Evidence for extensive pleiotropy among pharmacogenes. M. T. Oetjens1, W. S. Bush1,2, J. C. Denny2,3, K. A. Birdwell2, H. H. Dilks1, S. A. Pendergrass4,5, M. D. Ritchie4,5, D. C. Crawford1,6 1) Center for Human Genetics Research, Vanderbilt University, Nashville, TN; 2) Department of Biomedical Informatics, Vanderbilt University, Nashville, TN; 3) Department of Medicine, Vanderbilt University, Nashville, TN; 4) Center for Systems Genomics, The Pennsylvania State University, University Park, PA; 5) Department of Biochemistry and Molecular Biology, The Pennsylvania State University, University Park, PA; 6) Department of Molecular Physiology and Biophysics, Vanderbilt University, Nashville, TN.
Genetic variants in drug-metabolizing enzymes and drug transporters are exemplars of pleiotropy as a result of their diverse roles in the metabolism of drugs, pollutants, and endogenous compounds. Variants in pharmacogenes that change systemic levels of these compounds can lead to an altered risk of disease in carriers. For instance, genome-wide association studies (GWAS) revealed the dualistic role of the SLCO1B1 rs4149056 in serum bilirubin levels and statin induced myopathy. However, results from GWAS may underestimate the roles of pharmacogenes in disease risk. To systematically identify pleiotropic relationships among pharmacogenes, we performed phenome-wide association studies (PheWAS) using 6,092 European-descent subjects with DNA samples linked to de-identified electronic medical records as part of Vanderbilt University Medical Centers biorepository BioVU. All samples were genotyped on the Illumina ADME Core Panel, which was specialized for assaying 184 functional variants across 34 pharmacogenes. The ICD-9 codes present in these data were aggregated into 808 categories and matched control groups. We performed single SNP tests of association using logistic regression with each ICD-9 derived trait as an outcome adjusted for age and sex assuming an additive genetic model. We replicated five previously reported associations from the literature. For example, we robustly reproduced two associations between ADME genes and physiological traits: ABCG2 rs2231142 and gout p = 1.33 x 10-7 (OR = 1.74, 95% CI = 1.46 - 2.07) and SLCO1B1 rs4149056 and jaundice p=2.44 x 10-4 (OR = 1.67, 1.33 - 2.11). Epidemiological studies have reported an association between CYP2C19 variants and development of gastrointestinal cancer, and we observed an association between CYP2C19 rs4244285 and gastric cancer at p < 0.05 (OR = 1.69,1.09 - 2.63) and atrophic gastritis at p= 2.44 x 10-4 (OR= 1.99, 1.46 - 2.70). For novel associations, we set a Bonferroni corrected PheWAS significance threshold at p < 5.76 x10-5. We detected one novel association between SLC15A2 rs1143672 and renal osteodystrophy p=2.29 x 10-6 (OR = 0.60, 0.51 - 0.72). SLC15A2 encodes PEPT2, a peptide transporter expressed in the proximal tubule of the kidney. To our knowledge, this is the first systematic screen for phenotypic associations using functional variants in pharmacogenes. Collectively, this ADME PheWAS suggests pharmacovariants may have systemic disease risk as well as altered drug response.