Hypoxia-Inducible Transcription Factor 2α Promotes Steatohepatitis Through Augmenting Lipid Accumulation, Inflammation and Fibrosis

Abstract

Oxygen dynamics in the liver is a central signaling mediator controlling hepatic homeostasis, and dysregulation of cellular oxygen is associated with liver injury. Moreover, the transcription factor relaying changes in cellular oxygen levels, hypoxia-inducible factor (HIF), is critical in liver metabolism and sustained increase in HIF signaling can lead to spontaneous steatosis, inflammation, and liver tumorigenesis. However, the direct responses and genetic networks regulated by HIFs in the liver are unclear. To help define the HIF signal transduction pathway, an animal model of HIF overexpression was generated and characterized. In this model, overexpression was achieved by Von Hippel-Lindau (Vhl) disruption in a liver-specific temporal fashion. Acute disruption of Vhl induced hepatic lipid accumulation in a HIF-2α-dependent manner. In addition, HIF-2α activation rapidly increased liver inflammation and fibrosis demonstrating that steatosis and inflammation are primary responses of the liver to hypoxia. To identify downstream effectors, a global microarray expression analysis was performed using livers lacking Vhl for 24-hours and 2-weeks, revealing a time-dependent effect of HIF on gene expression. Increase in genes involved in fatty acid synthesis were followed by an increase in fatty acid uptake-associated genes, and an inhibition of fatty acid β-oxidation. A rapid increase in pro-inflammatory cytokines and fibrogenic gene expression was also observed. In vivo chromatin immunoprecipitation assays revealed novel direct targets of HIF signaling that may contribute to hypoxia-mediated steatosis and inflammation. These data suggest that HIF-2α is a critical mediator in progression from clinically manageable steatosis to more severe steatohepatitis and liver cancer, and may be a potential therapeutic target.

Introduction

Oxygen is a critical signaling molecule that regulates the metabolic activities of the liver(1, 2). Dysregulation of the normal oxygen gradient in the liver can induce liver steatosis and inflammation(2). Decreased cellular oxygen impacts gene expression through the transcription factor hypoxia-inducible factor (HIF). During normal cellular oxygen levels HIFα subunits are rapidly degraded by the ubiquitin proteasome system in which Von Hippel-Lindau (VHL) tumor suppressor protein is the critical E3 ubiquitin ligase required for HIF degradation(3–8). HIF-1α and HIF-2α regulate expression of genes that are critical for adaptation to low oxygen levels. Targeted disruption of Vhl in the liver increased HIF-1α and HIF-2α expression and this mouse model has demonstrated that HIFs are critical in erythropoiesis, iron metabolism, hepatic lipid homeostasis, glucose metabolism, and tumor formation in the liver(9–14). Since overexpression of HIF through disruption of Vhl has many robust pleiotropic effects, it is difficult to assess which are the direct responses of the liver following hypoxia. Furthermore, finding direct mediators of HIF signaling in the liver, which contribute to the phenotype has been difficult. To overcome this problem the present study describes a liver-specific temporal disruption of Vhl using a cre-ER^T2 system, which activates a liver-specific cre recombinase expression in the presence of the estrogen analog tamoxifen. Acute disruption of Vhl resulted in a robust accumulation of lipids in the liver and an increase in liver inflammation and fibrosis. Using a compound double deletion of Vhl and Hif-1α or Hif-2α, liver steatosis, inflammation, and fibrosis were mediated in a HIF-2α-dependent manner. To assess direct signaling pathways activated by HIF, global gene expression analysis was performed in the livers of mice with a temporal disruption of Vhl for 24-hours or 2-weeks. Gene expression profiles demonstrated that HIF rapidly regulates a large battery of genes important for fatty acid synthesis, uptake, and β-oxidation. Moreover, several pro-inflammatory mediators and pro-fibrogenic genes were rapidly activated following Vhl deletion. These data demonstrate that the liver injury due to hypoxia is a primary response mediated by HIF-2α.

Experimental Procedures

Luciferase assay

The mouse angiopoietin-like 3 (Angptl3)-promoter luciferase was previously described(15). Mouse transglutaminase 2 (Tgm2)-reporter plasmid was constructed by cloning the upstream regions into pGL3-basic vector (Promega, Madison WI) using primers listed in Supplementary Table 1. These luciferase reporters were transfected into Hepa-1 cells and luciferase assays were performed as previously described(16).

Animals and diets

Vhl^F/F, Vhl^F/FHif-1α^{F/F, and} Vhl^F/FHif-2α^F/F were previously described(16). For temporal hepatocyte-specific disruption Vhl^F/F, Vhl^F/FHif-1α^F/F, and Vhl^F/FHif-2α^F/F mice were crossed with mice harboring the cre-ER^T2 recombinase under control of the albumin promoter, SA-Cre-ER^T2(17). The mice are a mixed Sv129 and C57BL/6 background and wild-type littermate control mice were used as comparison for each experiment. The mice were used between the ages of 6- to 8-weeks-old for all experiments. For activation of the SA-Cre-ER^T2 recombinase for short term experiments (1- and 3-days) the mice were treated with one dose of tamoxifen (2mg/mouse in corn oil) by intraperitoneal (IP) injection and killed 24-hours or 3-days following tamoxifen treatment. For the 7-day and 2-week experiments, mice were fed tamoxifen in the diet for 2-days and then replaced with regular chow and killed at 7-days or 2-weeks after initial tamoxifen administration. For the alcohol treatment the mice were treated with tamoxifen by IP injection on 2 consecutive days and then were fed ad libitum a 4% alcohol-containing liquid diet (Lieber-DeCarli Diet, Dyets, Inc., Bethlehem, PA) and killed 2-weeks following alcohol administration. The mice were housed in temperature and light-controlled rooms, and were given water and pelleted chow ad libitum. All animal studies were carried out in accordance with guidelines and approved by the National Cancer Institute and University of Michigan Animal Care and Use Committee.

