Commercially available human iPSC-derived cardiomyocytes (Cardiosight®-S) were obtained from NEXEL Co., Ltd. (Seoul, Republic of Korea). Four differentiation batches for hiPSC-CMs (Lot# 12VAEZL, 12VVEZL, C2-6I1, C2-6U1) were used in this study. According to the COA (Certificate of Analysis) provided by the cell provider with the cell, all hiPSC-CM batches satisfied the provider’s QC criteria which are purity (>90% cTnT), MEA functionality, and drug reactivity for the reference drugs including E-4031 and nifedipine. The cells were cultured according to the instructions of the manufacturer. In brief, the cryopreserved hiPSC-CMs are thawed and seeded on the cell culture plate and then placed in an incubator at 37°C supplemented with 5% CO₂. The cells were maintained for 1 week to induce time-dependent maturation and electrophysiological stabilization. After then, the monolayered hiPSC-CMs were dissociated and harvested with the STEMdiff™ Cardiomyocyte Dissociation Kit (Stemcell Technologies, Vancouver, Canada) and replated to a 3-D culture microwell (SpheroFilm, 500 μm inner diameter, INCYTO, Republic of Korea). The media in the microwell were changed daily, and the spheroids were cultured for 3 days. After 3 days, the hiPSC-CM spheroids were collected for MEA experiments. In the case of the discoidal models of hiPSC-CMs, the detached cells from culture plates on day 7 were directly plated onto the fibronectin pre-coated MEA plates to create a dome shape at a density of ∼70,000 plated cells per well.

Before the sample plating on MEA plates, the recording areas of the 12-well MEA plate (Axion Biosystems, Atlanta, GA) were coated with human fibronectin (Sigma-Aldrich, St. Louis, MO) at 5 μL of 50 μg/mL in phosphate-buffered saline (PBS). Then, the plate was incubated at 37°C for 30 min. Then, the prepared hiPSC-CM spheroids were plated on the fibronectin pre-coated area of MEA plates. As previously mentioned, the dissociated and harvested cells from culture plates were directly plated on the MEA plates. Next, the MEA plates containing the hiPSC-CM spherical and discoidal models were incubated in a 37°C, CO₂ incubator for 1 h to attach the sample to the multi-electrodes on each well. After 1 h, each well was filled with 1 mL of the culture medium. The medium was changed every 2 days thereafter for 3–7 days to obtain the spherical and discoidal models of hiPSC-CMs with spontaneous and synchronous electrical activity on the MEA plates.

To evaluate the drug responses of the hiPSC-CM spheroids for ion channel-specific blockers, the effects of E-4031 (for a blocker of the hERG channel currents) and nifedipine (for a blocker of the calcium channel currents) on cardiac FPs of the hiPSC-CM spheroids were studied. The full of media on the plates was changed at least 4 h before recordings. FPs were recorded from spontaneously beating hiPSC-CM spheroids. For the FP recording, the MEA plate was placed on the Maestro MEA system (Axion Biosystems, GA, USA) keeping 37°C and 5% CO₂ conditions. Data were filtered with a Butterworth 0.1–2 kHz band-pass filter. The drugs were prepared with 1000X in dimethyl sulfoxide (DMSO) of the final target concentration and treated in each well as a cumulative method from the lowest concentration to the highest concentration every 20 min. The FP waveforms of the last 1 min of each dose were recorded and analyzed for data analysis. The beat per minute (BPM), FP amplitude (FPA), and FPD_cF (FP duration (FPD) corrected with Fridericia’s formula: FPD_cF = FPD/Beat Period^0.33) were analyzed with the AxIS program (Axion Biosystems, ver. 2.3.2). Experimental procedures for forming the models and field potential recording are shown in Figure 1.

