Potential conflict of interest: Nothing to report.
Hypoxia inducible factors (HIFs) are activated in many tumors and show either promoter or suppressor activity, depending on tumor cell biology and background. However, the role of HIF member HIF-2α remains unclear in hepatocellular carcinoma (HCC). Here, HIF-2α expression was measured in HCC and paired peritumoral tissues by quantitative real-time polymerase chain reaction, western blotting, and immunofluorescence assays, and the clinical significance was explored in 246 HCC patients. In cell culture, HIF-2α levels were up-regulated or down-regulated by use of expression or short hairpin RNA recombinant plasmid, respectively. Cells were analyzed by immunoblotting, chromatin immunoprecipitation coupled with microarray, coimmunoprecipitation, and immunohistochemical staining. In vivo tumor growth was analyzed in nude mice. We found that the average expression of HIF-2α was relatively low in HCC tissues, and the decreased level was associated with lower overall survival (P = 0.006). High HIF-2α expression in HCC cells induced higher levels of apoptosis and expression of proapoptotic proteins and inhibited cell and tumor growth. Furthermore, HIF-2α inhibited expression of the novel target gene, transcription factor dimerization partner 3 (TFDP3). TFDP3 protein was found to bind with E2F transcription factor 1 (E2F1) and inhibit its transcriptional activity through both p53-dependent and -independent pathways. Reintroduction of TFDP3 expression reversed HIF-2α-induced apoptosis. Conclusions: Data gathered from cell lines, tumorigenicity studies, and primary HCC samples demonstrate a negative role of HIF-2α in tumors, which is mediated by the TFDP3/E2F1 pathway. Our study provides evidence supporting a possible tumor-suppressor role for HIF-2α and has uncovered a mechanism that links HIF-2α to a fundamental biological regulator, E2F1. (HEPATOLOGY 2013)
Hepatocellular carcinoma (HCC) is one of the most common, aggressive malignancies and is the third leading cause of cancer-related deaths. The poor prognosis of HCC is mainly because of the rapid progression of this disease.1-3 Advances in treatment of this disease are likely to stem from a better understanding of its biology and behavior. As with most solid tumors, the hypoxic microenvironment exists in HCC as a result of an imbalance between oxygen supply and consumption in proliferating tumors.1 Significant evidences indicate that the hypoxia inducible factors (HIFs) play an important role in the pathogenesis and pathophysiology of HCC. Becasue HIF inhibitors are currently undergoing clinical evaluation as cancer therapeutics, a more-thorough understanding of the unique roles performed by HIFs in human tumor is warranted.4, 5
Cellular responses to low oxygen tension are mainly mediated by the activation of HIFs, which consist of a constitutively expressed subunit (aryl hydrocarbon nuclear translocator) and an oxygen-regulated subunit, mainly HIF-1α and HIF-2α. HIF-1α and HIF-2α promote adaptation of tumor cells to hypoxic stress by regulating the expression of genes involved in metabolism, angiogenesis, erythropoiesis, cell proliferation, and apoptosis.6-9 HIF-1α is found to be expressed at a higher level in dysplastic nodules and implicates a malignant transformation. We and others have previously shown that HIF-1α expression is significantly associated with an advanced stage and aggressive phenotype and indicates a poor prognosis in HCC.2, 3 An approximately 48% amino-acid sequence homology is shared by HIF-1α and HIF-2α10; however, much less is known about the HIF-2α isoform, and there have been inconsistent, and even conflicting, reports about its role in HCC. Some studies conclude that HIF-2α promotes hypoxic cell proliferation by enhancing c-Myc transcriptional activity,11 whereas others argue that knockdown of HIF-2α increases cell viability and growth by autophagy.1 It is known that HIF-2α correlates with vascular endothelial growth factor (VEGF) expression and indicates a poor prognosis in melanoma and non-small-cell lung cancers.12, 13 Other studies show that HIF-2α acts as a tumor suppressor in breast and brain cancer.14, 15
To date, there are limited published reports describing HIF-2α in HCC and the role of this protein is still controversial.1, 16 Here, we provide evidence for a possible tumor suppressor role for HIF-2α in HCC. Chromatin immunoprecipitation coupled with microarray (ChIP-on-chip) and additional assays indicate that transcription factor dimerization partner 3 (TFDP3) is a novel target of HIF-2α and inhibits E2F transcription factor 1 (E2F1) transcriptional activity through multiple pathways. This study has uncovered another mechanism connecting HIF-2α to E2F1 in the induction of apoptosis.
