|Home | About | Journals | Submit | Contact Us | Français|
Folate-associated one carbon metabolism (FOCM) may play an important role in colorectal carcinogenesis. Variation in FOCM genes may explain some of the underlying risk of colorectal cancer.
This study utilized data from 1,805 population-based colorectal cancer cases and 2,878 matched sibling controls from the Colon Cancer Family Registry (C-CFR). We used a comprehensive tagSNP approach to select 395 tagSNPs in 15 genes involved in folate and vitamin B12 metabolism. Genotyping was performed using the Illumina GoldenGate or Sequenom platforms. Risk factor and dietary data were collected using self-completed questionnaires. MSI status was determined using standard techniques and tumor subsite was obtained from pathology reports. The association between SNPs and colorectal cancer was assessed using conditional logistic regression with sibships as the matching factor and assuming a log additive or co-dominant model.
In the log additive model, two linked (r2=0.99) tagSNPs in the DHFR gene (rs1677693 and rs1643659) were associated with a significant decrease in CRC risk after correction for multiple testing (OR=0.87; 95% CI=0.71 – 0.94; P=0.029 and OR=0.87 95% CI=0.71 – 0.95, P=0.034 for rs1677693 and rs1643659 respectively. These two linked (r2=0.99) tagSNPs and one tagSNP in the MTR gene (rs4659744) were significantly associated with reduced CRC risk only among individuals not using multivitamin supplements.
Overall, we found only moderate evidence that genetic variation in 15 folate pathway genes may affect CRC risk except in non multivitamin users.
This study suggests that multivitamin supplement use may modify the association between folate pathway genes and CRC risk in a post folic acid supplemented population.
Folates carry most of the one carbon groups essential for hundreds of intracellular transmethylation reactions including those involved in DNA methylation and DNA synthesis. A role for these reactions and the carcinogenic changes important in colorectal cancer has been defined, both in vitro and in vivo, in both humans and animals (1–3).
Folates are a family of molecules that have a common structure based on a pteridine ring conjugated to one or more glutamate side chains. This basic structure is modified by the binding of various single carbon groups to the pteridine ring. Folate-associated one carbon metabolism (FOCM) provides the 1-carbon groups for numerous critical intracellular reactions including methionine synthesis, the de novo synthesis of dTMP from dUMP and de novo purine synthesis (2). Methionine synthesis is the first step in the synthesis of S-Adenosylmethionine (SAM). SAM is the primary methyl donor for hundreds of intracellular methylation reactions including the methylation of DNA. Aberrant DNA methylation, either global hypomethylation or hypermethylation of tumor suppressor genes, plays a key role in colorectal carcinogenesis (4). Global hypomethylation is associated with chromosomal instability and aneuploidy (5, 6) while hypermethylation is associated with loss of transcription which can affect the activity of tumor suppressor genes (4). Low folate conditions have also been associated with increased uracil in DNA (7, 8) and increased DNA damage (9).
Dietary folates have been associated with decreased colorectal neoplasia risk (1). However, their role in populations exposed to folic acid fortification of grain products is less clear (10, 11). Folic acid is the monoglutamated folate form, not found in nature, that is used in supplements and for fortifying cereal grain products in North America. In experimental preclinical studies the effect of folic acid supplementation on the risk for colon tumor formation depended upon the timing of its administration. Increased dietary folic acid decreased the number of intestinal tumors when given to APC min/− and APC min/MSH2−/− mice before the onset of pre-neoplastic foci but the number of tumors was increased significantly when given after the establishment of pre-neoplastic foci (12, 13). Some recent human studies also suggest that neoplasia risk may be increased in those taking high doses of folic acid from supplements (14–18) in the context of a folic acid-supplemented diet. In this regard Mason et al documented a significant trend toward increasing CRC incidence in the US and Canada co-incident with the fortification of grain products with folic acid (19). The present study population was recruited after fortification of the US and Canadian food supplies and so represents one of the first post-fortification study cohorts in which to assess associations between folate associated one carbon metabolism (FOCM)-associated genes and CRC risk.
