Human Induced Pluripotent Stem Cell–Derived Engineered Heart Tissue as a Sensitive Test System for QT Prolongation and Arrhythmic Triggers

Introduction

Impaired repolarization of cardiac action potentials (APs) is a critical factor in the development of early afterdepolarizations (EADs) and life-threatening arrhythmias. Long-QT syndrome (LQTS) is one such heritable arrhythmia disease, which is associated with prolonged QT because of a reduced repolarization reserve. The 2 most common LQTS types are caused by an impaired function of the slowly activating potassium current (I_Ks, LQT1) or the rapidly activating potassium current (I_Kr, LQT2) and cause ≈44% or ≈35% of all genetically identified LQTS, respectively.¹ Drug-induced LQTS is caused mainly by an unintentional I_Kr block. Therefore, new drugs have to be tested for their proarrhythmic potential. Nowadays, the compound effects on a single ion channel can be estimated in expression systems allowing automated measurements. However, the net effect on repolarization must be studied at AP level, especially if >1 channel is blocked by a compound.

Finding a valid model system to predict arrhythmia in humans is crucial for studying (drug induced) LQTS. Access to human ventricular tissue is limited and largely restricted to patients undergoing heart surgery with structural heart disease; accessing tissue with genetic proarrhythmic heart disease is nearly impossible. Therefore, animal models have been used to study the mechanisms of AP regulation in LQTS. Genetic heart disease can be studied in transgenic mice, but repolarization is remarkably different when compared with humans. The repolarization in mice is nearly independent of I_Kr. Heart tissue from larger mammals (rabbit, canine) is used for safety studies as their repolarization depends predominantly on I_Kr. The substantial interspecies differences in the contribution of individual repolarizing currents to AP duration (APD) regulation lead to different patterns of repolarization reserve.² For example, in canine left ventricle (LV), the I_Kr contribution is smaller than in nonfailing human LV. In guinea pigs, repolarization depends more on I_Ks, and the inhibition of a single component can be compensated by the remaining currents. Rabbit LV preparations display peculiarities in AP dynamics, such as AP shortening on low-frequency pacing in the physiological range of humans (<1.5 Hz),^3,4 which may disqualify rabbits as a universal model for human electrophysiology. Purkinje fibers (PFs) from rabbits are unique in their strong dependence on I_Kr, leading to marked AP prolongation⁴ and EADs⁵ in the presence of I_Kr blockers. Consequently, rabbit PFs are the most frequently used model in the pharmaceutical industry, where they are used to detect even small-QT-prolonging effects of new compounds suspected to block repolarizing currents.

It is expected that human induced pluripotent stem cell–derived cardiomyocytes (hiPSC-CMs) might overcome the problems of interspecies differences in animal models. The comprehensive in vitro proarrhythmia assay initiative was established to expand the understanding of torsadogenic mechanisms building on the emergence of new technologies with hiPSC-CMs. In addition, hiPSC-CMs open the possibility of studying genome-specific phenotypes of patients and could pave the way for individualized drug treatment. First, studies in hiPSC-CM generated from patients with monogenic arrhythmogenic diseases, for example, LQTS^6–8 and Brugada syndrome,⁹ support this notion.

Still, there are concerns that hiPSC-CMs show a low degree of maturity, which may reduce the predictive value of this model for adult patients with cardiac diseases.^9–11 In the context of electrical stability, the repolarization reserve is not well characterized in hiPSC-CM since longtime AP measurements in single hiPSC-CM without artificial current injection or dynamic clamping are lacking.

Here, we use engineered heart tissue (EHT) to study the repolarization reserve of hiPSC-CM. This multicellular format allows the application of sharp microelectrodes^12,13 measuring APs without enzymatic dissociation. We evaluated CM from 3 different hiPSC lines cultured in EHT format in comparison to human LV muscle preparations. We then investigated whether hiPSC-CM may serve as a valid model for (drug induced) LQTS and whether it may be suitable as a sensitive test system for arrhythmic triggers.

Methods

An expanded method section is available in the Data Supplement. Data, analytic methods, and study materials will be made available to other researchers for purposes of reproducing the results or replicating the procedures on request within the limits required for patient confidentiality.

