HIF-2α upregulation mediated by hypoxia promotes NAFLD-HCC progression by activating lipid synthesis via the PI3K-AKT-mTOR pathway

HIF-2α, but not HIF-1α, was upregulated in NAFLD-HCC patients

To investigate whether HIFs are involved in NAFLD-HCC, experiments were performed with human HCC tissues. The mRNA levels of HIF-2α (Figure 1B), but not HIF-1α (Figure 1A), were significantly higher in NAFLD-HCC patient tissues than in other HCC tissues. We further analysed HIF-2α protein expression in other HCC and NAFLD-HCC patient tissues by Western blotting (Figure 1C). These results indicated that the mRNA and protein expression levels of HIF-2α were significantly upregulated in NAFLD-HCC patient tissues. The IHC results (Figure 1D) show the same trends. A representative picture showing upregulated HIF-2α mRNA and protein levels in NALFD-HCC patients suggest that HIF-2α may play a role in NAFLD-HCC.

Figure 1. HIF-2α was upregulated in NAFLD-HCC patients and correlated with poor survival. (A, B) Quantitative RT-PCR assessment of HIF-1α and HIF-2α expression in tissues from NAFLD-HCC patients (n=7) and other HCC patients (n=21). Transcription levels were normalized to those of β-actin. (C). Western blot analysis of HIF-2α expression in the patients’ tissues. β-Actin was used as the loading control. Densitometric analyses of the band intensity ratios for HIF-2α/β-actin. (D) IHC determination of HIF-2α expression in tissues from HCC and NAFLD-HCC patients (magnification: 200X, 400X). H score between NAFLD-HCC and HCC (E, F) Survival curves of 139 patients with HCC and NAFLD-HCC. (G, H) Survival curves of 139 patients stratified according to HIF-2α protein expression.

HIF-2α expression levels in HCC were correlated with clinical progression and poor survival

IHC was used to study the expression of HIF-2α in 139 paraffin-embedded HCC tissues. The cut-off value was set as previously described. The number of cases with high HIF-2α expression was 67, and the remaining cases were deemed low HIF-2α expression. Moreover, to determine the possible relationship between HIF-2α expression and clinicopathologic features in our cohort, we analysed the clinical data by Pearson Chi-squared test (Table 1). As shown in Table 1, high HIF-2α expression was correlated with tumour size (P=0.001), tumour number (P=0.021) and alanine aminotransferase (P<0.001). The correlation between HIF-2α expression and clinical factors was further investigated with univariate and multivariate analyses. As shown in Table 2, high HIF-2α levels were significantly associated with tumour size (95%CI 1.611-8.178), tumour number (95%CI 1.289-22.174), alanine aminotransferase (95%CI 2.775-15.409) and age (95%CI 1.203-6.068).

We investigated the survival of other HCC and NAFLD-HCC patients (Figure 1E, 1F). NAFLD-HCC patients have a worse survival than other HCC patients, which is consistent with a previous study [22]. We further explored the possible correlation between HIF-2α level and patient survival. We used Kaplan-Meier survival analysis and discovered that patients with increased HIF-2α expression experienced worse outcomes than patients exhibiting low HIF-2α expression (Figure 1G, 1H). The correlation between HIF-2α overexpression and clinical survival in NAFLD-HCC was shown in Supplementary Figure 1D, 1F. The clinicopathological data and survival curves suggest that HIF-2α plays a role in HCC clinical progression and poor survival.

Hypoxic microenvironment induced lipid accumulation in steatotic HCC cells by upregulating HIF-2α

Because a hypoxic microenvironment is common in tumour tissues, we utilized a hypoxia chamber to mimic this microenvironment. To investigate the role of HIF-2α in steatotic HCC, we transferred HCC and steatotic HCC cells to a hypoxia chamber at different times. The mRNA and protein levels of HIF-2α are displayed in Figure 2A–2D. Compared with standard cell culture (control group), cells grown in a hypoxic microenvironment showed increased HIF-2α expression. Moreover, steatotic HCC presented more lipid accumulation in the hypoxic environment, and HCC showed a similar tendency (Figure 2F–2G). We also detected the triglyceride level with Oil red O straining, which followed the same trend (Figure 2E). These results suggest that hypoxia modulated HIF-2α upregulation and promoted lipid accumulation in steatotic HCC.

