Hypoxia‐inducible factor 2α drives hepatosteatosis through the fatty acid translocase CD36

Abstract Background & Aims Molecular mechanisms by which hypoxia might contribute to hepatosteatosis, the earliest stage in non‐alcoholic fatty liver disease (NAFLD) pathogenesis, remain still to be elucidated. We aimed to assess the impact of hypoxia‐inducible factor 2α (HIF2α) on the fatty acid translocase CD36 expression and function in vivo and in vitro. Methods CD36 expression and intracellular lipid content were determined in hypoxic hepatocytes, and in hypoxic CD36‐ or HIF2α ‐silenced human liver cells. Histological analysis, and HIF2α and CD36 expression were evaluated in livers from animals in which von Hippel‐Lindau (Vhl) gene is inactivated (Vhlf/f‐deficient mice), or both Vhl and Hif2a are simultaneously inactivated (Vhlf/fHif2α/f‐deficient mice), and from 33 biopsy‐proven NAFLD patients and 18 subjects with histologically normal liver. Results In hypoxic hepatocytes, CD36 expression and intracellular lipid content were augmented. Noteworthy, CD36 knockdown significantly reduced lipid accumulation, and HIF2A gene silencing markedly reverted both hypoxia‐induced events in hypoxic liver cells. Moreover livers from Vhlf/f‐deficient mice showed histologic characteristics of non‐alcoholic steatohepatitis (NASH) and increased CD36 mRNA and protein amounts, whereas both significantly decreased and NASH features markedly ameliorated in Vhlf/fHif2αf/f‐deficient mice. In addition, both HIF2α and CD36 were significantly overexpressed within the liver of NAFLD patients and, interestingly, a significant positive correlation between hepatic transcript levels of CD36 and erythropoietin (EPO), a HIF2α ‐dependent gene target, was observed in NAFLD patients. Conclusions This study provides evidence that HIF2α drives lipid accumulation in human hepatocytes by upregulating CD36 expression and function, and could contribute to hepatosteatosis setup.


| INTRODUC TI ON
Overnutrition is a major contributor to the development of nonalcoholic fatty liver disease (NAFLD) because a high consumption of saturated fatty acids, cholesterol and fructose along with a low intake of polyunsaturated fatty acids, featuring NAFLD patients, alters hepatic lipid metabolism homeostasis leading to an excessive fat accumulation within the liver which activates inflammation, hepatocellular damage and fibrogenesis. 1,2 There is extensive clinical and experimental evidence indicating that chronic intermittent hypoxia, featuring a respiratory disorder of growing prevalence worldwide termed obstructive sleep apnoea, could contribute to the progression of NAFLD from simple steatosis, also termed non-alcoholic fatty liver (NAFL) or hepatosteatosis, to non-alcoholic steatohepatitis (NASH), 3-6 the most clinically relevant form of NAFLD with a significant risk to progress into cirrhosis and hepatocellular carcinoma, 7,8 as well as increasing the cardiovascular morbidity and mortality and the incidence of extrahepatic cancers. 9,10 Hypoxia-inducible factors (HIFs), particularly HIF1α and HIF2α, are master regulators of hypoxia-induced cellular adaptive responses elicited to restore cell metabolism and survival. 11 HIFs are implicated in numerous physiological and pathological conditions, and it has been reported that HIF2α promotes NASH in mice, 12,13 and dysregulates lipid metabolism in HepG2 cells. 14 Since lipotoxicity due to free fatty acids (FFAs) overload within hepatocytes plays a central role in NAFLD pathophysiology, 15 and that this process is largely regulated by membrane-bound FFA transporters, it is conceivable that HIFs might contribute to NAFLD pathogenesis by upregulating the expression and function of FFA transporters in the membrane of hepatocytes.
Among membrane-bound FFA transporters, the fatty acid translocase CD36 (CD36) is the best characterized. 16 CD36 functions as a high affinity receptor for long-chain FFAs contributing under excessive fat supply to lipid accumulation and metabolic dysfunction. 17 This FFA receptor is involved in several aspects of lipid metabolism including fat taste perception, fat intestinal absorption and FFA utilization by muscle, adipose tissues and liver. 18 Regarding the latter, hepatic CD36 expression is normally weak but it increases by a number of different stimuli such as cytokines or insulin. 18,19 Noteworthy, experimental studies have demonstrated that CD36 plays a key role in the hepatosteatosis setup in rodents 20,21 and, more interestingly, NAFLD patients present high hepatic CD36 mRNA levels, 22 and this FFA transporter is largely overexpressed in the plasma membrane of hepatocytes. 23 While there are evidences that HIF1α upregulates CD36 expression and function in retinal epithelial cells and macrophages, 24,25 whether HIF2α is able to regulate CD36 gene expression in hepatocytes still remains to be elucidated. Therefore, the aim of the present study was to determine the impact of HIF2α on CD36 expression and function as well as on lipid content in hepatocytes submitted to hypoxic conditions, in livers from genetically-modified mice in which von Hippel-Lindau (Vhl) gene is inactivated (Vhl f/f -deficient mice), a murine experimental model which displays NAFLD features due to an overexpression of HIF1 and HIF2, in livers from mice in which both Vhl and Hif2a are simultaneously inactivated (Vhl f/ f Hif2α f/f -deficient mice), and in livers from patients with biopsy-proven NAFLD.

