In the current study, we assessed a comprehensive set of SNPs that characterized the genetic variation of the MTHFR gene, including 10 kb 5′ of the transcription start site and 5 kb into the 3′ UTR. To our knowledge, this is the most comprehensive analysis of genetic variation in the MTHFR gene in relation to colorectal cancer risk completed to date. Our data from both the population-based and clinic-based series are consistent with inverse associations for the 677 TT (rs1801133; A222V) and the 1298 CC genotypes (rs1801131; E429A). Stratification by folate- and multivitamin supplement-use suggested that 677 TT and 1298 CC genotypes may be associated with decreased CRC risk only among non-users of folate or multivitamin supplements, but neither interaction was statistically significant. Our data also support a positive association between the 677 TT genotype and the MSI-H phenotype and tumors of the proximal colon (wherein most MSI-H tumors are located); conversely, the results suggested an inverse association between this SNP and MSS/MSI-L tumors and tumors in the distal colon and rectum. No other SNPs were significantly associated with risk overall or in any subgroup.
Our data are consistent with four recent meta-analyses that suggested 15%–18% reductions in CRC risk for the 677 TT genotype (5
). The A1298C polymorphism has been also well studied. Although the functional effects of the C allele are unclear (2
), two meta-analyses (5
) suggest that homozygosity for the enzyme with a C at nucleotide 1298 is associated with an approximately 10–15% decrease in CRC risk, consistent with the estimate in the current study. Several studies published after the meta-analyses were completed have reported increased risk associated with the 677 TT genotype (23
) while others reported no association (26
) or a decreased risk for those with the variant genotype (29
). Chang et al. reported a significant increase in risk for the 677 TT genotype in those with low folate intake (9
). For the 1298 CC genotype, 2 studies reported an increase in risk (27
) one reported a decreased risk (29
) and one reported no association with risk (9
). Additionally, one recent study of a Hereditary Non-Polyposis Colon Cancer (HNPCC) cohort reported an increase in the age of CRC onset among those with the 1298 CC or joint 677T/1298C genotypes, suggesting a protective effect of the these variant alleles in HNPCC as well (30
). Variability in folate availability in these different source populations may explain the diverse findings, suggesting that future meta-analyses should account for such differences.
Data from observational studies suggest that the phenotype of the MTHFR
valine protein (677 TT genotype) depends significantly on folate availability (31
). A recent in vitro
study in HCT116 colon carcinoma cells reported that the valine protein (TT genotype) was associated with increased genomic DNA methylation in the setting of adequate folate but a significant decrease in genomic methylation when folate was deficient (35
), supporting the impression that folate availability is a modifier of genotype effect. This is consistent with the biochemical changes in the valine-containing enzyme, which show that the enzyme is stabilized by the addition of 5-MTHF to the culture medium (2
). Therefore, we considered whether folate availability might modify associations between other SNPs and CRC risk as well. While there was no heterogeneity by dietary or total folate consumption in the subgroup of participants with food frequency data, in data from the whole study population homozygosity for the MTHFR
677 T allele was significantly associated with a decreased risk only in non-multivitamin users, a result that was suggested for non-users of folate supplements as well. These results conflict with most previous reports of non-folate-supplemented populations, which suggest that the 677 TT genotype is protective mainly for those with higher folate availability (9
), although other studies have not observed this difference (29
). However, in the current study neither the interactions nor the tests for heterogeneity were statistically significant. Therefore this finding may be due to chance. On the other hand, one difference between our study population and those of the earlier studies is the likely higher folic acid levels to which most of our study population had been exposed during the 2-years preceding their recruitment, due to fortification of the food supply. It is likely that non-supplement users in the current population have a greater folate intake than subjects in pre-fortification study populations (39
) with levels more similar to those with higher folate intakes in previous studies. Subjects taking supplements in the current population showed little association with CRC risk, as one would expect if there was stabilization of the enzyme in the presence of high folate availability. Whether post-supplement folate levels are relevant to the lower risk we observed in 677 TT homozygotes not taking folate or multivitamin supplements is unclear and must be assessed in future studies of populations with similar levels of fortification.
In stratified analyses, we found that the 677 TT genotype was associated with a decreased risk of MSS tumors and tumors in the distal colon or rectum while being associated with an increased risk for MSI-H tumors and tumors in proximal colon. The literature on the relationship between the 677 TT genotype and MSI status is mixed. Most studies have reported an increased risk for MSI-H tumors in those homozygous for the 677 TT genotype (9
) but some studies have reported no difference between MSI subgroups (43
) or a decrease in the risk of MSI-H tumors in those homozygous for the 677 T allele (45
). Our data are also consistent with those of a recent study of Australian CRC patients, which reported a significant increase in the risk of proximal and decreased risk of distal colorectal cancers in those with the 677 TT genotype (47
) adding to the general consensus that those with the 677 TT genotype may have an increased risk for tumors with the MSI-H phenotype and for tumors of the proximal colon. These two findings are likely to reflect the same association since data strongly suggest that the majority of MSI-H tumors develop in proximal colon and rarely in distal colon or rectum (47
). We did not observe any modification by MSI status for the 1298 CC genotype, a result that is consistent with those of other reports (41
). It is unclear why the association between the 1298 genotype and MSI status should be different from that of the 677 genotype.
This study has several strengths. In addition to the large sample size, we had a comprehensive approach to identifying genetic variation in the MTHFR gene, a family-based design which minimized the probability of population stratification, included detailed risk factor information on all our subjects and the ability to assess possible heterogeneity of genotype effects by folate nutrition, multivitamin use, MSI status and tumor subsite. The ability to assess genotype associations in two separate samples is another strength.
Weaknesses of this study include the possibility that we have missed an important source of genetic variability since we used public data bases to define SNPs and these data bases are incomplete. Additionally, although the case-unaffected sibling design is more powerful for assessing gene-environment interactions, it is less powerful for detecting main effects (49
). Thus our study may have been underpowered in the main effects analyses. Also, we did not have dietary data from all the participants and not all cases provided tumor tissue for MSI analysis, limiting statistical power for analyses involving diet and MSI status; however, it is unlikely that the availability of dietary or MSI data is associated with genotype, so this should not have resulted in any bias in the observed odds ratios. As in any case-control study, dietary intake information was assessed after the diagnosis in cases and so may be affected by recall bias. We were unable to genotype the rare R593Q (G1793A, rs2274976) polymorphism and so could not assess whether this SNP was associated with risk overall or in any subgroup. The functional consequences of the R593Q polymorphism are not completely clear but some studies have suggested associations with various outcomes (50
). Several studies have shown the A allele for this SNP to be in cis with the 1298 Callele (53
). Finally, many of the protocols used in the C-CFR were designed to over-sample cases with a greater risk for a family history of CRC which may decrease the generalizability of our findings. Additionally, in the population-based sample almost 90% of our study population was Caucasian and from the US or Canada while this percentage was nearly 100% in the clinic-based subjects. Only 19% of the population-based sample was not of North American origin and none of the subjects in either population came from Latin America, South America or Western Europe, all countries with significantly lower folate availability. To the extent that MTHFR genotype modifies risk differently in different countries, perhaps associated with differences in folate intake and the prevalence of smoking and alcohol use, our results may not generalize to all relevant populations.