RNA analysis

RNA was extracted from tissues, reverse transcribed and quantitative Real-Time RT-PCR (qPCR) was performed using primer sequences listed in Supplementary Table 1.

Western blot analysis

Liver whole cell or nuclear extracts were prepared. The membranes were incubated with an antibodies against HIF-1α, HIF-2α (Novus Biologicals, Littleton, CO), ANGPTL3 (Santa Cruz Biotechnology Inc, Santa Cruz, CA), and smooth muscle actin (SMA) (Sigma), phophorylated and total acetyl-CoA carboxylase (ACC) (Cell Signaling Technology, Beverly, MA) the signals obtained were normalized to GAPDH (Santa Cruz) for whole cell extract and histone H1 (Santa Cruz), pregnane X receptor (PXR), and hepatic nuclear factor 4 (HNF4α) (Abcam, Cambridge, MA) for nuclear extracts.

cDNA microarray analysis

Liver cDNAs were hybridized to an Agilent 44 K mouse 60-mer oligo microarray (Agilent Technologies, Santa Clara, CA). The data was processed and analyzed by a Genespring GX software package (Agilent Technologies).

Immunohistochemistry

Hematoxylin & Eosin (H&E) staining and Masson’s Trichrome staining were performed on formalin fixed paraffin embedded sections. Oil red O staining was performed on frozen liver sections or adherent hepatoma-derived Hepa-1 cells. For quantification of oil red O in Hepa-1 cells, isopropanol was added to the cells following staining. Absorbance was measured at 510nm in the isopropanol extracts and the values were normalized to protein content.

Triglyceride and Cholesterol analysis

Hepatic lipids were extracted using 2:1 part chloroform-methanol solution. Liver and serum triglycerides were measured using Serum Triglyceride and Cholesterol Determination Kit according to manufacturer’s recommendation (Wako, Richmond, VA)

ChIP assays

Livers were crosslinked in 1% formaldehyde in 1X PBS at 37 C for 20 min. ChIP assays were performed for HIF-2α as previously described(16). Primers for qPCR ChIP are available upon request. The primers for Tgm2 ChIP are listed in Supplementary Table 1.

Data analysis

Results are expressed as mean ± S.D. P values were calculated by Independent t-test. p < 0.05 was considered significant.

Results

Generation of a mouse model containing a temporal hepatocyte-specific disruption of Vhl

Vhl^F/F mice were crossed with SA-Cre-ER^T2 transgenic mice to generate a temporal and conditional disruption of Vhl (Vhl^F/F;AlbERcre). The tamoxifen-inducible cre provides an advantage of assessing immediate downstream pathways controlled by VHL, and eliminates the confounding developmental effects of Vhl deletion. To confirm the inducibility and hepatocyte-specific disruption, Vhl^F/F and Vhl^F/F;AlbERcre mice were treated with one dose of vehicle or tamoxifen and livers and extrahepatic tissues were isolated 24-hours post-treatment. Vhl^F/F and Vhl^F/F;AlbERcre mice treated with vehicle did not demonstrate a decrease in Vhl gene expression, whereas tamoxifen treatment dramatically decreased Vhl gene expression in the Vhl^F/F;AlbERcre mice, but not the Vhl^F/F mice (Figure 1A). Moreover, the decrease was specific for the liver; no other tissues assessed demonstrated a tamoxifen-dependent decrease in Vhl expression (Supplemental Figure 1). Western blot analysis of nuclear extracts demonstrated an increase in HIF-1α and HIF-2α expression (Figure 1B). Consistent with HIFα subunit expression, an increase in pyruvate dehydrogenase kinase 1 (Pdk1) and erythropoietin (Epo) two well-characterized HIF-1α and HIF-2α target genes were observed (Figure 1C). In mice that contained a conditional disruption of Vhl, increased liver and spleen weights were noted at 6–8 weeks of age(9, 11). Therefore, to assess if these were early events following loss of VHL, liver and spleen weights were measured in mice in which Vhl was disrupted for 14-days. A significant increase in liver and spleen weights were observed (Figure 1D–F). Together this data demonstrates that the tamoxifen-inducible Vhl disruption is an optimal system to assess the primary responses, which are critical in hypoxia-induced liver injury.

Figure 1. Conditional and temporal disruption of Vhl in hepatocytes leads to HIF-1α and HIF-2α activation.

