Gestational weight gain in pregnant women with obesity is associated with cord blood DNA methylation, which partially mediates offspring anthropometrics

Obesity and excessive gestational weight gain (GWG) influence the offspring's future health, and DNA methylation, an epigenetic mark occurring on CpG sites, constitutes a potential mechanism underlying this relationship.1, 2 GWG has been associated with birthweight, fat mass in newborns, and childhood obesity.1 Additionally, other suggested offspring long-term health effects linked to GWG are cancers and neurodevelopmental outcomes such as attention-deficit/hyperactivity disorder.1 To our knowledge, associations between GWG and DNA methylation in cord blood have only been investigated in normal-weight populations using either epigenome-wide association studies (EWAS)3, 4 or a panel of ∼1500 sites5. In the EWAS of normal-weight women, no associations between GWG and DNA methylation in cord blood were found.3, 4 We, therefore, examined if GWG in women with obesity, with a wide range of GWG (-5.0–34.1kg), associated with cord blood DNA methylation in the offspring of 232 newborns participating in the Treatment of Obese Pregnant Women study using EWAS (Figure S1 and Table 1, see Methods in Supporting Information). Approximately half of the women exceeded, and a quarter either stayed within or gained less weight than the Institute of Medicine recommendations regarding GWG. To test if cord blood DNA methylation is associated with GWG independent of treatment allocation, a linear regression model adjusted for maternal and offspring confounders and cell-type composition using a reference-free method was used (Model 1, Figure 1A). We additionally examined three models: Model 2, an unadjusted model without cell-type composition adjustment; Model 3, adjusted for maternal and offspring confounders, without cell-type composition adjustment; and Model 4, adjusted for maternal and offspring confounders and cell-type composition using a reference-based method (Figure 1A, see details in Supporting Information, Methods). In model 1, we found GWG to associate with differential DNA methylation at 441 sites, annotated to 352 genes, for example, ABCC8, FOXA2, GATA3, GRB10, NEUROD2, SMAD2, and TUB in cord blood based on a false discovery rate [FDR] <5% (Model 1; Figure 1A,B and Table S1). Six sites remained statistically significant after the Bonferroni correction (Figure 1B–H). DNA methylation of 13 of these 441 sites was also associated with gestational age (GA) and one with the lifestyle intervention, while none was associated with age, body mass index (BMI), or sex (FDR<5%, Table S1). Additionally, one of our found GWG-associated methylation sites, cg14663510 (HMX3), has previously been linked to a low pre-pregnancy BMI (Table S1)3. GWG was associated with methylation at 410, and 413 of the 441 sites in Model 2 and 3, respectively, and all 441 methylation sites were nominally associated with GWG in Model 4 (Figure 1A and Table S1). We further validated the results from Model 1 by randomly splitting the cohort into a discovery and validation cohort (60:40). In the discovery cohort (n = 125), 438 of our 441 methylation sites (99%) associated with GWG based on P = 8.77 × 10−9–4.00 × 10−2, while in the validation cohort (n = 83), 328 of our 441 methylation sites were identified with P = 1.98 × 10−10–4.9 × 10−2 (Table S1). Furthermore, 74% of the sites found in the discovery cohort could be confirmed in the validation cohort. Our results differ from previous studies performed on normal-weight women.3-5 These discrepancies may be due to differences in maternal BMI (normal-weight vs. obese), large variations in GWG in our study, different methods for cell-type composition adjustment, and different time points for measuring GWG. We proceeded to study the genetic influence on DNA methylation in cord blood of the 441 GWG-associated sites using the methylation Quantitative Trait Loci (mQTL) database.6 4911 single nucleotide polymorphisms (SNPs) have been associated with cord blood DNA methylation of 111 of our 441 sites, so-called mQTLs (Figure 1I and Tables S2–S5). Among these mQTLs, 39 SNPs were associated with disease traits in the genome-wide association studies catalogue, including asthma (e.g. genes SIK2 and WDR36) and waist-to-hip ratio adjusted for BMI (in gene ATP6V0A2) (Figure 1I and Table S2).7 Moreover, methylation of several of these mQTLs have been linked to BMI (cg12338137 in the gene body of