Genetic analyses of the plasma proteome and metabolome in the same cohort from participants with African and European ancestry identify ancestry‐specific AD risk associated molecular traits

Abstract

AbstractBackgroundAlzheimer disease (AD) has different prevalence across different ancestries. However, the genetic factors underlying AD pathogenesis between diverse populations are largely unknown. The recent AD genome‐wide association study (GWAS) (Bellenguez et al 2022) reported 75 significant genetic loci in European ancestry, while the latest African‐based AD GWAS (Kunkle et al 2021) only reported two loci. It will be important to map these genetic loci into the functional molecular traits, e.g. RNA expression, protein abundance, metabolite levels. Multiple studies integrated such molecular traits with AD risk and termed them as TWAS, PWAS, and MWAS. However, few studies reported more than one molecular trait and in multiple ancestries.MethodWe first identified the plasma proteomic (SomaScan 7K) and metabolomic (Metabolon HD4) QTLs (quantitative trait loci) from participants with African (AFR, N = 400) and European (EUR, N = 2,300) ancestry, respectively. We next associated the genetic variants with the AD risk GWAS (EUR‐Bellenguez‐2022 & AFR‐Kunkle‐2021) using the protein or metabolite as the intermediate molecular trait. We extended the TWAS framework to test the disease association with the genetically predicted levels of proteins (PWAS) and metabolites (MWAS).ResultWe identified over 3,000 pQTLs and 400 mQTLs. Out of these findings, approximately 40% were novel QTLs compared to the previous studies. We found ancestry‐specific proteins and metabolites associated with AD risk. For example, proteins BIN1, TREM2, APOE, were only significant in EUR‐PWAS results but not AFR‐PWAS using the ancestry‐matched disease associations. Conversely, Spondin‐1 and RNF24 were only nominally significant in AFR‐PWAS. Similarly, metabolites glucuronide and sphingosine were only significant in EUR‐MWAS, whereas ethylmalonate was only nominally significant in AFR‐MWAS. There were 146 and 10 proteins associated with AD‐risk in EUR and AFR, respectively. 55 proteins can be used as drug targets. 119 proteins were not nominated by previous AD risk GWAS. Three and one metabolites were associated with AD‐risk in EUR and AFR, respectively. One metabolite can be used as drug target.ConclusionWe uncovered ancestry‐specific genetic architectures for proteomics and metabolomics. Our study serves as the first multi‐omic genetic study on multi‐ancestries for AD and can help guide the future ancestry‐specific therapy.