Epigenetics and regulation of gene activity

Epigenetics is defined as processes that act to regulate heritable changes in gene activity that are transmitted through meiosis and mitosis but that are not accompanied by changes in the DNA coding sequence. Epigenetic signals (or marks) control gene expression through remodelling of chromatin. Chromatin consists of DNA and DNA-binding proteins including histones, which interact to form repetitive nucleoprotein units called nucleosomes. Epigenetic control of gene expression is critically important in regulating embryonic development, cellular differentiation and organogenesis and for key biological processes such as imprinting, and silencing of large chromosomal domains such as the X-chromosome in females⁽Reference Jones and Baylin¹⁾. Conversely, dysregulation of epigenetic processes may be causal for several human diseases including cancer and heart disease⁽Reference Jones and Baylin^1, Reference Turunen, Aavik and Yla-Herttula²⁾ (Fig. 1).

Fig. 1. Epigenetics, diet and human disease. A simplified scheme illustrating the structure of mammalian chromatin and defining the role of epigenetic modifications both in normal human development and disease.

Fig. 2. Folate deficiency, cancer and CVD: a common mechanism? A simplified scheme describing how folate deficiency may alter normal DNA methylation and how this could impact both on risk of colon cancer and atherosclerosis.

Epigenetic mechanisms include nucleosome remodelling, histone modifications (including acetylation, methylation and ubiquitination), regulatory non-coding (small and micro) RNA and DNA methylation. These are linked processes and all contribute to controlling gene activity by changing chromatin architecture and/or access by transcription factors⁽Reference Jones and Baylin¹⁾.

Dysregulation of any of these processes has severe consequences for the phenotype and function of the cell. During malignant transformation, cancer cells can exhibit global changes in the chromatin structure affecting the whole epigenome, altering the expression of several hundreds of genes and perturbing entire metabolic pathways. Consequently, it has been proposed that epigenetic dysregulation can be as detrimental to the cell as mutations in DNA coding⁽Reference Jones and Baylin¹⁾.

Epigenetics, diet and disease

Although epigenetic marks remain comparatively stable during mitotic transmission, they are influenced by ageing and by the environment. In an elegant study of monozygotic twins aged between 3 and 74 years (n 40 pairs), genome-wide cytosine methylation, gene-specific DNA methylation and histone acetylation (H3 and H4) patterns in lymphocytes were found to correlate strongly in young twins, but these associations weakened both with age and depending on how divergent the twin's health and lifestyles had become⁽Reference Fraga, Ballestar and Paz³⁾. Approximately one-third of twins harboured epigenetic differences. Global gene expression, measured by microarray analysis, differed more that fourfold in the oldest twin pair compared to gene activity in the youngest⁽Reference Fraga, Ballestar and Paz³⁾.

Diet can profoundly alter epigenetic patterns in animals. The Agouti mouse has been used extensively to investigate the impact of maternal nutrition on the fetal epigenome and the phenotype of the offspring. In this model, coat colour is linked to the methylation status of the agouti gene, which is highly dependent on maternal diet. Most importantly, altered epigenetic marking is associated with an increased risk of diseases including a diabetes-like condition, obesity and tumorigenesis⁽Reference Yen, Gill and Frigeri^4, Reference Dolinoy⁵⁾. The influence of diet on epigenetics and fetal programming in relation to health and disease will be covered in greater detail in a subsequent paper (KA Lillycrop, in this issue).

How diet contributes to epigenetic control of gene expression and human disease risk is poorly understood. It is acknowledged that abnormal epigenetic control of gene expression increases disease risk (certainly for cancer) and that diet can also profoundly influence risk. What remains to be established is a causal link between diet and epigenetics in the development of human disease and whether diet is differentially effective at various stages in the human lifespan.

