Proarrhythmia Risk Assessment in Human Induced Pluripotent Stem Cell-Derived Cardiomyocytes Using the Maestro MEA Platform

Transcription

1 TOXICOLOGICAL SCIENCES, 147(1), 2015, doi: /toxsci/kfv128 Advance Access Publication Date: June 27, 2015 Research Article Proarrhythmia Risk Assessment in Human Induced Pluripotent Stem Cell-Derived Cardiomyocytes Using the Maestro MEA Platform Yusheng Qu 1 and Hugo M. Vargas Integrated Discovery and Safety Pharmacology, Amgen Inc., Thousand Oaks, California To whom correspondence should be addressed at Integrated Discovery and Safety Pharmacology, Amgen Inc., Thousand Oaks, CA yqu@amgen.com ABSTRACT Evaluation of stem cell-derived cardiomyocytes (SC-CM) using multi-electrode array (MEA) has attracted attention as a novel model to detect drug-induced arrhythmia. An experiment was conducted to determine if MEA recording from human induced pluripotent SC-CM (hipsc-cm) could assess proarrhythmic risk. Ten herg blockers, 4 Na þ blockers, and 1 IKs blocker were evaluated blindly. Eight drugs are associated with Torsades de Pointes (TdP) and 4 are not. Multiple parameters, including field potential duration (FPD), Na þ slope, Na þ amplitude, beat rate (BR), and early after-depolarization (EAD) were recorded. Minimum effective concentrations (MEC) that elicited a significant change were calculated. Our results determined that FPD and EAD were unable to distinguish torsadogenic from benign compounds, Na þ slope and amplitude could not differentiate Na þ channel blockade from herg blockade, BR had an inconsistent response to pharmacological treatment, and that hipsc-cm were, in general, insensitive to IKs inhibition. A ratio was calculated that relates MEC for evoking FPD prolongation, or triggering EAD, to the human therapeutic unbound C max (MEC/C max ). The key finding was that the ratio was sensitive, but specificity was low. Consistently, the ratio had high positive predictive value and low negative predictive value. In conclusion, MEA recordings of hipsc-cm were sensitive for FPD and EAD detection, but unable to distinguish agents with low- and high-risk for TdPs. Although some published reports suggested great potential for MEA recordings in hsc-cm to assess preclinical cardiac toxicity, the current evaluation implies that this model would have a high false-positive rate in regard to proarrhythmic risk. Key words: human induced pluripotent stem cells; cardiac myocytes; field potential; early after-depolarization; multielectrode array; minimal effective concentrations; Torsades de Pointes; proarrhythmia To assess drug-induced proarrhythmic risk, especially Torsades de Pointe (TdP), the pharmaceutical industry has been conducting nonclinical and clinical studies according to ICH S7B (ICH S7B), and E14 (ICH E14) guidelines, respectively, for the last decade. The primary focus of the ICH S7B guideline has been the evaluation of delayed ventricular repolarization, ie, in vivo QT prolongation due to herg current blockade. A novel paradigm called the Comprehensive In Vitro Proarrhythmia Assay (CiPA) was proposed as a new nonclinical strategy for assessing proarrhythmia risk (Sager et al., 2014). The CiPA approach has 3 components: (1) potency determination at multiple cardiac ion channels; (2) in silico action potential modeling for multi-ion channel data integration, and (3) examination of proarrhythmic endpoints in human stem cell-derived cardiomyocytes (hsc- CM). Although current regulatory guidelines focus on herg-mediated repolarization defects, CiPA will presumably provide an improved approach for proarrhythmia risk assessment by moving away from a herg-centric approach, given that all herg blockers do not cause QTc prolongation and even those that do cause QTc prolongation do not consistently cause TdP. The availability of hsc-cm (Kehat et al., 2001), especially those derived from induced pluripotent stem (ips) cells (Takahashi and Yamanaka, 2006) has sparked interest in their usefulness to improve proarrhythmia risk assessment. A variety VC The Author Published by Oxford University Press on behalf of the Society of Toxicology. All rights reserved. For Permissions, please journals.permissions@oup.com 286

