Association of cord blood methylation fractions at imprinted insulin-like growth factor 2 (IGF2), plasma IGF2, and birth weight
Altered methylation at Insulin-like Growth Factor 2 (IGF2) regulatory regions has previously been associated with obesity, and several malignancies including colon, esophageal, and prostate adenocarcinomas, presumably via changes in expression and/or loss of imprinting, but the functional significance of these DNA methylation marks have not been demonstrated in humans. We examined associations among DNA methylation at IGF2 differentially methylated regions (DMRs), circulating IGF2 protein concentrations in umbilical cord blood (UCB) and birth weight in newborns.
Questionnaire data were obtained from 300 pregnant women recruited between 2005 and 2009. UCB DNA methylation was measured by bisulfite pyrosequencing. UCB plasma concentrations of soluble IGF2 were measured by ELISA assays. Generalized linear regression models were used to examine the relationship between DMR methylation and IGF2 levels.
Lower IGF2 DMR methylation was associated with elevated plasma IGF2 protein concentrations (β = −9.87, p < 0.01); an association that was stronger in infants born to obese women (pre-pregnancy BMI > 30 kg/m2, β = −20.21, p < 0.0001). Elevated IGF2 concentrations were associated with higher birth weight (p < 0.0001) after adjusting for maternal race/ethnicity, pre-pregnancy BMI, cigarette smoking, gestational diabetes, and infant sex. These patterns of association were not apparent at the H19 DMR.
Our data suggest that variation in IGF2 DMR methylation is an important mechanism by which circulating IGF2 concentrations, a putative risk factor for obesity and cancers of the colon, esophagus, and prostate, are modulated; associations that may depend on pre-pregnancy obesity.
KeywordsIGF2 H19 ELISA DNA methylation Epigenetics Epidemiology
The ‘developmental origins of adult disease susceptibility’ hypothesis posits that environmental exposures early in the development permanently alter the phenotype, influencing susceptibility to lifetime risk of obesity and several cancers . These alterations are thought to be mechanistically driven by epigenetic alterations. The most stable and well-studied epigenetic mechanism is DNA methylation, which involves the covalent attachment of a methyl group to the 5-carbon position of the pyrimidine ring of cytosine residues when in 5′-CpG-3′ context. This epigenetic modification is faithfully transmitted to nascent DNA molecules during DNA replication and is maintained during somatic cell proliferation. DNA methylation contributes to the induction and maintenance of transcriptional gene silencing, propagating shifts in gene expression that orchestrate long-term susceptibility to several adenocarcinoma including colon [2, 3], ovary , prostate [5, 6], stomach , and esophagus . Epigenetic mechanisms have also been hypothesized to link birth weight, a proxy for the well-being of the offspring in utero, with some of these cancers in adulthood [9, 10, 11, 12, 13, 14, 15]. However, genomic regions in which DNA methylation increases risk of these chronic diseases are still unknown. Ultimately, cancer prevention will require identification of aberrant epigenetic profiles driving risk, demonstration of their functional significance as well as environmental exposures associated with their perturbation and ‘windows’ of susceptibility during life’s course.
Identification of aberrant DNA methylation marks has been hampered by a lack of stable targets with known baseline levels, allowing for deviations from baseline to be detected. The need for epidemiologic studies to use specimens obtained from non-invasive means, such as peripheral blood leukocytes or buccal cells, which are often the only tissues available in population-based studies, has presented additional hurdles; epigenetic regulation and dysregulation can vary by tissue and cell type. Genomically imprinted genes are among several major classes of genes that are regulated by DNA methylation . Unlike most autosomal genes that follow Mendelian laws of inheritance where both parentally derived alleles are expressed, imprinted genes are monoallelically expressed, with their functional haploidy controlled by parental allele-specific DNA methylation at differentially methylated regions (DMRs). Because these methylation marks are stable covalent modifications that are established before germ layer specification , these methylation profiles should be maintained in all germ layers, resulting in quantifiable systemic homogeneity in locus-specific epigenetic regulation [17, 18]. As such, imprinted DMRs stably exhibit ~50% methylation in somatic tissues.
