Introduction
Low birth weight is associated with an increased risk of chronic disease in later life. Studies around the world have shown that people who were born with low birth weight have increased rates of type 2 diabetes, hypertension and cardiovascular disease at adult age.Reference Huxley, Owen and Whincup1–Reference Whincup, Kaye and Owen3 These associations are thought to reflect the process by which fetal responses to limited food supply induce permanent alterations in the offspring's physiology and metabolism, optimizing its chances of survival in the short term in a poor nutritional environment but possibly being detrimental in later life, especially if food is abundant. The first direct evidence in humans that such undernutrition during gestation increases the risk of chronic non-communicable disease has come from the Dutch famine birth cohort study. People whose mothers had been exposed to this 5-month period of acute food shortage in early pregnancy had impaired glucose regulation,Reference de Rooij, Painter and Roseboom4, Reference Ravelli, van der Meulen and Michels5 a more atherogenic lipid profileReference Roseboom, van der Meulen and Osmond6 and increased rates of cardiovascular diseaseReference Painter, de Rooij and Bossuyt7, Reference Roseboom, van der Meulen and Osmond8 and women in this cohort had increased rates of breast cancer in adult life.Reference Painter, de Rooij and Bossuyt9 The effect size was striking; those exposed to famine in early gestation had a doubled rate of cardiovascular disease compared with those who had not been exposed to famine prenatally.Reference Painter, de Rooij and Bossuyt7
One of the mechanisms that may play a role in inducing such effects is epigenetic regulation of gene expression. Animal studies have shown that reduced food intake during pregnancy leads to changes in the epigenetic regulation of the expression of transcription factors that play a key role in regulating glucose and lipid metabolism in the offspring. For instance, feeding a protein-restricted diet to rats during pregnancy induces hypomethylation of the PPARα (peroxisome proliferator-activated receptor alpha) and GR (glucocorticoid receptor) promoters, increased expression of the GR and PPARα and their target genes and an increase in the metabolic processes that they control, namely β-oxidation and gluconeogenesis in the liver of the offspring.Reference Burdge, Slater-Jefferies and Torrens10, Reference Lillycrop, Phillips and Jackson11
The first evidence that undernutrition during gestation alters epigenetic regulation in humans has come from Heijmans et al.Reference Heijmans, Tobi and Stein12 who have shown that prenatal exposure to the Dutch famine is associated with decreased methylation levels of the differentially methylated region of the imprinted insulin-like growth factor-2 gene (IGF2 DMR) in peripheral blood. The mean level of methylation of exposed individuals was 49% compared with 52% in unexposed sibling controls.Reference Heijmans, Tobi and Stein12 The functional implication of this small, but statistically significant difference is unclear. Analysis of 15 additional candidate genes revealed that methylation of six of these loci was associated with prenatal exposure to famine, of which three were sex specific.Reference Tobi, Lumey and Talens13 The authors suggested that these epigenetic changes could provide a mechanism by which the prenatal nutrition affects risk of later chronic diseases. The authors, however, did not report whether such variations in methylation were associated with any phenotypic characteristics. In addition, there is an increasing body of evidence to suggest that adult lifestyle factors can also affect the epigenome.Reference Mathers, Strathdee and Relton14 We have previously shown that prenatal famine exposure affects lifestyle choices,Reference Lussana, Painter and Ocke15 which might also contribute to disease risk. Genomic imprinting is a phenomenon whereby a gene is expressed in a parent-origin manner, with one active and one silent allele. Many methylation studies have been limited to imprinted genes, but also non-imprinted genes can been affected by epigenetic programming effects.
We therefore selected four non-imprinted candidate genes with relevance to cardiovascular and metabolic functions. PPARγ is involved in insulin and lipid metabolism.Reference Auwerx16 GR 1-C regulates the stress response, developmental processes, immune responses and glucose metabolism. The p85α regulatory subunit of phosphatidylinositol 3 kinase (PI3kinase) is involved in placental development and fetal growthReference Forbes and Westwood17 and is a key component in the insulin-signaling pathway. Lipoprotein lipase (LPL) hydrolyzes lipids into lipoproteins. An LPL deficiency leads to hypertriglyceridemia.Reference Mead, Irvine and Ramji18 The expression of all four of these genes has been shown to be persistently altered by maternal dietary restriction in animal models of maternal programming.Reference Lillycrop, Slater-Jefferies and Hanson19–Reference Burdge, Slater-Jeffries and Grant22 First, we investigated whether methylation status of their promoter regions differed between individuals exposed to famine at different periods of gestation compared with unexposed individuals. Second, we investigated whether methylation status of these four genes was associated with markers of metabolic and cardiovascular disease or adult lifestyle factors, and thus whether they could explain the increased rates of disease in people who were prenatally exposed to the Dutch famine.
