Association between TNFA Gene Polymorphisms and Helicobacter pylori Infection: A Meta-Analysis

Background Several host genetic factors are thought to affect susceptibility to Helicobacter pylori infection-related diseases, including tumor necrosis factor (TNF)-α. Previous studies have evaluated the association between TNFA gene polymorphisms and H. pylori infection, but the results were inconclusive. We conducted this meta-analysis to clarify the association between TNFA polymorphisms and H. pylori infection. Methods Published literature within PubMed, Embase, and the Cochrane Library were used in our meta-analysis. Data were analyzed with the Stata13.1 software package using pooled odds ratios (ORs) with 95% confidence intervals (CI). Results A total of 24 studies were included in our study. The TNFA -308G>A polymorphism was associated with decreasing H. pylori infection (AA vs. AG+GG, OR = 0.64, 95% CI = 0.43–0.97; AA vs. GG, OR = 0.64, 95% CI = 0.43–0.97). A significantly decreased risk was also found for -1031T>C polymorphism (CC vs. CT+TT, OR = 0.61, 95% CI = 0.44–0.84). -863C>A polymorphism was associated with increasing risk of H. pylori infection (AA+AC vs. CC, OR = 1.47, 95% CI = 1.16–1.86; A allele vs. C allele, OR = 1.40, 95% CI = 1.14–1.72). There was no significant association between -857C>T polymorphism and H. pylori infection. When stratified analysis was conducted on H. pylori infection detection methods, -857C>T and -863C>A polymorphisms were associated with H. pylori infection for the non-ELISA subgroup. When stratified for ethnicity or study design, -863C>A significantly increased the risk and -1031T>C decreased the risk for the Asian subgroup and hospital-based subgroup. Conclusion Results of our meta-analysis demonstrate that TNFA -308G>A and -1031 T>C polymorphisms may be protective factors against H. pylori infection, and -863C>A may be a risk factor, especially in Asian populations. Further studies with larger sample sizes are required to validate these results.


Introduction
Helicobacter pylori, one of the most common pathogens worldwide, has proven to be associated with gastritis, peptic ulcers, gastric cancer, and mucosa-associated lymphoid tissue (MALT) lymphoma [1]. Some individuals when exposed to H. pylori may escape from persistent infection, even if they live in regions where H. pylori infection is highly prevalent. Previous studies indicate that host factors may play an important role during H. pylori infection [2]. Host cytokines and their gene polymorphisms may be host factors that affect an individual's susceptibility to H. pylori-related diseases [3,4]. H. pylori infection can induce production of some cytokines, including interleukin (IL)-1, -2, -4, -6, -8, -10, -17, interferon (IFN)-β, and tumor necrosis factor (TNF)-α [5]. These host cytokines affect the occurrence and development of the gastric mucosal inflammatory response, which is a key event of H. pylori infection [6].
TNF-α, a host cytokine induced by H. pylori in gastric mucosal, is supposed to be involved in H. pylori infection [7]. TNF-α is encoded by the TNFA gene, which is clustered on the short arm of human chromosome 6 (6p21.3), between HLA-B and HLA-DR [8]. The TNFA gene is known to have four single nucleotide polymorphisms in the regulatory sequences that may affect its expression: -308G>A, -857C>T, -863 C>A, and -1031T>C. TNF-α can inhibit gastric acid secretion and influence the immune response, which may be associated with persistent H. pylori infection [9].
A number of studies have focused on the association between TNFA gene polymorphisms and H. pylori-related diseases [10][11][12]. Previous meta-analysis have demonstrated that TNFA gene polymorphisms are associated with gastric cancer and have no association with peptic ulcers [13,14]. Many studies conducted on gastric diseases have investigated the relationship between TNFA gene polymorphisms and H. pylori infection simultaneously; however, results from these studies are inconclusive. Therefore, we performed this meta-analysis to clarify the association between TNFA gene polymorphisms and H. pylori infection.

Search strategy
Pubmed, Embase and Cochrane Library databases were searched up to August 2015. The following terms were used for searching: (TNF-α OR tumor necrosis factor-α OR TNF-A OR tumor necrosis factor-A OR TNF-alpha OR tumor necrosis factor-alpha) AND (polymorphism OR polymorphisms OR SNP) AND (Helicobacter pylori OR H. pylori OR HP). Searches were restricted to English. In order to identify potentially relevant studies, the reference lists of retrieved articles were also examined. In addition, the related citations of results in Pubmed were searched. We also contacted the authors to get more data as possible as we can. When more than one of the same case series was involved in several studies, only the study with the largest sample sizes was selected in our meta-analysis.

Selection criteria
Studies were included if the following conditions were met: (1) A relationship between the TNFA gene polymorphisms and H. pylori infection was described; (2) Case-control designed; (3) Objective H. pylori infection detection methods were used; (4) Sufficient genotype data to calculate the odd ratios (ORs) with a 95% confidence interval (CI) was available.

