Esophageal and Gastric Cardia Cancer Risk and Folate- and Vitamin B₁₂-related Polymorphisms in Linxian, China

Low folate has been associated with an increased risk for a number of gastrointestinal cancers (1), including esophageal (2, 3, 4) and stomach cancers (5, 6, 7, 8). There are few studies examining vitamin B₁₂ and esophageal or gastric cancer risk. One American study found a positive association between vitamin B₁₂ intake and esophageal and gastric cancer (5). Several others found that subjects with pernicious anemia, a condition characterized by vitamin B₁₂ deficiency, had greater risk for both cancers (9, 10). An ecological study in South Africa reported that subjects living in high ESCC⁷ risk districts had significantly lower folate and B₁₂ blood concentrations and nutrient intakes than did those living in low ESCC risk districts (11).

MTHFR and methionine synthase are two enzymes involved in folate and methyl group metabolism. MTHFR catalyzes the irreversible reduction of 5,10-methylenetetrahydrofolate to 5-methyltetrahydrofolate. Five-methyltetrahydrofolate is the carbon donor for the major reaction that regenerates methionine from homocysteine. In addition to folate, this reaction requires methionine synthase and its coenzyme, B₁₂. MTRR maintains adequate concentrations of activated B₁₂ for this reaction to occur. The integrity of these reactions has at least two important consequences. First, methionine in the form S-adenosylmethionine is the principle methyl donor for over 100 biological methylation reactions including DNA methylation. Second, the formation of methionine reconverts 5-methyltetrahydrofolate to tetrahydrofolate. Tetrahydrofolate is the precursor of 5,10-methylenetetrahydrofolate, which is required for the synthesis of nucleotides and purines for DNA. Deficiencies in folate and B₁₂ and alterations in MTHFR, MS, and MTRR functions may contribute to carcinogenesis through altered DNA methylation (e.g., DNA hypomethylation) and impeded thymidylate synthesis, resulting in nucleotide imbalances, increased uracil misincorporation in DNA, DNA strand breaks, and impaired excision repair, which may increase the susceptibility of DNA to mutations and damage (1).

Two common polymorphisms identified in the MTHFR gene have been associated with cancer (1). The best characterized is a missense mutation consisting of a C→T substitution at bp 677 that produces an alanine to valine amino acid substitution within the predicted catalytic domain of the MTHFR enzyme. The second is a missense mutation consisting of an A→Csubstitution at bp 1298 that produces a glutamate to alanine substitution. Several case-control studies from China have observed that both MTHFR 677TT and 1298CC variant genotypes were associated with ESCC (12), whereas only the MTHFR 677T allele was associated with gastric cancer (13, 14, 15). A more recently identified MTRR gene polymorphism involves an A→G substitution at bp 66, resulting in a methionine to isoleucine substitution. The variant GG genotype has been positively but nonsignificantly associated with colorectal cancer (16) and significantly associated with premature coronary heart disease, spina bifida, and Down’s Syndrome (1).

Since 1985 we have been studying a cohort from Linxian, China, a rural county in North Central China, which has among the world’s highest rates of ESCC and GCA. Originally, these individuals were participants in one of two randomized nutrition intervention trials conducted from 1986 to 1991, the General Population and Dysplasia Trials (17, 18, 19). The former included 29,584 subjects from four communes, and the latter included 3,318 subjects selected based on screening-detected cytologic esophageal abnormalities (17, 18, 19, 20). Detailed descriptions of the trial designs, subject selection criteria, subject characteristics, methods of disease ascertainment, and primary endpoints have been published (17, 18, 19, 20).

We recently measured pre-intervention levels of serum folate and B₁₂ on a random sample of over 3000 individuals from this cohort, and we found that 90% of this population had marginal folate status (serum folate <6 ng/liter), and 75% had marginal B₁₂ status (serum B₁₂ < 200 pg/ml; Ref. 21). The goal of this current study was to examine the relationship between incident ESCC and GCA and three polymorphisms in two genes (MTHFR C677T and A1298C and MTRR A66G). Both genes produce enzymes for which folate and B₁₂ are cofactors. Because the high rates for both cancers in Linxian suggest a common etiological factor, we examine risks separately for the end points ESCC and GCA as well as the combined end point ESCC/GCA (17, 18, 19, 20).