Insulin-like-growth factor-2 (IGF2) is perhaps the most intensively studied imprinted gene and is paternally expressed, with at least two regulatory DMRs located upstream of exon 3 and upstream of neighboring maternally expressed H19 . Hypermethylation (methylation fractions higher than the expected ~50%) at the H19 DMR [2, 4] and CpG hypomethylation (methylation fractions lower than the expected ~50%) at the IGF2 DMR  have been associated with higher IGF2 mRNA levels suggesting increased transcriptional activity, presumably via relaxation of imprint controls. Such increased transcriptional activity is a common feature in adenocarcinomas of the ovary, esophagus and prostate [4, 8, 20, 21, 22, 23, 24, 25]. IGF2 DMR loss of methylation was observed in DNA that was obtained from peripheral blood leukocytes of adults exposed to severe caloric restriction periconeptionally, suggesting methylation shifts in response to adverse events occur early and are stable over many decades . Based on these observations, methylation status at regulatory sequences of imprinted genes has been proposed as ‘archives’ or ‘biosensors’ of early exposure [27, 28]. However, the functional significance of these DMR methylation marks has not previously been demonstrated in epidemiologic studies. The objective of these analyses was to examine variation in DMR methylation at sequences regulating IGF2/H19, in relation to UCB plasma IGF2 protein concentrations. We also determined whether IGF2 protein concentrations are associated with variation in birth weight, an indicator of maternal and fetal well-being.
A detailed description of study participants has been previously reported . Briefly, between April 2005 and June 2009, 1101 eligible pregnant women seen at one of two Durham, North Carolina, obstetric facilities were identified through appointment logs and were invited to enroll in the Newborn Epigenetics STudy (NEST), a multiethnic cohort study. Approximately 85% (n = 940) provided written consent to participate. Eligibility criteria were as follows: age ≥18 years, English speaking, and an intent to use one of two Durham County obstetric facilities which enabled specimen collection at delivery. We excluded women infected with HIV and those planning to have their newborns adopted. Women were identified through appointment logs of two prenatal clinics serving the above two obstetrics care facilities: Duke maternal fetal medicine (MFM) and Durham Regional Hospital Obstetrics, Durham County, NC, as part of the NEST. Gestational ages at enrollment were 19–42 weeks (median = 28 weeks). This report is limited to 300 (32%) deliveries in whom DNA methylation analyses were performed and protein measurements are available. The 300 women are comparable to the 940 women with respect to maternal age (mean age 29 years), obesity (body mass index [BMI], 28 kg/m2), and sex of offspring (49% men). This study was approved by the Institutional Review Board at Duke University Medical Center.
Questionnaire data collection
A standardized questionnaire was either self-administered or completed with a trained interviewer. The questionnaire items included socio-demographic information including marital status, educational attainment, race/ethnicity, smoking history, and current smoking status, a history of common chronic diseases and conditions, dietary supplement use, weight and height before pregnancy (from which BMI and weight gain was computed). BMI was calculated as pre-pregnancy weight in kilograms divided by height in square meters (kg/m2), and weight gain was the difference between usual weight and weight at the delivery visit. At delivery, medical records were abstracted to verify cigarette smoking during pregnancy, reported morbidity and age, and to obtain additional parturition data including blood pressure and mode of delivery, body temperature and chorioamnionitis (the latter two as indicators of infection which may influence protein measurements), as well as offspring information including sex, anthropometric measurements, and 5-min Apgar scores.
Specimen collection and processing
Umbilical cord blood specimens were collected in vacutainer tubes containing EDTA within minutes of delivery and inverted gently to mix the anticoagulant with the blood. Specimens were stored in the hospital refrigerator until transported to the laboratory usually within 12 h of collection where they were centrifuged to isolate plasma and buffy coat, which was aliquotted and stored at −80°C. DNA was extracted from the white blood cells in the buffy coat using Puregene reagents according to the manufacturer’s protocol (Qiagen, Valencia, CA).
