Early life origins of metabolic diseases

Folate and C₁ metabolism is now seen as an integrator of nutrient status through which changes in nutrient inputs have multiple health effects via multiple changes in cell processes⁽ Reference Locasale ^{10 )} (Fig. 1). For example, because its role as a source of methyl groups for the re-methylation of homocysteine to methionine (Fig. 2), low folate supply leads to elevated concentrations of homocysteine⁽ Reference Lamprecht and Lipkin ^{11 )} which are associated with increased risk of NTD, CVD, cancers, dementias and osteoporosis⁽ Reference Clarke, Ward and Strain ^{12 )}. In India and other Asian countries, the prevalence of type 2 diabetes and other metabolic diseases is rising rapidly, apparently as a consequence of the double nutritional insults of poor maternal nutrition followed by overnutrition associated with urbanisation and the adoption of higher fat diets⁽ Reference Yajnik ^{13 )}. More specifically, the observation that raised maternal homocysteine concentration in pregnancy is associated with low birth weight has been confirmed by Mendelian randomisation analysis supports a causal role for dysregulated C₁ metabolism in poor fetal growth⁽ Reference Yajnik, Chandak and Jogle ^{14 )}. In addition, abnormal folate and homocysteine concentrations in mothers during pregnancy associate with both small birth size and increased likelihood of childhood insulin resistance in the offspring, emphasising the potential for nutritionally driven disturbances in C₁ metabolism to enhance fetal programming of diabetes and other metabolic diseases⁽ Reference Krishnaveni, Veena and Karat ^{15 )}. These observations fit with the ‘predictive adaptive response’ hypothesis which proposes that mismatch between the environment anticipated by the fetus, based on early (in utero) environmental exposures, and the environment encountered postnatally may predispose to the early development of metabolic, and other, diseases⁽ Reference Gluckman, Hanson and Spencer ^{16 )}. Therefore the adverse consequences of inadequate nutrient (folate) supply during early developmental may be exacerbated by overnutrition postnatally. Whilst the mechanisms responsible for the lifelong consequences of adverse nutritional exposures in early life are poorly understood, it seems probable that they include epigenetic processes⁽ Reference Vaiserman ^{17 )}.

Fig. 1.

(Colour online) Overview of C₁ metabolism indicating its role as an integrator of nutrient status (from Reference Locasale ¹⁰ ). ROS, reactive oxygen species; NADPH, nicotinamide adenine dinucleotide phosphate-oxidase.

Fig. 2.

(Colour online) Role of folate in C₁ metabolism supplying methyl groups for multiple purposes (from Reference Lamprecht and Lipkin ¹¹ ). SAM, S-adenosyl methionine; THF, tetrahydrofolate; DNMT, DNA methyl transferase; SAH, S-adenosyl homocysteine; CBS, cystathionine-beta-synthase; MS, methionine synthase; MTHFR, methylene tetrahydrofolate reductase; TS, thymidylate synthase.