Figure 2. The hypoxic microenvironment induces lipid accumulation in HCC and steatotic HCC cells by upregulating HIF-2α. (A, B) Quantitative RT-PCR assessment of HIF-2α expression in HCC and steatotic HCC cells under hypoxic conditions. Transcription levels were normalized to those of β-actin. (C) Western blot analysis of HIF-2α expression in HCC and steatotic HCC cells under hypoxic conditions. β-Actin was used as the loading control. (D) Densitometric analyses of the band intensity ratios for HIF-2α/β-actin. (E) Oil red O staining and quantification in HCC and steatotic HCC cells with or without hypoxia treatment. (F, G) Triglyceride levels in HCC and steatotic HCC cells subjected to different durations of hypoxia.

Hypoxia-induced HIF-2α expression activated mTOR signalling pathways

To study the underlying mechanism, we downloaded the list of genes correlated with HIF-2α in HCC from The Cancer Genome Atlas (TCGA) database (http://gepia.cancer-pku.cn) (Supplementary Table 1). These data suggested that HIF-2α is associated with PI3K, AKT and mTOR (Figure 3A).

Figure 3. Hypoxia-mediated upregulation of HIF-2α expression activates mTOR signalling pathways. (A) Pearson correlation analysis of the correlation between HIF-2α expression and relative genes in the patients with liver hepatocellular carcinoma (LIHC) from the TCGA database (http://gepia.cancer-pku.cn). (B) Western blot analysis of HIF-2α expression and expression of PI3K-AKT-mTOR pathway factors in HCC and steatotic HCC under hypoxic conditions. β-Actin was used as the loading control. Densitometric analyses of the band intensity ratios for HIF-2α/β-actin, PI3K/β-actin, p-GSK3β/GSK3β, p-AKT/AKT and p-mTOR/mTOR.

Because the PI3K-AKT-mTOR pathway has been shown to control tumour development and lipid synthesis, we tested whether HIF-2α affects PI3K-AKT-mTOR signalling in steatotic HCC cells. We also detected the expression of PI3K, AKT, mTOR, GSK3B and the phosphorylation level of AKT at Ser473, mTOR at S2448, and GSK3B at Ser9 in steatotic HCC cells (Figure 3B). When cells were treated with MHY1485, an mTOR selective agonist, p-mTOR was upregulated, proving its effectiveness. Under normoxic conditions, HIF-2α expression was also elevated in steatotic HCC compared with that in HCC. We found that, in hypoxic conditions, HIF-2α overexpression is accompanied by increases in p-AKT, p-GSK3B and p-mTOR. After transfection with sh-HIF2α, the levels of PI3K, p-AKT, p-mTOR and p-GSK3B decreased significantly. This result may indicate that hypoxia upregulates HIF-2α expression, activating PI3K-AKT-mTOR signalling.

Steatotic HCC growth, migration, invasion and angiogenesis can be ameliorated by intervention of the HIF-2α/PI3K-AKT-mTOR signalling axis

To further investigate the effect of HIF-2α on tumour biological function, we inhibited HIF-2α expression by transfecting cells with sh-HIF2α. Western blotting and quantitative RT-PCR determined that sh-HIF2α effectively downregulated HIF-2α expression (Supplementary Figure 1B, 1C).