Certain bioactive food components, such as sulforaphanes from broccoli, diallyl disuphides from garlic and resveratrol in wine, have been shown (in vitro and in vivo) to alter epigenetic processes with positive consequences for cell function, including control of proliferation, up-regulated apoptosis and a reduction in inflammation⁽Reference Ross, Dwyer and Umar⁶⁾. All of these food components are associated with a reduced risk of human cancer at various sites. Inhibition of histone deacetylase activity increases the transcription factor access to DNA and induces gene reactivation. Isothiocyanates present in cruciferous vegetables inhibit histone deacetylase activity in colon, breast and prostate cancer cells in vitro. Sulforaphane, a powerful inducer of phase 2 detoxification enzymes, increases histone acetylation and inhibits intestinal polyp formation in Apc min mice. Moreover, histone deacetylase activity (H3 and H4) is rapidly inhibited (3–6 h) in blood cells from human volunteers fed a relatively small single bolus (68 g) of broccoli sprouts (reviewed in⁽Reference Dashwood and Ho⁷⁾). Conversely, diallyl disulfide from garlic increases H3 and H4 histone acetylation, tumour suppressor (p21) gene expression and inhibits colon cancer cell growth in culture (reviewed in⁽Reference Druesne-Pecollo, Chaumontet and Latino-Martel⁸⁾). The polyphenols curcumin (from the spice turmeric) and resveratrol (from red wine), which have strong antioxidant, anti-inflammatory and anti-carcinogenic properties, also modify histone acetylation patterns in circulating inflammatory cells and induce histone deacetylase activity⁽Reference Rahman⁹⁾.

In contrast, the impact of diet on recently identified epigenetic processes, for instance, endogenous non-coding or regulatory (small and micro) RNA that suppress gene activity at the transcriptional level and where dysregulation may increase the risk of cancer remain unexplored⁽Reference Guil and Esteller¹⁰⁾.

Most studies investigating how diet alters the epigenome have explored how dietary components alter DNA methylation.

Aberrant DNA methylation and abnormal gene expression

Post-replicative DNA methylation in mammals involves addition of a 1-C methyl group at the fifth carbon position of cytosine residues within CpG dinucleotides. S-adenosylmethionine (SAM) acts as methyl donor in this reaction, which is catalysed by DNA methyltransferase (DNMT) enzymes. DNMT 3 (3a and 3b) regulates de novo methylation during development, whereas DNMT 1 maintains DNA methylation patterns during cell replication. Approximately 4% of cytosines in DNA are modified to 5-methylcytosine, with most occurring in the sequence, CpG. The genome methylation pattern is precisely inherited during mitosis and is highly tissue specific and species specific. Cytosine methylation changes the structure of the major groove in the DNA molecule and disrupts the attachment of DNA-binding proteins and transcription factors. In general, genes methylated at specific sites (for example upstream of a promoter region) are either not transcribed into mRNA or are transcribed at a reduced rate, and translation into the protein for which the gene encodes is reduced. Epigenetic DNA methylation therefore contributes to the control of gene and ultimately protein expression⁽Reference Costello and Plass¹¹⁾.

Most information on how dysregulated DNA methylation affects human disease risk comes from research into cancer. Two types of aberrant DNA methylation are universally present in tumours. DNA, both from benign and malignant carcinomas, is substantially undermethylated (hypomethylated) compared to DNA from adjacent normal tissue. DNA hypomethylation is associated with increased transcription and expression of proto-oncogenes that stimulate malignant cell growth. Genome-wide DNA hypomethylation occurs early in carcinogenesis and precedes mutation and deletion events. Conversely, increased DNA methylation (hypermethylation) is associated with gene silencing. Hypermethylation of cytosines in the promoter region of tumour suppressor (TS) genes (for example DNA repair genes and genes that down-regulate cell proliferation) is linked with cancer progression. Gene-specific DNA hypermethylation occurs both in early and advanced stages of malignant transformation. Aberrant DNA methylation and gene expression patterns may be causal in tumorigenesis⁽Reference Gaudet, Hodgson and Eden¹²⁾.

Many bioactive food compounds including folate, selenium, flavonoids, alcohol and fatty acids are associated with an altered risk both of human colon cancer and heart disease. All of these dietary components modify cytosine methylation. The impact of several of these phytochemicals and micronutrients on DNA methylation in colon carcinogenesis are reviewed comprehensively elsewhere⁽Reference Johnson and Belshaw^13, Reference Arasaradnam, Commane and Bradburn¹⁴⁾.

The remainder of this review describes how specific dietary polyphenols reactivate TS genes by reversing DNA hypermethylation and the consequence of this for risk of cancer and how the vitamin B folate modifies cytosine methylation in relation to both heart disease and colon carcinogenesis.

