Tumor Necrosis Factor Alpha Extended Haplotypes and Risk of Gastric Carcinoma

Gastric carcinoma is one of the most common malignancies worldwide. Although decreasing in incidence in the past 50 years, it remains the second most common cause of cancer-related death. Estimates predict 1.1 million new cases in the year 2010 (1, 2). The Helicobacter pylori bacterium colonizes the human stomach and is the most recognized and well-known etiologic risk factor for gastric carcinoma (3). It possesses a unique array of features that makes it highly adapted to this ecological niche and causes a long-term infection of the gastric mucosa. The strong inflammatory response by the human host usually does not eradicate the bacteria then causing damage on the gastric epithelial cells (4). The infection first induces chronic superficial (non-atrophic) gastritis, which can progress through chronic atrophic gastritis, intestinal metaplasia, and dysplasia towards gastric carcinoma (5).

The tumor necrosis factor (TNF) α is a pleiotropic cytokine mostly produced by activated monocytes and macrophages, which play a key role in the inflammatory response (6, 7). Thus, it is a strong candidate for mediating the risk of developing immunologically related malignant disease. The TNFA gene, which encodes the TNFα cytokine, is located within the human leukocyte antigen class III region of the major histocompatibility complex on chromosome 6 (section 6p21). This region is highly polymorphic and contains a number of immune response genes such as lymphotoxin α (also known as TNFB) and the lymphotoxin β (also known as TNFC).

Several single-nucleotide polymorphisms (SNP) have been identified in the TNFA gene, mainly in the 5′-promoter region. Five microsatellites (TNFa, b, c, d, and e), occurring together with those SNPs in extended haplotypes, have also been described (8-10). Many studies have investigated the relation between TNF polymorphisms and the susceptibility to or outcome of various infectious (11, 12), inflammatory (13, 14), and neoplastic diseases (15-17). Although several promoter polymorphisms have been analyzed, most studies have focused on the TNFA-308G>A (rs1800629). Since the first reports of TNFA association with gastric carcinoma (18-22), the TNFA-308*A allele has emerged as an important candidate accounting for the increased risk of gastric carcinoma development (23, 24). Nevertheless, outcomes of the numerous studies on whether polymorphisms in or near TNF genes do underlie these associations are conflicting. While TNFA alleles have been associated with the genetic predisposition to differential secretion of TNFα, the functional assays described in the literature remain contradictory (25, 26).

Numerous association studies of disease susceptibility target candidate genes based only on their putative functionality. The polymorphisms are then analyzed independently of the genetic structure of the chromosome where they are located. An alternative strategy is the establishment of the haplotypic structure inferred from allelic variants at several loci. The main advantage of this alternative is that, with this genetic information, gene-disease association studies can be performed without prior assumptions on the functional relevance of specific variants.

Several studies have described associations between TNF microsatellites and infectious diseases as well as cancer. However, none have studied TNF microsatellites extended haplotypes and the risk of gastric carcinoma. In the present investigation, we performed a case-control study for the TNFA-308G>A polymorphism in order to determine the association with the risk of gastric carcinoma development. The subsequent aim was to establish the haplotypic structure associated with the TNFA-308G>A polymorphism in both cases and controls by genotyping five polymorphic microsatellite loci (TNFa, b, c, d, and e) flanking the TNFA gene over a 20 kb region. Our results indicate that the detected association between the pro-inflammatory TNFA-308*A allele and the increased risk of gastric carcinoma is at least partially influenced by linkage disequilibrium with an as yet unidentified locus.

A total of 1,221 subjects from the North of Portugal were enrolled in this study, comprising gastric carcinoma patients (n = 508) and controls (n = 713). The control group consisted of healthy blood donors (n = 318) and a representative sample of the non-institutionalized adult population of Porto (n = 395), Portugal, selected during the assembling of the EpiPorto cohort (mean age, 37 y; median age, 35 y; range,18-64 y; male:female ratio, 1.7:1). A detailed description of the selection procedures and participants was published previously (27). Patients with gastric carcinoma (mean age, 56 y; median age, 58 y; range, 26-90 y; male:female ratio, 1.4:1) had been diagnosed and treated by tumor resection at Hospital S. João, Porto, Portugal.

