Article Text


Reversal of DNA hypomethylation by folic acid supplements: possible role in colorectal cancer prevention
  1. J C Mathers
  1. Correspondence to:
    Professor J C Mathers
    Human Nutrition Research Centre, School of Clinical Medical Sciences, University of Newcastle, Newcastle upon Tyne, NE1 7RU, UK;

Statistics from

Low folate intake may predispose to a greater risk of colorectal cancer

The deranged gene expression that is causal for colorectal cancer (CRC) development results from unrepaired genomic damage which accumulates in successive generations of mucosal cells and which provides the neoplasm with a growth and/or survival advantage. Such damage includes disabling mutations in tumour suppressor (TS) genes and facilitates inappropriate expression of oncogenes. Approximately 25 years ago, evidence began to appear that changes in methylation status of DNA might be responsible for changes in gene expression. The potential importance of aberrant DNA methylation in tumorigenesis was signalled when, on 6 January 1983, Nature carried an article by Feinberg and Vogelstein1 reporting the first evidence of hypomethylation of some genes in tumours compared with “normal” tissue from the same individuals. That same year, Gama-Sosa and colleagues2 used high performance liquid chromatography to demonstrate that metastatic tumours had a significantly lower content of 5-methylcytosine than did benign neoplasms or normal tissues. Somewhat surprisingly, site specific hypermethylation of particular genes, including the classical TS the retinoblastoma gene, was observed a few years later (see Feinberg and Tycko3 for an excellent review). It is now clear that global genomic hypomethylation and hypermethylation of specific genes coexist in the same tumours but whether the genesis of the two phenomena is related remains uncertain.


Laird4 has argued that cancer may be as much a disease of misdirected epigenetics as it is of genetic mutations. Epigenetics describes non-coding changes to the genome which are transmitted through mitosis and alter gene expression. Although changes in other epigenetic markings, especially in post translational modifications of histones, are likely to be of considerable significance in tumour pathogenesis, much of the work to date has focussed on DNA methylation. Approximately 1% of DNA bases in the human genome is accounted for by 5-methylcytosine which occurs predominantly in CpG dinucleotides. The promoter regions of about half of human genes contain CpG islands which are contiguous windows of 500 or more base pairs in which the G:C content is at least 55% and the observed over expected CpG frequency is at least 0.65.4 During cell division, methylation patterns in the parental strand of DNA are maintained in the daughter strand by the action of DNA methyltransferase 1 (DNMT1) which catalyses the transfer of a methyl group from S-adenosyl methionine (SAM; the universal methyl donor) to the 5′ position on cytosine residues by a relatively complex mechanism.5

Unlike cytosines elsewhere in the genome, those in CpG islands are normally unmethylated. Methylation of these CpG islands is accompanied by the transcriptional silencing of a range of genes important in tumorigenesis, including the colorectal mucosal gatekeeper APC, the cyclin dependent kinase inhibitor encoded by CDKN2A, and DNA repair genes most notably MLH1 and MGMT although whether hypermethylation is causal in this process continues to be debated.5,6 Given the TS roles of the protein products of these genes, the aetiological consequences of promoter hypermethylation for tumour development are likely to be indistinguishable from the effects of transcriptional silencing by mutation. Not all genes with CpG islands in their promoters become hypermethylated in cancers. For example, MLH1 and MSH2 are both DNA mismatch repair genes with CpG islands but only MLH1 is hypermethylated (observed in up to one fifth of bowel cancers). This suggests that, for reasons unknown, some genes are differentially sensitive to methylation.6

Early studies of DNA hypomethylation in tumorigenesis suggested a role in activation of oncogenes such as HRAS and KRAS7 which has been confirmed for several additional genes.3 Knockout of the housekeeping DNA methyltransferase Dnmt1 in mouse somatic cells resulted in widespread gene activation.8 However, oncogenic activation is unlikely to be the only means by which global hypomethylation contributes to neoplasia. Hypomethylation appears to induce genomic instability through abnormalities in chromosomal segregation processes9 and elevated mutation rates.10 However, neither point mutations nor genomic rearrangements were increased in a separate study of Dnmt1 deficient mouse embryonic stem cells.11


Diet, lifestyle, and other non-genetic factors have a strong influence on CRC risk, with obesity and possibly red or processed meat increasing the risk while higher intakes of plant foods may be protective.12 Several components of plant foods may inhibit, reverse, or retard tumorigenesis, with intracellular signalling cascades as the common molecular targets.13 Higher intakes of the B vitamin folate or higher folate status are associated with a lower risk of bowel cancer (see Kim14 for review). Plausible mechanistic arguments can be advanced for this observation since (i) folate in the form of 5′-methyltetrahydrofolate is required for the synthesis of SAM and, therefore, for DNA methylation and (ii) 5′, 10′-methylenetetrahydrofolate donates a methyl group to uracil converting it to thymine for DNA synthesis. When folate status is low and thymine supplies are limited, dUTP is incorporated into DNA in place of dTTP during DNA replication and repair. The resulting U:A mismatches initiate a futile cycle of DNA repair with single and double strand breaks and chromosomal damage as probable sequelae.15