Polyphenols, DNA methyltransferase activity and tumour suppressor gene re-expression

Hypermethylation of CpG islands in the gene promoter contributes to gene silencing. TS gene hypermethylation is a universal event in early cancer, with more genes becoming hypermethylated as the disease progresses. Drugs such as 5-aza-2′-deoxycytidine (decitabine) that reverse DNA hypermethylation and re-establish TS gene expression are used in cancer therapy. However, they are genotoxic to normal cells. Maintenance of normal DNA methylation and gene expression patterns by dietary phytochemicals and reversal of methylation-induced inactivation of TS genes may be an alternative approach for the prevention and treatment of cancer.

Polyphenols, a group of plant phytochemicals consumed in significant amounts in the human diet modulate genomic stability and risk of cancer in vitro and in vivo ⁽Reference Duthie, Duthie and Kyle¹⁵–Reference Duthie¹⁷⁾. There is strong evidence that polyphenols from tea, soft fruits and berries, vegetables, apples, and even from wine, are potent anti-carcinogenic agents in vitro and in animal models that prevent DNA instability at several sites in the carcinogenic pathway⁽Reference Duthie, Duthie and Kyle¹⁵–Reference Duthie¹⁷⁾. Genoprotective mechanisms include modulation of carcinogen metabolism (either by inhibiting the activation of cytochromes P450 or by inducing detoxification by glutathione transferase and glucuronyltransferase), decreased binding of the carcinogen to DNA, inhibition of oxidative DNA damage, alteration in cell signalling and gene expression, reduced inflammation, increased apoptosis and inhibition of malignant transformation, cell invasiveness, angiogenesis and metastasis⁽Reference Duthie, Duthie and Kyle¹⁵⁾. Moreover, several polyphenols are potent inhibitors of DNMT activity in vitro, capable of reversing DNA hypermethylation and reactivating TS gene activity. Polyphenols inhibit DNMT activity and DNA methylation in two ways. First, by direct insertion into the binding pocket of DNMT (competitive inhibition) and second, indirectly by decreasing intracellular SAM concentrations and non-competitively inhibiting DNMT activity⁽Reference Fang, Chen and Yang¹⁸⁾. Tea and soya are two of the most widely consumed plant products worldwide. Tea components (especially green and black tea) and specific soya isoflavones inhibit DNMT activity in human cancer cells. Genistein from soyabean and (−)-epigallocatechin 3-gallate (EGCG) from green tea are the most potent DNMT inhibitors. EGCG (and other green tea metabolites) dose-dependently inhibit DNMT activity in human oesophageal cancer cells⁽Reference Fang, Chen and Yang¹⁸⁾. Molecular modelling reveals that the gallate moiety on the D ring of EGCG interacts strongly with the cytosine-active site on the DNMT enzyme. Additionally, hydrogen bonds formed between hydroxyl groups of the A and B rings and Sr1229 and Cys 1225 on the protein contribute to the high-affinity binding of EGCG and inhibition of activity⁽Reference Fang, Chen and Yang¹⁸⁾. Genistein (from soya) was also found to interact strongly with DNMT, dose-dependently inhibiting enzyme activity⁽Reference Fang, Chen and Yang¹⁸⁾. Certain polyphenols are substrates for catechol O-methyltransferase, which is structurally analogous to DNMT. SAM is consumed in the reaction and intracellular S-adenosylhomocysteine (SAH), a potent inhibitor of DNMT is elevated. Polyphenols including EGCG, quercetin, myricetin and fisetin may decrease DNA methylation by altering the cellular ratio SAM:SAH and indirectly inhibiting DNMT activity. There is currently little evidence in support of this mechanism. Feeding mice green tea extract containing EGCG (65% (v/v) for 7 d) or transgenic Apc min mice (0·16% (v/v) EGCG for 9 weeks) did not alter intestinal or liver SAH⁽Reference Fang, Chen and Yang¹⁸⁾.

