Folate-dependent one-carbon metabolism is essential for nucleotide synthesis, methionine synthesis and for DNA methylation and depends on a number of enzymes that are functionally polymorphic. SNPs and other polymorphisms of the genes MTHFR, MTRR and TYMS have been associated with cardiovascular disease, neural tube defects, Down syndrome and leukaemia 1234. Adequate dietary folate, and its efficient metabolism have well-established roles in disease prevention; deficiency is associated with increased risk of neural tube defects, vascular disease, cancer and anaemia 5.

Proposals for the mandatory fortification of flour with folate in the United Kingdom are driven by the importance of this B vitamin to the development of the early embryo 6. Embryo development and viability has been the subject of many epidemiological studies of genetic variants, including, and possibly predominantly, in those genes involved in folate metabolism 2347. B-vitamin status and folate metabolism have been shown to influence the outcome of in vitro fertilisation (IVF) 89. Women who are homozygous for a variant of the gene MTHFR (1298A>C) were less likely to produce a live-birth than other women undergoing IVF and the chances of a twin birth increased with plasma and red cell folate concentrations 8. The authors proposed that the woman's genotype is linked to her potential to produce good quality embryos and that this, coupled with folate status, increases the likelihood of IVF resulting in live births. Haggarty et al also observed an association between raised concentrations of homocysteine and an increase in the risk of miscarriage in their study of women undergoing IVF and elevated homocysteine levels has been linked recurrent pregnancy loss previously 10.

A meta-analysis has demonstrated that maternal hyperhomocysteinaemia (a metabolic consequence of low folate status) is associated with orofacial cleft and congenital heart defects 11, but the authors found no independent association between these congenital malformations and MTHFR genotype in either the mothers or children. However, the negative in utero effect of raised homocysteine has been proposed by Lucock and Yates as one of the mechanisms by which the genetic variant, MTHFR 677C>T, compromises the viability of a pregnancy when dietary folate levels are inadequate 12. These authors also hypothesise that when folate levels are unrestricted, there is a survival advantage in utero for homozygotes for the 677C>T allele because that genotype will increase the levels of 5,10 methylene tetrahydrofolate favouring DNA synthesis and stability. This hypothesis is supported by the authors of a recent study of the prevalence of this genotype in adult populations and in DNA samples taken from spontaneous abortions, they present evidence for the selection of the 677C>T allele associated with increased folate intake by women in the peri-conceptional period 13.

The MTHFR 677C>T variant is associated with raised homocysteine concentrations and lower enzyme concentrations 14. Association studies of the MTHFR 677C>T variant and adverse pregnancy outcomes have in the main concluded that this variant is important in the pathogenesis of adverse pregnancy outcomes 15. Less compelling evidence is available for TYMS genetic variants 16, one of which is a 5'UTR repeat variant (2R/3R) and the rare 2R variant is associated with lower TYMS expression and plasma homocysteine 17. The simultaneous analysis of multiple genetic variants allows the consideration of additive, synergistic and compensating variants of folate metabolism. Disease association studies of two or more enzyme variants of folate metabolism are few, but suggest the possibility of gene-gene interactions or epistasis. One study has demonstrated a multiplicative effect on phenotype for the MTHFR 677C>T and MTRR 66G>A genotype combinations in relation to Down syndrome 4. Another study demonstrated that three genotype combinations of the loci MTRR/FOLH1; MTHFR 677/CBS and MTHFR 677/MTRR increase the risk of neural tube defects 7.

A functional study of MTHFR and TYMS has shown that these enzymes compete for limiting supplies of folate required for homocysteine methylation 17; which illustrates the likelihood that certain steps in one-carbon metabolism present bottlenecks for the distribution of folate species, and enzyme variants in these steps may be important.

We and others have previously found evidence for depletion of individuals homozygous for MTHFR 677C>T relative to younger cohorts, and postulated it to be due to genotype-specific survival 181920. Because of observations made in cross-sectional population studies to explore the impact of genetic variants of the enzymes MTHFR and TYMS on survival in elderly cohorts, we have studied the prevalence of two polymorphisms in younger populations. The MTHFR and TYMS loci are found on chromosomes 1 and 18 respectively so linkage disequilibrium will not explain any digenic phenomena observed.

