Hypoxia-inducible factor 2α drives nonalcoholic fatty liver progression by triggering hepatocyte release of histidine-rich glycoprotein

HUMAN SUBJECTS

The study on NAFLD patients was approved by the ethics committee of the Azienda Ospedaliera Universitaria Città della Salute (Turin, Italy). For this study we analyzed liver biopsies from NAFLD patients (n = 27) ranging from early disease (stage F0-F1) to more advanced conditions of fibrosis (stage F2-F3) or cirrhosis (F4) and referring to the Division of Gastro-Hepatology of the University of Turin. All samples were collected at the time of first diagnosis. Patients were characterized by anthropometric, clinical, and biochemical data; and liver biopsies were evaluated for the severity of steatohepatitis and fibrosis according to Kleiner et al.25 All subjects gave informed consent to the analysis, and the study protocol, which conformed to the ethical guidelines of the 1975 Declaration of Helsinki, was planned according to the guidelines of the local ethics committee. The clinical and biochemical features of the patients are reported in Supporting Table S1. Control human liver tissue (n = 5), as defined for normal histological structures in hematoxylin and eosin sections and the absence of inflammation in the portal tract and parenchyma, was obtained from diagnostic biopsies or from resection samples taken at a distance of more than 5 cm from the border of liver metastasis of colon carcinoma.

ANIMAL EXPERIMENTATION

Mice carrying a hepatocyte-specific deletion of HIF-2α (hHIF-2α^–/–) were obtained by breeding HIF-2α^fl/fl C57BL/6 mice with mice on the same genetic background expressing Cre-recombinase under the control of the albumin promoter (Alb/Cre^+/+ mice; Jackson Laboratories, Bar Harbor, ME). Eight-week-old male hHIF-2α^–/– mice and related control sibling littermates not carrying the HIF-2α deletion were fed either a methionine/choline deficient (MCD) diet for 4 or 8 weeks or a choline-deficient L-amino acid–defined (CDAA) diet (Laboratorio Dottori Piccioni, Gessate, Italy) for 12 and 24 weeks.26, 27 Control littermates received the corresponding methionine/choline-sufficient or the choline-sufficient L-amino acid–defined (CSAA) diet. In preliminary experiments 8-week-old normal C57BL/6 mice were fed the MCD diet or the methionine/choline-sufficient control diet for 4 and 8 weeks. The experiments complied with national ethical guidelines for animal experimentation, and the experimental protocols were approved by the Italian Ministry of Health.

BIOCHEMICAL ANALYSES

Plasma alanine aminotransferase (ALT) and liver triglycerides were determined by spectrometric kits supplied by Radim S.p.A. (Pomezia, Italy) and Sigma Diagnostics (Milan, Italy), respectively. Circulating interleukin 12 (IL-12) was evaluated by commercial enzyme-linked immunosorbent assay kits supplied by Peprotech (Milan, Italy).

IMMUNOHISTOCHEMISTRY, SIRIUS RED STAINING, AND HISTOMORPHOMETRIC ANALYSIS

Immunohistochemistry (IHC),28, 29 picro-sirius red staining, and histomorphometric analysis30, 31 were performed on paraffin-embedded human liver biopsies and/or murine liver specimens as reported. More details are in the Supporting Information.

QUANTITATIVE REAL-TIME PCR

RNA extraction, complementary DNA synthesis, and quantitative real-time PCRs were performed on human and murine liver specimens as described.28-30 mRNA levels were measured by quantitative real-time PCR, using the SYBR green method as described.28-30 More details and oligonucleotide sequences of primers used for quantitative real-time PCR are available in the Supporting Information. The murine transcripts for tumour necrosis factor-α (TNFα), cluster of differentiation 11b (CD11b), IL-12p40, C-C-motif chemokine ligand 2 (CCL2), C-X-C-motif chemokine ligand 10, HRGP, and β-actin liver RNA were retrotranscribed with the High Capacity cDNA Reverse Transcription Kit (Applied Biosystems Italia, Monza, Italy). Quantitative real-time PCR for these transcripts was performed in a Techne TC-312 thermacycler (TecneInc, Burlington, NJ) using TaqMan Gene Expression Master Mix and TaqMan Gene Expression probes for indicated murine genes (Applied Biosystems Italia) as described.26, 31 Human sample analysis was performed using SsoFast EvaGreen Supermix (Bio-Rad, Hercules, CA) following the manufacturer's instructions.

INTRAHEPATIC MONONUCLEATED CELL ISOLATION AND FLOW-CYTOMETRIC ANALYSIS

Hepatic mononucleated cells were isolated from the livers of naive and CDAA–fed mice and purified on a density gradient (Lympholyte-M; Cedarlane Laboratoires Ltd., Burlington, Canada) as described.31 Cells were washed with Hank's medium and incubated for 30 min with decomplemented mouse serum to block unspecific immunoglobulin binding. The cells were then stained with fluorochrome-conjugated antibodies for CD45, CD11b, Ly6C (eBiosciences, San Diego, CA), and F4-80 (Invitrogen, Abingdon, UK) and analyzed with a Attune NxT acoustic focusing cytometer (Thermo Fischer Scientific, Waltham, MA) following prior gating for CD45 and the absence of cell aggregates. Intracellular staining for IL-12 was performed using a specific fluorochrome-conjugated antibody (eBiosciences).

