Genetic effects on serum and urine metabolites across ancestries and their links to disease

Abstract

Genome-wide association studies of the metabolome have predominantly examined European-ancestry populations using a single biofluid, constraining both the discovery of metabolite quantitative trait loci (mQTLs) and the resolution of causal variant finemapping. Here, we report a multi-ancestry genome-wide association study of 1,955 serum and urine metabolites in 4,431 European American and African American participants from the Atherosclerosis Risk in Communities study. We identify 969 mQTLs for 553 metabolites, 30% unreported in studies of matched biofluid and ancestry, and replicate 91-99% of testable mQTLs in two independent cohorts. In the African American sample with balanced sample sizes, 36% of metabolites with mQTLs are serum-specific, 39% urine-specific, and 25% shared, demonstrating complementary genetic architecture. Only 23 of 632 candidate ancestry-specific mQTLs show heterogeneous effects beyond variant availability and power, yet allele frequency differences expose functional variants such as an SLC28A1 splice variant, virtually absent in Europeans, that increases urine uridine excretion, implicating impaired CNT1-mediated renal nucleoside reabsorption. Multi-ancestry finemapping improves credible set resolution, and colocalization with the Million Veteran Program phenome links the catalog to cardiometabolic, kidney and liver traits consistent with prioritized gene function. Integrating biofluids and ancestries enhances genetic discovery and mechanistic characterization of the human metabolome.