We have observed a consistent gene-gene interaction in 5 different cohorts. Certain genotypes are differentially represented in these populations and this could be evidence of genetic selection during early development; in gamete formation or in utero, with some gametes or pregnancies more or less viable because of variation in the way folate is handled at the cellular level, the metabolism of folate being the link between these enzymes and their genetic loci.

There is published evidence for significant genome-wide transmission distortion 25 which could be caused by genetic and/or environmental influences on gamete selection and embryo viability. Ebisch et al demonstrated that sperm concentration can be raised with folic acid and zinc sulphate supplementation, but only in males who are homozygous for the 677C genotype 26. Lucock and Yates have hypothesised that when folate levels are unrestricted, there is a survival advantage in utero for homozygotes for the 677C>T allele because that genotype will increase the levels of 5,10 methylene tetrahydrofolate favouring DNA synthesis and stability 12. The investigations by Mayor-Olea et al found that the wild type, 677CC, genotype is almost absent in spontaneous abortions suggesting that this genotype may protect against pregnancy loss. However, they also interpret the increase in the prevalence of the T allele alongside the use of peri-conceptional folic acid as an indication that an environmental factor can lead to the genetic selection for the T allele or permit the survival of deleterious genotypes 13.

Our observation could not be caused by any simple genotyping error because the relationship observed between MTHFR and TYMS is independent of heterogeneity in the allele frequencies for the individual polymorphisms between the cohorts. During the course of these experiments we investigated the possibility of technical artefacts because it can be seen from table 2 that some of the individual gene frequencies were out of HWE (see MTHFR frequency for cohort 4 and TYMS frequencies for cohorts 1 and 3) and genotyping errors could explain this. Undigested PCR product in the analysis of the MTHFR 677 polymorphism, and preferential amplification of the TYMS allele with 2 rather than 3 repeats could produce erroneous data for the individual genotypes. Although only a single laboratory assay was used for each genotype analysis, the DNA quality and data interpretation checks described in the methods provide confidence in the reproducibility of our methods. Futhermore, while genotyping errors could result in perturbation of HWE they would not generate the significant co-segregation of MTHFR and TYMS genotypes described.

Deviation from HWE is also possible if there are selective advantages for particular genotypes. A selective advantage for particular MTHFR genotypes could lead to deviations from HWE in the general population; heterozygote advantage occurs when environmental situations lead to heterozygous individuals for some disease alleles having increased fitness over both homozygous genotypes. In the older cohorts (1, 2 and 4) in this study both homozygous groups showed a reduced fitness relative to the heterozygotes for MTHFR 677 (see table 2). We and others have previously found evidence for depletion of individuals homozygous for the MTHFR 677C>T relative to younger cohorts, and postulated it to be due to genotype-specific fatal disease 18192024. However, given that the interaction between MTHFR and TYMS is found in all age groups, this interaction cannot be interpreted as an age-related survival factor; i.e. the death of some individuals from genotype-specific fatal disease does not explain the observation.

Given that all of the subjects from our study were born before the influence of folate related public health initiatives, could B-vitamin supplementation affect the biological phenomena that we and others have described? Haggarty et al report that folic acid intervention has not resulted in genetic selection in favour of the MTHFR 677T allele 27, however this does not take into account potential gene-gene interactions at other folate-related loci and does not discount the possibility that numerous genetic variants may act in concert to exert selective pressure.

Our data may suggest a gene-gene interaction and could be evidence of genetic selection, with some pregnancies more or less viable as a consequence of genetic variation. If these genetic phenomena affect the way folate is handled at the cellular level in utero it is possible that maternal folic acid intake may over-ride such genetic selection. Our findings should now be tested in independent population cohorts.

HUGO: approved nomenclature used for the gene loci; CBS: cystathionine-beta-synthase; FOLH1: folate hydrolase (prostate-specific membrane antigen) 1; MTHFR: 5,10-methylenetetrahydrofolate reductase (NADPH); MTRR: methyltetrahydrofolate-homocysteine methyltransferase reductase; TYMS: thymidylate synthetase; HWE: Hardy-Weinberg Equilibrium.

The authors declare that they have no competing interests.

GW, BJ and CR participated in the design of the study and contributed to the drafting or preparation of the manuscript. GW designed and optimised most of the PCR assays, collected and analysed most of the data. BJ also analysed the PCR data. CR also analysed the genotyping data by HWE. JS advised on and completed the statistical analyses. All authors contributed to and approved the final manuscript.