An unbalanced maternal diet in pregnancy associates with offspring epigenetic changes in genes controlling glucocorticoid action and foetal growth


  • Amanda J. Drake,

    1. Endocrinology Unit, University/BHF Centre for Cardiovascular Science, Queen's Medical Research Institute, University of Edinburgh, Edinburgh, UK
    Search for more papers by this author
  • Rhoanne C. McPherson,

    1. Endocrinology Unit, University/BHF Centre for Cardiovascular Science, Queen's Medical Research Institute, University of Edinburgh, Edinburgh, UK
    Search for more papers by this author
  • Keith M. Godfrey,

    1. Medical Research Council Lifecourse Epidemiology Unit, University of Southampton, Southampton, UK
    2. NIHR Southampton Biomedical Research Centre, Southampton General Hospital, University of Southampton and University Hospital Southampton NHS Foundation Trust, Southampton, UK
    Search for more papers by this author
  • Cyrus Cooper,

    1. Medical Research Council Lifecourse Epidemiology Unit, University of Southampton, Southampton, UK
    Search for more papers by this author
  • Karen A. Lillycrop,

    1. NIHR Southampton Biomedical Research Centre, Southampton General Hospital, University of Southampton and University Hospital Southampton NHS Foundation Trust, Southampton, UK
    Search for more papers by this author
  • Mark A. Hanson,

    1. NIHR Southampton Biomedical Research Centre, Southampton General Hospital, University of Southampton and University Hospital Southampton NHS Foundation Trust, Southampton, UK
    Search for more papers by this author
  • Richard R. Meehan,

    1. MRC Human Genetics Unit, IGMM, Western General Hospital, Edinburgh, UK
    2. Breakthrough Breast Cancer Research Unit, Western General Hospital, University of Edinburgh, Edinburgh, UK
    Search for more papers by this author
  • Jonathan R. Seckl,

    1. Endocrinology Unit, University/BHF Centre for Cardiovascular Science, Queen's Medical Research Institute, University of Edinburgh, Edinburgh, UK
    Search for more papers by this author
  • Rebecca M. Reynolds

    Corresponding author
    • Endocrinology Unit, University/BHF Centre for Cardiovascular Science, Queen's Medical Research Institute, University of Edinburgh, Edinburgh, UK
    Search for more papers by this author

Correspondence: Rebecca M. Reynolds, Endocrinology Unit, University/BHF Centre for Cardiovascular Science, Queen's Medical Research Institute, University of Edinburgh, 47 Little France Crescent, Edinburgh EH16 4TJ, UK. Tel.: +44 131 2426762; Fax: +44 131 2426779; E-mail:



In epidemiological studies, adverse early-life conditions associate with subsequent cardiometabolic disease. Hypothesized causes include maternal malnutrition, foetal glucocorticoid overexposure and reduced growth factors. Animal studies suggest a role for epigenetic processes in maintaining early-life effects into adulthood, but human relevance is unknown. We aimed to investigate relationships between an unbalanced maternal diet in pregnancy, neonatal and adult anthropometric variables with methylation at key genes controlling tissue glucocorticoid action and foetal growth.


We studied 34 individuals aged 40 from the Motherwell cohort study whose mothers ate an unbalanced diet in pregnancy, previously linked with elevated blood pressure and cortisol in adult offspring.


DNA methylation at 11β-hydroxysteroid dehydrogenase type 2 (HSD2), glucocorticoid receptor (GR) and insulin-like growth factor 2 (IGF2) was measured by pyrosequencing on buffy coat DNA.


Methylation at specific CpGs in the HSD2 promoter and at one of the IGF2 differentially methylated regions (H19 ICR) correlated with neonatal anthropometric variables. CpG methylation within HSD2, GR and H19 ICR was positively associated with increased adiposity and blood pressure in adulthood. Methylation at GR (exon 1F) was increased in offspring of mothers with the most unbalanced diets in pregnancy.


Alterations in DNA methylation at genes important in regulating circulating cortisol levels, tissue glucocorticoid action, blood pressure and foetal growth are present in adulthood in association with both early-life parameters and cardiometabolic risk factors. The data indicate a persisting epigenetic link between early-life maternal diet and/or foetal growth and cardiovascular disease risk in humans.