DNMT 1 inhibition demethylates CpG islands in the promoter regions of silenced TS genes including p16^ink4a, retinoic acid receptor β, methylguanine methyltransferase, mMLH1 and glutathione S-transferase π in cancer cells. Reactivation of TS genes is associated with a corresponding increase in mRNA and protein expression. EGCG (20–50 μm for up to 6 d) reverses DNA hypermethylation and increases expression of methylguanine methyltransferase, p16 and hMLH1 in cultured KYSE 510 oesophageal cancer cells. Changes in cytosine methylation and mRNA expression were detected after only 48 h exposure to the green tea polyphenol and were progressive with time. Moreover, reversal of DNA hypermethylation was associated with increased protein expression (determined by Western blotting). EGCG can also reactivate retinoic acid receptor β in prostate and breast cancer cells, p16 in colon cancer cells and glutathione S-transferase π in prostate cells⁽Reference Fang, Wang and Hou¹⁹–Reference Shukla, Trokhan and Resnik²¹⁾. Genistein (at the same concentrations) partially reversed DNA hypermethylation and reactivated p16, retinoic acid receptor β and methylguanine methyltransferase gene expression⁽Reference Fang, Chen and Sun²²⁾. The isoflavones biochanin A and daidzein and the flavonoids, myricetin, quercetin, hesperitin, naringenein, apigenin and luteolin were active, but less effective at altering DNA methylation and reactivating TS gene expression⁽Reference Fang, Chen and Sun²²⁾. Reactivation of TS gene expression by polyphenols is associated with reduced cancer cell progression. Malignant transformation involves phenotypic changes in initiated cells that promote invasion into extracellular tissue matrices. The TS matrix metalloproteinase (MMP) inhibitor RECK inhibits angiogenesis, invasion and metastasis. RECK activity is down-regulated in several human cancers and is associated with promoter hypermethylation in the gene. In an elegant study, Kato et al.⁽Reference Kato, Long and Makita²³⁾ investigated how EGCG altered RECK DNA methylation and expression in four human oral squamous cell carcinoma cell lines. Critically, they investigated whether reversing hypermethylation and increasing RECK gene expression could inhibit oral squamous cell carcinoma migration and invasiveness. EGCG (20–50 μm for up to 6 d) dose- and time-dependently decreased RECK DNA methylation and increased mRNA in two of the four cell lines. Significantly, this was associated with a subsequent decrease in MMP-2 and MMP-9 expression and in the ability of the cancer cells to invade a three-dimensional collagen model. Cell proliferation, depth of migration and frequency and size of invasive foci were all reduced by EGCG⁽Reference Kato, Long and Makita²³⁾. However, the ability of EGCG to decrease DNA hypermethylation and reactivate TS gene expression and activity is inconsistent, with some studies reporting no effect of EGCG on DNA methylation in several human cancer cell lines⁽Reference Chuang, Yoo and Kwan^24, Reference Stressman, Brueckner and Musch²⁵⁾.

Evidence for a modulating effect of dietary polyphenols on DNA methylation and re-expression of TS genes in vivo is less convincing. Consumption of genistein by mice appears to correlate positively with changes in prostate DNA methylation at CpG islands⁽Reference Day, Bauer and DesBordes²⁶⁾. Conversely, feeding a composite of green tea polyphenols (0·1, 0·3 and 0·6% (v/v) in drinking water for 12 or 24 weeks) to wild-type or TRAMP (transgenic adenocarcinoma of mouse prostate) mice did not change normal or cancer-specific DNA methylation patterns either in target prostate tissue or in gut and liver tissue⁽Reference Morey-Kinney, Zhang and Pascual²⁷⁾. In this model, genome-wide DNA hypomethylation occurs at an early stage in cancer progression with loci-specific hypermethylation detected in late-stage disease. No significant dose- or time-dependent effect of GTP on either global- or gene-specific DNA methylation (measured by several different biomarkers) was observed⁽Reference Morey-Kinney, Zhang and Pascual²⁷⁾.

In general, the concentrations of polyphenols employed to inhibit DNMT activity both in vitro and in vivo are considerably higher than can be achieved nutritionally. The levels of EGCG shown to be effective in reversing DNA hypermethylation and TS gene silencing (generally 10–50 μm) are 50-fold higher than blood and tissue concentrations measured after tea drinking⁽Reference Fang, Chen and Yang¹⁸⁾. The highest dose of GTP studied in the TRAMP mouse model (0·6% (v/v)) is equivalent to a human drinking eighteen cups of green tea per day. Similarly, the concentration of genistein needed to alter cytosine methylation and TS gene activity (5–20 μm) are 3–10-fold higher than can be achieved by eating soya products⁽Reference Fang, Chen and Yang¹⁸⁾. Dietary polyphenols are rapidly metabolized intracellularly through several biochemical pathways, including glucuronidation, sulfation and methylation. The lack of effect of polyphenols on DNMT activity, DNA methylation and gene expression in vivo is probably due to limited functional bioavailability post-metabolism. Whether dietary polyphenols can modify DNA methylation and TS gene expression in human subjects is unknown, but it is doubtful whether these compounds, consumed at nutritional levels, will have a major impact on DNA methylation and gene reactivation.