(A) qPCR analysis measuring liver Vhl expression in Vhl^F/F;AlbERcre or Vhl^F/F mice treated with vehicle (VEH) or tamoxifen (TM) and killed 24-hours post-treatment. Expression was normalized to β-actin. 7–9 mice were assessed per each treatment group. (B) Western blot analysis in liver nuclear extracts in Vhl^F/F;AlbERcre mice treated with vehicle (VEH) and tamoxifen (TM) and killed 24-hours and 3-days post treatment. Expression was normalized to Histone H1 protein expression. (C) qPCR analysis measuring liver pyruvate dehydrogenase kinase 1 (Pdk1) and erythropoietin (Epo) expression in Vhl^F/F;AlbERcre mice treated with vehicle (VEH) or tamoxifen (TM) 3-days post treatment. Expression was normalized to β-actin. 7–9 mice were assessed per each treatment group. (D) Liver weight analysis in Vhl^F/F;AlbERcre or Vhl^F/F mice treated with vehicle (VEH) or tamoxifen (TM) and killed 14-days post-treatment. Liver weights were normalized to body weights. 7–9 mice were assessed per each treatment group. (E and F) Spleen weight analysis in Vhl^F/F;AlbERcre or Vhl^F/F mice treated with vehicle (VEH) or tamoxifen (TM) and killed 14-days post-treatment. (F) Spleen weights were normalized to body weights. 7–9 mice were assessed per each treatment group. Each bar graph represents the mean value ± S.D. (*)= P<.05. 7

HIF-2α increases inflammation and lipid accumulation in the liver

Conditional inactivation of Vhl in hepatocytes results in liver inflammation and hepatic steatosis(9, 11, 14). However, it is not clear if inflammation and lipid accumulation are early events following disruption of Vhl or are due to the developmental or chronic effects from loss of Vhl. To address these questions livers were analyzed following disruption of Vhl for 2-weeks, a robust increase in liver inflammation was observed by H&E staining and qPCR analysis of two pro-inflammatory mediators, interleukin(Il)-1β and Il-6 (Figure 2A–C). The increase in Il-6 and Il-1β gene expression was evident as early as 3-days following tamoxifen treatment (Figure 2D). Overt inflammation as observed by H&E staining was evident at 7-days following tamoxifen treatment (Supplemental Figure 2A). To assess the influence of HIF-dependent pathways on inflammatory gene expression in the liver, mice with a double disruption of Vhl and Hif-1α or Hif-2α were generated. The double disruption of Vhl and Hif-2α (Vhl^F/FHif2a^F/F;AlbERcre+TM) ameliorated the increase in Il-6 and Il-1β compared to littermate controls (Vhl^F/FHif2a^F/F+TM) (Figure 2E). In contrast, a significant increase in Il-6 and Il-1β gene expression was observed in mice with a double disruption of Vhl and Hif-1α compared to littermate controls (Supplemental Figure 2B). Furthermore, 2-weeks following loss of Vhl, a dramatic increase in liver lipid accumulation was observed by oil red O staining (Figure 3A and B). The increase in lipid accumulation could be observed as early as 24-hours following Vhl disruption (Figure 3C and D). The compound disruption of Vhl and Hif-1α or Hif-2α demonstrated that the increase in lipid accumulation was due to HIF-2α but not HIF-1α. (Figure 3E and F). Consistent with oil red O staining, hepatic triglycerides and cholesterol increased following disruption of Vhl for 2-weeks (Figure 3G). Together this data demonstrates that HIF-2α is a direct regulator of liver inflammation and lipid accumulation in the liver.

Figure 2. HIF-2α induces liver inflammation.

H&E stained liver sections from (A) Vhl^F/F and (B) Vhl^F/F;AlbERcre mice treated with tamoxifen (TM) and killed 14-days post-treatment. 7–9 mice were assessed per each treatment group. (C) qPCR analysis measuring liver Il-6 and Il-1β expression in Vhl^F/F and Vhl^F/F;AlbERcre mice treated with vehicle (VEH) or tamoxifen (TM) and killed 14-days post treatment. Expression was normalized to β-actin. 7–9 mice were assessed per each treatment group. (D) qPCR analysis measuring liver Il-6 and Il-1β expression in Vhl^F/F and Vhl^F/F;AlbERcre mice treated with tamoxifen (TM) and killed 3-, 7-, and 14-days post treatment. Expression was normalized to β-actin. 7–9 mice were assessed per each treatment group. (E) qPCR analysis measuring liver Il-6 and Il-1β expression in Vhl^F/FHif2α^F/F and Vhl^F/FHif2α^F/F;AlbERcre mice treated with tamoxifen (TM) and killed 14-days post treatment. Expression was normalized to β- actin. 7–9 mice were assessed per each treatment group. Each bar graph represents the mean value ± S.D. (*)= P<.05.

Figure 3. HIF-2α increases lipid accumulation in the liver.

Oil red O staining of livers isolated from Vhl^F/F and Vhl^F/F;AlbERcre mice treated with tamoxifen (TM) and killed (A and B) 14-days or (C and D) 24-hours post-treatment. (E and F) Oil red O staining of livers isolated from (E) Vhl^F/FHif1α^F/F;AlbERcre and (F) Vhl^F/FHif2α^F/F;AlbERcre mice treated with tamoxifen (TM) and killed 14-days post treatment. (G) Hepatic lipid analysis was performed in livers isolated from Vhl^F/F and Vhl^F/F;AlbERcre mice treated with tamoxifen (TM) and killed 14-days post treatment. Each bar graph represents the mean value ± S.D. (*)= P<.05. 5–7 mice were assessed per each treatment group.

