Cardiotoxicity is one of the leading causes of drug withdrawal from the market (Laverty et al., 2011; Weaver and Valentin, 2018; Mamoshina et al., 2021). Concerns about cardiovascular safety have led to a substantial reduction in drug approvals from different therapeutic areas in the last decade, leading to a sharp increase in costs associated with such failure (Ferdinandy et al., 2019). Mechanisms behind drug-induced cardiotoxicity are multiple and not yet completely understood. One of the most common and well-investigated aspects of cardiotoxicity is drug-induced prolongation of the QT_c interval of the electrocardiogram (ECG). Drug-induced long QT syndrome is caused by an alteration of the action potential in cardiomyocytes, most often linked to repolarization prolongation due to inhibition of the human Ether-à-go-go-Related Gene (hERG) encoded delayed rectifier potassium channel (I_Kr), which can lead to life-threatening cardiac arrhythmias, such as Torsade de Pointes (TdP) (Tisdale et al., 2020). Furthermore, impairment of structural integrity and cardiac dysfunction also represent common cardiotoxic effects, such as in the seminal case of the hepatitis C drug BMS-986094, which induced heart failure in patients (Ahmad et al., 2015). In some circumstances, drug-induced cardiotoxicity might only occur upon chronic drug treatments, such as in the case of anthracycline treatment for cancer. Doxorubicin, one of the most commonly used anthracyclines in the treatment of hematologic and breast malignancies, can induce heart failure months to years after administration (Wenningmann et al., 2019). Compounds can also affect the trafficking, assembly and function of key proteins including ion channels (e.g., hERG channel) in cardiomyocytes. A representative drug that impairs hERG trafficking is pentamidine, an antiprotozoal agent used to treat trypanosomiasis, leishmaniasis and Pneumocystis carinii pneumonia in HIV-infected patients (Rybniker et al., 2010). Cases of QT_c prolongation and even TdP were reported to occur only after several days following intravenous administration of pentamidine (Eisenhauer et al., 1994; Girgis et al., 1997). Subsequent mechanistic studies determined that pentamidine inhibits the trafficking and maturation of hERG, with an IC₅₀ close to therapeutic doses, upon 24 h incubation in hERG-expressing HEK293 cells (Kuryshev et al., 2005). Arsenic trioxide (As₂O₃) is a naturally occurring metalloid used in the treatment of relapsed or refractory acute promyelocytic leukemia (Zhu et al., 2019). Environmental or clinical exposure to arsenic can lead to serious side effects, such as ventricular arrhythmias (including TdP), and sudden cardiac death (Barbey and Soignet, 2001; Westervelt et al., 2001). As shown in guinea-pig ventricular myocytes, As₂O₃ reduces the membrane expression of hERG and, therefore, prolongs the action-potential duration at clinically relevant concentrations, thus causing cardiotoxicity (Ficker et al., 2004).

In view of the effort of reducing the number of animal studies (in line with the 3Rs -Replacement, Reduction and Refinement-principles) (Han, 2023), the use of in vitro models to predict cardiotoxicity has grown exponentially (Narkar et al., 2022). Human-induced pluripotent stem cell-derived cardiomyocytes (hiPSC-CMs) were able to successfully recapitulate the acute and chronic drug-induced cardiotoxicity of many classes of compounds. For example, extensive characterization of doxorubicin-induced cardiotoxicity was conducted and found to be recapitulated in hiPSC-CMs, showing good translational value of in vitro studies (Wenningmann et al., 2019). Another class of anticancer drugs linked to cardiotoxicity are tyrosine-kinase inhibitors (TKi). TKi are small molecules that inhibit the phosphorylation activity of hyperactive receptor tyrosine kinases, often overexpressed in cancer cells (Brandão et al., 2022). TKi-induced cardiotoxicity does not appear to be a class-specific effect, with some compounds linked to adverse outcomes in patients (e.g., sunitinib, nilotinib), whilst others have a relatively safer profile (e.g., erlotinib). While the mechanisms behind this are not yet completely understood, hiPSC-CMs recapitulated the phenotype expected from toxicities in man, as shown by Sharma et al. with a panel of 21 FDA-approved TKi (Sharma et al., 2018).

