A Targeted Metabolomics-Based Assay Using Human Induced Pluripotent Stem Cell-Derived Cardiomyocytes Identifies Structural and Functional Cardiotoxicity Potential

Abstract

Implementing screening assays that identify functional and structural cardiotoxicity earlier in the drug development pipeline has the potential to improve safety and decrease the cost and time required to bring new drugs to market. In this study, a metabolic biomarker-based assay was developed that predicts the cardiotoxicity potential of a drug based on changes in the metabolism and viability of human induced pluripotent stem cell-derived cardiomyocytes (hiPSC-CM). Assay development and testing was conducted in 2 phases: (1) biomarker identification and (2) targeted assay development. In the first phase, metabolomic data from hiPSC-CM spent media following exposure to 66 drugs were used to identify biomarkers that identified both functional and structural cardiotoxicants. Four metabolites that represent different metabolic pathways (arachidonic acid, lactic acid, 2'-deoxycytidine, and thymidine) were identified as indicators of cardiotoxicity. In phase 2, a targeted, exposure-based biomarker assay was developed that measured these metabolites and hiPSC-CM viability across an 8-point concentration curve. Metabolite-specific predictive thresholds for identifying the cardiotoxicity potential of a drug were established and optimized for balanced accuracy or sensitivity. When predictive thresholds were optimized for balanced accuracy, the assay predicted the cardiotoxicity potential of 81 drugs with 86% balanced accuracy, 83% sensitivity, and 90% specificity. Alternatively, optimizing the thresholds for sensitivity yields a balanced accuracy of 85%, 90% sensitivity, and 79% specificity. This new hiPSC-CM-based assay provides a paradigm that can identify structural and functional cardiotoxic drugs that could be used in conjunction with other endpoints to provide a more comprehensive evaluation of a drug's cardiotoxicity potential.

Keywords: in vitro; cardiotoxicity; drug discovery and development; hiPSC-CM; metabolites.

Figure 1.

Diagram outlining the development of the targeted biomarker assay for predicting cardiotoxicity potential.

Figure 2.

Graphical representation of the prediction model. A, The prediction distance (PD) for each metabolite is calculated at each concentration. These results are used to determine the composite prediction distance (CPD), which is the value used in the composite model. The area above and below the “Metabolite-Specific Cardiotoxic Response Thresholds” are associated with cardiotoxicity and the area between the “Metabolite-Specific Cardiotoxic Response Thresholds” are associated with noncardiotoxicity. B, The concentration-response curve for the composite model is illustrated with the black line. The concentration predicted by the point where the concentration-response curve of the composite model crosses the cardiotoxicity threshold (horizontal line) indicates the exposure level where a compound has the potential to cause cardiotoxicity (cardiotoxicity potential concentration, black bordered circle). For (A) and (B), the x-axis is the drug concentration. The y-axis is the solvent control-normalized (fold change) values for the metabolite response (A) or the composite prediction distance (B). C, Scoring algorithm employed for known cardiotoxicants (■) and noncardiotoxicants (●) utilizing the response at 10× C_max (x-axis) to determine the performance of the composite model. The color image is available in the online version of this article.

Figure 3.

Representative concentration-dependent effects on lactic acid metabolism in hiPSC-CM following cardiotoxicant exposure. Lactic acid (▲) and cell viability (●) concentration-response curves are shown for (A) dofetilide, (B) astemizole, and (C) amphotericin B. The x-axis is the drug concentration (µM) and the y-axis is the solvent control-normalized (fold change) value for lactic acid or cell viability. Data represent mean ± SEM (n = 3). If not shown, error bars are smaller than the size of the symbol.

Figure 4.

Representative concentration-dependent effects on arachidonic acid metabolism in human induced pluripotent stem cell-derived cardiomyocytes (hiPSC-CM) following cardiotoxicant exposure. Arachidonic acid (■) and cell viability (●) concentration-response curves are shown for (A) rosiglitazone, (B) ondansetron, (C) verapamil, (D) amiodarone, (E) arsenic trioxide, and (F) cisapride. The x-axis is the drug concentration (µM) and the y-axis is the solvent control-normalized (fold change) value for arachidonic acid or cell viability. Data represent mean ± SEM (n = 3). If not shown, error bars are smaller than the size of the symbol.

Figure 5.

Representative concentration-dependent effects on thymidine metabolism in hiPSC-CM following cardiotoxicant exposure. Thymidine () and cell viability (●) are shown for (A) busulfan, (B) levomethadyl acetate, (C) imatinib, and (D) doxorubicin. The x-axis is the drug concentration (µM) and the y-axis is the solvent control-normalized (fold change) value for thymidine or cell viability. Data represent mean ± SEM (n = 3). If not shown, error bars are smaller than the size of the symbol.

Figure 6.

Representative concentration-dependent effects on 2′-deoxycytidine metabolism in hiPSC-CM following cardiotoxicant exposure. 2′-Deoxycytidine (◆) and cell viability (●) are shown for (A) daunorubicin, (B) clozapine, and (C) pergolide. The x-axis is the drug concentration (µM) and the y-axis is the solvent control-normalized (fold change) value for 2′-deoxycytidine or cell viability. Data represent mean ± SEM (n = 3). If not shown, error bars are smaller than the size of the symbol.

Figure 7.

Determination of the predictivity of the (A) BAC-trained and (B) SEN-trained composite models. The highest prediction distance obtained from the 4 metabolites in the composite model (arachidonic acid, 2′-deoxycytidine, lactic acid, and thymidine) was subjected to ROC analysis. The dotted line represents the line of unity, which shows the results of random assignment of cardiotoxicity.