Methods
Selection procedures
The Dutch famine birth cohort consists of 2414 men and women born as term singletons in the Wilhelmina Gasthuis in Amsterdam between 1 November 1943 and 28 February 1947. The selection procedure and subsequent loss to follow-up have been described in detail elsewhere.Reference Ravelli, van der Meulen and Michels5, Reference Painter, de Rooij and Hutten23 At age 58, 1423 of the 2414 original cohort members (58%) were still alive, living in the Netherlands at a known address. The study was approved by the local Medical Ethics Committee and carried out in accordance with the Declaration of Helsinki. All participants gave written informed consent.
Exposure to famine
Exposure to famine was defined according to the official daily rations for the general population older than 21 years. The official rations accurately reflect the variation over time in the total amount of food available in the west of the Netherlands.Reference Trienekens24 An individual was considered to be prenatally exposed to famine if the average daily food ration of the mother during any 13-week period of gestation contained less than 1000 calories. On the basis of this definition, babies born between 7 January 1945 and 8 December 1945 were exposed in utero. We used periods of 16 weeks each to differentiate between people who had been exposed in late gestation (born between 7 January 1945 and 28 April 1945), in mid gestation (born between 29 April 1945 and 18 August 1945), and in early gestation (born between 19 August 1945 and 8 December 1945). People born before 7 January 1945 (and were thus born before the famine) and those born after 8 December 1945 (and who had thus been conceived after the famine) were considered unexposed.
Study parameters
The medical birth records provided information about the mother, the course of the pregnancy and the body size of the baby at birth.Reference Ravelli, van der Meulen and Michels5 For the adult measurements of the participants, we measured adult height using a portable stadiometer and weight with a portable Tefal scale. Blood pressure was measured in duplo on two occasions (morning and afternoon) using an automated device (Omron 705 CP/IT; Omron Healthcare UK, West Sussex, UK) and appropriate cuff sizes. Mean systolic and diastolic blood pressure was calculated using all available measurements.
We performed an oral glucose tolerance test (OGTT) after an overnight fast with a standard load of 75 g. Participants with pre-existing diabetes (defined as taking glucose-lowering medication) were excluded from the OGTT. Plasma glucose concentrations were measured by standardized enzymatic photometric assay on a Modular P analyzer (Roche, Basel, Switzerland) and plasma insulin concentration by immunoluminometric assay on Immulite 2000 analyzer (Diagnostic Product Corporation, Los Angeles, USA). Type 2 diabetes was defined as a 2 h glucose level of >11.0 mmol/l (in accordance with our previous studies and the 1999 World Health Organization recommendationsReference de Rooij, Painter and Roseboom4, Reference Ravelli, van der Meulen and Michels5, Reference de Rooij, Painter and Philips25, 26) or taking anti-diabetic medication.
A fasting blood sample was drawn for analysis of low-density lipoprotein (LDL)-cholesterol, high-density lipoprotein (HDL)-cholesterol and triacylglycerol. HDL-cholesterol and triacylglycerol were measured using an enzymatic colorimetric agent (Roche) on a P-800 Modular (Roche). LDL-cholesterol was calculated using the Friedewald formula. Standard 12-lead electrocardiograms (ECG) were made of all participants. Coronary heart disease was defined as the presence of one or more of the following: angina pectoris according to the Rose/WHO questionnaire;Reference Rose27 Q waves on the ECG (Minnesota codes 1–1 or 1–2) or a history of coronary revascularization (angioplasty or bypass surgery).
We performed B-mode ultrasound examinations of the arterial walls of the common, bulb and internal carotid artery segments. Details of the measurements are described elsewhere.Reference Painter, de Rooij and Hutten23 Mean carotid intima media thickness (IMT) was defined as the mean IMT in mm of the right and left common artery, common bulb and the internal carotid far wall segments. If either the right or left value was missing for any given carotid segment, the remaining available segment was used to calculate the mean carotid IMT.
A Dutch translation of the HADS (Hospital Anxiety and Depression Scale) was administered to all participants to measure subclinical depression and anxiety symptoms.Reference Zigmond and Snaith28 The HADS consists of two subscales: a depression subscale (HADS-D, seven items, score range 0–21, Cronbach's α 0.80) and an anxiety subscale (HADS-A, seven items, score range 0–21, Cronbach's α 0.82). Cronbach's α was 0.88 for all 14 HADS items, indicating good internal consistency.