Epidemiological studies show an association between exposure to an adverse early-life environment and subsequent cardiometabolic disease.[1] These effects extend across the normal range of birthweights and are not confined to those born very small or premature. Although the mechanisms are unclear, animal studies suggest a role for epigenetic processes[2-4] and emerging human data indicate altered DNA (cytosine) methylation in neonates in association with intrauterine growth retardation[5] and in babies born to obese mothers.[6] Reports showing altered DNA methylation in adulthood in healthy men born with low birthweight[7] and in adults exposed to the Dutch Hunger Winter in utero[8, 9] suggest that such epigenetic changes may persist into adult life.[10]

One major issue with the identification of epigenetic changes associated with human disease is that of the tissue and cell-type specificity of DNA methylation patterns. Thus, patterns of epigenetic modifications in accessible tissues such as peripheral blood may not necessarily represent those present at a tissue level and which may be involved in a particular disease process.[11] Nevertheless, a number of studies have identified alterations in DNA methylation at candidate genes in peripheral blood in association with early-life environmental factors. For example, exposure to maternal depression during pregnancy associates with altered DNA methylation at the glucocorticoid receptor (GR)[12] and at the transmembrane serotonin transporter SLC6A4[13] in neonatal cord blood cells. As exposure to maternal depression in the antenatal period is associated with an increased risk of neurobehavioural disorders in childhood,[14, 15] these studies suggest that some epigenetic alterations in peripheral blood may be of relevance to the target organ – here the brain. Additionally, there is increasing interest in the potential for epigenetic analysis to provide useful biomarkers of disease risk across the lifespan,[16, 17] so that the identification of alterations in DNA in peripheral blood that associate with early-life factors may allow early identification and intervention in individuals at risk of disease in later life.

Glucocorticoids play a crucial role in tissue development and maturation and have well-described effects on metabolism. Increased activation of the hypothalamic–pituitary–adrenal (HPA) axis is a recognized association of exposure to adverse environment in early life.[18] Glucocorticoid action is mediated by GRs, which are expressed in most tissues from the early-foetal stages onwards,[19] under the control of multiple untranslated first exons.[20, 21] Studies in animal models and in humans have suggested that the early-life environment can affect DNA methylation at the GR promoter.[12, 20, 22-24] The access of glucocorticoids to GR is modulated by the 11β-hydroxysteroid dehydrogenase enzymes (HSD1 and 2). HSD2 is expressed in mineralocorticoid target tissues and is responsible for selective access of aldosterone to renal mineralocorticoid receptors; reduced activity of HSD2 is associated with hypertension,[25, 26] and data from an animal model of programming suggest that adverse prenatal events are associated with epigenetic modifications at the HSD2 promoter.[27] Studies in humans and in animal models showing that nutrition in early life is associated with altered DNA methylation at the differentially methylated regions (DMRs) controlling the expression of the insulin-like growth factor 2 (IGF2) gene,[8, 28] encoding one of the major prenatal growth factors,[29] suggest that this gene may also represent a key target for early-life programming effects.

We have previously studied individuals in Motherwell, Scotland, whose mothers were advised to eat one pound (0·45 kg) of red meat per day and to avoid carbohydrate-rich foods during pregnancy. The adult offspring of mothers who ate the most unbalanced diets have higher blood pressures,[30] higher cortisol levels[31] and heightened HPA axis responses to stress.[32] Here, we report findings of a pilot study investigating the relationships between early-life factors and adult phenotype with methylation status at the candidate genes GR, HSD2 and IGF2 in the Motherwell cohort.


Ethics statement

The study was approved by the Research Ethics Committee of Lanarkshire Health Board, and subjects gave informed written consent.

Study population

We have previously described the sample of men and women who were born in Motherwell Maternity Hospital during 1967–1968 for whom details of obstetric history and maternal diet during pregnancy are available.[30, 31] The mother's height and parity were abstracted from the antenatal records, together with her weight measurements from booking (median 111·5 days) and at the end of pregnancy (median 269 days). In addition, we abstracted summaries of the mothers' dietary intakes in ‘early’ (≤20 completed weeks) and ‘late’ (>20 weeks) pregnancy recorded by staff at the antenatal clinic. These described the number of portions eaten daily/weekly of 10 foods: meat, fish, eggs, cheese, green vegetables, potatoes, bread, cakes/scones/biscuits, sweets and milk. Consumption of protein rich foods almost doubled, whilst that of carbohydrate-rich foods fell to a third.[30] The baby's birthweight and gestation at birth were abstracted from the original obstetric records. In 2008, we invited 70 individuals who had participated in a recent study[32] to attend a local morning clinic after an overnight fast, avoiding caffeine, alcohol and smoking on the study morning. Those taking any form of glucocorticoids in the previous 3 months were excluded. Height, weight, waist circumference and blood pressure were measured and a venous blood sample taken.

