Abstract / Summary
Background: Both fetal and maternal genomes contribute to birthweight, but the timing and relative magnitude of their effects during gestation remain unclear. Polygenic risk scores (PRSs), constructed from genome-wide association studies (GWAS), capture the cumulative influence of common birthweight-associated variants to quantify genetic predisposition to fetal growth. We constructed fetal and maternal PRSs for birthweight in two UK birth cohorts, the Pregnancy Outcome Prediction Study (POPS) and Born in Bradford (BiB), to evaluate their associations with estimated fetal weight (EFW) throughout pregnancy and with birthweight at delivery. We also investigated whether PRSs could improve risk stratification of infants born below the 10th centile by distinguishing those with pathological intrauterine growth restriction (IUGR) from those who are constitutionally small-for-gestational-age (SGA). Methods: In POPS and BiB, we used LDpred2-auto to estimate PRSs based on the largest available maternal and fetal GWAS for birthweight. We longitudinally assessed EFW associations with PRSs using linear mixed effects models in POPS and BiB. We evaluated birthweight associations with PRSs using linear models, where birthweight was adjusted for gestational age and fetal sex. We estimated R2 and conducted variance decomposition analyses for all models. Genetically-predicted birthweights were estimated from joint models of fetal and maternal PRSs plus ten genetic principal components (PCs). BiB analyses were conducted separately for genetically-inferred European (EUR) and South Asian (SAS) ancestry. Results: While PRSfetus was significantly associated with EFW at baseline ([~]22 weeks gestation) across cohorts, PRSmother was not. Both the interaction of PRSfetus and PRSmother with gestational age (GA) were significantly associated with EFW. The effect of PRSmother on EFW increased more quickly across gestation than the effect of PRSfetus in POPS, but this finding was not replicated in BiB. Jointly, fetal and maternal polygenic risk explained 4% of variance in EFW in POPS and 1% in BiB-EUR. In POPS, fetal and maternal PRSs showed independent contributions to EFW variance of 3% and 0.5%, respectively. Joint and univariable linear models showed significant positive associations of PRSfetus and PRSmother with observed birthweight. In POPS, PRSfetus and PRSmother accounted for [~]10% and [~]4% of variance in birthweight over covariates alone, respectively, which was replicated in BiB-EUR. Maternal and fetal genetic effects jointly explained 11% of birthweight variance in POPS, similar to BiB-EUR, at 12%; however, the variance explained was reduced by half in BiB-SAS, at 6%. Among POPS babies born <10th birthweight centile, the magnitude of the difference between observed and genetically-predicted birthweight was greater in the group judged clinically to be IUGR vs. SGA (p=0.008; point estimate=0.29, 95%CI: 0.12, 0.47). Conclusions: Maternal and fetal PRSs jointly account for [~]11-12% of birthweight variance in European ancestry groups, but much less in a South Asian cohort. In POPS, the effect of birthweight-associated SNPs on estimated fetal weight increases more rapidly across pregnancy for maternal than for fetal genetic effects. Deviation from genetically-predicted birthweight may be a useful metric to help distinguish growth-restricted from constitutionally small babies.