For 2D growth assays, a live real-time IncuCyte was utilized. In response to hypoxia treatment, HIF-2α-KD cells exhibited less survival in low O₂ concentrations and presented a relatively lower proliferation rate than did the scramble group (Figure 4A). After 24 h of hypoxia treatment, the cells were transferred to a normoxia environment. HIF-2α knockdown also suppressed the growth of HCC under normal oxygen conditions after hypoxia treatment (Figure 4B). With regard to the overall growth curve, HIF-2α-KD cells had less growth than cells transfected with a control lentivirus (Figure 4C). The same trend could be found in steatotic HCC (Figure 4D–4F). For the transwell assay, HIF-2α-KD cells reduced tumour cell migration and invasion (Figure 4G). The in vitro tube formation assay showed more vessels in the scrambled group than in the HIF-2α-KD group (Figure 4H, P<0.005). We also activated p-mTOR with MHY1485, a mTOR-selective agonist, to study the role of the mTOR pathway in this process. In the HIF-2α-KD group, the effect of sh-HIF2α was reversed after MHY1485 treatment. Moreover, steatotic HCC showed the same tendency. Transwell assays and tube formation were all performed in a hypoxic environment. The results of the functional assays suggest that HIF-2α promotes growth, migration, invasion and angiogenesis of HCC and steatotic HCC under hypoxic conditions, and these effects can be rescued by MHY1485. Furthermore, HIF-2α influenced the biological function of HCC via the PI3K-AKT-mTOR pathway.

Figure 4. Steatotic HCC growth, migration, invasion and angiogenesis in a hypoxic microenvironment needs HIF-2α. (A–C) Summary graph showing the HCC cell 2D growth rate in hypoxic conditions (A), after 24 hypoxia treatments (B) and all conditions (C). (D–F) Summary graph showing the steatotic HCC cell 2D growth rate under hypoxic conditions (D) after 24 hypoxia treatments (E) and all conditions (F, G). Transwell assays of HCC, HIF-2α-KD HCC, steatotic HCC and HIF-2α-KD steatotic HCC cells under hypoxic conditions in the presence or absence of MHY1485 treatment. h. Representative images of HUVEC tube formation; HUVECs were treated with the supernatant of HCC, HIF-2α-KD HCC, steatotic HCC and HIF-2α-KD steatotic HCC cell cultures.

HIF-2α induced lipid synthesis and exacerbated lipid accumulation in steatotic HCC cells under a hypoxic microenvironment

As HIF-2α can regulate cell biological function under hypoxic conditions and mTOR complexity may modulate cell metabolism, we further studied the effect of HIF-2α on lipid synthesis.

Upon hypoxia treatment, HIF-2α-KD cells exhibited ameliorated triglyceride accumulation and steatosis compared to cells in the scramble group, and the effect was reversed by MHY1485 (Figure 5A–5C). To assess the link between HIF-2α upregulation and lipid synthesis, we employed Western blotting to evaluate the level of lipid synthesis-related proteins, such as FASN and ACC1 (Figure 5D). The results demonstrated that increases in HIF-2α expression are accompanied with the upregulation of lipid synthesis-related protein in response to hypoxia. Moreover, cells transfected with sh-HIF2α showed dramatically decreased FASN and ACC1 expression. When cells were treated with MHY1485, FASN and ACC1 was rescued and elevated to the previous endogenous levels. These results suggest that HIF-2α promotes lipid synthesis in steatotic HCC under hypoxic conditions.

Figure 5. HIF-2α induces steatotic HCC lipid synthesis under a hypoxic microenvironment (A) Oil red O staining and quantification in HCC and steatotic HCC under hypoxic conditions in the presence or absence of MHY1485. (B, C) Triglyceride levels in HCC and steatotic HCC under hypoxic conditions in the presence or absence of MHY1485. (D) Western blot analysis of lipid synthesis-related proteins in HCC and steatotic HCC under hypoxic conditions. Densitometric analyses of the band intensity ratios for ACC1/β-actin and FASN/β-actin.

Hypoxia-induced HIF-2α overexpression promoted tumour development and lipid accumulation in STAM mice in vivo, which can be rescued by an mTOR inhibitor

STAM mice, a mouse model characterized by its resemblance to consistent progression to NALFD-HCC in humans, were treated to induce diabetes and fed a high-fat diet. To investigate the long-term effect of hypoxia-induced HIF-2α, we mimicked the hypoxia environment in the cages and treated mice with INK-128.