Human Materials and hiPSC-CMs in EHT

This investigation conforms to all principles outlined by the Declaration of Helsinki and the Medical Association of Hamburg. All materials from patients were taken with informed consent of the donors. LV free wall samples were obtained from patients undergoing implantation of an LV assist device, heart transplantation, or aortic valve replacement. Patients having taken the medication amiodarone before the study were excluded. As previously described,¹⁴ single cell suspensions of hiPSC-CM were subjected to EHT generation in a 24-well format (1×10⁶ hiPSC-CM/EHT).

AP Measurements

APs were recorded as described previously^12,15 with standard sharp microelectrodes in intact EHTs (25–100 days old) or the endocardial site of LV tissue, which were field stimulated at 1 Hz at 50% above the stimulation threshold. AP parameters were stable for at least 60 minutes (Table I in the Data Supplement). We did not find a systematic influence of EHT culture duration on APD at 90% repolarization (APD₉₀), on effect size of E-4031 and HMR-1556 (Figure I in the Data Supplement), or on transcript levels of ion channels (Figure II in the Data Supplement). To measure rate adaptation, we allowed preparations to equilibrate for at least 500 beats before APD₉₀ was measured. Short-term variability (STV) was calculated as previously described (STV=Σ│APD_90(n+1)−APD_90(n)│/[30√2]),¹⁶ always in the absence of EADs.

Expression Analysis

Total RNA was extracted from snap-frozen LV and EHT using the RNAeasy mini Kit (Qiagen, CA). Fifty nanograms of total RNA was used for expression analysis (Table II in the Data Supplement) by NanoString nCounter SPRINT Profiler according to the manufacturer’s instructions.

Statistics

GraphPad Prism 5 (GraphPad Software, San Diego, CA) was used for data analyses. Curves were fitted to data points from individual experiments, and all data were compared using unpaired or paired t tests when appropriate and for >2 groups 1-way ANOVA followed by Tukey corrections. All analyses were 2-tailed, and a P<0.05 was considered to be statistically significant. Group data are presented as mean±SEM.

Mathematical Modeling and Computer Simulations of CM Function

To test the influence of different ion current densities on the AP and repolarization reserve in hiPSC-CM, we used an established computer model for endocardial LV CM¹⁷ and adjusted current densities based on previous publications (Table III in the Data Supplement).

Results

AP Characteristics Under Basal Condition

All EHTs beat spontaneously with an average frequency of 1.19±0.10 Hz (n=40), whereas LV tissue was quiescent. To compare AP parameters (usually measured at 1 Hz), we selected EHT beating <1 Hz (27 of 40 EHTs). The AP shape in EHT was ventricular like, which could be quantified by a similar repolarization fraction ([APD₉₀–APD₅₀]/APD₉₀) in LV and EHT (0.26±0.01, n=34 versus 0.29±0.02, n=27; P=0.07; unpaired t test). Maximum diastolic potential (MDP; LV, −76.5±0.8 versus EHT, −77.4±1.2 mV; P=0.53), take-off potential (−77.0±0.9 versus −75.5±1.0 mV; P=0.11), AP amplitude (108.4±1.5 versus 105.8±1.8 mV; P=0.11), and maximum upstroke velocity (V_max: 230±17 versus 283±16 mV; P=0.07) did not differ between LV and EHT, but AP duration was shorter in EHT (APD₅₀, 165±13 versus 258±10 ms in LV, P<0.001; APD₉₀, 233±13 ms versus 343±9 ms in LV, P<0.001; Figure 1).

• Download figure
• Download PowerPoint

We compared the expression of genes coding for different ion channels in EHT (n=12) with LV (n=12; Figure 1C). All major repolarizing and depolarizing ion channels of LV were expressed in EHT. However, transcript levels of the channel subunit carrying the funny current (I_f; HCN4, 345±56 for EHT versus 64±7 for LV; P<0.001) and T-type calcium current (I_CaT; CACNA1G, 401±97 versus 13±2; P<0.001) were much higher in EHT than in LV. The expression of L-type calcium current (I_CaL; CACNA1C, 648±95 versus 380±40; P=0.019) and I_Kr subunits (KCNH2, 522±44 versus 388±34; P=0.033) was also higher in EHT than in LV, whereas those for I_Na (SCN5A, 969±133 versus 732±85; P=0.15), I_Ks (KCNQ1, 382±34 versus 309±40; P=0.18), and I_K1 (KCNJ2, 355±82 versus 432±36; P=0.40) did not differ. Expression data for the 3 individual cell lines are given in Figure III in the Data Supplement.