Several studies have reported on the potential significance of polymorphic variation in FOCM-associated genes and colorectal cancer risk but all of these studies have been restricted to a few variants in a limited number of genes (3, 20). We conducted a systematic investigation of genetic variation in 15 folate pathway genes involved in nucleotide synthesis, methionine metabolism, uptake and distribution of vitamin B12 and polyamine synthesis (Table 1) and CRC risk using a comprehensive tagSNP approach. Previous analyses in this population include variation in MTHFR (21), where we reported an association only for the C677T TT and A1298C CC genotypes in non-multivitamin supplement users, and four genes involved in folate uptake and distribution; the reduced folate carrier 1 (RFC1/SLC19A1), folate receptor 1 (FOLR1), γ-glutamyl hydrolase (GGH), and folylpolyglutamate synthase (FPGS). None of these genes were associated with CRC risk in any subgroup (22).
The methods used for this study are described in detail in Levine et al (21): 1,806 cases and their sibling controls (n=2,879) were enrolled in the six sites of the Colon Cancer Family Registry (C-CFR), an NCI-supported consortium, initiated in 1997 (23). Details of subject recruitment can be found in Newcomb et al (24). Briefly, cases were recruited in two phases, from 1998 – 2002 (Phase 1) and from 2002 – 2007 (Phase 2). Phase 2 subjects were enriched in cases more likely to have a family history of CRC. In this study we used only cases that were identified through population-based cancer registries as described by Newcomb et al. (24). Controls were the unaffected siblings of cases. All subjects signed an informed consent before providing data to the C-CFR.
TagSNPs were selected using Haploview Tagger (25) for the CEU population as described (21). The linkage disequilibrium blocks were determined using data from HapMap data release #16c.1, June 2005, on NCBI B34 assembly, dbSNP b124. For each gene, we extended the 5'- and 3'-UTR regions to include the 5'- and 3'-most SNP within the LD block (approximately 10 kb upstream and 5 kb downstream). In regions of no- or low-LD, SNPs with MAF > 5% at a density of approximately 1 per kb were selected from either HapMap or dbSNP. Finally, non-synonymous SNPs and expert-curated SNPs regardless of MAF were included. A total of 395 SNPs met our criteria of a MAF ≥ 0.05 and a p-value for Hardy Weinberg equilibrium ≥ 0.00013.
SNPs were genotyped on the Illumina 1536 GoldenGate platform (26). We implemented a series of quality control checks based on Illumina metrics, as described previously (21). We performed additional genotyping using Sequenom’s iPLEX Gold for 13 SNPs that were not successfully genotyped on the Illumina platform as described (21). Polymerase chain reaction (PCR) and extension primers for these SNPs are available upon request.
All available tumors were assayed for instability at as many as 10 microsatellite markers: BAT25, BAT26, BAT40, BAT34C4, D5S346, D17S250, D18S55, D10S197, ACTC and MYCL as described (24). MSI data were available for 1,200 (66.4%) cases including all phase one tumors in probands and their relatives with tumors that had clear results for at least 4 markers, and a sample of phase 2 tumors as described by Newcomb et al (24). Instability at ≥30% of the tested loci was defined as MSI-H, instability at ≥10% but less than 30% of loci was defined as MSI-L and those with instability at 0 loci were categorized as MSS. MSI-L and MSS cases were combined in the analysis.
Colon subsite was obtained by reviewing pathology reports. Right colon was defined as occurring in the cecum through the splenic flexure, left colon included the descending colon through the sigmoid colon and rectal tumors included the recto-sigmoid junction and the rectum.
A standard risk factor questionnaire, described in Newcomb et al (24), was administered to all participants at recruitment and available for approximately 98% of the study population. Family history of CRC was defined as any 1st-degree relative with CRC. Folate- and multivitamin-supplement use was defined as ever use at least twice a week for more than a month.
Estimated dietary folate intake, available for 585 cases, 837 controls (about 1/3 of the study population) was estimated from a validated food frequency questionnaire developed at the University of Hawaii (27) and available only for subjects recruited in Ontario, Hawaii and the USC Consortium. All food frequency data were collected after folic acid fortification of the North American food supply in 1998 and analysed, in separate analyses, assuming either pre- or post-fortification folate values as dietary folate equivalents. In the analysis of total dietary folate intake we assumed that both multivitamin and folic acid supplements contained 400 µg of folic acid per pill or tablet.
We used multivariable conditional logistic regression with sibship as the matching factor to estimate main effects assuming a log-additive model. As a secondary analysis we assessed all genotype effects in a co-dominant model and a 2 df likelihood ratio test to estimate p-values for each comparison. We controlled for age and sex in all analyses. Additional control for other variables did not change the results by more than 10%. For the log-additive model and within each gene, p-values for all SNPs were adjusted for multiple testing taking into account correlated tagSNPs using a modified test of Conneely and Boehnke (Pact) (28). We report both the observed likelihood ratio p-value and Pact. For a test of a single gene, a α-level of 0.05 for Pact was used to determine statistical significance. To define significance across all the genes tested, a Bonferroni corrected p-value of (α = 0.05/15 = 0.0033) may also be considered. For the 2 degree of freedom co-dominant model and all stratified analyses we used the Bonferroni significance level of 0.00013 (0.05/395 tagSNPs).