Description of methylation measurements and primers used has been described elsewhere . Briefly, cord blood samples from the first 438 subjects were analyzed for methylation status at two DMRs controlling the expression of IGF2. The steps comprised modification of genomic DNA by treating it with sodium bisulfite using a high-throughput method previously described [4, 31] to convert unmethylated cytosines to uracils, leaving methylated cytosines unchanged. We then obtained a quantitative measure of methylation at each DMR using a Pyromark Q96 MD Pyrosequencing instrument (Qiagen; Valencia, CA). Three CpGs were measured at the IGF2 DMR , located upstream of the imprinted IGF2 promoter in exon 3, [2, 26] (chr 11p15.5; CpG site 1: 2,169,518; CpG site 2: 2,169,515; and CpG site 3: 2,169,499; NCBI Human Genome Build 37/hg19). We also measured DNA methylation at four CpG dinucleotides at the H19 DMR at a sequence motif that binds the CTCF zinc-finger protein (chr 11p15.5; CpG site 1: 2,024,261, CpG site 2: 2,024,259, CpG site 3: 2,024,257, and CpG site 4: 2,024,254; NCBI Human Genome Build 37/hg19). Methylation data were 98% complete. We validated both pyrosequencing assays using defined mixtures of plasmids containing the fully methylated or unmethylated versions of the bisulfate-modified sequence.
Plasma concentrations of IGF2 were measured using microplate enzyme-linked immunosorbent assays (Diagnostic Systems Laboratories, Webster, TX). Procedures for measurement followed the manufacture’s instructions. Absorbance was read using a microplate reader DSL, Webster, TX. The plasma IGF2 concentrations ranged from 50 to 2,000 ng/ml, and assay sensitivity was 2.2 ng/ml.
For analyses, maternal age at enrollment was categorized as 18–30, 30–39 versus 40 years or older; marital status was dichotomized into married/living with partner versus not married; educational level was categorized into groups defined as up to high school, some college to college graduate versus graduate/professional school; race/ethnicity was categorized into African Americans versus non-Hispanic White versus other, cigarette smoking was dichotomized into smoking during pregnancy versus not smoking; birth weight was treated as both a continuous variable and dichotomized at less < 2500 versus ≥2500 g; maternal BMI was categorized into <25, 25–29, 30–34, and 35 + kg/m2; weight gain during pregnancy was dichotomized at <30 pounds versus ≥30 pounds; delivery mode was classified as vaginal vs. C-section; and histological evidence of chorioamnionitis was dichotomized as yes or no.
We examined the distribution of methylation fractions for normality at each CpG dinucleotide using Kolmogorov–Smirnov tests followed by principal components and confirmatory factor analyses to determine whether methylation at the three CpGs measured in the IGF2 DMR and the four CpGs measured at the H19 DMR were correlated to justify the use of a single mean for each DMR. Cronbach’s alpha values for CpGs were 86 and 87% at the IGF2 DMR and H19 DMR, respectively. We also examined the distributions of birth weight, IGF2 proteins, maternal weight gain and usual BMI for normality and found that with the exception of maternal BMI and weight gain, these variables were normally distributed among the 300 participants under study. Thus, subsequent analyses dichotomized maternal BMI at ≤30 kg/m2, and weight gain at the 75th percentile.
The main goal of our analyses was to determine whether DNA methylation fractions at each DMR were associated with IGF2 protein concentrations. Although race/ethnicity and adult BMI have been previously associated with IGF2 concentrations and will be considered potential confounders, F-tests were used to identify additional potential confounders, by identifying factors associated with statistically significant variation (p ≤ 0.20) in IGF2 protein concentrations, and methylation levels at each DMR. In addition, chorioamnionitis was examined for potential effect modification and confounding as previous studies of predictors of IGF2 protein in otherwise healthy adults suggest acute infection may affect protein concentrations . Mode of delivery was evaluated for potential confounding or effect modifiers also because animal data suggests DNA methylation patterns may vary by delivery mode . Because 75% of African Americans in this study were >30 kg/m2 compared to 25% of Whites, we fitted linear regression models to examine the relationship between IGF2 protein and DMR methylation for African Americans and Whites and also for obese and non-obese women, separately. To further examine the association IGF2 DMR methylation and the protein product in the context of pre-pregnancy BMI or weight gain during the course of pregnancy, we repeated our analyses restricted to infants born to women >30 versus ≤30 kg/m2 and also gain weight below and above the 75th percentile. Cross-product terms for maternal BMI and DNA methylation at each DMR were also computed and included in statistical models where differences were observed in stratified analyses. Only variables with p value <0.05 were retained in final statistical models. To examine whether IGF2 protein concentrations were associated with birth weight, we also utilized linear regression models since birth weight was normally distributed. These latter models were adjusted for gestational diabetes, maternal BMI, race/ethnicity, and cigarette smoking, as these factors are known to independently influence birth weight in the general population. All statistical analyses were conducted in SAS version 9.3 (SAS Institute, Cary, NC, USA).