Low Concentrated Ivabradine Allows Slow Pacing of EHT Without Affecting APD

Low beating frequency is necessary to unmask the large contribution of I_Kr to repolarization.¹⁸ To slow down the beating rate in EHT (Figure 1B), we used the I_f blocker ivabradine. Even low concentrations of ivabradine (300 nmol/L) reduced spontaneous frequency effectively, as previously reported.¹⁴ The beating frequency was reduced from 1.15±0.20 to 0.31±0.03 Hz (n=11; P<0.001; paired t test). At 1 Hz pacing, ivabradine reduced diastolic depolarization (DD) from 7.0±1.0 to 0.9±0.6 mV/s (P<0.001; paired t test; Figure 2). As a consequence, take-off potential was lowered (−75.3±2.2 to −79.5±2.5 mV; P<0.01) and AP amplitude increased from 109±2 to 116±2 mV (P=0.017), whereas upstroke velocity was not significantly raised from 302±27 to 360±47 V/s (P=0.22). MDP was not affected by ivabradine (from −82.6±1.3 to −82.3±1.3 mV). APD was neither affected by ivabradine in EHT (APD₅₀, 146±19 to 161±22 ms, P=0.08; APD₉₀, 224±23 ms to 242±28 ms, P=0.20) nor in LV (APD₅₀, 279±33 to 278±31 ms, P=0.89; APD₉₀ 368±34 to 370±32 ms, P=0.62; n=5; paired t test; Figure IV in the Data Supplement).

• Download figure
• Download PowerPoint

Exaggerated Response to I_Kr Block in EHT

To test the contribution of I_Kr to the repolarization of hiPSC-CM in EHT, we used the I_Kr blocker E-4031. During AP measurements, all tissues were exposed to increasing concentrations of E-4031 (1 nmol/L to 1 µmol/L) in half-logarithmic steps (15 minutes incubation time per concentration step at 1 Hz pacing; Figure 3). E-4031 prolonged APD in all LV and EHT and slowed repolarization at membrane potentials <0 mV in LV and EHT. The half-effective concentration was not different between LV and EHT (73 nmol/L, n=8 versus 58 nmol/L in LV, n=12; P=0.42; Figure 3C). However, maximum prolongation of APD₉₀ by E-4031 (measured at 1 µmol/L) was significantly larger in EHT (n=26) than in LV (n=14) in absolute and also relative terms (+281±18 versus +109±14 ms, P<0.001; +126±9% versus +37±5%, P<0.001). In a subanalysis of LV, APD₉₀ was not significantly longer in tissues of female versus male patients (364±14 versus 324±12 ms; n=16 and 18; P=0.13), and there was no difference in APD₉₀ prolongation on 1 µmol/L E-4031 (+121±22 ms, +35±8%, n=7 versus +107±19 ms, +38±8%, n=7; P=0.79; P=0.47). In the human heart, AP prolongation by I_Kr block is typically larger at slower beating rates (so-called reverse use-dependence).^3,18–20 Therefore, we studied the effects of E-4031 on APD₉₀ at different pacing rates in a subset of experiments. The prolongation was larger at 0.5 Hz compared with 1 Hz in EHT (ΔAPD₉₀, +566±46 versus +273±35 ms, n=6; P<0.001) and in LV (ΔAPD₉₀, +176±42 versus +95±33 ms, n=6; P<0.001), demonstrating that both preparations have reverse use-dependence. However, the difference of ΔAPD₉₀ at 1 and 0.5 Hz was bigger in EHT than in LV (+293±24 versus +81±22 ms; P<0.001) indicating a stronger reverse use-dependence of I_Kr block in EHT than LV.

• Download figure
• Download PowerPoint

We suspected that high I_CaL density in hiPSC-CM²¹ contributes to pronounced AP prolongation by E-4031 in EHT. To test this hypothesis, we used verapamil to block I_CaL. Verapamil shortened APD in EHTs from 240±6 ms (basal) to 212±10 ms at 1 µmol/L (n=5; P=0.038). In the presence of 1 µmol/L verapamil, 1 µmol/L E-4031 prolonged APD to only 282±8 ms (n=5; Figure 4), which is significantly less prolongation (ΔAPD₉₀, +42±13 ms; Figure 4) than in the absence of verapamil (+218±18 ms, P<0.001; Figure 3B).