Thanks to the relatively good versatility of hiPSC-CMs, diverse technologies have been applied to investigate drug-induced cardiotoxicity in vitro (Guo et al., 2015). For example, functional recordings of hiPSC-CMs electrophysiological parameters and Ca²⁺-handling can be conducted through voltage- and Ca²⁺-sensitive dye measurements. Fluorescent dye-based recordings work particularly well for acute measurement but show some limitations in a chronic setting and with repeated measure studies (Kopljar et al., 2018). Motion field imaging (Kopljar et al., 2017) and impedance-based assays (Guo et al., 2018) have been demonstrated to detect the delayed effect of compounds on hiPSC-CMs, but these platforms provide an index of contractile function and no information on the underlying electrophysiological effects of test compounds. multielectrode array (MEA) platforms allow the recordings of the electrical waveform signals (field potential, FP) generated and shaped by monolayers of hiPSC-CMs (Sala et al., 2017). The FP contour correlates with the cardiac AP and, to some extent, with body surface electrocardiogram (ECG) recordings (Tertoolen et al., 2018). One of the main advantages of the MEA is the possibility of performing repeated measurements in a non-invasive way in a medium/high throughput fashion, rendering it ideal to investigate electrophysiological alterations induced by chronic exposure to compounds. Cytotoxic compounds such as doxorubicin, often cause the complete cessation of beating of hiPSC-CMs monolayers or the disappearance of the repolarization wave (also called T-wave-like repolarization wave analogous to human ECG), impairing the analysis of FP parameters. In the present study, we therefore developed and validated a functional assay that combines electrophysiological recordings of hiPSC-CMs (CDI iCell²) by MEA with cytotoxic readouts (impedance), by assessing acute (2 h) and chronic effects (up to 96 h) of thirteen reference compounds. Additionally, we used cytotoxicity assays on hiPSC-CMs, to determine the level of activation of necrotic or apoptotic cell death and to confirm the functional findings on the MEA assay with three reference compounds, with well-known cardiac outcomes in human.

In the present study of in vitro assessment of chronic drug-induced cardiotoxicity, both functional electrophysiological readouts and cytotoxicity measurements were integrated at acute and chronic time points (up to 96 h) using a set of 13 well published clinical reference drugs. The study describes an in vitro approach using hiPSC-CMs to capture different mechanisms of drug-induced cardiotoxicity (including impaired ion-channel trafficking and cytotoxicity) from both electrical and impedance signals and confirmed by live-cell count, apoptosis and necrosis activation assays.

Drug-induced cardiotoxicity is associated with complex and diverse phenomena that can arise both from an impairment of cardiac electrophysiology and the direct loss of cardiac function (or a combination thereof). Traditionally, within the pharmaceutical industry, the primary focus of the risk assessment process is the evaluation of compounds effect on both ion channels in vitro and the QT interval of the ECG in vivo, as prolongation of the latter interval is known to potentially predispose to life-threatening arrhythmias. Currently, scrutiny of potential delayed drug-induced cardiac effects [which may even arise with unidentified modes of action with new therapeutic modalities, such as in the case of monoclonal antibodies (Dempsey et al., 2021)] is on the increase. It seems therefore important to assess compounds both acutely after dosing and at chronic time points. Preferably this is done at the earliest stages of drug development as possible, to avoid high development costs and potential risks in clinical studies.

In the present study, hiPSC-CMs were used in a functional in-vitro assay based on MEAs technology and the changes in the field-potential duration (FPD) and impedance over time, as a surrogate of the QT interval and, impedance, as an indicator of cytotoxicity were assessed up to 96 h. The interpretation of the effects of compounds vs. vehicle-treated wells and the choice of parameters to focus on was based on non-parametric tolerance intervals of vehicle experiments taken from 15 plates and using four different hiPSC-CMs batches.

The first example of a drug that induces delayed electrical changes is pentamidine, an antimicrobial compound used in the treatment of protozoal infections (especially in HIV patients). This drug was reported to induce QT prolongation, ventricular arrhythmias (including TdP) and sudden cardiac death in man (Girgis et al., 1997). The majority of the TdP cases occurred after prolonged intravenous administration of the drug, and only hours or days after the infusions (Eisenhauer et al., 1994), with exposures ranging from 1 to 5 µM (Sands et al., 1985; Conte et al., 1986; Cordes et al., 2005). In fact, in vitro studies revealed that pentamidine can reduce I _Kr density with an IC₅₀ of 5–8 µM in heterologous expression systems, only after (at least) 24 h incubation, due to an inhibition of hERG channel trafficking (Cordes et al., 2005; Kuryshev et al., 2005). In this study, pentamidine induced a time- and dose-dependent prolongation of FPD and EAD-like events starting at 1 μM, in line with clinically reported exposures. A stronger effect on FPD was observed at 3 μM, a concentration that also induced cytotoxic effects at 48 h post compound addition, consistent with previously reported in vitro data (Cordes et al., 2005; Asahi et al., 2019). Interestingly, inhaled pentamidine (nasal spray formulations) exposures are often in the nM range and those have not been associated to either QT prolongation or TdP (Thalhammer et al., 1993). Correspondingly, we demonstrated that the lowest doses used in the study (i.e.,: 0.3–0.1 µM) did not result in a significant prolongation of the FPD.