Information on socioeconomic status, medical history and lifestyle (smoking, exercise) and the use of medication was retrieved from a standardized interview. Current socioeconomic status was coded according to International Socio-Economic Index 92, which is a numeric scale based on the person's or his or her partner's occupation, whichever was higher.Reference Bakker and Sieben29
Methylation-sensitive polymerase chain reaction (PCR)
DNA was extracted from a fasting blood sample and stored at 4°C. To ensure maximum comparability with rodent models of maternal undernutrition and epigenetic changes in the offspring, we assessed methylation status by applying methylation-sensitive PCR.Reference Lillycrop, Phillips and Jackson11
For analysis of promoter methylation, genomic DNA (400 ng) was incubated with the methylation sensitive restriction endonucleases AciI and HinfI as instructed by the manufacturer (New England Biolabs, Hitchin, Hertfordshire, UK). The resulting DNA was amplified using real time PCR, which was performed in a total volume of 25 μl with SYBR® Green Jumpstart Ready Mix (Sigma) as described by the manufacturer. A fragment of the human PPARα exon 7 that does not contain AciI or HinfI cleavage sites was used as an internal control gene. Primers were designed to amplify regions containing HinfI and/or AciI cutting sites in the proximal promoters of LPL (chr8: 19,796,366-19,796,515) PI3kinase (chr5: 67,521,933-67,522,282), PPARγ (chr3: 12,392,392-12,392,591) and the CpG island spanning the GR 1-C promoter (chr5: 142,782,821-142,783,152; see Lillycrop et al.Reference Lillycrop, Slater-Jefferies and Hanson19 for an overview). Cycle parameters were 94°C for 2 min then, 40 cycles of 95°C for 30 s, 60°C (GR1-C, PPARα) or 55°C (PPARγ, LPL) or 65°C (PI3 kinase p85) for 1 min and 72°C for 1 min. Primer sequences are shown in Table 1. All Ct values were normalized to the internal control and each sample analyzed in duplicate. Methylation was expressed as percentage methylation compared with the control gene.
Table 1 Primer sequences (5′–3′) used in the measurement of promoter methylation levels by methylation-sensitive PCR
PCR, polymerase chain reaction; PPAR, peroxisome proliferator-activated receptor; GR, glucocorticoid receptor; LPL, lipoprotein lipase; PI3 kinase, phosphatidylinositol 3 kinase.
Statistical methods
Distributions of methylation status of the promoter regions for the four candidate genes were highly skewed and we used logarithmic transformations to better approach normality. We used linear and logistic regression analysis to compare maternal, birth and adult outcomes and methylation levels of those exposed in late, mid or early gestation and those unexposed to famine during gestation. Adjustments were made for maternal age, sex and parity. In investigating possible associations between methylation status and markers of disease, we report regression coefficients and P-values of linear and logistic regression analyses using log-transformed methylation values. We considered effects to be statistically significant if P-values were ⩽0.05. SPSS 16.0 (SPSS Inc., Chicago, IL, USA) was used for all statistical analyses.
Results
Characteristics of the study population
DNA was available from 759 participants. Their mean age was 58 years (s.d. 1); 349 (46%) were men. A total of 319 (42%) had been exposed to famine in utero, and 440 (58%) had not been exposed to famine prenatally. Table 2 shows that mothers exposed to famine in late gestation were older and less often primiparous than mothers who had not been exposed to famine in pregnancy. Babies exposed to famine in late- and mid-gestation had lower birth weights than unexposed babies. At age 58, there were no significant differences between exposed and unexposed groups in smoking habits and socio-economic status. Relative methylation levels of the four candidate genes are shown in Table 3.
Table 2 Maternal, birth and adult characteristics according to timing of prenatal exposure to the Dutch famine
BMI, body mass index; SES, socio-economic status.
Data are means ± s.d., except where given as numbers and percentages.
*P < 0.05 compared with people unexposed to famine in utero.
Table 3 Relative methylation according to timing of prenatal exposure to the Dutch famine
GR, glucocorticoid receptor; LPL, lipoprotein lipase; PI3 kinase, phosphatidylinositol 3 kinase; PPARγ, peroxisome proliferator-activated receptor gamma.
aGeometric mean.
Methylation and birth and maternal characteristics
Birth weight was positively associated with GR methylation but not with the methylation status of the other three genes. A 1 kg increase in birth weight was associated with 22% (95% CI 4–43) increase in GR methylation (P = 0.02). There was no association between methylation of the four genes and gender, maternal age or parity.