• Download figure
• Download PowerPoint

Minor Contribution of I_Ks to Repolarization Reserve in EHT

We used HMR-1556 as an I_Ks blocker. In human nonfailing ventricle, contribution of I_Ks to AP is restricted to situations when repolarization reserve is reduced and β-adrenoceptors (β-ARs) are activated.¹⁹ The same holds true for failing LV in our hands. APD₉₀ was not prolonged by the I_Ks blocker HMR-1556 (1 µmol/L) alone (basal, 337±24 ms versus HMR-1556, 335±25 ms; n=9; P=0.87) and also not in EHT (248±43 versus 249±44 ms; n=7; P=0.77; paired t test; Figure 5). Even when repolarization reserve was reduced by pretreatment with E-4031 (1 µmol/L), HMR-1556 did not prolong APD₉₀ significantly in LV (465±34 ms in E-4031 versus 476±41 ms in E-4031+HMR-1556; n=6; P=0.50; paired t test) or in EHT (526±49.0 versus 537±51 ms; n=4, P=0.20; Figure V in the Data Supplement). β-AR activation by isoprenaline (100 nmol/L) was necessary to demonstrate a prolongation of APD₉₀ on HMR-1556 tested in the presence of E-4031 pretreatment in LV (428±29 ms in E-4031+isoprenaline versus 449±30 ms in E-4031+isoprenaline+HMR-1556, n=5; P<0.01) and EHT (496±38 ms in E-4031+isoprenaline versus 528±35 ms in E-4031+isoprenaline+HMR-1556; n=4; P=0.013; paired t test; Figure 5). The effect size of I_Ks block was small and did not differ between LV (+5.0±1.1%) and EHT (+6.6±1.7%; P=0.42; unpaired t test). In EHT, β-AR activation by isoprenaline (100 nmol/L) alone shortened APD₉₀ (basal, 264±44 ms versus isoprenaline, 229±37 ms; n=5; P=0.033; paired t test) and consecutively administered HMR-1556 prolonged APD₉₀ (isoprenaline, 212±52 ms versus isoprenaline+HMR-1556, 230±48 ms; n=3; P=0.048; paired t test, data not shown).

• Download figure
• Download PowerPoint

Early Afterdepolarizations

The occurrence of EADs during in vitro testing is an alarming finding, suggesting a high probability of a drug to induce torsade de pointes (TdP). Therefore, we measured the inducibility of EADs and triggered activity in LV and EHT. I_Kr block alone by 1 µmol/L E-4031 was not sufficient to evoke EADs in LV (0/18), whereas EHTs developed EADs and triggered activity in 50% of cases (16/32; P<0.001, Fisher exact test). To induce EADs in LV, several additional interventions were necessary to reduce repolarization reserve (pacing rate at 0.33 Hz, lowering extracellular potassium from 5.4 to 2.7 mmol/L; β-AR stimulation with 100 nmol/L isoprenaline, I_Ks block with 1 µmol/L HMR-1556 and I_K1 block by max; 300 µmol/L barium; Figure 6; Figure VI in the Data Supplement). Under these conditions, we could induce EADs in all 7 preparations of LV. In both, EHT and LV, EADs were preceded by incremental AP prolongation (Figure VII in the Data Supplement) or APD alternans (Figure 6). We analyzed APD₉₀ of the last (or the second last in case of alternans) AP before the first EAD to define a critical APD value for EAD development. EHTs developed EADs at much shorter APD₉₀ values than LVs (EHT, 531±49 ms, n=16 versus LV, 1072±63 ms, n=7; P<0.001; Figure 6B).

• Download figure
• Download PowerPoint

Specificity of EAD Development

Because test systems should provide not only a high sensitivity but also a high specificity, we tested 2 I_Kr blockers without proarrhythmic potential. Prior studies have shown that the antibiotic moxifloxacin induced EADs at 100 µmol/L in single CM but not in cardiac tissue or animal models.²² In EHT, moxifloxacin prolonged APD₉₀ from 202±36 ms (basal) to 252±52 ms at 10 µmol/L (P<0.01) and to 322±58 ms at 100 µmol/L (P=0.017; n=5) without the development of EADs (Figure 4). Verapamil is another I_Kr blocker without torsadogenic risk. In EHT, verapamil did not prolong APD or induce EADs (Figure 4), but reduced APD, in line with the well-known I_CaL blocking effect which seems to dominate at (clinically used) lower concentrations.