| Nile Red staining
The cells were fixed with paraformaldehyde 4% for 30 minutes at 4ºC, and resuspended in Nile Red working solution to 0.4 µL/mL (Sigma-Aldrich Inc). The fluorescence was determined using a flow cytometer Cytomics FC500 MPL TM (Beckman-Coulter Inc).

| Animals
All animal experimentation was conducted in accordance with

| Gene expression analysis by real-time quantitative PCR
Total RNA from cells or liver samples was extracted using TRIzol reagent (Vitro, Sevilla, Spain) and was reverse transcribed using a

| Extraction of nuclear protein liver extracts
Liver biopsy samples were homogenized in 4 volumes (w/v) of cold buffer A (0.3 M Sucrose solution with protease inhibitors).
Samples were centrifuged and the supernatant containing the cytosolic fraction was stored at −80ºC. The pellet containing the nuclear fraction was washed for a total of three times with buffer A. Samples were incubated in rotation for 1 hour with cold buffer B (1% Trion x100, 1% Sodium Deoxycholate, 0.1% SDS, 5 mM EDTA, 200 nM NaCl, 20 mM Tris HCl pH 8 with protease inhibitors). After sonication, cellular debris was removed by centrifugation and the supernatant fraction containing the nuclear fraction was stored at −80ºC.

| Western blot analysis
After protein content determination with Bradford reagent, 50 µg of total protein or100 µg of nuclear protein was boiled in Laemmli sample buffer and submitted to 8% SDS-PAGE gels.
Proteins were transferred to Inmunoblot nitrocellulose membrane (BioRad Inc) and, after blocking with 5% non-fat dry milk, Densitometric analysis of the bands was performed using Image J software (NIH, Bethesda, MD).

| Histopathology assessment
Paraffin-embedded liver biopsy sections (4 µm thick) were stained with haematoxylin/eosin and evaluated by a

| Hypoxia induces lipid accumulation and CD36 expression in both human and murine hepatocytes
To explore the molecular mechanisms involved in the regulation of CD36 by hypoxia, Huh7 human liver cells were maintained under normoxic conditions (Nx, 21%O2) or submitted to hypoxia (Hp, 1%O2) for 36h. To assure that hypoxia was achieved in Huh7 cell cultures, we assessed the expression of HIFα protein subunits by western blot. As Figure 1A shows, HIF1α and HIF2α expression was markedly induced by hypoxia. Accordingly, PHD3 upregulation, a well-recognized HIF responsive gene, was found in hypoxia-exposed Huh7 liver cells ( Figure 1B).
Next, we investigated whether Huh7 cells submitted to hypoxic conditions would increase their intracellular lipid content performing Nile Red staining experiments. After flow cytometry analysis, we observed a significant increase in the lipid content of Huh7 cells submitted to hypoxia, compared to those maintained under normoxic conditions ( Figure 1C). Interestingly, a parallel increase in CD36 mRNA and protein levels was found in hypoxic Huh7 cells ( Figure 1D). Similar findings were observed in mouse hepatocytes (AML12 cell line) submitted to hypoxia in which both intracellular lipid content and CD36 expression were significantly augmented when compared to normoxic cells ( Figure 1E-G).

