Polymorphisms in the Reduced Folate Carrier, Thymidylate Synthase, or Methionine Synthase and Risk of Colon Cancer

Folate is an essential micronutrient in humans, the primary function of which is as a carrier of single-carbon units. Folate-dependent reactions include the biosynthesis of thymidylate, purines, methionine, and glycine thus linking it to nucleotide synthesis as well as the provision of methyl groups (1). High dietary folate intakes, or biomarkers thereof, have been associated with a reduced risk of colon cancer or its precursors in most, although not all, studies (2-5). Several studies showing associations with genetic polymorphisms in folate-metabolizing enzymes lend support to a causal relationship between folate and colorectal carcinogenesis (6-10). Biological mechanisms linking folate to colorectal carcinogenesis include an altered provision of S-adenosylmethionine for methylation reactions, including DNA methylation, and changes in the availability of nucleotides, such as thymidylate, for DNA synthesis and repair (11, 12).

We have previously reported on associations between polymorphisms in 5,10-methylenetetrahydrofolate reductase (MTHFR) and risk of colon cancer (13). Here, we extend this work to common genetic variants in thymidylate synthase (TS), the reduced folate carrier (RFC), and methionine synthase (MTR) in relation to colon cancer risk. TS is a key enzyme in folate metabolism that catalyzes the conversion of dUMP to dTMP for the provision of thymidine, a rate-limiting nucleotide essential for DNA synthesis and repair (see Fig. 1). TS is also a primary target for major chemotherapeutic agents, including 5-fluorouracil. We investigated the role of a polymorphism in the 5′-untranslated region (5′-UTR) enhancer region (three or two repeats of a 28-bp sequence), resulting in reduced TS expression among those with fewer repeats (14) and a 6-bp insertion or deletion (1,494 bp in the 3′-UTR) that affects mRNA stability (15, 16).

RFC is responsible for the active transport of 5-methyl-tetrahydrofolate from plasma to cytosol. A polymorphism in the RFC gene (80G>A, Arg²⁷His) seems associated with a higher affinity for folate (17). Among 169 healthy individuals who were stratified by MTHFR 677C>T genotype, the variant A allele was consistently and linearly associated with higher plasma folate concentrations (17). Furthermore, concentrations of methotrexate 24 to 48 hours after administration were higher among children with acute leukemia homozygous for the variant A allele, providing additional support for differential carrier activity among those with variant genotypes (18).

MTR catalyzes the methylation of homocysteine to methionine with simultaneous conversion of 5-methyl-tetrahydrofolate to tetrahydrofolate (Fig. 1). A variant in the MTR gene (2756A>G, Asp⁹¹⁹Gly; ref. 19) may affect plasma homocysteine concentrations. Some studies (20, 21) but not others (22-24) have found that homocysteine concentrations tend to decrease linearly across genotypes, with the AA genotype associated with the highest homocysteine concentrations. Studies on colorectal neoplasia have been inconsistent: the GG genotype has been associated with a somewhat reduced risk of colorectal cancer (22, 25) yet a possible increased risk of colorectal adenoma (26).

In this large population-based case-control study of colon cancer, we sought to evaluate the role of these polymorphisms in defining colon cancer risk, either alone or in interaction with specific nutrient intakes and other genotypes. Furthermore, as some of the associations of folate metabolism may differ by estrogen exposure (13), possibly because of mechanisms attributable to hypermethylation of the estrogen receptor (27), we evaluated interactions with postmenopausal hormone (PMH) use.

