|Home | About | Journals | Submit | Contact Us | Français|
Helicobacter pylori (H. pylori), a known risk factor of gastric cancer, rarely colonize the deeper portion of normal gastric glands, where the mucus is rich in alpha-1,4-linked N-acetylglucosamine (A4GN) capped O-glycans, that strongly inhibit H. pylori growth in vitro.
We investigated the association between genetic variation in the O-glycan transferase encoding gene (A4GNT) and H. pylori infection and gastric cancer risk using a Polish population-based case-control study (273 gastric cancer patients and 377 controls).
A haplotype at the rs2622694-rs397266 locus was associated with H. pylori infection, with the A-A haplotype associated with higher risk compared with the most frequent G-G haplotype (odds ratio 2.30; 95% confidence interval 1.35 – 3.92). The association remained significant after correction for multiple tests (global P value: nominal 0.002, empirical 0.045). Neither this haplotype nor the tagSNPs were associated with overall gastric cancer risk.
A4GNT genetic variation may be relevant to H. pylori infection, but not to gastric cancer risk.
Gastric cancer remains the second leading cause of cancer-related death worldwide (1). Helicobacter pylori (H. pylori) infection is the most important acquired risk factor for gastric cancer (2). The past decade has seen many studies on genetic susceptibility to gastric cancer, particularly for genes in the inflammatory and detoxifying pathways (3-5). Few studies of genetic susceptibility have focused on host defenses against H. pylori within the gastric mucus (4, 5).
The gel layer covering the gastric mucosa is the major reservoir of H. pylori. An uneven distribution of H. pylori among mucins of varied physiochemical properties, as observed by electron microscopy (6), provides clues to understand how the host environment may influence H. pylori survival. H. pylori rarely colonize the deeper portions of the normal gastric mucosa (6, 7), where the mucins are rich in O-glycans capped with terminal alpha-1,4-linked N-acetylglucosamine (A4GN) (8, 9). Glycans possessing the A4GN residue, but not those without, suppressed H. pylori growth in vitro (10, 11). The transfer of A4GN to beta-Gal residues with alpha1,4-linkage, forming GlcNAc alpha-1, 4-Gal beta-R structures, is mediated by a transferase encoded by the gene A4GNT (8). We hypothesized that genetic variation in A4GNT, which could lead to varied expression or function of this transferase, may be related to the ability of H. pylori to establish colonization, and consequent gastric cancer risk.
A population-based case-control study on gastric cancer was conducted in Warsaw, Poland between 1994 and 1996, as described (12, 13). Cases (n=464) were patients newly diagnosed with histologically confirmed gastric cancer. Controls (n=480) were randomly selected from the computerized population registry and frequency matched to the cases by sex and age (±5 years). Blood samples were obtained from 305 (65.7%) cases and 427 (90.0%) controls. Genomic DNA was available from 273 cases and 377 controls. Informed consent was obtained from each participant. The study was approved by Institutional Review Boards at the US National Cancer Institute, the Cancer Center and Institute of Oncology of Health in Warsaw, Poland, and Regional Ethics Committee of Karolinska Institutet in Sweden.
A total of 59 single nucleotide polymorphisms (SNPs) spanning the A4GNT gene and flanking region (100kb) with minor allele frequency ≥ 5% were identified from the HapMap2 – CEU population database (http://www.hapmap.org/). SNPs were tagged using the Carlson approach such that a resulting tagSNP had r2 ≥ 0.8 with all other SNPs grouped in the same bin (14). A total of 59 SNPs were grouped into 9 bins, yielding 9 tagSNPs for assay by the SNPlex Genotyping System (Applied Biosystems 3730 DNA Sequencer) at Core Genotyping Facility of National Cancer Institute, MD. Two (rs2622723, rs329379) failed during the genotyping process, since they were singleton-tags, lacking information about other SNPs. The remaining seven tagSNPs (rs996432, rs405265, rs11928535, rs329386, rs2246945, rs2622694, rs397266), which carry information on 57 SNPs across the 100kb region, were available for final analysis. Details about assays, primers, probes, and procedures are available on the National Cancer Institute's SNP500 website (http://snp500cancer.nci.nih.gov).
Serum levels of IgG antibodies against H. pylori whole cell antigen and antibodies against CagA were measured using ELISA, as described (15).
