Clinical and Prognostic Significance of HIF-1α, PTEN, CD44v6, and Survivin for Gastric Cancer: A Meta-Analysis

Purpose This study was to quantitatively summarize published data for evaluating the clinical and prognostic significance of four proteins involved in hypoxia-inducible factor-1 (HIF-1α) regulation of the metastasis cascade. Methods Searches were performed using the MEDLINE, EMBASE, Cochrane Library, and Chinese Biomedicine databases without any language restrictions. Studies were pooled and either the summary risk ratio (RR) or odds ratio (OR) was calculated. Potential sources of heterogeneity were sought out via subgroup and sensitivity analyses, and publication bias was also performed. Results Seventeen studies evaluated HIF-1α, 20 studies evaluated phosphatase and tensin homolog (PTEN), 20 studies evaluated Survivin, and 16 studies evaluated CD44v6. Our results showed that increased HIF-1α expression was linked to a poor 5-year overall survival (RR = 1.508; 95% confidence interval (CI) 1.318–1.725; P<0.001). Decreased survival was heavily influenced by advanced tumor invasion (OR = 3.050; 95% CI 2.067–4.501; P<0.001), lymph node metastasis (1415 patients; OR = 3.486, 95% CI 2.737–4.440; P<0.001), distant metastasis (OR = 6.635; 95% CI 1.855–23.738; P = 0.004), vascular invasion (OR = 2.368; 95% CI 1.725–3.252; P<0.001), dedifferentiation (OR = 2.112; 95% CI 1.410–3.163; P<0.001), tumor size (OR = 1.921; 95% CI 1.395–2.647; P<0.001), and a higher TNM stage (OR =  2.762; 95% CI 1.941–3.942; P<0.001). Similarly, aberrant expression of PTEN, CD44v6, and Survivin were also observed in tumors that correlated with poor OS. The higher ORs of death at 5 years were 1.637 (95% CI = 1.452–1.845; P<0.001), 1.901 (95% CI = 1.432–2.525; P<0.001), and 1.627 (95% CI = 1.384–1.913; P<0.001), respectively, with an OR>2 for the main stratified meta-analyses of clinical factors. Conclusions Our findings indicate that HIF-1α/PTEN/CD44v6/Survivin, as measured by immunohistochemistry, can be used to predict the prognosis and potential for invasion and metastasis in Asian patients with gastric cancer. The development of strategies against this subset of proteins could lead to new therapeutic approaches.


Introduction
Gastric cancer is one of the most aggressive tumors and tends to be associated with peritoneal dissemination, lymph node metastasis, and hematogenous metastasis. Although recent advances in its diagnosis and treatment have offered increased long-term survival for patients diagnosed at early stages of gastric cancer, the prognosis of advanced cancer remains dismal, with a 5-year survival rate of only 10-15% [1,2]. A majority of patients with advanced disease die due to complications induced by metastasis but not the primary tumor [3]. Recently, a series of rate-limiting steps have been proposed for tumor cells to become metastatic [4]. The multi-step processes consist of loss of cellular adhesion, local invasion, motility, angiogenesis, intravasation, circulation, extravasation, homing and the premetastatic niche, and organotropic colonization in specific organs [5]. Therefore, identifying novel markers in the key steps of metastasis will help to predict recurrence and survival for patients in the early stages of gastric cancer.
Hypoxia has been reported to contribute directly to many critical aspects of cancer biology, including angiogenesis, epithelial-mesenchymal transition, invasion, metastasis, stem cell maintenance, energy metabolism, autocrine growth factor signaling, and refractory to targeted therapies [6,7]. The best characterized hypoxia response pathway is mainly mediated through a transcription factor called hypoxia-inducible factor-1 (HIF-1a) [8]. Currently, the number of target genes, which are controlled by HIF-1a, is greater than 1000 and can be divided into the following five categories: transcription factors and histone modifiers; matrix degradation enzymes; receptor, receptor-associated kinases, and transporters; microRNA targets; and cell-adhesion molecules and membrane proteins [9,10]. In addition, routine phase 1 and phase 4 clinical trials that target HIF-1a function or expression have been completed, including a pilot trial of oral Topotecan for the treatment of refractory advanced solid neoplasms expressing HIF-1a and the effects of Dutasteride on HIF-1a and vascular endothelial growth factor (VEGF) in the prostate (Clinical Trial: NCT00117013, NCT00880672; http://clinicaltrials.gov/). The positive results from these clinical trials have further reinforced the interest in drug development targeting HIF-1a signaling.
Despite the clinical development of anti-HIF-1a therapies, the prognostic and clinical value of HIF-1a overexpression in gastric cancer cells remains unclear. It is essential to explore whether tumors in which HIF-1a is overexpressed are associated with reduced survival. As the incidence and mortality rate of gastric cancer are extremely higher in Eastern Asian especially China, Japan and Korea, we present a meta-analysis evaluating the prognostic impact of one subset of proteins in HIF-1a signaling in gastric cancer patients in subgroup of different continents. We hope that our meta-analyses will provide a framework for hypoxia regulation of the metastasis cascade and further uncover the role of hypoxia/HIF-1a-regulated key target genes on the prognosis based on various steps of metastasis. Most importantly, the analyses of gene expression profiles on prognosis may lead to the development of clinical methods that can be used to predict the outcome of individual patients in a clinical setting.

