Signaling via VDR/SMAD3 Genomic Crosstalk

(A) Venn diagram depicting overlap of VDR and SMAD3 genomic binding sites in LX-2 cells treated as in Figure 3. (B) Intensity plots showing hierarchical clustering of ChIP-fragment densities as a function of distance from the center of statistically significant SMAD3 binding peaks (23,532 peaks, FDR=0.0001). Intensity around position 0 of VDR (blue) indicates overlapping VDR/SMAD3 sites with SMAD3 (red) acting as a positive control. (C) ChIP-re-ChIP of treated LX-2 cells analyzed by qPCR at VDR and SMAD3 co-bound sites. Occupancy is expressed relative to input chromatin. (D) Common human phenotypes enriched in genes co-occupied by VDR and SMAD3. (E) The number of TGFβ1/VDR-coregulated pro-fibrotic genes harboring genomic sites co-occupied by VDR and SMAD3. (F) The number of VDR/SMAD3 co-occupied sites observed in pro-fibrotic genes coregulated by TGFβ1 and VDR. LX-2 cells treated as in Figure 3. Data represents the mean ± SEM of at least three independent experiments performed in triplicate. Asterisks denote statistically significant differences (Student’s t-test, *p < 0.05, **p < 0.01). See also Suppl Table 2 &amp; Fig. S5.

Next, if anti-TGFβ signaling is mediated by a VDR/SMAD genomic intersection, then pro-fibrotic genes in HSCs should be overrepresented in jointly-bound regulatory elements. Indeed, GO analysis designating human phenotypes showed significant enrichment of “abnormal scarring” response (67%) for loci co-occupied by VDR and SMAD3 (Figure 4D) leading us to examine potential VDR/SMAD3 co-occupancy with the earlier identified 39 pro-fibrotic genes (Figure 2B). Within this subset, 34 were found to contain VDR/SMAD3 co-occupied sites (Figure 4E). Furthermore, many of these genes were found to contain multiple VDR/SMAD3 co-occupied sites (Figure 4F &amp; Table S2). Finally, we engineered luciferase reporter plasmids bearing VDR/SMAD3 co-bound sites on the COL1A1 gene and showed that these genomic elements could at least partially recapitulate the opposing actions of calcipotriol and TGFβ1, suggesting that these cis-elements function as enhancers of pro-fibrotic gene expression (Figure S5A).

VDR/SMAD Genomic Antagonism

Informatic analysis of the spatial relationships between VDR and SMAD3 in co-occupied genomic regions confirm that their respective response elements were co-localized within one nucleosomal window (≥200 base pairs) (Figure S5B), further supporting the possibility of genomic antagonism by proximal DNA binding (Barish et al., 2010; Hua et al., 2009).

The presence of VDR/SMAD genomic antagonism can be visualized by plotting the average ChIP-Seq signal intensity of VDR and SMAD3 to the center of their co-occupied sites. This demonstrated that, in the presence of calcipotriol, TGFβ-induced recruitment of SMAD3 was globally compromised by ~1.5 fold, whereas binding of VDR to these sites was globally enhanced by nearly 10 fold (Figure 5A&amp;B). In addition, the proposed genomic antagonism was illustrated by examining its impact along a pro-fibrotic gene harbouring VDR/SMAD co-occupied regulatory elements such as COL1A1. Visualization of sequencing tracks revealed that calcipotriol promoted VDR occupancy at all three major VDR/SMAD3 co-bound sites on the COL1A1 gene (Figure 5C, middle 2 tracks). In contrast, TGFβ-induced SMAD3 binding was typically diminished along the gene upon calcipotriol treatment (Figure 5C, top 2 tracks, and independently validated by ChIP-qPCR, Figure 5D&amp;F). Similar loss of SMAD3 coupled with VDR recruitment was also observed at the regulatory regions of other pro-fibrotic genes such as COL1A2, TGFB1, TGFB2, TIMP1, TIMP2 and LOXL2 (Figure S6A–F). Furthermore, RNAi-mediated depletion of VDR and SMAD2/3 reversed the calcipotriol-dependent loss of SMAD3 recruitment and TGFβ1-induced VDR binding to co-occupied regulatory elements respectively, demonstrating VDR and SMADs are required to mediate this genomic antagonism (Figure 5E&amp;G).