DNA methylation

DNA was extracted from buffy coat using the Gentra Puregene Blood Kit (Qiagen, Crawley, UK), and 1 μg of DNA was subjected to bisulphite conversion using the Epitect Bisulphite Kit (Qiagen). Bisulphite treatment converts unmethylated cytosines to thymine, whilst methylated cytosines remain unaffected. Pyrosequencing of bisulphite converted DNA utilizes the sequencing by synthesis principle to measure the % methylation at individual cytosines in a given sequence. Pyrosequencing primers were designed for exon 1(C) and 1(F) of the GR promoter region and for the promoter region of 11βHSD2, including the areas at which DNA methylation correlates negatively with gene expression[33] using PyroMark Assay Design 2.0 software (Qiagen). The positions of these assays are shown in Fig. 1, and primer sequences are given in Box 1. Published assays were used for the DMRs that are known to control the expression of IGF2 (DMR0, DMR2 and the H19 imprinting control region (ICR).[34] All primers were purchased from Eurogentec (Southampton, UK). Pyrosequencing was carried out using SQA reagents (Qiagen) on the PSQ HS-96A. Data were analysed using Pyro Q-CpG Software (Qiagen). Background nonconversion levels were in the region of 1–3%.

Figure 1.

Schematic diagram areas of methylation assessed methylation across the HSD2 promoter (a) and at exons 1(C) (b) and 1(F) (c) of the glucocorticoid receptor (GR) promoter. (a) – underlined area demonstrates the two regions of the HSD2 promoter assessed. (b) – For GR exon 1C, exons are shown in small letters. Assay 3 situated within exon 1-C1, and assays 1 and 4 overlap within 1-C3. (c) Analysed area of GR promoter 1F (capitals) and start exon 1F.

Box 1 Primer sequences for pyrosequencing assays

GenePCR Primers and hybridization temperatureSequencing primerProduct size (bp)
GR exon 1(C)
Region 1



Reverse (biotinylated)





Region 2



Reverse (biotinylated)





GR exon 1(F)



Reverse (biotinylated)






Region 1

−763 to −704



Reverse (biotinylated)





Region 2

−655 to −560



Reverse (biotinylated)





Statistical analysis

The percentage methylation at the CpG sites was measured across the whole buffy coat sample: at an individual allele, a specific cytosine is either methylated or not (i.e. 0% or 100%), but at the whole-sample level, the proportion of cells in which an individual site is methylated varies between individuals. Thus, the overall level of methylation may vary in a continuous fashion from 0% to 100%. As the distributions of methylation levels of 11 βHSD2 and H19 ICR were skewed, we used natural log to transform these variables and verified normal distribution using the Shapiro–Wilk Test. Other variables were normally distributed. Pearson correlation was used to examine associations between offspring epigenetic variables and offspring characteristics, having confirmed a similar pattern of correlation using untransformed data (for methylation of 11 βHSD2 and H19 ICR) using Spearman correlation. As methylation at individual CpGs within a gene were also highly correlated, we used principal components analysis (PCA), a statistical technique producing new variables that are uncorrelated linear combinations of the CpG variables that maximize the explained variance. PCA was performed for each gene using Z-scores for each CpG site calculated using the natural log-transformed values. Multiple linear regression analyses were performed adjusting for potential confounding variables. Residuals were checked for normality. Data were analysed using Statistica Release 6.0 and spss version 19 (SPSS Inc., Chicago, IL, USA).


Of 70 potential participants, 53 replied to invitation letters (76% response rate), of whom 49 were interested in participating. On further contacting, 15 of these individuals were either uncontactable, unwilling to attend, or failed to attend the clinic. 34 individuals aged 40 (0·12) years completed the study (Table 1). Nonparticipants were of similar age to the participants, and there were no differences in early-life or maternal variables between participants and nonparticipants.

