This study showed no evidence that IGF2/H19 ICR DNA methylation in peripheral blood at age 17 years is associated with birth anthropometry. On the other hand, a negative association with aggregated head circumference from 1 to 10 years and a positive association with skin fold thicknesses from multiple sites and subcutaneous adiposity at 17 years were observed.
Our study found repeated positive associations of greater IGF2/H19 ICR methylation with greater subcutaneous adiposity measured by multiple methods, including direct ultrasound and by calliper assessed skin folds at multiple body sites. The positive association of IGF2/H19 ICR methylation with subcutaneous adiposity in this study adds to the recent observations by Perkins et al. showing greater IGF/H19 ICR methylation in overweight or obese one-year-olds compared to normal weight counterparts
. The current study shows that an association between adiposity and methylation at this site persists to young adult life. It is also consistent with current understanding that the IGF axis increases the proportion of subcutaneous fat to visceral fat
[32, 33], particularly in younger age groups. In neonates, levels of IGF1 protein have been positively associated with subcutaneous fat and sum of skin folds, but not with visceral fat
. Interestingly, Russell Silver syndrome patients, who have decreased expression of IGF2, have a striking lack of subcutaneous fat as part of their generalized growth retardation
. Furthermore, genome wide association studies have identified the 11p15.5 region inclusive of the IGF2 gene to be associated with abdominal subcutaneous fat
In general, activation of the IGF axis decreases overall fatness. Many
[7, 9–12], but not all
 candidate studies have shown that IGF2 polymorphisms are associated with overall fatness. Specific to visceral adiposity, some
, but not all
 studies show an inverse correlation with IGF1 protein levels. Despite this, contrary to our initial hypotheses, no relationships were seen between IGF2/H19 ICR methylation and either BMI, waist circumference or visceral fat at age 17 years. BMI is relatively easy to measure, but does not always correlate with total fat mass and may be confounded by muscle mass
. As activation of the IGF axis also increases muscle mass
, the lack of discrimination of the BMI variable for fat and muscle body compartments may account for its poor correlation with IGF2 ICR methylation.
The lack of associations between IGF2/H19 ICR methylation and visceral fat or abdominal circumference (a marker for central adiposity) are also interesting to note. We speculate that the lack of association seen might be due to the relative youth of our cohort where the early response to obesogenic influences results in preferential fat deposition in the protective subcutaneous compartment
[41, 42], imparting a lesser metabolic risk than the visceral compartment
. At 17 years of age, DNA methylation at the IGF2/H19 ICR does not appear to be associated with detrimental fat distribution. However, no definitive conclusion can be drawn as to whether the altered DNA methylation is a consequence or cause of the altered subcutaneous adiposity in our study.
Based on the knowledge that suboptimal intrauterine environments leading to fetal programming are associated with epigenetic changes
[14, 15], we hypothesized that birth anthropometry would be associated with IGF2/H19 ICR methylation. Contrary to this, we did not detect an association of DNA methylation with any birth size parameters. This finding is not so unexpected given that fetal programming caused by altered intrauterine environment has been shown repeatedly to occur without disturbance in birth size
[15, 44, 45]. Sheep experiments isolating the effect of maternal under- and over-nutrition in the periconceptional period, show that weight at birth is not affected
, despite subsequent development of increased obesity and adverse metabolic parameters. In the Dutch Famine cohort, only modest differences in birth weight of 50 to 100 g were observed for exposed individuals overall
 and epigenetic differences were only present in those of normal birth weight after being exposed to periconceptional starvation
. Those with exposure to starvation later in gestation did demonstrate reduced birth weight, but in the absence of changes in DNA methylation
. On a molecular level, epigenetic marks may be particularly vulnerable during very early development
 a finding supported by experimental evidence of altered H19 methylation and expression in mouse 2 cell embryos cultured to the very early stage of blastocysts
. In addition, there is evidence that intrauterine growth retardation may not be associated with epigenetic changes at IGF2[49–51]. This suggests that birth weight may be a poor surrogate for a suboptimal environment during pregnancy when testing for fetal programming via DNA methylation mechanisms. It is likely that the intrauterine effects that cause DNA methylation changes in a modern, well-nourished population are subtle and do not cause great birth weight changes.