Participants were African American, Caucasian, or Hispanic subjects from the Kaiser Permanente Medical Care Program of Northern California, an eight-county area in Utah, and the metropolitan Twin Cities area of Minnesota. Eligibility criteria for cases included diagnosis with first-primary incident colon cancer (International Classification of Diseases for Oncology, 2nd edition codes 18.0, 18.2-18.9) between October 1, 1991 and September 30, 1994; between 30 and 79 years of age at time of diagnosis; and mentally competent to complete the interview. Proximal tumors were defined as cecum through transverse colon; tumors in the splenic flexure and descending and sigmoid colon were categorized as distal. Cases with adenocarcinoma or carcinoma of the rectosigmoid junction or rectum (defined as the first 15 cm from the anal opening) or with known familial adenomatous polyposis, ulcerative colitis, or Crohn's disease were not eligible. Of all cases identified, 65% of those contacted consented to participate in the study. Controls who had never had a previous colorectal tumor were randomly selected in proportion to the cases within the geographically defined areas from Kaiser Permanente Medical Care Program membership lists in California; driver's license lists, random digit dialing, or Centers for Medicare and Medicaid Services lists, formerly known as the Health Care Finance Administration, for Utah; and driver's license or state identification lists in Minnesota. Controls were frequency matched to cases by sex and 5-year age group. These methods have been described in detail (28). Of all controls selected, 64% participated.

Trained interviewers collected diet and lifestyle data in person using laptop computers. Study quality control methods have been described (29, 30). The reference period for the study was the calendar year ∼2 years before date of diagnosis (cases) or date of selection (controls). Dietary intake data were ascertained using an adaptation of the validated Coronary Artery Risk Development in Young Adults diet history questionnaire (31). Participants were asked to determine which foods were eaten and the frequency with which foods were eaten. Nutrients were calculated using the Minnesota Nutrition Coordinating Center's nutrient database, version 19.

Of 4,403 cases and controls with valid study data, 3,680 (84%) had blood collected primarily during the in-person interview, or during a clinical visit (83% of cases and 85% of controls). Genomic DNA was extracted using methods described in refs. (14, 32). All samples were genotyped for two polymorphisms in the TS gene (TSER, 3′-UTR 1494delTTAAAG), MTR D919G, and RFC 80G>A. A total of 3,562 (97% of cases and 97% of controls with blood collected) had genotype information for both TSER and 3′-UTR 1494delTTAAAG. 5′-Nuclease assays that had been previously used to genotype other polymorphisms in the folate pathway (MTHFR 677C>T, MTHFR 1298A>C, and MTR D919G) have been described (13, 26).

Both TS polymorphisms were analyzed using fluorescent size discrimination. For the analysis of the TSER 28-bp repeat polymorphism, a fragment containing the repeats was amplified using the following primers: forward primer, 5′-6FAM-GTGGCTCCTGCGTTTCCCCC-3′; reverse primer, 5′-GGCTCCGAGCCGGCCACAGGCATGGCGCGG-3′(14). The PCR reactions contained 1× GeneAmp buffer (Applied Biosystems, Foster City, CA), 1.5 mmol/L MgCl₂, 200 μmol/L deoxyribonucleotide triphosphates, 100 nmol/L each primer, 10% DMSO, 1 unit AmpliTaq DNA polymerase (Applied Biosystems), and 100 ng of genomic DNA. Cycling conditions were one cycle of 94°C for 2 minutes; 35 cycles of 94°C for 30 seconds, 63°C for 30 seconds, and 72°C for 30 seconds; and a final extension at 72°C for 5 minutes. The amplified fragments were analyzed on an ABI 3100 genetic analyzer. A fragment containing the 3′-UTR deletion was amplified using the following primers: forward primer, 5′-6FAM-CAAATCTGAGGGAGCTGAGT-3′; reverse primer, 5′-CAGATAAGTGGCAGTACAGA-3′. The PCR reactions contained 1× GeneAmp buffer, 2 mmol/L MgCl₂, 150 μmol/L deoxynucleotide triphosphates, 300 nmol/L each primer, 1 unit AmpliTaq DNA polymerase, and 50 ng genomic DNA. Cycling conditions were one cycle of 94°C for 5 minutes; 30 cycles of 94°C for 30 seconds, 60°C for 45 seconds, and 72°C for 60 seconds; and a final extension at 72°C for 10 minutes. The amplified fragments were analyzed on an ABI 3100 genetic analyzer. For both TS polymorphisms, the correlation between fragment size and repeat number was confirmed by sequencing.

