Discovery and refinement of genetic loci associated with cardiometabolic risk using dense imputation maps

Abstract

Large-scale whole-genome sequence data sets offer novel opportunities to identify genetic variation underlying human traits. Here we apply genotype imputation based on whole-genome sequence data from the UK10K and 1000 Genomes Project into 35,981 study participants of European ancestry, followed by association analysis with 20 quantitative cardiometabolic and hematological traits. We describe 17 new associations, including 6 rare (minor allele frequency (MAF) < 1%) or low-frequency (1% < MAF < 5%) variants with platelet count (PLT), red blood cell indices (MCH and MCV) and HDL cholesterol. Applying fine-mapping analysis to 233 known and new loci associated with the 20 traits, we resolve the associations of 59 loci to credible sets of 20 or fewer variants and describe trait enrichments within regions of predicted regulatory function. These findings improve understanding of the allelic architecture of risk factors for cardiometabolic and hematological diseases and provide additional functional insights with the identification of potentially novel biological targets.

Figure 1. Study design.

Summary of traits and studies investigated in this study. Study-specific information is given in Supplementary Table 1. HDL = high-density lipoprotein cholesterol; LDL = low-density lipoprotein cholesterol; TC = total cholesterol; TG = triglycerides; FG = fasting glucose; FI = fasting insulin; HOMA-B = homeostatic model assessment of β-cell function; HOMA-IR = homeostatic model assessment of insulin resistance; CRP = C-reactive protein; IL6 = interleukin-6; HGB = haemoglobin; RBC = red blood cell count; MCH = mean cell haemoglobin; MCHC = mean cell haemoglobin concentration; MCV = mean cell volume; PCV = packed cell volume or hematocrit; PLT = platelet count; WBC = white blood cell count.

Figure 2. Allelic spectrum of cardiometabolic trait variants.

(a) For each variant surpassing the genome-wide threshold in this study, the effect size (measured in standard deviations) is plotted as a function of the minor allele frequency proportion (MAF%). Loci discovered in this study are plotted with larger symbols. Associations for HDL (black) and PLT (purple) for which novel variants were discovered are shown. The dotted line represents the curve for 80% power with a sample size of 31,749 (for HDL) and an alpha of 8.31x10^–9. The power line for PLT (sample size 31,555) is similar and therefore not shown here. (b) Plot of smallest detectable effect size for a range of MAF%. Power calculations were performed for 4 traits of different trait groups and with different sample sizes: IL6, HOMA-B, LDL, and HGB.

Figure 3. GARFIELD functional enrichment analyses.

Wheel plot displaying functional enrichment of associations with PLT within DHS hotspot regions in ENCODE and Roadmap studies. Radial axis shows the fold enrichment (FE) values calculated at each of eight GWAS P-value thresholds (T<10^–1 to T<10^–8) for each of 424 cell types. Cell types are sorted by tissue, represented on the outer circle with font size proportional to the number of cell types from that tissue. Boxes and circles next to the tissue labels are coloured with respect to tissue (right legend). FE values at the different thresholds T areplotted with different colours on the inner side of the circle (e.g. T<10^–8 in black, bottom-left legend).Dots in the outer side of the circle denote significant enrichment (if present) for a given cell type at T<10^–5 (outermost) to T<10^–8 (innermost) (bottom-right legend). Results show overall well spread enrichment, with largest FE values obtained in blood, fetal spleen and fetal intestine tissues.

Figure 4. Fine mapping experiments.

Regional association plots for loci showing fine-mapped variants. (a) Numerical summary of 59 loci that were fine-mapped. (b) Example of fine-mapping and annotation at the LIPC locus for association with HDL. Panels show the regional association locuszoom plot, the PP statistics from the fine-mapping methods, the CATO and deltaSVM scores, VEP genic annotations and overlap of regulatory annotations found significant (coloured in blue) from GARFIELD enrichment analysis. Circles sizes and colours for all scores have been scaled with respect to score type (i.e. PP, CATO or deltaSVM) and numbers have been plotted below each circle.

Figure 5. Fine-mapping summary of variant consequences.

(a) Number of fine-mapped trait-region pairs containing at least one variant in the 95% credible set with consequences (i) coding; (ii) functional score and overlapping annotation significantly enriched for the given trait; (iii) functional score only; (iv) significant enrichment overlap only; (v) none of the above. (b) Distribution of the top posterior probability (PP) per region for variants with significant enrichment overlap and predicted functional score (after removing regions containing a coding variant). Boxplots depict the median (thick horizontal line), 1^st and 3^rd quartiles (coloured box), maximum and minimum values (whiskers) and outlying values (circles). (c) Proportions of variants within credible sets with coding, regulatory or no annotations.