In-Silico Investigation of I_Kr Block and EADs

Simulation results obtained with adult CM and hiPSC-like CM in-silico models corroborated our experimental findings on I_Kr block (Figure 7). That is, AP prolongation was more pronounced in the hiPSC-like CM model than in the adult LV CM model. In line with our experimental findings, even a 100% block of I_Kr did not induce EADs in the adult CM model. In contrast, even incomplete block of I_Kr was sufficient to evoke EADs (Figure 7B) in the hiPSC-CM model. The new hiPSC-like CM model differs to the LV model in 4 current densities (4× I_Kr, 0.5× I_K1, 1.5× I_CaL, and additional I_CaT; Table III in the Data Supplement). Therefore, we tried to estimate individual contribution of each current to the low EAD threshold. We created 4 additional hiPSC-like CM models in which each current was separately reverted back to LV values (Figure 7C). In a next step, we applied increasing I_Kr block until the first EAD occurred (Figure 7D) to get a threshold value of I_Kr block to induce EAD. With LV I_CaL density,²¹ APD of hiPSC-like CM was shorter (Figure 7C), and EAD development on I_Kr block was completely prevented (Figure 7D). Reverting I_K1 or I_CaT density to adult values resulted in slight AP prolongation, but the EAD threshold of I_Kr block was only marginally lifted from 83% to 90% with adult I_K1 density or to 86% without contribution of I_CaT. In contrast, the reduction of I_Kr density to LV level induced strong AP prolongation at basal condition and lowered the EAD threshold dramatically to 25% of I_Kr block.

• Download figure
• Download PowerPoint

Short-Term Variability

STV of APD (STV=Σ|APD_90(n+1)−APD_90(n)|/[30√2]) has been proposed as a parameter that predicts EAD development.⁵ We compared the STV of APD₉₀ in EHT and LV under baseline conditions and with E-4031 and formed subgroups of EHTs with and without EADs. At baseline, STV of APD₉₀ did not differ between EHT and LV (1.46±0.38; n=14 versus 0.80±0.17; n=16; P=0.11; unpaired t test). STV was increased by E-4031 only in EHTs which subsequently developed EADs from 1.49±0.54 to 8.01±3.31 (n=7; P=0.049; paired t test), whereas LV and EHT without EADs showed no increase in STV (Figure 8). Moxifloxacin (100 µmol/L) did not increase STV in EHT (from 2.11±1.03 to 2.93±0.87; n=4; P=0.48; paired t test).

• Download figure
• Download PowerPoint

Differences Between hiPSC-CM Control Cell Lines

We designed this study with CMs from 3 different hiPS cell lines (2 academic: C25,hPSCreg name: TUMi001-A, and ERC018; and 1 commercial: iCell2), to identify general electrophysiological mechanisms of hiPSC-CMs. Although similarities prevailed, some aspects differed between cell lines (Figures III, VIII, and IX). ICell2 showed shorter APD₂₀, APD₅₀, and APD₉₀ values than the other cell lines, as well as having higher mRNA levels of SCN5A, CACNA1G, and KCNJ2 (Figure III in the Data Supplement). The AP shape differed between and within hiPSC-CMs of different lines. Figure VIII in the Data Supplement shows 3 representative APs for each cell line to illustrate individual differences in AP shape. Nevertheless, absolute AP prolongation on E-4031 was not different between hiPSC-CM cell lines (ΔAPD₉₀; C25, +303±37 ms; ERC018, +293±34 ms; iCell2, +248±20 ms), whereas each cell line showed a stronger prolongation than LV (ΔAPD₉₀, +114±14 ms; Figure IX in the Data Supplement).

Discussion

The study compared AP characteristics and responses to repolarization-slowing drugs in hiPSC-CM–derived EHTs and human LV muscle preparations. The main results were (1) that the I_f blocker ivabradine at low concentrations is an effective tool to reduce DD and spontaneous beating rate in hiPSC-CM–based EHTs without affecting repolarization, (2) that similar to human LV, the contribution of I_Ks to repolarization in EHT was detectable only in the presence of β-adrenergic stimulation and concomitant I_Kr inhibition, and (3) that EHT showed exaggerated AP prolongation, stronger reverse use-dependence, and a high susceptibility to EADs on I_Kr block when compared with human LV revealing a high sensitivity and specificity for arrhythmic triggers.