To date, at least seven relatively small (1–20 subjects) human intervention trials of the effects of folate supplementation on putative biomarkers of CRC risk measured in the colorectal mucosa have been published (see Pufulete and colleagues16 for details of some) but all of these have used relatively large doses of folic acid (1–15 mg/day). In the randomised, double blind, placebo controlled intervention study reported by Pufulete et al in this issue of Gut,16 global genomic DNA methylation and folate status were assessed in patients with adenomatous polyps before and after supplementation with 400 μg folic acid/day (twice the reference nutrient intake for this vitamin) or placebo for 10 weeks (see page 648). As expected, circulating concentrations of folate increased and homocysteine concentrations fell following folate supplementation and there was a 31% increase (p<0.001) in global DNA methylation in leucocytes. Earlier studies by others have shown that a few weeks of moderate folate depletion resulted in global hypomethylation of circulating white cells but the effects on DNA methylation of repletion with folate for similar time periods were inconsistent.17–19 However, the usefulness of leucocyte DNA methylation as a surrogate for that in other tissues and especially for the colonic mucosa is not known. Pufulete and colleagues16 have made the novel observation that supplementation with physiologically normal amounts of folate increased genomic DNA methylation in biopsies of colonic mucosa by 25% (albeit, not significantly, p = 0.09).


Having observed a similar degree of hypomethylation of DNA in benign polyps as in malignant tissue, Goelz and colleagues20 proposed that methylation changes precede malignancy and “could be a key event in the initiation of malignancy”. While it is probable that changes in global methylation which result in activation of oncogenes or genomic instability contribute to the development of neoplasia, it is not clear that all changes in genomic DNA methylation will have such an effect. Much of what is known about the functional effects of altering global DNA methylation comes from studies in which very large changes have been induced by treatment of cells with demethylating drugs (for example, 5-azacytidine)9 or by knockout of Dnmt1.8,10 Because of their potential importance in advancing understanding of the mechanistic basis of dietary protection against CRC, the functional consequences of the more subtle effects produced by alterations in nutrient (for example, folate) supply16 is now a priority for further research. Such research should include gene expression studies employing transcriptomics and/or proteomics approaches to reveal global changes in expression which may accompany the global changes in methylation.

The most widely used methods for assessing global DNA methylation (including the in vitro acceptance assay21 and the cytosine extension assay22) yield estimates of the proportion of cytosines which are available for methylation but give no indication of how these cytosines are distributed across the genome. It is possible that alterations in cellular folate status result in methylation changes within “junk” DNA regions with no obvious implications for genomic stability or for gene expression. More informative techniques are needed for mapping changes in genomic methylation in response to nutritional or other interventions. Possible approaches to this problem include developments of the enzymatic regional methylation assay23 or exploitation of the power of pyrosequencing following bisulphite modification of DNA.24,25 Characterisation of regional changes in DNA methylation patterns should be linked with investigations of expression of associated genes.

Although CRC is a focal disease arising from genomic damage to mitotically competent cells within individual crypts, it is widely assumed that there are “field” changes across large regions of the macroscopically normal mucosa which predispose to, or are biomarkers of the risk of, tumour development. This is the basis for attempts to develop surrogate end points which can be measured in pinch biopsies of colorectal mucosa.26 Support for this concept comes from studies of crypt cell proliferation which demonstrate markedly different values between individuals but relatively low inter biopsy variation within subjects.27 Before global DNA methylation can be accepted as a potential biomarker of bowel cancer risk it will be important to undertake studies of the heterogeneity of this putative marker between biopsies taken at the same site and between anatomical sites along the colorectum.

In respect of global DNA methylation, Pufulete’s observations16 suggest that leucocytes may be a more easily obtained surrogate for colonocytes but we do not know how closely matched are baseline values or changes in response to supplementation in the two tissues. Exploitation of faeces as a non-invasive source of information about global DNA status in the colorectal mucosa is an obvious target following the recent demonstration that the methylation status of promoter regions of genes can be quantified in human DNA recovered from stool samples.28,29 However, the huge preponderance of bacterial over human DNA in stool means that the commonly used techniques for assessing global DNA methylation will be inappropriate unless complete separation of the two DNA sources can be achieved or polymerase chain reaction based techniques are used to pull out specific human sequences.

In summary, observational studies in human populations together with investigations in animal models and cell systems provide the evidence for the hypothesis that low folate intakes predispose to greater risks of CRC. Aberrant DNA methylation with its consequential effects on gene expression and on genomic stability offer a plausible causal link between cellular folate status and neoplastic development and therefore it will be important to build on the findings of Pufulete and colleagues16 to determine optimal intakes of folate to minimise the risk of bowel cancer. Given the recent observation that high dose folate supplementation during pregnancy may increase the risk of death from breast cancer30 it would be unwise to assume that, in respect of folate, more of a good thing is necessarily better.

Low folate intake may predispose to a greater risk of colorectal cancer


View Abstract


  • Conflict of interest: None declared.

Request permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

Linked Articles