Table 1. Characteristics of study participants and their mothers
 Participants n = 34 (12M, 22F)Nonparticipants n = 36 (19M, 17M)
  1. Data are mean (SD).

Adult variables
Age (years)40·0 (0·6)40·0 (0·5)
Weight (kg)73·6 (15·6) 
BMI (kg/m2)26·1 (4·2) 
Waist (cm)87·9 (14·5) 
Systolic BP (mmHg)122·7 (18·2) 
Diastolic BP (mmHg)80·6 (11·1) 
Early-life variables
Birthweight (g)3101 (324)3097 (352)
Birth length (cm)48·4 (2·8)48·3 (2·9)

Ponderal index at

birth (kg/m3)

27·7 (5·1)27·7 (5·1)

Gestation at delivery


272·5 (5·5)273·0 (5·6)
Maternal variables
Maternal age at birth (years)28·2 (6·1)27·9 (6·0)
Antenatal weight at booking (kg)59·6 (7·3)59·2 (9·7)
Antenatal BMI at booking (kg/m2)23·6 (2·7)23·3 (3·7)

Candidate gene methylation and neonatal anthropometry

Methylation across the HSD2 gene promoter was low (0·8–9·6%). There were significant positive correlations between methylation at HSD2 region 1 and birthweight (Table 2), which remained significant in regression analysis with birthweight as the dependent variable adjusting for potential confounding factors including gestation at delivery, parity, gender and maternal antenatal BMI (β = 0·43, partial correlation coefficient = 0·48, P = 0·01). There were also inverse correlations between methylation at specific CpGs within HSD2 region 2 and neonatal ponderal index (Table 2), which remained significant in regression analyses. GR promoter methylation was also low [mean 2·4 (0·08)%], and there were no correlations with size at birth (Table 2). Mean methylation across the 12 CpG sites of H19 ICR was 53·6 (3·9)%. In univariate analyses, there was a significant inverse relationship between birth length and H19 ICR methylation but no associations with birthweight (Table 2). These findings remained significant in regression analyses with birth length as the dependent variable controlling for potential confounding factors (β = −0·57, P < 0·0001). Mean methylation of IGF2 DMR0 and DMR2 was 47·6 (0·8)% and 58·3 (0·5)%, respectively, but this was not related to neonatal anthropometry.

Table 2. Pearson correlationsb of methylation status of H19, HSD2 and GR with early-life variables
 HSD2 Region 1HSD2 Region 1HSD2 Region 2HSD2 Region 2H19 ICRH19 ICRGR Exon 1FGR Exon 1C
  1. a

    GR, glucocorticoid receptor; ICR, imprinting control region.

  2. b

    Data presented are Pearson correlation coefficients. Methylation levels for HSD2 and H19 ICR were ln-transformed prior to analysis.

  3. c

    P < 0·05.

CPGBirthweightPonderal indexBirthweightPonderal indexBirth lengthBirthweightBirthweightBirthweight
7  −0·26−0·03−0·43c−0·090·07−0·02
8  0·0060·33−0·44c−0·090·110·17
9  −0·001−0·09−0·44c−0·100·06 
10  0·140·28−0·43c−0·080·06 
11  0·11−0·04−0·42c−0·08  
12  0·12−0·04−0·42c−0·07  
13  −0·030·30    
14  −0·03−0·38c    
15  0·060·30    
16  0·190·04    
17  −0·03−0·11    
18  −0·040·22    
19  −0·25−0·34c    
20  −0·08−0·15    
21  0·12−0·07    

Methylation, adult anthropometry and blood pressure

In univariate analyses, increased methylation at several of the individual CpG sites within HSD2 region 1 was associated with offspring adiposity. This finding was confirmed using PCA that showed significant associations between increased methylation across CpGs 1–4 and 6 (Factor 1) and increased weight, waist circumference and BMI, and increased methylation at CpG 5 (Factor 2) in association with increased weight and waist circumference (Table 3). There were significant correlations between systolic blood pressure and methylation at specific CpGs in HSD2 region 2 (Fig. 2). Increased methylation at GR exon 1C CpG5 was associated with increased waist circumference (r = 0·43, P = 0·035) and BMI (r = 0·45, P = 0·025), whereas methylation at GR exon 1F CpG1 was inversely associated with systolic and diastolic blood pressures (r = −0·40, P = 0·037; r = −0·37, P = 0·046 respectively). Increased H19 ICR methylation at most of the individual CpG sites was associated with increased weight, waist circumference, BMI, systolic and diastolic blood pressures. Principal components analysis demonstrated that increased methylation across the majority of the region tested (CpGs 1–4, and 6–12) was associated with increased weight, waist circumference, BMI, systolic and diastolic blood pressures (Table 3). All findings remained significant after adjustment for gender.