| Silencing of CD36 attenuates hypoxia-induced lipid accumulation in liver cells
To elucidate whether CD36 is involved in hypoxia-induced lipid accumulation in hepatic cells, we infected Huh7 cells with scrambled (control, shC) or CD36 shRNA (shCD36) lentiviral particles. With this approach, we obtained an average of 70% decrease of mRNA and protein levels of CD36 (Figure 2A,B). As depicted in Figure 2F, silencing of CD36 significantly reduced lipid accumulation in Huh7 submitted to hypoxia without altering the induction of hypoxia markers ( Figure 2C,D) and partly blocking hypoxia-induced CD36 increase ( Figure 2E).
These data propose that CD36 might play a major role in the onset of hepatosteatosis under hypoxic conditions.

| HIF2A silencing markedly reduces both lipid accumulation and CD36 upregulation in hypoxic human hepatocytes
We next wanted to determine whether HIF2α might be linked to hypoxia-induced CD36 upregulation. To this end, we infected Huh7 cells with scrambled (control, shC) or HIF2α shRNA (shHIF2) lentiviral particles achieving a 75% decrease in HIF2A mRNA levels ( Figure 3A). Under hypoxia conditions, it was observed a reduced HIF2α protein stabilization ( Figure 3B), as well as PHD3 and EPO mRNA levels, while the hypoxia-induced increased expression of PGK1, a major HIF1α target gene, remained unchanged ( Figure 3C).
Noteworthy, the reduction of HIF2α significantly decreased both lipid accumulation and CD36 upregulation observed in Huh7 cells submitted to hypoxia for 36 hours (Figure 3D,E). Taken together these data suggest that HIF2α is the responsible for the hypoxiainduced CD36 upregulation and, ultimately, for the increase of lipid content in hypoxic hepatocytes.

| Lack of HIF2α ameliorates NASH features and decreases CD36 content in livers from
Vhl f/f -deficient mice to those from Vhl f/f -deficient mice, suggesting that the nuclear signal observed correspond to endogenous HIF2α ( Figure 5B). Indeed, we also analysed HIF2α protein content in nuclear extracts by western blot, and we found similar results which indicate that hepatic HIF2α translocation into the nucleus was nearly blocked in Vhl f/ f Hif2α f/f -deficient mice ( Figure S1A).
In parallel, an increase in CD36 protein expression was observed in the livers of Vhl f/f -deficient mice detected by western blot ( Figure 5A) and by immunostaining ( Figure 5C). In these animals, the intensification of CD36 immunostaining was also enhanced in the plasma membrane of hepatocytes. Indeed, a co-localization between CD36 and N-cadherin, a well-characterized hepatocyte plasma membrane marker, 29 was observed ( Figure S1B). Noteworthy, hepatic CD36 expression was also significantly lower in Vhl f/f Hif2α f/f -deficient mice than in Vhl f/f -deficient mice ( Figure 5A,C).
Taken together, these findings strongly suggest that HIF2α could play an important role on hepatic lipid homeostasis by regulating