IN VITRO EXPERIMENTS AND WESTERN BLOT ANALYSIS

In vitro experiments in this study were performed in normal HepG2 cells (American Type Culture Collection) as well as in cells stably transfected in order to overexpress HIF-2α in either normoxic or hypoxic conditions, as reported.28 Western blot analysis was performed on total cell or tissue lysates as described.28, 30 More details on in vitro studies and western blot analysis are available in the Supporting Information.

DATA ANALYSIS AND STATISTICAL CALCULATIONS

Statistical analyses were performed by SPSS statistical software (SPSS Inc., Chicago, IL) using one-way analysis of variance with Tukey's correction for multiple comparisons or the Kruskal-Wallis test for nonparametric values. Significance was taken at the 5% level. Normality distribution was preliminarily assessed by the Kolmogorov-Smirnov algorithm.

HIF-2α IS OVEREXPRESSED IN HUMAN AND MURINE HEPATOCYTES IN NAFLD-RELATED LIVER SPECIMENS

IHC analysis of liver biopsies from a cohort of well-characterized NAFLD patients (n = 27, see Supporting Table S1 for patient characterization) at different stages of the disease evolution showed an evident increase of HIF-2α immune-staining compared to undamaged liver samples (n = 5) from patients undergoing resection for hepatic metastasis of colon carcinoma. In the majority of NAFLD patients (25 out of 27, 92%) HIF-2α positivity was detectable almost exclusively in hepatocytes (Fig. 1A,B) and involved both the cytoplasm and the nuclei, although the number of positive nuclei varied within patients and in different areas of the same biopsy (Fig. 1A). The nuclei of nonparenchymal cells, mainly inflammatory cells or myofibroblast-like cells in fibrotic septa, were essentially negative for HIF-2α. Occasionally, HIF-2α staining was observed in portal/periportal cholangiocytes, particularly in subjects with advanced fibrosis. Semiquantitative assessment of HIF-2α-positive hepatocyte nuclei confirmed a significant increase among NAFLD patients irrespective of the stage of the disease (Fig. 1B). These findings were reproduced in a rodent model of NAFLD-related liver fibrosis based on mice fed the MCD diet up to 8 weeks (Supporting Fig. S1A,B). In fact, while in control mice HIF-2α staining was only detected in the cytoplasm of hepatocytes closely surrounding centrilobular veins (Supporting Fig. S1A), extensive HIF-2α positivity was evident in liver sections from mice receiving the MCD diet (Supporting Fig. S1B). In these animals quantitative real-time PCR analyses confirmed that both HIF-2α as well HIF-1α transcripts were up-regulated in mice with NAFLD, but only those of HIF-2α paralleled the disease progression (Supporting Fig. S1C-F).

HEPATOCYTE-SPECIFIC DELETION OF HIF-2α IMPROVES NAFLD-ASSOCIATED LIVER INJURY, STEATOSIS, INFLAMMATORY RESPONSE, AND FIBROSIS

Based on previous data, we evaluated whether HIF-2α deletion in hepatocytes might modify NAFLD evolution. To this aim, progressive NAFLD was induced in hHIF-2α^–/– mice as well as related control littermates by two dietary protocols using the MCD diet, administered for 4 and 8 weeks, and the CDAA diet, administered for 12 and 24 weeks. In preliminary experiments, we first checked for effective HIF-2α depletion in the livers of hHIF-2α^–/– mice receiving the MCD diet. Quantitative real-time PCR analysis revealed a very significant down-regulation of liver HIF-2α transcripts in hHIF-2α^–/– mice compared to control littermates (Supporting Fig. S2A-C), whereas HIF-1α and vascular-endothelial growth factor transcript levels were not affected. HIF-2α deletion was also confirmed at the protein level by western blotting of total liver proteins (Supporting Fig. S2D).

Interestingly, in hHIF-2α^–/– mice receiving the MCD diet for 4 weeks the severity of steatohepatitis as evaluated by histology, ALT release, and hepatic triglyceride accumulation was significantly lower than in HIF-2α-sufficient animals (Fig. 2A-C). The extent of lobular inflammation and the hepatic expression of inflammatory markers TNFα, CD11b, and CCL2 were also decreased in hHIF-2α^–/– mice receiving the MCD diet (Fig. 2D). On the same line, using CDAA-fed mice we observed that the deletion of hepatocyte HIF-2α (which was confirmed also by means of IHC in these mice versus control littermates; see Supporting Fig. S3) slowed down the time-dependent progression of NAFLD (Fig. 3A,B,E) without affecting the gain in body weight and the development of insulin resistance (Fig. 3C,D). In this experimental model, hHIF-2α^–/– mice also showed less liver fibrosis at both 12 and 24 weeks, as evidenced by sirius red staining (Fig. 4A,B). This paralleled a decrease in hepatic transcript for transforming growth factor β1, α-smooth muscle actin (α-SMA), and collagen 1A1 (Fig. 4C), as well as in the number of α-SMA-positive myofibroblasts detected by IHC (Fig. 4D). Remarkably, the protection from fibrosis was also evident in hHIF-2α^–/– mice receiving the MCD diet for up to 8 weeks compared to control HIF-2α-sufficient littermates (Supporting Fig. S4A-D).