Search strategy and selection criteria
The meta-analysis was performed by means of preferred reporting items for meta-analyses statement [11,12]. The PUBMED, EMBASE, Cochrane Library, and Chinese National Knowledge Infrastructure (CNKI) databases were searched (up until June 2013) without language restrictions. Various combinations of the following terms were used to screen for potentially related studies: ''prognosis'' and ''survival'' and ''gastric'' or ''stomach'' as well as ''cancer'' or ''carcinoma'' or ''tumor''.
Studies were included in the meta-analysis if they met the following criteria: (1) diagnosed gastric cancer and normal gastric epithelial mucosa in humans; (2) evaluated proteins by using immunohistochemistry (IHC) methods; (3) used Asian cohorts from medical centers, and (4) had a follow up time exceeding 5 years. The study selection was based on the association of proteins related to HIF-1a-mediated tumor metastasis signaling and prognosis. References of retrieved articles were cross-searched to identify any studies missed by the computerized literature search. Authors of eligible studies were contacted for additional data relevant to the meta-analysis. However, not all authors responded.

Data extraction and methodological assessment
Data retrieved from all full publications included author, year of publication, country, antibody used for evaluation, and cut-off for diagnosis based on abnormal protein expression. In addition, data was collected on the number of readers, blinded readings, number of controls and cases, depth of invasion, lymph node status, distant metastasis, TNM stage, vascular invasion, histo-differentiation, tumor size, sex, and age of gastric cancer patients. Overall survival is defined as the time elapsed from surgery to death of patients with gastric cancer. In all cases, the data of interest for 5-year survival rates were extracted from Kaplan-Meier curves.
We tried to carefully avoid the duplication of data by examining each publication, the names of all authors, and the different medical centers involved. When an individual author published several articles obtained from the same or overlapping population, only the newest or most complete article was included in the analysis; otherwise independent data were analyzed. All data were extracted independently by three investigators (Chen J, Li T, and Liu XX), and any disparities were resolved by discussion.

Statistical analysis
In the present study, we analyzed three categories of stratified models. The first stratified multivariate model was performed to confirm whether each protein was abnormally expressed in gastric cancer compared to the normal gastric mucosa. The second outcome of the meta-analysis was to measure the impact of aberrant protein expression on 5-year overall survival. The third          Based on clinical characteristics, the following variables were compared: T 1 and T 2 vs. T 3 and T 4 ; Stage I and Stage II, vs. Stage III and Stage IV; well and moderate differentiation vs. poor and undifferentiation; tumors larger than 5 cm in size vs. tumors less than 5 cm; and patients older than 60 years vs. patients younger than 60 years.
Data were combined to perform meta-analysis using STATA version 9.0, and a two-tailed p-value of less than 0.05 was considered to be statistically significant. Estimates of risk ratios (RRs) or odds ratios (ORs) were weighted and pooled using different models based on their extent of heterogeneity. The heterogeneity across studies was quantified by using the I 2 statistic, which is generally considered significant for values $50%. In the case of heterogeneity, meta-analysis was performed using the random effects model after exploring the causes of heterogeneity. Otherwise, the fixed-effects model was applied. In addition, oneway sensitivity analysis was conducted to assess the stability of the results, namely, a single study in the meta-analysis was deleted one at a time to check the influence of the individual data set on the pooled RR (or OR) [13,14]. Both Begg's funnel plot and Egger's test were used to determine any publication bias for each of the pooled study groups.

