The authors have declared that no competing interests exist.
Conceived and designed the experiments: HH DAE TL GML AB BS OM VM SH PYK DSS YF. Performed the experiments: SH PYK. Analyzed the data: PLW CMS. Wrote the paper: PLW HH CMS DAE DSS YF.
Maternal pre-pregnancy body-mass index (ppBMI) and gestational weight gain (GWG) are associated with cardiometabolic risk (CMR) traits in the offspring. The extent to which maternal genetic variation accounts for these associations is unknown.
In 1249 mother-offspring pairs recruited from the Jerusalem Perinatal Study, we used archival data to characterize ppBMI and GWG and follow-up data from offspring to assess CMR, including body mass index (BMI), waist circumference, glucose, insulin, blood pressure, and lipid levels, at an average age of 32. Maternal genetic risk scores (GRS) were created using a subset of SNPs most predictive of ppBMI, GWG, and each CMR trait, selected among 1384 single-nucleotide polymorphisms (SNPs) characterizing variation in 170 candidate genes potentially related to fetal development and/or metabolic risk. We fit linear regression models to examine the associations of ppBMI and GWG with CMR traits with and without adjustment for GRS. Compared to unadjusted models, the coefficient for the association of a one-standard-deviation (SD) difference in GWG and offspring BMI decreased by 41% (95%CI −81%, −11%) from 0.847 to 0.503 and the coefficient for a 1SD difference in GWG and WC decreased by 63% (95%CI −318%, −11%) from 1.196 to 0.443. For other traits, there were no statistically significant changes in the coefficients for GWG with adjustment for GRS. None of the associations of ppBMI with CMR traits were significantly altered by adjustment for GRS.
Maternal genetic variation may account in part for associations of GWG with offspring BMI and WC in young adults.
Increasingly, the enduring consequences of maternal perinatal obesity (pre-pregnancy obesity and excess gestational weight gain) are being recognized. A recent analysis by our group demonstrated associations of maternal pre-pregnancy body-mass index (BMI) and gestational weight gain (GWG) with adulthood obesity-related risk factors for metabolic and cardiovascular diseases in the offspring
Factors that account for the associations of maternal perinatal obesity with offspring cardiometabolic (CMR) traits are poorly understood. Maternal variation in genes involved in pathways related to metabolic risk and intrauterine growth may account for these associations through (1) direct transmission of genetic susceptibility to the offspring or (2) their influence on the intrauterine environment, critical in early life programming, growth, and development. Whether maternal genetic variation accounts, at least in part, for associations of maternal perinatal obesity with offspring CMR traits has not been fully examined. Previous analyses have provided limited evidence for an effect of maternal genetic variation on the association of measures of maternal perinatal obesity with offspring obesity at birth
We sought to determine whether maternal genetic variation in a set of candidate genes thought to be associated with intrauterine growth and metabolic risk accounts, at least in part, for the associations of maternal size before and during pregnancy with CMR traits in their young adult offspring.
The design and implementation of the Jerusalem Perinatal Study (JPS) have been described in detail previously.
The JPS Family Follow-Up study includes a sample of 1,249 mother-offspring pairs with available clinical and genotypic data from the original 1974–1976 birth cohort, who were interviewed and examined between 2007 and 2009. The sampling frame included only singleton deliveries after 36 weeks gestation without congenital malformations. We obtained a stratified sample of eligible pairs, where the strata were defined by ppBMI and birth weight. Pairs with low (≤2500 grams) and high (≥4000 grams) birth weight offspring as well as overweight or obese mothers (BMI≥27) were over-sampled. Approximately 80% of those sampled participated in a detailed interview and clinical examination.
The study was approved by the committee on research involving human subjects of the Hebrew University-Hadassah Medical School, by the Israeli National Genetic IRB committee, and by the institutional review board of the University of Washington, and all subjects provided written informed consent.
Data on physical activity, ethnic origin, education, and smoking status were collected. Standing height in cm was measured without shoes. Body weight in kg was measured in indoor light clothes, and body mass index (BMI) was defined as kilograms divided by meters squared (kg/m2). Waist circumference (WC) was an average of two measurements taken at the midpoint between the lower ribs and iliac crest, and an average of resting blood pressure in the right arm was obtained in mmHg from three consecutive measurements after five minutes of rest (Omron M7 automated sphygmomanometer). Blood samples were taken after an eight-hour fast, and analyses of glucose, total cholesterol, LDL-C, HDL-C, and triglycerides are reported in mg/dL. Insulin levels were determined by radioimmunoassay using a double-antibody method (Millipore) in mU/L and reflect an average of two values.