An external file that holds a picture, illustration, etc.
Object name is nihms474458f5.jpg
Genomic Antagonism between VDR and SMAD

(A) and (B) Plots of VDR and SMAD3 ChIP-Seq signal intensity relative to the center of VDR/SMAD3 co-occupied sites in LX-2 cells (TGFβ1 (1ng/ml) ± calcipotriol (100nM) for 4 hours). (C) Representative ChIP-Seq reads aligned to COL1A1 for VDR and SMAD3 in treated LX-2 cells (Vehicle (DMSO), Calcipotriol (Cal, 100nM), TGFβ1 (1ng/ml), or TGFβ1+calcipotriol). The three co-occupied sites are designated as 1, 2 and 3. (D) and (F) ChIP-qPCR at COL1A1 regulatory region #1 co-bound by VDR and SMAD3 in LX-2 cells treated as above. (E) and (G) ChIP-qPCR at COL1A1 regulatory region #1 of control (siCNTL), VDR-specific (siVDR), or SMAD3-specific (siSMAD3) siRNA transfected LX-2 cells treated as above. Occupancy is expressed relative to input chromatin. Data represents the mean ± SEM of at least three independent experiments performed in triplicate. Asterisks denote statistically significant differences (Student’s t-test, *p < 0.05, **p < 0.01). See also Fig. S6 &amp; S7.

Since recruitment of histone-modifying cofactors such as CBP and p300 and hyperacetylation of histone H3 have been established as landmark events of activation of TGFβ signaling (Massague et al., 2005), we asked whether VDR/SMAD genomic antagonism could restrain TGFβ signaling by interfering with this epigenetic pathway. We therefore examined the status of histone H3 acetylation as well as recruitment of CBP and p300 to VDR/SMAD co-occupied sites in cells treated with either calcipotriol or TGFβ1 or both. ChIP-qPCR demonstrated that TGFβ1 induced recruitment of p300 and CBP and histone H3 hyperacetylation at the VDR/SMAD co-occupied regulatory region of COL1A1. This effect was lost in cells co-treated with calcipotriol and TGFβ1 (Figure S7A), suggesting that VDR/SMAD genomic antagonism limits TGFβ activation by compromising coactivator recruitment and histone hyperacetylation.

Ligand-dependent corepressor recruitment or “transrepression” has been proposed as the major mechanism for nuclear receptors such as PPARγ and LXR to negatively regulate inflammatory gene expression (Glass and Saijo, 2010). To test whether transrepression contributes to the antagonism, we examined potential induced recruitment of co-repressors including NCoR, SMRT, HDAC3, CoREST, LSD1, and G9a to VDR/SMAD3 co-occupied regulatory regions of pro-fibrotic genes such as COL1A1 and COL1A2 in response to calcipotriol and TGFβ1. However, altered binding of these corepressors to these sites could not be detected (Figure S7B), suggesting that the loss of transcriptional activation complexes from these sites is not due to increased co- repressor recruitment.

TGFβ Unmasks a Signal Dependent VDR Cistrome

While establishing VDR/SMAD3 genomic antagonism, we noticed that TGFβ/SMAD signaling appears to enhance liganded VDR recruitment to the cis-regulatory regions of COL1A1 (Figure 5F&amp;G). To determine whether this effect is observed at other VDR binding sites of pro-fibrotic genes, we analyzed the VDR cistrome ± calcipotriol in the presence and absence of TGFβ1. Examination of binding data demonstrated that TGFβ1 promotes binding of liganded, but not unliganded VDR to cis-regulatory regions at all pro-fibrotic genes (Figure 5B &amp; S6A–F, lower 4 tracks).

Next, we compared calcipotriol-induced VDR global binding patterns in the presence or absence of TGFβ1. While 6,281 binding sites comprise the de novo VDR cistrome in the absence of TGFβ1, a new cistrome comprised of 24,984 sites was induced in the presence of TGFβ1 (Figure 6A). Interestingly, only 3,537 sites were shared by both cistromes and 85% (21,447 sites) of the TGFβ-induced liganded VDR binding sites were unique (Figure 6A), suggesting that TGFβ results in a dramatic shift of genome-wide binding locations of liganded VDR.

An external file that holds a picture, illustration, etc.
Object name is nihms474458f6.jpg
TGFβ Unmasks a Signal Dependent VDR Citrome

(A) Venn diagram displaying overlapping VDR cistromes in treated LX-2 cells (FDR<0.0001). (B) Plot of VDR ChIP-Seq peak locations depicted in (A) categorized as VDR Cal/TGFβ1+Cal (3,537 overlapping), VDR Cal only (2,744 calcipotriol-only), or VDR TGFβ1+Cal only (21,447 calcipotriol+TGFβ1-only) relative to the center of SMAD3 binding sites in LX-2 cells. (C) Plot of VDR ChIP-Seq signal intensity relative to the center of VDR/SMAD3 co-occupied sites in LX-2 cells treated as indicated. (D) Western blot for VDR in nuclear and whole cell extracts (NE, WCE) from LX-2 cells treated as above. TFIIH (p89) was used as a loading control. (E) The percentages of calcipotriol-only, calcipotriol+TGFβ1-only or calcipotriol/calcipotriol+TGFβ1-overlapping VDR ChIP-Seq peaks containing VDREs. (F) Plot of histone H3 ChIP-Seq signal intensity relative to the center of VDR/SMAD3 co-occupied sites in LX-2 cells treated as indicated.