Hardy–Weinberg equilibrium was tested using Pearson's χ2-test in cancer-free controls. In single locus analysis, we compared different genotypes, with the most common genotypes among H. pylori-negative persons or controls served as references, and performed the Armitage trend tests. Because circulating CagA antibody is able to reflect infection even with negative values in the whole cell assay (16, 17), we defined H. pylori infection by positivity in anti-H. pylori and/or CagA assays. Odds ratios (ORs) with 95% confidence intervals (95% CIs) derived from unconditional logistic models were used to assess relative risks. All estimates were adjusted for sex and age. Firth's penalized maximum likelihood estimation was used in case of data separation (18). A priori, we also performed stratified analyses by anatomic location (cardia / non-cardia) and histological classification (intestinal / diffuse) of the tumour. To account for increased type I errors of multiple testing, statistical significance was assessed by empirical P values derived from Westfall and Young permutation (n=10,000) with minimum P and step-down method(19), where case-control status of all subjects were randomly permuted, then the set of seven SNPs analyses were performed for each permutation dataset, and the minimal P value of the seven analyses were recorded. The distribution of the 10,000 minimal P values obtained from 10,000 permutation datasets was used to derive the empirical significance of the observed test statistic (empirical P value equals to the percentile of the observed P in the distribution of the 10,000 minimal P values).
Haplotypes were inferred with comparison groups (infected vs. uninfected, or cancer cases vs. controls) jointly by an Expectation-Maximization algorithm, and analyzed by a “sliding-window” approach with varied window sizes ranging from 2 to 4 tagSNPs (20). Rare haplotypes, i.e. < 1% among the controls, were combined as one group. The probabilities, inferred from the EM algorithm, of having certain haplotypes for each individual were used as weights in a logit binomial model with sandwich covariance (21). The most common haplotype among uninfected or controls was used as a reference for each sliding window. Global P values were used to assess the difference in haplotype profiles between comparison groups. Permutations (n=10,000) were conducted to adjust global P values of multiple tests for all sliding windows. All analyses were conducted using the SAS 9.2 (SAS Institute, Cary, NC) package. (A SAS macro for sliding-window haplotype construction and permutation is available from the author)
All seven tested SNPs were in Hardy-Weinberg equilibrium (Table 1). Age and gender distributions were comparable between the case (mean age 63.4 ± 10.5, male 67.8%) and control (mean age 63.0 ± 10.3, male 65.0%) groups, as these were frequency matching factors. Compared with the controls, gastric cancer cases were more likely to have lower education, smoke more, and have a familial history of gastric cancer (data not shown). The majority of cases had the intestinal histological type (69.1%) and originated from the non-cardia region of the stomach (72.2%).
SNP rs2622694 heterozygous genotype AG was associated with a 49% reduced risk of H. pylori infection compared with the most common genotype AA (OR 0.51, 95% CI 0.26-1.01, Table 2). A similar magnitude of reduced risk was observed for the homozygous variant genotype GG (OR 0.52, 95% CI 0.21-1.31). For SNP rs397266, the heterozygous genotype AG conferred a 68% higher risk of H. pylori infection than the most common genotype GG. All carriers (n=29) of the homozygous variant genotype AA had H. pylori infection. Armitage trend test revealed a P value of 0.003, which remained significant after correction for multiple tests (0.036).
Among the fifteen sliding windows, three (rs2622694-rs397266, rs2246945-rs2622694-rs397266 and rs329386- rs2246945-rs2622694-rs397266) had haplotype profiles that differed between the infected and uninfected groups (Table 3). These three windows all pointed to a significant effect of haplotype at loci rs2622694-rs397266. Compared with the most common (G-G) haplotype, haplotype A-A was associated with a significantly higher risk of H. pylori infection (OR 2.30, 95% CI 1.35-3.92, global P 0.0019). This association remained significant after correction for multiple tests (empirical global P value 0.045).
Overall, none of the tested SNPs was associated with gastric cancer risk (Table 4). When limiting to the intestinal type of gastric cancer, the SNP rs2622694 G allele was associated with a reduced risk compared with the A allele (nominal trend test P value 0.029, data not shown); GG carriers had only half the risk compared with AA carriers (OR 0.53; 95% CI 0.28-1.01). However, the association was non-significant after correction for multiple tests (empirical trend test P value 0.24). For the diffuse type gastric cancer, despite the small sample size (n = 46), SNP rs996432 T allele was associated with a higher risk compared with C allele (TT vs. CC: OR 2.51, 95% CI 1.01-6.20), but became non-significant after correction for multiple tests (data not shown). Stratified analyses by tumour location did not reveal materially different results between cardia and distal gastric cancer (data not shown).
