Association Between hMLH1 Promoter Methylation and Risk of Gastric Cancer: A Meta-Analysis

Background: Human mutL homolog 1 (hMLH1) is located on chromosome 3q21-23. As a classic tumor suppressor gene, many researchers have studied the association between hMLH1 promoter methylation and gastric cancer, but their conclusions were not always consistent. Therefore, we performed a meta-analysis to make a more integrated and precise estimate of the associations. Method: PubMed, EMBASE, and Cochrane Library were retrieved without language restrictions. Data were analyzed by Review Manager 5.2 and Stata 12.0 software. Odds ratio (OR) with 95% confidence interval (95%CI) was used to assess the statistical associations. Result: A total of 39 studies published before January 20, 2018 were included in this study. The results indicated that the frequency of hMLH1 promoter methylation in gastric cancers was substantially higher than that in non-cancer controls (OR = 7.94, 95%CI = 4.32–14.58, P < 0.001). Furthermore, hMLH1 promoter methylation had considerable associations with lymph node metastasis, microsatellite instability (MSI), and low expression of hMLH1 protein (OR = 1.53, 95%CI = 1.04–2.26, P = 0.03; OR = 15.33, 95%CI = 9.26–25.36, P < 0.001; OR = 37.86, 95%CI = 18.03–79.50, P < 0.001, respectively). No association was found between hMLH1 promoter methylation and Lauren classification or Helicobacter pylori (HP) infection status. Conclusion: The present study provides evidence that promoter methylation of hMLH1 is a major causative event in the occurrence and development of human gastric cancer.


INTRODUCTION
Gastric cancer, also known as stomach cancer, continues to be a vital heath threat as the fifth leading cause of cancer and the third leading cause of death from carcinoma globally according to World Health Organization in 2014 (Stewart and Wild, 2014). Owing to inadequate early diagnosis and unclear pathogenesis, gastric cancer was considered to be a high-mortality disease and its 5-year survival rate was reported to be <10% (Orditura et al., 2014). Although the pathogenic mechanisms have not been fully elucidated, some factors including Helicobacter pylori (HP) infection, smoking, and excessive drinking, have been identified to contribute to the tumorigenesis of gastric cancer. In addition, epigenetic silencing of tumor suppressor genes was also thought to play an important part in the genesis of gastric cancer (Tahara and Arisawa, 2015). DNA methylation is the most common epigenetic mechanism. It is catalyzed by a family of DNA methyltransferases (DNMT) that transfer a methyl group from S-adenyl methionine (SAM) to the fifth carbon of a cytosine residue to form 5-methyl cytosine (5 mC), which is unstable and can spontaneously deaminate to form thymine, thereby affecting gene expression (Moore et al., 2013). Much evidence showed that hypermethylation of normally unmethylated CpG islands in the promoter regions of tumor suppressor genes was strongly related to carcinomas, for instance, BRCA1 promoter methylation in breast cancer (Zhang and Long, 2015), GSTP1 promoter methylation in prostate cancer (Jerónimo et al., 2001), hMLH1 methylation in gastroenteric tumors . hMLH1 is one of the human mismatched repair (MMR) genes, which is located on chromosome 3q21-23. As a classic anti-oncogene, the protein encoded by this gene is a component of the DNA mismatch repair pathway which can effectively repair mismatched bases and prevent the accumulation of DNA damage.
During the past decades, the associations between hMLH1 promoter methylation and the risk or clinicopathological characteristics of stomach cancer have been reported by many researchers. However, the conclusions were not always consistent and some results were unconvincing because of the small sample size. Therefore, we performed a meta-analysis to clarify the role of hMLH1 gene promoter methylation in the tumorigenesis and development of gastric cancer.

Literature Search Strategy
PubMed, EMBASE, and Cochrane Library were retrieved to obtain literatures concerning the association between gastric cancer and hMLH1 promoter methylation without language restrictions. We used the terms as follows: ("hMLH1" or "human mutL homolog 1") and ("methylation") and ("stomach" or "gastric") and ("cancer" or "neoplasms" or "carcinoma"). The search results were updated until January 20, 2018. In addition, we also performed manual search for other relevant literatures.

Selection Criteria
The following criteria were used in selecting eligible articles: (a) articles dealing with the association of hMLH1 promoter methylation with gastric cancers; (b) case-control studies. The major reasons for exclusion of studies: (a) reviews, letters, or case-only articles (b) articles with insufficient data or duplicated data.