Folate deficiency and human diseases

Folates, a family of water-soluble B vitamins, play a crucial role in the development of human diseases such as cancer and heart disease in adults, cognitive dysfunction and dementia in the elderly and congenital defects in babies⁽Reference Czeizel²⁸–Reference Duthie, Whalley and Collins³⁰⁾. Suboptimal folate status is widespread with 40% of 15–18 year olds in the UK exhibiting marginal folate status and overt folate deficiency common in people over 65 years of age, especially in the institutionalized elderly.

Low-folate status has been implicated in cancer development, notably of the cervix, lung, breast, brain, colorectum and pancreas. The evidence linking folate deficiency and human cancer is strongest for the colon. Data from the majority of human studies (retrospective, case–control and prospective) suggest that individuals with the highest habitual folate intake or with the highest circulating folate concentrations have a reduced risk of developing colon polyps or tumours⁽Reference Giovannucci³¹–Reference Kim³³⁾. Folate acts to maintain genomic stability by regulating DNA biosynthesis, DNA repair and DNA methylation. Folate deficiency both initiates and accelerates cancer by disrupting each of these processes.

Poor folate status is also associated with an increased risk of CVD. Historically, this has been attributed to the metabolic relationship between low folate and high homocysteine (hyperhomocysteinemia). Adverse effects of homocysteine include impairment of endothelial and smooth muscle cell (SMC) function, changes to vasodilation, altered extracellular matrix biochemistry, endoplasmic reticulum stress and increased coagulation and impaired fibrinolysis⁽Reference McNulty, Pentieva and Hoey³⁴⁾. Most human observational studies support a positive relationship between homocysteine and CVD⁽³⁵–Reference Wierzbicki³⁷⁾. However, several recent large intervention trials have failed to find any benefit of homocysteine-lowering therapies against death from cardiovascular causes in patients with pre-existing vascular disease or diabetics⁽³⁸⁾. Consequently, there is considerable debate whether homocysteine is causal for vascular dysfunction⁽Reference Splaver, Lamas and Hennekens³⁶⁾. In two recent prospective studies of men without prior CVD, high-serum folate was independently and significantly associated with a reduced risk of vascular events after adjusting for relevant confounders⁽Reference Voutilainen, Lakka and Porkkala-Sarataho^39, Reference Voutilainen, Virtanen and Rissanen⁴⁰⁾. Blood homocysteine levels did not correlate with risk⁽Reference Voutilainen, Virtanen and Rissanen⁴⁰⁾. Several alternative theories of how low folate may independently increase CVD have been proposed. Folate can directly influence vascular endothelial cell function. Nitric oxide production and vasodilation is impaired in CVD. Folates are critical in maintaining nitric oxide production and preventing oxygen radical formation by physically stabilizing tetrahydrobiopterin in the conversion of l-arginine to l-citrulline and preventing endothelial nitric oxide synthase uncoupling. Folic acid supplementation improves endothelial cell function in hyperhomocysteinemic patients with coronary artery disease, in diabetics, in subjects with hypercholesterolemia and in healthy subjects fed a high methionine load⁽Reference Doshi, McDowell and Moat^41, Reference Usui, Matsuoka and Miyazaki⁴²⁾. However, in addition to deregulating vascular endothelial cells, low folate may perturb vascular SMC gene expression, growth and function by disrupting DNA methylation⁽Reference Turunen, Aavik and Yla-Herttula²⁾.

