We explored the potential effect of parental obesity on IGF2/H19 DMR methylation in newborns. We found a significant decrease in methylation among newborns of obese fathers at the IGF2 DMR in DNA extracted from cord blood leucocytes. This finding remained significant after controlling for potential confounders (β-coefficient = -5.28, P = 0.003). Hypomethylation at the IGF2 DMR has been associated with an increased risk of developing cancers, such as Wilms' tumor , colorectal cancer  and ovarian cancer . We found no significant changes in methylation levels associated with paternal obesity at the H19 DMR region.
Obesity is a metabolic condition that has paradoxically been associated with poverty, low quality of life, malnutrition and an imbalanced intake of nutrients [21, 22]. Few epidemiological studies indicate potential correlations between obesity or food supplies in the paternal line and offspring's birth weight , body-fat in prepubertal girls , or mortality from chronic diseases [31–33]. Epidemiological data regarding associations between maternal obesity and the offspring's birth weight vary by study (reviewed by McDonald et al. ). We examined possible associations between paternal or maternal obesity and birth weight but detected no associations. Epidemiological studies on maternal obesity-related exposures generally show a positive association between abnormally high BMI and congenital anomalies (reviewed by Stothard et al. ). These harmful effects are mostly attributed to in utero exposures to malnutrition or overnutrition; while very often, data indicate the importance of exposures at the very early stages of development, even before conception. Research in animal models suggests that potential diet-dependent transgenerational effects may be explained by changes in the establishment of epigenetic gene regulatory marks [12, 36–38]. Analyses of adults born to mothers exposed to poor nutrition during the Dutch famine indicated a 5% decrease in methylation at the IGF2 DMR compared to the same sex non-exposed siblings. Interestingly, the magnitude of this effect is similar to the effect we observe in offspring from obese fathers at the same locus. In the Dutch famine cohort, exposure during the periconceptional period was more important than during the second or the third trimester of gestation . Furthermore, not only do exposures during early gestation cause harmful health outcomes, but famine prior to conception has also been associated with poor health . Similar results were found among Gambian children; they exhibit altered DNA methylation at several metastable epialleles according to the seasonal nutritional circumstances in which they were conceived . Both cohorts described above did not directly examine the relationship with the fathers' dietary patterns; although they were most likely exposed to the same famine or nutritional circumstances as the mothers. Analysis of the Framingham Heart Study indicates that early-onset paternal obesity, and not maternal obesity, increases the odds of aberrant serum levels of the metabolic biomarker ALT (alanine transaminase) in the offspring . Studies on animal models have shown similar associations. In rats, a high fat paternal diet results in offspring with early onset of impaired insulin secretion, altered expression of multiple genes related to normal pancreatic beta-cell function, and altered methylation at a putative regulatory region of the Interleukin 13 receptor, alpha 2 gene . Male mice whose mothers were exposed to a high-fat diet were not only obese, insulin insensitive and diabetic, they were also capable of passing part of this phenotype to the next generation, suggesting an underlying epigenetic mechanism transmitted through germ cells . To our knowledge, our study is the first epidemiological study that suggests a similar underlying epigenetic mechanism conferred by harmful paternal dietary patterns or obesity.
Obesity is associated with elevated IGF2 circulating levels  and increased estrogen levels . Although we did not include IGF2 protein levels in parents or offspring in our current analyses, we earlier showed that hypomethylation at the IGF2 DMR is associated with higher circulating IGF2 levels in the offspring [25, 46]. This association was strongest in offspring from obese mothers, independent of race. In brief, a decrease of 5% at the IGF2 DMR methylation corresponded to an increase of at least 10% in serum concentration of IGF2 . In addition, other studies have shown that small aberrant methylation changes at the IGF2 or H19 DMR is linked to increased expression of IGF2 [6, 7, 24, 25, 47], as well as an increased susceptibility to chronic diseases [48–51]. Similar small effects on DNA methylation have also been associated with the use of assisted reproductive technologies , the use of psychotropic drugs during pregnancy  and smoking . These subtle epigenetic changes have been described as adaptive responses to the environment, while major epigenetic shifts during development would cause more detrimental consequences . Furthermore, exploring the CpG sites at the IGF2 and H19 DMRs may represent only a fraction of changes occurring elsewhere in the genome. Environmental factors, among which is diet, have been associated with changes in DNA methylation and may have profound effects on genomic imprinting; accumulation of these effects may result in disturbed metabolic homeostasis . Follow-up studies on the anthropometric and other developmental factors of the NEST children are underway to further examine the influence of small changes in DNA methylations at several DMRs on childhood obesity or other adverse consequences. Evidence in animal studies indicates that DNA methylation at the IGF2/H19 locus in sperm might be under tight control of estrogen , produced by adipocytes. This suggests a mechanism by which increased exposure to estrogen could lead to inadequate establishment of methylation at the IGF2 DMR in sperm. Alternatively, obesity-related factors may also disrupt functioning of other components of the epigenetic machinery leading to an inability to appropriately establish imprint marks during spermatogenesis, which is ongoing through adult male life . Offspring of obese fathers may, therefore, demonstrate incomplete methylation. In order to further explore these hypotheses, more research on the epigenetic effects of obesity on human germ cells is needed.