The 80G>A polymorphism in RFC was detected by allelic discrimination using the 5′ nuclease assay on a 7900HT sequence detection system (ABI). The 5′-nuclease genotyping assay was validated by genotyping 100 individuals by both 5′-nuclease assay and RFLP. There were no discrepancies between the two assays. Genotyping of the 80G>A polymorphism was done in 20-μL reactions containing 1× Taqman PCR core reagents (ABI), 3 mmol/L MgCl₂, 200 nmol/L each PCR primer (forward primer, 5′-AGCCCAGCGGTGGAGAAG-3′ and reverse primer, 5′-AGCCGTAGAAGCAAAGGTAGCA-3′), 150 nmol/L MGB probe 5′-VIC-TCCTGGCGGCGCC-3′ (Applied Biosystems; G allele), 100 nmol/L MGB probe 5′-6-FAM-TGGCGGCACCTCG-3′ (A allele), 0.5 unit AmpliTaq Gold, 0.2 unit AmpErase UNG, and 5 ng genomic DNA. The amplification cycles were 50°C for 5 minutes, 95°C for 10 minutes, and 40 cycles of 95°C for 15 seconds and 60°C for 1 minute. Positive controls for all the genotypes as well as four negative controls were included in each plate. For quality control of all the polymorphisms, genotyping for 94 randomly selected samples was repeated. There were no discrepancies.

Logistic regression models were used to estimate associations in various ways. We stratified the data by sex and estimated the risk of colon cancer given a certain TS, MTR, or RFC genotype and examined risk estimates further stratified by other population characteristics (e.g., tumor site and age). The combined effects of TSER and 3′-UTR 1494delTTAAAG were calculated using individuals who were homozygous for the common allele at both loci as the reference group. We assessed the joint interaction between genotype and level of nutrient intake by using those with low nutrient intake and homozygous for the most common (wild type) allele for TSER, 3′-UTR, MTR, or RFC as a common reference point. We also assessed gene-gene interactions in the folate pathway using the homozygous genotype for the most common allele as the reference. Similarly, the interaction between genotype and recent estrogen status in postmenopausal women was assessed using as the reference group no PMH use and wild-type TS, MTR, or RFC genotype.

Maximum likelihood estimates of population TS haplotype frequencies from unphased genotype data were obtained from an expectation maximization algorithm, assuming Hardy-Weinberg equilibrium, according to Excoffier and Slatkin (33) using SAS/Genetics software, 2002 (SAS/Genetics, Cary, NC).

Odds ratios (OR) and 95% confidence intervals (95% CI) were calculated from unconditional logistic regression models. In these models, age at diagnosis or selection, body mass index reported for the reference period (kg/m²), long-term vigorous leisure time physical activity, total energy intake, dietary fiber, dietary calcium, and number of cigarettes smoked per day on a regular basis were included as covariates to adjust for potential confounding. Haplotype-specific relative risks were assessed according to methods described in Stram et al. (34) using logistic regression software (SAS, release 8.2).

Separate analyses were done for men and women to determine whether differences existed by sex, as most of the literature has focused on either men or women. Assessment of interactions among genotypes, diet, and the risk of colon cancer were based on departure from additive risks using the relative risk due to interaction formulation of Hosmer and Lemeshow extended to more than two allelic combinations and/or environmental exposures (35). Interaction using a multiplicative scale was also examined. Interactions between genotypes and PMH use in postmenopausal women were assessed using a Wald χ² test of the difference between slopes from the (assumed linear) change in ORs, keeping the wild-type TS genotype constant across the varying genotypes for the respective other TS polymorphism.

Selected characteristics of the study population and MTHFR genotype frequencies by case or control status are presented in Table 1. The study participants were predominantly self-identified as non-Hispanic Caucasian (92%), with the remainder Hispanic (4%) and African American (4%). All genotypes were in Hardy-Weinberg equilibrium (assessed separately for cases and controls), and allele frequencies were consistent with previous reports (36). The TS polymorphisms were in linkage disequilibrium (D′ = 0.46 among non-Hispanic Caucasians).