Genomic DNA from the blood samples of control individuals and from gastric carcinoma patients was isolated using standard proteinase K digestion and phenol/chloroform extraction. Genotyping of the TNFA-308G>A SNP was done by the 5′-nuclease PCR assay (TaqMan). TaqMan (SNP Genotyping Assays) primers, probes, and amplification conditions are available upon request. In a subset of 50 samples, the genotypes were confirmed by direct sequencing. The five TNF-linked microsatellites (TNFa, b, c, d, and e; Fig. 1) were genotyped using PCR and GeneScan in a total of 291 individuals (169 controls and 122 gastric carcinoma patients randomly selected from the total sample) and following methods previously described (28).

Evidence for deviation from Hardy-Weinberg equilibrium of alleles at individual loci was assessed by exact tests using GENEPOP (ref. 29; ftp://ftp.cefe.cnrsmop.fr/pub/pc/msdos/genepop/). Differences in the proportion of the TNFA-308G>A genotype between cases and controls was assessed using the χ² test. Crude odds ratios (OR) with 95% confidence intervals (95%`CI) were estimated by unconditional logistic regression analysis using SPSS (SPSS for Windows, Rel. 14.0.0. 2005, SPSS Inc.). Differences were considered to be significant at a P <0.05. All statistical tests were two-sided. Haplotypes defined by the association of alleles from the TNFA-308G>A SNP and the five microsatellite loci were inferred by the maximum likelihood approach implemented in the ARLEQUIN v.2.0. software package (30). Population differentiation was tested using ARLEQUIN v.2.0. Haplotypes associated with the TNFA-308*G allele were designated as follows: H1G-TNFa1b1c2-308*Gd5e1; H2G-TNFa6b5c1-308*Gd4e3; H3G-TNFa10b4c1-308*Gd4e3; H4G-TNFa11b4c1-308*Gd4e3; H5G-TNFa2b1c2-308*Gd4e1; H6G-TNFa7b4c1-308*Gd4e3; haplotypes associated with the TNFA-308*A allele were designated as follows: H1A-TNFa2b3c1-308*Ad2e3; H2A-TNFa4b5c1-308*Ad4e3; H3A-TNFa2b3c1-308*Ad1e3; H4A-TNFa4b5c1-308*Ad3e3; H5A-TNFa5b7c1-308*Ad5e3; H6A-TNFa3b5c1-308*Ad4e3; H7A-TNFa4b7c1-308*Ad4e3; H8A-TNFa2b1c2-308*Ad5e3.

Genotype frequencies of the TNFA-308G>A polymorphism in the control group did not deviate significantly from those expected under Hardy-Weinberg equilibrium (Table 1). In this group, the -308*G allele had a frequency of 0.811 and the -308*A allele had a frequency of 0.189. In the gastric carcinoma group the frequencies of the -308*G allele and the -308*A alleles were, respectively, 0.748 and 0.252. Genotype frequencies among the gastric carcinoma cases and controls are summarized in Table 2. Both control groups have similar TNFA-308G>A genotype frequencies. The comparison of TNFA-308G>A genotype frequencies between cases and controls confirmed the existence of a significant association of the TNFA-308*A allele and the increased risk of gastric carcinoma with an OR of 1.7 (95% CI, 1.3-2.2; Table 1).

The observed distributions for the TNFa, b, c, d, and e microsatellites are similar to those previously reported (13, 31). The TNFa microsatellite is the most polymorphic with 13 out of 14 known alleles present in our study population. TNFb, c, d, and e exhibited, respectively, 5, 3, 7, and 4 alleles. Table 1 shows the allelic frequencies at the five TNF microsatellite loci. TNF microsatellite alleles and the TNFA-308G>A SNP were phased into haplotypes by maximum likelihood. Haplotype inference disclosed a higher number of haplotypes associated with the TNFA-308*G allele than with the TNFA-308*A allele in both cases (75 and 22, respectively) and controls (84 and 19, respectively). This result supports the hypothesis that the TNFA-308*A allele is the most recent one. The frequencies of the TNFA-308*A–associated haplotypes are represented in Fig. 2. No significant differences could be observed relatively to the TNFA-308*G allele–associated haplotypes (designated H1G to H6G, Fig. 2). Regarding the TNFA-308*A allele–associated haplotypes (designated H1A to H8A), the most prominent difference was that found for the H1A haplotype, which was observed in 33.1% of the cases and in 12.5% of the controls. We have also found two other haplotypes (H7A and H8A) associated with the TNFA-308*A allele that are present only in the cases (although in very low frequencies) and three haplotypes present only in the control population (H3A, H4A, and H6A). A population differentiation test was performed showing that the control and the gastric carcinoma groups are significantly different (P < 0.005).

