Background

Impaired fetal growth and low birth weight increase the risk of short and long-term morbidity and mortality [1, 2]. A suboptimal prenatal environment contributes to fetal endocrine and metabolic adaptations with permanent effects [3, 2]. An important modifiable and adverse prenatal exposure is tobacco smoking by parents-to-be. The deleterious effects of maternal tobacco smoking on fetal growth and newborn weight are well-established [4]. The underlying mechanisms are not fully known but include vasoconstriction, reduced placental and fetal perfusion, and hypoxia [5]. Accumulating data suggest that derangements in epigenetic fetal and placental (re)programming may be one of the underlying mechanisms explaining the association between prenatal tobacco smoking and impaired fetal growth [6].

Epigenetic alterations are changes in gene expression potential that are not directly related to the deoxyribonucleic acid (DNA) sequence itself. DNA methylation modifications are one of the best understood epigenetic mechanisms [7]. During pregnancy, cells undergo major epigenetic reprogramming making them more susceptible for derangements during this period [7]. Both animal and human studies have shown that changes in the prenatal environment can lead to variations in epigenetic profiles [8–10]. Previous studies have shown global and gene specific differences in DNA methylation in different tissues in children prenatally exposed to tobacco smoking [11–13]. Insulin growth factor 2 (IGF2) and H19 are maternally and paternally imprinted genes located next to each other, playing an important role during fetal growth and development (Additional file 1: Figure S1) [14–17]. Moreover, the IGF2/H19 locus is a well-studied region that is known to be regulated through epigenetic mechanisms and sensitive to environmental exposures [9]. Previous research demonstrated that men may be more susceptible to detrimental early prenatal exposures than women [18].

Less is known about the influence of prenatal paternal tobacco smoking on gene-specific DNA methylation in newborns. Moreover, information on paternal tobacco smoking can be used to elucidate whether an effect on DNA methylation in newborns is due to the direct intrauterine effect of maternal smoking.

From this background, we aim to study whether prenatal parental tobacco smoking is associated with DNA methylation of IGF2 differentially methylated region (DMR) and H19 in umbilical white cord blood cells of newborns, with a focus on differences in sensitivity between genders. Furthermore, we examined mediation of DNA methylation in the association between tobacco smoking and the risk of being born small for gestational age (SGA).

Results

Maternal and newborns characteristics are presented in Table 1. Of the included mothers, 9.3 % (n = 47) reported tobacco smoking only in the first trimester and 24.7 % (n = 125) continued smoking during pregnancy. Mothers who continued tobacco smoking were younger, lower educated, and less often used a folic acid supplement during the periconception period than mothers who never smoked during pregnancy. Paternal tobacco smoking was more frequent in families with mothers who continued smoking (72.8 % in continued smoking versus 37.4 % in non-smoking mothers). Median (90 % range) birth weights of newborns from non-smokers, first trimester only and continued smoking mothers were 3500 (2536–4383), 3410 (2154–4219), and 3195 (2348–4081) grams. The median (90 % range) DNA methylation levels of IGF2DMR and H19 were 53.2 % (44.3–61.3) and 30.1 % (25.6–34.1), respectively. IGF2DMR and H19 were correlated (Pearsons r = 0.147, P value = 0.001). The DNA methylation levels of IGF2DMR and H19 in the non-smoking group were median (90 % range), 54.0 % (44.6–62.0), and 30.0 % (25.5–34.0), in the first trimester only smoking group 52.2 % (44.5–61.1) and 30.8 % (27.1–34.1), and in the continued smoking group 51.6 % (43.9–61.3) and 30.2 % (23.7–34.8), respectively. In addition, the methylation levels of IGF2DMR and H19 were plotted according to each of the smoking categories (Additional file 2: Figure S2).

