Abstract
High-throughput in vitro drug assays have been impacted by recent advances in human induced pluripotent stem cell-derived cardiomyocytes (hiPS-CMs) technology and by contact-free all-optical systems simultaneously measuring action potential (AP) and Ca2+ transient (CaTr). Parallel computational advances have shown that in silico models can predict drug effects with high accuracy. In this work, we combine these in vitro and in silico technologies and demonstrate the utility of high-throughput experimental data to refine in silico hiPS-CM populations, and to predict and explain drug action mechanisms. Optically-obtained hiPS-CM AP and CaTr were used from spontaneous activity and under pacing in control and drug conditions at multiple doses.
An updated version of the Paci2018 model was developed to refine the description of hiPS-CM spontaneous electrical activity; a population of in silico hiPS-CMs was constructed and calibrated using the optically-recorded AP and CaTr. We tested five drugs (astemizole, dofetilide, ibutilide, bepridil and diltiazem), and compared simulations against in vitro optical recordings.
Our simulations showed that physiologically-accurate population of models can be obtained by integrating AP and CaTr control records. Thus constructed population of models predicted correctly the drug effects and occurrence of adverse episodes, even though the population was optimized only based on control data and in vitro drug testing data were not deployed during its calibration. Furthermore, the in silico investigation yielded mechanistic insights, e.g. through simulations, bepridil’s more pro-arrhythmic action in adult cardiomyocytes compared to hiPS-CMs could be traced to the different expression of ion currents in the two.
Therefore, our work: i) supports the utility of all-optical electrophysiology in providing high-content data to refine experimentally-calibrated populations of in silico hiPS-CMs, ii) offers insights into certain limitations when translating results obtained in hiPS-CMs to humans and shows the strength of combining high-throughput in vitro and population in silico approaches.
Significance We demonstrate the integration of human in silico drug trials and optically-recorded simultaneous action potential and calcium transient data from human induced pluripotent stem cell-derived cardiomyocytes (hiPS-CMs) for prediction and mechanistic investigations of drug action. We propose a population of in silico models i) based on a new hiPS-CM model recapitulating the mechanisms underlying hiPS-CM automaticity and ii) calibrated with all-optical measurements. We used our in silico population to predict and evaluate the effects of 5 drugs and the underlying biophysical mechanisms, obtaining results in agreement with our experiments and one independent dataset. This work supports the use of high-content, high-quality all-optical electrophysiology data to develop, calibrate and validate computer models of hiPS-CM for in silico drug trials.
Introduction
Both, new in silico methods and the use of human induced pluripotent stem cell-derived cardiomyocytes (hiPS-CMs) have become increasingly important in tackling the challenge of assessment and prediction of drug effects and their potential cardiotoxicity, as supported by the Comprehensive In Vitro Proarrhythmia Assay (CiPA) initiative (1, 2). Many in silico studies on this topic have been published in recent years, showcasing a variety of methodologies, including electrophysiological models of cardiac cells, machine learning algorithms, and a combination of both (3–8). The potential of hiPS-CMs for drug-induced pro-arrhythmia predictions in vitro has been shown in many experimental studies (9, 10) despite certain outstanding limitations. Concerns lie with their high inter-lab and inter-batch variability and level of maturity compared to adult cardiomyocytes (11), e.g. spontaneous beating, cell morphology, disorganization of their contractile elements (12), and different ion channel expression (13). Nevertheless, hiPS-CMs represent the best experimental platform to date to study human cardiac electrophysiology and drug action in a rigorous and scalable/high-throughput way. In silico models of hiPS-CMs have emerged (14–17) as an invaluable tool to better understand the distinct ionic mechanisms underlying hiPS-CM’s drug response (18, 19). The robustness of in silico models depends on the amount and the quality of the experimental data used in their calibration and validation. Traditionally, such data have been acquired from a limited number of isolated cells (outside of their multicellular environment), through time-demanding and tedious manual patch-clamp techniques.
Limited experimental data present challenges of not being able to capture the genotypical and the phenotypical variability observed in a cell population, which is especially relevant for the highly-variable hiPS-CMs. These challenges have been partially addressed through modelling and data curation. In silico population of models approaches have been developed to reflect the wider range of parameters beyond the limited experimental data (20, 21). Database merging has also been used in the desire to expand the experimental data needed to tune the model parameters, e.g. in (19, 22) we merged 6 in vitro datasets of action potential (AP) biomarkers to generate a population of in silico hiPS-CMs. Using data from different laboratories widen the data variability considerably.
On the technology side, the problem of limited experimental data has been tackled by new experimental techniques with increased throughput and amenable to automation, e.g. automated patch-clamp platforms (23, 24) or microelectrode arrays (MEAs) (13). However, these techniques still suffer the limitations of probe-sample physical contact, which limits their performance with hiPS-CMs (25). Contact-free optical recordings overcome these limitations and offer comprehensive characterization. Calcium and contraction-measurement systems have been leveraged for cardiotoxicity testing (26). Ahola et al. (27, 28) developed a video-based contact-free method to quantify the biomechanics of beating hiPS-CMs, by processing simultaneous recording of motion and Ca2+ transients (CaTr) from fluorescence videos. However, AP signals represent key aspects of cardiotoxicity responses that may not be captured by field potentials, CaTr or mechanical contractions. All-optical electrophysiology (29, 30) approaches offer contactless interrogation and high-throughput records of voltage and calcium in an attempt to increase information content. Application of these techniques to drug screening with hiPS-CMs have been successfully demonstrated (25, 31, 32), including our OptoDyCE that combines optical pacing and simultaneous optical records of voltage and calcium or contractions. The use of optical systems with hiPS-CMs preparations provides an abundance of in vitro data with the potential to provide an excellent basis to construct experimentally-calibrated population of in silico hiPS-CMs. The value of such high-content optical recordings of CaTr and AP (without ion channel level data) to constrain in silico populations of models remains to be tested.