Epigenetics as a mechanism linking early life nutrition with later health

Epigenetics describes an integrated system of marks (DNA methylation and post-translational modification of histones) and molecules (including small non-coding RNA) which is responsible for regulating the transcriptional state of individual cells. Importantly, epigenetic mechanisms are responsive to the cell's environment and so are modulated by dietary and other exposures⁽ Reference Mathers, Strathdee and Relton ^{18 )}. The importance of the maternal intake of folate and other methyl donors during pregnancy on DNA methylation patterns and offspring phenotype was established more than a decade ago using the agouti (A(vy)) mice which have a transposable element in the agouti gene⁽ Reference Waterland and Jirtle ^{19 )}. Maternal supplementation with folic acid, vitamin B₁₂, choline and betaine increased CpG methylation at the A(vy) locus in the offspring and increased the proportion of leaner offspring with darker coats⁽ Reference Waterland and Jirtle ^{19 )}. The agouti locus is an example of a metastable epiallele (ME) that is variably expressed in genetically identical individuals due to epigenetic modifications that were established during early development⁽ Reference Rakyan, Blewitt and Druker ^{20 )}. In human subjects, Waterland et al. were the first to establish that the maternal environment around the time of conception influences the methylation status of ME in the offspring⁽ Reference Waterland, Kellermeyer and Laritsky ^{21 )}. This work was undertaken in rural Gambia where seasonal differences in work patterns, food availability and other environmental factors provide a natural experiment for testing the effects of maternal exposures on pregnancy outcomes. Although in such natural experiments it is impossible to determine which of the many factors is responsible for a given outcome, it seems likely that nutrition is important. For example, seasonal differences in maternal consumption of methyl-donor nutrients influenced the maternal plasma concentrations of multiple substrates for C₁ metabolism during pregnancy and these changes were reflected in altered methylation of ME in DNA extracted from lymphocytes and from hair follicles in their infant offspring⁽ Reference Dominguez-Salas, Moore and Baker ^{22 )}. More recent study of these Gambian children revealed that methylation of the tumour suppressor gene VTRNA2-1 differed according to season of conception⁽ Reference Silver, Kessler and Hennig ^{23 )}. In addition, once altered in utero, the methylation status of VTRNA2-1 was stable over at least 10 years⁽ Reference Silver, Kessler and Hennig ^{23 )}. VTRNA2-1 is an ME which appears to influence both cancer risk and immune function and the authors argued that this season of birth-related epigenetic change is a plausible candidate pathway to explain their earlier observation that season of birth predicts adult mortality from infection-related causes in rural Gambians⁽ Reference Moore, Cole and Poskitt ^{24 )}. Whilst these exciting findings should stimulate further investigations of the impact of peri-conceptual nutrition on health outcomes and the possible mediating effects of epigenetic mechanisms, the results should be interpreted with caution since they demonstrate associations but not causality.

Towards a mouse model for testing the effects of maternal folate inadequacy on the metabolic health of the offspring

Direct testing in human subjects of the effects of maternal folate inadequacy on the metabolic health of the offspring is fraught with obvious ethical and practical difficulties. In addition, attempts to investigate the epigenetic mechanisms through which such nutritional insults during early development produce their long-term effects on health⁽ Reference McKay, Waltham and Williams ^{25 )} in human subjects would require access to tissue such as liver and placenta. This is because the patterns of epigenetic marks, notably DNA methylation, are cell and tissue specific so that methylation marks assessed in a surrogate tissue e.g. blood or buccal cells may not reflect those in the cell or tissue of interest⁽ Reference McKay, Xie and Harris ^{26 )}. For example, we quantified DNA methylation by pyrosequencing^® for a panel of genes including Esr1, Igf2 and Slc39a4 using DNA extracted from blood, liver and kidney from female mice and observed tissue-specific differences in methylation at all loci⁽ Reference McKay, Xie and Harris ^{26 )}. Such inter-tissue differences in DNA methylation patterns in human subjects were confirmed for IGF2, GNASAS and IL10 by Waterland et al.  who examined methylation patterns in post-mortem samples of brain, liver and kidney obtained from Vietnamese motor vehicle accident victims⁽ Reference Waterland, Kellermeyer and Laritsky ^{21 )}. In addition, we observed that folate depletion during pregnancy altered Igf2 methylation in a tissue-specific manner (P < 0·05)⁽ Reference McKay, Xie and Harris ^{26 )}. In this context, ME are a special case since their methylation status is very similar in all tissues investigated and methylation changes induced in early development appear to be stable indefinitely⁽ Reference Silver, Kessler and Hennig ^{23 )}. As a consequence, suitable animal models are necessary to test for causality, to undertake investigations of possible epigenetic mechanisms and to augment the epidemiological evidence available from human studies.

In recently completed work, we set out to test three hypotheses: Maternal folate depletion during pregnancy and lactation may contribute to the development of obesity and the Metabolic Syndrome; These effects may be exacerbated by provision of a high-fat (HF) diet post-weaning (double nutritional insult); The altered phenotype may be due to altered gene expression through epigenetic mechanisms.