Although our bivariate analysis did not indicate an association between maternal obesity and DNA methylation at the IGF2 DMR in newborns, the regression analysis showed that controlling for paternal obesity resulted in an opposite effect, meaning that while paternal obesity was associated with a decrease in methylation, maternal obesity tended to be associated with an increase in methylation, but this was only significant at one CpG site. Using pre-pregnancy BMI instead of obesity in our multivariate analysis strengthened this association. At the H19 DMR, the bivariate analysis showed an increase of 4.1% when mothers were obese (P = 0.01). An increase in DNA methylation by maternal pre-gestational BMI has also been reported earlier in cord blood samples, more particularly at the PPARG promoter . Hypermethylation at the IGF2 or H19 DMR has been associated with loss of imprinting of IGF2, and several disorders. For instance, IGF2 imprinting defects have been implicated in Silver-Russell syndrome , Wilms' tumor [24, 58], hepatoblastoma  and ovarian cancer . Our regression analyses at the H19 DMR showed that adjusting for potential confounders, including paternal obesity or BMI, diminished this positive association (Table 2, Model 3). We attribute this to the fact that maternal and paternal obesity are closely related and the fact that methylation outcomes for both parental exposures are in the same direction. A larger study is necessary to further explore the potential impact of parental obesity on DNA methylation at the H19 DMR. Furthermore, the exposure from oocyte stage till birth to maternal obesity or related lifestyles is complex. Hormonal factors that may influence DNA methylation cannot be ruled out. It has been shown that the rat H19 DMR has an estrogen responsive element, suggesting that estrogen can form a complex with Dnmt1, a DNA methyl transferase, leading to DNA methylation at a normally unmethylated maternal allele .
A potential limitation of our study includes the use of cord blood as a marker for the newborn's epigenetic status. However, we used isolated leucocytes and IGF2 is a well studied imprinted gene whose germline DMRs should be similarly methylated across all cell types, given the establishment of the epigenetic profile prior to conception. IGF2 and H19 DMR methylation profiles were verified in DNA from different cell fractions from umbilical cord blood and we found no differences across the cell types . Another possible limitation is proof of paternity, and the reliability of the paternal anthropometric data, which were reported by pregnant mothers. However, the questions regarding the anthropometric data were detailed and verified for consistency. We do not expect that methylation outcomes are differential with respect to the potential misclassification of exposures. Nineteen percent of our population had missing data about the father. The methylation outcomes of these newborns were not included in our final analytical study group. However, we compared methylation outcomes by missing and non-missing characteristics of the fathers and found no differences. We also compared all measured maternal and newborn characteristics in both groups and most characteristics were similarly distributed. We found no significant differences when mothers were single or not, but found significant differences by education and race. Most missing data were among the lower educated and African American mothers. As far as we could test, education was not associated with obesity in either of the parents, and race was only associated with obesity of the mother. It is very likely we missed a number of obese fathers from African Americans. We cannot verify if paternal obesity is equally distributed in all subgroups of missing and non-missing paternal data, and, therefore, selection bias cannot be excluded. However, when reanalyzing our regression models with race included as an independent variable, our results remained the same; only the effect of maternal obesity attenuated in Model 2 (Table 2) at H19 DMR, the β-coefficient for maternal obesity became +2.18 (SE = 1.65, P = 0.19) (data not shown). Possibly, race is on the same causal pathway as obesity, regarding the effect on methylation outcomes. As mentioned in the results section, race by itself was not associated with methylation outcomes at neither of the two DMRs studied in this cohort. Nevertheless, given our earlier analyses on maternal exposures showed a race dependent effect at the same imprinted loci , we further stratified our data by race. The outcomes were in the same direction as the complete analytical cohort, but given the smaller numbers, the results represented unstable estimates. In general, the small sample size remains a limitation in our study which may partly explain why our results at the H19 DMR do not reach statistical significance. Nevertheless, the data concerning associations between IGF2 DMR methylation and parental BMI or obesity reached sufficient power, especially when studying offspring from fathers with high BMI.