When exposed to a hypoxic environment, STAM mice developed 2.9±1.9 surface tumours compared to 0.8±1.0 tumours in the untreated STAM group (p=0.005) (Figure 6C). After INK-128 treatment during the hypoxia intervention, the number of surface tumours decreased to 1.1±1.3 compared with the hypoxia group (p=0.015) (Figure 6C). Moreover, after PT-2385 treatment, STAM mice developed less tumours (1.4±1.2) compared with the hypoxia group. Similar to hypoxia treatment, tumour sizes were significantly increased in hypoxic STAM mice compared with untreated STAM mice (p=0.005) (Figure 6D). Tumours in the INK-128-treated group were also significantly smaller (p=0.011) (Figure 6D). Tumour sizes were decreased in PT-2385 group compared with Hypoxia STAM mice. As shown in Figure 6B, hypoxia STAM mice represent more tumour cell atypia and lipid accumulation than untreated STAM mice. HE staining display pathological changes in different mice group. The steatotic cells and the tumors happened in mice. Moreover, these behaviours can be rescued by INK-128 and PT-2385.

Figure 6. Hypoxia-induced HIF-2α overexpression promotes STAM mouse tumour development and lipid accumulation in vivo, which can be rescued by an mTOR inhibitor. (A). Gross specimens from STAM mice of different groups. (B) Images of liver stained by HE and Oil red O staining. (C) The numbers of tumours on the liver surface of different STAM mice (n = 10 mice/group). d. In the tumour size analysis, the largest three tumours found at a random location in the liver were used.

Hypoxia-induced HIF-2α upregulation altered NAFLD-HCC lipid accumulation via the PI3K-mTOR pathway in vivo

To determine the underlying mechanism of how HIF-2α manipulates NAFLD-HCC lipid accumulation, we detected HIF-2α, PI3K and mTOR expression by Western blotting. After hypoxia intervention, HIF-2α was upregulated, accompanied by PI3K and mTOR elevation (Figure 7A). Moreover, with INK-128 treatment, STAM mice suppressed mTOR expression and lipid accumulation, which contribute to restraining the progression of NAFLD-HCC.

Figure 7. Hypoxia-induced HIF-2α overexpression promotes STAM mouse tumour development via activating the PI3K-AKT-mTOR signalling axis. (A) Western blot analysis of HIF-2α expression and expression of PI3K-mTOR pathway factors in STAM mice of different groups. β-Actin was used as the loading control. Densitometric analyses of the band intensity ratios for HIF-2α/β-actin, PI3K/β-actin and mTOR/β-actin. (B, C) The liver sections were stained with DAPI to visualise nuclei (blue), Alexa Fluor 488 to visualise the distribution of HIF-2α (green) and Alexa Fluor 568 to PI3K/mTOR (red) proteins. Scale bar: 10 μm.

We further investigated HIF-2α, PI3K, mTOR and Akt expression and distribution in STAM mouse tumour frozen sections. HIF-2α expression was significantly elevated with hypoxia intervention and shown to colocalize with PI3K by confocal microscopy (Figure 7B). With INK-128 treatment, mTOR expression was suppressed (Figure 7C). These results suggested that hypoxia-induced HIF-2α upregulation participated in altering NAFLD-HCC lipid accumulation via the PI3K-mTOR pathway. Moreover, we also explore de novo fatty acid synthesis level in mice liver tissues (Figure 8A). With INK-128 or PT-2385 treatment, de novo fatty acid synthesis was suppressed significantly. Therefore, HIF-2α could be a biomarker and target in developing specific therapeutic measures for NAFLD-HCC patients. INK-128 (mTOR inhibitor) and PT-2385 (HIF-2α inhibitor) have a definite role in suppressing NAFLD-HCC progression.

Figure 8. HIF-2α induced lipid synthesis and exacerbated lipid accumulation in STAM mice in vivo. (A), Western blot analysis of FASN and ACC1 expression in STAM mice of different groups. β-actin was used as the loading control. Densitometric analyses of the band intensity ratios for FASN/β-actin and ACC1/β-actin.

