Identifying causal serum protein-cardiometabolic trait relationships using whole genome sequencing

Abstract

Cardiometabolic diseases, such as type 2 diabetes and cardiovascular disease, have a high public health burden. Understanding the genetically determined regulation of proteins that are dysregulated in disease can help to dissect the complex biology underpinning them. Here, we perform a protein quantitative trait locus (pQTL) analysis of 248 serum proteins relevant to cardiometabolic processes in 2893 individuals. Meta-analyzing whole-genome sequencing (WGS) data from two Greek cohorts, MANOLIS (n = 1356; 22.5× WGS) and Pomak (n = 1537; 18.4× WGS), we detect 301 independently associated pQTL variants for 170 proteins, including 12 rare variants (minor allele frequency < 1%). We additionally find 15 pQTL variants that are rare in non-Finnish European populations but have drifted up in the frequency in the discovery cohorts here. We identify proteins causally associated with cardiometabolic traits, including Mep1b for high-density lipoprotein (HDL) levels, and describe a knock-out (KO) Mep1b mouse model. Our findings furnish insights into the genetic architecture of the serum proteome, identify new protein-disease relationships and demonstrate the importance of isolated populations in pQTL analysis.

Figure 1

Chromosomal location of cis- (red) and trans-pQTLs (blue) plotted against the chromosomal location of the gene encoding the proteins of interest. Cis-pQTLs were defined as variants lying within 1 Mb of the start of the gene encoding the target protein.

Figure 2

Two-sample Mendelian randomization between proteins (exposure) and cardiometabolic traits (outcome), using only downloaded summary statistics. Points represent the effect size (beta) and direction of each causal association, with errors bars representing ±SE. Arrows indicate beta coefficients that are below −1. Actual beta and SE values are given to the right of each plot. Traits marked with an asterisk (*) indicate that a Wald ratio test was performed; otherwise, the inverse-variance weighted method was used. Full MR results with MRBase traits are given in Supplementary Material, Table S6.