On the basis that better models facilitate better hypothesis testing, we developed a mouse model of maternal folate depletion which was uncomplicated by other factors. In addition, because epigenetic processes may be particularly plastic in early embryonic/fetal life, we considered it important that the mouse dams were folate depleted at mating. To minimise potential confounding by other factors, the degree of folate depletion should be sufficient to impose a nutritional stress but not be so severe as to limit reproduction. Finally, by feeding an HF diet from weaning, we aimed to simulate a common secondary nutrition stress. Such a model may yield data which may be more readily interpretable mechanistically and which are likely to be of greater relevance to human nutrition, i.e. more potential for translation.

Our first question was how long would it take to induce folate-depletion in young female mice? We addressed this question by feeding female mice a folate-free diet for up to 7 weeks⁽ Reference McKay, Waltham and Williams ^{25 )}. Blood was collected from pairs of culled mice weekly and erythrocyte folate concentration measured. There were two distinct phases in erythrocyte folate kinetics; an initial phase lasting approximately 2 weeks during which erythrocyte folate concentration remained largely unchanged followed by exponential ‘decay’ in erythrocyte folate concentration towards a new equilibrium after 4–5 weeks⁽ Reference McKay, Waltham and Williams ^{25 )}. For comparison, Leamon et al. reported that erythrocyte folate reached a new (lower) plateau in BALB/c mice approximately 6 week after transfer to a low folate diet⁽ Reference Leamon, Reddy and Dorton ^{27 )}. Further studies showed that a milder nutritional insult (feeding 0·4 mg folic acid/kg diet) reduced maternal and offspring erythrocyte folate concentrations (measured at weaning) by about 50 %⁽ Reference McKay, Waltham and Williams ^{25 )}.

For the remaining studies described in this report, we adopted a 2×2 factorial design in which female C57/Bl6 mice were fed a folate-normal or folate-depleted diet (2 and 0·4 mg folic acid/kg diet, respectively) for 4–5 weeks before mating and throughout pregnancy and lactation. At weaning, offspring were randomised to either a control diet or an HF diet (50 and 200 g fat/kg diet, respectively) from weaning until age 6 months (Fig. 3)⁽ Reference Langie, Achterfeld and Gorniak ^{28 ,} Reference McKay, Xie and Manus ^{29 )}. For dams with successful pregnancies, duration of dietary exposure did not differ between those fed the normal and low-folate diets (P = 0·42). However, dams fed the low-folate diet were more likely to experience reproductive failure due to miscarriage or postpartum litter death⁽ Reference McKay, Xie and Manus ^{29 )}. At weaning, maternal serum folate concentration was reduced by approximately two-thirds (P < 0·001) in the dams fed the low-folate diet confirming the magnitude of the maternal dietary insult⁽ Reference McKay, Xie and Manus ^{29 )}. Whilst the low-folate diet had no effect on litter size, body weight at weaning was 6 % lower for offspring of mothers fed the low-folate diet⁽ Reference McKay, Xie and Manus ^{29 )}.

Fig. 3.

Study design investigating effects of maternal folate depletion and high-fat feeding from weaning (from Reference Langie, Achterfeld and Gorniak ²⁸ ).

As expected, offspring randomised to the HF diet were heavier and contained more body fat at ages 3 and 6 months than those randomised to the lower-fat (control) diet, but there was no effect of maternal folate depletion on offspring adult weight or adiposity⁽ Reference McKay, Xie and Manus ^{29 )}. In addition, there was no effect of maternal folate supply on the gross anatomy of the 6-month-old offspring as gauged by organ weights and gut lengths⁽ Reference McKay, Xie and Manus ^{29 )}. However, adult offspring from dams fed the folate-depleted diet had significantly raised plasma TAG concentrations when given the HF diet from weaning, whereas the HF diet had no effect on TAG concentrations in the offspring of mothers with adequate folate supply before mating and during pregnancy and lactation (P _interaction = 0·005)⁽ Reference McKay, Xie and Manus ^{29 )}. This provided evidence that the early life nutritional insult (maternal folate depletion during pregnancy and lactation) had long-term adverse metabolic consequences for the offspring, which were revealed only when the offspring were exposed to a second nutritional insult, i.e. HF feeding from weaning. This adverse metabolic phenotype in the adult offspring occurred in the absence of a detectable effect on adiposity which suggests a different aetiological origin from the frequently reported phenotype characterised by enhanced adiposity and other symptoms of the Metabolic Syndrome in the adult offspring of dams exposed to malnutrition during pregnancy⁽ Reference Burdge and Lillycrop ^{30 )}.