Genomic DNA from mothers and offspring was extracted and amplified at Hebrew University using conventional methods. Genotyping was performed at University of California, San Francisco using an Illumina, Inc., BeadArray. For this analysis, data on only maternal genotype were used.
We examined 1,384 single-nucleotide polymorphisms (SNPs) from 170 candidate genes in molecular pathways related to fetal development and metabolic risk in adults (
For our initial analyses, we fit linear regression models to examine the associations of pre-pregnancy BMI and GWG with each CMR trait. All traits that were not normally distributed were log-transformed. All analyses used inverse probability weighting to account for our sampling scheme, and all models were adjusted for offspring sex and ethnic background. Ethnic background of offspring was classified based on country of origin of all four grandparents, using nine strata (Israel, Morocco, Other North Africa, Iran, Iraq, Kurdistan, Yemen, Other Asia and the Balkans, and Ashkenazi).
Additionally, we included covariates thought to be associated with the exposures and outcomes. Maternal covariates included: parity (continuous); mother’s age at childbirth (continuous); maternal smoking during the pregnancy (ever/never smoked); socioeconomic status based on father’s occupation at time of birth and categorized as low, medium, and high; mother’s years of education at time of birth (continuous); a dichotomous variable for mother’s health during pregnancy, reflecting the presence of diabetes, hypertension, or pre-eclampsia; and gestational age in weeks (continuous). Offspring covariates were birthweight (continuous); smoking status (ever/never smoked), a dichotomous variable for physical activity (based on the question: during leisure time are you engaged in moderate or vigorous physical activity that lasts at least 20 minutes, 3 or more times a week?), and years of education (continuous). Pre-pregnancy BMI models were also adjusted for GWG, and GWG models for pre-pregnancy BMI. Because of the potential that covariates that occurred after conception might mediate the associations of interest, models unadjusted for birthweight, offspring smoking status, offspring physical activity, and offspring years of education were also fit. The first part of this analysis replicates an analysis recently published in Circulation on a similar dataset
We created genetic risk scores, using the Lasso shrinkage and selection method
To examine effects of variation in genes participating in these
Characteristics of the cohort are summarized in
Women (N = 614) | Men (N = 635) | Total (N = 1249) | ||||
Maternal pre-pregnancy BMI, kg/m2 | 24.3 | (3.9) | 23.7 | (3.9) | 24.0 | (3.8) |
Gestational weight gain, kg | 10.8 | (4.6) | 11.4 | (4.6) | 11.1 | (4.6) |
Maternal smoking in pregnancy, % | 16.8 | 18.3 | 17.5 | |||
Maternal ethnic origin, % | ||||||
12.3 | 13.5 | 13.0 | ||||
28.5 | 24.7 | 26.6 | ||||
23.0 | 23.5 | 23.2 | ||||
36.2 | 38.3 | 37.2 | ||||
Maternal years of education | 11.8 | (3.4) | 12.0 | (3.4) | 11.9 | (3.4) |
Parity | 2.9 | (2.0) | 2.8 | (1.8) | 2.9 | (1.9) |
Mother’s age | 28.4 | (5.6) | 28.0 | (5.2) | 28.2 | (5.4) |
Socioeconomic status, % | ||||||
21.1 | 23.4 | 22.3 | ||||
41.7 | 32.0 | 36.7 | ||||
37.1 | 44.6 | 40.9 | ||||
Birth weight, % | ||||||
13.4 | 8.8 | 11.1 | ||||
70.3 | 58.4 | 64.1 | ||||
16.6 | 32.8 | 24.8 | ||||
Gestational age at delivery, weeks | 40.0 | (1.5) | 40.0 | (1.5) | 40.0 | (1.5) |
Mothers with any medical condition |
8.6 | 7.2 | 7.9 | |||
Education, years | 14.9 | (2.6) | 15.4 | (3.6) | 15.2 | (3.2) |
Smokers, % | 18.5 | 35.3 | 27.0 | |||
Physically active |
49.4 | 54.9 | 52.2 | |||
Systolic BP, mmHg | 98.8 | (9.5) | 113.9 | (10.5) | 106.4 | (12.6) |
Diastolic BP, mmHg | 68.6 | (8.0) | 74.8 | (7.9) | 71.7 | (8.6) |
Waist circumference, cm | 81.2 | (13.3) | 91.4 | (12.3) | 86.4 | (13.8) |
BMI, kg/m2 | 25.9 | (5.4) | 26.9 | (4.8) | 26.4 | (5.2) |
HDL-C, mg/dL | 57.1 | (15.0) | 43.3 | (10.9) | 50.0 | (14.8) |
LDL-C, mg/dL | 107.9 | (28.3) | 117.3 | (28.2) | 112.7 | (28.6) |
Triglycerides, mg/dL | 91.9 | (49.5) | 121.2 | (82.0) | 106.9 | (69.7) |
Values are expressed as mean (SD) or percent.
