The VEGF -634G>C promoter polymorphism is associated with risk of gastric cancer

The case-control study consisted of 171 patients with newly diagnosed and histologically confirmed gastric cancer at The University of Texas MD Anderson Cancer Center (Houston, TX) between February 1990 and July 2008. The 353 cancer-free controls subjects were selected from an ongoing molecular epidemiology study of the head and neck cancer since 1995 [23] by frequency matching by age, sex and ethnicity. All cancer-free control subjects were recruited from persons who were not hospital patients, nor seeking health care but accompanying the patients visiting the clinics, and who were genetically unrelated to the cases. The study protocol and informed consent form used in this study were approved by our institutional review board.

Genomic DNA was extracted from the buffy coat fraction of each blood sample by using a Blood Mini Kit (Qiagen, Valencia, CA) according to the manufacturer's instructions. DNA purity and concentrations were determined by spectrophotometric measurement of absorbance at 260 and 280 nm by UV spectrophotometer.

The selected three TGFB1 SNPs [-509C>T (rs1800469) in the promoter, +869T>C (rs1800470) and +915G>C (rs1800471) in the exon 1] and three promoter VEGF SNPs [-1498T>C (rs833061) in the promoter and -634G>C (rs2010963) in the 5'-untranslated region, and VEGF +936C>T (rs3025039) in the 3'-untranslated region] were genotyped using polymerase chain reaction-restriction fragment length polymorphism (PCR-RFLP) method. These six SNPs are relatively common (a minor allele frequency > 0.05) in Caucasian populations. Genotypes of TGFB1 SNPs were determined as previously described [24]. Assays on the VEGF -1498T>C (rs833061), -634G>C (rs2010963), and +936C>T (rs3025039) SNPs were previously reported [17]

For the PCR-RFLP assay, the genotyping was performed without knowing the subjects' case and control status, and almost the same number of cases and controls were assayed in each 96-well PCR plate. Two research assistants independently read the gel pictures and performed the repeated assays, if they did not reach a consensus on the tested genotype. In addition, 10% of the samples were randomly selected to perform the repeated assays, and the results were 100% concordant.

The χ² tests were used to compare the distribution of demographic variables and selected risk factors between cases and controls. The Hardy-Weinberg equilibrium (p²+2pq+q² = 1), where p is the frequency of the variant allele and q = 1-p, was tested by a goodness-of-fit χ² test to compare the observed genotype frequencies with the expected genotype frequencies in cancer-free controls. Unconditional logistical regression analysis was used to calculate odds ratios (ORs) and their 95% confidence intervals (CIs) with and without adjustment for age, sex, ethnicity, smoking status and alcohol use. Age was also dichotomized according to the median age (57 years) of the controls. The logistic regression model was also used for the trend test of the genotypes with different number of the minor alleles. The 2-linkage disequilibrium (LD) program and Proc ALLELE in SAS/Genetics software were used to detect the LD of any pair of SNPs. Proc HAPLOTYPE in SAS/Genetics software using the expectation-maximization (EM) algorithm was applied to generate maximum likelihood estimates of haplotype frequencies. Haplotype were combined when their frequency was <5%. The joint effects of the TGFB1 and VEGF SNPs and their interactions with smoking and drinking on gastric cancer risk were also evaluated. Finally, we calculated the false-positive report probability (FPRP) to test for false-positive associations, using prior probabilities of 0.0001, 0.001, 0.01, 0.1, and 0.25. The OR was set close to the observed value obtained in our study, and a probability of <0.2 was considered a noteworthy association. All statistical tests were two-sided, with a P value of 0.05 considered significant, by using SAS software (version 9.1; SAS Institute, Cary, NC).

This analysis included 171 gastric cancer cases and 353 cancer-free controls. Because we used frequency matching, there were no significant difference in distributions of age, sex and ethnicity between the cases and controls (P = 0.135, 0.444 and 0.486, respectively). For the tumor types, more than 50% of the cases had intestinal type, about 125 had difuse type, and the remaining was mixed. There were more current and former smokers in cases (19.9% and 39.2%, respectively) than in controls (13.6% and 38.5%, respectively), but the difference was not statistically significant (P = 0.126). However, there were significantly more alcohol drinkers among cases than among controls (P = 0.037) (Table 1). These variables were further adjusted for in multivariate logistic regression models to control for any residual effect of their confounding on the main effect of selected SNPs.

The genotype and allele distributions of the three TGFB1 SNPs and three VEGF SNPs between the cases and controls are summarized in Table 2. The observed genotype frequencies of these SNPs were all in agreement with the Hardy-Weinberg equilibrium in the control subjects, except for TGFB1 -509C>T and VEGF -634G>C, indicating some possible selection bias in the hospital-based controls. Although none of the minor alleles had a different frequency distribution between the cases and the controls, the VEGF-634G>C genotype frequencies were statistically significantly between the cases (40.4% for GG, 42.1% for CG, and 17.5% for CC) and the controls (51.0% for GG, 28.0% for CG and 21.0% for CC) (P = 0.003) (Table 2).

Further logistic regression analysis revealed that, of the six SNPs examined, only the VEGF -634CG heterozygous and the combined -634CG+CC variant genotypes were associated with a significantly elevated risk of gastric cancer (adjusted OR = 1.88, 95% CI = 1.24-2.86 and adjusted OR = 1.56, 95% CI = 1.07-2.27, respectively). In further stratified analysis by age, sex, ethnicity, alcohol use and smoking status, the risk associated with the combined -634CG+CC variant genotypes was more evident in subgroups of ever smokers (adjusted OR = 1.75, 95% CI = 1.06-2.90), never drinker (adjusted OR = 2.00, 95% CI = 1.14-3.54) and non-white subjects (adjusted OR = 2.39, 95% CI = 1.15-4.98) (Table 3). This kind of significant association was not found for the three TGFB1 SNPs and other two VEGF SNPs (Table 2).

To further analyze possible combined effects of the TGFB1 and VEGF SNPs, we estimated the haplotypes on the basis of the observed genotypes. As shown in Table 4, the four common TGFB1 haplotypes (CTG, CCG, TCG and CCC) represented 85.6% and 89.9% of the chromosomes of the controls and cases, respectively. For the VEGF haplotypes, the common alleles included CGC, TCC, CCC and CGT. However, the frequency distributions of these haplotypes between the cases and controls was not significantly different for both TGFB1 (P = 0.759) and VEGF (P = 0.808), nor were they associated with risk of gastric cancer.

Finally, because the sample size of this study was small, we calculated FPRP for the positive findings. We found that under the assumption of a prior probability of 0.1, the FPRP of VEGF -634CG vs. -634GG for yielded values of 0.13, suggesting that the VEGF -634CG vs. -634GG findings was noteworthy. However, the FPRPs of VEGF -634CG+CC vs. -634GG for all subjects, non-whites, ever smokers and never drinkers yielded values of 0.31, 0.54, 0.49 and 0.38, respectively, which were likely to be false positive findings. Therefore, our findingds need to be validated in larger studies.