Folate deficiency, cancer and CVD: a common mechanism

Within the methionine cycle, 5-methyltetrahydrofolate remethylates homocysteine to methionine, which is subsequently metabolized to SAM. As described earlier, SAM serves as 1-C donor in the methylation of DNA. Under conditions of low-dietary folate, SAM concentrations are depleted causing hypomethylation of newly synthesised DNA and increased proto-oncogene expression (see aberrant DNA methylation and abnormal gene expression above). While abnormal DNA methylation and gene expression is a consistent event in tumorigenesis, aberrant cytosine methylation and proto-oncogene activation may also be a common mechanism linking malignant transformation in cancer and vascular dysfunction in heart disease (Fig. 2). The atherosclerotic plaque and the tumour are both monoclonal in origin with unregulated cell proliferation providing a growth advantage for selected clones. During atherogenesis, SMC (which make up approximately 90% of the aorta cell population) transform from quiescent, contractile cells to synthetic cells that migrate to the intima and rapidly multiply. Overexpression of growth factors and pro-proliferative oncogenes such as c-myc and p53 (which are commonly dysregulated in cancer) also drive SMC invasiveness and plaque formation. Moreover, dysregulated expression of genes involved in lipid accumulation, connective tissue formation, calcification and inflammation further accelerate plaque progression. Hence, changes in the phenotype and invasiveness of SMC in the vascular plaque are similar to changes in monoclonality and cell proliferation described for tumour formation⁽Reference Turunen, Aavik and Yla-Herttula²⁾.

Low-folate status is associated both with alterations in cytosine methylation in experimental and human studies (see later) and an increased risk of malignant transformation. Whether folate status can similarly influence DNA methylation and risk of vascular disease remains to be established.

The remainder of this review will describe evidence that abnormal DNA methylation may be causal in the development of atherosclerosis and that folate status influences genome-wide DNA methylation and the incidence of both colon cancer and vascular disease. How gene-specific DNA hypermethylation influences disease risk was covered in a paper also presented at this symposium (E Lund, unpublished results).

Folate, abnormal genome-wide DNA methylation and colon cancer

While extreme methyl donor depletion alters genome-wide cytosine methylation (primarily in the liver) and induces hepatocarcinogenesis in animal models⁽Reference Wainfan and Poirier^43, Reference Pogribny, James and Jernigan⁴⁴⁾, the effect of folate deficiency alone on global DNA methylation in vitro and in vivo is strongly influenced by the treatment regime, rodent species, tissue and genes examined (reviewed in⁽Reference Kim³³⁾). Folate depletion decreases global DNA methylation in some human and animal cell lines in vitro, but not in others. Mouse NIH/3T3 fibroblast and CHO-K1 DNA became hypomethylated after 12 d in folate-depleted medium, but not in human colon adenocarcinoma HCT116 and Caco-2 cells⁽Reference Stempak, Sohn and Chiang⁴⁵⁾. Conversely, global and p53 TS gene DNA was hypomethylated in human colon adenoma cells grown in folate-depleted medium but was restored by folic acid repletion⁽Reference Wasson, McGlynn and McNulty⁴⁶⁾. We have shown that global cytosine methylation is decreased in SV40-immortalised human colonocytes grown in folate-free medium for 14 d⁽Reference Duthie, Narayanan and Blum⁴⁷⁾ but not in NCM460 non-malignantly transformed human colon cells under the same conditions⁽Reference Duthie, Mavrommatis and Rucklidge⁴⁸⁾. Severe and prolonged folate depletion in rodents induces global DNA hypomethylation in the liver and certain regions of the colon⁽Reference Balaghi and Wagner⁴⁹–Reference James, Pogribny and Pogribna⁵¹). Moderate folate deficiency does not consistently induce global DNA hypomethylation in blood, liver and colon, despite a significant depletion in blood and tissue folate and liver SAM⁽Reference Sohn, Stempack and Reid⁵²–Reference Duthie⁵⁴⁾.