It is well established that gastric carcinoma depends on the combined effects of H. pylori infection, environmental factors, and host susceptibility. The TNFA-308G>A promoter SNP is one of the most frequently reported candidate host factors associated with the risk for gastric carcinoma. However, the increasing number of published studies has revealed a conspicuous lack of reproducibility among the various populations (23, 24). So far, it is not known whether the TNFA-308G>A SNP is acting as an etiologic susceptibility locus or if it is associated in linkage disequilibrium with a causal sequence variant that has not yet been identified. In order to help clarify the association between the TNFA-308G>A polymorphism and the risk of gastric carcinoma, we used a haplotype-based approach by genotyping five microsatellites surrounding that SNP.

Although a similar strategy using the same markers has been used for the study of susceptibility to a number of conditions like Chagas' disease (12), pancreatitis (32), Crohn's disease (13), Alzheimer's disease (33), and rheumatoid arthritis (14), this study provides the first description of TNF microsatellite allele frequencies and derived haplotypes in the study of gastric carcinoma susceptibility. Moreover, our TNFA-308G>A SNP association study was based on a larger sample size than any of the previously published studies. Relatively to this SNP, we have confirmed the established association (22) between individuals carrying the TNFA-308*A allele and increased risk for developing gastric carcinoma (OR, 1.7; 95% CI, 1.3-2.2). In fact, two recent meta-analyses (23, 24) identify the -308G>A as the single TNFA SNP displaying a statistically significant association with increased risk of gastric carcinoma. In those reports, evaluations of association with gastric carcinoma risk limited to anatomic and histologic subtypes or populations in Hardy-Weinberg equilibrium did not yield heterogeneous results and remained consistently strong. However, when studies considered Caucasian and Asian populations separately, the association was statistically significant only for the Caucasian. In their overall analysis, these studies did not rule out counter-examples in Caucasian populations like those reported by García-González et al. 2007 (34) and Kamangar et al. 2006 (35). The present understanding of the role of the TNFA-308G>A polymorphism on susceptibility to disease only emphasizes the importance of extending single-SNP association studies to haplotype-based approaches.

Our study shows that the TNFA-308*A–associated haplotypic structure is significantly different between cases and controls. Although the haplotype inference disclosed a higher number of haplotypes associated with the *G allele than with the *A allele in both cases and controls, no significant differences between groups could be observed in the TNFA-308*G haplotypic structure. In contrast, a population differentiation test showed that the control and gastric carcinoma groups are significantly different for the TNFA-308*A haplotypic structure as opposed to that of the TNFA-308*G. The most prominent difference was found for the H1A (TNFa2b3c1-308*Ad2e3) haplotype that showed a 3- fold increase in cases. We have also observed haplotypes associated with the TNFA-308*A allele that were found only in cases or controls. Taken together, these results suggest that the association between the promoter polymorphism and increased risk of gastric carcinoma is dependent on linkage disequilibrium. Thus, we cannot rule out the possibility that the TNF chromosomal region may contain a stronger genetic causal factor for gastric carcinoma development than any of the TNFA analyzed polymorphisms. The studies referred to above as using the TNF microsatellite haplotype-based approach have also found haplotypes that are more frequent in affected individuals than in controls, but none of those matched our susceptibility haplotypes.

In our search for a more suitable candidate gene and/or locus, we also did a TNF microsatellite single-allele association study (data not shown). Approximately 60% of the observed microsatellite alleles showed a higher frequency in the gastric carcinoma group than in the control group. Microsatellite markers were considered individually, and the association with gastric carcinoma risk was determined. We found higher magnitude associations between the TNF microsatellite markers and the increased risk of gastric carcinoma than that observed for the TNFA-308*A allele; however, only the TNFd2 allele was found to be significantly associated with the development of gastric carcinoma with an OR of 2.9 (95% CI, 1.4-5.7). We can conclude that the allele 2 of the TNFd microsatellite is also dependent on linkage disequilibrium and may be a better marker for increased risk of development of gastric carcinoma in our studied population.

In summary, the TNFA-308*A allele is associated with increased risk of gastric carcinoma development in our population. Additionally, we reported that this allele is in linkage disequilibrium with TNF microsatellite alleles and that a possible gastric carcinoma causal sequence variant is present in the same chromosomal region. This might represent a fundamental finding as it highlights the relevance of using haplotype-based approaches to provide a better understanding of susceptibility to gastric carcinoma and the consequent differences in incidence among populations.

No potential conflicts of interest were disclosed.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.