Introduction
The periconception, prenatal and early postnatal period are of significant importance in the development and growth of the embryo, foetus and child. Epidemiologic studies have shown numerous associations between adverse periconception exposures, such as smoking and foetal growth restriction.Reference Villalbi, Salvador, Cano-Serral, Rodriquez-Sanz and Borrell 1 Next to these evident short-term consequences there have been reports on long-term health consequences of prenatal exposure to smoking for the offspring including but not limited to respiratory illness, type 2 diabetes and cardiovascular disease.Reference Doherty, Grabowski, Hoffman, Ng and Zelikoff 2
Epigenetics is suggested to be an important mechanism linking periconceptional, prenatal and postnatal exposures in association with health and disease risks in later life. It is described as a biological mechanism to explain gene–environment interactions, which allows (heritable) changes in gene expression to occur without changing the DNA sequence.Reference Jirtle and Skinner 3 , Reference Jaenisch and Bird 4 DNA methylation is one of the best-understood epigenetic marks, in which the differentiation of cells and tissue is programmed. Furthermore, it is hypothesized that changes induced on this mark by past intrauterine exposures may be a mechanism to adapt a priori to future environment exposures.Reference Barker 5
Of major importance in both prenatal and postnatal growth is the insulin pathway in which the insulin growth factor signalling system plays an important role. It comprises products of the maternally imprinted insulin growth factor 2 receptor (IGF2R), which in general inhibits growth, and the paternally imprinted insulin growth factor 2 (IGF2) which enhances growth.Reference Baker, Liu, Robertson and Efstratiadis 6 Polymorphisms in IGF2R and IGF2 DMR (IGF2 differentially methylated region) are associated with birth weight.Reference Adkins, Somes and Morrison 7 Moreover, INSIGF, the overlapping region of IGF2 and insulin (INS), has been associated with small-size-for-gestational-age (SGA) at birth.Reference Adkins, Krushkal and Klauser 8 Of interest is that the methylation of IGF2 DMR and INSIGF have been shown to be related to exposure to famine in utero.Reference Tobi, Lumey and Talens 9 Although prenatal famine shares similarities with SGA at birth regarding health problems in later life, this similarity was not observed on the level of DNA methylation of IGF2 DMR and INSIGF.Reference Godfrey and Barker 10 – Reference Tobi, Heijmans and Kremer 12
Evidence is increasing that maternal smoking during pregnancy is associated with differences in DNA methylation both globally and gene specific, both increases and decreases of DNA methylation have been reported.Reference Breton, Byun and Wenten 13 – Reference Toledo-Rodriguez, Lotfipour and Leonard 16 A recent genome-wide methylation analysis has shown significantly differential methylation among the placental genome of smokers, and a significant reduction in the child's birth weight.Reference Suter, MA and Harris 17 None of the before mentioned studies examine dose–response relationships or looked specifically at periconception smoking. From this background, we studied associations between periconception maternal smoking and DNA methylation of IGF2 DMR, IGF2R and INSIGF genes in children at the age of 17 months. Second, socioeconomic status (SES), as measured by its proxy education, may be seen as an additional and cumulative determinant of harmful exposures and may therefore add additional insight. In our study we used low education defined by a maximum of 12 years of on-going education from the age of 4, as a proxy for low socioeconomic status (SES) and considered it as a cumulative determinant and the strongest marker of SES.Reference Desai and Alva 18 Recently an exploratory paper showed widespread differences in DNA methylation on a genome wide level associated with SES.Reference Borghol, Suderman and McArdle 19 Therefore, we investigated and included additional analyses on the influence of SES on these loci and the relation between DNA methylation and smoking.
Method
Study sample
We examined 120 children (boys n = 70, girls n = 50) at a mean age of 17 months (s.d. 2.5) and their mothers. These subjects served as controls in the HAVEN study, they were previously described in detail.Reference Steegers-Theunissen, Obermann-Borst and Kremer 20 Mothers and their children were recruited from the public child health care centres of ‘Thuiszorg Nieuwe Waterweg Noord’ currently known as ‘Careyn’ in the Rotterdam (The Netherlands) area. Public child health care centres are part of the Dutch Health Care system where physicians specialized in child health care regularly check all new-borns at standardized moments on health, growth and development. Children were eligible as controls if they did not have a major congenital malformation or chromosomal abnormality according to the medical records from the regular check up at the child health centre up to the study moment. The total number of responding mothers and their children to serve as controls in the HAVEN study between October 2003 and December 2009 and willing to be contacted was 794 of which 490 (61.7%) were included in the study. Reasons for exclusion were: mother pregnant or lactating; the mother not being the birth mother; the child having a congenital defect or chromosomal abnormality; not willing to have blood drawn. At the standardized study moment of 17 months of age, we retrospectively collected periconception, prenatal and postnatal data of the children and their mothers by self-administered questionnaires sent before the hospital visit, which were checked by the researcher for completeness and consistency at the hospital. We extracted the following data: maternal education level, periconception risk factors, for example, folic acid supplement use, smoking, the number of cigarettes smoked, child's birth weight and health status. The latter was divided in the following categories: renal disease; liver disease; disease of the gastro-intestinal tract; thyroid disease; cardiovascular disease; epilepsy; thrombosis; malignancy; disease of the urinary tract; malaria; dermatologic disease; allergy; haematological disease; disease of the respiratory tract; or other.