Lymphocyte DNA is hypomethylated in women made experimentally folate deficient over several weeks⁽Reference Jacob, Gretz and Taylor^55, Reference Rampersaud, Kauwell and Hutson⁵⁶⁾ and low-dietary folate intake (<200 μg/d) correlates with hypomethylation of long-interspersed nucleotide element repeats (LINE-1; used as an indicator of global DNA methylation) in human colon tumours⁽Reference Schernhammer, Giovannucci and Kawasaki⁵⁷⁾. Conversely, LINE-1 methylation was not associated either with intake or blood folate in colon biopsy tissue collected from the Aspirin/Folate Polyp Prevention study⁽Reference Figueiredo, Grau and Wallace⁵⁸⁾. Little association between folate and genome-wide DNA methylation is apparent in healthy individuals with sufficient blood folate⁽Reference Kim^33, Reference Fenech, Aitken and Rinaldi^59, Reference Ingrosso, Cimmino and Perna⁶⁰⁾. The effect of supplementation with synthetic folic acid on DNA methylation is highly inconsistent and dependent on initial folate status, the level and duration of intervention, the genes and tissue reported and the health status of the subject⁽Reference Duthie⁵⁴⁾. DNA hypomethylation is reversed in lymphocytes from folate-deficient volunteers repleted with synthetic folic acid⁽Reference Jacob, Gretz and Taylor^55, Reference Rampersaud, Kauwell and Hutson⁵⁶⁾ and in leucocytes from colorectal adenoma patients given folic acid (400 μg/d for 10 weeks⁽Reference Pufulete, Al-Ghnaniem and Khushal⁶¹⁾). However, DNA methylation remained unchanged in the colonic mucosa of these patients⁽Reference Pufulete, Al-Ghnaniem and Khushal⁶¹⁾. Pharmacological (rather than nutritional) concentrations of folic acid (5–10 mg/d for 3–6 months) increased DNA methylation in the colorectum⁽Reference Cravo, Fidalgo and Pereira^62, Reference Cravo, Pinto and Chaves⁶³⁾, but intervention with a more moderate dose (1 mg folic acid/d for 3 years) did not⁽Reference Figueiredo, Grau and Wallace⁵⁸⁾. Supplementing healthy volunteers (with initial circulating folate concentrations within the normal range) has little impact on blood cell DNA methylation despite significantly elevated whole blood, plasma and lymphocyte folate⁽Reference Fenech, Aitken and Rinaldi^59, Reference Basten, Duthie and Pirie⁶⁴⁾.

Folate, DNA methylation and vascular disease

While there is good evidence that abnormal DNA methylation is causal for human carcinogenesis, decreased genomic methylation is also observed in cultured vascular cells and diseased vascular tissue. 5-methylcytosine levels drop rapidly (about 65%) in healthy rabbit aorta SMC cultured from aortic explants. Moreover, these cells change from a contractile to an invasive phenotype⁽Reference Hiltunen, Turunen and Hakkinen⁶⁵⁾. Hypomethylated DNA is also detectable in atherosclerotic plaques. Global genomic DNA is under methylated in advanced atherosclerotic lesions isolated from ApoE knockout mice and denuded New Zealand White rabbit aorta⁽Reference Hiltunen, Turunen and Hakkinen⁶⁵⁾. Most importantly, changes in DNA methylation patterns in mononuclear cells and in the aorta of ApoE mice occur early (4 weeks of age) and prior to appearance of vascular lesions, supporting a pathogenic role for abnormal DNA methylation in atherosclerosis⁽Reference Lund, Andersson and Lauria⁶⁶⁾. Cytosine methylation is reduced (9%) in advanced human atherosclerotic lesions compared to normal arteries or arteries with fatty streaks and at a level similar to that detected in human tumours⁽Reference Hiltunen, Turunen and Hakkinen⁶⁵⁾. Both global genomic DNA and gene-specific DNA is hypomethylated in blood cells from uraemia patients (with hyperhomocysteinemia) compared to healthy subjects⁽Reference Ingrosso, Cimmino and Perna⁶⁰⁾. Gene-specific hypomethylation has also been measured in human atherosclerotic advanced lesions and is associated with expression of the SMC proto-oncogene platelet-derived growth factor⁽Reference Hiltunen, Turunen and Hakkinen⁶⁵⁾ which may increase SMC proliferation in the atherosclerotic lesion. In addition to increasing SMC proliferation and invasiveness, DNA hypomethylation may also accelerate atherogenicity by up-regulating genes involved in lipid deposition and inflammation and increasing plaque instability. Several CpG islands are hypomethylated in the 15-lipoxygenase gene promoter and mRNA expression is correspondingly increased in human plaques⁽Reference Hiltunen, Turunen and Hakkinen⁶⁵⁾. As in cancer, gene-specific DNA hypermethylation is also present in atherosclerosis. Methylation of the estrogen receptor-α gene (which restricts cell proliferation) is increased (about two-fold) in arteries from patients with severe atherosclerosis compared to healthy subjects⁽Reference Post, Goldschmidt-Clermont and Wilhide⁶⁷⁾. In addition, estrogen receptor-α hypermethylation is associated with a switch from a contractile to synthetic phenotype in human SMC explants⁽Reference Ying, Hassanain and Roos⁶⁸⁾. Aberrant expression of other genes implicated in atherosclerotic plaque progression include interferon-γ, MMP-2, MMP-7, tissue inhibitor of metalloproteinases (TIMP)-3, p53 and EC-SOD (extracellular superoxide dismutase)⁽Reference Hiltunen and Yla-Herttuala⁶⁹⁾.

