BackgroundInsulin is known to be associated with a higher risk of coronary artery disease (CAD), but molecular mechanisms remain unclear. This study aimed to explore protein-mediated pathways linking fasting insulin to CAD using Mendelian randomization (MR).MethodsThis MR study examined the association between fasting insulin and CAD using genome-wide association study (GWAS) data from MAGIC and CARDIoGRAMplusC4D. To investigate underlying mechanisms, a two-step proteome-wide MR analysis was conducted. First, associations of fasting insulin with 2940 circulating proteins were assessed using GWAS of proteomics from UKB-PPP. Proteins affected by insulin were then analyzed for their association with CAD risk. Proteins selected in both steps were considered as potential mediators. Sensitivity analyses to test whether associations are robust to pleiotropy and replication using other GWAS data, including GWAS of proteomics from deCODE and GWAS of CAD from FinnGen Biobank, were performed.ResultsGenetically predicted insulin was associated with a higher risk of CAD (odds ratio 1.79, 95% confidence interval 1.34 to 2.40). At a false discovery rate of 0.05, insulin affected 355 proteins, ten of which were both increased by insulin and linked to a higher risk of CAD. After sensitivity and replication analyses, PLA2G7, GZMA, LDLR, AGRP, and HHEX were identified as reliable mediators. Mediation analyses using non-pleiotropic instruments showed that PLA2G7, GZMA, LDLR, and AGRP explained 19.50%, 6.91%, 19.31%, and 29.66% of insulin's total effect on CAD, respectively.ConclusionsThis study identified five protein mediators linking insulin to CAD. These proteins could be considered as potential targets to mitigate insulin-related cardiovascular risk, providing novel insights for drug repurposing.