Association between rs2853669 in TERT gene and the risk and prognosis of human cancer: a systematic review and meta-analysis

The polymorphism rs2853669 within the promoter of telomerase reverse transcriptase gene (TERTp) has been debated about its role in cancer risk and prognosis. Additionally, several studies report inconsistent results concerning the modifying effect of rs2853669 on the prognostic value of TERTp mutations in cancer patients. Here, we performed this meta-analysis to comprehensively evaluate the role of rs2853669 in the risk and prognosis of human cancer, and further assess its modifying impact on TERTp mutations in the survival of cancer patients. We systematically searched literature via PubMed, Web of Science, and EMBASE through July 2016, and included 22 eligible studies. The overall analysis (64,119 cases and 78,988 controls) demonstrated that rs2853669 did not increase or decrease the overall cancer risk. Subsequent analyses also did not reveal any association between rs2853669 and overall cancer prognosis. However, we identified a modifying effect of rs2853669 on TERTp mutations in that, among cancer patients with TERTp mutations, only those carrying the TT genotype had a poor survival (Hazard ratio = 1.56, 95% confidence interval = 1.06-2.28); subgroup analyses by cancer type also supported these results. In conclusion, our findings suggest that rs2853669 could be important for assessing the prognostic value of TERTp mutations. Future large studies are required to further validate our results.


INTRODUCTION
Human cancer is a major global health problem. There were approximately 14 million new cancer cases in 2012 worldwide, and by 2025, the predicted global cancer incidence will rise to 20 million [1]. Cancer is the leading cause of death in China [2] and the second leading cause of death in the United States [3]. Although environmental risk factors are important in cancer pathogenesis, genetic predisposition also has a confirmed crucial role in the risk and prognosis of cancer. Notably, genetic alterations in the proximal promoter of the telomerase reverse transcriptase gene (TERT) are significantly associated with many cancer types [4].
The TERT gene encodes the key catalytic subunit of telomerase, which is important for the maintenance of chromosome stability [5]. Deregulation of TERT often results in abnormal telomerase activation and causes unlimited cell proliferation and even malignancies [4]. An increasing number of studies suggest that somatic mutations (e.g., -124C>T and -146C>T) and single nucleotide polymorphisms (SNPs) within the TERT promoter (TERTp) could influence the susceptibility and prognosis of human cancers [6][7][8][9][10][11][12]. In particular, the SNP rs2853669 (T > C), located at -245 bp from the TERT ATG site, is associated with the risk of various cancers such as breast cancer [6], lung cancer [13], and hepatocellular carcinoma [14]; however, these results remain inconclusive. Some studies also supported a prognostic effect of rs2853669 on cancer prognosis [15,16] whereas other studies have disproved this conclusion [17,18]. This SNP may have a modifying role in the prognostic value of TERTp mutations. For example, Ko et al. [11] showed that C carriers of rs2853669 in www.impactjournals.com/oncotarget/ Oncotarget, 2017, Vol. 8, (No. 31), pp: 50864-50872 Research Paper conjunction with TERTp mutations had a poor prognosis in liver cancer. By contrast, Nagore et al. [16] reported a poor melanoma survival only in cancer patients with both TERTp mutations and the TT genotype.
These conflicting results may be because of the insufficient power of individual studies. Therefore, we comprehensively searched the relevant literature and performed a meta-analysis to systematically evaluate the individual effect of rs2853669 on cancer risk and prognosis, and further assess its modifying impact on TERTp mutations with regard to the survival of cancer patients.

Association between rs2853669 and cancer risk
Thirteen eligible studies involving 16 datasets were pooled to evaluate the association between rs2853669 and cancer risk; these studies comprised 64,119 patients and 78,988 controls [6, 7, 12-14, 19, 23-29]. As shown in Table 1, four studies focused on breast cancer [6,12,19,29], two studies focused on lung cancer [13,27], one study focused on breast cancer and ovarian cancer [7], one study focused on prostate cancer and breast cancer [26], and the remaining studies each focused on a different cancer type [14,[23][24][25]28]. Nine studies were conducted in Caucasians [6, 12, 19, 23-26, 28, 29], three studies in Asians [13,14,27], and one study in a mixed-ethnic population [7]. Most studies were of high quality (i.e., scores >9), except for three studies (i.e., scores ≤9) [7,14,23]. In the overall analyses, rs2853669 was not significantly associated with cancer risk in the allelic, dominant, recessive, codominant, or additive model ( Table 2). Because of significant heterogeneity among the studies, we performed a metaregression analysis to explore the source of heterogeneity by considering as potential covariates the publication year, cancer type, ethnicity, sample size (>1000 vs. ≤1000) and quality scores (>9 vs. ≤9). The results did not identify any significant covariate contributing to heterogeneity in all genetic models (data not shown). We further conducted subgroup analyses by cancer type (Supplementary Figure  1a), and observed that the rs2853669 C allele seemed to    Figure 3).

