XPG gene polymorphisms and cancer susceptibility: evidence from 47 studies

Xeroderma pigmentosum group G (XPG) is a single-strand-specific DNA endonuclease that functions in the nucleotide excision repair pathway. Genetic variations in XPG gene can alter the DNA repair capacity of this enzyme. We evaluated the associations between six single nucleotide polymorphisms (SNPs) in XPG (rs1047768 T>C, rs2296147 T>C, rs2227869 G>C, rs2094258 C>T, rs751402 C>T, and rs873601 G>A) and cancer risk. Forty-seven studies were identified in searches of the PubMed, Scopus, Web of Science, China National Knowledge Infrastructure, and WanFang databases. Crude odds ratios (ORs) and 95% confidence intervals (CIs) were calculated using a fixed or random effects model. We found that rs873601 G>A was associated with an increased overall cancer risk (AA vs. GG: OR = 1.14, 95% CI = 1.06–1.24; GA/AA vs. GG: OR = 1.08, 95% CI = 1.02–1.15; A vs. G: OR = 1.06, 95% CI = 1.02–1.10). In a stratified analysis, rs1047768 T>C was associated with an increased risk of lung cancer, rs2227869 G>C was associated with a decreased risk of cancer in population-based studies, and rs751402 C>T and rs873601 G>A were associated with the risk of gastric cancer. Our data indicate that rs873601 G>A is associated with cancer susceptibility.


INTRODUCTION
There were an estimated 14.1 million new cancer cases and 8.2 million cancer-related deaths in 2012 worldwide [1,2]. Although recent advances in the diagnosis and treatment of various cancers have improved patient prognosis, most malignancies still impose a heavy burden on society. Cancer is a multifactorial, chronic disease caused by both endogenous (genetic, immune, and endocrine disorders) and exogenous factors (environmental carcinogens and unhealthy behaviors) [1]. Among these etiological factors, gene-environment interactions have been shown to play key roles in cancer development.
The maintenance of genomic integrity is essential for human health. However, DNA damage can occur due to exposure to various chemicals, environmental agents, and ultraviolet radiation. DNA damage can also occur naturally. For example, metabolic processes can generate compounds that damage DNA, which include reactive oxygen and reactive nitrogen species. There are five major DNA damage repair pathways in humans: nucleotide excision repair (NER), base excision repair, doublestrand break repair, mismatch repair, and homologous  [3]. Failure to properly repair DNA damage can lead to tumorigenesis. The versatile NER pathway is responsible for excising DNA lesions including cross-links, bulky adducts, thymidine dimers, alkylating damage, and oxidative DNA damage [3].
There are at least eight core functional genes in the NER pathway. These include Excision repair cross complementing group 1 (ERCC1) and Xeroderma pigmentosum group (XP) A-G. XPG, also known as ERCC5, is located on chromosome 13q22-q33 [4]. The XPG gene encodes a single-strand specific DNA endonuclease of 1,186 amino acids that cleaves the damaged DNA strand at the 3' end [5]. Defects in the XPG gene can impair DNA repair resulting in genomic instability and carcinogenesis [6]. Single nucleotide polymorphisms (SNPs) in the XPG gene have been associated with various cancers including colorectal [7], lung [8,9], gastric [10,11], and laryngeal [12]. However, different studies have achieved conflicting results. For example, Duan et al. found that rs2296147 T>C in XPG was associated with an increased risk of gastric cancer [13], but this association was not replicated in other studies [10,11]. The discordances might be attributed to the limited sample sizes of individual studies, different sources of controls, and ethnic variation. In this study, we performed a meta-analysis of the associations between six potentially functional SNPs: rs1047768 T>C, rs2296147 T>C, rs2227869 G>C, rs2094258 C>T, rs751402 C>T, and rs873601 G>A in the XPG gene and the risk of cancer.

