The Clinicopathologic and Prognostic Value of Hypoxia-Inducible Factor-2α in Cancer Patients: A Systematic Review and Meta-Analysis

Oxygen is essential for energy metabolism, which drives cellular biological activity (1). Region of hypoxia serves as a common microenvironment in the development and progression of many solid tumors (2). Hypoxia is associated with a series of molecular biology of tumors, such as proliferation, invasion, migration, and chemoradiotherapy resistant (3, 4). The mediator of hypoxia in tumor is hypoxia-inducible factors (HIF), which are heterodimeric complex composed of a constitutive subunit (HIF1β) and a regulated subunit including HIF1α, HIF2α, or HIF3α (5). Among them, so far HIF1α and HIF2α have received more attention (6), and increasing evidence has demonstrated the structure, function, and regulation of HIF2α in human cancer (7, 8).

HIF2α shares approximately 48% of amino-acid sequence homology with HIF1α (9). As opposed to HIF1α protein, HIF2α is active under mild or physiologic hypoxia (<5% O₂; ref. 10). During the process of tumor development, HIF2α shows different functions, such as angiogenesis, proliferation, and tumor stem cells. In addition, HIF2α exhibits distinct roles in hypoxic activation of target genes compared with HIF1α. In clinical specimens, HIF2α involved in the development of many human cancers, such as pancreatic cancer (11), colorectal cancer (12), and ovarian cancer (13). Moreover, HIF2α was directly or indirectly related to the regulation of progression of tumors (14, 15). It was reported that HIF2α expression was associated with poor prognosis in patients with various cancers, including lung cancer (16), gastric carcinoma (17), pancreatic cancer (18), and breast cancer (19). However, some studies indicated that there was no significance between HIF2α expression and the prognosis of tumors (20, 21). Hence, the results of different studies are controversial, and the prognostic value of HIF2α expression in cancer remains unknown.

In this study, a meta-analysis from eligible studies was performed to investigate the relationship between HIF2α expression and the overall survival (OS) on both univariate and multivariate analyses in cancer patients. Furthermore, we made subgroup analysis to assess the roles of HIF2α in clinicopathologic factors of cancer.

This analysis was conducted according to the Preferred Reporting Items for Systematic Reviews and Meta-analyses (PRISMA) guidelines (22).

This meta-analysis was limited to assess the prognostic implication of HIF2α expression in human cancer patients. A comprehensive literature search of the following databases was performed prior to May 27, 2018: PubMed, Web of Science, Cochrane Library databases, Chinese National Knowledge Infrastructure, and Wanfang databases. The articles were identified using the search strategy: (cancer OR sarcoma OR tumor OR neoplasm) AND (HIF2α OR HIF2α OR HIF2 OR HIF2 OR hypoxia-inducible factor 2α OR hypoxia-inducible factor-2α) AND (prognosis OR survival OR outcome). The language of enrolled studies was restricted to English and Chinese. Citation lists of included studies were screened manually to ensure sensitivity of the search strategy.

All studies included in this meta-analysis were accorded with the following inclusion criteria: (i) case–control studies focus on the relationship of HIF2α expression with the prognosis and clinicopathologic characteristics of human cancer patients; (ii) studies were reported HIF2α expression status, with availability of OS data on univariate or multivariate analysis, reported either as hazard ratios (HR) or as Kaplan–Meier (KM) curves; (iii) studies were published in English or Chinese with the full text. In addition, the major exclusion criteria were as follows: (i) overlapped articles or studies with overlapping data; (ii) reviews or meta-analysis; (iii) no sufficient data to estimate the HR and 95% confidence intervals (95% CI).

The following information was independently extracted by two investigators: first author's name, publication year, region, histologic type, expression type of HIF2α, patient cases, number of HIF2α positive, percentage of HIF2α positive, analysis type, and follow-up time. The Newcastle–Ottawa scale (NOS) score was used for assessing the quality, which ranged from 0 to 9. Studies with scores ≥5 were regarded as high-quality studies; otherwise, studies were considered to have a low quality. In addition, clinicopathologic characteristics included gender (male vs. female), tumor differentiation (poor vs. well/moderate), tumor–node–metastasis (TNM) stage (III/VI vs. I/II), and lymph metastasis (yes vs. no).

