DNA methylation of imprinted genes at birth is associated with child weight status at birth, 1 year, and 3 years

Understanding factors that influence the risk of obesity in children is crucial to the development of new strategies for obesity prevention. Obesity in early childhood is a risk factor for obesity later in life [1–3] and for a number of chronic diseases in both childhood [4] and adulthood [5]. Birth weight has been associated with weight outcomes later in life, particularly for those who are on the extremes of the birth weight distribution [6–9]. Early identification of obesity or its risk factors will inform interventions to prevent the progression of obesity and its consequences later in life [10]. Consistent with the developmental origins of disease hypothesis, the intrauterine environment is hypothesized to influence an individual’s later susceptibility for chronic diseases [11], including obesity [12, 13].

Epigenetic modifications have been proposed as a mechanism for the in utero origin of later obesity, and a growing literature has found supporting evidence [14–16]. DNA methylation is the most studied epigenetic mechanism in humans, due in part, to its stability. DNA methylation that controls the monoallelic expression of imprinted genes is established during gametogenesis and is stably maintained throughout somatic division [17–21] and therefore provides a stable “register” of early in utero exposures. A study of famine survivors found that adults who experienced famine in utero had hypo-methylation of the imprinted IGF2 gene compared to their same sex siblings who had not experienced famine in utero [22]. The significance of this locus was reported to not have been replicated in this cohort using alternate techniques, including RRBS. However, this technique is generally biased toward CG-rich areas and may not have covered the specific and limited number of CpGs that comprise the IGF2 DMR. Additional genes, such as INSR and CPT1A, have also been identified in association with exposure to the Dutch famine [23]. Another study found that maternal nutrition, affected by striking seasonal variations in food intake in the Gambia, influenced methylation at RBM46 [24]. A colorectal cancer study found that methylation status of the IGF2/H19 imprinted locus of adult controls was maintained 3 years later [25]. Moreover, a study of NEST children between birth and age 1 year found similar results at the IGF2/H19 locus [26].

Select imprinted genes have been identified as playing a role in the development of fetal over and undergrowth caused by imprinting defects. The IGF2 locus is used in clinical diagnostic settings to identify Beckwith-Wiedemann syndrome, which is characterized by overgrowth [27], and the H19 locus has been used in the diagnosis of Silver-Russell syndrome (SRS), which is characterized by undergrowth [28].

Although epigenetic data linking DNA methylation and childhood obesity has increased exponentially in the last 5 years [29, 30], few regions agnostically identified have been replicated. This could be in part due to differences in the ethnic composition; however, differences could also be due to the sex composition. At imprinted loci, weight has been associated with the IGF2 locus. Studies have found a relationship between the IGF2 domain and fetal growth [31–34] and children’s body composition or weight [32, 35, 36]. Data with directional consistency in associations between additional differentially methylated regions (DMRs) and weight gain are required.

This study aims to assess the association between methylation at nine DMRs of imprinted genes and birth weight for gestational age (BW/GA) z score, weight-for-length (WFL) z score at 1 year, and BMI z score at 3 years. In this analysis, we include the following DMRs: MEG3 and MEG3-IG, which are involved in regulating the delta-like 1 homolog/maternally expressed gene 3 imprinted domain on chromosome 14q32.2; IGF2 and H19, which are involved in the imprinting of the insulin growth factor 2/H19 domain on chromosome 11p15, which are located upstream of the imprinted promoters of IGF2 and at the imprinting control region for the IGF2/H19 imprinted domain near the H19 promoter, respectively; PLAGL1 at the pleiomorphic adenoma gene-like 1 locus at 6q24.2; MEST at the mesoderm-specific transcript promoter at 7q32.2; NNAT at the neuronatin locus at 20q11.23; PEG3 at the paternally expressed gene 3 promoter region at 19q13.43; and PEG10 at the epsilon sarcoglycan and paternally expressed gene 10 promoter region at 7q21.3. We selected these regions for their association with infant and child growth [32, 34, 37], chronic disease [22, 38], and parental obesity [39].

In these analyses, we examined DNA methylation of nine regulatory regions at birth and anthropometric measures at birth and age 1 and 3 years. No DMR showed a consistent association between methylation and anthropometric z scores at all 3 time points explored. Our key findings were that high methylation of the sequences regulating the PEG10 DMR at birth was associated with a higher age 1-year WFL z score and 3-year BMI z score in boys, low methylation at the NNAT DMR was associated with higher BMI z scores at age 3 in girls, and high methylation was associated with lower WFL z scores at 1 year and BMI z scores at 3 years. These associations persisted after excluding SGA infants. Additional findings included associations between methylation at the PLAGL1 DMR and WFL z scores in girls at age 1, as well as an association at the IGF2 DMR at age 1 among boys. At age 3, we also observed an association between methylation at the PEG3 DMR and BMI z scores in girls. These results suggest that methylation of imprinted genes at birth is associated with anthropometric measures at ages 1 and 3 years, with the PEG10 and NNAT DMRs potentially indicating an early risk for obesity, and MEST potentially indicating a lower risk for obesity.