Comparative studies of the two VDR cistromes revealed that TGFβ1+calcipotriol sites (but not calcipotriol-only sites) were highly enriched at SMAD3 binding sites (Figure 6B). Moreover, binding of VDR to these genomic sites was enhanced by TGFβ signaling (Figure 6C) and this effect was not likely due to a change of VDR expression (Figure 6D). We next analysed the DNA sequences of different subsets of VDR genomic loci and found more than 70% contain de novo VDR regulatory sites (Figure 6E), suggesting that VDR acts directly on the DNA, as opposed to SMAD-dependent tethering. Interestingly, we observed that TGFβ induced significant depletion of nucleosomes at VDR-SMAD3 co-bound sites (Figure 6F), indicating TGFβ-SMAD signaling may promote binding of VDR to its adjacent sites by potentiating local chromatin remodelling and resultant accessibility.

The Genomic Circuit between VDR and SMAD

Our findings suggest a dynamic relationship between VDR and TGFβ-SMAD signaling: perhaps, TGFβ induction of SMAD binding to chromatin creates a new genomic landscape that now becomes accessible to liganded VDR which could enable temporally delayed SMAD repression. To explore this spatio-temporal relationship, we determined the kinetics of SMAD3 and VDR recruitment to co-occupied cis-regulatory elements of fibrotic genes (such as COL1A1) in the presence of either calcipotriol or TGFβ1 or both. Specifically, ChIP-qPCR was employed to monitor binding of VDR and SMAD3 to the cis-regulatory region of COL1A1 at multiple time points (0, 1, 2, 4, 6, 16 hr). Notably, binding of both liganded VDR and SMAD3 to this site were maximally promoted by TGFβ1 after 4 hours of treatment, followed by a gradual decrease to basal levels after 16 hours (Figure 7A&amp;B), confirming the role of TGFβ1 in facilitating recruitment of VDR to chromatin. Interestingly, the binding curve of SMAD3 upon TGFβ1 stimulation was dramatically shifted by the presence of calcipotriol, with the maximum binding of SMAD3 observed just 1 hour post-TGFβ1 treatment. After 4 hours, SMAD3 recruitment was significantly reduced by 70% (Figure 7B). Furthermore, normalization of VDR and SMAD3 binding in the presence of both calcipotriol and TGFβ1 to their basal levels revealed that the occupancy of VDR and SMAD3 were inversely correlated (Figure 7C), suggesting TGFβ-induced chromatin accessibility produces a genomic architecture that facilitates VDR to reverse SMAD activation. Together, this VDR/SMAD genomic circuit provides a chromatin based mechanism for VDR to block fibrosis by antagonizing TGFβ signaling in HSCs.

An external file that holds a picture, illustration, etc.
Object name is nihms474458f7.jpg
VDR/SMAD Genomic Circuit

(A) and (B) Time course of VDR and SMAD3 binding at the COL1A1 regulatory region #1 in treated LX-2 cells (vehicle (DMSO), calcipotriol (100nM), TGFβ1 (1ng/ml), TGFβ1 (1ng/ml) + calcipotriol (100nM)) determined by ChIP-qPCR. LX-2 cells were pretreated with calcipotriol (100nM) for 16 hours prior to time course assay and occupancy is expressed relative to input chromatin. Data represents the mean ± SEM of at least three independent experiments performed in triplicate. Asterisks denote statistically significant differences compared to calcipotriol–induced VDR occupancy or TGFβ1-induced SMAD3 occupancy of corresponding time point (Student’s unpaired t-test, *p < 0.05, **p < 0.01). (C) Time course of TGFβ1+calcipotriol-induced VDR and SMAD3 binding, normalized to calcipotriol alone or TGFβ1 alone, respectively. Data represents the mean ± SEM of at least three independent experiments performed in triplicate. (D) Model depicting proposed VDR/SMAD genomic circuit controlling pro-fibrogenic responses in HSCs.

DISCUSSION

The establishment of HSCs as the primary effector cell for the deposition of ECM in normal and fibrotic liver in the early 1990s was a milestone discovery in understanding the pathogenesis of hepatic fibrosis (Friedman, 1993). Since then, a wide spectrum of cellular signaling molecules, hormones, cell membrane receptors and transcription factors in HSCs have been investigated and found to promote hepatic fibrogenesis (Hernandez-Gea and Friedman, 2011). However, the factors and signaling cascades that actively prevent this pathological process are poorly understood. Here, we demonstrate that pharmacological activation of VDR attenuates the progression of liver fibrosis in an experimental animal model while genetic abrogation of VDR expression results in the spontaneous development of liver fibrosis, thus implicating VDR in an endocrine checkpoint that negatively modulates the wound healing response in liver. Mechanistically we delineate a previously unrecognized and temporally controlled genomic circuit composed of the opposing action of VDR and SMAD transcription factors that is able to restrain the intensity of the fibrogenic response in HSCs and govern fibrogenesis in liver. Specifically, in response to liver injury, HSC activation by TGFβ1 induces pro-fibrotic gene expression via SMAD translocation to the nucleus and chromatin remodeling. By increasing accessibility to adjacent vitamin D response elements (VDREs), SMAD activation facilitates VDR recruitment to previously cryptic genomic sites. Liganded VDR subsequently antagonizes SMAD residency on chromatin and compromises acetylation of histone H3 to ultimately suppress pro-fibrotic gene expression (Figure 7D). Notably, the proximal location of nearly 10,500 TGFβ1-induced SMAD and VDR binding sites identifies a global chromatin architecture and suggests that the integrated VDR/SMAD genomic circuit functions as a master regulator of the hepatic fibrotic response.