IGF2 proteins and covariable
Distribution of offspring IGF2 concentrations by maternal and offspring characteristics
IGF2 protein (ng/μl)
Maternal age (in years)
Some college/college graduate
Non-smoker during peri-conception
Smoker during peri-conception
≤2,500(≤5 lb 8.2 oz)
>2,500 (>5 lb 8.2 oz)
Sex of offspring
Maternal BMI (kg/m2)
Weight gain during pregnancy
IGF2 and H19 DMR methylation and covariables
IGF2 and H19 DMR methylation fractions and maternal and offspring characteristics
IGF2 DMR % (sd)
H19 DMR % (sd)
p = 0.742
p = 0.199
Sex of offspring
p = 0.289
p = 0.478
≤2,500(≤5 lb 8.2 oz)
>2,500 (>5 lb 8.2 oz)
p = 0.875
p = 0.194
Maternal BMI (kg/m2)
p = 0.618
p = 0.390
Weight gain during pregnancy
p = 0.487
p = 0.079
p = 0.180
p = 0.128
p = 0.634
p = 0.950
H19 DMR methylation and IGF2 protein concentrations
Associations between methylation fractions at IGF2 and H19 DMRs and IGF2 circulating concentrations stratified by race and maternal obesity
IGF2 CpG methylation
H19 CpG methylation
All participants (n = 300)
African Americans (n = 156)
Whites (n = 114)
BMI < 30 kg/m2(n = 190)
BMI ≥ 30 kg/m2(n = 76)
IGF2 DMR methylation and IGF2 protein concentrations
Table 3 also shows that hypomethylation at the IGF2 DMR was significantly associated with higher IGF2 protein concentrations (β = −9.87, p = 0.01). Adjusting for maternal BMI pre-pregnancy strengthened this association (β = −11.36, p = 0.01) and further adjusting for chorioamnionitis, birth weight and race/ethnicity did not alter this association. This association may be stronger in African Americans (p = 0.06) than Whites (p = 0.31). Analyses stratified by maternal BMI before pregnancy revealed an association between IGF2 DMR methylation and IGF2 protein levels, which was considerably stronger in infants born to obese women (β = −20.41, p = 0.003) corresponding to ~110 ng/μl increase in IGF2 protein concentration for every 5% decrease in DNA methylation levels at the IGF2 DMR. The magnitude of this association was markedly weaker in infants born to non-obese women (β = −4.75, p = 0.37). The cross-product term for obesity and IGF2 DMR methylation, however, was not statistically significant (p = 0.78). Because weight gain during pregnancy is strongly associated with pre-pregnancy BMI, repeating these analyses within each stratum of pre-pregnancy weight gain also did not reveal any insight to clarify the relative importance of these factors as the associations were much stronger in the 71 infants born to women with the largest weight gain (>75th percentile), when compared to those who gained ≤75th percentile. Intriguingly, including maternal weight gain in models with pre-pregnancy BMI and its cross-product term did not suggest weight gain was an important factor in the association between IGF2 DMR methylation and IGF2 protein concentration.