HCC specimens from patients

Paraffin-embedded specimens from 139 patients diagnosed with HCC who underwent hepatectomy at the Fifth Affiliated Hospital, Sun Yat-Sen University, China between 2009 and 2015 were investigated. All patient diagnoses were histologically and clinically confirmed. NAFLD-HCC was further proven by an experienced pathologist who evaluated the biopsy samples. Fresh HCC tissues (n=21) and NAFLD-HCC tissues (n=7) were received from the same hospital and stored in liquid nitrogen for further use. The Correlations between the clinicopathologic variables and HIF-2α expression in HCC are presented in Table 1. For the sake of using these patients’ specimens for experimental purposes, all specimens were obtained with informed consent, and the study was approved by the Institutional Ethical Review Board of the Fifth Affiliated Hospital of Sun Yat-Sen University. Moreover, all patients approved the use of their specimens in the following experiments and the analysis of their clinicopathological data.

Cell lines and cell culture

The human HCC cell line HepG2 was purchased from the Chinese Academy of Sciences, Shanghai Branch (Shanghai, China). HCC cells were cultured in Dulbecco’s modified Eagle’s medium (DMEM; Gibco) supplemented with 10% foetal bovine serum (FBS; Gibco), 50 U/mol penicillin and 50 mg/ml streptomycin (Sigma-Aldrich, USA) and maintained at 37°C with 5% CO₂.

To investigate SH-HCC, HepG2 cells were treated with free fatty acids (FFAs) (oleate/palmitate, 2:1 ratio); oleate and palmitate (Sigma-Aldrich) were diluted to different concentrations in DMEM containing 10% FBS [31–34]. Cells were serum-starved for 24 h and then treated with different concentrations of FFAs. We chose 1 mmol/L FFAs as the best concentration to establish SH-HCC (Supplementary Figure 1A). To mimic the hypoxia microenvironment, cells were transferred to a hypoxia chamber containing 1% O₂ [46]. MHY1485, a selective agonist, was utilized to activate the mTOR complex according to the manufacturer’s suggestion.

Establishment of knockdown HCC cells

A lentiviral-mediated shRNA approach was used to generate knockdown HCC cells. Lentiviral particles expressing either shRNA targeting HIF-2α or a control sequence were constructed by Shanghai Genechem. Lentiviral vectors were used to infect HepG2 cells. After infection, cells were selected with puromycin (Sigma). HIF-2α expression under hypoxic conditions was confirmed by qPCR and Western blotting (Supplementary Figure 1B, 1C).

Animals and animal models

Pregnant female C57BL/6J mice (180-250 days old) were purchased from the Laboratory Animal Centre of Guangzhou University of Chinese Medicine (Guangzhou, China) and used to establish a model of STAM in which mice are treated to induce diabetes and fed a high-fat diet [47, 48]. In brief, pups birthed within 5 days were injected with 200 μg of streptozotocin (Sigma-Aldrich) and fed with a diet with 60% fat content (D12492; Research Diets). To mimic hypoxia, mice were exposed to 60-s IH cycles (30 s at 5% fraction of inspired oxygen and 30 s at 21% FIO2) for 8 h in their cages between 6 A.M. and 2 P.M. as described in a previous study [45]. Animal experimental protocols were approved by the Animal Ethics Committee of Forevergen Company and performed in the line with the National Institutes of Health guidelines.

RNA extraction and quantitation and quantitative RT-PCR

Total RNA was extracted from the cultured cells and freshly resected HCC tissues using TRIzol (Life Technologies) according to the manufacturer’s instructions. cDNA was synthesized from total RNA using a RevertAid RT Reverse Transcription kit (K1691). To explore the relative expression levels of different genes, SYBR Green PCR kits (TAKARA) were utilized with the β-actin gene as an internal reference. PCR was performed using a T100™ Thermal Cycler (Bio-Rad) instrument. To ensure the repeatability and authenticity of the results, the experiments were performed in triplicate. Sequences of primers are as follows: β-actin: forward 5′CATCTG CTGGAAGGTGGACA3′, reverse 5′CGACAGGATG CAGAAGGAGA3′; HIF-2α: forward 5′GAGCAAGGA GACGGAGGTGT3′, reverse 5′TGTGTTCGCAGGA AGCTGAT3′.