Ion Channel Expression and Basic Characteristics of APs

The repolarization fraction of APs, that is, the relation between early and late repolarization, was similar in EHT and LV, suggesting that hiPSC-CM display a rather uniform ventricular-like repolarization pattern in EHT. On the other hand, automaticity of EHTs driven by DD is the most obvious electrophysiological difference in comparison to the quiescent LV tissues. A strong expression of HCN4 (Figure 1), high current densities,¹¹ and the effectiveness of ivabradine on DD and beating frequency (Figure 2) indicate a relevant contribution of I_f to spontaneous pacemaking in hiPSC-CM. The robust expression level of SCN5A in EHT reflects the high I_Na density in hiPSC-CM¹² and the fast upstroke velocity, although DD slightly depolarizes the take-off potential and, thereby, reduces the functional availability of Na⁺ channels. The high expression levels for I_CaT conforms with recent current recordings.²¹ This I_CaT current may be relevant because the coexistence of I_Ca,L and I_Ca,T should generate more window Ca²⁺ current²¹ during the plateau phase, increasing depolarizing forces and thereby reducing the repolarization reserve.

We found higher expression levels of KCNH2 in hiPSC-CM than in LV in line with the higher I_Kr densities reported for hiPSC-CM (0.55–1.9 pA/pF)^6,7,11 versus LV (0.25–0.31 pA/pF).^2,23,24 Why differences in I_Kr density between hiPSC-CM and LV are more pronounced than those in transcript levels remains unclear. The expression of KCNQ1 was similar, which is in line with I_Ks densities reported in a similar range for hiPSC-CM (0.3–1 pA/pF)^6,11,25,26 and LV (0.2–0.6 pA/pF).^2,27,28 As reported previously,¹² we measured physiological MDP in EHT corresponding to similar expression levels for KCNJ2 in EHT and LV. Although I_K1 densities are consistently reported to be 2- to 8-fold lower in hiPSC-CM (−0.85 to 2.9 pA/pF)^10,11,29 than in human LV (6–7 pA/pF)^2,23,30 at similar conditions (at −100 mV with 5.4 mmol/L K⁺), several groups have measured physiological MDP <−74 mV^11,29,31 in hiPSC-CMs. This is in line with in-silico investigations, which suggests only a minimal depolarization of MDP if I_K1 density is reduced by half in hiPSC-CM (Figure 7) and in LV.² These data underline that the contribution of I_K1 to MDP is smaller than often anticipated.¹³

Low Concentrations of Ivabradine Allow AP Recording in EHT at Slow Rate

We have tried to reduce the spontaneous beating rate of EHT via pharmacological intervention because the effect of ivabradine on spontaneous beating frequency of hiPSC-CMs is controversial.^14,32–34 Here, we demonstrated for the first time that a reduction of spontaneous frequency by a low concentration of ivabradine is associated with a reduction of DD, suggesting a relevant contribution of I_f to pacemaking in hiPSC-CM in EHT. In the normal development of the heart, Ca²⁺-clock pacemaking predominates the first phase of development, whereas the membrane clock’s influence rises during development to the adult state.³⁵ If one interprets the strong dependence of spontaneous beating on ivabradine as a sign of dominance of the membrane clock over the Ca²⁺-clock in EHT, this observation may indicate a later stage of maturation of hiPSC-CM. However, it should be noted that hiPSC-CM culture in EHT does not reduce the spontaneous rate per se,¹² and the definitive source of automaticity in EHT is difficult to identify.¹⁴ Nevertheless, the omnipresent, ivabradine-sensitive DD and the ventricular-like repolarization fraction point to ventricular-like CMs with an abnormally high I_f.

Ivabradine does not affect AP shape in healthy³⁶ and failing LV (Figure IV in the Data Supplement). In contrast, ivabradine can prolong APs in canine LV,³⁶ when repolarization reserve is impaired. Despite the clear evidence for a smaller repolarization reserve in EHT than in LV, ivabradine did not affect AP shape, neither under basal conditions nor in the presence of E-4031. To avoid a potential bias by pretreating only EHT with ivabradine, we also performed experiments in LV with 300 nmol/L ivabradine and E-4031. As seen in EHT and LV, maximum E-4031–induced AP prolongation with or without ivabradine did not differ (Table IV in the Data Supplement). Therefore, we feel confident in the use of ivabradine to facilitate AP recordings in EHT at slow rates without reducing the repolarization reserve.