Heterogeneity and sensitivity analysis
Study heterogeneity was observed for the association between rs1047768 T>C and overall cancer risk under homozygous, dominant, and allele contrast models (P = 0.010, P = 0.038, and P = 0.012, respectively); rs2094258 C>T under homozygous and allele contrast models (P = 0.025 and P = 0.015, respectively); rs751402 C>T under homozygous, recessive, dominant, and allele contrast models (P < 0.001, P = 0.006, P < 0.001, P < 0.001, respectively); and rs873601 G>A under a recessive model www.impactjournals.com/oncotarget (P = 0.035). These data indicated that the removal of any individual study from the analysis did not qualitatively change the pooled ORs (data not shown).

False-positive report probability (FPRP) analysis and trial sequential analysis (TSA)
All significant findings remained significant at a prior probability of 0.1, with all the FPRP values less than 0.20 with the exception of the population-designed studies of rs2227869 G>C (Table 4). TSA indicated that the cumulative z-curve crossed the trial sequential monitoring boundary, suggesting that the sample size was sufficient and that no further analysis was required to confirm the results (Figure 4).

DISCUSSION
The NER pathway is critical for the repair of bulky DNA lesions resulting from exposure to chemical carcinogens as well as ionizing radiation in order to maintain genomic integrity and prevent carcinogenesis [55]. Because the XPG gene is an indispensable component of the NER pathway, SNPs in XPG may alter the expression or function of XPG thereby modifying the risk of cancer. Most previous meta-analyses of the association between SNPs in XPG and cancer risk have focused on rs17655 G>C [56][57][58][59]. However, recent studies have shown that other SNPs in XPG may also be associated with cancer risk. For example, Chen et al. found that rs873601 G>A was associated with an increased risk of gastric cancer in a Chinese Han population [36]. Wang et al. found that rs751402 C>T was protective against breast cancer in Chinese Han women [47]. Additionally, the T allele of rs2296147 was associated with an increased risk of prostate cancer [35]. However, the results of previous studies have been inconsistent, possibly due to variations in the study populations and limited sample sizes. We therefore performed a meta-analysis of 47 studies to comprehensively evaluate the associations between six SNPs in XPG: rs1047768 T>C, rs2296147 T>C, rs2227869 G>C, rs2094258 C>T, rs751402 C>T, and rs873601 G>A and cancer risk.
The rs873601 G>A polymorphism is located in a miRNA binding site in the XPG gene. Thus, it may alter XPG expression by modulating the miRNA-mRNA interaction, which could play a role in carcinogenesis [10]. We demonstrated that rs873601 G>A was significantly associated with overall cancer risk. Individuals with the AA genotype of rs873601 had a     1.14-fold higher risk of cancer compared to individuals with the GG genotype. Similar results were obtained for gastric cancer. The A allele of rs873601 was previously shown to result in reduced mRNA expression of XPG in both adjacent normal gastric cancer tissue and normal cell lines in a recessive manner [10]. These findings provide insight into the molecular mechanisms by which the AA genotype of rs873601 may increase the risk of gastric cancer. The rs751402 C>T polymorphism is located in the E2F1/YY1 binding and response site in the proximal promoter region of XPG [60]. This variant might reduce the DNA repair capacity of XPG by disrupting the DNA binding motifs and altering transcription factor affinities [47]. In our study, rs751402 C>T was significantly associated with overall cancer risk. The TT genotype of rs751402 was associated with an 18% increase in cancer risk compared to the CC genotype. Moreover, a significant association was observed between rs751402 C>T and gastric cancer risk under all genetic models. The rs751402 C>T polymorphism is likely to influence cancer risk by regulating XPG expression, but its effect on XPG function is not yet clear [47].
The rs2094258 C>T polymorphism is located in a transcription factor binding site in the 5' region of the XPG gene. We found that the association between rs2094258 C>T and overall cancer risk was borderline significant. Individuals with the TT genotype of rs2094258 had a 9% higher risk of cancer compared to those with the CC genotype. However, the association was not significant in gastric cancer, indicating that it may not impact gastric cancer risk. Significant associations were observed among some subgroups for all other selected SNPs. We found that the C allele of rs1047768 may increase the risk of lung cancer. Moreover, the C allele of rs2227869 significantly reduced cancer risk in population-based studies. No statistically significant association was observed between rs2296147 T>C and overall cancer risk.
Although we found significant associations between SNPs in the XPG gene and cancer risk, our study had several limitations. First, although Egger's tests showed no obvious publication bias, some bias was unavoidable since only studies published in English and Chinese were included in our meta-analysis. Second, we observed significant heterogeneity in some of our analyses, which is a common drawback of a meta-analysis. Third, due to a lack of sufficient individual data, we were unable to perform multivariate analysis with adjustment for potential confounding factors such as tobacco use, alcohol consumption, and other carcinogenic factors.
Our study is the first meta-analysis of the association between the six selected SNPs in XPG gene and cancer risk. The results indicate that the AA genotype of rs873601 increases overall cancer risk. Additionally, rs751402 C>T and rs873601 G>A were associated with gastric cancer risk. Finally, rs1047768 T>C was found to confer susceptibility to lung cancer. Further epidemiological investigations with larger sample sizes are warranted to validate our findings. Functional studies are also required to elucidate the mechanisms by which these SNPs modify cancer risk.