The main goal of this work was to demonstrate the utility of in silico simulation trials informed by all-optical cardiac electrophysiology (optically-obtained high throughput measurements of AP and CaTr from hiPS-CMs under spontaneous and optically-triggered conditions) for prediction and mechanistic understanding of drug action. Optically-obtained AP and CaTr measurements are used to guide and improve the design and calibration of a population of in silico hiPS-CMs. We then test the performance of in silico simulation trials with the populations of models against in vitro drug trials for 5 reference compounds, both in terms of their consistency and to deepen the mechanistic insights unravelled. In detail: i) We present an improved version of the Paci2018 hiPS-CM model (15), providing improved simulation of the Na+/Ca2+ exchanger (INCX) role in sustaining the automaticity of AP. ii) We use high throughput optical measurement of AP, CaTr alone and both to calibrate an in silico population of hiPS-CMs models. iii) We challenge this population by applying 5 reference compounds at multiple concentrations, and comparing the results against in vitro data, not used for the calibration step. We investigate the mechanisms underlying the different response to bepridil in hiPS-CMs (both in vitro and in silico) compared to adult cardiomyocytes.
Materials and Methods
Experimental dataset
The experimental dataset consists of AP and CaTr recordings from hiPS-CMs syncytia (CDI iCell2 cardiomyocytes) obtained with the all-optical OptoDyCE system (25) in a 384-well plate format at room temperature (21°C) and with extracellular concentrations Nao = 135.0, Ko = and Cao = 1.33mM, in both paced and non-paced conditions. Recordings were performed in control conditions (0.1% DMSO) and following application of 5 reference compounds: astemizole (antihistamine), dofetilide (antiarrhythmic agent, class III), ibutilide (antiarrhythmic agent, class III) bepridil (antiarrhythmic agent, class IV) and diltiazem (antiarrhythmic agent, class IV).
Control recordings were performed on 10 plates (384-well format). Voltage and calcium-derived biomarkers were obtained from 5 independent multicellular samples per plate (each sample having at least 200 cells). The following biomarkers were considered: AP and CaTr cycle length (AP CL and CaTr CL), duration at 30%, 50% and 90% of AP repolarization (APD30, APD50 and APD90) and of CaTr decay (CTD30, CTD50, CTD90), AP and CaTr triangulation (AP Tri90-30=APD90-APD30 and CaTr Tri90-30=CTD90-CTD30) and CaTr time from CaTr onset to peak (CaTr tRise0,peak). Each measurement was characterized by its mean value (mean) and its standard deviation (SD) over a variable number of beats for each multicellular sample. Some acquisitions failed, and were discarded from the dataset, leading to a total of 42 control non-paced and 49 control paced multicellular samples, thus integrating responses from over 8400 cells. Min and Max experimental ranges for each biomarker were computed by defining a lower and upper bounds (LB = min(mean − 2 ∗ SD) and UB = max(mean − 2 ∗ SD), respectively), for non-paced and paced measurements, as reported in Table 1.
Reference compounds were tested in 5 plates (one for each drug), considering 4 increasing doses (D1, D2, D3 and D4) and 6 multicellular samples for each dose (thus integrating responses from at least 1200 cells per drug dose). After discarding failed recordings, we used the same methods as in control to compute the experimental biomarker ranges.
Updated version of the Paci2018 hiPS-CM model
A limitation of the Paci2018 hiPS-CM model (15) was noted - namely, failure to reproduce the cessation of the spontaneous electrical activity following strong block of the INCX, as shown by recent in vitro and in silico experiments (16, 33). A very large window current in Paci2018 for the fast Na+ current (INa) was identified as the key to sustaining the automaticity upon INCX block. We improved the Paci2018 model to reproduce this specific mechanism, while preserving all its good features. We kept the same structure of the Paci2018: the model includes two compartments, namely cytosol and sarcoplasmic reticulum (SR), and it follows the classical Hodgkin & Huxley formulation, which describe the membrane potential as where C is the membrane capacitance V the membrane voltage and Istim the stimulus current. The ion current/pumps in the model are: INa, the late Na+ current (INaL), the funny current (If) the L-type Ca2+ current (ICaL), the transient outward K+ current (Ito), the rapid and slow delayed rectifier K+ currents (IKr and IKs), the inward rectifier K+ current (IK1), the Na+/Ca2+ exchanged (INCX), the Na+/K+ pump (INaK) the sarcolemmal Ca2+ pump (IpCa) and the Na+ and Ca2+ background currents (IbNa and IbCa). The SR compartment exchanges Ca2+ with cytosol through three fluxes: RyR-sensitive release current (Irel), SERCA pump (Iup) and leakage current (Ileak).
To develop the Paci2019 model (details in the Supporting Material):
we updated the formulations for INa and If with the ones proposed in (16);
we optimized the model parameters to fit the same dataset of in vitro AP and CaTr biomarkers used for (15), which have been recorded at 37°C;
we validated the model against the same experimental protocols used for (15).
As a result, we obtained an improved version of our hiPS-CM model (Paci2019), where the spontaneous electrical activity is triggered both by If and Ca2+ release from the sarcoplasmic reticulum, which in turn depolarize the membrane potential via INCX. Details on the optimization procedure are reported in the Supporting Material, together with the model parameter values and equations.