Gestational age (GA) was based on the first day of the last menstrual period or calculated from the first trimester foetal ultrasound scan. Because periconception folic acid use was previously associated with IGF2 DMR methylation, we include folic acid use in the analysis as potential effect modifier.Reference Steegers-Theunissen, Obermann-Borst and Kremer 20 We included intake of folic acid according to the Dutch recommendation of a daily intake of a folic acid containing preparation of 400 μg from at least 4 weeks before until 8 weeks after conception. We used educational level as a proxy for SES and considered it as a cumulative determinant and the strongest marker of SES, dichotomized in two categories: low education defined by a maximum of 12 years of on-going education from the age of 4 (primary/lower vocational/intermediate secondary education), and other (higher secondary/intermediate vocational education or higher vocational/university education).Reference Desai and Alva 18 We additionally derived information on weight and length from the child's record at the public child health care centres of Careyn. We calculated body mass index (BMI) by weight in kilograms divided by squared height in centimetres. Of every child with available growth data from birth onwards, we calculated the growth rate for which the length was plotted against the square of the root of the age. The slope of each individual regression line (β) was used as variable of the postnatal growth rate. The study protocol was approved by the Central Committee for Human Research (CCMO) in The Hague, The Netherlands, and the Medical Ethical and Institutional Review Board of the Erasmus MC, University Medical Centre in Rotterdam, The Netherlands. All mothers gave their written informed consent and mothers and their partner on behalf of their participating child.
DNA methylation measurements
The selection of 120 mother–child pairs is described in our previous study.Reference Steegers-Theunissen, Obermann-Borst and Kremer 20 In the present study we explored DNA methylation of IGF2 DMR, IGF2R and INSIGF measured in 120 children. Genomic DNA was isolated from whole blood using the salting out method. One-half microgram of genomic DNA was bisulfite-treated using the EZ 96-DNA methylation kit (Zymo Research) on one of two 96-well plates. Bisulfite-converted DNA-specific polymerase chain reaction (PCR) primers were used to amplify the investigated region. DNA methylation of the CpG dinucleotides was measured in triplicate using by a mass spectrometry-based method (Epityper, Sequenom). The quantitative nature, accuracy and reproducibility of this method has been shown extensively.Reference Heijmans, Tobi and Stein 11 , Reference Coolen, Statham, Gardiner-Garden and Clark 21 The measurements of IGF2 DMR, IGF2R and INSIGF are part of an on-going study of a total of eight loci including IL10, TNF, LEP, KCNQ1OT1 and FTO. Details of the measured amplicons, including details of functional relevance and PCR primers were published before.Reference Tobi, Lumey and Talens 9 , Reference Talens, Boomsma and Tobi 22 In short, the region analysed for the IGF2 DMR includes five CpG dinucleotides (chr11:2126035-2126372, in NCBI build 36.1) separated in four CpG units, in other words, fragments of DNA, because of two adjacent CpG's that could not be resolved individually. The region for IGF2R includes 10 CpG dinucleotides (chr6:160346346-160346595) resulting in four CpG units measured, because of adjacent CpG sites that could not be resolved individually and CpG's containing fragments of too little or high mass for the mass spectrometer to resolve. The region for INSIGF, located in the promoter of the imprinted INSIGF transcript originating from the INS promoter, includes six CpG sites in total, of which four could be resolved and all individually (chr11:2138912-2139216).
Statistical analysis
We applied linear mixed models on the raw data without imputation of missing values to calculate exposure-specific differences and associations.Reference Tobi, Lumey and Talens 9 All the analyses accounted for bisulfite plate and the correlation between the CpG dinucleotides. The CpG sites of the studied locus were simultaneously entered with variables under study as fixed effects, with overall methylation as outcome. The linear mixed model was chosen over a standard paired t-test, because it allows for the analysis of all individual CpG dinucleotides of the locus in one test, accounts for the correlation between adjacent CpG dinucleotides, includes relevant adjustments within the model on the raw data and uses all available data. Absolute changes in DNA methylation are presented as regression coefficient with standard error (s.e.). Relative changes in percentage of DNA methylation were calculated by dividing the mean methylation with risk factor by mean methylation without risk factor. The association between the CpG sites and variables under study were studied separately with t-tests. Before testing the independent association of birth weight with methylation, Z-scores were calculated so that the resulting estimated effect size indicates the methylation change per standard deviation (s.d.) change in birth weight. Z-score does not affect a variable otherwise than standardizing the mean to zero and the s.d. to one and assists in interpreting the results. All P-values reported are two-sided. IBM SPSS Statistics 19.0 software was used for all analyses.
