Transcriptome prediction in relevant tissues reveals mechanisms of drug-induced peripheral neuropathy. H. E. Wheeler1, B. P. Schneider2, D. L. Kroetz3, K. Owzar4, D. L. Hertz5, H. L. McLeod6, E. R. Gamazon1, K. P. Shah1, K. D. Miller2, G. W. Sledge2,7, N. J. Cox1, M. E. Dolan1, H. K. Im8 1) Dept of Medicine, University of Chicago, Chicago, IL; 2) Dept of Medicine, Indiana University, Indianapolis, IN; 3) Dept of Bioengineering and Therapeutic Sciences, University of California San Francisco, San Francisco, CA; 4) Dept of Biostatistics and Bioinformatics, Duke University, Durham, NC; 5) Dept of Clinical, Social, and Administrative Sciences, University of Michigan College of Pharmacy, Ann Arbor, MI; 6) Personalized Medicine Institute, Moffitt Cancer Center, Tampa, FL; 7) Dept of Medicine, Stanford University Medical Center, Stanford, CA; 8) Dept of Health Studies, University of Chicago, Chicago, IL.

   The biological mechanisms underlying associations discovered in GWAS are often not well understood because few associated variants fall in the protein-coding regions of genes. For many traits, including chemotherapeutic toxicity, gene regulation is likely to play a crucial mechanistic role given the consistent enrichment of eQTLs among trait-associated variants. Our approach, called PrediXcan, harnesses the regulatory knowledge generated by eQTL studies to directly test for genes associated with complex traits. The Genotype-Tissue Expression (GTEx) Project has increased the number of relevant tissues for which genotype and expression data are available. Chemotherapy-induced peripheral neuropathy is the major dose-limiting toxicity for several anticancer drugs. Using genotype and tibial nerve expression data from the GTEx pilot phase, we computed predicted levels of gene expression in six genotyped clinical neuropathy cohorts and tested the predictions for association with the neuropathy phenotypes. An advantage of this gene-based approach is that the results are biologically interpretable, guiding follow-up experiments and future drug development. Positive correlation between predicted levels and phenotype indicates high expression of the gene is associated with neuropathy risk, and thus the gene represents a potential drug target, while negative correlation indicates low expression is associated with neuropathy risk. We performed a meta-analysis of the results from six clinical neuropathy cohorts and one gene, TMED4, reached genome-wide significance (Z-score = 4.8, unadjusted P = 1.4 x 10-6, Bonferroni P = 0.02). TMED4 (transmembrane emp24 protein transport domain containing 4) is involved in vesicular protein trafficking and sensitizes cells to oxidative damage and cell death. Thus, peripheral nerves expressing high levels of TMED4 may be more sensitive to drug-induced oxidative stress and apoptosis. The second most significant gene in the meta-analysis was MARK3 (Z-score = -4.0, unadjusted P = 6.7 x 10-5, MAP/microtubule affinity-regulating kinase 3). These two genes were also the top two hits from a meta-analysis that only included the three (of six total) taxane-induced peripheral neuropathy cohorts (TMED4 P = 1.1 x 10-5; MARK3 P = 2.1 x 10-5). Given that taxanes target the microtubules, MARK3 also represents a promising candidate for functional investigation. PrediXcan is applicable to GWAS of other complex traits and pharmacological phenotypes.

You may contact the first author (during and after the meeting) at