Ancestrally diverse genome-wide association analysis highlights ancestry-specific differences in genetic regulation of plasma protein levels

Abstract

Fully characterizing the genetic architecture of circulating proteins in multi-ancestry populations provides an unprecedented opportunity to gain insights into the etiology of complex diseases. We characterized and contrasted the genetic associations of plasma proteomes in 9,455 participants of European and African (19.8%) ancestry from the Atherosclerosis Risk in Communities Study. Of 4,651 proteins, 1,408 and 2,565 proteins had protein-quantitative trait loci (pQTLs) identified in African and European ancestry respectively, and twelve unreported potentially causal protein-disease relationships were identified. Shared pQTLs across the two ancestries were detected in 1,113 aptamer-region pairs pQTLs, where 53 of them were not previously reported (all trans pQTLs). Sixteen unique protein-cardiovascular trait pairs were colocalized in both European and African ancestry with the same candidate causal variants. Our systematic cross-ancestry comparison provided a reliable set of pQTLs, highlighted the shared and distinct genetic architecture of proteome in two ancestries, and demonstrated possible biological mechanisms underlying complex diseases.

Figure 1.. Flowchart displaying the strategy for the protein Quantitative Trait Loci (pQTLs) analysis in ARIC

Figure 2.. Genomic locations of protein Quantitative Trait Loci (pQTLs)

European ancestry (EA) signals are represented by a “+” symbol with cis signals represented in red and trans signals in magenta. African ancestry (AA) signals are represented by a “o” symbol with cis signals represented in blue and trans signals in green. The x axis indicates the positions of the sentinel variants of the pQTLs, and the y axis indicates the genomic locations of the coding genes of the proteins. Highly pleiotropic genomic regions are annotated at the top (red: EA only; green: AA only; black: both EA and AA).

Figure 3.. Visual summary of protein Quantitative Trait Loci (pQTLs) results by ancestry groups

From left to right: Upper Panel, significance of cis associations versus distance between sentinel variant to Transcription Starting Site (TSS); Number of significantly associated loci per aptamer; Number of conditionally significant associations within each pQTL; Bottom Panel, histogram of variance explained by conditionally significant variants; Effect size versus minor allele frequency.

Figure 4.. Colocalization of protein Quantitative Trait Loci (pQTLs) and genome-wide association signals of phenotypes

(A) Posterior probability of colocalizations of all four types of traits (Heart Failure (HF), Atrial Fibrillation (AF), sleep and others including fractional anisotropy, retinal vascular density, and retinal vascular fractal dimension) with pQTLs identified from European (EA) and African (AA) ancestries. (B) Significant associations of protein levels of LRIG1 in EA and AA and AF (3) in a region around the identified common causal variant, rs2306272, a missense variant in LRIG1. (C) Significant associations of protein levels of sPLA(2)-XIII encoded by PLA2G12B in EA and AA and HF (90) in a region around the identified common causal variant, rs12740374, a downstream gene variant at PSRC1 and 3’ UTR variant at CELSR2. (D) Significant associations of protein levels of LRRN1 in EA and AA and sleep (84) in a region around the identified common causal variant, rs429358, a missense variant at APOE.

Figure 5.. Mendelian Randomization (MR) analyses of plasma proteins’ effects on human diseases

The plot shows 12 statistically significant (P<3.62×10⁻⁷) Mendelian Randomization (MR) findings of selected human diseases, with each protein-disease pair being represented by a filled circle. The diseases were grouped into five categories (cancer, digestive, immunological, metabolic, and liver, and neurological) and are represented by different colors. The sizes of the circles emphasize the strength of the association. The colors of background square boxes represent the direction of effect. The x-axis corresponds to the diseases listed in alphabetical order and the y-axis corresponds to the gene symbol of the proteins.