In this prospective cohort study, we did not find a positive association between maternal dietary intake of methyl donors and first trimester, second trimester or cord blood LINE-1 leukocyte methylation, despite adequate statistical power as reflected in narrow confidence intervals. We did find an inverse association between maternal periconceptional betaine intake and cord blood LINE-1 global DNA methylation, but this association appeared to be confounded by intake of cadmium.
Our findings of no other associations between methyl donor intake and LINE-1 methylation are consistent with some studies in non-pregnant adults.14,15
For example, Friso et al. did not find an association between plasma folate and global genomic methylation in lymphocytes of 198 non-pregnant adults except among subjects with MTHFR minor variants, in whom low folate was associated with lower methylation.15
There are several potential explanations for the lack of association of methyl donor intake with DNA methylation in our study. First, due to folic acid fortification of the US food supply and generally high plane of nutrition among the mothers in Project Viva, there are few individuals in this population who were deficient in methyl donor nutrients, especially by the second trimester of pregnancy. For example, only 96 (9%) of women had second trimester folate intake <600 µg/day. It is possible that a threshold exists above which consuming more of these nutrients does not alter methylation levels. However, a relatively high proportion of women had low daily folate intake from weeks 0 to 4 of pregnancy [422 (39%) had folate <600 µg], so if there were truly an effect at levels currently defined as deficient, we would likely have observed it. Second, because much about methylation processes is still unknown, there may be unmeasured confounding by other factors. Authors of a recent combined analysis of five human investigations have identified covariates that affect LINE-1 methylation16
and may operate as confounders in epidemiology investigations. However, we have assessed the extent of confounding by those and many other carefully measured variables and have not identified many variables that substantially affect this association. Third, using an FFQ for dietary assessment likely introduced some measurement error into our analyses, especially since use of this FFQ in pregnancy was not validated for the nutrients we examined; however, the FFQ should accurately rank individuals' intake of these nutrients. We also used a detailed interview about maternal use of nutritional supplements periconceptionally. Fourth, LINE-1 methylation in leukocytes may not be a sensitive long-term measure of global genomic methylation, and the lack of association may be due to statistical noise of the methylation variable. However, we used the most technologically advanced and validated methods currently available for LINE-1 methylation analysis and have already demonstrated a relationship between higher LINE-1 and male sex and several other obesity-related risk factors in offspring in early life.17
This method for LINE-1 methylation analysis has been shown in previous investigations to be sensitive to environmental and host-related factors, including age,18
and tobacco smoke components,21
and environmental pollution.23–26
Also, we measured DNA methylation only in blood DNA and we cannot exclude that methylation changes due to dietary components might be present in other tissues in the same individuals. Finally, we do not have gene- or tissue-specific methylation data, and there may be differences in gene- or tissue-specific methylation despite no differences in global LINE-1 in leukocytes.
Our null findings are important because they suggest that at the levels of folic acid intake found in a healthy US population, global DNA methylation in blood is not related to intake. This is reassuring considering the concern about potential adverse effects, possibly through epigenetic changes, of folic acid fortification of the food supply in the US and elsewhere.27,28
One exception to our null findings was lower cord blood methylation in relation to higher periconceptional intakes of betaine and choline in males only. We consider these sex-specific observations to be hypothesis-generating. This was a post hoc analysis based on our initial observation that males had higher levels of cord blood methylation than females, rather than an a priori hypothesis. Nevertheless, given that this pattern is consistent with some, but not all, animal experiments,7,9,10
we would recommend that investigators search for sex-specific determinants, as well as outcomes, of DNA methylation.
These inverse associations among males, and the initial finding that periconceptional maternal betaine intake was inversely related to cord blood methylation in the overall cohort were counter to our hypothesis, considering that some previous studies have found positive relationships between maternal methyl donor intake and offspring methylation.9,11,12
The inverse association could possibly be due to changes in the expression or activity of DNA methyltransferase (DNMT), the enzyme that attaches methyl groups to DNA. This would be consistent with a study in rats in which a deficiency of choline, a precursor of betaine, during gestation caused increased global methylation in the brain and liver of offspring by upregulation of DNMT1 expression.29
However, the fact that we did not find other associations of methyl donor intake at two timepoints with DNA methylation at three timepoints suggests that the findings may be due to chance. If we had corrected for multiple testing, these associations would not have been statistically significant by conventional standards (p < 0.05). Also, the betaine association was substantially attenuated after adjustment for maternal dietary cadmium intake, which itself was directly associated with first trimester and inversely associated with cord blood LINE-1.
We found that dietary cadmium was associated with first trimester and cord blood LINE-1 methylation. In the human diet, cadmium, a toxic heavy metal if ingested at high doses, comes mostly from plant-based foods grown in soils treated with fertilizer.30
Betaine comes from similar sources, and intake was moderately correlated (rs
= 0.50) with cadmium intake in our study. Additionally, women with higher cadmium intakes in our study were older and consumed more methyl donors in general. Cadmium interferes with the activity of DNMT in animals and human cell cultures,31
so dietary intake of this mineral could affect methylation in human populations as well. One study found that long-term low exposure to cadmium increased DNMT activity and genomic methylation in human embryo lung fibroblast cells.32
A study in mouse testicular Leydig cells found that 24-h exposure to cadmium led to decreased expression of genes for DNMTs.33
Another study in rat liver cells found that exposure to cadmium initially led to decreased, then later to increased DNMT activity and methylation.34
Finally, investigators found that cadmium exposure in chick embryos caused downregulation of DNMT3A/3B gene expression.35
More research must be conducted about the validity of cadmium intake measured by FFQ, cadmium's effects on DNMT activity and methylation, and potential intergenerational health effects in humans in vivo.
Strengths of this study include its prospective design, detailed dietary information, measurement of DNA methylation in both maternal and cord blood, adjustment for multiple confounding factors, and large sample size compared with previous studies. The findings from this study may or may not be generalizable to other populations.
In conclusion, in this folate-replete population, we did not find a positive association between maternal dietary intake of methyl donors and first trimester, second trimester or cord blood LINE-1 leukocyte methylation. Cadmium intake was a predictor of maternal and cord blood methylation. Future research should focus not only on effects of methyl donors in other populations but also on maternal cadmium exposure as a potential determinant of DNA methylation and offspring health.