Plasma lipid levels are heritable risk factors for coronary heart disease. Deleterious coding variants in known lipid genes are present in a very small fraction of individuals with extreme quantitative lipid levels. Whole genome sequencing (WGS) permits simultaneous assessment of Mendelian coding variants and the entirety of the genome for comprehensive evaluation of genetic drivers of extreme lipid levels. We estimated the contribution of monogenic and polygenic determinants of the extremes (upper and lower 5th percentiles) of LDL, HDL, and triglycerides in 5,910 white and in 4,380 black participants from the Framingham Heart Study (FHS), Jackson Heart Study (JHS), and the Multi-Ethnic Study of Atherosclerosis (MESA). Subject samples underwent deep-coverage WGS as part of the NHLBI TOPMed program. We catalogued presence of rare coding deleterious variants in known lipid Mendelian genes and calculated polygenic risk scores (PRS) utilizing 2M variants from prior plasma lipid GWAS summary statistics (Willer CJ et al, Nat Genet 2013). We defined individuals as having high polygenic risk if their PRS was in the upper 5th percentile of the distribution. Among whites, carrying a monogenic deleterious variant was associated with a 30 mg/dl increase in LDL (P=2x10-4); further, having a high PRS was associated with a 33 mg/dl increase in LDL (P=1x10-57). In blacks, those with a monogenic deleterious variant had 41 mg/dl higher LDL (P=2x10-7); further, those having a high PRS had 17 mg/dl greater LDL (P=6x10-10). Having a monogenic variant resulted in a larger increase in LDL in blacks than observed among whites, yet having a high PRS yielded a smaller increase in LDL in blacks than in whites. These results translate to individuals carrying a monogenic variant having an increased odds of extremely high LDL in both whites (10.9; 95% CI: 3.7-32.1) and blacks (7.4; 95% CI: 3.0-18.4); further, subjects having a high PRS is associated with an elevated odds of high LDL, greater in whites (7.7; 95% CI: 5.6-10.5) than in blacks (3.2; 95% CI: 2.1-4.9). In conclusion, we found that both rare coding variants and a PRS composed of common variants contribute to high LDL. In particular, a PRS comprised of >2M common variants substantially altered the odds of having an extreme lipid value. These results will be extended to larger sample sizes (>10,000 for whites and >9,000 blacks) for validation and other ancestries (Hispanics and Pacific Islanders) for transferability.

Introduction
Genetic analyses of plasma lipids (total cholesterol, LDL-C, HDL-C, and triglycerides) have yielded fundamental biological, clinical, and therapeutic insights for coronary heart disease (CHD). Whole genome sequencing now permits the most comprehensive genetic analysis of plasma lipids across large sample sizes.

Methods
Deep-coverage (>30X) whole genome sequences were generated as a part of the NHLBI Trans-Omics for Precision Medicine (TOPMed) Program. Plasma lipids were obtained for these individuals. Variants with minor allele count (MAC) > 20 were each individually associated with plasma lipids accounting for age, age^2, sex, cohort, genotypically-derived kinship, and prevalent Amish founder mutations in the dataset. Analyses were performed using MMAP and OASIS. Based on prior simulation analyses, we assigned statistical significance if P < 1x10-8.

Results
Whole genome sequences and plasma lipids were obtained for 28,541 ethnically-diverse individuals across 14 cohorts (FHS, JHS, Amish, MESA, GENOA, GeneStar, SAS, SAFS, GOLDN, DHS, CFS, WHI, ARIC, CHS) and combined into a single dataset. 35.9M high quality, MAC>20 genomic variants were included. Of observed associations, four novel sites were detected (outside of a +/- 500 kb window of 250 previously reported significant variants). While being just outside of the defined window, the 15q15.3 lead variant, associated with triglycerides, was in strong linkage disequilibrium with a known associated variant. The 9q31.1 lead variant, associated with LDL-C, is a low-frequency (MAF 0.4%) synonymous SNP in RNF20. The 11q14.1 (MAF 0.4%) and 11q23.3 (MAF 0.2%) lead variants were in non-coding sequences and associated with triglycerides. The 11q14.1 variant is near PRCP, whose product is a regulator of energy expenditure and fat mass. The 11q23.3 variant is ~511 kb away from the APOC3-A4-A1 cluster but is not in linkage disequilibrium with previously associated variants.

Conclusions
Deep-coverage whole genome sequence association with plasma lipids in 28,541 ethnically-diverse individuals yields putatively novel associations even at sample sizes much smaller than larger array-based genome-wide association analyses.