Cardiotoxicity is defined as toxicity that affects the heart. Drug-induced cardiotoxicity remains an important cause of pre-clinical and clinical drug failure. At Cyprotex, we are developing cutting-edge strategies to effectively predict toxicities at an early stage in the drug development process to guarantee the progression of safe novel pharmaceuticals and reduce later-stage attrition.
The mechanisms by which drugs can induce cardiotoxicity are diverse, ranging from functional (acute alteration of the mechanical function of the myocardium) to structural impairment (morphological damage to cardiomyocytes), and the clinical manifestations are wide-ranging, spanning from arrhythmia to myocardial dysfunction, to terminal heart failure. Cardiotoxicity generally results from the disruption of key cardiomyocyte processes affecting contractility, electrophysiology (ion channel trafficking), mitochondrial function, growth factors and cytokine regulation. Consequently, our assays have been developed to cover various readouts by combining multiple approaches to obtain a complete understanding of toxic effects.
At the 58th Congress of the European Societies of Toxicology (Copenhagen, Eurotox 2024), we presented our latest work in cardiotoxicity prediction in the form of a poster, titled: “High-throughput transcriptomics combined with in vitro assays for cardiotoxicity risk assessment and mechanistic understanding”. Here, we investigated the effects of 148 reference compounds, (including structural and/or functional cardiotoxicants as well as non-cardiotoxicants) on human induced pluripotent stem cell derived cardiomyocytes (hiPSC-CMs) using high-throughput transcriptomics assessing the entire transcriptome (HT-transcriptomics), high-content imaging (HCI) and kinetic monitoring of calcium transients (CaT). Our compound set covered a broad range of mechanisms of action including ion channel inhibitors (Na+, K+, Ca2+), receptor modulators (adrenergic, dopamine, serotonin, histamine, acetylcholine, glucocorticoid, sulfonylurea), enzyme activities (COX, phosphodiesterase) and DNA metabolism.
Calcium transients, assessed by fast kinetic fluorescent readings, allowed a series of Ca2+ peak parameters to be studied including amplitude, frequency, full rise and decay time and peak width, which taken together revealed the effects of compounds on cardiomyocyte contraction. Since the calcium transients are closely associated with muscle contraction and ventricular action potentials, they can help us understand the in vivo cardiotoxicity effects of some compounds including electrocardiogram alterations such as QT interval prolongation. Additionally, HCI was used to assess any structural damage to the cardiomyocytes upon analysis of nuclei impairment, calcium homeostasis and mitochondrial function. Finally, HT-transcriptomics shed light on the transcriptional responses triggered upon compound treatment, which were further analysed for pathway enrichment and differential gene expression.
This multi-parametric approach allowed the identification of the readout showing the lowest minimum effective concentration (MEC). Compounds were then classified as cardiotoxic if the MEC value was below a specific maximum plasma concentration (Cmax) threshold calculated using in vivo literature cardiotox classifications. Additionally, AI/Machine Learning (ML) models were developed to predict cardiotoxicity using a 20x Cmax threshold and a tox score threshold. This allowed the classification of compounds as cardiotoxic if the true Cmax (historical in vivo response) was above the predicted safe Cmax, giving excellent prediction metrics with 78.7% sensitivity, 86.7% specificity and 81.4% accuracy.
Finally, testing dynamic Cmax thresholds and different assay combinations proved useful to effectively predicting cardiotoxicity risk with excellent accuracy, whilst assigning more weight to specificity over sensitivity to avoid losing valuable drug candidates due to false-positive risks. The best predictions were achieved by combining HT-transcriptomics AI/ML modelling (20x Cmax), HCI (1x Cmax) and CaT (2x Cmax) endpoints, with 85.9% sensitivity, 84.1% specificity and 85.3% accuracy.
Future work will involve further expansion of our reference compound list to cover an even larger range of mechanisms and chemical space, and to explore the transcriptomics pathway endpoints by performing a Point of Departure (PoD) using Benchmark Dose (BMD) analysis approach, for both hiPSC-CMs and organotypic 3D models which are likely to be more representative of the in vivo tissue structurally and functionally.
Interested in learning more?
Contact us to discuss your project.