Table 1 Baseline characteristics
Full size table

The associations between parental tobacco smoking and DNA methylation of IGF2DMR and H19 are shown in Table 2. Although IGF2DMR methylation of mothers who smoked only during the first trimester showed a tendency to be lower compared to non-smokers, no significant differences were observed. However, continued prenatal smoking was associated with lower IGF2DMR methylation (crude model, β = −1.14, 95 % CI −1.81; −0.47, P value = 0.001; adjusted model, β = −1.03, 95 % CI −1.76; −0.30, P value = 0.006). Expressed as relative to the standard deviation, the difference found in the adjusted model corresponds with a standardized effect size in DNA methylation of −0.13 standard deviation score (SDS). Of the mothers who continued smoking, 77.6 % (n = 97) provided information concerning the amount of cigarettes they smoked, of which 42.3 % (n = 41) reported to have smoked less than five cigarettes per day and 57.7 % (n = 56) reported to have smoked five or more cigarettes per day. No significant association was found between the amount of cigarettes and IGF2DMR methylation. However, a dose-dependent association of the number of cigarettes smoked was assessed in mothers who continued smoking (P value for trend = 0.030). Although no association was observed between mothers who continued smoking and H19 methylation, a significantly inverse association was revealed in mothers smoking up to five cigarettes per day (crude model, −0.90, 95 % CI −1.70; −0.11, P value = 0.026; adjusted model, β = −0.96, 95 % CI −1.78; −0.14, P value = 0.021). Expressed relative to the standard deviation, the difference found in the adjusted model corresponds with a standardized effect size in DNA methylation of −0.13 SDS. No significant associations were observed in mothers who reported to have smoked five or more cigarettes per day with H19 methylation. Among mothers who did not smoke during pregnancy, information on prenatal paternal tobacco smoking was provided by 99.4 % of the mothers (n = 332/334). Of the fathers who smoked, 45.6 % (n = 57/125) smoked less than five cigarettes per day and 53.6 % (n = 67/125) smoked five or more cigarettes per day. No association was observed between paternal smoking and IGF2DMR or H19 methylation. We repeated the analyses after exclusion of the attention deficit hyperactivity disorder (ADHD) cases (see Additional file 3: Table S3), which did not change substantially the effect estimates.

Table 2 Prenatal parental tobacco smoking habits and DNA methylation
Full size table

Table 3 shows the associations between maternal tobacco smoking and DNA methylation, stratified by gender. In mothers who smoked during the first trimester, an inverse association among newborn girls was shown with IGF2DMR methylation (crude model, β = −1.40, 95 % CI −2.79; −0.01, P value = 0.048), which attenuated in the adjusted model (β = −1.25, 95 % CI −2.71; 0.21, P value = 0.093). This association was not observed among boys. In addition, first trimester smoking was positively associated with H19 methylation among girls but lost significance after adjustment (crude model, β = 1.04, 95 % CI 0.04; 2.03, P value = 0.041; adjusted model, β = 1.00, 95 % CI −0.05; 2.03, P value = 0.063). The association between continued maternal tobacco smoking and IGF2DMR methylation was slightly more profound among girls (crude model, β = −1.46, 95 % CI −2.59; −0.33, P value = 0.011; adjusted model, β = −1.38, 95 % CI −2.63; −0.14, P value = 0.029) than in boys (crude model, β = −0.89, 95 % CI −1.73; −0.06, P value = 0.037; adjusted model, β = −0.72, 95 % CI −1.68; 0.24, P value = 0.142).

Table 3 Prenatal maternal tobacco smoking and DNA methylation, stratified by gender
Full size table

Associations between maternal tobacco smoking, the risk of a newborn being SGA, and gene-specific DNA methylation were analyzed using mediation analyses. We hypothesized that epigenetic alterations can mediate the association between maternal prenatal tobacco smoking and impaired fetal growth (Fig. 1). The indirect effect, defined as the product of the coefficient for the effect of maternal tobacco smoking on DNA methylation (path a), and the coefficient for the effect of DNA methylation on the risk of a newborns being SGA (path b), was calculated (Fig. 2). The observed significant effect (β = −0.095, 95 % CI −0.249; −0.018) substantiates the notion that the association between maternal prenatal smoking and the risk of a newborn being SGA is partially mediated through IGF2DMR methylation. The remaining direct effect of smoking on SGA (c’) seems to be explained by the slightly younger age and lower educational level of the smoking mother. No use of a folic acid supplement, higher body mass index (BMI) of the mother, nulliparity, and the male gender of the newborn did not significantly explain this effect.