The identification of a chromatin basis for inhibiting TGFβ signaling places a direct focus on SMAD-dependent transcription as a regulatory target. This is relevant as TGFβ-SMAD signaling plays an essential role in almost every aspect of metazoan biology and its dysregulation can result in a diversity of human diseases ranging from autoimmunity to fibrosis and cancer (Hernandez-Gea and Friedman, 2011; Li and Flavell, 2008; Massague, 2008). Our finding of genomic antagonism between VDR and SMAD not only establishes VDR as the first DNA-binding transcription factor that attenuates TGFβ-SMAD signaling at a chromatin interface but also adds specificity (a cistromic layer) for the more general concept of ‘transcriptional crosstalk’.

The observation that TGFβ-SMAD activation enables subsequent recruitment of ligand-bound VDR to repress SMAD targets reveals a means by which two endogenous signaling pathways can cross-regulate each other’s activity. Thus, this genomic relay allows positive activation by SMAD to be subsequently inhibited by VDR and thus constitutes a self-adjusting genomic circuit, which is highly distinguishable from the previously reported genomic crosstalk between transcription factors in a mutually exclusive manner (Barish et al., 2010; Hua et al., 2009). It seems logical that this circuit may confer on HSCs the ability to orchestrate ECM synthesis in both the normal and fibrotic liver.

In addition to the TGFβ-SMAD pathway, fibrosis is almost always preceded by persistent inflammation clinically (Hernandez-Gea and Friedman, 2011; Lee and Friedman, 2011). Hence, a broader anti-inflammatory role for VDR signaling might conceivably contribute to its anti-fibrotic property in liver. In this regard, VDR has been documented for its expression in several cell types central to the inflammatory response (Barish et al., 2005; Griffin et al., 2001; von Essen et al., 2010), and both vitamin D deficiency and polymorphisms of VDR itself as well as genes involved in vitamin D metabolism have been linked to both risk and severity of inflammatory diseases (Agmon-Levin et al., 2012; Janssens et al., 2011; Munger et al., 2006; Ramagopalan et al., 2011). However, the role of VDR signaling’s anti-inflammatory action in the context of hepatic fibrogenesis is far less clear. On one hand, the dysregulated inflammatory response coupled with the spontaneous development of liver fibrosis in Vdr−/− mice suggests that VDR signaling might control hepatic fibrogeneis through an anti-inflammatory mechanism (Figure 1H, right). On the other hand, this notion is blunted by the modest peri-sinusoidal liver fibrosis phenotype without any inflammatory response found in Vdr+/− mice (Figure 1H, center). Furthermore, the causable relationship between inflammation and fibrosis remains to be fully established and the major pro-fibrogenic role of inflammation during hepatic fibrogenesis appears to be to sensitize HSCs for TGFβ-SMAD activation (Seki et al., 2007; Seki and Schnabl, 2012). It is therefore unlikely that the anti-inflammatory property of VDR signaling plays a major role in its anti-fibrotic function.

Our studies further serve to clarify an unappreciated function of VDR signaling in liver pathophysiology. Due to its exceptionally low expression, VDR has received much less attention than its highly expressed cognate clade members that include FXR, PXR and CAR that impact nearly every aspect of hepatic function including lipid and glucose metabolism, drug disposition, cholesterol efflux and bile acid homeostasis (Bookout et al., 2006; Chawla et al., 2001). However, recent studies showing that low vitamin D levels are linked to increased hepatic fibrosis in patients with chronic liver disease (Abramovitch et al., 2011; Lim and Chalasani, 2012; Petta et al., 2010; Terrier et al., 2011) and that vitamin D can inhibit liver fibrosis in rats (Abramovitch et al., 2011) suggest a potential physiologic role for hepatic VDR. However, whether and how VDR directly or indirectly regulates hepatic fibrogenesis remained unresolved. Our findings that VDR promotes HSC quiescence and controls TGFβ signaling identify a new mechanism through which vitamin D can exert its anti-fibrotic effects. It is noteworthy that our delineation of a VDR-signaling pathway to inhibit fibrosis is also consistent with recent studies suggesting that a polymorphism in VDR is correlated with increased progression of liver fibrosis and evolution of cirrhosis (Baur et al., 2011; Tanaka et al., 2009).