Table 2 shows the demographic characteristics of the study population. The study population was over 87% non-Hispanic White. The mean age of cases was 53.5 (± 10.9) and the mean age of controls was 54.0 (±11.8). Fifty one percent of cases and 44% of controls were male and 81% of the cases were from sites in North America. Thirty percent of cases reported a family history of CRC. Multivitamin supplement use was reported for 52.8% of cases and 46.8% of controls while 11% of cases and 9.5% of controls reported taking a folic acid supplement.
The results of the single SNP analysis, log additive model, are shown in Figure 1 and Supplementary Table 1. Only 2 tagSNPs were associated significantly with CRC risk after correction for multiple testing. Table 3 shows the associations for the 10 tagSNPs with nominally significant associations with risk of CRC. The two significant tagSNPs, DHFR rs1677693 and rs1643659, were strongly linked (r2=0.99) and associated with a decrease in CRC risk (per minor allele OR= 0.87; 0.71–0.94, Pact=0.029 and 0.87; 0.71–0.95, Pact=0.034 for rs1677693 and rs1643659 respectively).
We assessed possible heterogeneity of effects by MSI-status, tumor subsite (proximal, distal, rectal), folate and multivitamin supplement use (yes or no), dietary folate intake (continuous), time to interview (≥ 2 years or > 2 years), ascertainment site (Australia or North America) and family history of CRC (yes or no). There was no significant heterogeneity except with multivitamin use (Table 4). The two DHFR tagSNPs (rs1677693 and rs1643659) and one tagSNP in MTR (rs4659744) were significantly associated with risk only in individuals not using multivitamin supplements. In individuals who did not use multivitamin supplements the per allele OR for both rs1677693 and rs1643659 was 0.69 (95% CI 0.58–0.82; p = 0.00002 and 0.00003 respectively). For individuals using multivitamin supplements these OR’s were 0.91 (0.78–1.07) for both tagSNPs. The p-values for heterogeneity between the two strata were 0.000899 and 0.000986 respectively. For MTR rs4659744 the OR in non-multivitamin supplement users was 0.76 (0.67–0.88; p=0.00011) and 0.97 (0.85–1.11) in supplement users (p for heterogeneity = 0.000265). There was also significant heterogeneity between non-multivitamin and multivitamin supplement users for 4 tagSNPs in MAT2A (rs1446669,rs699664, rs10179195, rs6739015) and one tagSNP in MTRR (rs716537). In each case the per allele OR was non-significantly decreased in non-users and slightly increased in users of multivitamin supplements (Table 4).
In this large family-based case-control study of CRC risk we found that common variants in DHFR and MTR may be associated with a decrease in CRC risk in non-users of multivitamin supplements while there was no association in multivitamin users. No other significant associations were observed for common variants in these 15 FOCM-associated genes involved in nucleotide synthesis, methylation, vitamin B12 transport or polyamine synthesis. There was significant heterogeneity by multivitamin use for 4 tagSNPs in MAT2A and 1 tagSNP in MTRR although neither association was statistically significant.
Dihydrofolate reductase (DHFR) reduces dihydrofolate to tetrahydrofolate (THF). Failure of this reaction can trap folate moieties in an oxidized form that cannot be used for further one-carbon transfers. Inhibition of the DHFR reaction has been exploited for chemotherapy by the drug methotrexate (MTX) and increased DHFR activity is a common cause for resistance to MTX (29). Reduction by DHFR is also involved in the utilization of folic acid, the fully oxidized folate form used for folate supplementation (30).
Variation in DHFR seems to affect folate metabolism. Homozygosity for a 19bp deletion polymorphism in the 1st intron of the DHFR gene has been associated with increased DHFR mRNA (31, 32), a 14% decrease in plasma homocysteine (33), increased red blood cell and plasma folate in women (34), increased circulating folic acid in those with high folic acid intake (>500 µg/day) and decreased red blood cell folate in those consuming less than 250 µg/day (35). A study that assessed the association between this polymorphism and CRC risk reported no association for either CpG island methylator positive (CIMP) or negative tumors (36). Other promoter region (37) or 3’UTR SNPs (38) have been identified in DHFR but no SNPs in the coding region of the gene have been reported (33).