IGF2 protein concentrations and birth weight
Regression model β-coefficients and standard errors of the association between umbilical cord blood IGF2 protein concentrations and birth weight (n = 278)
IGF2 protein concentrations (ng/μl)
Race/ethnicity (African American vs. Caucasian)
Maternal cigarette smoking
BMI before pregnancy (obese vs. non-obese)
Sex of infant
We examined DNA methylation within two regulatory regions of imprinted IGF2 in relation to circulating IGF2 concentrations in UCB of newborns. We observed a significant increase in IGF2 protein concentrations with decrease in methylation levels at the IGF2 DMR. This association was strongest among infants born to women with a pre-pregnancy BMI > 30, similar in African Americans and Caucasians. Effect sizes of a similar magnitude have been previously reported at the IGF2 DMR and other regions [26, 34, 35]. We also evaluated IGF2 protein concentrations in relation to birth weight, and observed elevated IGF2 protein concentrations were associated with a higher birth weight. We found no evidence for associations between CpG methylation at the H19 DMR and IGF2 protein concentration after adjusting for maternal obesity and race/ethnicity. We provide the first evidence for an association between DNA methylation at the IGF2 DMR and IGF2 protein concentrations. These findings are consistent with the interpretation that DNA methylation at the IGF2 DMR is functionally relevant to the production of the potent mitogenic growth factor IGF2 and that soluble IGF2 independently contributes to variation in birth weight.
The exact mechanism by which loss of methyl groups at the IGF2 DMR increases IGF2 protein concentrations, exacerbated by maternal obesity, is still unknown, although changes in DNA methylation on the maternally derived allele has been shown to relax imprint controls, reactivating the otherwise silent maternally derived allele, which has been shown in previous studies to increase IGF2 mRNA and presumably protein products. Because imprinted genes are usually found in clusters  and their regulation may be networked , methylation alterations of this DMR have the potential for regulating (and dysregulating) multiple genes on different chromosomes. Therefore, methylation at this single DMR can provide information on epigenetic dysregulation elsewhere in the genome . Using DNA obtained from leukocytes, methylation differences of a similar magnitude were reported among Gambians  and Dutch  adults who were exposed to nutritional challenges periconceptionally. Together, these findings support the idea that methylation marks acquired in the periconceptional period is similar across tissue types and is stable over many years. The stability of DNA methylation marks at the IGF2 DMR during a 3-year period was also recently reported among adult population controls . The small magnitude of differences is also consistent with the contention that methylation marks acquired as a result of adjustments to perceived environment may not be as large as those found in cancer [39, 40]). Strong inverse associations between DNA methylation levels at the IGF2 DMR using DNA obtained from leukocytes, biallelic expression of this otherwise monoallelically expressed gene, and elevated colon cancer risk [2, 38, 41] have been reported. In tissue specimens, hypermethylation at the H19 DMR is also frequent in cancers of the ovary , breast , esophagus , and prostate . The IGF2 and H19 DMRs have been previously proposed as biomarkers, biosensors, or archives of early exposures [2, 27, 28]. Therefore, our findings of a strong relationship between lower methylation at the IGF2 DMR and elevated IGF2 protein levels support the use of DNA methylation levels at this and perhaps other DMRs as biosensors or archives of early exposure. Differences observed by maternal obesity may be partly explained by genetic factors that also play a role in modulating IGF2 levels in the offspring.
Findings that protein levels of the mitogenic IGF2 are strongly associated with decreased methylation at the IGF2 DMR and with birth weight support the interpretation that variation in birth weight is modulated, at least in part, by IGF2 plasticity. This may also be occurring at other DMRs, if hypothesized co-regulation with other imprinted regions  can be demonstrated. Previous studies have reported an association between circulating IGF2 protein concentrations and high birth weight  as well as risk of obesity  and several cancers [48, 49] in adulthood, suggesting early exposures may contribute to the variation in birth weight via epigenetic alterations.
Our inability to find a consistent association between H19 DMR methylation and IGF2 protein levels except among the 188 infants born to women with BMI < 30 was unexpected. Current thinking for transcriptional control of the imprinted IGF2/H19 domain correlates increased DNA methylation at the H19 DMR with failure to bind methylation sensitive CTCF, which can transcriptionally activate IGF2 on the normally silent maternal chromosome [50, 51, 52]. We cannot exclude the possibility that these were chance findings, since the statistical power to conduct multivariable analyses may have been inadequate. However, others have also failed to find a relationship between methylation at the H19 DMR and IGF2 transcription among children conceived in vitro and in vivo  suggesting that at the H19 DMR, other processes including genetic factors or other unmeasured environmental stimuli may contribute to increased IGF2 protein concentrations and body size, independent of DNA methylation alterations.