Study selection and characteristics
After screening a collection of publications on proteins associated with hypoxia regulation of the metastasis cascade and prognosis, we identified 73 studies that used IHC techniques to assess the expression of four proteins on our prior criteria in the search strategy and study selection. Of these, 17 studies  Table 1). The median age of patients was 57.7 years old. The main characteristics of included studies are shown in Table 1 and Table S1, and the publications used to perform meta-analyses are listed in the References S1.

Meta-analysis results
Correlation of HIF-1a expression between gastric cancer and normal gastric mucosa. Our analyses, combining 8 independent studies that included 923 patients and 898 controls, revealed that HIF-1a overexpression was frequently observed in Table 2. Cont.   Table 2).

Sensitivity analysis and publication bias
In the present study, sensitivity analysis indicated that the pooled RR (or OR) was not significantly influenced by omitting any single study at a given time. In addition, the results showed there was no evidence of publication bias for most subgroup analyses (Table 2 and 3). However, the potential biases observed for depth of invasion for the HIF-1a group (P bias = 0.017), casecontrol study (P bias = 0.034) and overall survival (P bias = 0.006) for the PTEN group, and case-control study (P bias = 0.011), depth of invasion (P bias = 0.027), and overall survival (P bias = 0.008) for the Survivin group could be ruled out by Begg's and Egger's tests ( Table 2 and 3).

