GWAS-by-subtraction reveals new genetic architecture and health implications of type 2 diabetes-independent gestational diabetes mellitus

Abstract Background The etiology of gestational diabetes mellitus (GDM) is constituted by both type 2 diabetes (T2D)-dependent and T2D-independent mechanisms. However, existing studies have evaluated the impact of T2D-dependent loci only for GDM, which limits the power to assess to what extent genetic variants or biological pathways are specific to GDM. Methods We subtracted the genetic effects of a genome-wide association study (GWAS) data for T2D from a GWAS data for GDM to reveal loci linked with T2D-independent components of GDM using genomic structural equation model (SEM). In the discovery stage of GWAS-by-subtraction, we used GWAS summary statistics of GDM and T2D from the FinnGen study as input latent variables (16,802 GDM cases and 237,816 controls; 71,728 T2D cases and 369,007 controls). In the replication stage, we adopted the summary statistics from a previously reported GWAS for GDM and T2D as input (21,263 GDM cases and 301,918 controls; 50,409 T2D cases and 523,897 controls). We functionally annotated variants with genome-wide significance, and performed comprehensive analyses including transcriptome-wide and proteome-wide associations, summary-data-based mendelian randomization, linkage disequilibrium score regression, and Mendelian randomization, to explore genetic patterns for T2D-dependent and -independent components of GDM. Results We found 69 independent genome-wide significant loci associated with the T2D-independent components of GDM for discovery stage, of which 49 SNPs are not significant in the GWAS for GDM used as input. T2D-independent components of GDM showed only genetic correlation with birth weight dependent on maternal genetic effect (rg = 0.19; P = 0.01), whilst T2D-dependent components of GDM showed genetic correlation with birth weight dependent on fetal genetic effect (rg = -0.22; P = 4.76 × 10 − 7 ). Loci of T2D-independent components of GDM effects map to genes related to islets of Langerhans, neural stem cells, blood levels of mannose, abundance of Streptococcus thermophilus and Bacteroides vulgatus , glucose and carbohydrate homeostasis, and the MAPK cascade and adenylate cyclase-activating G protein-coupled receptor signaling pathway. Conclusions We identified independent genetic components, loci and genes of GDM that may have been previously masked by T2D, providing more accurate targets for early detection and management of GDM.