Association between rs2853669 and cancer prognosis
There were eight eligible studies, which included 2,234 cases, used to evaluate the association between rs2853669 and cancer prognosis (Table 3) [15-18, 20, 21, 24, 25]. Of these, five studies focused on glioblastoma  [15,17,18,21,24], and the remaining studies focused on a separate cancer type [16,20,25]. Two genetic models, including dominant and codominant models, were used in these studies; therefore, we pooled the survival data, based on the two genetic models. The results were that rs2853669 was not associated with the overall survival (OS) of cancer in the dominant or TC vs. TT codominant model (Table 2 and Supplementary Figure 4a and 4b), and the sensitivities in these models exhibited robust results ( Supplementary Figure 5a and 5b). There seemed to be a significant association in the CC versus TT codominant model between rs2853669 and cancer OS (hazards ratio (HR) = 1.80, 95%CI = 1.09-2.97, P heterogeneity = 0.105, I 2 = 55.7%); however, the small number of studies (n = 3) limited its credibility (  Figure 5c) [25]. We then performed subgroup analyses by cancer type in the two genetic models, and found that rs2853669 also had no association with the OS of glioblastoma (Supplementary Figure 4). In addition, no publication bias was identified in these two genetic models (Table 2 and Supplementary Figure 6). Because only two studies assessing the prognostic effect of rs2853669 on cancer-free survival were identified in the literature, we failed to conduct a meta-analysis on this issue. However, these two studies did not find an association of rs2853669 with progression-free survival (PFS) [21] or with disease-free survival (DFS) [16] in cancer patients.

The modifying effect of rs2853669 on TERTp mutations
Eight studies comprising 1,775 cases were included to evaluate the modifying effect of rs2853669 on the prognostic value of TERTp mutations for cancer survival (Table 3) [8, 11, 15-18, 21, 22]. Of these, four studies focused on glioblastoma [15,17,18,21], and the other studies focused on bladder cancer [8], clear cell renal cell carcinoma [22], hepatocellular carcinoma [11], and localized invasive cutaneous melanoma [16]. Interestingly, we observed that TERTp mutations conferred worse OS only in cancer patients carrying the rs2853669 TT genotype (Figure 2a In hepatocellular carcinoma, the rs2853669 seemed to play an opposite role than in other cancer type. But the only one relevant study [11] hindered us to perform subgroup analysis to explore the potential effect of rs2853669 on this cancer type. However, in cancer patients who were rs2853669 C carriers, we did not observe any significant result in the analysis of OS (Figure 3a We also performed sensitivity analyses and found that the pooled HR was not materially changed for cancer OS in rs2853669 C carriers (Supplementary Figure 8b), but the pooled HRs were not stable for cancer OS in rs2853669 TT genotypes (Supplementary Figure 8a) or for cancer DFS ( Supplementary Figure 8c and 8d).
We observed that rs2853669 alone did not increase or decrease the risk and prognosis of cancer. However, when we performed subgroup analyses by cancer type, we interestingly found that rs2853669 may decrease breast cancer risk in the allelic model, which is in accordance with the findings of two published meta-analyses [30,31]. A possible explanation for the contradictions between the overall results and subgroup results is that rs2853669 may have a tissue-specific effect. Several studies have shown its function on affecting TERT expression and telomerase activity [8,21]. In addition, the TERTp mutations are well established to increase TERT expression and its activity, with a tissue-specific distribution in several cancers. Killela et al. [32] reported that the frequency of TERTp mutations is different in various cancer types: constantly self-renewing cancers, including breast cancer, harbor few of these telomere-maintaining mutations, whereas cancers arising from seldom self-renewing cells such as gliomas often harbor these mutations. Therefore, the role of rs2853669 likely depends on the frequency of TERTp mutations in different cancers, thereby demonstrating a tissue-specific effect on cancer risk. Several other cancers such as lung cancer and glioblastoma were also analyzed by stratification; however, each type of cancer only contained one or two studies, which unfortunately provided insufficient study numbers for a meta-analysis. Thus in the future, more large-scale studies are warranted to further elucidate the role of rs2853669 in different cancers.
Several recent studies [8, 11, 15-18, 21, 22] reported inconsistent results about the modifying effect of rs2853669 on TERTp mutations in cancer survival, so we also conducted a meta-analysis to address this issue. Results revealed that the poor prognosis contributed by TERTp mutations occurs only in cancer patients carrying the rs2853669 TT genotype; whereas in patients carrying the rs2853669 C allele (i.e., the TC or CC genotype), TERTp mutations showed no effect on the OS or DFS. Subgroup analyses based on cancer type was consistent with the overall analysis result. The rs2853669 polymorphism is located within a preexisting Ets2 transcription factor binding site in the promoter region of the TERT gene [8]; therefore, the C allele variant may impair the Ets2 binding site and then prevent c-Myc from binding to the TERT E-box. This factor decreases TERT expression and lowers telomerase activity, thereby blunting the detrimental effect of TERTp mutations [21,33,34].
Our study included 143,107 subjects and 2,234 cases for evaluating the effect of rs2853669 alone on cancer risk and prognosis, respectively, and included 1,775 cases for further evaluating the modifying effect of rs2853669 on TERTp mutations in the survival of cancer patients. A large sample size and systematic assessment could provide the exact role of rs2853669 in cancer onset and development. However, several limitations need to be addressed. First, many types of cancer were included in our meta-analysis, but the majority only contained one or two studies; therefore, the subgroup analyses were not fully implemented for each type of cancer. Second, one-way sensitivity analysis showed that the significant result obtained for the modifying effect of rs2853669 was not stable. This finding should therefore be treated with caution. Third, our pooled results may be biased by residual or unmeasured confounders in the original studies. Thus it also should be cautious about the prognostic role of rs2553669 in cancer, because age, TNM stages, and treatment were not fully adjusted in original studies.
In conclusion, our meta-analysis demonstrated that rs2853669 alone does not increase or decrease the overall risk and prognosis of cancer. Moreover, the prognostic value of TERTp mutations significantly depends on the rs2853669 status: cancer patients with TERTp mutations who carry the TT genotype have a poor survival. Our findings reveal the potential role of rs2853669 in carcinogenesis, and provide evidence for the clinical utility of the combination of rs2853669 and TERTp mutations as biomarkers of cancer prognosis. However, more cohort studies with refined designs are required to further validate our results.