Tissues were collected and tissue microarray analyses were performed as previously described.17 Follow-up procedures are described in the Supporting Materials and Methods.
Cell Culture and Recombinant Plasmid DNA.
HCC cell lines used in this study, construction of short hairpin RNA (shRNA), and expression of HIF-2α recombinant plasmids are described in the Supporting Materials and Methods. The pcNDA3-TFDP3 plasmid was a gift from Yu Zhang's lab at Peking University. To select the stably transformed monoclonal cell lines, selective medium containing neomycin (pcDNA3.1 plasmids) or puromycin (shRNA plasmids) was used.
Real-Time Polymerase Chain Reaction Assay and Western Blotting.
RNA extraction, complementary DNA synthesis, quantitative real-time PCR (qRT-PCR) reactions, and western blotting were performed as previously described.18
Immunofluorescence Staining and Apoptosis Analysis.
Primary antibodies (Abs) used were as follows: anti-HIF-2α from Novus Biologicals (Littleton, CO); anti-Ki67 from Dako (Carpentaria, CA); and anti- caspase-3 from Cell Signaling Tecnology (Boston, MA).
To quantify the extent of apoptosis, cells were harvested with trypsin without ethylenediaminetetraacetic acid and visualized by Annexin V/fluorescein isothiocyanate/propidium iodide (PI) staining, according to the manufacturer's protocol (BD Biosciences, Rockville, MD).
ChIP-on-chip, ChIP-PCR, and Immunoprecipitation.
A complete protocol of ChIP-on-chip provided by NimbleGen Systems (Madison, WI) is included in the Supporting Materials and Methods. Predicted binding sites were confirmed using ChIP-PCR in the MHCC97L cell line. Immunoprecipitation (IP) was performed as previously described.19
Data were analyzed with SPSS 16.0 software (SPSS, Inc., Chicago, IL), as previously described.17 A P value <0.05 was considered statistically significant.
Expression Level of HIF-2α in HCC Patients and Its Clinical Significance.
To explore effects of HIF-2α in HCC, we first detected HIF-2α levels in tumor and peritumoral tissues. HIF-2α showed a mainly cytoplasmic staining in cancer cells, as previously indicated (Fig. 1A),16 and the average levels of HIF-2α on both protein and messenger RNA (mRNA) were significantly lower in tumor than peritumoral tissues (Figs. 1B,C). We further explored the clinical significance of HIF-2α expression in 246 HCC cases. The clinical and pathological characteristics of these patients were listed in Supporting Table 1. The percentage and intensity of positively staining cells were largely varying among cases, and tumors were classified as strong in 118 cases (47.9%) and moderately or weakly positive in 128 cases (52.1%) (Fig. 1D).13 Duplicate tumor samples showed a good level of agreement with respect to intensity and percentage of positively stained cells. Kaplan-Meier analysis revealed that patients with high HIF-2α expression levels in tissues had significantly longer overall survival (OS) rates than those with low expression (P = 0.006) (Fig. 1E).
HIF-2α Inhibits HCC Cell Growth Mainly through Apoptosis.
To establish whether HIF-2α plays a suppressor role in HCC tumor growth, human liver cancer cell lines MHCC97H and SMMC-7721 were studied. Both HIF-1α and HIF-2α expression were up regulated in the MHCC97H cell line, and the proliferation rate was lowered under hypoxia condition (Supporting Figs. 1A,B). To explore the role of HIF-2α in overcoming the effects of HIF-1α and other factors, we isolated high- or low-HIF-2α expression clones that had been transfected with expression vector pcDNA3.1-HIF-2α or shRNA inhibition vector pT2sh-HIF-2α, respectively (Fig. 2A,B). In these clones, HIF-1α levels did not vary significantly (Supporting Fig. 1C).