Another example of a drug with similar delayed effects is arsenic trioxide (As₂O₃), which is used to treat acute promyelocytic leukemia. This compound has also been associated with cardiac arrhythmias and sudden cardiac death (Westervelt et al., 2001; Vineetha and Raghu, 2019). Like pentamidine, As₂O₃ impairs the trafficking of hERG thus explaining the link to life-threatening cardiac arrhythmias (Vineetha and Raghu, 2019). In our study, after 24 h treatment, As₂O₃ induced a prolongation in FPD, more pronounced in beat-rate corrected FPD, at relevant therapeutic concentrations. In fact, FPDc was significantly prolonged at 1 and 3 μM, consistent with a therapeutic fC_max of 0.3 µM. A mild cytotoxic effect, demonstrated by a significant reduction in impedance, was also observed at the highest concentration of 3 µM and at the latest timepoint.

A third reference drug we tested was probucol, which is used in the clinic to lower LDL and HDL cholesterol. This compound is also reported to interfere in vitro with hERG and K_V7.1 channel trafficking (Taniguchi et al., 2012). However, in our study we did not observe a significant prolongation of FPD or FPDc, even at the highest concentration (i.e., 10 μM, 25-times the fC_max). Although probucol is reported to reduce hERG expression in the plasma membrane without directly blocking hERG channel (Shi et al., 2015) and to prolong QT and cause TdPs in patients and in animal studies (Nogawa et al., 2013), the prolongation of the QT interval is not as evident as after treatment with other drugs, such as from arsenic trioxide and/or pentamidine. In men, Hayashi et al. (2004) reported the occurrence of TdP and excessive QT prolongation during a 3-month treatment with probucol in a long-QT syndrome patient carrying a missense hERG mutation. In an older study, authors suggested that probucol-induced lengthening of the QTc and subsequent arrhythmias occurrence is linked to several risk factors, such as female sex, longer QTc before treatment and low serum albumin level (Ohya et al., 1993). Taken together, it appears that QTc prolongation and TdP induced by probucol might be more evident only in patients with concomitant underlying risk factors (e.g., congenital long QT syndrome, female sex, etc.). An in vitro study using human embryonic stem cell-derived cardiomyocytes showed that 24 h-treatment with probucol (10 µM) only prolonged FPDc by ∼16%, compared to 4% in vehicle-treated wells, while in contrast large prolongation (62%) was observed with pentamidine (10 µM) (Hihara et al., 2014). From a mechanistic point of view, we can speculate that the increase in the beating rate and shortening of the FPD reported in our study might be caused by an effect of probucol on the composition of the membrane lipid content. It has been shown, in fact, that in rabbit ventricular myocytes, cholesterol contributes to the regulation of L-type Ca²⁺-current and its modulation by β-adrenergic signaling (Tsujikawa et al., 2008). In general, cholesterol plays a crucial role in the modulation of ion-channels membrane localization and function (Zakany et al., 2020), therefore the effect on hERG trafficking might have been concealed by a reorganization of ion-channels macromolecular complexes. Furthermore, as demonstrated by Taniguchi et al., probucol causes, additional to its effect on I _Kr, an inhibition of I _Ks-encoding K_V7.1 channel trafficking (Taniguchi et al., 2012). In hiPSC-CMs, I _Ks only marginally contributes to the repolarization process (Zeng et al., 2019), thus the effect of probucol on repolarization prolongation due to the trafficking of I _Ks and I _Kr currents is more pronounced in freshly isolated cardiomyocytes or in vivo.

Taken together, the assay presented in the study was able to recapitulate the effect of two major hERG-trafficking inhibitors, pentamidine and arsenic trioxide, showing a prolongation of FPD and arrhythmic events, at relevant therapeutic exposures, but not probucol, which may be consistent with the mixed reports of trafficking effect in the literature.