The main limitation of this study is the small sample size which, although more than adequate for examining crude relationships among DMR methylation fractions, protein IGF2 concentrations and birth weight, may have been inadequate for multivariable analyses, especially when testing interaction by race and maternal BMI. Despite the need for larger studies to clarify these relationships, the analyses raised the possibility that maternal obesity, which may be driven by other factors including genetic variation, also plays a role in modulating IGF2 concentrations. This suggests that a more detailed understanding of these epigenetic targets will require genetic analyses within the sequences under evaluation. Another limitation is that DNA obtained from unfractionated blood containing multiple cell types may have distinct epigenetic profiles depending on the cell population analyzed. However, we have previously described the analysis of fractionated cord blood for methylation at the same H19 and IGF2 DMRs, and the major cell types showed no significant differences in methylation levels . Additionally, overall white blood cell counts vary and could confound the methylation fraction measures. However, inclusion of chorioamnionitis during parturition into statistical models did not alter our findings, suggesting variation in total cell counts may not contribute to differences in methylation at these DMRs. Furthermore, although we demonstrated an association between DMR methylation and protein levels downstream, we were limited by a lack of transcription data. However, a positive correlation between the levels of IGF2 mRNA and DNA methylation at these DMRs has been previously reported. These limitations notwithstanding, our findings suggest that IGF2 DMR plasticity is an important mechanism by which IGF2 protein levels are modulated.
In summary, among 300 newborns, we found a strong inverse association between DNA methylation at the IGF2 regulatory sequence and elevated plasma protein concentrations of the encoded mitogenic growth factor. This relationship appeared to be stronger in infants born to women who were obese before pregnancy, but not weight gain. Higher IGF2 protein concentrations were also associated with higher birth weight. Herein, we provide evidence in support of the functional significance of aberrant methylation profiles at regulatory elements controlling the expression of imprinted IGF2, supporting the use of IGF2 DMR methylation as an archive or biosensor of early exposure. If co-regulation of imprinted domains can be demonstrated in humans, and transcripts are also associated with DNA methylation and protein levels for other imprinted genes in the same population, our findings would support the use of methylation marks in an imprint regulatory network to improve exposure assessment in studies of cancer and other chronic diseases.
We thank the parents and children of the Newborn Epigenetics Study, research nurse Tammy Bishop, and research personnel: Stacy Murray, Carole Grenier, Darby Kroyer, Natasha Duggan, Suba Narasimhan, and James Nathan Yarnall for enrolling and tracing participants, and processing the specimens; and the obstetrics faculty and staff at Duke University and Durham regional Hospitals, Durham, NC. This work was supported in part by R21ES014947, R01ES016772, R01DK085173, K01CA104517 the American Cancer Society (grant number ACS-IRG-83-006) and the Fred and Alice Stanback Foundation. Dr. Susan Murphy was also supported by the Duke Comprehensive Cancer Center and Duke Environmental Health Sciences Research Center.
Conflict of interest
The authors declare that they have no conflict of interest.
This article is distributed under the terms of the Creative Commons Attribution License which permits any use, distribution, and reproduction in any medium, provided the original author(s) and the source are credited.
- 23.Nakagawa H, Chadwick RB, Peltomaki P, Plass C, Nakamura Y, de La Chapelle A (2001) Loss of imprinting of the insulin-like growth factor II gene occurs by biallelic methylation in a core region of H19-associated CTCF-binding sites in colorectal cancer. Proc Natl Acad Sci U S A 98:591–596PubMedCrossRefGoogle Scholar
- 38.Cruz-Correa M, Zhao R, Oviedo M et al. (2009) Temporal stability and age-related prevalence of loss of imprinting of the insulin-like growth factor-2 gene. Epigenetics 4:114–118Google Scholar
- 43.Yang R, Wang Y, Zhu X, Gao X (2009) Study on genomic imprinting of IGF2 in breast cancer. Wei sheng yan jiu. J Hyg Res 38:423–425Google Scholar