Multi-ancestry GWAS of severe pregnancy nausea and vomiting identifies risk loci associated with appetite, insulin signaling, and brain plasticity

Abstract

While most pregnancies are affected by nausea and vomiting, hyperemesis gravidarum (HG) is at the severe end of the clinical spectrum and is associated with dehydration, undernutrition, and adverse maternal, fetal, and child outcomes. Herein we performed a multi-ancestry genome-wide association study (GWAS) of severe nausea and vomiting of pregnancy of 10,974 cases and 461,461 controls across European, Asian, African, and Latino ancestries. We identified ten significantly associated loci, of which six were novel (SLITRK1, SYN3, IGSF11, FSHB, TCF7L2, and CDH9), and confirmed previous genome-wide significant associations with risk genes GDF15, IGFBP7, PGR, and GFRAL. In a spatiotemporal analysis of placental development, GDF15 and TCF7L2 were expressed primarily in extra villous trophoblast, and using a weighted linear model of maternal, paternal, and fetal effects, we confirmed opposing effects for GDF15 between maternal and fetal genotype. Conversely, IGFBP7 and PGR were primarily expressed in developing maternal spiral arteries during placentation, with effects limited to the maternal genome. Risk loci were found to be under significant evolutionary selection, with the strongest effects on nausea and vomiting mid-pregnancy. Selected loci were associated with abnormal pregnancy weight gain, pregnancy duration, birth weight, head circumference, and pre-eclampsia. Potential roles for candidate genes in appetite, insulin signaling, and brain plasticity provide new pathways to explore etiological mechanisms and novel therapeutic avenues.

Figure 1.

Manhattan plot for HG multi-ancestry GWAS. The plot shows −log10 transformed p-value against genomic positions. The red line represents genome-wide significant threshold p < 5×10⁻⁸. The significant SNPs are annotated with the nearest candidate gene. In the annotation, red indicates novel SNPs and black indicates SNPs replicated in previous studies.

Figure 2.

Effect size estimates for the lead SNP at each locus in the MoBa cohort. Forest plot of the effect size estimates when encoding no NVP, NVP, and HG as 0, 1, and 2, respectively, for (A) maternal effect estimates as obtained from the GWAS against the maternal genome and (B) conditioned maternal and fetal effects obtained after WLM-adjustment of effect estimates in GWASes against the fetal, maternal, and paternal genomes. For the sake of readability, paternal effect sizes are not represented. Points represent effect size estimates and error bars represent 95% confidence intervals of the estimate.

Figure 3.

Enrichment analysis suggests evolutionary constraint shapes genetic architecture of HG.

Figure 4.

Phenome-wide association study (PheWAS) on lead HG variants. Phenotypes were identified for 3 / 10 variants (p < 1.48×10⁻⁶ after adjusting for multiple testing; red dashed line).

Figure 5.

Cross sectional analysis in MoBa. Maternal effect size estimates of the lead SNPs obtained in GWASes of NVP reported by mothers stratified by gestational week. The size of the point represents the significance of the association and its color the effect size estimate.

Figure 6.

Spatiotemporal candidate gene expression during early placental development. A) Model of fetal and maternal tissue during trophoblast differentiation, invasion, and spiral artery remodeling. Regions of interest (ROI) were selected corresponding to one of the following categories: artery (N=17), decidua (N=10), floating villi (N=12), VCT (N=15), interstitial EVT (N=23), anchoring EVT (N=20), endovascular EVT (N=8), endometrial glands (N=4) from 30 patients between 6–20 weeks’ gestation. Normalized expression (individual gene counts divided by counts for all genes) for each ROI is shown for B) GDF15, C) TCF7L2, D) PGR, and E) IGFBP7.