Methylation after exposure to famine in utero
Crude and adjusted methylation differences between subjects exposed in late-, mid- or early-gestation and those unexposed to famine are shown in Table 4. Exposure to famine during gestation was not associated with altered methylation status of the GR, LPL, PPARγ and PI3kinase promoter. This also applied after adjustment for sex, maternal age or parity. The associations did not differ between men and women (P-value for all interaction terms >0.05).
Table 4 Methylation differences (%) and 95% CIs compared with unexposed subjects using a linear regression model
GR, glucocorticoid receptor; LPL, lipoprotein lipase; PI3 kinase, phosphatidylinositol 3 kinase; PPAR, peroxisome proliferator-activated receptor.
aAdjusted for sex, maternal age and parity.
Methylation and markers of disease
We tested for associations between DNA methylation status and markers of cardiovascular and metabolic disease as well as lifestyle factors predisposing for cardiovascular disease. Regression coefficients of these markers and lifestyle factors according to methylation levels are shown in Table 5. The regression coefficients depict the size and direction of the association between the disease markers and lifestyle factors and the methylation levels. Increased GR-1C promoter methylation status was associated with lower score on HADS subscale anxiety and depression, lower self perceived health and lower body mass index (BMI), higher levels of physical activity and non-smoking. Increased methylation status of PPARγ promoter was associated with lower plasma triglyceride level, higher HDL levels, higher levels of physical activity and non-smoking.
Table 5 Health outcomes according to methylation status
GR, glucocorticoid receptor; LPL, lipoprotein lipase; PI3 kinase, phosphatidylinositol 3 kinase; PPAR, peroxisome proliferator-activated receptor; CHD, coronary heart disease; HDL, high-density lipoprotein; LDL, low-density lipoprotein; RRsys, systolic blood pressure; RRdia, diastolic blood pressure; IMT, intima-media thickness; HADS, Hospital Anxiety and Depression Scale; BMI, body mass index.
Health outcomes according to methylation status. β: regression coefficients from linear or logistic regression models in which methylation values were log transformed. Regression coefficients depict the size and direction of the association between health and lifestyle factors and methylation levels of the four different candidate genes.
Significant associations are depicted in bold print.
aValues were log transformed.
bMeasurements at baseline and 120 min.
Discussion
In the present study, we examined the epigenetic state of non-imprinted genes in peripheral blood of adults exposed to famine during different periods of perinatal development in comparison with unexposed controls. We did not observe differences in methylation levels of the promoters of our chosen candidate genes between men and women who had been exposed to famine in either late-, mid- or early-gestation and those who had not been exposed. We did find that increased methylation status of PPARγ promoter was associated with lower plasma triglyceride level, higher HDL levels and higher levels of exercise and non-smoking. Hypermethylation of the GR promoter was associated with lower score on HADS subscale anxiety and depression, lower self-perceived health, lower BMI and higher levels of exercise and non-smoking.
These data contrast with those we previously reportedReference Burdge, Slater-Jefferies and Torrens10, Reference Lillycrop, Phillips and Jackson11 in young adult rat offspring born to dams exposed to a similar level of unbalanced nutrition throughout the whole of pregnancy. There are several possible reasons for this difference.
First, the lack of effects of prenatal nutritional environment on methylation of the candidate genes may be due to the variable levels of postnatal diet and lifestyle factors that humans experience. In comparison with the rat offspring that were fed a standardized diet, such dietary variations might induce a variable degree of pre- and postnatal mismatch and thus of epigenotype. Such masking of prenatal epigenetic changes, if present, may be greater in older individuals, as was the case for the adults studied here as compared with previous rat studies. Support for this concept comes from studies of monozygotic twinsReference Fraga, Ballestar and Paz30 in which changes in the epigenome occurred with age. Our finding that lifestyle factors in adults were associated with the methylation of the promoters of GR1-C and PPARγ is in agreement with this.
A second consideration is that epigenetic processes are known to be tissue specific.Reference Byun, Siegmund and Pan31 Although in the present study we only had access to DNA isolated from peripheral white blood cells, studies in rodents used target tissues such as liver.Reference Lillycrop, Phillips and Jackson11 In newborn twin pairs, methylation level of four DMRs associated with the IGF2/H19 locus varied in a tissue-specific manner in five different tissues representing the different germ layers.Reference Ollikainen, Smith and Joo32 However, other studies may suggest that DNA methylation patterns are similar between tissues.Reference Byun, Siegmund and Pan31 It may be that if the environmental constraint occurred early on in development, gene methylation may be altered in three germ layers and an imprint of this altered epigenetic mark will be detectable in all tissues. Methylation patterns in blood and perhaps in other readily available tissue may therefore provide useful proxy markers of methylation in more metabolically relevant tissues. Indeed, Heijmans et al.Reference Heijmans, Tobi and Stein12 did find changes in DNA methylation in peripheral white blood cells, although the changes were small, occurring in imprinted genes, and their functional significance is not known. More recently, Godfrey et al.Reference Godfrey, Sheppard and Gluckman33 has reported in two independent cohorts the methylation status of a single CpG site in the promoter region of the transcription factor RXRA in umbilical cord was strongly related to childhood adiposity in both boys and girls.