Table 2 describes the main associations seen with the polymorphisms, stratified by gender. Among men, TSER variant genotypes were associated with a significantly decreased risk (3rpt/2rpt: OR, 0.8; 95% CI, 0.6-0.98; 2rpt/2rpt: OR, 0.7; 95% CI, 0.6-0.98). No such risk reduction was observed among women. When combined genotypes (or diplotypes) for both TS polymorphisms were considered, almost all of the ORs for variant genotypes among both men and women were below 1.0, with statistically significantly reduced risks for women. Risk estimates for TS haplotypes, including variant alleles, compared with wild-type haplotype were not significantly different from 1.0 (data not shown).

Neither the MTR D919G nor RFC 80G>A polymorphism was associated with altered colon cancer risk among men or women (Table 2).

Folate metabolism involves circulation of folate metabolites through multiple cycles, as well as feedback mechanisms between these cycles (Fig. 1). Therefore, we evaluated gene-gene interactions between the polymorphisms investigated here, as well as those we have reported on previously (8, 13). ORs different from 1.0 were seen largely for stratifications of TS, RFC, or MTR by MTHFR 677C>T or 1298A>C genotypes, and these are presented in Table 3. Among men, reduced risks associated with variant TS genotypes (e.g., the presence of TSER 2rpt/2rpt or TS 3′-UTR deletion) were most pronounced for those with MTHFR TT genotypes. The MTHFR 1298CC genotype was associated with a decreased risk among women; however, this risk reduction seemed independent of TS genotypes. There was no evidence for interactions between MTR D919G or RFC 80G>A and MTHFR genotypes.

We also investigated whether risk estimates associated with polymorphisms in these folate-metabolizing enzymes differed by dietary intakes of folate, methionine, alcohol, or vitamins B6, B2, or B12. Consistent with our previous report on colorectal adenoma (37) among men, the TSER variant conferred a reduced risk in the presence of low folate intake (lowest tertile < 318 μg/d; TS 2rpt/2rpt or 2rpt/3rpt: OR, 0.7; 95% CI, 0.5-0.9 compared with wild-type 3rpt/3rpt). A similar risk reduction with the TSER variant genotypes was observed among men with low methionine intakes (<2.0 g/d): TS 2rpt/2rpt or 2rpt/3rpt (OR, 0.6; 95% CI, 0.4-0.9) compared with wild-type 3rpt/3rpt. However, none of these patterns was observed among women for the TS genotypes nor for TS haplotypes in either sex. There were no clear patterns or associations following stratification by vitamin B6 or B12 intake. No meaningful differences in risk were observed for MTR genotypes when stratified by nutrient intakes.

Among women in the lowest tertile of folate intake (≤273 μg/d), the RFC variant genotypes were associated with a decreased risk (wild-type GG: OR, 1.0; GA or AA: OR, 0.7; 95% CI, 0.5-1.0). Among women, we observed a significant gene-nutrient interaction in that only those with the GG genotype benefited from a diet higher in folate, whereas no difference in risk with variable folate intake was seen among those with the combined GA or AA genotypes (P_interaction = 0.04, multiplicative scale; P = 0.01, additive scale). This pattern was not seen among men.

Because of the observed differences in risk patterns among men and women and our past findings regarding an interaction between postmenopausal hormone use (PMH use) and MTHFR genotype, we investigated whether risk estimates of TS, MTR, or RFC genotypes differed by PMH use. Among PMH users, the variant TS genotypes were associated with substantially reduced risk of colon cancer, whereas much weaker associations were observed among non-PMH users (Table 4). No such interactions were observed for MTR or RFC (data not shown).