The optically-obtained in vitro data in this paper were recorded at 21°C and with extracellular concentrations (Nao = 135.0, Ko = 5.4 and Cao = 1.33mM instead of Nao = 150.0, Ko = 5.4 and Cao = 1.8mM). Consequently, we implemented temperature correction of the new Paci2019 model to these conditions. Temperature difference was managed by setting the correct temperature in the specific model parameter affecting the Nernst potentials and ion currents such as INCX or INaK, rescaling the time constants of the other main ionic currents by means of the Q10 factors reported in (34–37), and summarized in Table S1 in the Supporting Material.
hiPS-CM in silico population calibrated with optical AP and CaTr recordings
The new Paci2019 model, adapted to the temperature and extracellular concentrations of the optical recordings, was used as baseline to construct a population of in silico hiPS-CMs, based on the population of models methodology (19, 20, 38). We sampled a total of 22 parameters in the [50-150]% range compared to their original values. Parameters were chosen similarly to (39), to include all the main ionic conductances, as well as key kinetics parameter, known to impact both AP and CaTr biomarkers: (i) the maximum conductances of INa, INaL, If, ICaL, Ito, IKs, IKr, IK1, INCX, INaK, IpCa, Irel, Iup; (ii) activation and inactivation time constants of INa, ICaL and Irel; (iii) adaptation time constant and half inactivation Ca2+ concentration of Irel; (iv) Iup half saturation constant. An initial population of 30,000 hiPS-CMs was generated, and then calibrated based on the optical recordings, i.e. only the models whose biomarkers were in agreement with the in vitro data were maintained. Biomarkers were computed in steady state (after 800s), as the average on the last 20 beats. The lack of absolute amplitude values for AP in the optically-recorded data was handled by an additional biomarker to constrain the amplitude of the non-paced AP (AP peak between 17.0 and 57.7 mV), as in (19).
Three different calibration options were performed considering both paced and non-paced biomarkers, thus generating three different experimentally-calibrated populations: i) All AP and CaTr biomarkers (AP_CaTr population); ii) AP biomarkers only (AP_only population); iii) CaTr biomarkers only (CaTr_only population). The three populations were compared to investigate how the choice of AP and CaTr biomarkers affect the calibration process and the coverage of the biomarker space compared to experimental ranges.
In silico drug trials
In silico drug trials were performed for 5 compounds (astemizole, dofetilide, ibutilide, bepridil and diltiazem) considering the 4 concentrations for each tested in vitro. Drug simulations were run for 400s from steady state conditions. Models were not paced, to also investigate drug-induced effect on the spontaneous beating frequency. We used a simple pore-block drug model as in (3, 19, 38), consisting of IC50 and Hill’s coefficients from literature and reported in Table S2 in the Supporting Material. The experimental concentrations for each drug are reported in Table S3 in the Supporting Material, together with the corresponding percentage of residual currents following drug application and the maximal effective free therapeutic concentration (EFTPCmax), for comparison.
Because of the discrepancy between hiPSC and adult CMs observed for bepridil ((40) vs (3, 41)), only for bepridil 10μM, we run additional tests, reducing its ICaL blocking action to half (64% residual ICaL instead of 32%) and to zero (100% residual ICaL), while preserving its blocking action on the other ion channels. This test was done on 4 models selected among the ones that showed a pro-arrhythmic behaviour when administered astemizole.
Following drug application, we assessed the drug-induced changes on AP and CaTr biomarkers, as well as the occurrence of abnormalities. Single and multiple early after-depolarizations (EADs) were defined as extra-peaks greater than −55mV in between two consecutive AP upstrokes. Repolarisation failure were identified when a stable (dV/dtmax<0.1 V/s) membrane potential > −40 mV was observed during the last 15 s of simulation. Irregular rhythm was identified when the difference in cycle length between two consecutive AP greater than 150%.
We looked also for two additional phenotypes, that we did not consider as abnormalities: quiescence (40) and residual activity (42), mainly occurring during diltiazem administration (see Results). If a model reacted to drug by producing AP whose peaks were greater than −40mV but smaller than 0mV, we labelled the model as residual activity. Conversely, we considered the model quiescent, i.e. not producing anymore spontaneous AP, if during the last 15s the average membrane potential was smaller than −40mV or a potential residual activity had all the peaks smaller than −40mV.
Results
The new Paci2019 hiPS-CMs model
The automated optimization process successfully identified a new Paci2019 model in agreement with the in vitro AP and CaTr biomarkers used in (15), as shown in Table 2. Figure S1 in the Supporting Material shows a detailed comparison between the new model (in black) and the Paci2018 model (in red) (15). Parameter values are reported in the Supporting Material.
The main difference between the two models is the shape of the INCX current. Before the upstroke, the new INCX provides an additional inward contribution (−0.5A/F) that is added to If (−0.25A/F), supporting the membrane depolarization and allowing the opening of the INa channels. Figure 1 illustrates the contribution of INCX to the hiPS-CM automaticity, as reported in (16, 33): blocking INCX reduces its inward component slowing down the rate of spontaneous AP, up to suppression. In particular, an issue in the Paci2018 model was that AP suppression did not happen, in disagreement with in vitro data by Kim et al. (33) in response to 2μM SEA0400, an inhibitor of the forward INCX in a cluster of hiPS-CMs. The large INa window current was identified as a key factor in supporting the automaticity, thus making the Paci2018 model unable to capture the aforementioned mechanism.