HIF regulated genetic program in the liver

To understand the critical genes regulated following Vhl disruption, gene expression profiles of Vhl^F/F and Vhl^F/F;AlbERcre were assessed in livers isolated 24-hours or 2-weeks following Vhl disruption. 3597 significantly regulated changes were identified following 2-weeks of Vhl deletion, whereas 470 genes were identified 24-hours following Vhl disruption (Figure 4A and the full gene list with an average change of 1.5 fold is in Supplementary Tables 2 and 3). The data suggested that a rapid increase in genes critical for lipid synthesis was followed by an increase in genes important for fatty acid uptake. Consistent with the microarray data, an increase was observed in the expression of fatty acid synthase (Fasn) and sterol regulatory element binding factor-1C (Srebp-1c) at 3-days following Vhl disruption. Interestingly, at 14-days following Vhl disruption, a significant repression of Fasn and Srebp-1c was observed (Figure 4B). Whereas, a rapid repression of Cd36 gene expression was observed following 3-days of Vhl disruption followed by a dramatic increase in gene expression 14-days following loss of Vhl (Figure 4B). In addition, a significant decrease was observed in genes critical in fatty acid β-oxidation, a decrease in carnitine palmitoyltransferase 1A (Cpt1a), carnitine palmitoyltransferase 2 (Cpt2), acyl-CoA oxidase 1 (Acox), and peroxisome proliferator-activated receptor alpha (Pparα) were observed following 2-weeks of Vhl disruption; Pparα expression did not reach statistical significance (Figure 4C). Interestingly, the expression of PPARα protein was significantly decreased 2-weeks following Vhl disruption suggesting enhanced degradation (Figure 4D). Expression of two other important nuclear receptors, PXR and HNF4α, were unchanged. The decrease in β-oxidation genes were not observed at 3-days following Vhl disruption, but were dependent on HIF-2α expression (Supplemental Figure 3). These data suggest that HIF-2α regulates fatty acid synthesis, uptake and β-oxidation in a time-dependent manner.

Figure 4. HIF-2α modulates lipid homeostatic genes in the liver.

(A) Global gene expression profiling was assessed in liver RNAs isolated from Vhl^F/F and Vhl^F/F;AlbERcre mice treated with tamoxifen (TM) and killed 24-hours or 14-days post-treatment. 4 mice were used for each treatment group. (B) qPCR analysis of sterol regulatory element binding factor-1C (Srebp-1c), fatty acid synthase (Fasn), and Cd36 in livers from Vhl^F/F and Vhl^F/F;AlbERcre mice treated with tamoxifen and killed 3- or 14-days post-treatment. Expression was normalized to β-actin. (C) qPCR analysis of β-oxidation genes, carnitine palmitoyltransferase 1A (Cpt1a), carnitine palmitoyltransferase 2 (Cpt2), acyl-CoA oxidase 1 (Acox), and peroxisome proliferator-activated receptor alpha (Pparα) in livers from Vhl^F/F and Vhl^F/F;AlbERcre mice treated with tamoxifen and killed 14-days post-treatment. Expression was normalized to β-actin. 7–9 mice were assessed per each treatment group. (D) Western blot analysis measuring PPARα, pregnane X receptor (PXR) and hepatic nuclear factor 4 (HNF4α) expression from livers in Vhl^F/F and Vhl^F/F;AlbERcre mice treated with tamoxifen and killed 14-days post-treatment. Each bar graph represents the mean value ± S.D. (*)= P<.05.

HIF-2α is a novel regulator of Angptl3

SREBP-1c, FASN, CD36, and PPARα have critical roles of in fatty acid homeostasis in the liver, however their gene expression patterns suggest these genes may not be direct targets for HIF-2α in the liver. Interestingly, angiopoietin-like 3 (Angptl3) demonstrated rapid and sustained increase following Vhl disruption (Figure 5A). ANGPTL3 is specifically expressed in the liver and is a direct regulator of lipid homeostasis(18–20). Mutations in Angptl3 in mice or humans are associated with low serum lipid levels, whereas overexpression of ANGPTL3 increases circulating lipid levels(18, 20). In mice with a double disruption of Vhl and Hif-2α demonstrated that the induction of Angptl3 was due to HIF-2α increase (Supplemental Figure 4). Gene expression data correlated to an increase in protein expression, tamoxifen treated Vhl^F/F;AlbERcre mice demonstrated an increase in liver ANGPTL3 protein expression compared to tamoxifen treated Vhl^F/F mice (Figure 5B). Since mouse models which overexpress ANGPTL3 demonstrated an increase in serum lipid levels(20), serum triglycerides were assessed in mice 2-weeks after loss of Vhl. Vhl^F/F;AlbERcre treated with tamoxifen had elevated serum triglycerides compared to similarly treated Vhl^F/F mice (Figure 5C). In addition, liver-derived Hepa-1 cells, which overexpress ANGPTL3, demonstrated a dose-dependent increase in oil red O accumulation, suggesting that ANGPTL3 may play a critical role in HIF-mediated lipid accumulation (Figure 5D). To assess whether ANGPTL3 could be a novel direct target of HIF-2α, Angptl3-promoter luciferase assays were performed. A 1.7kb Angptl3 proximal promoter luciferase construct was transfected into Hepa-1 cells. Hypoxia (1% O₂) induced luciferase expression (Figure 5E), and co-transfection with a mammalian expression plasmid for HIF-1α moderately increased luciferase expression, while co-transfection with HIF-2α expression plasmid strongly increased the luciferase expression. The HIF-1α and HIF-2α increase in luciferase expression was further potentiated in cells incubated in 1% O₂ (Figure 5E). Deletion analysis showed that the HIF responsive site on the Angptl3 promoter was within the first 100bp of the proximal promoter, however no consensus HIF response element (HRE) was found with in this site (Figure 5F). Furthermore, in vivo ChIP assays failed to demonstrate HIF-2α binding to the promoter (data not shown). Together, this data suggest that Angptl3 is a rapid HIF-2α responsive gene through yet unknown mechanism.

