Identification of new rare coding variants associated with hemoglobin levels and platelet counts. P. Livermore Auer1,4, N. Chami2, U. Schick1,3, S. de Denus2, C. Carlson1, M. Dube2, R. D. Jackson5, J. D. Rioux2, C. L. Kooperberg1, U. Peters1, J. C. Tardif2, L. Hsu1, A. P. Reiner1,3, G. Lettre2 1) Fred Hutchinson Cancer Research Center, Seattle, WA; 2) Universite de Montreal and Montreal Heart Institute, Montreal, Quebec, Canada; 3) University of Washington, Seattle, WA; 4) University of Wisconsin-Milwaukee, Milwaukee, WI; 5) The Ohio State University, Columbus, OH.

   Introduction: Hematological traits are clinically important diagnostic and prognostic parameters. Inter-individual variation in these blood-related traits is heritable and genome-wide association studies (GWAS) have already implicated dozens of loci. The development of new genotyping arrays that build on variant catalogues from deep re-sequencing efforts now allows extension of the search for associations with blood cell phenotypes to low-frequency and rare coding variants. Materials and Methods: We genotyped 18,388 participants from the Women Health Initiative study and 6,796 patients from the Montreal Heart Institute biobank on the Illumina ExomeChip array. We analyzed hemoglobin levels, hematocrit, and white blood cell and platelet counts as quantitative phenotypes for association with single nucleotide variants (SNV) using linear regression and gene-based tests (T1, SKAT-O). To minimize the impact of phenotypic outliers on rare variant association results, extreme values of blood phenotypes were winsorized (bottom 0.5% and top 99.5%). Results: We replicated several associations with common SNVs previously identified by GWAS: TMPRSS6 and HFE variants with hemoglobin levels (P<5x10-17); variants within the 17q12 locus, the HLA region and the ABO gene with white blood cell count (P<1x10-7); and SNVs near ARHGEF3, SH2B3 and JMJD1C with platelet counts (P<1x10-16). We also found potential causal variants at several loci previously identified by GWAS, including a splice site mutation in C1orf150 (minor allele frequency (MAF)=7%, P=2x10-11) and a SH2B3 Glu400Lys missense variant (MAF=0.1%, P=2x10-8) associated with platelet count. Finally, we identified novel associations with rare coding alleles: a rare missense variant in the erythropoietin (EPO) gene (MAF=0.4%, effect=-0.3 g/dL per rare allele, P=7x10-7) and a splice site beta-thalassemia variant (HBB) (MAF=0.02%, effect=-2.3 g/dL per rare allele, P=9x10-9) associated with hemoglobin levels, and the somatic JAK2 V617F mutation associated with platelet count (MAF=0.05%, effect=124x109 platelets/L, P=3x10-22). Conclusions: Our results confirm the utility of exome-wide genotyping in large populations to validate and fine-map previously identified genetic association signals. It is also a powerful tool to discover new genes and biological pathways involved in complex diseases and traits.