TNS1), birthweight (e.g. cg22441770 in the gene body of CRTC2, and cg24796852 in GMFG promotor), and asthma (e.g. cg21689291 in the gene body of TMEM106A) in published EWAS (Tables S2–S5). Furthermore, we have previously shown that offspring of mothers with obesity are born with higher fat mass8 and that carbohydrate intake in late gestation in pregnant women with obesity is positively associated with fat mass in their offspring at birth.9 However, when women with obesity underwent lifestyle interventions during pregnancy, the offspring were born with more lean mass than the offspring of women assigned a control intervention.10 Therefore, we explored if GWG is also associated with lean mass at birth in 139 offspring. We found a negative correlation between GWG and offspring lean mass at birth (Figure 2A). When performing a linear model adjusting for lifestyle intervention, smoking, GA, and sex, it was estimated that with every kilogram of GWG, lean mass at birth decreased in the offspring by 0.23 ± 0.05 percentage points (95% CI: -0.33; -0.13). We next analyzed whether GWG was associated with offspring birthweight in 208 offspring and found a positive correlation between GWG and offspring birthweight (Figure 2B), in line with published data1. After adjustments, birthweight was estimated to increase by 21.1 ± 5.0 g (95% CI: 11.3; 30.9) for every unit of GWG. Thereafter, we tested whether cord blood methylation of our 441 sites was associated with offspring lean mass and birthweight. DNA methylation at 62 sites was associated with offspring lean mass, while methylation at 21 sites was associated with offspring birthweight (Tables S6 and S7). Methylation of 16 of the sites associated with both offspring anthropometric measurements. We then performed causal mediation analyses to investigate whether the effect of GWG on the offspring's lean mass and/or birthweight was partially mediated through cord blood DNA methylation of any of the lean mass (62 sites) and/or birthweight (21 sites) associated sites (Figure 2C–E). The mediation analysis breaks down the total effect of exposure (GWG) on the outcome (offspring's lean mass/birthweight) into two parts: first, the indirect effect acting via the mediator of interest (DNA methylation), and second, the direct effect acting directly or via a mediator other than what is under study (Figure 2C,D). We found DNA methylation at 21 and 17 sites to partially mediate the effect of GWG on offspring's lean mass and birthweight, respectively, including methylation sites annotated to CCN4, KDM1B, and MUC5AC (FDR<5%, Table 2). Gene symbol, hg38, GENCODE version 36 ACME estimate of mediator CpG (95% CI) ACME q-value Total Effect (95% CI) Proportion mediated by CpG (95% CI) −0.047 (−0.092; −0.013) −0.188 (−0.297; −0.086) −0.235 (−0.343; −0.134) 0.198 (0.055; 0.440) −0.061 (−0.123; −0.018) −0.174 (−0.273; −0.070) −0.235 (−0.343; −0.134) 0.259 (0.082; 0.588) −0.042 (−0.082; −0.011) −0.193 (−0.298; −0.098) −0.235 (−0.343; −0.134) 0.178 (0.052; 0.385) −0.053 (−0.097; −0.015) −0.181 (−0.284; −0.087) −0.235 (−0.343; −0.134) 0.228 (0.067; 0.448) −0.051 (−0.102; −0.017) −0.184 (−0.286; −0.070) −0.235 (−0.343; −0.134) 0.217 (0.078; 0.535) −0.063 (−0.114; −0.020) −0.172 (−0.284; −0.063) −0.235 (−0.343; −0.134) 0.268 (0.084; 0.610) −0.052 (−0.110; −0.011) −0.183 (−0.286; −0.077) −0.235 (−0.343; −0.134) 0.223 (0.055; 0.515) −0.052 (−0.103; −0.015) −0.183 (−0.281; −0.082) −0.235 (−0.343; −0.134) 0.223 (0.066; 0.473) −0.046 (−0.096; −0.010) −0.189 (−0.293; −0.090) −0.235 (−0.343; −0.134) 0.196 (0.049; 0.417) −0.046 (−0.085; −0.013) −0.189 (−0.293; −0.091) −0.235 (−0.343; −0.134) 0.197 (0.058; 0.393) −0.035 (−0.082; −0.007) −0.200 (−0.309; −0.095) −0.235 (−0.343; −0.134) 0.151 (0.031; 0.396) −0.051 (−0.094; −0.014) −0.184 (−0.290; −0.087) −0.235 (−0.343; −0.134) 0.217 (0.067; 0.467) −0.049 (−0.099; −0.015) −0.186 (−0.292; −0.081) −0.235 (−0.343; −0.134) 0.209 (0.069; 0.496) −0.040 (−0.081; −0.008) −0.195 (−0.306; −0.096) −0.235 (−0.343; −0.134) 0.171 (0.037; 0.381) −0.041 (−0.083; −0.009) −0.194 (−0.298; −0.096) −0.235 (−0.343; −0.134) 0.175 (0.039; 0.378) −0.059 (−0.117; −0.011) −0.176 (−0.290; −0.063) −0.235 (−0.343; −0.134) 0.251 (0.044; 0.574) −0.039 (−0.092; −0.007) −0.196 (−0.309; −0.079) −0.235 (−0.343; −0.134) 0.168 (0.026; 0.499) −0.036 (−0.082; −0.004) −0.199 (−0.302; −0.100) −0.235 (−0.343; −0.134) 0.151 (0.021; 0.387) −0.045 (−0.094; −0.010) −0.189 (−0.303; −0.093) −0.235 (−0.343; −0.134) 0.194 (0.043; 0.434) −0.038 (−0.087; −0.004) −0.196 (−0.298; −0.099) −0.235 (−0.343; −0.134) 0.164 (0.018; 0.388) −0.030 (−0.064; −0.004) −0.205 (−0.305; −0.108) −0.235 (−0.343; −0.134) 0.128 (0.015; 0.297) −0.038 (−0.079; −0.005) −0.197 (−0.291; −0.098) −0.235 (−0.343; −0.134) 0.163 (0.025; 0.376) −0.033 (−0.071; −0.005) −0.202 (−0.310; −0.099) −0.235 (−0.343; −0.134) 0.139 (0.023; 0.347) −0.031 (−0.066; −0.004) −0.204 (−0.310; −0.100) −0.235 (−0.343; −0.134) 0.132 (0.018; 0.336) −0.046 (−0.094; −0.007) −0.189 (−0.303; −0.076) −0.235 (−0.343; −0.134) 0.196 (0.031; 0.496) −0.038 (−0.080; −0.005) −0.197 (−0.304; −0.093) −0.235 (−0.343; −0.134) 0.162 (0.020; 0.395) −0.034 (−0.072; −0.004) −0.201 (−0.296; −0.107) −0.235 (−0.343; −0.134) 0.146 (0.019; 0.304) −0.031 (−0.073; −0.003) −0.204 (−0.309; −0.104) −0.235 (−0.343; −0.134) 0.131 (0.014; 0.330) 4.764 (1.699; 9.521) 16.323 (3.741; 27.365) 21.087 (9.363; 31.862) 0.226 (0.077; 0.643) 4.198 (1.233; 8.178) 16.889 (5.059; 27.542) 21.087 (9.363; 31.862) 0.199 (0.057; 0.520) 3.912 (0.697; 8.527) 17.175 (6.093; 27.767) 21.087 (9.363; 31.862) 0.185 (0.037; 0.454) 3.805 (0.759; 8.043) 17.282 (6.112; 27.558) 21.087 (9.363; 31.862) 0.180 (0.037; 0.461) 3.883 (0.899; 7.463) 17.204 (5.582; 27.884) 21.087 (9.363; 31.862) 0.184 (0.047; 0.482) 3.386 (0.582; 7.136) 17.701 (6.493; 28.443) 21.087 (9.363; 31.862) 0.161 (0.027; 0.424) 3.180 (0.455; 6.587) 17.907 (5.654; 29.069) 21.087 (9.363; 31.862) 0.151 (0.024; 0.423) 3.919 (0.569; 7.764) 17.168 (6.631; 27.663) 21.087 (9.363; 31.862) 0.186 (0.036; 0.445) 3.051 (0.641; 6.345) 18.036 (6.174; 29.420) 21.087 (9.363; 31.862) 0.145 (0.027; 0.404) 3.277 (0.446; 7.046) 17.810 (6.538; 28.048) 21.087 (9.363; 31.862) 0.155 (0.023; 0.420) 3.143 (0.491; 5.953) 17.944 (6.334; 28.594) 21.087 (9.363; 31.862) 0.149 (0.020; 0.390) 3.408 (0.257; 7.368) 17.678 (6.035; 28.248) 21.087 (9.363; 31.862) 0.162 (0.014; 0.436) 3.180 (0.356; 7.367) 17.907 (6.072; 28.262) 21.087 (9.363; 31.862) 0.151 (0.022; 0.415) 3.608 (0.410; 6.912) 17.479 (6.016; 27.682) 21.087 (9.363; 31.862) 0.171 (0.024; 0.402) 3.709 (0.248; 7.600) 17.378 (5.257; 27.569) 21.087 (9.363; 31.862) 0.176 (0.016; 0.485) 3.121 (0.170; 6.966) 17.966 (6.755; 28.466) 21.087 (9.363; 31.862) 0.148 (0.012; 0.394) 3.386 (0.175; 7.700) 17.701 (4.840; 28.923) 21.087 (9.363; 31.862) 0.161 (0.011; 0.515) In this study, we found for the first time associations between GWG in pregnant women with obesity and differential DNA methylation at individual sites in offspring's cord blood. Several identified epigenetic alterations were also associated with the offspring's lean mass and birthweight. Notably, we found that DNA methylation at 21 and 17 sites partially mediates the effect of GWG on lean mass and birthweight in the offspring, respectively. Six methylation sites are proposed to partially mediate the effect of GWG on both lean mass and birthweight. One site resides in KDM1B, encoding a histone demethylase that regulates histone lysine methylation.11 Additionally, in the present study, GWG was negatively associated with offspring's lean mass and positively associated with offspring birthweight, in line with previous work.1, 12 Interestingly, one in four of the found GWG-associated cord blood methylation sites has also been associated with genetic variation, so-called mQTLs, and traits such as asthma, birthweight, BMI, and type 2 diabetes based on previous EWAS. Hence, our novel data provide evidence that GWG in pregnant women with obesity impacts the methylome in offspring cord blood and on anthropometric measurements with probable importance for the health of the offspring (Figure 2G). Together these data imply that reducing GWG in a population with obesity is relevant to reduce potential long-term health effects in the offspring. In conclusion, this study provides evidence that GWG in pregnant women with obesity is associated with cord blood DNA methylation of sites previously linked to BMI, type 2 diabetes, and asthma. We also demonstrate that anthropometric measurements of importance for the future health of the offspring (i.e. lean mass and birthweight) are associated with GWG. This further supports that reducing GWG in women with obesity may be of value for the offspring's future health. These results also stress the importance of the intrauterine environment in humans and its ability to program the methylome, potentially affecting the offspring's metabolism. Please note: The publisher is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.