All analyses were performed using STATA 12 software (STATA Corp.). The effect of HIF2α expression on OS of cancer was calculated as HR with 95% CIs. If the study only showed KM graphs, KM curves were read by Engauge Digitizer version 4.1 (http://digitizer.sourceforge.net/). Risk ratios (RR) and corresponding 95% CIs were combined to evaluate the relationship between HIF2α expression and clinicopathologic characteristics, including gender, tumor differentiation, TNM stage, and lymph metastasis. A heterogeneity test of pooled HRs was conducted using χ²-based Cochran Q test and I² index. An I² value > 50% indicated significant heterogeneity among studies; then a random-effects model was used. Otherwise, a fixed-effects model was applied to pooled data (I² < 50%). To detect heterogeneous studies, a sensitivity analysis was performed to evaluate the influence of individual studies on the stability of pooled results. Publication bias was assessed by visually assessing a Begg funnel plot and by quantitatively performing Begg test and Egger test. All statistical tests were two sided, and statistical significance was defined as P value less than 0.05.

A total of 854 articles related with HIF2α and cancer prognosis were identified from online database searches. According to the literature selection criteria (Fig. 1), 40 studies were included in this meta-analysis (16–21, 23–56), including 4,345 cancer cases and 2,066 HIF2α-positive cases. The main characteristics and clinicopathologic factors of the included studies were summarized in Table 1 and Supplementary Table S1, respectively. All studies included in the meta-analysis were retrospective studies published between 2002 and 2017. Of these studies, 28 were Asian and 12 were non-Asian. The type of cancers included breast cancer, chondrosarcoma, colorectal cancer, diffuse large B-cell lymphoma (DLBCL), endometrial carcinoma, gastric and esophageal cancers, hepatocellular cancer (HCC), head and neck squamous cell cancer (HNSCC), lung cancer, malignant astrocytoma, neuroblastoma, osteosarcoma, oral squamous cell carcinoma (OSCC), pancreatic cancer, renal cell cancer (RCC), and salivary duct carcinoma (SDC). Among these studies, 38 studies detected the protein expression level of HIF2α, and 2 studies reported the mRNA expression of HIF2α. Overall, sample sizes ranged from 21 to 315 cases, and the follow-up of these studies ranged from 20 to 247 months. Patients’ OS on univariate and multivariate analyses were reported in 35 studies and 25 studies, respectively. Moreover, 23 studies were evaluated as high quality and 17 as low quality.

To further investigate the effect of HIF2α on the prognosis of cancer, we analyzed clinicopathologic features including gender, differentiation, TNM stage, and lymph metastasis. All the detailed data are summarized in Table 2 and Supplementary Fig. S1A–S1D. The results indicated that high expression of HIF2α was not related with gender (RR, 0.94; 95% CI, 0.87–1.02, P = 0.116; random-effects model: χ² = 15.67, I² = 0%, P = 0.476). However, the overexpression of HIF2α was significantly associated with tumor differentiation (RR, 1.42; 95% CI, 1.23–1.65, P < 0.001; random-effects model: χ² = 22.77, I² = 56.1%, P = 0.012), TNM stage (RR, 1.50; 95% CI, 1.35–1.66, P < 0.001; random-effects model: χ² = 10.41, I² = 13.5%, P = 0.319), and lymph metastasis (RR, 1.38; 95% CI, 1.24–1.54, P < 0.001; random-effects model: χ² = 8.35, I² = 4.2%, P = 0.400).

Thirty-five studies demonstrated the association between HIF2α expression and OS with univariate analysis in human cancer patients. As shown in Fig. 2, relevant results indicated that HIF2α overexpression was markedly correlated with a poor OS survival with univariate analysis among cancer patients (HR, 1.64; 95% CI, 1.41–1.92, P < 0.001; heterogeneity: random-effects model: χ² = 81.51, I² = 58.3%, P < 0.001). In addition, we further analyzed the relationship between expression of HIF2α and OS with multivariate analysis among 25 studies. As shown in Fig. 3, results indicated that there was significant association of high HIF2α expression with lower OS on multivariate analysis (HR, 2.21; 95% CI, 1.70–2.87, P < 0.001; heterogeneity: random-effects model: χ² = 90.39, I² = 73.4%, P < 0.001).

There was some evidence for heterogeneity in OS data with both univariate analysis and multivariate analysis; therefore, we created Galbraith plots to identify potential sources of heterogeneity. As shown in Supplementary Fig. S2A, the studies by Biswas, Sun, Szendroi, and Xie should be the main contributors to heterogeneity in OS data with univariate analysis. When these four studies were omitted, the statistical significance of the combined HRs was not substantially altered, but I² decreased from 58.3% to 0% (Supplementary Fig. S3A). Correspondingly, the main sources of heterogeneity in OS data with multivariate analysis were studies by Sun, Wang, Xie, and Yao (Supplementary Fig. S2B). After omission of the above four studies, I² decreased from 73.4% to 37.4% (Supplementary Fig. S3B).