The new Paci2019 model can simulate spontaneous Ca2+ release from the SR both with standard extracellular Ca2+ concentration (Cao = 1.8mM, Figure S2 in the Supporting Material) and Ca2+ overload (simulated by increasing the extracellular Ca2+ concentration to Cao = 2.8, 2.9 and 3.0mM, Figure S3 in the Supporting Material). Moreover, it reproduces well the in vitro data by Ma et al. (43) with ion channel blockers (Figure S4 in the Supporting Material), If block and hyperkalemia experiments as (33) (see Supporting Information) and alternans in ischemia-like conditions as (15) (Figure S5 in the Supporting Material). Finally, the CaTr amplitude of 160 nM is in agreement with data by Rast et al. (44), recorded from hiPS-CM ensembles incubated at 37°C (calibrated Fura-2-based photometry measures) and not used for model calibration.
After updates for the extracellular ion concentrations and temperature adjustment, as described in Methods, the Paci2019 model’s AP and CaTr biomarkers moved closer to the optical recordings reported in Table 1 obtained at room temperature. For example, spontaneous CL increased (from 1,712 to 4,144 ms) and APD90 prolonged from 390 to 1,119 ms. Figure 2 shows a comparison of the Paci2019 model (green traces) vs. the same model adapted for extracellular concentrations and temperature (blue traces).
Single dataset calibration vs combined dataset calibration
The Paci2019 model, adapted for the extracellular concentrations and room temperature used in the in vitro experiments, was deployed to generate an initial population of 30,000 models. As described in Methods, 3 different calibrations were performed (using AP only, CaTr only or both AP and CaTr biomarkers), leading to 3 calibrated populations: AP_only, CaTr_only and AP_CaTr, respectively.
A comparison of the AP and CaTr biomarkers for the 3 populations is shown in Figure 3. The AP_only population (green boxplots) consists of 969 models. As expected, it shows good agreement with the experimental AP biomarkers in addition to a good coverage of the experimental ranges, both non-paced and paced (Panel A and B). However, many models have CaTr biomarkers outside the experimental ranges, e.g. CTD90, CTD50 and CTD30 are often too short (Panel C and D). The CaTr_only population (black boxplots) consists of 5,030 models in good agreement with CaTr biomarkers, both non-paced and paced (Panels C and D). However, many models yield AP durations and triangulation outside the experimental ranges (Panels A and B). As expected, the AP_CaTr population, obtained by calibrating with both AP and CaTr biomarkers (blue boxplots), appears to be the best constrained, with 477 models showing good agreement and coverage of the biomarker space.
Figure 4 shows the distributions of the seven parameters with differential responses in the 3 experimentally-calibrated populations (|Δmedian| > 10% between AP_only/CaTr_only and AP_CaTr). Distributions of all parameters varied in the population are shown in Figure S6 in the Supporting Material. Adding AP biomarkers for calibration (AP_only and AP_CaTr populations vs. CaTr_only) helps adjust five key parameters in important ways (lowers their median values): GNa, INa inactivation time constants, GK1, INCX maximum current and ICaL inactivation time constant (Figure 4). The smaller GNa is due to the upper limit on AP peak. This also imposes a smaller INa inactivation time constant (faster inactivation), further contributing to reduced AP peak amplitude. A lower GK1 results in a slightly depolarized MDP, consequently reducing INa availability, and again limiting the AP peak. A reduced INCX maximum current prevents an excessively fast early repolarization phase, e.g. short APD30. Finally, a smaller ICaL inactivation time constant speeds up ICaL inactivation, thus limiting excessively long AP.
Considering CaTr biomarkers for calibration (CaTr_only and AP_CaTr, vs AP_only) increases the median values for two calcium-release parameters: Irel inactivation time constant and Iup half saturation value (Figure 4). The first causes a slower inactivation of Irel, and consequently a longer CaTr (Figure 3, Panels C-D). The latter, that appears in the denominator of the Iup formulation (15), causes a reduction of Ca2+ uptake, thus also contributing to a longer CaTr.
Overall, these results reveal important information contributed by the AP or CaTr biomarkers in the calibration process to better capture the experimental recordings. For the rest of this study, including the in silico drug trials, only the AP_CaTr population of 477 hiPS-CM models was considered. The AP and CaTr traces for this population are shown in Figure 5.
In silico drug trials
Using the population of 477 hiPS-CM models shown in Figure 5, calibrated with both experimental AP and CaTr biomarkers, we ran in silico drug trials for 5 reference compounds (astemizole, dofetilide, ibutilide, bepridil, diltiazem) at 4 increasing concentrations (D1-D4) each. Simulation results were validated against the corresponding in vitro experiments, which were not used during the calibration process. For each drug trial, we checked how the drug affects the AP and CaTr biomarkers compared to control (D0), and assessed the presence of drug-induced abnormalities. Figure 6 summarizes the drug effects on four AP and CaTr biomarkers (AP CL, APD90, CTD90 and CaTr Tri90-30). Shown are: i) in silico biomarker boxplots for the models that after drug administration still produce spontaneous AP and CaTr at room temperature and at the ion concentrations tested in vitro; and ii) in vitro optically-recorded biomarkers (green/purple diamonds) and their variability ranges (green/purple bars). Results for all biomarkers are shown in the Supporting Information, Figure S7-11 in the Supporting Material.
Our in silico population, calibrated with optically-recorded biomarkers in control conditions only, reproduces successfully the drug-induced changes in the AP and CaTr biomarkers. If no in vitro biomarkers are reported for a specific dose, it means that the drug stopped the spontaneous activity in in vitro hiPS-CMs.
