DNA methylation signatures of educational attainment

Abstract

Educational attainment is a key behavioural measure in studies of cognitive and physical health, and socioeconomic status. We measured DNA methylation at 410,746 CpGs (N = 4152) and identified 58 CpGs associated with educational attainment at loci characterized by pleiotropic functions shared with neuronal, immune and developmental processes. Associations overlapped with those for smoking behaviour, but remained after accounting for smoking at many CpGs: Effect sizes were on average 28% smaller and genome-wide significant at 11 CpGs after adjusting for smoking and were 62% smaller in never smokers. We examined sources and biological implications of education-related methylation differences, demonstrating correlations with maternal prenatal folate, smoking and air pollution signatures, and associations with gene expression in cis, dynamic methylation in foetal brain, and correlations between blood and brain. Our findings show that the methylome of lower-educated people resembles that of smokers beyond effects of their own smoking behaviour and shows traces of various other exposures.

Fig. 1

Manhattan plot showing the fixed effects meta-analysis p-values for the association between educational attainment and DNA methylation level. a EWAS not adjusted for individual smoking status. The red line denotes the significance threshold. b EWAS adjusted for individual smoking status. The red line denotes the significance threshold

Fig. 2

Overlap of educational attainment signatures and cigarette smoke signatures. a Venn-diagram showing the overlap between methylation sites associated with educational attainment in our meta-analysis and methylation sites previously reported to be associated with individual smoking or with maternal prenatal smoking in cord blood at birth. b Forest plot showing educational attainment meta-analysis effect sizes (betas) of significant CpGs, with and without adjustment for smoking status, and in never-smokers. c Educational attainment effect sizes (betas) in the NTR cohort for educational attainment top site cg05575921 (AHRR) while adjusting for various covariates, in all NTR subjects and in never smokers. SHS = second hand smoke exposure, indexed by plasma cotinine levels between 1 and 15. d Forest plot showing educational attainment fixed effects meta-analysis effect sizes (betas) of significant CpGs, with and without adjustment for self-reported smoking status, and adjusting for methylation level at cg05575921 (AHRR). The error bars display 95% confidence intervals. The sizes of squares are scaled by the size of the confidence intervals

Fig. 3

Heritability of CpGs associated with educational attainment and relationship between educational attainment and methylation level in twins. a Total heritability and proportion of variance explained by all genome-wide SNPs. b The forest plot shows the effect sizes (betas) of the 58 top sites for educational attainment adjusted for smoking status: 1) in the entire NTR cohort and 2) within twin pairs, derived by regressing the within-pair methylation difference on the within-pair educational attainment difference. c The forest plot shows the effect sizes (betas) of the 58 top sites for educational attainment in smoking discordant twin pairs (current smoker and never smoker) and smoking concordant twin pairs (both twins current smokers or both twins never smoked)

Fig. 4

Pearson correlation between effect sizes from previous epigenome-wide association studies and the meta-analysis effect sizes for educational attainment. a Methylation sites significantly associated with smoking. The effect sizes for smoking (current minus never smoking, y-axis) are plotted against the effect sizes (betas) for educational attainment, adjusted for smoking status (x-axis). b Methylation sites significantly associated with maternal smoking at birth. The effect sizes for maternal smoking in cord blood at birth (exposed minus non-exposed babies, y-axis) are plotted against the effect sizes (betas) for educational attainment, adjusted for smoking status (x-axis). c Methylation sites significantly associated with maternal smoking in older children. The effect sizes for maternal smoking in older children (exposed minus non-exposed children, y-axis) are plotted against the effect sizes (betas) for educational attainment, adjusted for smoking status (x-axis). d Methylation sites significantly associated with maternal plasma folate level. The effect sizes for maternal plasma folate level in cord blood at birth (y-axis) are plotted against the effect sizes (betas) for educational attainment, adjusted for smoking status (x-axis). e Methylation sites significantly associated with 28-day average fine matter particulate concentration. The effect sizes for fine particulate matter concentration in the environment (y-axis) are plotted against the effect sizes (betas) for educational attainment, adjusted for smoking status (x-axis). f Methylation sites significantly associated with alcohol intake. The effect sizes for alcohol intake (y-axis) are plotted against the effect sizes (betas) for educational attainment, adjusted for smoking status (x-axis). GW-significant Genome-wide significant

Fig. 5

Methylation level in the brain during foetal development at cg05575921 and the number of educational attainment top sites displaying significant changes in foetal brain, significant correlations between blood and brain, or significant associations with maternal smoking. a Methylation level in brain tissue (y-axis) is plotted against foetal age (x-axis) for cg05575921 (AHRR). b Venn-diagram for the overlap between educational attainment associated sites, sites significantly correlated between blood and one or more brain tissues, sites associated with maternal smoking (based on the previously published meta-analysis by Joubert et al.) and sites where methylation level in the brain correlates significantly with foetal age. All significant sites for educational attainment are included in the figure (N = 58).