To further explore the sources of high heterogeneity, we conducted subgroup analysis for OS data with univariate analysis (Table 3; Supplementary Fig. S4A–S4E) according to region, histologic type, number of cases, months of follow-up, and quality of the included study. Based on histologic type, HIF2α overexpression resulted in poor OS in lung cancer (HR, 1.89; 95% CI, 1.46–2.44, P < 0.001), colorectal cancer (HR, 1.72; 95% CI, 1.21–2.43, P = 0.002), pancreatic cancer (HR, 1.97; 95% CI, 1.42–2.75, P < 0.001), gastric and esophageal cancers (HR, 1.65; 95% CI, 1.21–2.26, P = 0.002), HNSCC (HR, 1.79; 95% CI, 1.26–2.53, P = 0.001), and neuroblastoma (HR, 2.53; 95% CI, 1.29–4.93, P = 0.007). Clinical associations between HIF2α and OS with univariate analysis were found in the Asian (HR, 1.75; 95% CI, 1.44–2.11, P < 0.001), non-Asian (HR, 1.40; 95% CI, 1.08–1.82, P = 0.011), smaller cases (n < 100; HR, 1.77; 95% CI, 1.35–2.33, P < 0.001), larger cases (n ≥ 100; HR, 1.57; 95% CI, 1.30–1.90, P < 0.001), shorter follow-up time (n < 100; HR, 1.80; 95% CI, 1.48–2.20, P < 0.001), longer follow-up time (n ≥ 100; HR, 1.36; 95% CI, 1.10–1.67, P = 0.004), high quality (HR, 1.70; 95% CI, 1.40–2.07, P < 0.001), and low quality (HR, 1.54; 95% CI, 1.18–2.02, P = 0.002).

In addition, to demonstrate the predictive role of HIF2α in multivariate analysis, subgroup analysis was conducted based on patients’ region, histologic type, number of cases, months of follow-up, quality of the included study, and expression type of HIF2α (Table 4; Supplementary Fig. S5A–S5F). After stratifying by histologic type, significantly poor OS were obtained in HCC (HR, 3.35; 95% CI, 1.32–8.50, P = 0.011), lung cancer (HR, 3.94; 95% CI, 1.47–10.58, P = 0.006), breast cancer (HR, 1.78; 95% CI, 1.19–2.67, P = 0.005), colorectal cancer (HR, 2.78; 95% CI, 1.23–6.31, P = 0.014), HNSCC (HR, 1.85; 95% CI, 1.10–3.11, P = 0.021), and sarcomas (HR, 2.42; 95% CI, 1.14–5.14, P = 0.022). The significant associations also exist in other subgroups, including the Asian (HR, 2.46; 95% CI, 1.68–3.62, P < 0.001), non-Asian (HR, 1.72; 95% CI, 1.40–2.12, P < 0.001), protein expression of HIF2α (HR, 2.23; 95% CI, 1.69–2.93, P < 0.001), smaller cases (n < 100; HR, 2.63; 95% CI, 1.75–3.96, P < 0.001), larger cases (n ≥ 100; HR, 1.91; 95% CI, 1.42–2.56, P < 0.001), shorter follow-up time (n < 100; HR, 2.71; 95% CI, 1.74–4.24, P < 0.001), longer follow-up time (n ≥ 100; HR = 1.90; 95% CI, 1.44–2.51, P < 0.001), high quality (HR, 2.28; 95% CI, 1.71–3.05, P < 0.001), and low quality (HR, 1.94; 95% CI, 1.14–3.31, P = 0.015).

Sensitivity analysis was conducted through the sequential omission of individual studies. Furthermore, no single study could essentially change the results in both univariate analysis (Supplementary Fig. S6A) and multivariate analysis (Supplementary Fig. S6B) of OS, demonstrating that the results of this meta-analysis were statistically stable. In addition, no evidence for significant publication bias was found for OS with univariate analysis (Fig. 4A; Begg P = 0.293 and Egger P = 0.345) or multivariate analysis (Fig. 4B; Begg P = 0.059 and Egger P = 0.060).

