MGMT Leu84Phe Polymorphism Contributes to Cancer Susceptibility: Evidence from 44 Case-Control Studies

Background O6-methylguanine-DNA methyltransferase is one of the few proteins to directly remove alkylating agents in the human DNA direct reversal repair pathway. A large number of case-control studies have been conducted to explore the association between MGMT Leu84Phe polymorphism and cancer risk. However, the results were not consistent. Methods We carried out a meta-analysis of 44 case-control studies to clarify the association between the Leu84Phe polymorphism and cancer risk. Results Overall, significant association of the T allele with cancer susceptibility was verified with meta-analysis under a recessive genetic model (P<0.001, OR=1.30, 95%CI 1.24-1.50) and TT versus CC comparison (P=0.001, OR=1.29, 95% CI 1.12-1.50). In subgroup analysis, a significant increased risk was found for lung cancer (TT versus CC, P=0.027, OR=1.67, 95% CI 1.06-2.63; recessive genetic model, P=0.32, OR=1.64, 95% CI 1.04-2.58), whereas risk of colorectal cancer was significantly low under a dominant genetic model (P=0.019, OR=0.84, 95% CI 0.72-0.97). Additionally, a significant association between TT genetic model and total cancer risk was found in the Caucasian population (TT versus CC, P=0.014, OR=1.29, 95% CI 1.05-1.59; recessive genetic model, P=0.009, OR=1.31, 95% CI 1.07-1.61), but not in the Asian population. An increased risk for lung cancer was also verified in the Caucasian population (TT versus CC, P=0.035, OR=1.62, 95% CI 1.04-2.53; recessive genetic model, P=0.048, OR=1.57, 95% CI 1.01-2.45). Conclusions These results suggest that MGMT Leu84Phe polymorphism might contribute to the susceptibility of certain cancers.


Introduction
Over the past decades, there has been an increasing understanding of the disease process in human carcinoma. It is now well established that carcinoma can be initiated by DNA damage from UV exposure, ionizing radiation, environmental chemical agents, and byproducts of cell metabolism. Normally, when DNA damage occurs, DNA repair systems recognize the DNA lesions, excise them, and restore the DNA to maintain genome stability and integrity [1]. However, if genetic alterations occur in genes encoding DNA repair proteins, the DNA repair process may be impaired, potentially contributing to an increased risk for developing cancers.
The O 6 -methylguanine-DNA methyltransferase (MGMT) is one of the most important proteins in the DNA repair process. It is a 207 amino acid zinc-bound protein which is encoded by MGMT gene located on chromosome 10 at 10q26 and spans approximately 300kb [2]. It has been shown that MGMT has basic methyl-transferring activity [3] and plays a central role in the cellular defense against alkylating agents within the human DNA direct reversal repair pathway.
Also known as O 6 -alkylguanine-DNA alkyltransferase (ATase, AGT, or AGAT), MGMT protein can directly remove alkyl or methyl adducts from the O 6 position of guanine to an internal cysteine residue at codon 145 of the protein [4]. By which, it protects cells against potential DNA alkylation damage from endogenous and exogenous alkylating species such as cigarette consumption, environmental contaminants, and diet [5]. Additionally, it seems that MGMT lacks the ability to dealkylate itself. MGMT therefore can take part only in a single reaction, in which it is irreversibly inactivated [6]. Hence, the reaction should be stoichiometric rather than catalytic. The MGMT expression shows significant variation not only among different body tissues [7], but also among individuals in the same specific tissue [8]. Though the causes of the interindividual differences in MGMT protein expression levels remain unclear to date, functional polymorphisms in the MGMT gene may have the potential to affect DNA repair capacity. Because of its important role in human DNA direct reversal repair pathway, MGMT has attracted significant attention as a candidate susceptibility gene for cancer.
A large number of molecular epidemiology studies have been carried out to assess the roles of the MGMT polymorphisms in various types of cancer, including lung cancer, head and neck cancer, and colorectal cancer [9,10,11,12,13,14,15,16,17,18,19,20,21]. The MGMTLeu84Phe substitution is the most widely studied polymorphism in MGMT due to a (C->T) transition at nt.262 (MGMT Leu84Phe, rs12917). However, numerous studies on the association of the MGMT Leu84Phe polymorphism with cancer risk have yielded inconsistent results and even partially contradictory conclusions. Several factors may contribute to the discrepancies among different studies. The differences of tumor sites, ethnicities or sample size may all cause the bias of the result of each individual study.
Since single studies may have been underpowered in clarifying the associations of MGMT polymorphisms with cancer susceptibility, to address the controversy among literatures, in the present study we conducted an evidencebased quantitative meta-analysis of the association between the MGMT Leu84Phe polymorphism and susceptibility to cancer.