Fig. 1
figure 1

Graphical representation of proposed mediation effect. Panel a illustrates the total effect (path c). Panel b represents the mediation design. The direct effect is depicted as path c’. The indirect effect (ab) is defined as the product of path a and path b

Full size image
Fig. 2
figure 2

Mediation analyses. Values represent unstandardized beta’s. *P < 0.05

Full size image

Discussion

Main findings

In 506 newborns, derived from a population-based birth cohort, we examined whether parental tobacco smoking is associated with IGF2DMR and H19 methylation in umbilical white cord blood cells. After adjustment for confounders, continued maternal prenatal tobacco smoking was associated in a dose-dependent manner with decreased IGF2DMR methylation in newborns. This association was stronger in newborn girls than boys. Decreased H19 methylation was also observed in mothers who continued to smoke <5 cigarettes/day. Prenatal paternal smoking was not associated with IGF2DMR or H19 methylation. Furthermore, the association between maternal smoking and the increased risk of being SGA may be partially mediated through IGF2DMR methylation.

Interpretation

Prenatal maternal tobacco smoking has been recognized as an important and modifiable detrimental exposure and has repeatedly been associated with disease [4, 19, 5]. Several studies have investigated associations between maternal prenatal smoking and global and gene-specific DNA methylation [20, 6]. Epigenome-wide methylation analyses revealed methylation changes at birth in newborns whose mother had smoked during pregnancy [21, 22]. Others investigated DNA methylation differences at the IGF2/H19 locus in relation to prenatal exposure to maternal tobacco smoking but found higher IGF2DMR methylation levels in boys born to smokers compared to non-smokers (+4.7 % IGF2DMR methylation) [15]. The region analyzed by Murphy et al. for the IGF2DMR included three CpG dinucleotides upstream of exon 3 (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) which overlapped with the region analyzed in our study (chr 11p15.5; CpG site 3, 2,169,709; CpG site 4, 2,169,660; CpG site 6/7, 2,169,518/2,169,515; NCBI Human Genome Build 37/hg19) [15]. The third CpG site analyzed by Murphy et al. overlapped with our eighth CpG site, which was excluded due to technical issues (see Additional file 4: Table S4) [15]. For the analyses of H19 DMR, the regions did not overlap. Other detrimental exposures, such as prenatal exposure to famine, have been previously linked to lower IGF2DMR methylation (−5.2 % IGF2DMR methylation when periconception exposure to famine), whereas advantageous exposures, such as periconceptional folic acid supplement use, have been associated with increased methylation (+4.5 % IGF2DMR methylation) [9, 8]. Our findings are also consistent with Guerrero-Preston et al., who also reported an inverse association between global DNA methylation and the nicotine biomarker cotinine however without stratification for gender [13]. In a previous study performed in the same population, we showed that SGA was associated with −1.13 % lower IGF2DMR methylation [17]. However, further research is warranted to further elucidate this difference.

In our study, the inverse association between maternal tobacco smoking from the first trimester onwards and IGF2DMR methylation seemed to be slightly more profound in girls. This finding is in contrast to previous research which demonstrated that boys may be more susceptible to detrimental exposures than girls [23, 18]. The first days after conception are important for the reprogramming of the epigenome. A limitation of our study is that we were not able to study difference in preconception and postconception smoke exposure on the methylation of the IGF2/H19 locus. This warrants further research.