SEM imaging and analysis for hiPSC-CMs spherical and discoidal models was conducted at Korea Research Institute of Bioscience and Biotechnology (KRIBB) according to the internal protocol. In brief, the samples were fixed in a 2.5% paraformaldehyde-glutaraldehyde mixture buffered with 0.1 M phosphate (pH 7.2) for 2 h, postfixed in 1% osmium tetroxide in the same buffer for 1 h, dehydrated in graded ethanol, and substituted by isoamyl acetate. Then they were dried at the critical point in CO₂. Finally, the samples were sputtered with gold in a sputter coater (SC502, POLARON) and observed using the scanning electron microscope, FEI Quanta 250 FEG (FEI, USA) installed in KRIBB.

Figure 2 shows optical micrographs of the discoidal and spherical models of hiPSC-CMs. The discoidal model has a diameter of approximately 1876.29 ± 64.59 µm (n = 3 from two batches) and is attached flatly to the bottom surface with a slightly thicker dome-like shape in the middle. In the case of the spherical models, it has a three-dimensional round shape and a diameter of 318.34 ± 10.25 µm (n = 15 from two batches). In both models, adjacent cells are overlapped and tightly attached to conduct electrical signals (Figure 2).

In silico 3-D models of hiPSC-CM spheroids were constructed in spherical and discoidal shapes (Figure 2). For the spherical model, it was slightly deformed such that a flat surface was generated, which was in contact with the bottom surface of the well of the MEA device. The diameters of the models were set to 1 and 2 mm for spherical and discoidal models, respectively, which were consistent with experimental observation. Tetrahedral meshes were generated inside the models for the computation of AP propagation. The number of meshes was increased until the simulation results converged. The propagation of the AP in the model was obtained by numerically solving the following monodomain equation: ∂ V m ∂ t = ∇ ∙ D ∇ V m − I i o n + I s t i m 1 C m where V _m is the membrane potential, t is time, D is the diffusion coefficient, I _ion and I _stim are ionic and stimulation currents, respectively, and C _m is the membrane capacitance. The diffusion coefficient was set to 0.0072 mm²/s which resulted in the experimentally observed conduction velocity of 0.12 ± 0.0028 mm/ms. The equation was numerically solved by using finite element method (Im et al., 2008). For the calculation of I _ion, the O’Hara-Rudy dynamic (ORd) human ventricular cell model was used (O'Hara et al., 2011). At the locations of the electrodes, the extracellular potential was calculated as described in a study (Ugarte et al., 2014). The method uses the ratio of the extracellular to intracellular conductivities which we set to three following the value used by Abbate et al. (2018). The FP measurement system was modeled as an equivalent circuit with an electrode capacitance, an electrode resistance, and the internal resistance of the measurement device as described in previous studies (Moulin et al., 2008; Raphel et al., 2018). The FP was calculated as the electric current through the equivalent circuit due to the extracellular potential, multiplied by the internal resistance of the measurement device (Raphel et al., 2018).

The effects of drugs E-4031 and nifedipine were incorporated by partly blocking I _Kr and I _CaL channels, respectively, more with increasing drug concentration. The percentages of each channel blockade were determined such that the simulated FPD were matched with experimentally measured data. Using the percentages of each channel blockade at multiple drug concentrations, the IC₅₀ and Hill Coefficient were calculated by applying the Hill equation (Goutelle et al., 2008).

All experimental results were statistically processed using GraphPad Prism (GraphPad, San Diego, United States), AxIS program (Axion Biosystems, ver. 2.3.2), and Excel (Microsoft, Redmond, WA, United States). All values are represented as mean ± Standard error of mean (SEM) and n represents the number of experimental replicates with individual samples. Statistical significance was determined using the Student’s t-test; p < 0.05 was considered to indicate statistical significance.