Initially, viability of transfected cells was investigated. High expression of HIF-2α in MHCC97H cells was associated with a slower growth rate, compared to the respective control, whereas cells transfected with anti-HIF-2α shRNA grew faster than control (Figure 2C). These differences were also observed in the SMMC-7721 cell line (Supporting Fig. 1D,E). HIF-2α also significantly slowed cell growth rate under hypoxic condition, regardless of whether with or without HIF-1α RNA interference (Supporting Fig. 1F,G). Next, we determined whether the altered growth rate is the result of an increased cell death rate. Annexin V/PI assays showed a dramatic apoptosis rate increase in the high-expression HIF-2α clone, compared to respective control (Fig. 2D). Confocal fluorescence microscopy revealed mitochondrial dysfunction and more cytochrome C released from HIF-2α-overexpressing cells (Fig. 2E). Multidomain proapoptosis B-cell lymphoma 2 (BCL-2) proteins Bax and Bak were expressed at higher levels in cells with high HIF-2α expression, relative to controls (Fig. 2F), which would facilitate the release of cytochrome C from mitochondria by forming pores in the outer mitochondrial membrane.20, 21 We also observed a higher amount of cleaved caspase-3, a key executer of apoptosis, in overexpressing HIF-2α cells (Fig. 2F), which would have been triggered by the cytochrome C release. The Bcl-2 homology domain 3 (BH3)-only proapoptotic proteins, including the Bcl-2-associated death promoter (Bad), p53 up-regulated modulator of apoptosis (Puma), and Bid were also found to be up-regulated in HIF-2α-overexpression cells (Fig. 2F). No significant differences were found in Ki67 staining between groups with different HIF-2α levels (Supporting Fig. 1H). Collectively, these data further confirmed that high levels of HIF-2α in HCC cells caused a cell-growth arrest through apoptosis.
Having established that high expression of HIF-2α in human HCC increases OS rates and induces HCC cell-growth arrest in vitro, we further tested the effect of HIF-2α in vivo. Clones with different HIF-2α expression were established (Supporting Fig. 2A). Tumors formed by cells with high HIF-2α level were smaller than mock controls, and tumors formed by cells with low HIF-2α level showed no significant changes in tumor weight, compared to respective controls (Fig. 3A,B). Cell proliferation and apoptosis measurements indicated that overexpression of HIF-2α induced more cleaved caspase-3 expression in tumor cells, compared to mock-treated cells (Fig. 3C). These in vivo data support HIF-2α's role of inducing apoptosis in HCC cells.
HIF-2α Augments Apoptosis by Inhibiting Expression of the Target Gene, TFDP3.
To uncover how HIF-2α affects apoptosis, we sought to identify possible targets using a ChIP-on-chip screen in MHCC97H cells (GEO accession no.: GSE37167). With a specific anti-HIF-2α Ab, we identified 470 target genes bound within their promoter regions, spanning 2.2 kilobases upstream and 500 base pairs downstream of the transcription start site, by HIF-2α. To validate screening results, we performed an independent ChIP experiment coupled with qRT-PCR (Fig. 4A).
The biological function and cellular component of HIF-2α-binding target genes were analyzed according to their ChIP-on-chip assay enrichment score. Genes with the highest enrichment scores were those that regulate cellular processes and are located in the nucleus (Supporting Fig. 3A,B). A substantial number of genes regulating apoptotic processes were found to be HIF-2α-bound genes. The top 10 genes were ranked according to their PeakFDR values and PeakScore values (Supporting Table 5). We observed a particularly strong suppression of TFDP3 mRNA (Fig. 4B) and protein levels in HIF-2α-overexpressing cells, compared to controls (Fig. 4C), and these effects were not noted when HIF-1α levels were varied (Supporting Fig. 3C,D).
The TFDP3 gene encodes a member of the dimerization partner (DP) family of transcription factors,22-24 which exerts a regulatory function by dimerization with the E2F protein. The TFDP3 protein is known to be highly expressed in HCC, but not in healthy liver tissues.23 We hypothesized that HIF-2α repression of TFDP3 expression could account for the high rate of apoptosis observed in the HIF-2α-overexpressing cell line. To test this idea, we overexpressed TFDP3 and/or HIF-2α in MHCC97H cells (Supporting Fig. 3E). The number of activated caspase-3-positive cells was decreased when TFDP3 expression was reestablished in HIF-2α-overexpressing cells (Fig. 4D,E). The expression of Bax and Bad was also down-regulated in these cells (Fig. 4F). TFDP1 and TFDP2 expression levels were tested to exclude the possibility of effects on E2F transcriptional activity (Fig. S3F). Collectively, these data show that HIF-2α significantly increases apoptosis by repressing the transcription of TFDP3, and the reintroduction of TFDP3 expression significantly reverses the effect.