Data Extraction and Quality Assessment
Articles were screened first by reading titles and abstracts, then two reviewers (Y Shi and P Ye) read the entire articles that seemed to fit the inclusion criteria and extracted basic information including the first authors' names, publication years, countries, methylation detection methods, sample sizes, and number of methylation in cases and controls from every eligible study. All included studies concerning hMLH1 methylation and gastric cancers risk were evaluated by the modified Newcastle-Ottawa scale (NOS) assessment (Jadad et al., 1996) (available at http:// www.ohri.ca/programs/clinical_epidemiology/oxford.asp). The latest NOS assessment for case-control studies consists of seven items of methodology which are grouped into three major categories: cases and controls selection, comparability of cases and controls, and ascertainment of exposure. The total score ranges from 0 to 8, and studies with more than four points are considered as qualified. Disagreement was resolved by discussion and consensus.

Statistical Analysis
Data were analyzed by Review Manager 5.2 and Stata 12.0. The strength of the association between hMLH1 promoter methylation and gastric cancers risk or clinicopathologic characteristics was assessed by pooled OR with corresponding 95% CI. Chi-square test based Q-test and I 2 -test were performed to assess heterogeneity among studies. It indicated a lack of heterogeneity if P > 0.10 and I 2 < 50%, then the pooled OR would be calculated by using the fixed-effect model in line with the Mantel-Haenszel method. Otherwise, the randomeffect model would be used according to the DerSimonian-Laird method. The subgroup analysis was further conducted based on different ethnicities, types of controls, specimen materials, and methods of detecting methylation. The stability of the pooled result was evaluated by sensitivity analysis and potential publication bias was assessed by Begg's test.

Meta-Analysis Results
Firstly, we analyzed the association between hMLH1 promoter methylation and risk of gastric cancer. Significant heterogeneity  of studies (I 2 = 68%, P < 0.001) was detected by X 2 test, so we employed random-effect model. In general, our result showed that the hMLH1 methylation frequency in gastric cancer was obviously higher compared with noncancer controls (OR = 7.94, 95%CI = 4.32-14.58, P < 0.001, Figure 2). We explored the sources of heterogeneity through subgroup analysis about the ethnicity, type of controls, specimen materials, and methods of detecting methylation ( Table 2). There was no significant heterogeneity between different races (Asians subgroup and Caucasians subgroup), different specimen materials (tissue subgroup or other materials subgroup), or different methylation detection methods (MSP subgroup and other methods subgroup). However, remarkable heterogeneity between subgroup whose controls were autologous and subgroup whose controls were heterogeneous has been observed (I 2 = 61.7%, P = 0.11). Then we explored the associations between hMLH1 promoter methylation and clinicopathological characteristics of gastric cancer. In short, there were significant statistical associations between hMLH1 methylation and lymph node metastasis (OR = 1.53, 95%CI = 1.04-2.26, P = 0.03, fixed-effect model), microsatellite status (OR = 15.33, 95%CI = 9.26-25.36, P < 0.001, fixed-effect model) and low hMLH1 protein expression (OR = 37.86, 95%CI = 18.03-79.50, P < 0.001, fixed-effect model) in gastric cancer patients, but not with Lauren classification (OR = 1.48, 95%CI = 0.86-2.55, P = 0.16, fixed-effect model) or HP infection status (OR = 1.18, 95%CI = 0.69-2.01, P = 0.54, fixed-effect model). The results were listed in Table 3

Sensitivity Analysis and Publication Bias
We conducted a sensitivity analysis to evaluate the stability of our result by sequentially omitting every study from pooled analysis. The sensitivity analysis confirmed that the result was stable since omission of each single study could not significantly alter the pooled OR (Figure 8). We applied Begg's funnel plot to assess potential publication bias of these eligible articles. The shape of funnel plot was symmetric and P = 0.561, which indicated that no obvious publication bias was found. The funnel plot for evaluating the association of hMLH1 promoter methylation with stomach cancer risk was shown in Figure 9.