| Expression of HIF2α and CD36 is increased within the liver of NAFLD patients
Finally, we wanted to explore whether this link between HIF2α and CD36 exists in human liver as well. Representative haematoxylin/eosin staining liver pictures and the mean of the NAFLD activity score from the study patients are shown in Figure 6A,B.
Furthermore, we estimated the hepatic protein content of nuclear HIF2α and total CD36 assessing their expression by immunohistochemistry. A higher expression of HIF2α was observed in NAFLD patients, such in NAFL as in NASH cases, than in NL individuals ( Figure 6C). Interestingly, HIF2α immunostaining was also more intense in the nucleus of hepatocytes in NAFLD patients, which is in line with nuclear HIF2 expression upon Vhl inactivation in mouse liver shown in Figure 5B. Moreover in agreement with previous results reported by our group, 23 CD36 was weakly expressed in liver biopsies from NL subjects, while is markedly expressed at the plasma membrane and cytoplasm of numerous hepatocytes in NAFL and NASH patients ( Figure 6D). Further immunofluorescence staining revealed that CD36 co-localized with N-cadherin in NAFLD human samples ( Figure S1C), which was also found in Vhl f/f -deficient mice ( Figure S1B). Finally, we measured hepatic mRNA levels of CD36 and EPO, being the latter one of the best-characterized HIF2αdependent gene targets. Therefore, we used EPO mRNA content as a surrogate marker of HIF2α activation. We found that both CD36 and EPO mRNA expression was elevated in NAFLD patients and, interestingly, a significant positive correlation between the mRNA levels of these genes was observed in the entire study population ( Figure 6E,F). There is a growing evidence indicating that hypoxia contributes to NAFLD development and progression to NASH likely by the effects that HIFs exert on target genes regulating glucose and lipid homeostasis in the adipose tissue, small intestine and liver. [30][31][32] Regarding the latter, HIF2α-dependent effects in the liver appear to be linked to its activation level, thus mild HIF2α activation will enhance insulin signalling and fatty acid oxidation whereas potent HIF2α activation will lead to liver dysfunction and steatosis. 33  To the best of our knowledge, this is the first study demonstrating that hypoxia upregulates CD36 expression and function in hepatocytes, and that CD36 gene knockdown markedly reduces the increased FFA uptake in these hypoxic hepatocytes. In addition, another novel finding of this study is the concomitant increase in HIF2α and CD36 protein content in the liver of NAFLD patients. In particular, noteworthy, such in NAFLD patients as in Vhl f/f -deficient mice, the increase in HIF2α expression was markedly observed in the nuclei of hepatocytes and that of CD36 was also detected at the plasma membrane of hepatocytes. More interestingly, we observed that histological features of NASH significantly ameliorated

| D ISCUSS I ON
in Vhl f/f Hif2α f/f -deficient mice along with a marked decrease of both mRNA and protein hepatic CD36 levels, supporting our assumption that HIF2α may contribute to NAFLD onset by upregulating CD36 expression and function in hepatocytes.
The subcellular distribution of CD36 is critical for the regulation of its functional activity as FFA transporter facilitating the uptake and influx of FFA to the cells, remaining functionally inactive at intracellular storage pools and active when translocated to the plasma membrane. 34 In line with our present findings and reinforcing others we previously reported, 23 Zhao et al 35 have recently confirmed that CD36 is largely located in the plasma membrane of hepatocytes such in NASH patients as in mice with histological features of NASH, providing further evidence indicating the key role of palmitoylation in regulating CD36 translocation to the plasma membrane of hepatocytes. Protein palmitoylation is mediated by a family of palmitoyl acyltransferases (PAT). 36 In humans, 23 genes encoding PAT have been described so far, whose expression and function could be regulated from transcriptional to post-translational level. 37 It is tempting to speculate that HIF2α might induce CD36 expression and its translocation to the plasma membrane of human hepatocytes by upregulating PAT gene expression, thus increasing CD36 palmitoylation which facilitates incorporation of CD36 into plasma membranes, but this hypothesis still remains to be addressed.
In conclusion, this study provides novel evidence indicating that within the liver and its detrimental effects on the outcome of NAFLD.

CO N FLI C T O F I NTE R E S T S TATE M E NT
Authors have not conflict of interest.

ACK N OWLED G EM ENTS
The authors thank Esther Fuertes Yebra for helpful technical assistance.