In the present study we used methylation-sensitive PCR to measure methylation across the entire proximal promoter region of our candidate genes. This method would not allow us to detect changes at the level of individual CpGs. Nevertheless, this is precisely the technique we have previously used extensively in rodentsReference Burdge, Slater-Jefferies and Torrens10, Reference Lillycrop, Phillips and Jackson11, Reference Lillycrop, Slater-Jefferies and Hanson19, Reference Burdge, Lillycrop and Phillips34, Reference Lillycrop, Phillips and Torrens35 and in this study we did find associations between the methylation status of GR1-C and PPARγ with either postnatal metabolic outcomes and/or postnatal lifestyle factors.
In addition, the small number of people exposed to famine in early-gestation (n = 73) – the group in whom we previously found a doubled rate of cardiovascular disease in adult lifeReference Painter, de Rooij and Bossuyt7 – may have limited our ability to detect differences. However, the number of subjects in the Hungerwinter families study was lower and did not prevent them from finding a highly significant difference in methylation levels between subjects exposed to famine in early-gestation and unexposed subjects.Reference Heijmans, Tobi and Stein12 Therefore, we do not consider our study underpowered to detect famine effects on methylation.
In rats, the hepatic GR110 promoter, which shows 70.6% homology with the human GR1-C promoterReference Turner, Pelascini and Macedo36 is hypomethylated after prenatal low protein diet.Reference Lillycrop, Slater-Jefferies and Hanson19 Hypomethylation of the promoter is associated with an 84% higher mRNA expression;Reference Lillycrop, Slater-Jefferies and Hanson19 therefore, GR1-C was a likely candidate gene in our study. Interestingly, in both our current study as well as in the previously mentioned Hungerwinter families study, differences in GR methylation were not found after prenatal exposure to famine.Reference Tobi, Lumey and Talens13 Unfortunately, mRNA was not available from the participants of this study.
Although this was not our primary research focus, and there is always a possibility of type 1 errors, we found an association between beneficial characteristics such as lower levels of BMI, non-smoking and physical activity in adult life – which tend to cluster – and increased methylation status of the GR1-C promoter. Smoking – which was most strongly associated with GR methylation – is a well-known lifestyle factor influencing epigenetic patterns, both in animal models and in humans.Reference Mathers, Strathdee and Relton14 For instance, exposure of human lung cancer cells to cigarette smoke resulted in demethylation of the promoter of the prometastatic oncogene synuclein-gamma.Reference Liu, Zhou and Boggs37 In DNA from peripheral blood, cigarette smoking has been linked to hypomethylation of genes involved in platelet activationReference Breitling, Yang and Korn38 and serotonin regulation,Reference Launay, Del and Chironi39 findings that fit with the idea that postnatal influences also alter epigenetic patterns. Whether our finding of smoking and increased methylation status of the GR1-C promoter has a clinical significance remains to be investigated. We now know from our data that GR methylation is associated with adult stress responsiveness, although these associations were largely explained by differences in lifestyle and education.Reference de Rooij, Costello and Veenendaal40 Our current analyses suggest that GR hypomethylation is also associated with higher levels of depression and a negative health perception. Recent studies have reported associations between methylation status at birth and body size in later childhood, suggesting a role for epigenetic factors as mediators for early life programming of disease in later life.Reference Godfrey, Sheppard and Gluckman33, Reference Relton, Groom and St. Pourcain41
On the basis of the results of this study, we can conclude that the increased rates of chronic degenerative disease found after famine exposure are not mediated by differences in methylation status of four genes of strong candidacy based on animal and other studies. This argues for further studies, in particular to address the interaction between pre- and postnatal nutritional and other environmental influences, and to explore the mechanistic pathways underlying the association between prenatal famine and later markers of cardiovascular and metabolic disease.
Acknowledgment
This work was supported by the Netherlands Heart Foundation (Grant number 2007B083).
The authors' responsibilities were as follows: R.C.P., K.A.L., M.A.H. and T.J.R. designed the study. M.V.V. and P.M.C. conducted the study. M.V.V. performed the statistical analyses and wrote the manuscript. All authors contributed to the revisions of the manuscript and read and approved the final version.