Within this large population-based study of colon cancer, we investigated polymorphisms in three folate-metabolizing enzymes (TS, MTR, and MTHFR), as well as the relevant carrier protein (RFC), thus addressing genetic variability in multiple key proteins in this biological pathway. There is strong evidence for the functional effect of MTHFR 677C>T and TSER variant genotypes (14, 38-41), with some, yet less well defined, evidence for the in vivo functional relevance of MTR, RFC, and the TS 3′-UTR variant (16-18, 20-23, 42). Our evaluations of colon cancer risk confirm this assessment, in that there were no significant alterations in risk for MTR, RFC, and TS polymorphisms, with the exception of a risk reduction associated with the TSER variants among men. The ORs for TSER are comparable with those from a previous report by Chen et al. (43) among male physicians (OR, 0.9; 95% CI, 0.6-1.3 for the 2rpt/3rpt genotype and OR, 0.6; 95% CI, 0.4-0.98 for the 2rpt/2rpt genotype). A statistically significant trend towards reduced risk with the variant TSER alleles was observed both here and in that population (43). These results indicate that the variant TSER 2rpt/2rpt genotype reduces risk of colon or colorectal cancer among men to a degree comparable with that of the MTHFR 677TT genotype (44). As risk reductions for TSER variants were only seen among women who are on postmenopausal hormones but not among women overall, estrogen status may play a role.

Further complexity is added as a result of the presence of a second functionally relevant polymorphism in TS (15, 16). We evaluated the combined effects of these genotypes as well as haplotypes to discern possible risk patterns. The combined TS wild type/wild type genotype constitutes only 8.3% of our population. When compared with this wild type/wild type reference group of putatively highest TS expression and TS mRNA stability, the variant TSER genotypes were associated with statistically significantly reduced risk among women (OR, 0.6; 95% CI, 0.4-0.9) yet not among men. Our sample sizes for these sex-specific associations were limited, and results should be followed up in large study populations that have the ability to investigate combined genotypes as well as sex-specific ORs. The presence of two common functional variants within TS suggests that it is essential to take both of these into account simultaneously.

The MTR polymorphism is less common (allele frequency = 0.20) and has been investigated in three epidemiologic studies, including a large Norwegian cohort (7, 25, 45). Similar to our findings, Le Marchand et al. (45) and Chen et al. (7) did not report any associations between this variant and colon cancer risk, whereas Ulvik et al. (25) observed a significantly reduced risk among those with a GG genotype compared with wild-type AA (OR, 0.65; 95% CI, 0.47-0.90). We observed reduced risks among men (OR, 0.7), but the 95% CI included 1.0. As only about 5% of the population have the homozygous variant genotype, very large studies are needed to quantify the strength of this association.

For gene-gene interactions, only combinations with the MTHFR polymorphisms showed interesting patterns. This is not surprising, because MTHFR is a key regulatory enzyme in folate-mediated one-carbon metabolism, the activity of which determines the distribution of folate metabolites toward nucleotide synthesis or methylation reactions. There is strong evidence that the MTHFR 677C>T variant alters the balance of metabolites within the pathway (39, 46). In combined analyses of TS and MTHFR polymorphisms, we observed that men carrying at least one variant TS allele (either TSER 2rpt or TS 1494del) in addition to the MTHFR 677TT genotype were at relatively lowest risk compared with all other groups (both OR, 0.6; 95% CI, 0.4-0.9). This confirms our previous observation in colorectal adenoma, where individuals with low TS expression and low MTHFR activity genotypes also experienced the lowest adenoma risk (OR, 0.56; ref. 10). If this statistical interaction reflects biological mechanisms, then we may hypothesize that the observed pattern suggests that a greater diversion of folate metabolites (specifically 5,10-methylene-tetrahydrofolate) toward purine synthesis is protective for the development of colorectal neoplasia. Recent findings by Quinlivan et al. suggest that folate depletion adversely affects purine synthesis in humans and a greater relative rate of adenine synthesis among individuals with the MTHFR TT genotype (46). Depurination is the most common type of DNA damage with ∼10,000 depurinations/cell/d (47, 48). Although efficiently repaired, apurinic sites are present in DNA. We recognize that one other study did not observe this risk pattern, but their sample size was limited to 270 cases, with consequent restricted power for studying gene-gene interactions (43).