Study identification
We searched multiple databases including PubMed, Scopus, Web of Science, CNKI, and the WanFang database using combinations of keywords such as "XPG", "polymorphism", and "cancer" as well as synonyms "Xeroderma pigmentosum group G, ERCC5 or Excision repair cross complementing group 5", "variant or variation", and "tumor, neoplasm, or carcinoma". Human studies published before December 20, 2016 in either English or Chinese were included. The reference lists in eligible studies and review articles were examined in order to identify additional relevant studies. In cases of study population overlap, the study with the largest sample size was selected.

Inclusion and exclusion criteria
All studies included in this analysis were required to meet the following criteria: (1) study of the associations between any of the six potentially functional SNPs: rs1047768 T>C, rs2296147 T>C, rs2227869 G>C, rs2094258 C>T, rs751402 C>T, and rs873601 G>A in the XPG gene and cancer risk; (2) case-control study; and (3) sufficient genotype data available to calculate ORs and 95% CIs. The exclusion criteria were: (1) studies conducted in the same or overlapping population and (2) review article or conference report.

Data extraction
Key information was independently extracted from eligible studies by two investigators and included the following items: the first author, year of publication, type of cancer, country, ethnicity, control source, number of cases and controls, the quantity of each genotype in cases and controls, minor allele frequency (MAF), and the Hardy-Weinberg equilibrium (HWE) test P value for the control subjects. Disagreements regarding these items were resolved through discussion.

Statistical analysis
Chi-square tests were used to test deviation from HWE in the study control groups. Genetic associations between the six selected SNPs in the XPG gene and cancer risk were assessed using the crude ORs and corresponding 95% CIs under homozygous, heterozygous, recessive, dominant, and allele contrast models. Heterogeneity between studies was assessed using the Q and I 2 values. A random effects model was adopted to calculate the pooled OR and 95% CI in the case of P het < 0.1 or I 2 > 50%. Otherwise, a fixed effects model was applied. Stratified analyses were conducted by ethnicity (Asians and Caucasians), source of control [population-based (PB) or hospital-based (HB)], and cancer type. Sensitivity analyses were performed to assess the influence of the individual studies on the pooled OR by sequentially removing one study at a time and recalculating the pooled OR. Egger's tests were used to evaluate publication bias. FPRP analysis [61,62] and TSA were performed as described previously [63]. All statistical analyses were performed using the STATA 12.0 software (Stata Corporation, College Station, TX, USA). All statistics were two-sided. P values < 0.05 were considered statistically significant.

CONFLICTS OF INTEREST
The authors declare that there are no conflicts of interest.

FUNDING
This study was supported by a grant from the Natural Science Foundation of Guangdong Province (No. 2015A030310324).