Assessing the cardiotoxicity of 31 compounds using a multiplexed kinetic image cytometry (KIC)-based assay: Harnessing the predictive power of human iPSC-cardiomyocytes for early drug development

Human-induced pluripotent stem cell (hiPSC) models are more species-relevant than animal models and are amenable to high-throughput scalability. Previously, our Kinetic Image Cytometry (KIC) proarrhythmia assay examining calcium kinetics in hiPSC-cardiomyocytes (CMs) had a clinical accuracy of ~90 % (Pfeiffer, et al., 2016); when augmented with deep learning, accuracy increased to 95 % (Serrano, et al., 2023). Expanding upon this work, we screened 31 compounds with known clinical effects (effector classes: sarcoplasmic reticulum/calcium release, plasma membrane/ion channel, mitochondrial/myosin, and negative controls) in a 7-point concentration-response format—bracketing clinical exposures—using a panel of patient-derived hiPSC-CMs. We loaded hiPSC-CMs with Cal-520 (a green-channel calcium indicator), BeRST-1 (a far red-channel membrane voltage dye), TMRM (a mitochondrial membrane potential dye), and Hoechst (for nuclei), then exposed them to test compounds acutely. We then acquired interleaved calcium and voltage movies at 60 Hz (30 Hz/channel) as well as single images of TMRM and Hoechst. Vala's image analysis software, CyteSeer, was used to simultaneously measure calcium/action potential (AP) kinetics on—and to derive contraction metrics from—those movies. Furthermore, we mathematically characterized and classified calcium/AP waveforms to quantify compound-induced cardiotoxic effects. Additionally, we assessed mitochondrial health (using TMRM fluorescence) as well as acute cytotoxicity (using nucleus morphology) in the same experiment. By multiplexing these readouts on a single-cell level, we were able to correctly classify 30/31 (96.8 %) compounds on the basis of earliest-observed adverse effects (if any) by primary effector class at clinically-relevant test concentrations. While the data were more complex than initially expected (e.g., due to pleiotropic clinical effects and/or multiple efficacy−/toxicity-related targets), this multiplexed approach has the potential to elucidate mechanistic classes or targets quickly; and some of the more unexpected results from our screen highlight the need for de-risking early in drug development using relatively inexpensive and highly scalable hiPSC models. A larger study is currently underway to examine a library of 300+ compounds in order to comprehensively validate and refine our multiplexed assay.