Our investigations of gene-diet interactions confirmed, to some extent, associations we have previously observed with respect to TSER, and folate intake that reduced TS expression (TSER 2rpt/2rpt) is associated with a reduced risk in the presence of a low folate intake (10). However, this pattern was seen only among men and also has not been observed in the Health Professionals study (43). Again, if that pattern reflects a biological mechanism, it would point toward purine synthesis as a key link between one-carbon metabolism and colorectal neoplasia. We were unable to confirm previously observed gene-diet interactions for MTR in colorectal adenoma (26) and did not see a clear pattern for RFC-diet interactions. However, the RFC is the transporter for naturally occurring folates (in the form of 5-methyl-tetrahydrofolate) but plays a smaller role in the transport of folic acid (49). Thus, in populations, such as the one described here, where folate intake from supplements in the form of folic acid comprises a substantial proportion of the overall folate intake, genetic variability in the RFC may not be as relevant for the overall supply of folate metabolites. Unfortunately, no quantitative information on supplement use was available for this population.

Lastly, we observed differences in risk patterns dependent on the past use of postmenopausal hormones. Interactions between folate metabolism and PMH use are not implausible, as there are links between homocysteine concentrations and PMH use (50-53), and methylation of the estrogen receptor is an early event in colorectal carcinogenesis, which may less frequently occur in the presence of PMH (27, 54). We have previously reported on a significant difference in risk patterns of PMH-associated risks by MTHFR genotypes (13). However, these interactions need to be confirmed by others, because sample sizes were in parts insufficient to yield stable estimates.

Although this study is quite comprehensive with respect to investigations of genetic variability in one-carbon metabolism and risk of colon cancer, there are three important limitations. First, our investigations did not include other genetic polymorphisms in folate-metabolizing enzymes that may be of possible relevance, such as methionine-synthase reductase (MTRR) or serine-hydroxymethyltransferase (SHMT). Thus far, MTRR does not seem related to colon cancer risk (45), and the functional relevance of the cSHMT polymorphisms is unclear. The study presented here focused on candidate polymorphisms in key enzymes with substantial evidence for functional effect; we hope to expand our investigations to other relevant candidate polymorphisms as they are reported.

Second, there is now strong evidence that a subset of colorectal cancer cases arises as part of a CpG island methylator phenotype (55, 56). Information on CpG island methylator phenotype status should be taken into account in future studies investigating links between genetic variability in folate metabolism and risk of colorectal cancer.

A final limitation is our current inability to integrate knowledge of biochemical relationships within the pathway into the statistical analysis. Although an approach that uses stratification for gene-nutrient or gene-gene interactions is valuable, in that it allows for an empirical investigation of the associations, it is also limited in that statistical power for higher-order interactions is lacking, even within this large study population. Because folate metabolism consists of several interconnected cycles (see Fig. 1), such interactions are to be expected. Our approach toward solving this problem is to use, in the future, results from a mathematical model of one-carbon biochemistry for investigations of multiple genetic variants on selected biomarkers. Although this model is still under development, preliminary results show that it replicates the biochemical relationships in the folate cycle and methionine cycle with reasonable accuracy (57, 58). Furthermore, our group and others are developing methods to address this key problem for molecular epidemiologic studies (59). Thus, we hope that in the future, we will be able to achieve closer integration of the biochemistry and statistical analysis. Because there is strong evidence that disturbances in this biochemical pathway can modify risk of several types of malignancies (60-63), birth outcomes (64-66), and possibly cardiovascular disease (67) and autism (68), a more thorough understanding of the interplay of multiple genetic polymorphisms under specific dietary conditions and their combined effect on biomarkers and disease end points will be highly relevant.

We thank Sandra Edwards and Leslie Palmer at the University of Utah for data collection efforts of this study, Clayton Hibbert and Linda Massey for word processing assistance, and Juanita Leija for genotyping assays.