Exome Genotyping Identifies Pleiotropic Variants Associated with Red Blood Cell Traits

Abstract

Red blood cell (RBC) traits are important heritable clinical biomarkers and modifiers of disease severity. To identify coding genetic variants associated with these traits, we conducted meta-analyses of seven RBC phenotypes in 130,273 multi-ethnic individuals from studies genotyped on an exome array. After conditional analyses and replication in 27,480 independent individuals, we identified 16 new RBC variants. We found low-frequency missense variants in MAP1A (rs55707100, minor allele frequency [MAF] = 3.3%, p = 2 × 10(-10) for hemoglobin [HGB]) and HNF4A (rs1800961, MAF = 2.4%, p < 3 × 10(-8) for hematocrit [HCT] and HGB). In African Americans, we identified a nonsense variant in CD36 associated with higher RBC distribution width (rs3211938, MAF = 8.7%, p = 7 × 10(-11)) and showed that it is associated with lower CD36 expression and strong allelic imbalance in ex vivo differentiated human erythroblasts. We also identified a rare missense variant in ALAS2 (rs201062903, MAF = 0.2%) associated with lower mean corpuscular volume and mean corpuscular hemoglobin (p < 8 × 10(-9)). Mendelian mutations in ALAS2 are a cause of sideroblastic anemia and erythropoietic protoporphyria. Gene-based testing highlighted three rare missense variants in PKLR, a gene mutated in Mendelian non-spherocytic hemolytic anemia, associated with HGB and HCT (SKAT p < 8 × 10(-7)). These rare, low-frequency, and common RBC variants showed pleiotropy, being also associated with platelet, white blood cell, and lipid traits. Our association results and functional annotation suggest the involvement of new genes in human erythropoiesis. We also confirm that rare and low-frequency variants play a role in the architecture of complex human traits, although their phenotypic effect is generally smaller than originally anticipated.

Figure 1

Quantile-Quantile Plots of Single-Variant Association Results in the All Ancestry Meta-analyses for the Seven Red Blood Cell Traits Analyzed (A) Distribution of the single variant results for all variants tested on the exome array. (B) Only markers with a minor allele frequency < 5% are shown here. (C) Variants outside of known RBC GWAS regions. Variants that are within 1 Mb from a previously published RBC GWAS locus were excluded for this QQ plot. Abbreviations are as follows: HCT, hematocrit; HGB, hemoglobin; RBC, red blood cell count; MCV, mean corpuscular volume; MCHC, mean corpuscular hemoglobin concentration; MCH, mean corpuscular hemoglobin; RDW, red blood cell distribution width.

Figure 2

CD36 Expression in Human Erythroblasts (A) In a dataset of 12 human fetal liver erythroblasts, all samples were homozygous at rs3211938 for the reference T-allele with the exception of one heterozygous sample (FL11). FL11 demonstrated strong allelic imbalance: we observed 705 reads for the reference allele (T) and 126 reads for the alternate allele (G) (binomial p = 4.9 × 10⁻⁹⁵). (B) FL11 (in green) shows the lowest CD36 expression level when compared to the other 11 samples. Abbreviation is as follows: FPKM, fragments per kilobase of transcript per million mapped reads.

Figure 3

Venn Diagram Summarizing Pleiotropic Effects for Genetic Variants Associated with Red Blood Cell Traits We considered variants only if their association p values with white blood cell (WBC) traits, platelet (PLT) traits, or with lipid levels was p < 1 × 10⁻⁴. Results for WBC and PLT are from the accompanying Blood-Cell Consortium articles., Results for lipids have previously been published (Table 3). Genes highlighted in red are novel RBC trait findings.