Genome-wide mapping of plasma protein QTLs identifies putatively causal genes and pathways for cardiovascular disease

Abstract

Identifying genetic variants associated with circulating protein concentrations (protein quantitative trait loci; pQTLs) and integrating them with variants from genome-wide association studies (GWAS) may illuminate the proteome's causal role in disease and bridge a knowledge gap regarding SNP-disease associations. We provide the results of GWAS of 71 high-value cardiovascular disease proteins in 6861 Framingham Heart Study participants and independent external replication. We report the mapping of over 16,000 pQTL variants and their functional relevance. We provide an integrated plasma protein-QTL database. Thirteen proteins harbor pQTL variants that match coronary disease-risk variants from GWAS or test causal for coronary disease by Mendelian randomization. Eight of these proteins predict new-onset cardiovascular disease events in Framingham participants. We demonstrate that identifying pQTLs, integrating them with GWAS results, employing Mendelian randomization, and prospectively testing protein-trait associations holds potential for elucidating causal genes, proteins, and pathways for cardiovascular disease and may identify targets for its prevention and treatment.

Fig. 1

Flowchart of Study Design. (1) Selection and measurement of 71 high-value plasma proteins for atherosclerotic CVD via multiplex immunoassays in 7333 FHS participants, (2) GWAS of the 71 proteins in 6861 FHS participants to identify genome-wide significant pQTL variants, (3) independent external replication of sentinel pQTLs in INTERVAL, KORA, and previous GWAS, (4) colocalization and functional enrichment analyses of the identified pQTL variants, (5) integrated analysis of pQTL variants that coincide with CHD SNPs from GWAS, (6) identification of causal proteins for CHD using Mendelian randomization, (7) association analyses of proteins from steps 5 and 6 with risk for new-onset CHD/CVD events in 3520 FHS participants 50 years of age or older with available long-term follow-up. CHD coronary heart disease, CVD cardiovascular disease, FHS Framingham Heart Study, GWAS genome-wide association study, pQTL protein quantitative trait locus, SNP single-nucleotide polymorphism

Fig. 2

Sentinel cis- and trans-pQTLs and their associated proteins. Circos plots of sentinel cis- (left panel) and trans-pQTL variants (right panel). Sentinel pQTL variants are listed in order of chromosomal locations (blue boxes in the left semicircle). pQTL variants previously identified in GWAS to be associated with CHD appear in red text. Proteins with genome-wide significant pQTLs are listed in the right semicircle. The following two conditions are summarized for each protein: (1) The corresponding protein-coding gene is a known drug target (green text). (2) GO biological processes for the protein-coding gene (green box denotes lipid metabolism pathways, blue box denotes inflammatory/immune response pathways, yellow box denotes coagulation/platelet/hemostasis pathways, and gray box denotes other pathways not included in the three most common, previously listed pathways). A single primary GO process was chosen when the protein-coding gene was included in multiple pathways. CHD coronary heart disease, GO Gene Ontology, GWAS genome-wide association study, pQTL protein quantitative trait locus, SNP single-nucleotide polymorphism

Fig. 3

pQTL variant minor allele frequency vs. effect size. Minor allele frequency of sentinel pQTL variants (X-axis) vs. effect size of variants on proteins. The average absolute estimated effect size (per standard deviation per allele) is significantly different (by the unequal variance t-test) between coding and non-coding pQTL variants (0.56 versus 0.31, P = 0.02), and also significantly different between cis and trans-pQTL variants (0.53 vs. 0.30, P = 0.017). pQTL protein quantitative trait locus

Fig. 4

pQTL-protein-coronary heart disease network. Network of proteins and significant pQTL variants that are also GWAS risk SNPs for CHD. For proteins with multiple pQTL variants that coincide with CHD GWAS SNPs, the pQTL variant with the lowest P value of association with its corresponding protein level is shown. The following two conditions are summarized: (1) Proteins that tested causal for CHD in Mendelian randomization (P < 0.05). (2) Proteins associated with new-onset major CHD/CVD events (P < 0.0038) in 3520 Framingham Heart Study participants 50 years of age or older with long-term follow-up. Proteins in green fulfill neither condition 1 nor 2; proteins in blue fulfill condition 1; proteins in red fulfill condition 2; proteins in purple fulfill conditions 1 and 2. CHD coronary heart disease, FHS Framingham Heart Study, MR Mendelian randomization, pQTL protein quantitative trait locus

Fig. 5

Protein effects on coronary heart disease from Mendelian randomization and observed protein-trait associations. A comparison of protein effects on risk of CHD estimated from Mendelian randomization vs. the observed protein-trait associations hazards ratios. CHD coronary heart disease, CI confidence interval; CHD risk (per standard deviation increase in inverse-normalized protein level)