Investigation of possible mechanism(s) for long-term metabolic effect on offspring of folate depletion during pregnancy and lactation

We hypothesised that maternal folate depletion alters folate supply to the embryo and the developing fetus with potential widespread effects because of its impact on C₁ metabolism. In particular, inadequate folate limits the availability of methyl groups for S-adenosylmethionine synthesis and this generates competition between cellular processes which use S-adenosylmethionine, notably the methylation of macromolecules including lipids, proteins and DNA. Further, we hypothesised that reduced S-adenosylmethionine availability would alter the pattern of DNA methylation which, because of its role in regulating transcription, would change the repertoire of expressed genes⁽ Reference Mathers ^{31 )}. Finally, the altered gene expression would reduce the capacity of adult offspring to cope metabolically with a second nutritional stress, e.g. HF feeding, leading to elevated plasma TAG concentrations. If these putative effects were induced in early life, then we anticipated that they might be detectable in the fetus late in pregnancy and that this might be an informative life-stage at which to test our hypothesis.

Using time-mated dams, we collected fetuses and their placentae at pregnancy day 17·5. Both tissues were used for investigation of genome-wide gene expression and, in addition, genome-wide promoter methylation was investigated in the fetal livers. The outcomes of those mechanistic studies will be published elsewhere.

Public health implications

Adequate folate intake pre-conception and during pregnancy and lactation is essential for good maternal and child health. Recent data from the National Diet and Nutrition Survey rolling programme shows widespread folate inadequacy among women of child-bearing age in the UK (Table 1)^{( 32 )}. Biochemical folate deficiency is more prevalent in younger, than in older, women and in Scotland and Northern Ireland than in the UK as a whole. In addition, overweight and obesity are common among women of child-bearing age^{( 33 )}. The Health Survey for England reported that 14 % of females aged 16–24 years were obese and that this rose to almost 25 % for women aged 35–44 years^{( 33 )}. If the observations from our mouse study⁽ Reference McKay, Xie and Manus ^{29 )} apply in human subjects, then the combination of folate deficiency and excess adiposity among women of child-bearing age may disadvantage the health of their offspring by increasing the risk of metabolic diseases exemplified by raised TAG concentrations. This risk may be exacerbated if the folate needs of obese women are greater than those of normal weight women. A study of short-term folate pharmacokinetics in women of child-bearing age showed that the area-under-the-curve for the absorption phase (0–3 h) and the peak serum folate concentration were both significantly lower in obese women following consumption of 400 µg folic acid, the recommended supplemental dose to lower NTD risk (Fig. 4)⁽ Reference da Silva, Hausman and Kauwell ^{34 )}. In a previous study in which the folic acid dose administered was calculated per total body weight, the area-under-the-curve was higher in obese women and the authors suggested that it would be preferable to define the folate dose in relation to lean body weight⁽ Reference Stern, Matok and Kapur ^{35 )}. In summary, folate recommendations for obese women of reproductive age may need to be reconsidered to ensure adequacy.

Fig. 4.

Serum folate responses following ingestion of 400 µg folic acid in obese and normal weight women (from Reference da Silva, Hausman and Kauwell ³⁴ ).

Table 1.

Percentages of women of child-bearing age with evidence of biochemical folate deficiency based on serum and erythrocyte folate concentrations (National Diet and Nutrition Survey rolling programme 2015)

* Percentage of women with serum folate concentration <10 nm ^{( 36 )}.

† Percentage of women with erythrocyte folate concentration <340 nm ^{( 36 )}.

In conclusion, our mouse studies showed that uncomplicated folate inadequacy peri-conceptually and during pregnancy and lactation predisposed the offspring to metabolic derangements when fed an HF diet. The widespread prevalence of folate deficiency and of overweight and obesity among women of child-bearing age highlights the possibility that this double nutritional insult may exacerbate the risk of metabolic disease in their offspring and points to the need for appropriate interventions.