Field potentials were recorded in the hiPSC-CMs-based discoidal and spherical models using the MEA system. To check the quality of the hiPSC-CMs used in model production, we confirmed that cardiac troponin T (cTnT) and α-cardiac actin (α-actinin) were expressed in >90% of hiPSC-CMs. Also, it was confirmed that the human ether-à-go-go-related gene (hERG) and cardiac L-type calcium channel (Cav1.2), which contribute to ventricular repolarization and are the leading causes of drug-induced cardiac arrhythmia in drug development, were also expressing in the cell membrane (Figure 3A). The field potential parameters were recorded and analyzed by attaching discoidal or spherically shaped multi-cellular hiPSC-CMs on the MEA plates. The average BPM was significantly higher in the discoidal model than the spherical model [52.8 (n = 15) vs. 39.3 bpm (n = 9), p = 0.0141]. The field potential amplitudes were 0.99 and 1.1 mV for the discoidal and spherical models, respectively, and there was no statistical difference between the two groups. In the case of the FPD_cF as well, there was no statistical difference between the two groups with the values of 436.4 and 459.2 ms for the discoidal and spherical models, respectively (Figure 3B).

Figure 4 shows simulated FPs obtained from the spherical model of hiPSC-CM spheroid. The electrical stimulations were applied at four different sites on the model surface from the bottom to the top of the model (Figure 4A). The simulated FPs obtained at four different electrode locations at the bottom surface (a, b, c, and d in Figure 4A) are shown in Figure 4B. The shape of the signal exhibits a large variation similar to the observation reported by Abbate et al. (Abbate et al., 2018). The signals showed inverted shape between the electrode locations a and c for the stimulation sites 1 (blue color). The amplitudes of the signals were relatively small at the electrode location d compared to the other locations. The positive deflection corresponding to the repolarization was relatively more noticeable for the stimulation sites near the top of the model. Figure 4C shows the AP propagation on the surface of the model for each stimulation site.

Figure 5 shows simulated FPs obtained from the discoidal model of hiPSC-CM spheroid. The shape of the signal showed a large variation as in the case of the spherical model. The positive deflection corresponding to the repolarization was relatively more noticeable at the electrode location c for the stimulation sites 1 and 2 (Figures 5A, B). Figure 5C shows the AP propagation on the surface of the model for each stimulation site.

Figure 6 shows experimentally measured and simulated FPs under the effect of E-4031 from the spherical model of hiPSC-CM spheroid. The use of the mid-myocardial cell model, for which the action potential duration (APD) is longer than endocardial or epicardial cell, resulted in similar FPD as the experimentally observed one before the drug was applied. The simulated APs are shown in Supplementary Figure S1. To match the experimentally observed FPD increase with increasing E-4031 concentration, the percentage of the I _Kr blockade was searched for each E-4031 concentration used in the in vitro experiment because E-4031 is an hERG channel blocker. The I _Kr had to be blocked 95% when E-4031 concentration was 0.1 μM, for which the FPD_cF increased by 80% (Figure 6). The IC₅₀ calculated using the percentages of the I _Kr blockade at each E-4031 concentration was 8 nM which is close to the experimentally determined value of 7.7 nM with a Hill coefficient of 1.0 reported in the literature (Zhou et al., 1998). The simulated positive deflection corresponding to the repolarization tended to be flattened with increasing drug concentration which was also observed in the study of Tertoolen et al. (Tertoolen et al., 2018) although the flattening was not observed in our in vitro experiment. The simulated FPA change with increasing drug concentration was negligible, which qualitatively agreed with experimental data (Figure 6).

Figure 7 shows experimentally measured and simulated FPs under the effect of nifedipine for the spherical model of hiPSC-CM spheroid. The simulated APs are shown in Supplementary Figure S2. Because nifedipine is a Ca²⁺ channel blocker, the percentage of the I _CaL blockade was searched for each nifedipine concentration used in the experiment to match the experimentally observed FPD change with increasing nifedipine concentration. The IC₅₀ calculated using the percentages of the I _CaL blockade at each drug concentration were 0.018 μM which is close to the experimentally measured value of 0.012 μM with a Hill coefficient of 1.02 reported in the literature (Kramer et al., 2013). The simulated FPA change with increasing drug concentration was negligible, although experimental data showed slightly increasing FPA with increasing drug concentration (Figure 7).