The four drugs (astemizole, dofetilide, ibutilide, bepridil), causing a strong IKr block, induced AP and CaTr prolongation. In particular, simulated APDs, CaTr tRise0,peak, and AP and CaTr Tri90-30 are well within the experimental ranges. Conversely, simulated AP and CaTr CL and CTDs tend to underestimate the prolongation observed in vitro. For diltiazem, a ICaL blocker, simulations reproduced a dose-dependent APD90 shortening. However, the CTD90 prolongation observed in vitro for intermediate doses (D2 and D3) was not captured in silico. Table 3 reports the occurrences of drug-induced repolarisation abnormalities and quiescent phenotypes, both in simulations and experiments.
The in vitro dataset showed overall less abnormalities in hiPS-CMs in response to drugs than the simulations. A likely reason for this could be that in silico results assume single-cell behavior with a wide range of ionic profiles, while syncytial structures were used in vitro, where good cell-cell coupling usually has damping effects on pro-arrhythmic behavior. For the drugs inducing AP prolongation (astemizole, dofetilide, ibutilide and bepridil), the abnormalities recorded in vitro were single or multiple early-afterdepolarizations (EADs), corresponding to the types A, B and C reported in (13). We also observed 3 cases of tachyarrhythmia (rate of spontaneous oscillations>2Hz), 2 for Dofetilide (D3 and D4, following EADs) and 1 for ibutilide (D4). Finally, 9 cases of irregular rhythm were observed: 4 for Dofetilide (D1, D2 and D3), 4 for ibutilide (D1 and D2) and 1 for Bepridil (D1). For diltiazem, the abnormalities observed in vitro were an irregular rhythm at D1, a multiple EAD and irregular rhythm at D2 and a tachiarrhythmic time course at D4 (in 6 out of 6 observations, 1 also with irregular rhythm).
In the simulations for 4 out of 5 tested compounds (astemizole, dofetilide, ibutilide and bepridil) we observed a variety of drug-induced phenotypes, as seen in vitro both in our experiments and in (13). Exemplary in silico traces are shown in Figure 7 and compared to in vitro experiments: single and multiple EADs (panels A, B, C, D), single EADs (panel E, F), repolarization failures (panels G, H) and irregular rhythms (panels I, J, K, L, M, N). Expanded and additional traces are reported in Figure S12 in the Supporting Material. In addition to AP shortening, for diltiazem we observed a residual electrical activity, characterized by low-amplitude oscillations and an EAD (Figure S13 in the Supporting Material).
For astemizole, in silico results reveal 9 abnormalities at D3, and 43 at D4 (mainly EADs and repolarization failures, but also 5 irregular rhythms per dose); the in vitro data show dose-dependent increase in pro-arrhythmic markers but no arrhythmia events per se at the tested doses. Again, the syncytial nature of the experimental samples and/or lower temperature may have dampened the arrhythmia events. Nevertheless, the simulation results are in agreement with the fact that at clinical doses, this drug is considered as intermediate risk in (40) on hiPS-CMs and at high risk in the in silico drug trials performed in (3) and in CredibleMeds (41). This highlights the value of in silico investigations with broader population of models to complement in vitro experiments, and ability to cover a wide range of ionic profiles.
Simulations of ibutilide and dofetilide closely agree with the experiments. A dose-dependent increase in abnormalities was seen, typical of drugs classified as high risk in CredibleMeds (41) and in hiPS-CMs in (40). The abnormalities in silico are mainly EAD and repolarization failures at the higher doses, together with few cases of irregular rhythm (ibutilide: 1 at D3 and 2 at D4; dofetilide: 1 at D1, 5 at D2, 6 at D3). For dofetilide, at D4 we observed in silico only 5 repolarization abnormalities and 9 irregular rhythms, while all 6 in vitro recordings showed single or multiple early EADs. Therefore, we tested in silico 3 additional doses higher than D4, as in (3), that triggered a considerable amount of EADs (up to 59 EADs/repolarization failures at D7).
Bepridil simulations are in agreement with our in vitro experiments. Bepridil’s main effect on hiPS-CMs is the suppression of spontaneous activity in a high percentage of the population (107/477 and 444/477 models, at D3 and D4, respectively). This is consistent with our in vitro experiments (6/6 observations at D4 did not produce AP) and with other reports (40). Conversely, only few abnormalities were observed in hiPS-CMs: in vitro only 1 irregular rhythm at D1 and in silico 5 and 6 abnormalities (2 irregular rhythms and the rest EADs) for D3 and D4, respectively, in agreement also with (40). However, this is in contrast with the high bepridil toxicity observed for adult cells in vitro and in silico, where it triggers many repolarization abnormalities (3, 41) and might be due to the different expression of ion currents in adult and hiPS-CMs, especially ICaL (13). Therefore, for bepridil only, we tested also the effect of modulating its ICaL blocking power, not changing the drug effect on INa, IKr and INaL. Figure 8 shows four different models that developed abnormalities with astemizole D4, but not with Bepridil D4 (black traces). However, reducing to half bepridil ICaL blocking power was already enough to trigger EADs. The same behavior was observed by fully inhibiting bepridil ICaL blocking effect.
For diltiazem, we observed in silico only 1 EAD at D4 (Figure S13, Panels C), but no tachyarrythmic rhythm, as in our in vitro experiments. In fact, most of our models (Table 3) stopped their spontaneous AP, in agreement with what was observed in (40). However, 20 models at D4 showed a strong decrease in AP amplitude (in few cases peaks were recorded below 0mV) and slight increase of frequency (Figure S13, Panels A and B). This low-amplitude oscillations (or residual activity) of the membrane potential were observed in Zeng et al. (42). Zeng et al. demonstrated that such residual electrical activity is due to a residual availability of INa, not fully blocked by drugs specifically designed to mainly block L-type Ca2+ channels. It is possible that such abnormal re-activation of INa may have triggered re-entrant (tachycardic) responses in our multicellular experiments. In silico results provide further insights that this residual spontaneous electrical activity is due to a combination of residual INa (partly blocked by diltiazem, but still able to trigger an AP), strong If and weak IK1 (Table S4 in the Supporting Material, column RESAC).