Western blotting and immunohistochemistry (IHC)

Protein from cells and tissues was extracted according to a protocol recommended by the manufacturer (P00013C, Beyotime), and the protein concentration was further determined with a BCA kit (Beyotime). For separation, 20 μg of protein was loaded. Primary antibodies targeting HIF-2α (7096), PI3K (4249), GSK-3B (4337), p-GSK3B (5558), AKT (9272), p-AKT (4060), mTOR (2972), p-mTOR (2971), FASN (3180), and ACC1 (4190) were purchased from Cell Signalling Technology. β-Actin (Abcam; ab6276; 1:5000) was used as a loading control. Quantitative analysis of the Western blot data was done using Image J software.

For the IHC assay, the tissues were incubated with HIF-2α at a 1:300 dilution at 4°C for 8 h. For negative controls, the primary antibody was replaced by normal goat serum. Two independent and experienced pathologists who were blinded to the biochemical and clinical patient information scored the immunohistochemically stained tissue. The scoring method was performed as described [49, 50]. For the purposes of quantification and comparison, specimens were assigned an H-score, which multiplies the score corresponding to the percentage of tumour cells stained by the score corresponding to the strength of staining (from 0 for negative staining to 3 for positive staining). The median value in 139 patients served as a cut-off to categorize all cases as high or low HIF-2α expression.

Immunofluorescence

Immunofluorescence was utilized to study the distribution and expression levels of HIF-2α, PI3K and mTOR in HCC tissues from STAM mice from three groups. We blocked tissues with 5% bovine serum albumin for 60 minutes and incubated them with primary antibodies against HIF-2α (1;200, Cell Signalling Technology), PI3K (1:200, Cell Signalling Technology) or mTOR (1:200, Cell Signalling Technology) overnight at 4°C. PBS was used to rinse the sections 3 times, which were then incubated with Alexa Fluor®568-conjugated goat anti-mouse IgG(H+L) (1:500, Thermo) or Alexa Fluor® 488-conjugated goat anti-rabbit IgG(H+L) (1:500, Thermo) antibodies for one hour at room temperature. DAPI was used to stain cell nuclei. The sections were visualized on a laser confocal microscope (Zeiss).

Metabolism characteristic assay

A triglyceride assay kit was purchased from Jiancheng Biology Institution PeproTech (Nanjing, Jiangsu, China). Oil red O staining was employed according to a protocol recommended by the manufacturer (Solarbio, G1262). The experiments were performed at least in triplicate.

Cell proliferation and transwell assay

To measure cell proliferation, a live real-time IncuCyte ZOOM was used as described [51, 52]. For the migration and invasion studies, HepG2 cells or steatotic HepG2 cells were suspended in 100 μl of serum-free medium and placed in a transwell chamber with or without Matrigel. After the cells were incubated at 37°C for 24 h, those attached to the lower surface of the chamber were fixed, stained with crystal violet and counted at a magnification of × 200 in eight randomly selected fields. The experiments were performed at least in triplicate.

Tube formation assay

Growth factor-reduced Matrigel (BD Biosciences) was thawed at 4°C. After the Matrigel changed to a liquid state, 50 μl per well was added to a 96-well plate for polymerization. Human umbilical vein endothelial cells (HUVECs) (2 × 10⁴/well) were seeded on top of the Matrigel and cultured with the supernatants from HepG2 or steatotic HepG2 cells for 24 h. Meanwhile, we set a program on the IncuCyte ZOOM to record an image every hour. The cumulative number of tubes in every well in 5 random microscopic fields was calculated by ImageJ software. The experiments were performed at least in triplicate.

Statistical analysis

Statistical data were analysed by SPSS version 13.0 (Chicago, IL. USA). The correlations between HIF-2α levels and clinical characteristics were evaluated using the Pearson Chi-squared test. Overall survival (OS) curves were plotted by Kaplan-Meier survival analysis and log-rank test. Moreover, predictive variables, such as tumour size, were estimated by univariate and multivariate analyses. Student’s t-test and ANOVA were used to evaluate the differences among groups. We considered differences statistically significant when the p values were less than 0.05.