Discussion
Almost two-thirds of the world's gastric cancer cases occur in Asia (China, Japan, and Korea) [15]. The mortality of cancer patients is largely caused by metastases rather than their primary tumors at the time of diagnosis. Therefore, identifying the risk of disease recurrence and mortality in gastric cancer patients is critical to monitor patients and select appropriate adjunctive therapies in clinical practice [16,17]. However, useful biomarkers for predicting the prognosis of gastric cancer patients have not been well studied. Here, we introduced one subset of potential clinically useful biomarkers, HIF-1a/PTEN/CD44v6/Survivin, and precisely estimated their prognosis and clinicopathological significance. Mounting evidence suggests that hypoxic tumor microenvironments, especially the overexpression of HIF-1a, are strongly implicated as the hallmark of a wide variety of human malignancies [18,19]. When activated by the novel tumor suppressor gene PTEN [5,20], HIF-1a can transcriptionally regulate a host of hypoxia-responsive molecules that contribute to drug resistance, epithelial-mesenchymal transition, survival, angiogenesis, and metastasis [9,10,21,22,23], including inducers of angiogenesis (e.g., VEGF), proliferation of regulatory proteins (e.g., Survivin), and mediators of metastasis (e.g., CD44v6, MMP, E-cadherin). In this study, we found that the overexpression of HIF-1a occurred at a median frequency of 54.1% in gastric cancer. Patients who expressed high levels of HIF-1a were associated with a worse outcome, with a pooled risk for overall survival (RR = 1.508) that was similar to that obtained in a recently published study on hepatocellular carcinoma (HR = 1.65) [24]. Moreover, aberrant expression of PTEN, CD44v6, and Survivin were also observed in tumors correlating with poor overall survival, with risk of death at 5 years of 1.637, 1.901, and 1.627, respectively (Table 2 and 3). Subgroup analysis confirmed that the reduced survival was significantly correlated with increased dedifferentiation, tumor size, advanced tumor invasion, lymph node spread, distant metastasis, vascular invasion, and higher TNM stage, indicating increased biological aggressiveness and a greater possibility of systemic diffusion.
Gastric tumors can trigger the substantial development of new blood vessels for tumor growth, maintenance, and metastasis [25,26]. The high proliferation of tumor cells can induce local hypoxia, which is a strong stimulus for HIF-1a. Loss of PTEN in gastric cancer promotes tumor angiogenesis and invasion by increasing expression of VEGF through the increase of the HIF-1a protein level, which is an active process that requires the degradation of the extracellular matrix, the increase of microvascular permeability both in the blood and lymphatic vessels, and interstitial fluid pressure (IFP) [27], favoring the progression of the intravasation and extravasation of tumor cells. This may offer a possible explanation for the observed strong statistical association of HIF-1a overexpression with advanced tumor invasion, lymph node spread, vascular invasion, and distant metastasis. Recent studies have shown that the percentage of CD44 positive cells expressing variant exons v6 (CD44v6) in tumor cells could be significantly increased by HIF-1a-mediated transcription under hypoxic conditions [28]. CD44v6, one cell adhesion molecule, has been proposed to function in the homing of lymphocytes, extracellular matrix binding, cell migration, and invasion. Therefore, a high level of CD44v6 expression may also contribute to the aggressive phenotype. In addition, the upregulation of Survivin by HIF-1a and PTEN contributed to cisplatin (CDDP) resistance, indicating that inhibition of these pathways may be a potential strategy for overcoming CDDP resistance in the treatment of gastric cancer [29].
These analyses have several important implications. First, we show that the abnormal expression of HIF-1a/PTEN/CD44v6/ Survivin is associated with worse outcome, which suggests that each protein may be a useful therapeutic target for drug development. Currently, some clinical trials targeting HIF-1a/ PTEN/Survivin at different phases are being developed, which will likely benefit populations with certain conditions (http://www. clinicaltrials.gov) (Table 4). Second, the analyses highlight the importance of developing multiple biomarkers for monitoring treatment response, clinical uses of HIF-1a inhibitors, and prognosis assessment. We have also noticed several new-ly- publicated meta analyses of estimating prognostic value of either HIF-1a or Survivin on gastric cancer patient [30,31,32,33]. Among them, three quarters of electronic databases were used to identify published studies before December, 2012, with the combined HR ,2. Because hypoxia is likely to have complex, and even opposing, effects during different stages of tumor development [5,8], any single molecule cannot be used to independently predict the full prognosis of patients. Combinations of proteins involved in HIF-1a regulation of the metastasis cascade should provide increased prognostic power over individual markers themselves. Moreover, systematically assessing the main prognostic factors in gastric cancer, both tumor-and patientrelated, may also have meaningful impacts at the time of diagnosis or surgical treatment, including the depth of tumor invasion, lymph node spread, venous invasion, TNM stage, differentiation, tumor size, as well as sex and age. In present study, the most key results of the meta-analysis of clinical variables among Asian patients showed a magnitude of effect size of OR .2, and in some cases .3. As a rule of the thumb, a prognostic factor with RR (or OR) .2 is considered to be of useful practical value [34]. Therefore, we believe our results will provide more useful and precise information for clinical decision-making regarding gastric cancer. Third, as shown by our previous reports, VEGF appears to be a significant prognostic factor for hematogenous metastasis of gastric cancer (RR = 2.45, P = 0.000) [34,35]. In addition, we have proposed three other genes, E-cadherin, Stat3, and MMP-9, as prognosis biomarkers of tumor metastasis. Therefore, analyses combining previous results may show a possible axis of action by HIF-1a and its oncogenic signaling pathway (Fig. 6), which could contribute to improvements in prognosis assessment, functional analysis, and drug-targeted therapy in the prevention and treatment of gastric cancer. From this perspective, we believe that our meta-analysis does indeed present positive significance and novelty.
There are also limitations that should be noted based on the present analysis. First, because this is a literature-based analysis from which predominantly positive results were published, our estimate for the association between HIF-1a/PTEN/CD44v6/ Survivin and poor outcome might be inflated. Therefore, the discrepancies in the conclusions of various studies encouraged researchers to publish their data regardless of the significance of their results, which may limit the publication bias. In the present study, we placed emphasis on assessing biases across studies and pinpointing any potential sources of heterogeneity. Subgroup analyses by ethnicity and clinical variables were also performed. We comprehensively assessed the publication biases by using Begg's and Egger's tests and did not detect any significant deviation among most studies, except for the several factors mentioned in the ''Results'' section. In view of this, we are confident that the results of our meta-analysis are reliable. In addition, we could not pool the hazard ratios of death because of a lack of time-to-death data. We only figured out the risk ratio of death at fixed time points. Although this measure is less robust because it does not consider the duration of survival until death, this is the only feasible method of the data available [36].
In conclusion, our analyses show that the aberrant expression of HIF-1a, PTEN, CD44v6, and Survivin, as measured by IHC, may predict the 5-year overall survival risk and potential for invasion and metastasis in gastric cancer patients, particularly in Asian patients. These data suggest that the development of strategies against this subset of proteins could lead to a reasonable therapeutic treatment program for gastric cancer. However, further large sample and non-Asian population-based studies are required. Author Contributions