Results
Characteristics of the mother and child
In this study, we included 120 mother and child pairs (Table 1). Thirty-two mothers (28.3%) smoked during the periconception period and 31 (25.8%) had a low education. Education level was not significantly different between periconception smokers and non-smokers (P = 0.814). Maternal smoking was independently and inversely associated with birth weight (−231 g; P = 0.021). Of the periconception smokers, 19 (59.3%) continued smoking during pregnancy and 21 (65.6%) were smokers at the study moment. In the children, disease of the respiratory tract was present in 10 (8.3%) who all used inhalation medication for asthma, dermatologic disease was present in three (2.5%), allergies in eight (6.7%) and ear infection was present in one child (0.8%).
Table 1 Characteristics of mothers and children
BMI, body mass index; IGF2, insulin growth factor 2; IGF2R, insulin growth factor 2 receptor; INSIGF, the overlapping region of IGF2 and insulin.
Data are presented in median (p25–p75) or number (percentage).
DNA methylation
The average DNA methylation percentages were 48.9%, 55.3% and 89.4% for IGF2 DMR, IGF2R and INSIGF, respectively. Methylation of the separate CpG sites is listed in Table 1.
DNA methylation, periconception exposures and child characteristics
We tested for an association in all measured individuals between DNA methylation, maternal low education, smoking, no use of a folic acid supplement, gender, birth weight, growth rate, BMI and having a disease of the respiratory tract such as asthma (Table 2). Periconception low education and smoking were independently associated with a higher INSIGF methylation of +1.6% (P = 0.021) and +1.3% (P = 0.043), respectively. Both factors showed an additive effect of +2.8% (P = 0.011) on DNA methylation. All four CpG sites of INSIGF showed higher methylation, tested individually for educational level this was significant for CpG #2 (P = 0.034), #5 (P = 0.042), #6 (P = 0.003) and for smoking for CpG #5 (P = 0.012). The number of cigarettes smoked was associated with an increase (P = 0.028) in overall methylation percentage of INSIGF. For no smoking the mean methylation percentage (s.e.) was 89.0 (0.3), for 1 to 10 cigarettes 89.9 (0.6) and for 10 to 25 cigarettes 90.8 (0.9).
Table 2 Relation between DNA methylation, periconception risk factors and child characteristics
BMI, body mass index; IGF2, insulin growth factor 2; IGF2R, insulin growth factor 2 receptor; INSIGF, the overlapping region of IGF2 and insulin.
aThe regression coefficient [β (s.e.)] from a linear mixed model adjusted for the correlation between individual CpG dinucleotides and bisulfite batch. The investigated variable was entered as a fixed effect with overall methylation as outcome.
bThe percentage relative change (s.e.) in methylation of the locus.
cA two-sided P-value resulting from the linear mixed model.
dAdditionally adjusted for smoking P = 0.023.
eInteraction term: low education × smoking, P = 0.911.
fAdditionally adjusted for low education P = 0.047.
gAdjusted for birth weight P = 0.007.
hGender was entered as a fixed effect.
iOne s.d. increase in birth weight (584 g), adjusted for gestational age.
jAdjusted for periconception folic acid use.
kThe slope (β) of each individual regression line of length plotted against the square of the root of the age was used.
lOne s.d. increase in BMI (1.3 point).
Maternal smoking and low education were not associated with overall methylation of IGF2 DMR and IGF2R, this was similar for the individual CpG sites of IGF2 DMR and IGF2R, except for low education and IGF2R CpG #20 and 21 (P = 0.031) and borderline significant for IGF2R CpG #8–10 (P = 0.050).
We previously reported in the same children the association between periconception folic acid use and birth weight with IGF2 DMR methylation,Reference Steegers-Theunissen, Obermann-Borst and Kremer 20 and repeated here this analysis for IGF2R and INSIGF. There was no association with overall methylation. Only methylation of INSIGF CpG #5 was inversely associated with birth weight adjusted for GA (P = 0.035). For none of the loci there was an association with gender, or significant additive or interaction effect of gender and smoking on DNA methylation. Growth, BMI and asthmatic condition of the child revealed no association with DNA methylation of the three loci.