DISCUSSION
Mismatch repairing is essential to ensure DNA replication fidelity, and mismatch repair deficiency will increase the chances of DNA mutation, which is important to the development of FIGURE 2 | Forest plot concerning hMLH1 methylation and gastric cancer risk. The squares and horizontal lines represent the corresponding OR and 95% CI for each study. The area of the squares reflects the weight of each study. The diamond represents the pooled OR and 95% CI. Random-effect model was used.
Frontiers in Physiology | www.frontiersin.org  a tumor from normal cells (Thomas et al., 1996). Mismatch repair genes are a group of highly conserved housekeeping genes playing a major role in the maintenance of genetic and epigenetic stability (Li, 2008). hMLH1 gene is one of the most widely studied. In recent years, some studies on cell lines suggested that methylation of hMLH1 gene promoter is directly related to lack of hMLH1 protein expression, which is an important event in initiation of many kinds of sporadic tumors (Kane et al., 1997;Yao et al., 2004). Previous studies have investigated the association between aberrant hMLH1 promoter methylation and gastric cancer, yet the results were not consistent probably due to different ethnic groups, types of control, methylation detection methods and specimen materials. In order to obtain consistent results, we performed a meta-analysis involving 2,182 cases of tumors and 2,319 controls. The results showed that aberrant methylation of hMLH1 promoter in stomach cancer tissues or bloods was positively associated with gastric cancer risk. Moreover, the frequency of hMLH1 methylation was higher in gastric cancer with lymph node metastasis, microsatellite instability and absence of hMLH1 protein expression, suggesting that hMLH1 promoter methylation might play a critical role in gastric cancer initiation and progression.
We confirmed that the frequency of hMLH1 promoter methylation in gastric cancer was 7.94-fold higher than that in control groups, meanwhile, the expression of hMLH1 protein substantially decreased in stomach cancer patients with hMLH1 hypermethylation. Our conclusion was similar to some results reported in other types of carcinomas (Mitchell et al., 2002;Han et al., 2016). hMLH1 promoter methylation often leads to transcriptional silencing accompanied by down-regulation of mRNA expression, resulting in decrease of hMLH1 protein expression and mismatch repair dysfunction which contribute to tumorigenesis. Because of the heterogeneity of the studies, we conducted a subgroup analysis to explore the sources of heterogeneity. We found that there was heterogeneity between the autologous controls group and the heterogeneous controls group. The pooled OR was higher when the heterogeneous samples from non-cancer patients were used as the control group. There may be two explanations: (1) Gastric cancer is a systemic disease, certain changes may occur in healthy tissues more or less.
(2) Different sampling methods among the studies: some used adjacent normal tissues as their autologous controls while the others chose remote normal tissues. We also found that hMLH1 methylation was more frequent in gastric cancer patients with lymph node metastasis, indicated that hMLH1 methylation may be implicated in the invasion and metastasis of gastric cancer.
Our meta-analysis also demonstrated that aberrant methylation of hMLH1 was closely related to microsatellite instability. MSI was first found in hereditary nonpolyposis colorectal cancer (HNPCC) (Aaltonen et al., 1993), while gastric cancer possessed the highest prevalence of MSI (Ottini et al., 1997;Keller et al., 1998). It comprises length mutations in tandem oligonucleotide repeats which was believed to be caused by the inability of the MMR protein to fix a DNA replication error (Lynch and de la Chapelle, 2003). Indeed, MSI can be a molecular hallmark of mismatch-repair-deficient-tumors and even serve as a tool for the classification of gastric cancer (Simpson et al., 2001). We didn't find associations between      hMLH1 methylation and HP infection or Lauren classification, upon which researchers had different views. Nevertheless, the outcomes might be due to small sample sizes and need to be confirmed by more studies with larger samples in the future. The sensitivity analysis and publication bias analysis results demonstrated that this meta-analysis was stable and had no obvious publication bias. However, our meta-analysis might still have some limitations. First of all, there was significant heterogeneity among these studies which were used to analyze the prevalence of hMLH1 promoter methylation in gastric cancers, but we could not provide a good solution about sources of heterogeneity; Also, there may be differential effects in hMLH1 methylation among different races, but these eligible studies did not contain all races, more researches are also needed to determine whether our outcomes are in consistent with studies about other ethnicities; Thirdly, we were not able to evaluate the associations between hMLH1 promoter methylation and other clinicopathological features because of insufficient data, prospective population-based studies are necessary for further research.
Above all, this is the first meta-analysis focused on the association between aberrant hMLH1 promoter methylation and gastric cancer, which provides evidence that silencing of the hMLH1 gene by promoter hypermethylation is a major causative event in the occurrence and development of human gastric cancer. Nevertheless, more efforts are still needed to be made before regarding hMLH1 promoter methylation as a potential diagnostic or prognostic biomarker.

AUTHOR CONTRIBUTIONS
The literature searching, data extraction, statistical analysis, and paper writing were conducted by PY and YS. AL reviewed the manuscript. All authors approved the final version of the manuscript.

FUNDING
This work was supported by grants from Natural Science Foundation of Hubei Province (2015CKB741).