Cancer and atherosclerosis are diseases of old age. Moreover, normal ageing is associated with altered DNA methylation patterns, making it difficult to clearly establish the effects of diet alone on DNA methylation and disease risk. Similarly, determining whether low folate induces plaque development through abnormal DNA methylation independently of homocysteine is complicated by the intimate metabolic relationship between these two metabolites. While there is evidence that homocysteine is associated with altered DNA methylation and vascular disease, currently there is only very limited evidence that folate alone alters cytosine methylation in atherosclerosis.

In heterozygous and homozygous variant methylenetetrahydrofolate reductase mice, folate concentrations are low, homocysteine is elevated and brain and ovary tissue DNA is hypomethylated. Lipid deposition is increased in the aorta of aged methylenetetrahydrofolate reductase variant mice⁽Reference Chen, Karaplis and Ackerman⁷⁰⁾. Similarly, in mice heterozygous for the cystathionine-β-synthase gene, DNA methylation patterns in the imprinted H19 gene are changed in response to hyperhomocysteinemia⁽Reference Devlin, Bottiglieri and Domann⁷¹⁾. However, here, alterations in methylation profiles are highly tissue specific and do not correlate consistently with H19 gene expression⁽Reference Devlin, Bottiglieri and Domann⁷¹⁾. Plasma homocysteine and DNA methylation in peripheral mononuclear cells are negatively associated in healthy human subjects⁽Reference Yi, Melnyk and Pogribna⁷²⁾, while supraphysiological folate supplementation (15 mg/d for 8 weeks) reverses genome-wide DNA hypomethylation and abnormal gene expression (H19, IGF2) in hyperhomocysteinemic patients⁽Reference Ingrosso, Cimmino and Perna⁶⁰⁾.

In an attempt to establish the independent effects of folate and hyperhomocysteinemia on vascular function, Brown et al. ⁽Reference Brown, Huang and Lu⁷³⁾ developed an in vitro model of prolonged folate deficiency in cultured human endothelial (EA.hy926) cells⁽Reference Brown, Huang and Lu⁷³⁾. Several intracellular folate vitamers were significantly depleted (up to 99%) but homocysteine, SAM and SAH were unchanged relative to cells grown in adequate folic acid. MCP-1, a cytokine up-regulated early in atherogenesis, was increased providing limited evidence that folate deficiency (independently of homocysteine) is associated with an atherogenic phenotype. However, global DNA methylation was unaltered and gene-specific methylation was not measured⁽Reference Brown, Huang and Lu⁷³⁾.

The effect of hyperhomocysteinemia and/or vitamin B deficiency on aortic plaque formation has been investigated extensively in rodent models. Generally, hyperhomocysteinemia is associated with a 2-fold increase in aortic plaque area⁽Reference Hoffman, Lalla and Lu⁷⁴–Reference Zhou, Moller and Ritskes-Hoitinga⁷⁶⁾ but vitamin B deficiency is not⁽Reference Hoffman, Lalla and Lu^74, Reference Troen, Lutgens and Smith⁷⁷⁾. Conversely, vitamin B supplementation (up to three times control levels in the diet) is reported to decrease endogenous and homocysteine-induced plaque formation in ApoE null mice⁽Reference Hoffman, Lalla and Lu^74, Reference Zhou, Moller and Ritskes-Hoitinga⁷⁶⁾. However, the data are inconsistent and profoundly dependent on the dietary regimen⁽Reference Hoffman, Lalla and Lu^74, Reference Zhou, Moller and Ritskes-Hoitinga⁷⁶⁾. DNA methylation was not measured in any of these studies. We have recently developed an ApoE mouse model of folate and combined folate, vitamins B₆ and B₁₂ deficiency that demonstrates either mild (2-fold increase in homocysteine) or moderate hyperhomocysteinemia (7-fold increase in homocysteine). Folate (but not combined vitamin B deficiency) significantly increased atherosclerotic plaque formation (approximately 17%) in the aorta of mice fed a high fat Westernised diet for 16 weeks. Moreover plaque area was not associated with plasma homocysteine. However, global DNA methylation in the mouse heart, aorta and liver was similar across all groups indicating that disease progression is probably not related to altered cytosine methylation in response either to vitamin B or hyperhomocysteinemia (SJ Duthie, unpublished results).