The vitamin B12-dependent enzyme methionine synthase (MTR) catalyses the transfer of a methyl group from 5-methyltetrahydrofolate (5-MTHF) to homocysteine to make methionine and tetrahydrofolate (THF). Interference with this process traps folates as 5-MTHF, which cannot be recycled for further one-carbon transfer reactions (39). The de novo synthesis of methionine is the first step in the synthesis of the universal methyl donor S-adenosylmethionine (SAM) suggesting a possible mechanism by which genetic variation in the MTR gene may influence CRC risk. In this population one intronic tagSNP in MTR (rs4659744) was associated with decreased CRC risk in non-multivitamin supplement users. The association between one non-synonymous SNP in the MTR gene (A2756G, D919G, rs1805087) and CRC risk has been studied by several groups (40–46). The reported associations were inconsistent although a recent meta-analysis reported a small but significant decrease in CRC risk for the GG genotype in European populations (47). The D919G genotype (not linked to rs4659744) was not associated with risk in this population.
TagSNPs in two other genes, MTRR and MAT2A were also modified significantly by multivitamin use, although the individual associations were not statistically significant in either multivitamin use group. MTRR reduces oxidized MTR and B12 after transfer of the methyl group from the B12 cofactor to homocysteine, while the MAT2A enzyme transfers the methyl group from methionine to SAM in an ATP-dependent reaction. The MTRR SNPs rs2303080 and rs2287780 were significantly associated with CRC risk in a recent study of 24 non-synonymous SNPs in 13 folate pathway genes (42) but were not associated with risk overall or in any subgroup in our study.
Martinez et al reported that the association between the MTHFR C677T TT genotype and adenoma recurrence was limited to non-multivitamin supplement users with risk increased in that largely post-fortification population (48). In the current study population, on the other hand, the MTHFR-677 TT genotype was associated with a significantly decreased CRC risk in individuals not using a multivitamin supplement but an OR of approximately 1.0 in those using a multivitamin supplement, results that are similar to those reported here for the other folate pathway gene variants (21).
Numerous other studies of the MTHFR C677T and A1298C polymorphisms and other FOCM-associated genes have assessed interactions between genotypes and indices of folate, alcohol, or other B-vitamins, although most of these studies were undertaken in populations not exposed to fortification of foods with folic acid (36, 41–46, 49–64). All but 5 of these studies (36, 46, 52, 57, 63) support the existence of such interactions. For the MTHFR C677T genotype the majority of studies reported that the TT genotype is more protective in those with a diet high in folate or other methyl-group nutrients or low in alcohol (42–44, 49, 50, 53–55, 60, 64, 65) although other studies have reported no such trends or trends in the opposite direction (45, 51, 59). Although none of these studies looked specifically at multivitamin supplement use, such supplement use is the major source of folates for those in the highest category (66).
Clearly, we do not know for certain which nutrients or what combination of nutrients in multivitamins may modify the association between these FOCM-associated genetic variants and colorectal cancer risk. However, folic acid, B6 and B12 are likely to play the dominant roles. Additionally, in this study we used a tagSNP approach to assess genetic effects so it is not possible to predict how the variations tagged by these SNPs might interact with methyl-group nutrient status. It is notable, however, that over 80% of our study population was recruited after fortification of the North American food supply with folic acid, the synthetic folate used in fortified foods and supplements.
In our population, those not taking multivitamin supplements may have had folate levels more similar to those taking multivitamin supplements than to the nonmultivitamin supplement users in pre-fortification populations. The significant increase in plasma and red blood cell folate in both supplement users and non-users after fortification is well documented (67–69). Thus, for genotype effects that are more relevant at higher folate levels we would expect to see associations in non-multivitamin supplement users in this population.