Identification and eligibility of relevant studies
To identify all studies that explored the association of MGMT Leu84Phe polymorphism with cancer risk, we carried out a computerized literature search of the PubMed database (up to July 20, 2012), using the following key words: 'MGMT,' 'polymorphism,' and 'cancer,' without any restriction on language or publication year. The searched papers were read and assessed for their appropriateness of including. All references cited in the articles were also read to identify relevant publications. Eligible studies should meet two criteria: (1) case-control studies; and (2) genotype frequencies in both cancer cases and controls were available. Exclusion criteria were as follows: (a) not relevant to MGMT Leu84Phe polymorphism; (b) not case-control study; (c) control population included malignant tumor cases; and (d) article was a review or duplication of previous publication.

Data extraction
The data was extracted by two investigators (Jun Liu and Fei Chen) from each article independently. Discrepancies were not solved until consensus was reached on every item. From each study, the following data were collected: author's name, year of publication, country of origin, racial descent, cancer type, source of the control population, genotyping methods, matched factors as well as adjusted factors, number of cases and controls, genotype frequencies for cases and controls, characteristics of cancer cases, and controls. If data of subpopulation from different ethnicities was available in one paper, we took each subpopulation as an individual study.

Statistical analysis
Hardy-Weinberg equilibrium (HWE) for each study was assessed using goodness-of-fit test (x 2 of Fisher's exact test) only in control groups [22]. Crude odds ratios (ORs) with 95% confidence intervals (CIs) were calculated to evaluate the strength of association between MGMTLeu84Phe polymorphism and cancer susceptibility. In the overall and subgroup meta-analysis, we evaluated the associations of genetic variants with cancer risk in homozygous genetic contrast (TT vs. CC), dominant geneticmodel (CT+TT vs. CC), recessive genetic model (TT vs. CT+CC) and T allele vs C allele. The significance of the pooled OR was assessed by the Z-test (P<0.05 shows a significant association). In addition to overall meta-analysis, stratified analysis on ethnicity (Asians, Caucasians, and the other ethnicities group) and tumor site was also performed A x 2 -based Q-test was carried out to assess the heterogeneity of the ORs [23]. If the result of heterogeneity test was P>0.1, ORs were pooled according to the fixed-effects model (Mantel-Haenszel model). Otherwise, the random-effects model (DerSimonian and Laird model) was applied [24]. The Egger regression test and Begg-Mazumdar test were utilized to measure the potential publication bias [25]. All statistical tests were conducted with the software STATA v. 10.0 (Stata Corporation, College Station, TX, USA) using twoside P values.

Characteristics of studies
The preliminary literature search yielded 46 articles that explored the association of MGMT polymorphisms with the susceptibility to different cancers. However, six articles [26,27,28,29,30,31] irrelevant to MGMT Leu84Phe polymorphism and four articles [32,33,34,35] without detailed MGMT Leu84Phe genotypes data were excluded. In addition, three articles [10,36,37] were included by literature reading and manual searching. Therefore, 39 articles [9][10][11][12][13][14][15][16][17][18][19][20][21] were identified and included in the final meta-analysis (Figure 1). Five papers [14], [18] [56], [59], and [61] presented data including more than one racial populations and each subgroup in these studies was taken as a separate study. Therefore, a total of 44 studies from 39 papers (18938 cancer patients and 28796 controls) were included. All of the cases were confirmed by histological or pathological examination. A classic polymerase chain reaction-restriction fragment length polymorphism (PCR-RFLP) assay was adopted only in 7 of 44 studies and some other genotyping methods were also used widely, such as Taqman, sequencing and Illumina SNP genotyping BeadLab platform. All the genotyping methods are valid for the present meta-analysis. All studies stated that the gender status and the age range were matched between case and control population. The characteristics of included studies are listed in Table 1. All studies were case-control studies or nested case-control studies within prospective cohort studies, including 9 upper aerodigestive tract squamous cell carcinoma (UADT SCC) studies, 7 colorectal cancer studies, 5 lung cancer studies, 4 brain cancer studies, 3 prostate studies and 13 studies on "other cancers". There were 15 studies of Caucasian ethnicity, 13 studies of Asian ethnicity, and 16 studies of "mixed ethnicities" (including studies of American, Australian, Black and unspecified population, which cannot be categorized as a unique group since it is mixed). The detailed MGMT Leu84Phe genotype distributions and allele frequencies for cancer cases and controls were presented in Table 2. The equilibrium of genotypes in the controls was consistent with HWE in all but five studies [9,10,17,21,45]

Quantitative synthesis
In overall analysis, significant associations between the T allele and cancer risk were found under the recessive genetic model (P=0.001, OR=1.28, 95%CI 1.11-1.47) and TT versus CC comparison (P=0.001, OR=1.28, 95% CI 1.11-1.47). And, after we excluded those studies whose genotype equilibrium was not consistent with HWE, significant associations between the T allele and cancer susceptibility was also uncovered under the recessive genetic model (P<0.001, OR=1.30, 95%CI 1.24-1.50) and TT versus CC comparison (P=0.001, OR=1.29, 95% CI 1.12-1.50). However, no significant association was found in the dominant genetic model (TT+TC versus CC) and T versus C comparison. These results were summarized in Table  3.   Table 3.
In most of the available studies, there was no difference of MGMT Leu84Phe genotype/allele distribution among different ethnicities. We also performed stratified analysis by ethnicity (Caucasians, Asians, and mixed ethnicities), and by ethnicity and tumor site together ( As shown in Table 3 and Table 4, heterogeneity widely existed in the present meta-analysis under the dominant genetic mode and T versus C comparison but not under the homozygous comparison and recessive genetic model.