Figure 2.

Associations of percentage methylation at individual CpG sites of HSD2 Region 2 with systolic blood pressure. Figure shows the significant correlations (P < 0·05) of % methylation at individual CpG sites of HSD2 Region 2 with systolic blood pressure in 34 individuals. Table insert shows results of regression analysis adjusting for gender.

Table 3. Associations of methylation of HSD2 Region 1 and H19 IC1 with anthropometry and blood pressurea
 WeightWaistBMISystolic BPDiastolic BP
βPartial correlation coefficientP-valueβPartial correlation coefficientP-valueβPartial correlation coefficientP-valueβPartial correlation coefficientP-valueβPartial correlation coefficientP-value
  1. a

    All analyses adjusted for gender.

  2. b

    Methylation levels for HSD2 and H19 ICR were natural log-transformed prior to analysis. As methylation levels between individual CpG sites for each gene were correlated, principal components analysis was performed for each gene using Z-scores calculated from the natural log-transformed values. For HSD 2 Region 1, 2 independent factors were identified; Factor 1 included CpG sites 1–4 and 6, and Factor 2 included CpG 5 alone. Two independent factors were also identified for H19 ICR; Factor 1 included CpGs 1–4 and 6–12, and Factor 2 included CpG5 alone. ICR, imprinting control region.

HSD2 Region 1b
Factor 1 (CpGs 1–4,6)0·370·470·030·370·500·020·450·490·020·070·070·750·220·230·33
Factor 2 (CpG 5)0·350·450·040·450·570·0070·310·360·100·110·120·620·010·010·97
H19 ICRb
Factor 1 (CpGs 1–4, 6–12)0·370·430·030·420·470·020·410·420·030·540·620·0020·440·450·03
Factor 2 (CpG5)0·030·030·890·110·140·520·050·060·80−0·03−0·050·830·110·140·56

Exploratory analysis of methylation status of candidate genes with maternal diet

We examined associations between methylation and the extent to which the subject's mothers had followed the specific dietary advice to increase protein and reduce carbohydrate intake during pregnancy. In all analyses, we used methylation levels as the dependent variable and adjusted for gender, BMI and birthweight as potential confounders. Offspring whose mothers reported higher meat/fish and vegetable intake and lower bread/potato intake in late pregnancy had higher mean methylation at GR exon 1F (Fig. 3); methylation was also increased at a specific CpG sites in HSD2 Region 2 with increased meat (β = 0·41, partial correlation coefficient = 0·42, P = 0·03) and fish (β = 0·37, partial correlation coefficient = 0·40, P = 0·04) intake in late pregnancy.

Figure 3.

Associations of percentage mean methylation at glucocorticoid receptor (GR) in the offspring with maternal diet (portions of food per week) during late pregnancy. Significant correlations (P < 0·05) of percentage mean methylation at GR in the offspring with maternal diet according to the specific dietary advice to increase protein intake and reduce carbohydrate intake. Figure shows portions of food eaten per week during ‘late’ pregnancy (>20 weeks gestation). (a) meat portions/week, (b) fish portions/week, (c) vegetable portions/week, (d) bread portions/week, (e) potato portions/week.


We show that DNA methylation at key genes involved in tissue metabolism and action of glucocorticoids and in the regulation of foetal growth associates with the early-life environment, including maternal diet, and separately with adult cardiometabolic risk factors. Our findings that events in early life correlate with methylation at specific candidate loci some 40 years later add to the growing literature, suggesting a role for epigenetic alterations in the programming of cardiovascular disease risk in humans.