Up to 45% of deaths in the developed world can be attributed to fibrotic diseases, yet few anti-fibrotic drugs are currently approved for clinical use (Wynn, 2008). Though therapies designed to neutralize TGFβ show broad anti-fibrotic activity (Rosenbloom et al., 2010) the benefits are compromised by unnecessarily blocking TGFβ in non-diseased tissue. Our discovery of the VDR/SMAD genomic circuit illuminates a potentially safer anti-fibrotic strategy by restricting TGFβ inhibition to VDR-positive cells instead of perturbing signaling body-wide.

In summary, our work describes an intersecting genomic circuit comprising VDR and SMAD transcription factors that governs hepatic fibrogenesis. This finding significantly extends our understanding of how two distinct signal-dependent transcription factors interact with each other to establish cell identity and function. Through the use of genetic and inducible models we provide new insight into how global programs responding to TGFβ1 signaling are established and regulated. Furthermore, these studies establish VDR as a potential drug target to treat liver fibrosis and provide a new paradigm of VDR-dependent gene expression regulation. Given the ubiquitous expression patterns of VDR and TGFβ, the VDR/SMAD genomic circuit is likely to be applicable to many other cell types and may impact the pathogenesis of a wide range of human diseases.

EXPERIMENTAL PROCEDURES

Primary HSCs Isolation and Culture

HSCs were isolated from 10-week old male C57BL/6J mice and Wistar rats by in situ pronase, collagenase perfusion and single-step Histogenz gradient as previously reported (Hendriks et al., 1985; Knook et al., 1982). Isolated HSCs were cultured in DMEM (Mediatech) containing 20% FBS (Hyclone) on 6-well plates for 40 hours prior to end-point assays.

Immunoprecipitation and Western Blot

The whole cell lysates were obtained through RIPA buffer lysis while isolation of nuclear extract was performed as previously reported (Ding et al., 2008). Total SMAD3 was immmunoprecipitated in nuclear extracts from LX-2 cells using anti-SMAD2/3 antibody (Santa Cruz, sc-133098) followed by SDS-PAGE and western blot detection by anti-SMAD3 (Cell Signaling, 9523) and anti-pSMAD3 (Cell Signaling, 9520) specific antibodies.

Cell Culture, Luciferase Assay and RT-qPCR

LX-2 cells, a generous gift from Professor Scott Friedman, Mount Sinai School of Medicine, New York, NY, were cultured as described previously (Xu et al., 2005). TGFβ1 (R&amp;D Systems), 1,25(OH)2D3 and calcipotriol (Tocris) were used at concentrations of 1ng/ml, 100nM and 100nM, respectively, except when otherwise indicated. For luciferase assays, DNA transfections were performed using Fugene 6 (Roche) following the manufacturer’s instructions. 24 hours following DNA transfections, cells were treated with vehicle, calcipotriol or TGFβ1 or both for another 24 hours prior to Luciferase/β-galactosidase assays (Promega). For RT-qPCR, total RNA was purified following TRIzol extraction and treated with DNaseI (Invitrogen). cDNA synthesis was carried out with iScript RT Supermix (Biorad). Quantitative PCR was performed in technical triplicates using SYBR Green reagent (Biorad). The relative standard curve method was used for quantitation (Biorad). Expression levels were calculated by normalization to either Gapdh (mouse) or U36B4 (human) quantities. The sequences of primers are listed in Table S3.

Transfection of siRNAs

Transfection was carried out at a concentration of 20nM of indicated siRNAs (in the case of SMAD2/3, 10nM of each siRNA was combined for transfection) using RNAiMax transfection reagent (Invitrogen). Transfected cells were cultured without perturbation for at least 48 hours prior to terminal assays.

CCl4 Model of Liver Injury and Fibrosis

8 week-old male C57BL/6J mice were IP injected with 0.5ml/kg body weight CCl4 (1:50 v/v in corn oil from Sigma) or vehicle (DMSO in corn oil) three times a week for 4 weeks. Calcipotriol (20μg/kg body weight) was administered by oral gavage 5 times a week, commencing 20 days after the first dose of CCl4. The animals were terminated 72 hours after the final CCl4 injection and whole livers and serum were collected for histological, cytological, biochemical and molecular analyses.

Vdr Knockout Mice

C57BL/6J mice heterozygous for targeted ablation of Vdr (Li et al., 1997) were obtained from The Jackson Laboratory (Stock Number 006133). Wild type controls, Vdr and Vdr mice were maintained on a Vdr−/− rescue diet (Amling et al., 1999) containing 21% calcium, and 0.67% phosphorus and 20% lactose supplemented with 4.4 units of vitamin D per gram diet for 6 months prior to sacrifice. Livers were collected for analysis as above.