At the end of the Linxian intervention trials in 1991, approximately 6000 individuals who were alive and cancer free were selected for a blood sampling study. We were able to extract DNA yields of ≥1.5 μg from the RBCs on 4005 of these individuals. The subjects for this study were selected from this group (n = 4005), in accord with a stratified case-cohort design (22). The six stratums were defined by sex and the following three age categories: <50 years; 50–59 years; and ≥60 years. Polymorphisms we measured on all incident ESCC (n = 131) and GCA (n = 90) cases that occurred between May 1991 and May 1996 and on 421 controls. Excluded from this analysis are 22 individuals who could not be genotyped (2 ESCC cases and 20 controls) and 3 controls who lacked information on height and weight, leaving us with 129 ESCC cases, 90 GCA cases, and 398 controls. In each stratum, the control:site-specific case ratio was greater than 2:1. Disease classifications were based on monthly end point surveillance (17, 18, 19, 20) and interview and examination of all living participants, or their next of kin, in May 1996 (>99% response rate). All cancer diagnoses from the 1991–1996 period were reviewed by our panel of United States and Chinese experts (20). Gastric cancers were defined as cardia cancers if they were in the proximal 3 cm of the stomach. These methods of assessment were identical to those used during the trial period (17, 18, 19, 20) _.

Genomic DNA was extracted from frozen blood, and the MTHFR C677T and A1298C and MTRR A66G polymorphisms were genotyped at a commercial laboratory (BioServe Biotechnologies, Ltd., Laurel, MD), using a PureGene kit (Gentra Systems, Inc., Minneapolis, MN) and real-time PCR analysis amplification (TaqMan; PE Biosystems, Foster City, CA). PCR primers and dual-labeled allele discrimination probes were designed using Primer Express software (version 1.5; Perkin-Elmer). All laboratory personnel were blinded to case-control status. A blinded repeat genotyping of 10% of the DNA samples yielded 100% concordance for all three polymorphisms.

Information on age, sex, height, weight, BMI (kg/m²), smoking (tobacco use for ≥6 months during lifetime: no, yes), alcohol use (any use during the past year: no/yes), and trial (Dysplasia or General Population Trial) were obtained from the 1985 questionnaire (17, 18, 19, 20, 23). As reported previously (17, 18, 19, 20, 23), this binary classification of smoking and drinking captures the variability of these characteristics in this population as assessed by more detailed measures. Ps for case and control differences were generated using the Wilcoxon rank-sum (continuous variables) or χ² tests (categorical variables). In the one instance with small cell counts (n ≤ 5), we used Fisher’s exact test. Tests for Hardy-Weinberg equilibrium and tests and estimates of correlation between polymorphisms (Spearman correlation coefficients) were performed using only the 431 individuals in the “subcohort” (21, 24).

RRs and 95% CIs were estimated using a stratified case-cohort Cox proportional hazards models (21, 24). Models included indicator variables for trial, smoking, and alcohol use and continuous variables for BMI, and stratum-specific age. Risks are estimated separately for ESCC, GCA, and the combined category ESCC/GCA. Nested models were compared using score tests.

RR was first estimated with each genotype categorized by the number of variant alleles (0, 1, or 2). Trend tests were based on this classification. When indicated by similarity of risk estimate and/or biological considerations, we also estimated risks in groups defined by combining heterozygous individuals with one of the homozygous groups. For MTHFR C677T, we combined the CC and CT genotypes; for MTHFR A1298C, we combined the AC and CC genotypes; and for MTRR A66G, we combined the AG and GG genotypes. Effect modification of the polymorphisms with each other and for each polymorphism with the covariates in Table 1 was evaluated by the addition of cross product terms and within subgroup estimates.

Statistical Analysis Systems (SAS) software version 8.2 (Cary, NC) was used for the cross-sectional analyses, and Peanuts in Epicure (25) was used for the prospective analysis. All statistical tests were two tailed and considered significant at P = 0.05.

Table 1 lists the baseline characteristic of subjects as measured in 1985 by case control status. Compared with the controls, ESCC and GCA cases were more often participants in the Dysplasia Trial. Female cancer cases had lower BMI than controls. When, as in the table, age was averaged overall strata, the ESSC cases were younger than the controls. However, when age comparisons are made within the six age-sex strata used in the analyses, no significant differences existed (data not shown).

Both MTHFR C677T (P = 0.24) and MTRR A66G (P = 0.82) were in Hardy-Weinberg equilibrium. MTHFR A1298C was not (P < 0.005): 0% had the CC genotype; whereas 1.7% was expected. The MTHFR C677T and A1298C polymorphisms were negatively correlated (r = −0.37, P < 0.0001). Fewer than 1% of subjects had both MTHFR 677TT and 1298AC/CC; no subjects had MTHFR 677TT and 1298CC (data not shown). Neither of the two MTHFR polymorphisms were correlated with the MTRR polymorphism (r = ∼0; C677T, P = 0.39; A1298C, P = 0.14).