Figure 5. HIF-2α increases angiopoietin-like 3 (Angptl3) expression in the liver.

(A) qPCR analysis of Angptl3 in livers from Vhl^F/F and Vhl^F/F;AlbERcre mice treated with tamoxifen and killed 3-, 7-, or 14-days post-treatment. Expression was normalized to β-actin. 7–9 mice were assessed per each treatment group. (B) Western blot analysis measuring ANGPTL3 expression in livers from Vhl^F/F and Vhl^F/F;AlbERcre mice treated with tamoxifen and killed 14-days post-treatment. Expression was normalized to GAPDH protein expression. (C) Serum triglycerides analysis measuring from Vhl^F/F and Vhl^F/F;AlbERcre mice treated with tamoxifen and killed 14-days post-treatment. 7–9 mice were assessed per each treatment group. (D) Hepa-1 cells were transfected with increasing amounts of mouse Angptl3. 48 hours post-transfection the cells were washed and stained with oil red O or Western analysis performed to verify overexpression (top panel). Oil red O accumulation was quantitated in isopropanolic extracts of the cells and the data was normalized to protein content (bottom panel). (E) Luciferase-reporter constructs under the control of the (E) 1.7kb regulatory region of the mouse Angptl3 gene or (F) 5’ deletion constructs of the Angptl3 promoter. Hepa-1 cells transiently transfected with the luciferase construct, and co-transfected with empty vector, HIF-1α or HIF-2α expression plasmids. Standard dual luciferase assays were performed on cells incubated in normal or hypoxic cell culture conditions. Each bar graph represents the mean value ± S.D. (*) = P<.05.

HIF-2α regulates liver fibrogenesis

A dramatic induction of genes that are important in liver fibrosis were observed in the gene expression profiling data. Increase in several fibrogenic genes were confirmed in Vhl^F/F;AlbERcre mice treated with tamoxifen compared to littermate control mice (Figure 6A). A specific increase in lysyl oxidase-like 1 (LOXL1), lysyl oxidase-like 2 (LOXL2), prolyl 4-hydroxylase alpha 1 (P4HA1), prolyl 4-hydroxylase alpha 2 (P4HA2), procollagen-lysine, 2-oxoglutarate 5-dioxygenase 2 (PLOD2), and transglutaminase 2 (TGM2) were observed. These genes are critical for the formation and stabilization of collagen(21–26). In addition, smooth muscle actin (SMA), a marker of stellate cell activation and fibrosis was significantly increased in Vhl^F/F;AlbERcre mice treated with tamoxifen compared to littermate control mice as assessed by qPCR and Western blot analysis (Figure 6A and B). To confirm an increase in fibrosis, Masson’s Trichrome staining was performed (Figure 6C and D). Livers isolated from Vhl^F/F;AlbERcre 14-days following tamoxifen treatment demonstrated a moderate increase in focal areas of fibrosis compared to similarly treated Vhl^F/F mice (Figure 6C). Moreover, Vhl^F/F and Vhl^F/F;AlbERcre mice were treated with tamoxifen and then put on liquid diet consisting of 4% ethanol for 2-weeks. Mice are resistant to alcohol-induced fibrosis as chronic treatment with alcohol (over 3-months) typically results in no marked liver fibrogenesis in mice(27). However, in mice with a disruption of liver Vhl, alcohol treatment caused marked fibrosis compared to littermate controls treated with alcohol (Figure 6D). The double disruption of Vhl and Hif-2α (Vhl^F/FHif2a^F/F;AlbERcre+TM) ameliorated the increase in SMA, whereas a significant increase in SMA expression was observed in mice with a double disruption of Vhl and Hif-1α (Vhl^F/FHif1a^F/F;AlbERcre+TM) (Figure 7A). Similarly, the increase in fibrosis observed in Vhl disrupted mice on alcohol diet was completely lost in the Vhl and Hif-2α double knockout, but not the Vhl and Hif-1α double knockouts (Figure 7B). Consistent with the role of HIF-2α in exacerbating fibrosis, fibrogenic gene expression levels were not increased in the Vhl and Hif-2α knockout as compared to mice with a Vhl disruption (Figure 7C). Together, these data demonstrate that HIF-2α is a critical transcription factor in exacerbating fibrosis in the liver.

Figure 6. Disruption of Vhl increases liver fibrosis.