In addition, as already mentioned, drug-induced cardiotoxicity can occur through loss of viable cardiomyocytes in the heart, leading to impaired cardiac function and heart failure. This is the case of for several classes of oncological compounds and antiviral drugs.

Doxorubicin is an effective chemotherapeutic agent used to treat different types of cancer (Octavia et al., 2012). Nevertheless, doxorubicin administration has been strongly correlated to heart failure in a significant portion of patients. Left-ventricular ejection fraction (LVEF) decreases in a dose-dependent manner, while congestive heart failure occurs in up to 48% of patients treated with doses of doxorubicin above 700 mg/m² (Wenningmann et al., 2019). In the present study, we demonstrated that doxorubicin induced a concentration-dependent drop in the impedance signal, together with cessation of beating at therapeutic exposures. Furthermore, the cytotoxicity assays also confirmed a reduction in the live cells count, driven primarily by apoptosis, as shown by caspases 3 and 7 activation. Interestingly, doxorubicin 0.3 µM, at clinically relevant free plasma concentrations, induced a strong decrease in the impedance signal at 72 h (of ∼70%), while the live cells assay showed a drop of ∼40%. Interestingly (as summarized in Supplementary Figure S5), significant changes in all electrophysiological and cytotoxicity parameters (except for necrosis) occur at concentrations lower than fC_max. This suggests that the impedance assay also detects changes in shape and morphology of the hiPSC-CMs monolayer, as well as cell detachment from the electrodes, as also demonstrated by others (Lei, 2014).

Another example of a compound with direct cardiotoxicity is BMS-986094, which was developed to treat hepatitis C infections. The compound was stopped after a phase II clinical trial in which several patients developed delayed alteration in the ECG and reduced LVEF within 6 months after administration of the compound, even leading to the death of one patient (Ahmad et al., 2015). In the present study, BMS-986094 induced changes in the FP morphology, including the flattening of the T-wave and shortening of FPD, compatible with the effect of this compound on calcium-transient amplitude reported earlier (Yanagida et al., 2021). We also detected a significant decrease in impedance at relevant concentrations, suggesting a cytotoxic effect of this compound on hiPSC-CMs. This is consistent with other in vitro data (Yanagida et al., 2021), with results in cynomolgus monkeys (Gill et al., 2017) and, most importantly, with pathological findings in patients involved in the phase II clinical trial (Ahmad et al., 2015). Due to the unavailability of exposures obtained in the phase II clinical trial, a comparison with the concentrations used in the present study is not possible. Nevertheless, an estimated dose of 3 µM caused cardiotoxicity in cynomolgus monkeys, in line with our results in vitro (Gill et al., 2017).

The third example of cardiotoxicity are a series of compounds all in the class of tyrosine-kinase inhibitors (TKi), which have been developed to treat several types of cancer (Huang et al., 2020). Toxicity can arise from on- and off-target effects, and not necessarily class-specific. In fact, clinical and in vitro studies demonstrated that some TKi are linked to severe cardiotoxicity, such as sunitinib, nilotinib and vandetanib (Orphanos et al., 2009). However, compounds of the class of TKis do not all have the same degree of toxicity and some clinical studies have shown that, for example, the TKi erlotinib has a safer profile (Brandão et al., 2022). In the present study we demonstrated that sunitinib, nilotinib and vandetanib have different cardiotoxic effects that included the generation of arrhythmic events, marked prolongation of the FPD and quiescence. Interestingly, while vandetanib induced a significant decrease in impedance at the highest concentration tested, sunitinib (up to 1 µM) did not show cytotoxic effects on the impedance readout, nor in the live cell count. In contrast, a profound effect on hiPSC-CMs viability (live cell count) was observed at 10 μM, mainly driven by necrosis pathway activation. As summarized in Supplementary Figure S5, significant changes in FPD, BR and FPDc occurred (for the 48 h time point) at 1 and 3-fold the therapeutic fC_max value (and within the minimum-maximum fC_max value reported), while a decrease in live-cell count only occurred at a concentration higher than 300-times the median fC_max. In this assay we did not detect any significant effect on hiPSC-CMs electrophysiology or cytotoxicity after erlotinib incubation, even up to ∼ three times the estimated fC_max, which is in agreement with clinical (Brandão et al., 2022) and preclinical data (Sharma et al., 2017).