Data extraction and quality appraisal
The following data were collected from each study: first author's name; year of publication; ethnicity; country; study design; number of cases and controls; H. pylori infection detection methods; and genotyping method. The Newcastle-Ottawa scale (NOS) [15] was used to assess the quality of studies included, according to three main criteria: selection of cases and controls; comparability of cases and controls; and exposure to risk factors. NOS scores ranged between 0 and 9 stars. Studies with a score of seven stars or greater were considered to be of high quality, while those that scored five stars or less were considered low quality. Two authors (XDS and YYX) of this meta-analysis independently extracted all information and conducted the quality appraisal. Disagreements were resolved by discussion with other authors.

Statistical analysis
Statistical analysis was performed using STATA 13.1 (STATA Corp, College Station, TX, USA). Pooled OR and corresponding 95% CI was used to measure the strength of associations between TNFA gene polymorphisms and H. pylori infection. Heterogeneity among studies was assessed by the Q-test and I 2 statistics. P < 0.10 or I 2 > 50% indicated significant heterogeneity [16]. If significant heterogeneity exists, the ORs were pooled with a random effect model. Otherwise, a fixed effect model was selected. Subgroup analyses were conducted based on H. pylori infection detection methods (ELISA or non-ELISA methods (including bacterial culture, rapid urease test (RUT), polymerase chain reaction (PCR), urea breath test (UBT), Helicobacter pylori stool antigen test (HpSAT) and histological examination)), study designs (hospitalbased (HB) or population-based (PB)) and ethnicity (Asian or Caucasian). Publication bias was examined using a Begg's funnel plot or Egger's plot, and the significance level was set at 0.05 for both. Hardy-Weinberg equilibrium was assessed by the χ 2 test for goodness of fit, with a P-value less than 0.05 considered a significant deviation.

Study characteristics
A total of 230 articles were retrieved from the initial search. From these, 164 articles were assessed for ineligibility after reading titles and abstracts, and 47 articles with insufficient data were excluded after reading the full texts. In addition, 5 papers were included through references. According to our inclusion and exclusion criteria, 24 articles were used for this metaanalysis finally . The study selection process is summarized in Fig 1. Of the studies included, 17 concerned -308 G>A, nine concerned -857C>T, four concerned -863C>A, 10 concerned -1031T>C; 14 were on Asians, five were on Caucasians, one was on Africans, four were on mixed ethnicity (Table 1).
Variable H. pylori infection detection methods were used in the studies included in this meta-analysis (Table 1). These methods were different in sensitivity and specificity, and various methods could cause various results of diagnosing H. pylori infection. ELISA method had special features during H. pylori epidemiological survey, so we performed a subgroup analysis for ELISA and non-ELISA methods. The TNFA -308G>A and -1031T>C polymorphisms had no association with H. pylori infection for ELISA or non-ELISA subgroups. -857C>T polymorphism significantly decreased the risk of H. pylori infection in allelic model for the non-ELISA subgroup, and -863C>A polymorphism increased the risk in dominant and allelic models for the non-ELISA subgroup. We also conducted a subgroup analysis on ethnicity. The results showed that the -863C>A polymorphism had a significant association with H. pylori infection in dominant and allelic models for the Asian subgroup, and -1031T>C polymorphism was associated with H. pylori infection in recessive model for the Asian subgroup too. -308G>A and -857C>T polymorphisms did not have significant association with H. pylori infection for  Asian or Caucasian subgroups. A stratified analysis on study design was also performed, and the results indicated that -863C>A significantly increased the risk and -1031T>C decreased the risk for HB subgroups. All results of the meta-analysis are shown in Table 2.

Heterogeneity and sensitivity analysis
Significant heterogeneity was observed in the TNFA -857C>T and -1031T>C polymorphism results. We then conducted sensitivity analysis to identify the results by omitting one study in turn. Heterogeneity decreased when a study by Saijo et al. [34] was excluded for the -857C>T polymorphism and a study by Ando et al. [21] was excluded for the -1031T>C polymorphism. The pooled ORs were not significantly altered in all investigated SNPs by sequential omission of included studies.