(A) qPCR analysis measuring expression of fibrogenic genes in livers of Vhl^F/F;AlbERcre or Vhl^F/F mice treated with vehicle (VEH) or tamoxifen (TM) and killed 14-days post-treatment (collagen 5a2; Col5a2, collagen 12a1; Col12a1, connective tissue growth factor; Ctgf, insulin-like growth factor binding protein-1; Igfbp1, lysyl oxidase-like 1; Loxl1, lysyl oxidase-like 2; Loxl2, α-2- macroglobulin; A2m, smooth muscle actin; SMA, collagen 1a1; Col1a1, collagen 3a1; Col3a1, collagen 4a1; Col4a1, collagen 4a2; Col4a2, prolyl 4-hydroxylase alpha 1; P4ha1, prolyl 4-hydroxylase alpha 2; P4ha2, procollagen-lysine, 2-oxoglutarate 5-dioxygenase 2; Plod2, transforming growth factor b1; Tgfb1, transglutaminase 2; Tgm2, tissue inhibitor of metallopeptidase 1; Timp1). Expression was normalized to β-actin. 7–9 mice were assessed per each treatment group. (B) Western blot analysis measuring SMA expression in livers from Vhl^F/F and Vhl^F/F;AlbERcre mice treated with tamoxifen and killed 14-days post-treatment. (C and D) Masson’s Trichrome stained livers from Vhl^F/F and Vhl^F/F;AlbERcre mice treated with tamoxifen and killed 14-days post-treatment on (C) normal chow diet or (D) liquid alcohol diet. Fibrotic cells are indicated by arrows or are outlined by a dashed line. 7–9 mice were assessed per each treatment group. Each bar graph represents the mean value ± S.D. (*) = P<.05 and n=7–9 mice per each treatment group.

Figure 7. HIF-2α activation increases liver fibrosis.

(A) Western blot analysis of SMA from livers of Vhl^F/FHif1α^F/F;AlbERcre and Vhl^F/FHif2α^F/F;AlbERcre mice treated with tamoxifen (TM) and killed 14-days post-treatment. Expression was normalized to GAPDH protein expression. (B) Masson’s Trichrome stained livers from Vhl^F/FHif1α^F/F;AlbERcre and Vhl^F/FHif2α^F/F;AlbERcre mice treated with tamoxifen and killed 14-days post-treatment on liquid alcohol diet. Fibrotic cells are indicated by arrows or are outlined by a dashed line. 7–9 mice were assessed per each treatment group. (C) qPCR analysis measuring expression of fibrogenic genes in livers of Vhl^F/FHif1α^F/F;AlbERcre or Vhl^F/FHif2α^F/F mice treated with tamoxifen (TM) and killed 14-days post-treatment (connective tissue growth factor; Ctgf, insulin-like growth factor binding protein-1; Igfbp1, lysyl oxidase-like 1; Loxl1, lysyl oxidase-like 2; Loxl2, α-2- macroglobulin; A2m, smooth muscle actin; SMA, collagen 1a1; Col1a1, prolyl 4-hydroxylase alpha 1; P4ha1, , prolyl 4-hydroxylase alpha 2; P4ha2, procollagen-lysine, 2-oxoglutarate 5-dioxygenase 2; Plod2, transforming growth factor b1; Tgfb1, transglutaminase 2; Tgm2). Expression was normalized to β-actin. 7–9 mice were assessed per each treatment group. Each bar graph represents the mean value ± S.D. (*) = P<.05

HIF-2α directly regulates several profibrogenic genes

To assess if HIF-2α could directly regulate fibrogenic genes in the liver, ChIP assays were performed using crosslinked liver DNA isolated from tamoxifen treated Vhl^F/F and Vhl^F/F;AlbERcre mice with the average shearing size of 1.5kb. Primers were designed to the center of the proximal promoter to assess HIF-2α occupancy. This method provides assessment of HIF-2α occupancy at promoters without defining the precise HIF response element. With this method, it was shown that HIF-2α was enriched at the promoters of several fibrogenic genes in Vhl^F/F;AlbERcre mice compared to control littermates (Figure 8A). To assess whether the low-resolution ChIP assays indeed did identify direct targets, TGM2 expression was further assessed. Increase in TGM2 protein expression was observed in Vhl^F/F;AlbERcre mice compared to control littermates following 2-weeks of Vhl disruption (Figure 8B). Next, a Tgm2 proximal promoter luciferase construct was co-transfected into liver-derived Hepa-1 cells with a mammalian expression plasmid for HIF-1α, HIF-2α or empty vector. HIF-2α specifically induced luciferase expression, whereas HIF-1α had no effect compared to empty vector transfected control (Figure 8C), and mutating the two putative HREs ablated HIF-2α activity (Figure 8D). Using primers flanking the HREs and sheared crosslinked liver DNA (shearing size 300bp) from tamoxifen treated Vhl^F/F and Vhl^F/F;AlbERcre mice demonstrated increased HIF-2α binding to the Tgm2 promoter in livers isolated from Vhl^F/F;AlbERcre mice compared to Vhl^F/F mice (Figure 8E). This data demonstrates the HIF-2α can directly regulate fibrogenic genes.

Figure 8. HIF-2α can directly bind to promoters of fibrogenic genes in the liver.