Fibrotic Score and Quantification Hepatic Collagen and Hydroxyproline Content

5μm sections of formalin-fixed liver were stained following standard H&amp;E and Sirius Red methods and reviewed by a pathologist who was blinded to the experimental conditions. Fibrosis was scored using the Ishak modified histological activity index (HAI) scoring system. Fibrosis was also quantified using Image J software on 10 non-contiguous Sirius Red stained sections. All images were obtained using a high-resolution Leica DFC420 digital camera mounted on an Olympus microscope equipped with ×4/0.13, ×10/0.30, ×20/0.50 and x40/0.75 UplanFL N plan objective lenses and processed with the Leica Application Suite. Hepatic hydroxyproline content was measured using a commercial colorimetric assay from Biovision (K555–100).

ChIP and ChIP-Re-ChIP

LX-2 cells were pretreated with calcipotriol (100nM) for 16 hours followed by incubation of calcipotriol (100nM) or TGFβ1 (1ng/ml) or both for an additional 4 hours. Cells were then harvested for ChIP assay. The experimental procedure for ChIP was as previously described (Barish et al., 2010). Briefly, after fixation, nuclei from LX-2 cells were isolated, lysed and sheared with a Diagenode Bioruptor to yield DNA fragment sizes of 200–1000 base pairs followed by immunoprecipitation using antibodies listed below: normal rabbit IgG (Santa Cruz, sc-2027), VDR (Santa Cruz, sc-1008), SMAD3 (Abcam, ab28379) and histone H3 (Abcam, ab1791). For ChIP-Re-ChIP, after first ChIP, the immunprecipitated DNA-protein complex was eluted from beads using 10mM DTT and diluted 100-fold followed by second ChIP.

ChIP-seq Data Analysis

The procedure was as previously described (Barish et al., 2010). Briefly, short DNA reads were aligned against the human hg18 reference genome (NCBI Build 36.1) using the Illumina Pipeline Suite v1.7. Reads were aligned using the Bowtie aligner allowing up to 2 mismatches in the read. Only tags that map uniquely to the genome were considered for further analysis. Subsequent peak calling and motif analysis were conducted using HOMER, a software suite for ChIP-Seq analysis. The methods for HOMER, which are described below, have been implemented and are freely available at http://biowhat.ucsd.edu/homer/ (Heinz et al., 2010). One tag from each unique position was considered to eliminate peaks resulting from clonal amplification of fragments during the ChIPSeq protocol. Peaks were identified by searching for clusters of tags within a sliding 200 bp window, requiring adjacent clusters to be at least 1 kb away from each other. The threshold for the number of tags that determine a valid peak was selected for a false discovery rate of <0.0001, as empirically determined by repeating the peak finding procedure using randomized tag positions. Peaks are required to have at least 4-fold more tags (normalized to total count) than input or IgG control samples and 4-fold more tags relative to the local background region (10 kb) to avoid identifying regions with genomic duplications or non-localized binding. Peaks are annotated to gene products by identifying the nearest RefSeq transcriptional start site. Visualization of ChIP-Seq results was achieved by uploading custom tracks onto the UCSC genome browser. Human phenotype analysis was performed using GREAT (Genomic Regions Enrichment of Annotations Tool) at http://great.stanford.edu/.

Microarray Data Analysis

Total RNA from primary rat or mouse HSCs was isolated using the RNeasy mini kit (Qiagen) according to standard protocols. RNA integrity and quality was assessed using the Agilent Bioanalyzer and prepared for hybridization to Illumina rat or mouse gene expression arrays according to standard Illumina protocols. Feature extraction was performed using the Illumina GenomeStudio software. Normalization and identification of differentially expressed genes from biological duplicates was performed using VAMPIRE at http://sasquatch.ucsd.edu/vampire/.

Primary HSCs Isolation and Culture

HSCs were isolated from 10-week old male C57BL/6J mice and Wistar rats by in situ pronase, collagenase perfusion and single-step Histogenz gradient as previously reported (Hendriks et al., 1985; Knook et al., 1982). Isolated HSCs were cultured in DMEM (Mediatech) containing 20% FBS (Hyclone) on 6-well plates for 40 hours prior to end-point assays.

Immunoprecipitation and Western Blot

The whole cell lysates were obtained through RIPA buffer lysis while isolation of nuclear extract was performed as previously reported (Ding et al., 2008). Total SMAD3 was immmunoprecipitated in nuclear extracts from LX-2 cells using anti-SMAD2/3 antibody (Santa Cruz, sc-133098) followed by SDS-PAGE and western blot detection by anti-SMAD3 (Cell Signaling, 9523) and anti-pSMAD3 (Cell Signaling, 9520) specific antibodies.

