Multiple associated variants increase the heritability explained for plasma lipids and coronary artery disease

Abstract

Background: Plasma lipid levels as well as coronary artery disease (CAD) have been shown to be highly heritable with estimates ranging from 40% to 60%. However, top variants detected by large-scale genome-wide association studies explain only a fraction of the total variance in plasma lipid phenotypes and CAD.

Methods and results: We performed a conditional and joint association analysis using summary-level statistics from 2 large genome-wide association meta-analyses: the Global Lipids Genetics Consortium (GLGC) study, and the Coronary Artery Disease Genome-Wide Replication and Meta-Analysis (CARDIoGRAM) study. There were 100 184 individuals from 46 GLGC studies for plasma lipids, and 22 233 cases and 64 762 controls from 14 studies for CAD. We detected several loci where multiple independent single-nucleotide polymorphisms were associated with lipid traits within a locus (12 out of 33 loci for high-density lipoprotein cholesterol, 10 of 35 loci for low-density lipoprotein cholesterol, 13 of 44 loci for total cholesterol, and 8 of 28 loci for triglycerides), reaching genome-wide significance (P<5×10(-8)), nearly doubling the heritability explained by genome-wide association studies (from 3.6 to 7.6% for high-density lipoprotein cholesterol, from 5.0 to 8.8% for low-density lipoprotein cholesterol, from 5.5 to 8.8% for total cholesterol, and from 5.7 to 8.5% for triglycerides). Multiple single-nucleotide polymorphisms were also associated with CAD (3 of 15 loci; an increase from 9.6% to 11.4% of heritability explained).

Conclusions: These results demonstrate that a portion of the missing heritability for lipid traits and CAD can be explained by multiple variants at each locus.

Keywords: coronary artery disease; genetics; genome-wide association study; lipids.

Figure 1

Regional association plots of three loci with significant joint associations for CAD. A, CDKN2A and CDKN2B on chromosome 9p21. B, APOA5-A4-C3-A1 on chromosome 11. C, LDLR on chromosome 19. SNPs are plotted as red diamonds with –log₁₀ P (y-axis) from the meta-analysis result for each trait. Significant joint association is marked as blue star corresponding to genomic position (x-axis) (See Supplemental Tables 1-6 for details). Color scheme illustrates linkage disequilibrium r² based the 1000 Genomes European Ancestry (CEU) data. Genes are shown at the bottom of each plot. Note that scale of y-axis is different from each figure.

Figure 2

Example of a locus with a secondary joint association signal conditioning on a primary signal. A, SNPs are plotted as red diamonds with −log₁₀ P (y-axis) from the meta-analysis result for CAD. The rs1122608 SNP is a primary signal at this LDLR locus. B, P values of the same SNP sets, which were calculated from conditional analysis on rs1122608, are plotted.