This study adds to a growing body of epidemiologic evidence on early postnatal growth associated with DNA methylation at birth and suggests that DNA methylation at multiple DMRs may be associated with WFL z score at age 1 year and BMI z score at age 3 years. PLAGL1, MEST, NNAT, and PEG10 have been associated with obesity or weight in previous literature. Paternal obesity has been previously associated with reduced PEG10 transcription in mouse placentas [46]. Our results indicating a higher level of PEG10 methylation is associated with a greater BMI z score show a similar pattern. Previous literature has also shown a possible association between increased MEST expression and inhibition of adipogenesis [47]. The results of this study echo these findings, as boys with higher than average methylation levels had lower BMI z scores. Methylation at the MEST, NNAT, and PEG10 DMRs has also been previously associated with paternal obesity [39], and small and large for gestational age [48, 49]. In addition, the NNAT gene has been associated with severe obesity in childhood and adulthood [50]. Higher methylation at PLAGL1 has also been associated with maternal obesity [39], and fetal and postnatal growth [37]. A previous study found a positive correlation between PLAGL1 methylation and BMI z scores at age 1 year. PLAGL1 is thought to be an imprint control region [51]; however, the implications of this in relation to a potential role in the risk of early obesity are not yet clear. More research is needed to gain a better understanding of the relationship between child BMI and methylation at these DMRs. Previous literature has supported the role of the IGF2 DMR in fetal growth [31–34] and birth weight [32, 52], as well as infant weight gain [32, 36] and child adiposity [35]. We observed an association between IGF2 methylation and lower BMI z scores; however, this association was not seen at birth or age 3 years.

We observed sex-specific differences in the associations between DNA methylation at birth and anthropometric z scores at ages 1 and 3 years. Sex-specific methylation has been previously observed in relation to nutrition and other environmental exposures [44, 45, 53], as well as in relation to outcomes, such as small for gestational age [54]. However, these studies did not find sex-specific differences in methylation in PEG10, NNAT, or MEST. This study also adds to the growing literature on sex-specific DNA methylation.

We conducted additional analysis to explore the influence of infants who are SGA, as their growth patterns may differ from those of infants who are not SGA, and found that SGA may be associated with DNA methylation. No associations between methylation and BW/GA z scores remained significant after SGA exclusion, suggesting that SGA may have been driving these associations. We also found that associations between PEG3 and BMI z scores at both ages 1 and 3 years became statistically significant after exclusion of SGA infants. This suggests that perhaps the association between PEG3 methylation and SGA is in the opposite direction of the association between PEG3 methylation and BMI z scores for non-SGA infants, thus attenuating the original association. However, a cautious interpretation is warranted, as exclusion of SGA infants also decreased the statistical power, which may have led to unstable estimates. Additional supplemental analysis including only the 166 infants who had non-missing anthropometric values at all 3 time points showed that many of our associations remained statistically significant, including our key findings at the PEG10 and NNAT DMRs. This suggests that these associations were not related to differences in the samples at each time point. Additional associations emerged as statistically significant at MEST, IGF2, PEG10, and NNAT; however, these results must be interpreted with caution, as the analysis was underpowered. Similarly, the results of the supplemental analysis including maternal alcohol consumption and the analysis stratified by race/ethnicity should be interpreted with caution, as our sample size was greatly reduced, and estimates may be underpowered.

The direction of the association between higher methylation at MEST and z scores changed from positive to negative from birth to ages 1 and 3. The reasons for this are unclear; however, it is possible that the modest increase in BW/GA z scores at birth, which we found to be associated with a higher level of methylation at MEST, is also associated with a greater likelihood of becoming lean as the child grows and becomes more mobile. MEST expression has been previously associated with obesity in mice [55]; however, a study in humans found MEST to possibly inhibit adipogenesis [47].

This study benefits from an ethnically diverse cohort, and prospectively collected data at multiple time points. This facilitates a better understanding of the timing of methylation with regard to our outcome of interest, weight gain. In addition, the use of BMI z scores provides widely accepted estimate of adiposity that accounts for a child’s age. However, it is not without limitations. Our study only included nine DMRs of imprinted genes. Although these genes were chosen because they have been linked to growth or chronic disease, it is possible that important genes have been left out of this study. In addition, our study’s small sample size may have limited our ability to see statistically significant differences associated with DNA methylation among our population. Although our analyses were hypothesis driven, multiple testing is a limitation in this study, as it may increase the possibility that our results are observed by chance. However, many associations remained (PEG10, MEST, PLAGL1, NNAT) even after the stringent Bonferroni correction. A final limitation was the use of weight status instead of weight gain. Much of the literature has pointed to weight gain in the first year of life as being associated with later obesity [8]; however, there is some literature indicating high weight status in early childhood as a risk factor for later obesity [56]. It is unclear whether or not these findings are related to later obesity.

In addition, the results of this study show differences in anthropometric z scores in association with DNA methylation that is either lower or higher than the “average” methylation of our study sample. Notably, the actual change in continuous methylation associated with our results is likely small (approximately 1%). However, even a 1% change in methylation has been previously shown to result in a doubling or halving of gene expression at these imprint regulatory regions [45]. Not assessing DNA methylation continuously may be a limitation of our study, as it creates a challenge in comparing our results to those of other studies. However, we believe the results presented in this study may be meaningful for public health, as it provides a range of methylation values that may be associated with anthropometric, and possibly even adiposity.

In summary, our study findings suggest that DNA methylation of the PLAGL1, MEST, PEG10, and NNAT DMRs at birth is associated with BMI z scores in early childhood and varies by sex. Longitudinal assessment of DNA methylation in these DMRs at older age time points is needed to determine whether or not methylation at these DMRs is associated with obesity later in life. Determining the associations between DNA methylation and early obesity risk is important, as DNA methylation of regulatory regions may serve as markers for the assessment of early obesity risk. However, gaining a better understanding of the exposures that affect methylation at these regions is also important, as exposures that modify methylation of regions that are associated with obesity risk may be a good target for early obesity prevention efforts.