Mammalian cells need oxygen to maintain efficient energy supply, and lack of oxygen can eventually lead to cell death due to impaired energy-requiring processes (57). A hypoxic microenvironment in tumors represents one of the major obstacles for effective cancer therapies (58). HIF2α is a type of HIF that mediates transcriptional response to hypoxia stress. Many lines of evidence suggest that the expression level of HIF2α was enhanced and predicted poor survival in a variety of cancers, such as lung cancer (16), colorectal cancer (26), breast cancer (59), and osteosarcoma (32). Conversely, some studies have indicated that HIF2α have a significant impact on the development of tumors, and increased HIF2α expression could be a favorable prognostic factor in RCC (43) and HCC (21). Moreover, clinicopathologic significance of HIF2α in cancer patients remains inconclusive. Some studies indicated that HIF2α overexpression might predict worse tumor differentiation, higher TNM stage, and higher lymph metastasis (17, 25, 54). However, some reports found that there was no obvious association between HIF2α and the above clinicopathologic features (16, 27, 36). Although one meta-analysis has reported on the prognostic role of HIF2α in cancer, it lacks test of heterogeneity, publication bias, and correlation analysis between HIF2α and clinicopathologic features.

In this meta-analysis, a total of 854 studies with 4,345 patients were obtained. Our data indicated that the overexpression of HIF2α was significantly correlated with the prognosis of cancer patients. High expression of HIF2α could predict unfavorable OS of cancer patients on both univariate and multivariate analyses (P < 0.001 and P < 0.001, respectively). Furthermore, significant results were observed in the subgroup analyses of OS on both univariate and multivariate analyses by region, histologic type, number of cases, months of follow-up, quality of the included study, and expression type of HIF2α. Meanwhile, we investigated the relationship between high HIF2α expression and clinicopathologic features. The results indicated that HIF2α overexpression was associated closely with worse tumor differentiation, higher TNM stage, and higher lymph metastasis. In addition, there was no obvious evidence for significant publication bias on OS with both univariate and multivariate analyses.

It is well known that oxygen is important for tumor cell survival and lack of oxygen would cause cell death. Tumor cell was able to adapt to the hypoxia environment through reducing energy consumption and increasing anaerobic metabolism. Hypoxia-inducible transcription factors, especially HIF2α, play a critical role in the adaptation process. Tumor hypoxia and high HIF2α protein levels are frequently related with aggressive disease and represent major obstacles to effective cancer therapies (60). A previous study has shown that targeting HIF2α had great clinical potential by affecting tumor spread and transition into advanced clinical stages (61).

HIF2α is known to regulate proteins involved in cell proliferation, cell growth, cellular glucose transport, and angiogenesis. Evidence indicates that HIF2α is involved mainly in metastasis and chemoresistance in advanced cancers (61). Recently, we have seen great progress being made in understanding HIF2α-mediated cancer pathways. Overexpression of angiogenic VEGF and EphA2 markers, which were mediated by HIF2α, is well correlated with tumor recurrence and progression in patients with HCC. Moreover, activation of HIF2α promoted PKC-mediated tumor cell migration (62). HIF2α could promote epithelial–mesenchymal transition by regulating the expression of N-cadherin and E-cadherin via activation of the MEK/ERK pathway in gastric cancer (63). Furthermore, HIF2α increased multidrug resistance of stomach cancer cells by directly activating the pregnane X receptor (PXR) signaling pathway (64). In addition, HIF2α promoted colon cancer growth by potentiating Yes-associated protein 1 (YAP1) activity, suggesting that this pathway might be targeted in potential anticancer approaches for treating colorectal cancer patients (65).

Although we executed exhaustive meta-analysis, our study had some limitations. First, all the studies included were retrospective articles, and randomized controlled trials were not found. Second, although being modified for heterogeneity by the application of random-effect model, subgroup analyses, and sensitivity analyses, there was still some heterogeneity in some subgroups. Third, the studies included in this meta-analysis were reported only in the English and Chinese languages, and studies reported in other languages were omitted.

In summary, this meta-analysis revealed that overexpression of HIF2α was positively associated with clinicopathologic features and poor OS of cancer patients, suggesting that HIF2α might be a significant biomarker for cancer diagnosis and prognosis.

No potential conflicts of interest were disclosed.

This work was supported by the Natural Science Foundation of Zhangzhou, Fujian, China (ZZ2017J36 to H.Y. Ren) and the Youth Nursery Foundation of the Affiliated Southeast Hospital of Xiamen University, Zhangzhou, Fujian, China (16Y012 to D.Q. Luo).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.