Publication bias
Begg's funnel plot and Egger's test were utilized to evaluate the publication bias of the literature. As shown in Figure 2, the contour-enhanced funnel plot for publication bias did not reveal any evidence of obvious asymmetry in allele contrast (T allele versus C allele), and, as expected, the Egger's test did not provide any obvious evidence for bias (t=0.12, P=0.902).

Discussion
This meta-analysis including a total of 18938 cancer patients and 28796 controls from 44 independent genetic studies implies that MGMT Leu84Phe polymorphism might contribute to the susceptibility of certain cancers Although the global analysis indicated that the T variant allele might increase the risk of cancer, the subgroup metaanalysis showed significant association at only two tumor sites (colorectal cancer and lung cancer) and two ethnicity subgroups (Caucasian subgroup and mixed ethnicities subgroup). This phenomenon suggests that the MGMT Leu84Phe polymorphism may play differing roles in cancerogenesis at different sites or in different ethnicities because of variability in genetic backgrounds [62].
Since cancer is a complex disease, it is highly possible that any single genetic factor has only weak effects on an individual's phenotype. It has been reported that the interaction of different combinations of polymorphisms in the same gene or between and among different genes might together have a pronounced effect on cancer risk [63,64,65]. Studies by Li et al. [66,67] have shown that MGMT is a transcriptional suppressor of ER-dependent signaling upon repair of the O 6methylguanine lesion and that the Lue84 and Ile143 residues lie in close proximity to three conserved leucines of the LXXLL ER-interacting helix. Therefore, it is possible that the ERdependent signalling could be differentially mediated by the variant 84Phe and 143Val residues. Some studies [9,10,13,40,42,48,49,54] have tried to investigate the combined effects of Lue84Phe, Ile143Val, and other polymorphisms in MGMT on cancer risk. Because the available data were not compatible, we could not evaluate the combined effects of MGMT Leu84Phe and Ile143Val on cancer susceptibility in our meta-analysis.
It is well established that genetic factors may play an important role in the development of tumors. However, there is no doubt that environmental factors such as alcohol consumption, cigarette use, and aging also participate in tumorigenesis. Several studies [11,39,42] reported that heavy cigarette smoking could aggravate the effects of MGMT variants on cancer risk. However, Chae et al. [10] did not find the same results. Li et al. [40] found that both drinking and smoking enhance genetic variants' effects on bladder cancer risk. It should be noted that alcohol consumption and cigarette use may play different roles at different tumor sites because of the different levels of alkylating agents and different tissue exposure concentrations. Unfortunately, owing to a lack of studies restricted to populations only exposed to alkylating agents, we could not obtain enough original data to further estimate the effects of the gene-environment interactions on cancer susceptibility. We note several limitations in the present study. First, there was wide heterogeneity due to the nature of our meta-analysis, and the results should be interpreted with caution. Second, our  results were based on unadjusted information, and the lack of original data limited estimation of the effect of confounding factors on cancer risk. Notably, confounding factors such as sex, age, alcohol drinking, smoking, and socioeconomic status may alter the association of genetic variants with cancer susceptibility. Third, the number of eligible studies in the subgroup analysis was limited. Subsequently, some subgroup meta-analysis might not have enough statistical power to accurately evaluate the association between the MGMT Leu84Phe polymorphism and cancer risk. More importantly, haplotype analysis has been regarded as a much better approach in genetic association research. However, since more detailed individual information on genotypes of the other polymorphisms of MGMT was unavailable, we were not able to conduct linkage disequilibrium and haplotype analysis in this study.
In conclusion, we observed several significant associations of the MGMT Leu84Phe polymorphism with cancer susceptibility. MGMT Leu84Phe variants may increase lung cancer risk, especially in Caucasians, but reduce colorectal cancer risk, indicating some differences among different tumor sites. In addition, MGMT Leu84Phe variants may increase cancer risk in Caucasians and in the mixed ethnicities group, which suggests an appreciable difference among different ethnic populations. Further well-designed study with greater sample size will be helpful in clarifying the haplotypes, genegene and gene-environment interactions on MGMT polymorphisms and tissue-specific cancer risk in ethnicity specific populations, and further mechanistic studies are warranted to elucidate the exact functional roles of MGMT variants.