We have focused on the hypotheses that epigenetic alterations mediate the association between maternal tobacco smoking and adverse birth outcomes. Our study supports the notion that DNA methylation can be one of the mechanisms through which tobacco smoking impairs fetal growth. As nicotine crosses the placenta, it can directly harm the fetus [24]. Cigarette smoke has been shown to contain high levels of reactive oxygen species (ROS) causing oxidative stress [25]. DNA ROS damage can interfere with the binding of DNA methyltransferases (DNMTs) to the DNA, resulting in DNA hypomethylation [26]. Tobacco smoking can also indirectly damage the fetus through the effects on the placenta. Nicotine causes vasoconstriction and thus can impair placental function [43]. To address this potential problem, we created standard curves constructed from DNA with low and high methylation (EpigenDx, Worcester, MA, USA) in steps of 10 % methylation difference on both amplicons. CpGs showing irregularities were excluded from the analyses. Last, CpG units with interference of single nucleotide polymorphisms (CEU) with a frequency of >5 % were also excluded, as this could change the weight of the CpG unit and therefore interfere with the measurement. Details concerning the success rate of the amplicons can be found in Additional file 5: Table S5.

Covariates

From self-administered questionnaires, data was available on maternal age and maternal educational level, parity, and folic acid supplement use before and during pregnancy. Maternal education level was assessed by the highest completed education and classified as (1) none/primary or “low”, (2) secondary or “medium”, and (3) college/university or “high”. Parity was classified as (1) nulliparous and (2) multiparous. Folic acid supplement use was categorized into (1) folic acid supplement use (pre- or postconception start) and (2) no folic acid supplement use. At enrolment (median 13.5 weeks, 90 % range 10.7–21.6), maternal weight and height were measured to calculate BMI (kg/m2). Information concerning date of birth, offspring sexm, and birth weight was obtained from community midwives and hospital registries.

Statistical analyses

Differences in maternal or newborns characteristics between maternal tobacco smoking categories were tested using ANOVA and chi-square tests. Thereafter, linear mixed models were used to examine the associations between maternal or paternal tobacco smoking (independent variable) and DNA methylation (dependent variable). This model was chosen as it can account for correlation between CpG dinucleotides, incorporates relevant adjustments within the models, and has the ability to accommodate missing data. The restricted maximum likelihood method was used for the model fitting. DNA methylation was treated as a continuous variable. To achieve normality, DNA methylation was square root transformed. Outliers per CpG (>3SDS) were excluded from further analysis. For all analyses, subject/person identifier was added as random effect and bisulphite batch and CpG site were added as fixed effects. In the crude analyses, maternal and paternal tobacco smoking were both entered as a fixed effects in separate models. In the adjusted analyses, potential confounders were additionally entered to the model at the same time as fixed effects.

Since maternal and paternal smoking are correlated, investigating paternal smoking among all mothers would overestimate the effect of paternal smoking if direct intrauterine mechanisms were present. Therefore, the association between paternal smoking and DNA methylation was examined among mothers who did not smoke during pregnancy. Prenatal exposures have shown to alter IGF2DMR methylation differently according to sex. [18] Therefore, the association between maternal tobacco smoking and DNA methylation was assessed in both boys and girls. The analyses were also repeated with exclusion of the ADHD (n = 88) cases. Two newborns were classified as both ADHD and SGA.

Additionally, we performed a mediation analyses to investigate whether DNA methylation of either IGF2DMR or H19 mediates the association between maternal tobacco smoking and the risk of a newborns being born SGA. The direct and indirect effects between the dependent (SGA) and independent variable (maternal tobacco smoking) as well as the mediator (DNA methylation) were tested (Fig. 1) using a bootstrap approach (1000 samples) described by Preacher and Hayes [44].

Missing data of potential confounders (maternal educational level (0.6 %), maternal BMI (0.2 %), and maternal folic acid supplement use (13.2 %) were completed using the Markov-Chain-Monte-Carlo multiple imputation technique [45]. Ten imputed datasets were created. The linear mixed model analyses were performed using the data measured in triplicates including imputed missing data. Overall, the effect estimates using the imputed dataset were slightly smaller compared to the original dataset. All analyses were performed using the Statistical Package for the Social Sciences version 21.0 for Windows (SPSS Inc, Chicago, IL, USA).