Publication bias
Begg's funnel plot of SNPs did not reveal any evidence of significant publication bias (Fig 3)

Discussion
Results of our meta-analysis indicate that TNFA -308G>A and -1031T>C polymorphisms might be associated with a decreasing risk of H. pylori infection, while the -863C>A polymorphism could increase the risk of H. pylori infection. When stratified analysis was conducted on ethnicity in our meta-analysis, only -863C>A and -1031T>C polymorphisms had significant association with H. pylori infection in Asian population. -308G>A and -857C>T polymorphisms had no significant association with H. pylori infection in Asian or Caucasian population. TNFA polymorphisms did not show up in a genome wide association study in Europeans  [41], which was consistent with the results of our meta-analysis in Caucasian subgroup. The association between TNFA polymorphisms and H. pylori infection may be more meaningful in Asian population. When stratified for study design, -863C>A significantly increased the risk and -1031T>C decreased the risk for the HB subgroups. Significant heterogeneity existed in meta-analysis results of -857C>T and -1031T>C polymorphisms, and heterogeneity decreased after excluding the study of Saijo et al. [34] for -857C>T polymorphism and the study of Ando et al. [21] for -1031T>C polymorphism, which suggests that the above two studies might be the source of heterogeneity. Subjects of the study by Saijo et al. were all healthy Japanese transit company employees whose ages ranged from 35-60 years, including 413 men and only 5 women. Specific gender, age and occupational composition of the subjects might lead to the difference between the study by Saijo et al. and other including studies. 41% of the subjects of the study by Ando et al. suffered from gastrooesophageal reflux disease, which might be the source of heterogeneity between the study by Ando et al. and other including studies. No significant difference with pooled ORs was shown in the sensitivity analysis. In our study selection process, two studies on -238G>A, one study on -555G>A, and one study on -806C>T investigated the association with H. pylori infection, and all reported no significant association. We did not conduct a meta-analysis in three TNFA SNPs [20,27,40].
Numerous methods have been developed for diagnosing H. pylori infection, such as bacterial culture, RUT, PCR, UBT, histological examination and serum antibody detection [42]. Bacterial culture, RUT, UBT and histological examination can be affected by biopsy location, bacterial density and morphology, fastidious growth requirements, and so on [43]. Serology could not distinguish between current and past H. pylori infection, but an IgG-positive sample can show that the host is susceptible to H. pylori [44]. Since variable H. pylori infection detection methods were used in studies included in our meta-analysis, which could cause different results of diagnosing H. pylori infection, we conducted subgroup analyses (ELISA vs. non-ELISA methods) to verify the association between TNFA polymorphisms and H. pylori infection. A significant association was found between the TNFA -863C>A polymorphism and H. pylori infection for the non-ELISA subgroup in dominant and allelic models, and between -857C>T and H. pylori infection for the non-ELISA subgroup in allelic model. -308G>A and -1031T>C polymorphisms had no association with H. pylori infection for ELISA or no-ELISA subgroups.
Gastric acid secretion is supposed to be inhibited by TNF-α, which was produced by macrophages in the gastric submucosa [45]. Since the TNFA -863A polymorphism is related to high transcriptional promoter activity [46], carriers of the TNFA -863A polymorphism may have a significantly higher level of TNF-α than those with the C allele. High concentrations of TNF-α could directly suppress gastric acid secreted by parietal cells, and simultaneously inhibit the functions of gastrin-stimulated enterochromaffin-like cells to decrease the secretion of histamine, which can elevate gastric secretion [46]. In addition, a high level of TNF-α could amplify inflammatory responses by activating neutrophils, T cells, and B cells. Low levels of gastric acid, and an aggressive inflammatory response, can facilitate the colonization of the gastric mucosa with H. pylori from the gastric antrum to the corpus [9]. This might increase the risk of developing atrophic gastritis, or even gastric cancer.
Although there are papers reporting that -308G>A and -1031T>C polymorphisms are also related to high transcriptional promoter activity [47][48][49], our meta-analysis revealed that -308G>A and -1031T>C polymorphisms could decrease the risk of H. pylori infection. This difference may be linked with the sample size and ethnicity. Moreover, TNF-α possibly regulates H. pylori infection through other mechanisms. Further studies are needed to confirm the mechanisms.
There were some limitations to this study. Firstly, most of the studies included for -857C>T, -863C>A, and -1031T>C polymorphisms were conducted on Asian populations, so further research with other ethnic populations are needed. Secondly, only a low number of studies were included. Therefore, more studies involving much larger sampling sizes should be conducted. Thirdly, it is also possible that language bias might exist, as our meta-analysis only included articles published in English.

Conclusions
This meta-analysis is the first to investigate the association between TNFA polymorphisms and H. pylori infection. Our conclusion suggests that TNFA -308G>A and -1031T>C polymorphisms may be associated with a decreasing risk of H. pylori infection, and the -863C>A polymorphism may be associated with an increased risk of H. pylori infection. There was no significant association between -308G>A and H. pylori infection for Asian or Caucasian subgroups. TNFA -863C>A and -1031T>C polymorphisms had significant associations with H. pylori infection for Asian and HB subgroups, and -857C>T and -863C>A polymorphisms had significant associations with H. pylori infection for non-ELISA subgroup. Further studies with different ethnicities and larger sample size are required to validate our results.