Similarly as for single locus analysis, haplotypes at the seven loci were not associated with gastric cancer risk overall (data not shown). In stratified analyses, the haplotypes constructed from rs329386-rs2246945-rs2622694 loci were borderline associated with the intestinal type of gastric cancer (global P value 0.057). Compared with the most frequent haplotype C-A-A (59.2% among controls), haplotype C-C-G (31.4% among controls) was associated with a non-significant (25%) reduced risk for intestinal type gastric cancer (OR 0.75, 95% CI 0.56 - 1.01), but haplotype G-C-G (6.4% among controls) was associated with a 50% risk reduction (OR 0.50, 95% CI 0.27-0.94). However, the global P value was non-significant in the permutation test (empirical P value 0.24), though. For the diffuse type of gastric cancer, none of the haplotypes was associated with risk (data not shown). Stratifying the analysis by tumour location did not reveal material differences.
Host genetic variation was estimated to contribute 57% of variation in acquisition of H. pylori infection (22). No prior epidemiological study had examined the role of variation in A4GNT in relation to risk of H. pylori infection or gastric cancer. Evidence from in vitro studies demonstrates that O-glycans capped with the residue A4GN alpha-1, 4Gal beta is capable of exerting an inhibitory effect on H. pylori growth by inhibiting the biosynthesis of cholesteryl-alpha-D-glucopyranoside, a major H. pylori cell wall component (10, 11, 23). The transfer of A4GN to betaGal with alpha1,4-linkage, forming A4GN alpha-1, 4-Gal beta-R structures, is mediated by a transferase encoded by A4GNT 9. In contrast to an expected beneficial effect of the A4GNT gene product, two small clinical series studies showed up-regulated expression of A4GNT mRNA in patients with gastric cancer (24). One explanation for this discrepancy may be that there exist functionally impotent subclasses of the A4GNT-encoded enzyme. This is supported by the observation that the expression of the A4GNT enzyme, but not of the A4GNalpha1-4Galbeta-R (a mucous gland specific mucin specific glycan that suppresses H. pylori growth), was up-regulated amongst patients with H. pylori gastritis and decreased to normal level after H. pylori eradication (7). The up-regulation of A4GNT enzyme accompanied by unchanged amount of A4GNalpha1-4Galbeta-R suggests that there exist subclasses of A4GNT with little or no transferase activity. Hosts with up-regulated expression of non-functional A4GNT subclasses would be less capable of preventing H. pylori from establishing colonization and, consequently, could have more severe clinical outcomes, such as gastric cancer (24).
Examining into the sequence of A4GNT (8) that had been used in the previous in vitro studies (10, 11), in which the A4GNT, producing a glycan suppressing H. pylori growth, had an alanine at position 218 (corresponding to a C allele at rs2246945), consistent with our finding that the C allele rs2246945 was associated with reduced risk for H. pylori infection. Since the effect of A4GNT only has been demonstrated in one sequence that came from the same research group, functional studies on subclasses of A4GNT encoded by other sequences are warranted.
In our study, A4GNT variation was associated with H. pylori seropositivity but was not associated with overall gastric cancer risk. The proportion of the risk haplotype A-A at loci rs2622694-rs397266 was 17% among the uninfected group in this study. Such a relatively low prevalence requires a larger sample-size study to examine for moderate or modest effects of A4GNT on gastric cancer risk.
In summary, this preliminary study, suggests that the A-A haplotype at rs2622694-rs397266 may be related with an increased risk for H. pylori infection, but no strong effect on gastric cancer risk was observed.
We thank Dr. Chuen Seng Tan for statistical consultancy.
Financial support: Supported in part by the Intramural Research Program of NIH, Division of Cancer Epidemiology and Genetics, a grant from the Swedish Research Council, and R01 GM 63270 from the National Institutes of Health.
No conflict of interests declared.
CONFLICT OF INTEREST
Guarantor of the article: Weimin Ye
Specific author contributions:
Conception and design: Weimin Ye, Zongli Zheng, Wong-Ho Chow, Meredith Yeager;
Acquisition of data: Wong-Ho Chow, Meredith Yeager, Weimin Ye;
Analysis and interpretation: Zongli Zheng, Meredith Yeager, Stephen J. Chanock, Wong-Ho Chow, Weimin Ye;
Drafting of manuscript: Zongli Zheng, Weimin Ye, Wong-Ho Chow;
Critical revision of manuscript: Yanbin Jia, Lifang Hou, Christina Persson, Jolanta Lissowska, Martin Blaser;
Statistical analysis: Zongli Zheng, Weimin Ye
Potential competing interests: None.