HIF-2α Induces Apoptosis Through the TFDP3/E2F1 Pathway.
It is well known that DP proteins bind to E2F1, forming heterodimers that are essential for E2F high-affinity DNA binding and efficient transcriptional activity.24, 25 The TFDP3 gene was shown to colocalize with E2F1 by confocal microscopy (Fig. 5A). The physical interaction between E2F1 and TFDP3 was further analyzed by co-IP. In an E2F1 pull-down assay, TFDP3 bound directly to E2F1. Reciprocally, in an IP assay using anti-TFDP3 Ab, E2F1 was found to be bound (Fig. 5B). Knockdown of E2F1 expression by siRNA in HIF-2α-overexpressing cells increased their growth rate and decreased some proapoptotic gene expression (Supporting Fig. 3G-I). All these data suggest that E2F1 plays an important role in the induction of apoptosis by HIF-2α.
E2F1 has been reported to induce apoptosis through both p53-dependent and -independent pathways.26 Therefore, expression levels of key genes in these two pathways were determined. Cells were transiently transfected with both HIF-2α and TFDP3 recombinant expression plasmids, either in combination or individually. Although results showed that the mRNA level of TP53 did not change significantly (Figure 5C), its protein level greatly increased in cells transfected with HIF-2α and decreased to a normal level when cells expressed high HIF-2α and TFDP3 levels (Fig. 5D). The p14ARF gene has been reported to be regulated by E2F1 and interacts with murine double minute 2 (Mdm2), thus preventing p53 ubiquitination and subsequent degradation.27 Our results showed that p14ARF levels increased more than 2-fold in the clonal line with high HIF-2α expression and decreased nearly to normal level in cells that expressed both HIF-2α and TFDP3 (Fig. 5C). No significant expression changes were found for Mdm2 and p53 kinase ataxia telangiectasia mutated, which induce p53 phosphorylation28 (Fig. 5C). Expression of the apoptosis-stimulating protein of p53 1 (ASPP1), which directs p53 activity toward apoptosis,29 increased significantly in cells with a low TFDP3 expression level and decreased to a nearly normal level in cells that coexpressed HIF-2α and TFDP3 (Fig. 5C). These results suggest that HIF-2α-TFDP3-E2F1 increases apoptosis by increasing p14ARF and ASPP1 transcription, which further increases the p53 protein level.
Expression of some genes in the p53-independent apoptotic E2F1 pathway, such as the p53 homolog, p73,30 BID,25 and several caspases,32 were found to be significantly up-regulated in cells with high HIF-2α levels, whereas the restoration of TFDP3 expression significantly reversed effects on expression levels of those genes (Fig. 5C). Effects on apoptosis protease-activating factor 1,33 Puma, and the BH3-only protein encoding the gene Noxa,34 regulated by both E2F1 and p53, also showed a similar trend (Fig. 5C,D). Overall, both p53-dependent and -independent pathways play a role in HIF-2α-induced apoptosis. We further measured expression levels of TFDP3 and some representative genes of the E2F1 pathway in HCC specimens, which were with different HIF-2α protein levels determined by immunohistochemistry (IHC) staining. We found that patients with high HIF-2α protein expression had low expression levels of TFDP3, but high levels of proapoptotic genes (Figure 5E). Collectively, data from in vitro models and HCC patient samples have been implicated for the mechanism of HIF-2α-induced apoptosis (Figure 5F).
HIFs are detected in most solid tumors and are known to correlate with poor patient prognosis. The identification of genes, such as VEGF and CA9, in a HIF-activated pathway reveals a clear mechanism by which HIFs contribute to the survival and progression of cancer cells, and clinical therapeutic targeting of HIFs has emerged as a rational approach to treat solid tumors.35 Recently, studies have shown that HIF isoforms differentially regulate gene expression and sometimes have opposing functions during tumor progression. Therefore, experiments elaborating the precise roles of each subunit is needed.4, 10 HIF-1α has been widely explored and shown to regulate cell proliferation and angiogenesis and to be correlated with OS in HCC.2, 3, 36 However, studies on HIF-2α function in HCC are limited and these results sometimes are inconsistent.1, 16, 31 Therefore, we explored clinical significance by measuring gene expression in tumor samples from HCC patients as well as putative mechanism by altering cellular HIF-2α expression levels by overexpression or knock-down experiments.