Compared with MTHFR 677CC homozygotes, neither the CT nor TT genotype was associated with cancer (Table 2). With the combined MTHFR 677CC/CT genotypes used as referent category, the TT genotype was associated with a 1.45-fold increased risk for ESCC/GCA. Similar but nonsignificant risk increases were observed for separate ESCC and GCA. Only three individuals, all with ESSC, had MTHFR 1298CC (Fisher’s exact P = 0.03). No associations were observed for the combined MTHFR 1298AC/CC genotypes. For MTRR A66G, ESCC risk increased with each additional variant G allele. The combined AG/GG genotypes were associated with a significant 1.59-fold increased ESCC risk. MTRR A66G was not associated with GCA or combined ESCC/GCA. There was no evidence for interaction between the three polymorphisms and cancer risk.

Alcohol modified the associations of MTHFR C677T with GCA and combined ESCC/GCA and of MTRR with ESSC (P interaction = 0.03, 0.04, and 0.01, respectively). Among those with MTHFR 677TT, alcohol drinkers had approximately 5- and 2-fold the risk of GCA and ESCC/GCA as nondrinkers, respectively (Table 3). In contrast, alcohol drinkers with MTRR 66GG had approximately half the risk of ESSC as nondrinkers. There were no other significant interactions of the polymorphisms with the covariates in Table 1.

We prospectively examined the relationship between three polymorphisms in two genes for enzymes for which folate and vitamin B₁₂ are cofactors (MTRR A66G and MTHFR C677T and A1298C), and incident ESCC and GCA. In this Chinese population deficient in folate and vitamin B₁₂, we found that individuals with the MTHFR 677TT genotype had a 45% greater risk of ESCC/GCA than those with the CC/CT genotypes. This elevated risk was higher among alcohol drinkers. Although only three individuals had the MTHFR 1298CC genotype, all had ESCC (Fisher’s exact P = 0.03). For MTRR, ESSC risk increased with each 66G allele: those with one variant allele had twice the risk of the wild-type group, and those with two alleles had 2.4 times the risk of the wild-type group.

Several retrospective studies (12, 13, 14, 15), all from China, have examined the association of the MTHFR polymorphisms and ESSC or gastric cancer. Concordant with our results, these studies found that individuals with the MTHFR 677TT genotype had increased risks of ESSC (Ref. 12; n = 240 cases) and GCA [Refs. 13 (n = 82 cases) and 14 (n = 217 cases)] of approximately 6- and 2-fold, respectively. The only study to examine whether alcohol modified the MTHFR 677TT association was a case-control study of gastric cancer, which, although published only in Chinese, reports in its abstract a finding similar to ours. Given the sample size of our study and the number of interactions we examined, our findings on this alcohol effect are more suggestive than conclusive.

The association of MTHFR 677TT with malignancy is best studied in colon cancer. The research suggests that the MTHFR 677TT genotype is positively associated with cancer only among those with low methyl-group nutrient intake and folate status (1). The MTHFR677T allele produces a less stable thermolabile protein with lower MTHFR activity. This is thought to impair folate metabolism, particularly in those with deficiencies. The thermolability may contribute to carcinogenesis in a manner similar to folate deficiency by altering DNA methylation and/or thymidylate synthesis, thus resulting in nucleotide imbalances, uracil misincorporation into DNA, DNA strand breaks, and impaired DNA excision repair (1).

For the MTHFR A1298C polymorphism, the study of ESSC found a significant 4.43-fold increased risk with the MTHFR 1298CC genotype (12). The GCA studies found nonsignificant increased (14) or decreased risk (13). Although these results agree with ours, each study had a small number of subjects with the CC genotype, thus limiting the conclusions of any results.

To our knowledge, ours is among the first study to examine the relationship between the MTRR A66G polymorphism and cancer. Because MTRR requires B₁₂ to support its function, and the majority of our population had marginal B₁₂ status (21), it is possible that our population is particularly susceptible to perturbations in MTRR. Others have shown that the positive association between MTRR 66G and spina bifida offspring is stronger among mothers with low B₁₂ status (1).

The greatest shortcoming of our study is the number of cancer cases observed. This limited our ability to examine the joint effects of the polymorphisms. The strength of our study is that it is prospective, based on a well-characterized cohort, with well-documented methods for end point assessment. Thus, it is free from the possible sampling, recall, and disease misclassification biases that can occur in case-control studies. Although studying a population with marginal folate and vitamin B₁₂ status may have increased our ability to detect effects of the polymorphisms, it may also limit the generalizability of these results to Western populations.

In conclusion, we observed significant positive associations between the MTHFR 677TT genotype and combined ESCC/GCA and between MTRR 66G allele genotypes and ESCC in a folate- and vitamin B₁₂-deficient population. The association that we observed with the MTHFR polymorphisms is consistent with that of other studies in methyl-group-deficient populations. Our results suggest that the MTHFR C677T and MTRR A66G polymorphisms influence the risk of ESCC and GCA and support the hypothesis that folate and vitamin B₁₂, cofactors for their respective enzyme products, may play a role in the carcinogenesis of these cancers.