In our study, head circumference (HC) at birth was not correlated with IGF/H19 ICR DNA methylation at 17 years. However, between 1 and 10 years HC was negatively correlated with IGF/H19 ICR DNA methylation (in all CpG units except for CpG 23). The key role of the IGF axis and methylation of its related genes in head size maintenance is evident; for example, IGF1-deficient subjects treated with exogenous IGF1 show a striking increase in head circumference
. Adult brain size has been shown to relate positively to IGF2 DMR2 (exon 9) methylation and be negatively related to IGF/H19 ICR (at the CTCF3 region 2kb from our ICR assay) DNA methylation in DNA extracted from the cerebellum
[53, 54]. Although the region within the ICR assayed in the latter study did not overlap with our assay, the negative direction of association was the same. Further, dietary interventions aimed at modifying DNA methylation (periconceptional folate supplementation) have been associated with both increases in IGF2 DMR DNA methylation
 and head size
. In-utero head circumference is genetically
 and, possibly also, epigenetically controlled.
The reason that an association was detectable in infancy/childhood, but not immediately post-delivery may be due to compromise in the accuracy of HC measurements by molding that inevitably occurs with a non-breech delivery. In addition, maternal uterine constraint may play a role, whereby both gain in size and stunting only become apparent after a period of time outside the limiting influences of the womb
. Differences might also be due to variation in the mechanisms of IGF2/H19 imprinting between blood and brain
. Alternatively, birth size may truly have no linear relationship with IGF2/H19 methylation. This is plausible given that fetal programming can occur without change in birth size
. It is also possible that in our cohort, fetal programming may have influenced DNA methylation at the IGF2/H19 locus, specifically in utero, for which there is prior evidence in relation to certain early life environments
[28, 29, 59].
No conclusion can be drawn as to whether the altered DNA methylation is a consequence or cause of the altered head size in our study. The DNA methylation was measured at a later age than some of the longitudinal anthropometry measurements. Nevertheless, there is evidence that imprinted domains, such as the studied IGF2/H19 ICR (having been established either in germ cells in the previous generation or somatic tissues in utero), are relatively stable over time postnatally
, into late adulthood
[61, 62] and equivalent across tissues in humans
. As further indirect confirmation we have shown that Ollikainen et al.’s
 measurements of the same region at birth are comparable to measurements at age 14 years in this study.
This is, to our knowledge, the first study to investigate the association between DNA methylation and longitudinal growth. Although RNA was not available from study members, we suggest that, due to the previously demonstrated association between IGF2/H19 DMR methylation, expression and levels of IGF2 protein
, this association will extend to levels of IGF2 transcripts and protein. A strength of the study is the large number of subjects and the serial prospective anthropometry. However, a limitation of the study is that DNA methylation has only been measured at one time point at age 17. Although methylation levels in peripheral blood (as used in our study) appear very similar to those seen at birth in the same tissue
 and IGF2/H19 imprinting is likely to be stable over time in adults
[60, 62], a greater understanding is required regarding the dynamics and influences on postnatal methylation in childhood and adolescence.
Further, the importance of tissue in which the DNA methylation is measured needs to be considered, as we have found that effects of intrauterine exposures can be different and even opposite in different tissues (JMC, unpublished data). DNA has been derived from peripheral blood samples in this study for ethical and practical reasons. To our knowledge, no study has compared methylation in blood, adipose and brain tissues, although altered DNA methylation has been observed with obesity in peripheral blood
. Furthermore, the region we studied (H19 CTCF6 DMR) has shown minimal variation in methylation between tissues, while IGF2 DMR showed greater variation
Other limitations are the subset of the Raine cohort used for this study, which is biased to higher BMI and extremes of cortisol measurements. Although adjustment for serum cortisol and adrenocorticotropic hormone (ACTH) in the models did not alter the direction or significance of our findings (data now shown), the significance of our findings would be enhanced with further analysis of the entire cohort and in other cohorts.
In a similar manner to genetic linkage, interaction between CpG sites needs to be resolved to clarify interpretation of epigenetic studies
[65, 66]. The statistical challenge of modeling potentially highly-correlated CpG sites raises the issue of collinearity, whereby use of multiple regression to isolate the individual contribution of each CpG site to the association under study becomes difficult. To overcome this issue, we took two approaches. First, we performed univariate regression, whereby each of the six CpG units was investigated individually. This allowed identification of individual CpG units potentially playing a greater or lesser role in the phenotypes, but has the drawback of multiple testing and the potential of for false positives. The second complementary approach, principal components analysis, minimized multiple testing, while reflecting average methylation across all the six CpG units. Both methods were applied to investigate the effect of IGF2 DMR methylation on the outcome of anthropometry at 17 years (Figure
2) and the outcome of head circumference (Table
3). For anthropometry at 17 years, the principal components analysis confirms the univariate results, indicating a general association with methylation in the IGF2 DMR region studied. For head circumference, the associations with individual CpG units appear to be of differing signs for different CpG units, indicating that any association with methylation in the region studied is more likely to be CpG unit specific.