We found that HIF-2α-induced HCC cell growth arrest and knockdown of HIF-2α expression increased cell viability in vitro. These effects have previously been observed in other cell lines under hypoxic conditions.1 Our clinical data further confirmed the effect. Through analysis of samples from hundreds of HCC patients, we found that HIF-2α expression in HCC tissues positively correlated with patient OS. Moreover, there was a significant negative correlation between tumor size and relative HIF-2α level (P < 0.001). Thus, our results suggest a possible tumor-suppressor role of HIF-2α in HCC.
With ChIP-on-chip screening and subsequent confirmatory assays, we showed that TFDP3 is a novel transcriptional target of HIF-2α, and its expression was significantly repressed by HIF-2α. Our findings link HIF's transcriptional activity to one of the least understood of the DP family, TFDP3. An important property that distinguishes TFDP3 from other DP members is its ability to inhibit E2F1-induced p53-mediated apoptosis.39 We found that the p53-independent pathway is also modulated by TFDP3. The TFDP3 protein is only expressed in HCC, but not in healthy liver tissues.23 Further exploring its role as a potential therapeutic target in HCC may uncover new applications.
The E2F family is best known for regulating cell cycle, proliferation, and apoptosis and is involved in signaling cascades induced by hypoxia.32 The level of E2F1 is up-regulated under hypoxic stimulation, and HIF-2α has been shown to increase E2F1 levels by c-Myc.11 Here, we showed that HIF-2α promotes E2F1 transcriptional activity and expression levels of E2F1-dependent target genes by inhibition of TFDP3 expression (Fig. 5F). This property of TFDP3 provides another connection between HIFs and E2F1.
HIF-2α exhibits a mixed nuclear/cytoplasmic localization pattern in HCC samples, and both the percentage and intensity vary greatly between patients.16, 41 HIF expression is known to be affected by many factors. For example, viral hepatitis infection and liver cirrhosis cause inflammatory cell infiltration and a hypoxic microenvironment, further modulating HIF expression patterns.37, 38 Therefore, the rate of viral hepatitis infection and the presence of liver cirrhosis need to be carefully considered when evaluating a study cohort's HIF expression profile. There were 82.9% hepatitis B virus–positive and 81.3% cirrhosis-positive patients involved in our study, whereas these rates were only 67.3% and 45.1%, respectively, in another HIF-2α study.16 In addition, there are other epidemiologic features, such as geographic regions, racial and ethnic groups, gender, and study methods, such as cell type, cultural method, and staining method used, that may also influence HIF-2α expression profile.
The role of HIF-2α in the development of cancer is not clear, and there is some discrepancy in the existing literature. One study showed that HIF-2α was correlated with angiogenesis and a poor outcome in HCC,16 whereas we and another group found that HIF-2α regulated autophagy and apoptosis, and high HIF-2α expression in HCC correlated with a good outcome.1 Diverse or complex roles of HIF-2α have also been found in non-small-cell lung cancer (NSCLC). HIF-2α appears to promote tumor epithelial-mesenchymal transition and is defined as a promoter of NSCLC.40 On the other hand, deletion of HIF-2α results in an increased NSCLC tumor burden by regulation of tumor-suppressor gene secretoglobin 3A1a expression.5
Reconciling these diverse roles is difficult. One possibility is that the tumor microenvironment markedly influences the active role of HIFs,1 as has been shown in astrocytoma. Loss of HIF-1α impairs astrocytoma growth subcutaneously, but it increases its proliferative and invasive properties in the brain.14 In HCC, hepatitis virus infection and liver cirrhosis are perhaps important factors, which significantly affect the tumor microenvironment.37, 38 Another possibility is that a tumor may employ a “stop and go” strategy to maintain its growth and survival, as suggested by Acker et al.15 When oxygen levels are limited, HIF-2α acts negatively on survival by inducing apoptosis, but, at the same time, promotes angiogenesis. Clearly, the mechanisms regulating HIF-2α function need to be further explored.
In conclusion, we have established that TFDP3 is a novel transcriptional target of HIF-2α and have uncovered another mechanism that links HIFs to a fundamental biologic regulator, E2F1. We have shown that HIF-2α induces HCC cell apoptosis by its effect on the TFDP3/E2F1 pathway and provided clinical evidences demonstrating a possible tumor-suppressor role of HIF-2α.