For the multivitamin supplement users the interaction between genotype and CRC risk may be more complex. First, genotype effects may become less important when mucosal folate levels are maximized, consistent with the OR’s close to 1.0 that we observed in multivitamin supplement users. Additionally, unmetabolized folic acid levels become measurable in the circulation at folic acid levels over 400 µg/d, due to saturation of the DHFR enzyme (30, 70, 71). Such high folic acid intakes may be more prevalent in multivitamin users eating folic acid fortified foods (72). The effect of this on cancer risk is unknown. One in vivo study (73) reported that increased unmetabolized folic acid was associated with a decrease in natural killer cell activity, suggesting a possible decrease in tumor surveillance. Additionally, sustained exposure to high levels of folic acid was associated with a significant decrease in folate uptake in human colon cancer cells in vitro (74). Whether circulating folic acid levels among multivitamin users in fortified populations are high enough to have this effect in vivo is not known but the possibility should be studied further as it suggests a paradoxical decrease in tissue folate stores among those with the highest folic acid intakes. Decreased folate uptake in those with higher folic acid intakes may nullify any protective effect of genotype. Alternatively, the lack of any effect of genotype on CRC risk in multivitamin users may be due to a higher progression risk in multivitamin users with pre-clinical lesions at the time fortification was instituted in 1998. Such an effect would be consistent with the increased neoplasia risk observed in recent studies of post folic acid fortified populations (14–18), the dual effect of folic acid supplementation observed in animal studies (12, 13) and the co-incident rise in CRC incidence in the US and Canada around the time of fortification (19). In this regard it may be of interest that the prevalence of multivitamin use was significantly higher in cases than controls in this study population (Table 2, p<0.01) suggesting increased CRC risk in supplement users. It is possible that this increase in CRC risk among supplement users ameliorated any protective effects attributable to genotype. Whether multivitamin use is a CRC risk factor in post-fortification populations requires further study.
The strengths of our study include the large number of subjects, the comprehensive approach to tagSNP selection and the use of a family-based design which limits the possibility of population stratification. This study has several limitations. In the main analysis we had limited statistical power for SNPs with a minor allele frequency (MAF) of 5% or less (60% power to detect an OR of 1.4). For SNPs with an MAF of 10% we had 80% power to detect an OR of 1.35. We did not have data on all potentially relevant FOCM pathway genes and may have missed some relevant tagSNPs. However, gene coverage ranged from 71% of all identified SNPs for CBS to 98% for SHMT1. The mean coverage was 89%. Although we did take multiple testing into account some possibility of a false positive result remains given the large number of parameters estimated. Similarly the stringent Bonferroni correction may have resulted in some false negatives in stratified analyses. Our study sample was selected to enrich the subject pool for those at higher CRC risk and recruited largely from populations with mandatory supplementation of foods with folic acid, potentially limiting the generalizability of our results.
In conclusion, in these data from a large population-based study we found significant associations between tagSNPs in DHFR and CRC risk in a study of 395 tagSNPs in 15 folate-pathway genes. Our data suggest that two linked tagSNPs in DHFR and one in MTR may mark genotypes that decrease CRC risk in non-multivitamin supplement users in this folate fortified population. Future studies of similarly fortified populations are required to replicate these results.
This work was supported by the National Cancer Institute, National Institutes of Health under RFA # CA-95-011 and through cooperative agreements with the Australasian Colorectal Cancer Family Registry (U01 CA097735), the USC Familial Colorectal Neoplasia Collaborative Group (U01 CA074799), the Mayo Clinic Cooperative Family Registry for Colon Cancer Studies (U01 CA074800), the Ontario Registry for Studies of Familial Colorectal Cancer (U01 CA074783), the Seattle Colorectal Cancer Family Registry (U01 CA074794), and the University of Hawaii Colorectal Cancer Family Registry (U01 CA074806) as well as NCI R01 CA112237 (RWH), NCI T32 CA009142 (J.N.P.) and NCI PO1 CA41108 CA-23074 and CA 956060 (M.E.M). P.T.C. and J.C.F. were supported in part by National Cancer Institute of Canada post-PhD Fellowships (#18735 and #17602).
We thank the following individuals for their support in data collection and management: Margreet Luchtenborg, Maj Earle, Barbara Saltzman, Darin Taverna, Chris Edlund, Matt Westlake, Paul Mosquin, Darshana Daftary, Douglas Snazel, Allyson Templeton, Terry Teitsch, Helen Chen, Maggie Angelakos and Paul Mosquin. We also thank all the individuals who participated in the Colon CFR.
Authors Contributions: AJL and RWH conceived of the study, participated in its design and in drafting the manuscript. JCF, WL, and DVC structured the statistical analyses. WL performed the statistical analyses. KK and DD conducted SNP selection and genotyping. JCF, DVC PTC, PN, JNP, MEM, JLH, LLM, JAB, PJL, CMU and RWH participated in the design of the study and made substantive comments in drafting the manuscript. All authors read and approved the final manuscript.
Potential Conflicts of Interest: D. Conti is a consultant for Pfizer Inc. and P. Limburg is a consultant for Genomic Health Inc.