Transcriptome-wide association study uncovers the role of essential genes in anthracycline-induced cardiotoxicity

Figures

• 

  Fig. 1. Overview of methods. A TWAS was initially conducted using both previous GWAS data and GTEx gene expression information to identify differentially expressed genes associated with ACT. From there, replication of the top associations in an independent cohort and a meta-analysis as well as subsequent functional validation were conducted. Gene set enrichment and pathway analyses identified gene lists and pathways overrepresented in the TWAS. Finally, compound and perturbation gene expression signature matching was conducted to identify compounds and perturbations with similar or dissimilar expression profiles to the TWAS genes.

• 

  Fig. 2. ACT TWAS identifies novel candidate genes based on expression profiles. a Analysis of diverse tissues identified four genes that passed an FDR correction for multiple testing in the TWAS and indicate that expression may modify ACT risk. Gene-level Z-scores, reflecting the number of standard deviations the mean expression of the gene (adjusted for the weight of each eQTL influencing expression of the gene) in ACT cases is from that of controls, across tissues where these genes were detected are plotted. b Association of top TWAS associated genes in tissues most relevant for ACT (i.e. heart and arterial). Z-score > 0 (green) indicates increased expression in ACT cases, while Z-score < 0 (purple) indicates decreased expression in these patients. c Gene interaction networks for the top TWAS associated genes identified using GeneMANIA.

• 

  Fig. 3. In response to increasing concentrations of doxorubicin, GDF5 undergoes gene expression changes and alters HCM cell viability when silenced. a Expression levels of GDF5 at increasing concentrations of doxorubicin were measured in HCM cells, normalised to the housekeeping gene HPRT1, and compared using one-way ANOVA to baseline (0 nM doxorubicin). Two independent experiments were performed (total n = 6) and their results combined. ****P < 0.0001 after Bonferroni multiple testing correction. Data are presented as mean and standard deviation. b Expression levels of GDF5 were silenced with RNAi and HCM cell viability at increasing concentrations of doxorubicin was measured using an MTT assay (n = 5). Dose-response curves were fitted to a non-linear regression, log(inhibitor) versus normalised response 4 parameter model and data are presented as mean and standard deviation. siNT represents negative control non-targeting siRNA. The logarithm of the half maximal inhibitory concentrations (IC50) are shown.

• 

  Fig. 4. Compounds with highly similar and dissimilar gene expression signatures to the TWAS. Expression signatures of compounds in CMap were compared with those of the TWAS. Compounds with a median connectivity score > 90 (similar gene signatures) or < −90 (dissimilar gene signatures) are shown. Anthracyclines are highlighted in red.