Simulation studies were used to better understand biophysical mechanisms underlying the drug-induced phenotypes. We observed that the abnormalities induced by astemizole, dofetilide and ibutilide are mainly repolarization abnormalities, while bepridil and diltiazem mainly stopped the spontaneous activity. Table S4 summarizes the ionic parameter differences, the amount of repolarization abnormalities and residual activity at the maximal dose tested in silico (D4, except D7 for dofetilide). For the cessation of the spontaneous activity, D3 had more balanced groups for bepridil and diltiazem. We focused our analysis only on those groups containing at least 20 models showing non-sinus rhythm. Models developing EADs and repolarization failures in response to astemizole, dofetilide and ibutilide show weak IKs and IK1 compared to the models not developing such abnormalities, highlighting a reduced repolarization reserve. Also IpCa, an outward flow of Ca2+ ions is very small, contributing to accumulation of positive charges in the cytosol. Conversely, a different pattern emerged for the models that terminated their spontaneous activity in response to bepridil and diltiazem. They show, compared to the models still developing AP at D3, a strong IK1 that stabilizes the resting potential. Furthermore, especially for bepridil, the stronger Iup half saturation constant Kup reduces the intake of Ca2+ by the SERCA pump and therefore the Ca2+ available to be released from the sarcoplasmic reticulum, impairing the Ca2+handling that it is now an important component of automaticity in the Paci2019 model. For diltiazem, we found that INa was smaller in models where the drug terminated spontaneous activity compared to the group that still showed spontaneous activity.
Discussion
Here we demonstrate the integration of human in silico drug trials and optically-recorded simultaneous AP and CaTr data from hiPS-CMs for prediction and mechanistic investigations of drug action. We report:
An improved version of the Paci2018 hiPS-CM model (15) was developed and validated. It better reflects the mechanisms underlying AP automaticity.
The value of comprehensive high-throughput optical measurements of cellular responses, especially combining AP and CaTr, is demonstrated in refining in silico populations of models.
The predictive power of the experimentally-calibrated population of hiPS-CMs models is demonstrated through in silico drug trials on 5 drugs and comparison to in vitro datasets.
Mechanistic insights are gleaned from in silico population runs to understand differential responses of hiPS-CM and adult cardiomyocytes to bepridil. Despite observed cardiotoxicity in adult cells (3, 41), in vitro experiments, in this dataset as well as in another independent in vitro dataset (40), showed low occurrence of proarrhythmic markers in hiPS-CMs. In silico trials with the hiPS-CMs models show a wide range of responses to drug action, which complement and explain the in vitro experiments.
Research on hiPS-CMs is rapidly developing, with new experimental data becoming available, which in turn serve as a driving force for the constantly evolving computational models to offer more accurate in silico tools to the scientific community. Based on in vitro (33) and in silico (16) tests, it was identified that our Paci2018 hiPS-CM model (15) did not properly reflect the role of INCX in automaticity, i.e. no cessation of spontaneous activity was seen in the model as consequence of a strong INCX block, as suggested by experiments. Therefore, we updated this hiPS-CM model to reproduce the specific mechanisms reported in (16, 33). (Figures 1, S1 and Supporting Material). In addition, the new Paci2019 model also qualitatively simulates the relationship between changes in CL and APD90 as consequence of the If modulation (Figure S16 in the Supporting Material). The model responds to If augmentation with shorter CL and APD90, while If reduction increases them. In Rast et al. (45), a similar relationship was observed in iCells (CDI) hiPS-CMs field potentials between the inter-beat interval and the field potential duration for ivabradine (for If reduction) and forskolin (for If augmentation).
Using the Paci2019 model to construct an in silico population based on our in vitro optical recordings, we showed that the combination of AP and CaTr biomarkers provides superior calibration, with a better coverage of the biomarker space (Figures 3). It is also interesting that the calibration with AP biomarkers was the most restrictive: AP_CaTr and the AP_only populations contained only 477 and 968 accepted models, respectively, while the CaTr_only population contained over 5,000, many of which were inadequate, e.g. presented extremely short or long APs (Figure S14 in the Supporting Material). Therefore, model calibration exclusively based on CaTr can easily lead to inclusion of more unrealistic models for hiPS-CM. For this cell type, AP biomarkers are preferred to obtain physiological (or semi-physiological) models, while combining both biomarkers clearly refines the calibration. These tests highlight the importance of the calibration process and one of the main advantages of comprehensive records, such as the ones obtained through all-optical cardiac electrophysiology systems like OptoDyCE, that allow the acquisition of AP and CaTr in large populations of cells in their multicellular context.
Figures 6 and S7-11 compare simulated and experimental biomarkers. Of note, the experimental drug trials were not used to calibrate the population of models; yet, the experimentally observed biomarker trends over increasing drug doses, in particular APDs, CTDs and Tri90-30, were successfully reproduced. Moreover, for CaTr tRise0,peak and AP and CaTr Tri90-30, simulations showed good reproduction of the experimental variability intervals. CTDs were generally underestimated at the various drug doses. A possible reason for this is the fact that in the control population (Figure 3) CTDs are included in the variability ranges, but they cannot cover the higher values. Physiologically-correct in silico drug-induced CaTr prolongation (except for diltiazem) was seen, as proven by the overlapping of the in silico and in vitro CaTr Tri90-30. However, the CTD90 and CTD30 absolute values after drug administration were overall smaller in silico than in vitro.