(A) ChIP assays of livers from Vhl^F/F;AlbERcre or Vhl^F/F treated with tamoxifen (TM) and killed 3-days post-treatment using HIF-2α specific antibody (collagen 5a2; Col5a2, collagen 12a1; Col12a1, connective tissue growth factor; Ctgf, insulin-like growth factor binding protein-1; Igfbp1, lysyl oxidase-like 1; Loxl1, lysyl oxidase-like 2; Loxl2, α-2- macroglobulin; A2m, smooth muscle actin; SMA, collagen 1a1; Col1a1, collagen 3a1; Col3a1, collagen 4a1; Col4a1, collagen 4a2; Col4a2, prolyl 4-hydroxylase alpha 1; P4ha1, , prolyl 4-hydroxylase alpha 2; P4ha2, procollagen-lysine, 2-oxoglutarate 5-dioxygenase 2; Plod2, transforming growth factor b1; Tgfb1, transglutaminase 2; Tgm2, tissue inhibitor of metallopeptidase 1; Timp1). The data is expressed as fold enrichment over control IgG and normalized to input. 7–9 mice were assessed per each treatment group. (B) Western blot analysis measuring TGM2 expression in livers from Vhl^F/F and Vhl^F/F;AlbERcre mice treated with tamoxifen and killed 14-days post-treatment. Expression was normalized to GAPDH. (C) Luciferase-reporter constructs under the control of the 1kb regulatory region of the mouse Tgm2 gene. Hepa-1 cells transiently transfected with the luciferase construct, and co-transfected with empty vector, HIF-1α or HIF-2α expression plasmids. Standard dual luciferase assays were performed. (D) Tgm2 promoter illustrating the HREs in the regulatory region, the upstream regions are numbered in relation to the transcription initiation site, which is designated +1 (top panel). HRE deleted Tgm2 Luciferase-reporter was co-transfected with empty vector or HIF-2α expression plasmids (bottom panel). Standard dual luciferase assays were performed. (E) ChIP assays of livers from Vhl^F/F;AlbERcre or Vhl^F/F treated with tamoxifen (TM) and killed 3-days post-treatment using HIF-2α specific antibody. Primers flanking the two HREs were used for PCR. Each bar graph represents the mean value ± S.D. (*) = P<.05. 3–5 mice were assessed per each treatment group.

Discussion

One third of adults in the United States are diagnosed with fatty liver disease mostly attributed to obesity or alcohol consumption. About 10% will proceed to develop steatohepatitis, and associated co-morbidities (fibrosis, cirrhosis and liver cancer)(28). Currently the mechanisms for the increased progression are not known. However, according to the two-hit hypothesis the initial insult is the fat accumulation within the liver and second insult being increased oxidative stress and inflammation and both are critical for steatohepatitis(29). The current study demonstrates that mice with a temporal hepatic disruption of Vhl have spontaneous fatty liver and liver inflammation that will progress to focal fibrosis and hepatomegaly in a HIF-2α-dependent manner. This demonstrates that hypoxia and HIF-2α play critical role in both insults needed for progression of fatty liver disease as suggested by the two-hit hypothesis.

Gene expression profiling demonstrated that several genes important in fatty acid synthesis, uptake, and β-oxidation are significantly altered after loss of VHL. Fasn and Srebp-1c were repressed in mice with a conditional disruption of Vhl, therefore fatty acid synthesis was not thought to be involved in increased lipid accumulation in the liver following Vhl disruption(14). However, the present data suggest that at early times points, lipid synthesis may contribute to steatosis since both Fasn and Srebp-1c are significantly increased following acute disruption and then are significantly repressed following long-term Vhl deficiency. To assess if indeed at early timepoints following HIF activation that fatty acid synthesis was increased, ACC activity was measured. However, both phosphorylated and total ACC were significantly repressed at 3-days following tamoxifen treatment in the Vhl^F/F;AlbERcre mice compared to Vhl^F/F mice, making the data difficult to interpret (Supplemental Figure 5). However, 14-days following tamoxifen treatment no change in phosphorylated or total ACC was observed in the Vhl^F/F;AlbERcre mice compared to Vhl^F/F mice, suggesting β-oxidation maybe a critical driver in fat accumulation at later time points. In addition, the gene expression analysis demonstrates a significant modulation of several nuclear receptor target genes (liver X receptor, farnesoid X receptor, and PPARγ). However, changes were not found in the expression of these nuclear receptors by qPCR or microarray analysis, suggesting nuclear receptors are not direct transcriptional targets of HIF. Interestingly, in mice with the conditional Vhl deletion, adipose differentiation-related protein (ADFP) was significantly induced and thought to be critical in the liver steatotic phenotype(14). However, in the Vhl^F/F;AlbERcre mice following tamoxifen treatment, no increase in ADFP was observed at any time point assessed (data not shown), suggesting that the increase in ADFP is a late secondary response or to due to developmental defects following conditional Vhl disruption. These data highlight the importance of temporal gene disruption of Vhl to identify direct mediators of response.

One important mediator of lipid homeostasis, ANGPTL3, an endogenous lipoprotein lipase (LPL) inhibitor(30–32), was identified as a HIF responsive gene. ANGPTL3 is important in regulating serum triglycerides levels(20). In tamoxifen-treated Vhl^F/F;AlbERcre mice, the increase of ANGPTL3 correlated to an increase in serum triglycerides, and ANGPTL3 directly increased lipid accumulation in Hepa-1 cells as assessed by oil red O staining. Currently, it is not known whether the increase in lipid accumulation is through the LPL inhibitor function of ANGPTL3, but is a clear point of emphasis for future studies. Angptl3 gene expression and promoter activity were rapidly induced by HIF-2α. However, no HREs were identified in the promoter, suggesting that its activation is HIF-2α-mediated through an indirect mechanism. The HIF responsive region was localized to a 100bp region directly proximal to the transcription initiation site and HIF-2α regulation of this sequence is being further assessed.