*Diabetes, hypertension, heart disease, or pre-eclampsia.
**Includes self-report of moderate or vigorous physical activity lasting at least 20 minutes, 3 or more time a week.
Offspring mean BMI was 26.4 kg/m2 (range 16.9 to 52.7) and mean waist circumference was 86.4 cm (range 58 to 150) at age 32. Mean blood pressure was 107/72 mmHg. Average fasting glucose was 80 mg/dL with an average insulin level of 13 mIU/mL. Average LDL-C was 113, HDL-C was 50, and triglycerides were 107 mg/dL respectively. All cardiometabolic risk factors other than insulin and triglycerides were normally distributed.
Phenotype | n | Models withoutgenetic scores | Models withgenetic scores | % change in β[95% CI] | ||
Estimatedcoefficient | p-value | Estimatedcoefficient | p-value | |||
Offspring BMI(kg/m2) | 1092 | 1.810 | <0.0001 | 1.744 | <0.0001 | −4% [−24%, 21%] |
Waistcircumference (cm) | 1093 | 3.244 | <0.0001 | 2.988 | <0.0001 | −8% [−36%, 25%] |
Glucose (mg/dL) | 940 | 0.726 | 0.155 | 0.403 | 0.337 | −45% [−279%, 348%] |
Log-transformedinsulin |
930 | 1.051 | 0.079 | 1.048 | 0.109 | −5% [−124%, 482%] |
Systolic bloodpressure (mmHg) | 1079 | 1.685 | 0.004 | 1.570 | 0.006 | −7% [−56%, 103%] |
Diastolic bloodpressure (mmHg) | 1079 | 1.195 | 0.011 | 1.195 | 0.015 | 0% [−77%, 146%] |
HDL-C (mg/dL) | 984 | −1.661 | 0.022 | −2.265 | 0.004 | −37% [−262%, 27%] |
LDL-C (mg/dL) | 974 | 2.473 | 0.123 | 1.514 | 0.261 | −39% [−182%, 316%] |
Log-transformedtriglycerides (mg/dL) |
984 | 1.057 | 0.040 | 1.063 | 0.033 | 10% [−73%, 273%] |
0.847 | 0.503 | |||||
1.196 | 0.443 | |||||
Glucose (mg/dL) | 940 | 0.354 | 0.542 | 0.509 | 0.219 | 44% [−99%, 1304%] |
Log-transformedinsulin |
930 | 1.020 | 0.512 | 0.999 | 0.971 | −105% [−1627%, 103%] |
Systolic bloodpressure (mmHg) | 1079 | 0.955 | 0.062 | 0.771 | 0.099 | −19% [−126%, 269%] |
Diastolic bloodpressure (mmHg) | 1079 | 0.739 | 0.105 | 0.338 | 0.438 | −54% [−514%, 13%] |
HDL-C (mg/dL) | 984 | −1.420 | 0.071 | −1.538 | 0.062 | −8% [−288%, 62%] |
LDL-C (mg/dL) | 974 | 1.103 | 0.509 | −0.647 | 0.634 | −159% [−2168%, 32%] |
Log-transformedtriglycerides (mg/dL) |
984 | 1.039 | 0.144 | 1.043 | 0.037 | 38% [−22%, 1007%] |
All coefficients are for a 1-standard deviation change in level of the exposure variable (PBMI 1SD = 3.71, GWG 1SD = 4.61).
*Values were log-transformed to better approximate a normal distribution. To simplify interpretation, back-transformed results are presented and show the ratio of geometric mean of the outcome associated with a one-SD increase in BMI or GWG.