We were able to obtain the same abnormality classes (Figure 7) observed in our in vitro data and in (13), i.e. single and multiple EADs (panels A, B, C, D, E, F), with the addition of repolarization failure (panels G and H), and irregular rhythms (panels I, J, K, L, M, N). Conversely, the in silico models did not show tachyarrhythmias observed e.g. in (13) or in 6 cases in our in vitro experiments in response to the highest dose of diltiazem. As discussed previously, these tachyarrhythmias may be syncytium-level events in vitro that could not have been captured in the simulations. Furthermore, a common response of the in silico hiPS-CMs, especially to administration of diltiazem and bepridil, is the suppression of spontaneous activity. Indeed, diltiazem administration at D3 and D4 also stopped the spontaneous AP in a big portion of our in silico population, 269 and 444 models out of 477, respectively. This is in agreement with the in vitro diltiazem experiments of 7 out of 15 laboratories involved in the multisite study reported in (40), where 100% of the hiPS-CMs tested did not produce spontaneous AP after administration of 10μM diltiazem (equals to our D4). Furthermore, in 5 laboratories a variable amount (20% - 70%) hiPS-CMs stopped beating. The same effect was observed for bepridil. In fact, as consequence of D3 and D4 bepridil administration, 107 and 444 models out of 477 stopped. Again, this is in agreement with our in vitro experiments (no spontaneous AP at D4), and with the experiments of (40) (50% hiPS-CMs stopped spontaneous AP in 4 laboratories (out of 15) with D3 bepridil, and over 80-90% hiPS-CMs in 15 laboratories with D4 bepridil).
It is interesting to note that in our in vitro experiments, despite the reliable AP and CaTr duration and triangulation increase, astemizole did not induce abnormalities, while they were observable in 9 in silico hiPS-CMs at D3 and 43 at D4. Astemizole is considered an intermediate risk drug in (40) and a high-risk drug both in vitro (46) and in the in silico drug trials performed in (3). Especially in Blinova et al. (40), 11/15 laboratories observed single and multiple EADs in 100% of their cells at 37°C, in response to 0.1μM astemizole (equivalent to our D4). The absence of EADs in our in vitro data (while showing pro-arrhythmic markers such as APD prolongation and increased APD triangulation), may be due to a number of reasons. One possibility is the lower temperature, though temperature-corrected in silico hiPS-CMs revealed repolarization abnormalities. Another reason could be potentially higher IK1 (and or IKs) in our high-density syncytial preparations compared to other studies.
Overall, hiPS-CMs proved to be an effective in vitro and in silico model to test drug-induced adverse cardiac effects. Unexpected results in vitro and in silico for bepridil, considered a highly cardiotoxic drug (3, 41), prompted further investigation. As reported in Table 3, bepridil triggered a very small amount of abnormalities in our in silico population. This is in agreement with our in vitro experiments and with the tests performed by Blinova et al. (40): in this multisite study, used here for comparison only, bepridil stopped the spontaneous AP in 80-90% hiPS-CMs in all the 15 laboratories at the highest bepridil dose 10μM (in agreement with our simulations); abnormalities were seen only in 2 out of 15 laboratories. Potential reason for this discrepancy can be the higher expression of L-type Ca2+ channels observed in vitro in hiPS-CMs than in adult cells (13). Blinova et al. (40) state: “Bepridil is a potent hERG blocker that also blocks L-type calcium and peak and late sodium currents at higher concentrations. High expression levels of calcium ion channels in hiPSC-CMs as compared to primary ventricular tissue may have contributed to more attenuated cellular proarrhythmic effects of the drug as compared to other drugs in the high TdP risk category.”. We were able to test this idea in silico: Figure S15A in the Supporting Material shows ICaL in the original O’Hara-Rudy model of human adult ventricular cell (34) (black trace) and in our hiPS-CM in silico population translated to 37°C (cyan traces). We tested if high levels of ICaL could have had a pseudo-protective effect against bepridil in hiPS-CMs, partially compensating the IKr block, resulting in a milder effect than in cells expressing less ICaL (e.g. adult cardiomyocytes). At room temperature, we tested bepridil D4 on 4 in silico hiPS-CMs that showed abnormalities with astemizole, by reducing bepridil ICaL blocking power first to half of its original value and then completely. This resulted in abnormalities in all 4 models (Figure 8), as expected. In addition, these 4 models in control conditions and 37°C showed ICaL higher than the adult one (Figure S15B). Table S2 shows the IC50 used for our in silico drug trials, taken from (3). Bepridil has the closest IKr and ICaL IC50 among APD-prolonging drugs. Therefore, an ICaL block comparable to IKr block in condition of highly expressed ICaL could indeed compensate APD prolongation and mask the occurrence of abnormalities, which may have occurred in adult cardiomyocytes (as reported in silico in (3, 47)). Our in vitro and in silico tests show the undeniable value of hiPS-CMs as models for drug testing and how in silico simulations could benefit the interpretation of the in vitro tests. The hiPS-CMs represent a potentially infinite pool of human cardiomyocytes and can capture key aspects of human cardiac electrophysiology in normal and diseased conditions (genetic mutations). Therefore, they are a great asset to predict the occurrence of adverse drug effects, in a unparallel manner that can be patient-specific.