We have previously shown that exposure to an unbalanced diet in pregnancy programmes increased blood pressure[30] and cortisol responses to stress[31, 32] in this cohort. We found significant associations between methylation at the promoter of HSD2 and adult variables including blood pressure. Methylation at the HSD2 promoter decreases gene expression[20, 33] and associates with hypertension in humans[35] and animal models.[27] Thus, the data are compatible with a role for altered methylation at the HSD2 promoter and the risk of hypertension in adults. Changes in GR promoter methylation also associate with altered blood pressure and with adiposity in adulthood and additionally with maternal diet in pregnancy, notably increased protein and reduced carbohydrate intake. Studies in rats have reported differences in DNA methylation at the alternate promoters/first (untranslated) exons of GR including at (exon 17) in the brain in association with variations in maternal care[24] or prenatal stress[22] and decreased methylation at exon 110 in liver following prenatal protein restriction.[36] More recently, altered methylation at exon 1F (equivalent to exon 17) has been described in human brain in association with early-life adversity[37] and in cord blood from infants whose mothers were depressed/anxious during pregnancy.[12]

IGF2 and H19 are linked imprinted genes,[38] and altered methylation at the differentially methylated regions at this locus is implicated in a number of human growth disorders with even modestly aberrant imprinting at this locus resulting in marked effects on foetal growth and development.[39] In our study, methylation at the H19 ICR, but not other DMRs, was associated with obesity in adulthood. We found no relationship between birthweight and DNA methylation, consistent with findings in Dutch Hunger Winter offspring, in which altered IGF2 DMR2 methylation occurred following exposure to famine in early gestation without birthweight changes.[8] Consistent with recent reports in animal models and in human studies, alterations in DNA methylation at all three genes were region and CpG specific.[12, 24, 37, 40] Although the precise mechanisms that account for these alterations in methylation are unclear, DNA methylation depends on the dietary supply of methyl donors, so that altered nutrition during prenatal development when the epigenome is particularly responsive to environmental stimuli may result in permanent alterations in DNA methylation patterns.[41] The mechanisms by which these site-specific changes in DNA methylation are mediated are unclear, but at least for GR, such changes have been shown to impact on transcription factor binding to the GR promoter.[24]

Limitations of our study include (i) the specific dietary advice given to mothers precludes direct extrapolation to other populations and (ii) the study size was small. However, recent reports suggest that important differences in DNA methylation at candidate genes can be identified in such sample sizes.[37] Methylation levels at HSD2 and GR were low, in agreement with other studies and consistent with the presence of CpG islands within both promoters.[10, 12, 42-44] Nevertheless, such methylation levels may be functionally important.[45] In addition, although the use of multiple tests raises the possibility of false positive results, the detailed phenotyping of subjects allowed us to adjust for potential confounding factors, thus addressing such statistical concerns.

It is rather surprising that any relationship between maternal diet or birth parameters and methylation of leucocyte DNA is still present after 40 years. Whether these changes in DNA methylation are programmed at birth or occur in later life is not known; however, studies reporting altered DNA methylation at birth at HSD2[27] in rodent models and in human neonates[5, 6] in association with exposure to an adverse environment in utero suggest that such changes may indeed be present from early life. Additionally, methylation at IGF2 is reported to be stable for many decades.[8, 46] Alterations in gene expression and DNA methylation in animal models may be tissue specific,[47-49] and these effects may not therefore be causative of disease but may plausibly be cocaused by exposure to an adverse environment in utero. Differences in DNA methylation in leucocytes may represent a ‘shadow’ of programmed methylation changes in key (inaccessible) target organs and/or of more major changes earlier in life and raises the possibility that such changes may indeed prove to be useful biomarkers of disease risk.[17]

In conclusion, we found separate alterations in DNA methylation in peripheral blood in association with both early-life parameters and with adult cardiometabolic risk factors, suggesting that programming of disease susceptibility may depend on epigenetic modifications at specific loci. Additionally, specific components of maternal diet can impact on DNA methylation and disease risk later in life. Further detailed longitudinal studies are needed to confirm whether such changes are identifiable at birth and hence may represent useful biomarkers of risk.


This work was supported by a grant from the Chief Scientist Office, Scotland. We thank Vanessa Cox for computing and administrative support and Clive Osmond for statistical advice. KMG is supported by the National Institute for Health Research through the NIHR Southampton Biomedical Research Centre. We thank the men and women who participated in the study.