Contribution of I_Kr to APD Regulation in EHT

Block of I_Kr by E-4031 induced AP prolongation in both EHT and LV, starting just <0 mV of the membrane potential. Although half-effective concentration values did not differ between EHT and LV (both ≈3-fold lower than reported for human atrial tissue; 181 nmol/L¹⁵), the total effect was ≈2.5-fold higher in hiPSC-EHT. The pronounced prolongation on I_Kr inhibition in EHT can be mainly explained by the higher I_Kr^{2,6,7,11,23,24} and I_CaL densities²¹ in hiPSC-CM than in LV. Our in-silico modeling suggests that enhanced depolarizing and repolarizing currents in hiPSC-CM-EHTs are balancing each other under basal conditions. When I_Kr is blocked (either in the in-silico model or experimentally), this balance is abolished, and APs are markedly prolonged. Both the modeling data (Figure 7) and the experiments with verapamil (Figure 4) show that this prolongation is carried by a strong Ca²⁺ influx.

The inhibition of I_Kr shows a reverse use-dependence, which might be dangerous for class III antiarrhythmic compounds, such as sotalol. Many models like papillary muscles from guinea-pig²⁰ and rabbits⁴ fail to demonstrate reverse use-dependence of I_Kr blockers. In contrast, hiPSC-CM displayed pronounced reverse use-dependence similar to LV, qualifying EHT as model to study bradycardia-associated EAD generation, which is the dominant trigger for TdP in the human heart.

Contribution of I_Ks to APD Regulation in EHT

There is no doubt about the relevance of I_Ks contribution to repolarization in human ventricle in vivo as LQT1 syndrome leads to life-threatening TdP arrhythmia. In stark contrast, under ex vivo conditions, I_Ks contribution to repolarization is difficult to demonstrate. Ex vivo measurements of human nonfailing LV showed that I_Ks does not contribute to repolarization under basal conditions.² Reducing the repolarization reserve by I_Kr block and the activation of β-AR was necessary to unmask a small increase in APD₉₀ by I_Ks block,¹⁹ which may be seen in line with the clinical effectiveness of sympathetic denervation to suppress TdP in LQT1 patients.³⁷ We could confirm this finding for failing LV and EHT (Figure 5). On the contrary, recent publications reported that APD of hiPSC-CM derived from LQT1 patients were markedly longer than those from control cell lines.^6,26 The reasons for this difference to normal human heart physiology (and our present data) are unknown, but the observation indicates a large nonphysiological contribution of I_Ks to repolarization in these hiPSC-CM cell lines or may result from the high variability in APD between control cell lines (Figure III in the Data Supplement).

EADs on I_Kr Block in EHT

EADs are one of the triggers initiating TdP in both genetic and drug-induced LQTS. Under ex vivo conditions, inhibition of I_Kr alone is not sufficient to induce EADs in human LV preparations^2,3 (and this study). Even the stimulation of β-AR in the presence of low extracellular potassium concentration was not sufficient to induce EADs in human LV³ but in 30% of rabbit heart.³ In our study, multiple interventions to reduce potassium conductance and activation of β-AR were necessary to induce EADs in LV (Figure 6 and Figure VI in the Data Supplement). In contrast, EADs developed in ≈50% of EHTs when I_Kr was blocked by 1 µmol/L E-4031 only. This finding again demonstrates an extraordinarily large contribution of I_Kr to electrical stability in hiPSC-CM⁷ embedded in EHT. Unlike in LV, the occurrence of EADs in EHT was not restricted to extraordinarily long APD (Figure 6). As previously reported for canine heart,¹⁶ we could confirm that STV works as a predictor of EAD development in EHT when I_Kr is blocked.

Isolated CMs are prone to develop EADs, limiting the specificity to predict clinically relevant TdP arrhythmia.²² For instance, moxifloxacin is known to induce EADs in single cardiomyocytes but not in cardiac tissue or intact animals²² and verapamil has no torsadogenic risk. The lack of EAD development in EHT on moxifloxacin and verapamil fits with clinical data and indicates a high specificity of hiPSC-EHTs in predicting proarrhythmic risk.

