Genome-Wide Identification of Rare and Common Variants Driving Triglyceride Levels in a Nevada Population

Abstract

Clinical conditions correlated with elevated triglyceride levels are well-known: coronary heart disease, hypertension, and diabetes. Underlying genetic and phenotypic mechanisms are not fully understood, partially due to lack of coordinated genotypic-phenotypic data. Here we use a subset of the Healthy Nevada Project, a population of 9,183 sequenced participants with longitudinal electronic health records to examine consequences of altered triglyceride levels. Specifically, Healthy Nevada Project participants sequenced by the Helix Exome+ platform were cross-referenced to their electronic medical records to identify: (1) rare and common single-variant genome-wide associations; (2) gene-based associations using a Sequence Kernel Association Test; (3) phenome-wide associations with triglyceride levels; and (4) pleiotropic variants linked to triglyceride levels. The study identified 549 significant single-variant associations (p < 8.75 × 10^-9), many in chromosome 11's triglyceride hotspot: ZPR1, BUD13, APOC3, APOA5. A well-known protective loss-of-function variant in APOC3 (R19X) was associated with a 51% decrease in triglyceride levels in the cohort. Sixteen gene-based triglyceride associations were identified; six of these genes surprisingly did not include a single variant with significant associations. Results at the variant and gene level were validated with the UK Biobank. The combination of a single-variant genome-wide association, a gene-based association method, and phenome wide-association studies identified rare and common variants, genes, and phenotypes associated with elevated triglyceride levels, some of which may have been overlooked with standard approaches.

Keywords: GWAS; PheWAS; rare variant analysis; triglycerides; whole exome sequencing.

FIGURE 1

Phenome-wide analysis in the HNPT_EU between triglyceride levels and EHR diagnoses. Each point represents the p-value of an individual association between triglyceride levels and incidence of one of 1,372 phenotype groups, with covariates age, sex, DM2, and PC1–PC4. The x-axis shows different phenotypic conditions, grouped into 11 groups. The y-axis presents the -log₁₀ transform of the p-value of each association. The significance level α = 3.8 × 10^{– 3} is shown by the horizontal red line. Comprehensive results can be found in Supplementary Table 2.

FIGURE 2

Manhattan plot of significant SKAT-based gene collapse results. The x-axis represents the genomic start position of 25,283 genes. The y-axis represents -log₁₀-transformed raw p-values of each genotypic association. For ease of viewing, only genes above the horizontal line, which indicates the significance level α = 2.0 × 10^{– 6}, are annotated.