During the preparation of this manuscript, others published similar findings in a temporally deleted liver-specific VHL mouse model in which disruption of Vhl was induced by tail vein injection of adenovirus encoding cre recombinase (ad-Cre)(33). Five days following injection of ad-Cre, the mice demonstrated dramatic steatosis and a decrease in PPARα signaling thus establishing as does the present study that HIF signaling has a primary role in liver lipid homeostasis. Furthermore, the present study demonstrates that these are immediate and rapid responses of HIF-2α signaling. Interestingly, following ad-Cre injection, the mice demonstrated rapid death in a HIF-dependent manner, where the median survival was 6 days(34). The increase in survival in the Vhl^F/F;AlbERcre mice following tamoxifen administration allowed further assessment of the livers, revealing increased progression of steatosis to inflammation. Thus, the Vhl^F/F;AlbERcre mice may be a valuable model of spontaneous steatohepatitis for use in preclinical drug development.

While the direct effectors that increase inflammation are not known, it is possible that HIF-2α can directly activate inflammatory mediators in the liver. Indeed, it was shown that Il-6 is direct HIF-2α target gene in macrophages(34). However, our data clearly show that HIF-2α can bind to the promoters of several pro-fibrogenic genes, consistent with data demonstrating that hypoxia can activate fibrogenesis in hepatocytes and stellate cells(35–37). Hepatic stellate cells initiate the fibrotic process. In liver, quiescent stellate cells are critical in storage of vitamin A. During liver injury, stellate cells become activated, proliferate and express a fibrogenic gene program(38). Following Vhl disruption, a robust activation of stellate cells is observed in the liver due to high activation of collagen gene expression and increase in SMA, both markers of stellate cell activation. The initiating factor in activation of stellate cells following Vhl loss is thought to be due to a sustained increase in lipid accumulation and inflammatory genes. In addition, the increase in fibrosis mediated by HIF-2α may be due to collagen matrix stabilization. P4HA1, P4HA2, and PLOD2 are required for hydroxylation of lysyl and prolyl residues on collagen(23, 26). The resultant hydroxylysyl and hydroxyproline groups are critical for the stability and synthesis of collagen matrixes. Loxl1 and loxl2 gene expression were also increased in the livers of tamoxifen treated Vhl^F/F;AlbERcre mice and their respective promoters were occupied by HIF-2α. Lysyl oxidase activity is critical in formation of insoluble collagen fibers, and HIF-1α has been shown to increase renal fibrosis through a lysyl oxidase-mediated mechanism(21, 22). Moreover, TGM2, a multifunction enzyme that covalently cross-links collagen matrixes, has been shown to be critical in inducing apoptosis by inactivation SP1 and c-met in injured livers following alcohol administration(24, 25). HIF-2α can directly regulate the promoter of Tgm2 in a distinct manner as observed with HIF-1α(39). It is not clear if Tgm2 is the key enzyme that regulates fibrosis since Tgm2-null mice are not protected in the carbon tetrachloride and the thioacetamide-induced fibrosis models(40). However, it is likely that the cumulative increase in several profibrogenic genes are needed to increase liver fibrosis and HIF-2α may be the critical transcription factor to integrate these signals.

The present study demonstrates that activation of HIF-2α in the liver regulates liver homeostasis and disease progression and establishes that steatosis, inflammation, and fibrosis are direct responses initiated by the liver following HIF-2α activation. In addition, the present work provides a novel animal model to study the precise molecular and genetic changes that are required for the progression of fatty liver disease to steatohepatitis. Together, these findings may lead to novel therapies for liver injury.

Abbreviations

A2M

α-2-macroglobulin

ACOX

acyl-CoA oxidase 1

ADFP

adipose differentiation-related protein

ANGPTL3

angiopoietin-like 3

ARNT

aryl hydrocarbon nuclear translocator

CPT1A

carnitine palmitoyltransferase 1A

CPT2

carnitine palmitoyltransferase 2

ChIP

chromatin immunoprecipitation

COL1A1

collagen 1a1

COL3A1

collagen 3a1

COL4A1

collagen 4a1

COL4A2

collagen 4a2

COL5A2

collagen 5a2

COL12A1

collagen 12a1

CTGF

connective tissue growth factor

FASN

fatty acid synthase

EPO

erythropoietin

H&E

hematoxylin & eosin

HIF

hypoxia-inducible factor

IL-1β

interleukin-1β

IL-6

interleukin-6

IGFBP1

insulin-like growth factor binding protein-1

LOXL1

lysyl oxidase-like 1

LOXL2

lysyl oxidase-like 2

PPARα

peroxisome proliferator-activated receptor alpha

P4HA1

prolyl 4-hydroxylase alpha 1

P4HA2

prolyl 4-hydroxylase alpha 2

PLOD2

procollagen-lysine, 2-oxoglutarate 5-dioxygenase 2

PDK1

pyruvate dehydrogenase kinase 1

qPCR

quantitative Real-Time RT-PCR

SMA

smooth muscle actin

SREBP-1C

sterol regulatory element binding factor-1C

TIMP1

tissue inhibitor of metallopeptidase 1

TGFB1

transforming growth factor b1

TGM2

transglutaminase 2

VHL

Von Hippel-Lindau tumor suppressor protein