HEPATOCYTE-SPECIFIC DELETION OF HIF-2α PREVENTS THE PROGRESSION OF EXPERIMENTAL NAFLD BY AFFECTING HEPATOCYTE-DEPENDENT RELEASE OF HRGP

To address the mechanisms by which hepatocyte HIF-2α up-regulation promotes inflammatory mechanisms in NAFLD, we focused our attention on the role of HRGP, a hepatocyte-released mediator (i.e., hepatokine) that has been recently shown to support liver macrophage activation in experimental and clinical NAFLD/NASH.32 To investigate the possible hypoxia-dependent and HIF-2α-dependent modulation of HRGP, we employed HepG2 cells, which are known to rapidly respond to hypoxia with HIF-2α stabilization and the up-regulation of HIF-2α target genes.28, 33 Incubation of HepG2 cells under hypoxic conditions, which is known to rapidly promote the nuclear translocation of HIF-2α,28 was associated with an appreciable increase in HRGP synthesis (Fig. 5A). A very significant up-regulation of HRGP transcription was also observed in HepG2 cells transfected with an HIF-2α-containing vector but not with an empty vector (Fig. 5B), confirming an HIF-2α-dependent modulation of HRGP in liver cells. According to the in vitro data, hepatic HRGP transcript levels increased in a time-dependent manner in control mice fed the CDAA diet in parallel with the up-regulation of the expression of another HIF-2α-dependent gene such as cysteine-X-cysteine receptor 4 (Fig. 5C). Conversely, HRGP up-regulation associated with NAFLD evolution was almost completely prevented at both the mRNA and protein levels in hHIF-2α^–/– mice (Fig. 5C,D). The relationships between HIF-2α and HRGP were further confirmed by the evidence that HepG2 stably transfected to overexpress HIF-2α also overexpressed transcript levels of other HIF-2α target genes such as cysteine-X-cysteine receptor 4 and erythropoietin (Supporting Fig. S5A,B) or, as previously reported, SerpinB3.28 Moreover, HRGP transcript and protein levels were down-regulated in HepG2 cells exposed to hypoxic conditions following efficient silencing for HIF-2α (Supporting Fig. S5C,D). IHC performed on liver from mice fed the CDAA diet confirmed that HRGP was selectively expressed by hepatocytes of control mice with steatohepatitis but not in those from hHIF-2α^–/– animals (Fig. 5E). Up-regulation of HIF-2α and HRPG was unrelated to choline deficiency and was also observed in mice with NAFLD induced by administration in the drinking water of a high-fat/fructose diet (Supporting Fig. S6A-D). Hepatocyte-derived HRGP has been recently shown to promote liver inflammation by stimulating M1 polarization of hepatic macrophages.32 In our hands, the impaired production of HRPG in mice receiving the CDAA diet decreased the liver infiltration by F4/80-positive macrophages as evidenced by both IHC (Fig. 6A) and flow cytometry (Supporting Fig. S7A). Furthermore, the fraction of inflammatory Ly6C^high hepatic macrophages and their capacity to produce IL-12 were significantly lower in hHIF-2α^–/– mice compared to control littermates (Fig. 6B). Consistently the hepatic expression of M1 cytokines/chemokines TNFα, IL-12p40, CCL2, and C-X-C-motif chemokine ligand 10 (Supporting Fig. S7B); and the serum levels of IL-12 were also significantly reduced in hHIF-2α^–/– mice receiving the CDAA diet (Fig. 6C), indicating a possible role of HRGP in the mechanisms by which the stimulation of HIF-2α in parenchymal cells supports proinflammatory responses of hepatic macrophages.

Finally, to further support the possible relevance of the relationships between HIF-2α and HRGP in the human disease, we analyzed for HIF-2α and HRGP immune-staining serial sections (4 μm thick) in our cohorts of NAFLD patients. Figure 7A shows that HRGP positivity was very low in sections from control livers. Conversely, in agreement with previous data,32 HPRG immune-staining was evident and increased in liver samples from NAFLD patients irrespective of the disease evolution (Fig. 7B,C) and colocalized with that of HIF-2α (Fig. 7A). Furthermore, we found a significant positive correlation between HIF-2α and HRGP mRNA levels in those patients of this cohort (n = 19) for whom frozen liver specimens were available (Fig. 7D).