Cardiotoxicity is a well-known side-effect for the patients who are treated with different classes of anticancer drugs. In order to prevent potential drug-induced adverse effects, it is crucial to develop predictable human-based models and assays for drug screening. To that end, human induced pluripotent stem cell-derived cardiomyocytes (hiPSC-CMs) are becoming promising and important for disease modeling and drug-induced toxicity screening.

It is, however, a challenge to set up a fully automated high-throughput screening and analysis system especially geared for cardiotoxicity analysis on hiPSC-CMs. The major obstacle is to handle clustered cell growth patterns as well as strong intracellular and intercellular variation of a fluorescent signal.

We reported on the development of a fully automated image analysis system for quantification of cardiotoxic phenotypes from hiPSCCMs which are treated with varied concentrations of two anticancer drugs: doxorubicin and crizotinib.

We make use of signals from the nuclear channel and the a-actinin channel so as to segment single-cells by propagating segmented single nuclear region in the cardiac a-actinin region. In order to manage the heterogeneous a-actinin signals, we use enhanced fuzzy C-mean clustering to segment cardiac a-actinin signal. Compared to manual segmentation, our approach generates precision and recall scores of 0.81 and 0.93, respectively. The results show the reliability of our single-cell segmentation method even with heterogeneous a-actinin signals. We further quantify related phenotypes for each single cardiomyocyte. This fully automated image analysis system is dedicated to analyze high-throughput images and is capable of determining cardiotoxicity based on phenotypic changes in hiPSC-CMs.

Cardiotoxicity, hiPSC-derived cardiomyocytes, High-throughput screening, Image analysis, Phenotype quantification