Cell Culture, Luciferase Assay and RT-qPCR

LX-2 cells, a generous gift from Professor Scott Friedman, Mount Sinai School of Medicine, New York, NY, were cultured as described previously (Xu et al., 2005). TGFβ1 (R&amp;D Systems), 1,25(OH)2D3 and calcipotriol (Tocris) were used at concentrations of 1ng/ml, 100nM and 100nM, respectively, except when otherwise indicated. For luciferase assays, DNA transfections were performed using Fugene 6 (Roche) following the manufacturer’s instructions. 24 hours following DNA transfections, cells were treated with vehicle, calcipotriol or TGFβ1 or both for another 24 hours prior to Luciferase/β-galactosidase assays (Promega). For RT-qPCR, total RNA was purified following TRIzol extraction and treated with DNaseI (Invitrogen). cDNA synthesis was carried out with iScript RT Supermix (Biorad). Quantitative PCR was performed in technical triplicates using SYBR Green reagent (Biorad). The relative standard curve method was used for quantitation (Biorad). Expression levels were calculated by normalization to either Gapdh (mouse) or U36B4 (human) quantities. The sequences of primers are listed in Table S3.

Transfection of siRNAs

Transfection was carried out at a concentration of 20nM of indicated siRNAs (in the case of SMAD2/3, 10nM of each siRNA was combined for transfection) using RNAiMax transfection reagent (Invitrogen). Transfected cells were cultured without perturbation for at least 48 hours prior to terminal assays.

CCl4 Model of Liver Injury and Fibrosis

8 week-old male C57BL/6J mice were IP injected with 0.5ml/kg body weight CCl4 (1:50 v/v in corn oil from Sigma) or vehicle (DMSO in corn oil) three times a week for 4 weeks. Calcipotriol (20μg/kg body weight) was administered by oral gavage 5 times a week, commencing 20 days after the first dose of CCl4. The animals were terminated 72 hours after the final CCl4 injection and whole livers and serum were collected for histological, cytological, biochemical and molecular analyses.

Vdr Knockout Mice

C57BL/6J mice heterozygous for targeted ablation of Vdr (Li et al., 1997) were obtained from The Jackson Laboratory (Stock Number 006133). Wild type controls, Vdr and Vdr mice were maintained on a Vdr−/− rescue diet (Amling et al., 1999) containing 21% calcium, and 0.67% phosphorus and 20% lactose supplemented with 4.4 units of vitamin D per gram diet for 6 months prior to sacrifice. Livers were collected for analysis as above.

Fibrotic Score and Quantification Hepatic Collagen and Hydroxyproline Content

5μm sections of formalin-fixed liver were stained following standard H&amp;E and Sirius Red methods and reviewed by a pathologist who was blinded to the experimental conditions. Fibrosis was scored using the Ishak modified histological activity index (HAI) scoring system. Fibrosis was also quantified using Image J software on 10 non-contiguous Sirius Red stained sections. All images were obtained using a high-resolution Leica DFC420 digital camera mounted on an Olympus microscope equipped with ×4/0.13, ×10/0.30, ×20/0.50 and x40/0.75 UplanFL N plan objective lenses and processed with the Leica Application Suite. Hepatic hydroxyproline content was measured using a commercial colorimetric assay from Biovision (K555–100).

ChIP and ChIP-Re-ChIP

LX-2 cells were pretreated with calcipotriol (100nM) for 16 hours followed by incubation of calcipotriol (100nM) or TGFβ1 (1ng/ml) or both for an additional 4 hours. Cells were then harvested for ChIP assay. The experimental procedure for ChIP was as previously described (Barish et al., 2010). Briefly, after fixation, nuclei from LX-2 cells were isolated, lysed and sheared with a Diagenode Bioruptor to yield DNA fragment sizes of 200–1000 base pairs followed by immunoprecipitation using antibodies listed below: normal rabbit IgG (Santa Cruz, sc-2027), VDR (Santa Cruz, sc-1008), SMAD3 (Abcam, ab28379) and histone H3 (Abcam, ab1791). For ChIP-Re-ChIP, after first ChIP, the immunprecipitated DNA-protein complex was eluted from beads using 10mM DTT and diluted 100-fold followed by second ChIP.

ChIP-seq Data Analysis

The procedure was as previously described (Barish et al., 2010). Briefly, short DNA reads were aligned against the human hg18 reference genome (NCBI Build 36.1) using the Illumina Pipeline Suite v1.7. Reads were aligned using the Bowtie aligner allowing up to 2 mismatches in the read. Only tags that map uniquely to the genome were considered for further analysis. Subsequent peak calling and motif analysis were conducted using HOMER, a software suite for ChIP-Seq analysis. The methods for HOMER, which are described below, have been implemented and are freely available at http://biowhat.ucsd.edu/homer/ (Heinz et al., 2010). One tag from each unique position was considered to eliminate peaks resulting from clonal amplification of fragments during the ChIPSeq protocol. Peaks were identified by searching for clusters of tags within a sliding 200 bp window, requiring adjacent clusters to be at least 1 kb away from each other. The threshold for the number of tags that determine a valid peak was selected for a false discovery rate of <0.0001, as empirically determined by repeating the peak finding procedure using randomized tag positions. Peaks are required to have at least 4-fold more tags (normalized to total count) than input or IgG control samples and 4-fold more tags relative to the local background region (10 kb) to avoid identifying regions with genomic duplications or non-localized binding. Peaks are annotated to gene products by identifying the nearest RefSeq transcriptional start site. Visualization of ChIP-Seq results was achieved by uploading custom tracks onto the UCSC genome browser. Human phenotype analysis was performed using GREAT (Genomic Regions Enrichment of Annotations Tool) at http://great.stanford.edu/.