All models were adjusted for sex, ethnicity, offspring birthweight, maternal disease during pregnancy, parity, maternal smoking during pregnancy, family SES during pregnancy, gestational weeks at birth, maternal age at pregnancy, maternal education, offspring physical activity, offspring smoking status, and offspring education.
Pre-pregnancy BMI models were also adjusted for gestational weight gain, and GWG models for BMI.
All models used inverse probability weighting to account for stratified sampling.
In the models with scores, genetic scores predicting exposure and outcome are included. Scores were generated using the lasso algorithm, incorporating SNPs with no more than 5% missingness.
Confidence intervals for the % change in β were obtained via bootstrapping and based on quantiles of bootstrap replicates.
As reported previously
Compared to models not adjusted for maternal genetic risk scores, the coefficient for the association of a one-standard deviation difference in GWG and offspring BMI decreased by 41% (95% CI −81%, −11%) from 0.847 to 0.503 and the coefficient for a one-standard deviation in GWG and WC decreased by 63% (95% CI −318%, −11%) from 1.196 to 0.443, with adjustment for the maternal genetic risk scores. For the association of GWG and offspring BMI, the amount of variability explained by the model without maternal genetic risk scores was 0.16. The increase in the amount of variability explained by the model was 0.22 when maternal genetic risk score terms were included. For the association of GWG and offspring WC, the amount of variability explained by the model without GRS was 0.29. The increase in the amount of variability was 0.16 when GRS terms were included.
For the other offspring CMR traits, there were no statistically significant changes in the coefficients for GWG with adjustment for the maternal genetic risk scores, and none of the associations of ppBMI and offspring CMR traits were significantly altered by adjustment of the maternal genetic risk scores. Additionally, we repeated the analyses excluding covariates that occurred after the exposure of interest (including birthweight, offspring smoking status, offspring education level, and offspring physical activity level). For example, in the association of maternal gestational weight gain with offspring BMI, with these offspring covariates
In this study, associations of GWG with adult offspring measures of central adiposity (waist circumference) and overall adiposity (BMI) were attenuated in models that accounted for variation in maternal genes associated with fetal development and metabolic risk. In contrast, associations of maternal ppBMI with offspring BMI and WC were not similarly attenuated when we accounted for variation in a similar set of maternal genes. The observed attenuation of the GWG-offspring body size association by maternal genetic variation was independent of ppBMI.
Previous large-scale epidemiologic studies without direct measurement of genetic information have used statistical techniques such as generalized-estimating equations or linear mixed models to explore the role of maternal genetic variation in maternal size-offspring CMR associations, specifically offspring size, at various time points
Several reasons may explain the observation that variation in a set of maternal genes associated with development and metabolic risk appear to account for–at least in part–the associations of GWG with measures of offspring body size but not similar associations of ppBMI with offspring size. Firstly, the relationships may be different because the exposures themselves measure very different constructs. Pre-pregnancy BMI is a measure of baseline maternal obesity, while GWG is a surrogate for a complex measure–change in maternal body composition over the course of pregnancy–and incorporates contributions from both maternal diet and physical activity
Conversely, the lack of attenuation of association between ppBMI and offspring body size by maternal genetic variation is surprising given the high heritability of obesity, but there are possible explanations. Because GWG can be influenced by both maternal and fetal contributions, accounting for common maternal genetic variation may reflect genetic influences from mother
Lastly, epigenetic changes may amplify, attenuate, or mediate the effect of genotype on maternal GWG-offspring CMR associations. Siblings born before and after mothers underwent gastric bypass demonstrated differential levels of methylation of genes in glucoregulatory and inflammatory pathways
This study has two major strengths: First, it uses detailed records of maternal and offspring characteristics obtained at birth and in young adulthood. This allowed for adjustment for pregnancy-related factors and socioeconomic characteristics. Second, unlike previous studies, which have used statistical techniques to estimate the contribution of maternal genetic variation to associations of maternal and offspring size
Better understanding of these mechanisms is critical to guide recommendations for healthy gestational weight gain. This analysis suggests several future avenues for investigation. First, our results require validation. Second, the role of maternal genotype in gestational weight gain has not been well established, and limited evidence suggests that SNPs strongly associated with adult body size in genome-wide analyses may not be associated with higher levels of GWG
Single nucleotide polymorphisms used to calculate maternal genetic risk scores.
(DOCX)