As all experimental models, the hiPS-CMs are not without limitations. For example, they have different ion current expressions than adult cardiomyocytes, potentially affecting INa, ICaL, IKr and IKs (see Figure 2 in (13)), i.e. currents for which IC50 values are commonly computed. It must be noted that extensive experimental datasets from healthy adult human cardiomyocytes are non-existent due to unavailability of such cardiac tissue. Thus, inferences could only be made based on donor heart-derived human cells (34, 47, 48) or well-studied adult cardiomyocytes from other species. Nevertheless, different ion channel expressions can lead to underestimation (as for bepridil) or overestimation of the actual toxicity of a drug. A variety of optimization approaches are being developed to improve the maturity of the hiPS-CMs and bring them closer to an adult phenotype. These include extracellular matrix optimizations, stimulation protocols, mass transport improvements, alignment, substrate and metabolic function optimizations etc. (49). Such advances can impact positively cardiotoxicity testing.
Overall, commercial hiPS-CMs (e.g. CDI) have demonstrated their utility and superiority to animal models, even in their current state of maturity. Here we show the suitability of optically-recorded data from hiPS-CMs to produce information that empowers in silico modelling. With suitably-high acquisition rates, optical data can provide accurate temporal biomarkers for in silico models. All available Ca2+ data is indeed obtained by optical means; with the development of new small-molecule and genetically-encoded voltage dyes, AP records may completely replace electrical measurements due to their contact-less nature, easy parallelization and ability to measure cell properties in multicellular context. However, absolute values remain a challenge for optical measurements as voltage and Ca2+-sensitive dyes are rarely calibrated, i.e. they cannot provide reliable amplitude information for AP or CaTr, i.e. mV or mM. Such absolute values were essential in (19) to calibrate our first hiPS-CM population; in fact, AP peak <57.7mV (19) was included as a biomarker here to avoid unrealistic membrane potentials.
During our in silico tests, three limitations emerged. Firstly, CTD90, CTD50 and CTD30 are underestimated during drug administration (Figure 6, rows 1-4). The reason is that the 477 models in the population show relatively short control CaTr despite correct inclusion in the variability ranges by calibration. While the in silico CaTr correctly captured the drug-induced trends, they underestimated the changes observed experimentally. The in silico CaTr Tri90-30 matched well the experimental values, i.e. CaTr triangulation during drug administration was captured. In case of diltiazem (Figure 6, last row) we observed a peculiar behavior of the in vitro measurements following drug administration, since the CaTr showed larger CTDs at D2 and D3 than at D1, in spite CTD90 shortening for increasing diltiazem doses is clear from D2 to D4. The second limitation is that up to D4 in silico dofetilide generated few abnormalities, while D4 dofetilide triggered in vitro EADs in all the measurements. We observed already in (22) that to induce a remarkable amount of EADs or repolarization failures in an in silico hiPS-CM population, we needed IKr block>90%. Conversely, D4 dofetilide blocks only 80% IKr. With higher doses, tested in (3), we obtained a considerable increase in AP abnormalities. Finally, we did not observe in our simulations tachyarrhythmias as seen in vitro in a few samples, perhaps due to difference in single vs. multicellular behavior. We observed higher spontaneous AP rates, e.g. in irregular rhythms (e.g. in Figure S12, panel I, AP rate goes to 0.59Hz or a cycle of 1700ms) or residual activity in case of diltiazem (in Figure S13, Panel B, rate up to 0.83Hz, corresponding to AP CL of 1200ms). However, we did not observe AP rates greater than 2Hz.
Conclusions
In conclusion, this work supports the use of high-content, high-quality all-optical electrophysiology data to develop, calibrate and validate computer models of hiPS-CM for in silico drug trials. We report that simultaneously-acquired AP and CaTr enhance the model calibration process to obtain a final population that better reflects the experimental recordings. Our population was able to reproduce the effect of 5 different compounds, including the drug-induced abnormalities observed in vitro. In silico models constrained by in vitro data can be used to expand the parameter space of the investigations and to glean mechanistic insights into drug action. Finally, our simulations highlight the importance of being aware and taking into account potential differences in ionic currents between hiPS-CMs and adult cardiomyocytes, which could result in differences between in vitro/in silico hiPS-CMs and in vivo outcomes for specific compounds.
Author Contributions
AK and EE recorded and analyzed the optical in vitro data. MP and SS designed the Paci2019 hiPS-CM model. MP, EP, SS, JH, BR and EE designed the in silico tests on the populations of models. MP implemented the models and software tools used to produce and analyze the in silico data. MP, EP and SS analyzed the in silico data. All authors contributed to the writing and reviewed the manuscript.
Acknowledgements
The authors thank Dr. Jussi Koivumäki for sharing his code and for the fruitful discussion about the updates to the hiPS-CM model. The authors wish also to acknowledge CSC – IT Center for Science, Finland, for generous computational resources. MP was supported by the Academy of Finland (decision number 307967). EP and BR were supported by an NC3Rs Infrastructure for Impart Award (NC/P001076/1), a Wellcome Trust Senior Research Fellowship in Basic Biomedical Sciences (100246/Z/12/Z, 214290/Z/18/Z), EPSRC Impact Acceleration Awards (EP/K503769/1), the CompBioMed project (European Commission grant agreement No 675451), the Oxford BHF Centre of Research Excellence (RE/08/004/23915, RE/13/1/30181) and the TransQST project (Innovative Medicines Initiative 2 Joint Undertaking under grant agreement No 116030, receiving support from the European Union’s Horizon 2020 research and innovation programme and EFPIA). EE was supported by the National Institutes of Health (R01HL144157) and the National Science Foundation (1827535 and 1830941).
Footnotes
Techniques: Electrophysiology; Fluorescence.