Discussion
In this study, we show associations of DNA methylation of the INSIGF gene in very young children with exposure to periconception maternal smoking and low education. Both exposures were associated with a higher methylation of INSIGF. Methylation of INSIGF CpG #5 was positively associated with periconception smoking and inversely with birth weight.
Based on our findings we hypothesize that adverse effects of the periconception environment, that is, factors related to low education, and smoking on birth weight could be related in part to silencing of the imprinted INSIGF gene in the child by slightly higher DNA methylation. Smoking is considered one of the major mediators in the association between low SES and adverse pregnancy outcome, and is also in our study highly associated with a decreased birth weight.Reference Villalbi, Salvador, Cano-Serral, Rodriquez-Sanz and Borrell 1 However, it remains important when studying outcomes related to low education and low SES, to analyse the relationship with smoking separately, since low education and smoking both had an independent association with increased methylation of INSIGF. This may suggest that there are other factors that affect the methylation of INSIGF, which are related to low education such as bad nutrition and stress. Our finding that low SES is significantly associated with higher methylation of INSIGF may be a signature of the early life environment, similar to a recent finding in adult males that revealed a distinct genomic methylation pattern related to their childhood SES.Reference Borghol, Suderman and McArdle 19
A recent study examined the association between maternal smoking and IGF2 DMR methylation in cord blood after birth, and reported a significantly higher methylation in children born to mothers who smoked throughout pregnancy compared with quitters or never smokers, which was most pronounced and only significant in males.Reference Murphy, Adigun and Huang 23 A finding that we do not replicate. In the same study, there was no association between methylation and smoking in early pregnancy compared with non-smoking, which is in line with our results in IGF2 DMR. In one of our own studies we did not find a significant association between maternal smoking throughout the pregnancy and methylation of several loci including IGF2 DMR and INSIGF at the age of 19 years.Reference Heijmans, Tobi and Stein 11 Important to note is that there are several differences between our studies, such as the younger age of the children in the present study. In our previous study, we studied children born before 32 weeks of gestation with and without SGA, sometimes accompanied with multiple pregnancy complications.Reference Heijmans, Tobi and Stein 11 Furthermore, in the present study we used periconception smoking and not smoking throughout pregnancy. Timing may be essential for INSIGF, as we previously reported an association with a decrease in INSIGF methylation and exposure to famine periconceptionally and not exposure later in gestation.Reference Tobi, Lumey and Talens 9 Supportive of the validity of our results is the relationship observed between the number of cigarettes and the increase in methylation of INSIGF, which may suggest a dose–response relationship. Further research on this locus and on other loci, would be interesting, especially since transgenerational effects of smoking of the mother but also of the grandmother has been reported to modify the risk for asthma in children, in which epigenetics is suggested to be the link.Reference Li, Langholz, Salam and Gilliland 24 Our group may have been too small to show an association of methylation with asthma.
Strength of our study is the fixed study moment relatively short after pregnancy in order to minimize recall bias regarding periconception exposures. Our overall power to detect statistically significant associations was limited by a relatively small sample of study subjects. In addition, many of the mothers who smoked in the periconception period were still smoking at the study moment. Current exposure may influence DNA methylation and we do not know to what extent their child was postnatally exposed to tobacco smoke.Reference Breitling, Yang, Korn, Burwinkel and Brenner 25
We measured DNA methylation of IGF2 DMR, IGF2R and INSIGF in whole blood; it would be very interesting to measure DNA methylation profiles in other tissues as well, such as subcutaneous fat or muscle tissue. Because the children were at a young age at the study moment, we are not able to assess diabetes, obesity or cardiovascular diseases in these children. Replication of our findings in larger groups and measuring DNA methylation of INSIGF in different tissues should be subject of future studies. This will be feasible because of new techniques where genome-wide methylation profiling on CpG sites and promoter regions is possible by high-throughput techniques. This will help us also to gain more insight into the intra- and interindividual distribution of methylation patterns throughout the genome in different tissues in different age categories.
In conclusion, the present study has shown that periconception maternal smoking and low education, as proxy for SES, are associated with methylation of the INSIGF gene in very young children. These effects may modify gene expression and future health and disease risks, but this has to be confirmed in larger birth cohorts.
Acknowledgements
The authors are grateful to the women and children who made this work possible. Furthermore, they thank Mr B. D. van Zelst and Mr P. H. Griffioen for laboratory assistance, and the project team of the HAVEN Study for data collection and management. Mrs C. Siebel is gratefully acknowledged for the cooperation with all the Child Health Care Centres of Careyn. Mr R. Talens, MSc, is acknowledged for assisting in obtaining and processing DNA methylation data. This research was financially supported by The Netherlands Heart Foundation (Grant 2002.B027) and the Bo Hjelt Foundation (Grant 2005).