The combined MEA assay presented in this study recapitulated several mechanisms of drug-induced cardiotoxicity. By digging into the molecular mechanisms of a selected set of compounds using cytotoxicity assays, we demonstrated that the functional impedance parameter predicts apoptosis- or necrosis-driven cell death and the morphological changes in hiPSC-CMs monolayers preceding cell death. This combined approach of electrophysiology and impedance offers the possibility to more comprehensively assessing the chronic effects of compounds, using a medium-throughput, human-based platform, suitable for further upscaling (up to 96-well MEAs plates are available). Furthermore, integration of the impedance measurement with MEA electrophysiological parameters allows for a deeper understanding of the mechanisms causing drug-induced cardiotoxicity, such as in the case of TKis presented in this study. In fact, sunitinib and vandetanib had a similar electrophysiological profile characterized by excessive FPD prolongation and arrhythmias, but only vandetanib induced cytotoxic effects. Similarly, doxorubicin caused cessation of beating and the disappearance of the T-wave, a phenomenon common to different mechanisms of action of drugs [such as hERG- or sodium-channel block (Blinova et al., 2018)]. The impedance readout allowed us to understand that cytotoxic effects were driving the changes in the FP, which was further corroborated by the live-cell count assay.

As cardiotoxicity is still one of the leading causes of drug withdrawal from clinical development (or in some cases even from the market), it is important to integrate a human-based chronic evaluation of compounds as early as possible in development, in order to predict potential effects on protein (e.g., ion channels) trafficking, drug-induced transcriptional modulations and cytotoxicity. Currently, the only opportunity to detect delayed cardiotoxicities is generally from long-term toxicology studies on animals. Often the design of such toxicity studies is suboptimal for sensitive QT-effect detection. Moreover, ex-vivo/in vitro cardiac tissues from animals have a limited acute physiological timespan within which to study the effects of a drug. Hence there are preclinical opportunities for sensitive in vitro human based assays to detect these effects and reduce animal use. In the future, integration of a chronic hiPSC-CMs assay with the classical approach to safety study from ICH E14/S7B (International Council for Harmonization) could contribute to a more timely detection of cardiac liabilities earlier in the drug development process, ultimately contributing to a reduction in development costs, impact of drug withdrawal and increase the level of safety for patients.

The present in vitro study used hiPSC-CMs derived from one single human donor and thus does not represent the genetic diversity of a population of humans and additionally does not mimic the complete in vivo cellular complexity of the 3D heart (including atrial/ventricular and other diversity, complex feedback/feedforward mechanisms involved in the cardiovascular functioning of the body). Furthermore, hiPSC-CMs cultured in a dish lack inflammatory and immune system interactions, which are also known to be linked to drug-induced cardiac toxicities in humans. In addition, compared to freshly isolated human cardiomyocytes, hiPSC-CM cells do have some electrophysiological and structural limitations. In fact, hiPSC-CMs beat spontaneously due to a reduction in the expression of Kir2.x channels, responsible for the inward rectifier potassium current I _K1 (Goversen et al., 2018), higher expression of HCN “pacemaker” channels, as compared to adult human cardiomyocytes (Ma et al., 2011). From a structural point of view, the sarcomeric-structure of hiPSC-CMs does not fully resemble the longitudinally organized one of adult cardiomyocytes (Ma et al., 2011). Finally, the actual presence of compounds in the wells was not examined and hence may be somewhat lower than nominal (given the cell culture medium contains 10% bovine serum). Consequently, although there is debate about how to best calculate margins from in vitro to in vivo, this may affect the accuracy of the margin comparison between the in vitro nominal concentrations and clinical plasma fC_max.

In the present study, we described and characterized a functional MEA assay using hiPSC-CMs evaluating the effect of vehicle (DMSO 0.1%) and 13 clinical reference compounds on field potential. By combining the assessment of electrophysiological parameters and cytotoxicity (through impedance measurement) we were able to capture different mechanisms of drug-induced cardiotoxicity, including impaired ion-channel trafficking and direct cytotoxicity, at relevant therapeutic concentrations. This chronic approach can complement acute (high throughput) assays for the prediction of proarrhythmic liabilities of developing compounds, used in for, e.g., the CiPA initiative (Blinova et al., 2018). The early deselection of compounds based on hazard identification of chronic cardiotoxicities will help to avoid late-stage detection in vivo studies or even in the clinic.