Microarray Data Analysis

Total RNA from primary rat or mouse HSCs was isolated using the RNeasy mini kit (Qiagen) according to standard protocols. RNA integrity and quality was assessed using the Agilent Bioanalyzer and prepared for hybridization to Illumina rat or mouse gene expression arrays according to standard Illumina protocols. Feature extraction was performed using the Illumina GenomeStudio software. Normalization and identification of differentially expressed genes from biological duplicates was performed using VAMPIRE at http://sasquatch.ucsd.edu/vampire/.

Supplementary Material

01

02

03

04

01

Click here to view.(2.5M, xlsx)

02

Click here to view.(15K, xlsx)

03

Click here to view.(11K, xlsx)

04

Click here to view.(44M, pdf)

Acknowledgments

We thank C. Brondos and E. Ong for administrative assistance, J. Nery for assistance with DNA sequencing, C. Benner for assistance with HOMER software, and H. Juguilon and J. Alvarez for technical assistance. N.D. was supported by a postdoctoral fellowship from Genentech Foundation, and G.D.B. by K08HL092298. This work was funded by grants from NIH ({"type":"entrez-nucleotide","attrs":{"text":"DK057978","term_id":"187700204","term_text":"DK057978"}}DK057978, {"type":"entrez-nucleotide","attrs":{"text":"HL105278","term_id":"1051677264","term_text":"HL105278"}}HL105278, {"type":"entrez-nucleotide","attrs":{"text":"DK090962","term_id":"187689227","term_text":"DK090962"}}DK090962, {"type":"entrez-nucleotide","attrs":{"text":"HL088093","term_id":"1051658502","term_text":"HL088093"}}HL088093, {"type":"entrez-nucleotide","attrs":{"text":"ES010337","term_id":"164124468","term_text":"ES010337"}}ES010337 and {"type":"entrez-nucleotide","attrs":{"text":"CA014195","term_id":"24291539","term_text":"CA014195"}}CA014195) and National Health and Medical Research Council of Australia Project Grants 512354 and 632886 (C.L. and M.D.), as well as the Helmsley Charitable Trust, Samuel Waxman Cancer Research Foundation and Ipsen/Biomeasure. R.M.E. and M.D. are supported in part by a Stand Up to Cancer Dream Team Translational Cancer Research Grant, a Program of the Entertainment Industry Foundation (SU2C-AACR-DT0509). R.M.E is an investigator of the Howard Hughes Medical Institute and March of Dimes Chair in Molecular and Developmental Biology at the Salk Institute.

Gene Expression Laboratory, Salk Institute for Biological Studies, La Jolla, California, USA
Howard Hughes Medical Institute, Salk Institute for Biological Studies, La Jolla, California, USA
The Storr Liver Unit, Westmead Millennium Institute and University of Sydney, Westmead Hospital, Westmead, New South Wales 2145, Australia
Diabetes and Transcription Factors Group, Garvan Institute of Medical Research, Darlinghurst, Sydney, New South Wales 2010, Australia
Correspondence: ude.klas@snave and ude.klas@senwod
Publisher's Disclaimer

SUMMARY

Liver fibrosis is a reversible wound-healing response involving TGFβ1 activation of hepatic stellate cells (HSCs). Here we show that vitamin D receptor (VDR) ligands inhibit HSC activation and abrogate liver fibrosis, while Vdr knockout mice spontaneously developed hepatic fibrosis. Mechanistically, we describe a pronounced redistribution of genome wide VDR binding sites (VDR cistrome) in HSCs elicited by a TGFβ1 pro-fibrotic insult. This TGFβ1-induced VDR cistrome overlaps extensively with SMAD3 binding sites, with co-occupancy at numerous cis-regulatory elements identified on a large set of pro-fibrotic genes. Addition of VDR ligand reduces SMAD3 occupancy at co-regulated genes, revealing an intersecting VDR/SMAD genomic circuit that regulates hepatic fibrogenesis. These results define a role for VDR as a endocrine checkpoint to modulate the wound healing response in liver, and suggest VDR ligands as a potential therapy for liver fibrosis.

SUMMARY
HIGHLIGHTS

Footnotes

Accession Numbers

The GEO accession number for the full data sets (ChIP-seq and microarray) is {"type":"entrez-geo","attrs":{"text":"GSE38103","term_id":"38103"}}GSE38103.

Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

Footnotes
Collaboration tool especially designed for Life Science professionals.Drag-and-drop any entity to your messages.