Potential Cellular Mechanisms for EADs

The high susceptibility of EHT to develop EADs indicates differences in the cellular electrophysiology when compared with LV. In a previous study, we found similar upstroke velocities that indicate physiological peak I_Na.¹² In the same report, we did not find late I_Na in hiPSC-CM, excluding this as a potential mechanism for EAD development. Ca²⁺ influx across the cell membrane is larger in hiPSC-CM, as I_CaL density is higher.²¹ By blocking I_CaL, the pronounced AP prolongation by I_Kr block was diminished (Figure 4C and 4D) and EAD development prevented. Furthermore, I_CaT, which is absent in human adult LV CMs, is active in hiPSC-CMs.²¹ However, its contribution to the total Ca²⁺ influx is much smaller than I_CaL (ratio, ≈1:50). In-silico investigation suggests that strong I_CaL density²¹ in the hiPSC-CM contributes substantially to lowering the EAD threshold on I_Kr block, whereas reduced I_K1 density and the contribution of I_CaT had only a small proarrhythmic effect (Figure 7D). The strong I_Kr current is suggested to be antiarrhythmic by counterbalancing strong I_CaL and by prohibiting early AP prolongation and EAD development by I_Kr block. However, when the repolarizing force of strong I_Kr is abrogated by a higher degree of I_Kr block, inward currents can be large enough to generate the EADs.

Potential Structural Mechanisms for EADs

Although the EHT format supports the structural maturation of hiPSC-CM to a more rod-shaped form and better end-to-end cell contacts,¹² the cell density in EHT is still lower than in adult cardiac tissue and may cause reduced coupling of CM. After the source-sink theory,³⁸ lower coupling means less (hyperpolarizing) sink. Assuming a heterogeneous EAD development (as the source) on I_Kr block in individual cells, this may explain the high susceptibility to EADs in EHT compared with LV because coupling (as the sink)³⁸ can buffer less in EHT than in LV.

EHT as a Model for PFs?

Strong AP prolongation and high susceptibility to the development of EADs on I_Kr block are classic hallmarks of PF. Therefore, PFs but not papillary muscles are used in safety pharmacology to unmask even small risks of arrhythmogenicity.^4,5 EHTs developed EADs on I_Kr block with a similar incidence (50%) as reported for rabbit and canine PF (40%–80%),^4,5,39 whereas LV preparations from rabbit⁴ and human^3,19 did not. Enhanced I_f, I_CaL, and I_CaT^40,41 might be responsible for this difference, all of which would be expected to contribute to ivabradine-sensitive DD³⁶, strong AP prolongation and EAD development on I_Kr inhibition. In fact, the same features were observed in the present study in hiPSC-CM EHTs, suggesting that this new model might be as sensitive as PFs to predict EAD development in (drug induced) LQTS. Detailed comparative studies are necessary to investigate whether hiPSC-CMs in EHT share the basic electrophysiological peculiarities of PF CMs.

Limitations

Current densities for in-silico analysis were taken from published data for hiPSC-CM cultured isolated or in monolayers (Table IV in the Data Supplement). Differences in current densities between CM cultured in EHT or monolayer, as known for I_Na,¹² cannot be excluded. All 3 iPS cell lines were generated from female donors. It remains unclear whether hiPSC-CMs reflect the well-known sex differences of AP. LV tissue was obtained from patients with aortic valve disease or heart failure as we do not have access to LV tissue of healthy humans (Table V in the Data Supplement). Thus, potential disease-specific differences in AP parameters cannot be excluded.

Conclusions

We have characterized the impact of the 2 repolarizing currents of LQT1 and LQT2 in hiPSC-CM–based EHT compared with human ventricular tissue. The exaggerated AP prolongation and the high susceptibility to EADs of hiPSC-CM on I_Kr block indicate hiPSC-CM as a sensitive test system for QT prolongation and arrhythmic triggers in I_Kr assays. hiPSC-CM in EHT might serve as a sensitive and specific model to study the individual response to I_Kr block also in respect to monogenic arrhythmogenic heart disease.

Acknowledgments

We gratefully acknowledge expert technical advice and help in providing engineered heart tissues (EHTs) and left ventricular (LV) tissue from Benjamin Kloth, Sharia Iqbal, Umber Saleem, Ingra Mannhardt, Anika Knaust, Tessa Werner, Marta Lemme, Klaus Söhren, Thomas Schulze, Birgit Klampe, Elisabeth Krämer, Aya Domke-Shibamiya, and Sandra Laufer. We also thank Alessandra Moretti and Dennis Schade for their kind contribution of materials, Herbert Himmel for expert advice in electrophysiological safety studies, and Jennifer N. Lohr for useful comments on the article, as well as language editing and proofreading.

Footnotes