Search strategy and eligibility criteria
We searched literatures through PubMed, Web of Science, and EMBASE up to July 2016. The search items included "rs2853669," "TERT or hTERT," "polymorphism or variant," and "cancer or tumor or carcinoma or neoplasm or malignancy". Our report was conducted in accordance with the guidelines of Preferred Reporting Items for Systematic Reviews and Meta-Analyses [35]. Eligible studies were included if they met the following criteria: (1) studies had a case-control or cohort design; (2) studies focusing on the association of TERT-rs2853669 with cancer risk or prognosis were published in English; and (3) studies reported the ORs or HRs with the corresponding 95%CIs, or studies provided sufficient information to calculate the ORs or HRs. If studies had overlapping case series, we used the study with the latest or largest sample size.

Data extraction and quality assessment
Two authors independently extracted data from each eligible study, and discrepancies were resolved through discussion and consensus. Extracted items included the first author, publication year, country, cancer type, sex, number of subjects and so on. We resolved any discrepancy by group discussion. For cancer risk, we assessed the quality of each study by using the quality assessment criteria (Supplementary Table 1) with scores ranging from "0" (i.e., worst) to "15" (i.e., best); scores ≤9 indicated a low quality and scores >9 indicated a high quality [36]. For cancer prognosis, we did not assign a quality scale for each study because there is no such score assessment having a general consensus in a prognostic meta-analysis for observational studies; we instead carried out the subgroup and sensitivity analyses to evaluate the potential effects of rs2853669, according to the previous meta-analyses [37,38].

Statistical analysis
We applied a random-effects model to pool data [39]. Adjusted effect measures were chosen preferentially, if available. When an individual study did not provide effect measures, we calculated the OR and 95% CI by the allele frequencies, and extrapolated the HR and 95%CI by the methods of Parmar [40] and Tierney [41]. Hardy-Weinberg equilibrium was re-examined by χ 2 test in controls. The heterogeneity was assessed by the Cochran's Q test and quantified by I 2 statistics, in which P < 0.10 or I 2 > 50% indicated significant heterogeneity. Meta-regression analysis was applied to investigate potential sources of heterogeneity. We also performed one-way sensitivity analyses to evaluate the stability of the pooled results. In addition, we performed the Begg's test [42] and Egger's test [43] to examine publication bias. All aforementioned statistical analyses were performed via Stata 12.1 software (StataCorp, College Station, TX, USA), using a two-sided P ≤ 0.05 as the significance level, unless otherwise specified.

Author contributions
L.C. conceived and designed this study. N.S. and Y. L performed literature search, study selection and data extraction. X.W. and J. P. conducted statistical analysis. N.S. and Y. Z. prepared tables and figures. N.S. and L.C. wrote the manuscript. All these authors completely consented with all the data in the study and approved the final manuscript. L.C. had the primary responsibility for final manuscript.