2 QU AND VARGAS 287 of technologies have been utilized to characterize hsc-cm, including action potential measurements (Caspi et al., 2009; Ma et al., 2011; Peng et al., 2010; Qu et al., 2013), calcium- and voltage-sensitive dyes (Cerignoli et al., 2012), and impedance measurements (Abassi et al., 2012; Guo et al., 2011). For proarrhythmic assessment, isolated ventricular myocytes are ideal cells for recording action potentials with conventional patch clamp (Caspi et al., 2009; Ma et al., 2011; Peng et al., 2010; Qu et al., 2013), but this technique is low throughput. On the other hand, cardiac myocyte monolayer cultures of large aggregates of cells are useful for recording extracellular field potentials using MEA, which provides a multiwell platform with higher throughput (Asai et al., 2010; Braam et al., 2010; Caspi et al., 2009; Clements and Thomas, 2014; Harris et al., 2013). Another advantage of monolayer (2D) culture is that myocytes are electrically coupled to adjacent myocytes; thus the multicellular matrix may be better for identifying proarrhythmic endpoints than isolated ventricular myocytes. Three-dimensional (3D) culture of ipsc-cm was shown to increase gene expression of excitation contraction coupling and Ca 2þ handling (Zhang et al., 2013). Therefore 3D culture may improve assessment of myocyte contractility, however, expression of ion channel genes that determine AP initiation, maintenance, and termination is not significantly different between 2D and 3D culture (Zhang et al., 2013). An additional report also demonstrated no significant difference in the spontaneous firing rate and APD between 2D and 3D culture (Rao et al., 2013). Therefore, for evaluation of proarrythmic risk, 2D culture should theoretically perform at the same level as 3D culture. Previously we evaluated the electrophysiological profile of 15 various ion channel blockers in hesc-cm (embryonic origin) and found that action potential duration was sensitive to herg blockade, but was less sensitive for identifying NaV1.5 and IKs inhibition (Qu et al., 2013). In this study, the same 15 compounds were evaluated (in a blinded manner) in hipsc-cm (icell cardiomyocytes; 2D culture) using MEA to assess the overall sensitivity of this high throughput platform for detection of proarrhythmic endpoints like field potential duration (FPD) and early afterdepolarization (EAD). MATERIALS AND METHODS The MEA studies were conducted at Cyprotex US, LLC (Watertown, Massachusetts) and lab personnel were blinded to the test articles used in the study. Cells. Human ips cell-derived cardiomyocytes (icell cardiomyocytes; Cellular Dynamics International, Madison, Wisconsin) were used and cultured according to manufacturer s instructions. As one of the first commercialized hipsc-cm cell lines, icell have been investigated substantially and there are more than 30 peer reviewed journal articles published using icell since They are a mixture of atrial-, nodal-, and ventricular-like myocytes with majority of the cells being ventricularlike myocytes (Ivashchenko et al., 2013; Ma et al., 2011). Electrophysiological characterization of icell indicated that these myocytes are not mature and exhibit slow action potential upstrokes and more depolarized resting membrane potentials (Ivashchenko et al., 2013; Ma et al., 2011). In addition, response to b-adrenergic stimulation and Ca 2þ handling properties are less developed (Ivashchenko et al., 2013). The gene transcription profile of icell is stabilized approximately 14 days post-thaw following cryopreservation and remained constant for the duration of 42 days post-thaw (Puppala et al., 2013). MEA plate preparation. Six-well tissue culture plates were precoated with 0.1 % gelatin. The gelatin was aspirated from each well and freshly thawed cardiomyocytes were dispensed into each well of the 6- well plate with icell plating medium and incubated in a standard cell culture incubator at 37 C, 5 % CO 2 for 48 h. After 48 h, nonadhered cardiomyocytes and debris were removed by rinsing with icell maintenance medium and the plate was incubated for an additional 5 days in maintenance medium, with replacement of the medium every other day. After 7 days, the cells were replated onto a fibronectin coated 48-well microelectrode array plates at a density of cells per well in icell maintenance medium. The plate was incubated for a further 5 8 days until a stable beating rate was attained and the cells were ready for drug treatment. The medium was decanted and cells were treated with vehicle (0.2% DMSO), quinidine (positive control), or test compound at 5 concentrations in duplicate. MEA recording. The activities of cells in a 48-well microelectrode array were recorded prior to (baseline) and 45 min after drug treatment using the Maestro MEA system (Axion Biosystems). The recording conditions were at 37 C using the standard cardiac settings (Axion Biosystems Maestro Axis software version 1.7.8). These settings have 130 gain and record from 1 to Hz, with a low-pass digital filter of 2 khz for noise reduction. The beat detection threshold was 300 mv and the field potential duration (FPD) detection threshold was 3 noise. Field potentials were analyzed with the platform software, and outputs included: beat period, fast Na þ slope (V/s), fast Na þ amplitude (mv) and FPD (ms) per well. The software had a pruning add-on (through MatLab) which analyzes each electrical beat waveform for quality of the T-wave determination, to calculate FPD. Because all of the electrodes in a well were synchronized, baseline electrodes and their corresponding post-treatment electrodes were removed from the calculation if beat rates were greater than 2 SDs away from the average of the electrodes. Reported results were calculated by averaging all of the electrodes in each well, then averaging the duplicate wells. The drugtreatment data were normalized to baseline initially, then expressed as a percent of vehicle-treated wells. Data analysis. A minimum effective concentration (MEC; the lowest concentration that caused an effect) was determined for each MEA parameter, including FPD, Na þ slope, Na þ amplitude, and beat period. An effect size of 6 10% compared with vehicle, or that reached statistical significance, was considered drugrelated. Field potential waveforms were also inspected for EAD formation. An EAD was defined as a small voltage spike which occurred after the Na þ spike (eg, spike n) but before the next Na þ spike (eg, spike nþ1). Detection of EAD required visual inspection of FP traces. For statistical analysis, a Student s t-test was used to assess significance and correlation analysis was performed using GraphPad Prism5 (GraphPad Software Inc.). Assessment of assay performance. Calculation of sensitivity, specificity, positive predictive value, and negative predictive value was conducted according to the following formulas, Sensitivity ¼ (true positive)/(true positive þ false negative) Specificity ¼ (true negative)/(true negative þ false positive) Positive predictive value ¼ (true positive)/(true positive þ false positive) Negative predictive value ¼ (true negative)/(true negative þ false negative)

3 288 TOXICOLOGICAL SCIENCES, 2015, Vol. 147, No. 1 Reagents. Sertindole (CAS ) and Moxifloxacin ( ) was purchased from ChemPacific (Maryland), L (Selnick et al., 1997) was purchased from Albany Molecular Research Inc. (New York), Cisapride ( ) from Tocris Bioscience (Missouri), Ranolazine, Alfuzosin, Mexiletine, Flecainide, Terfenadine, Lamotrigine, DL-sotalol, Terodiline from Sigma-Aldrich (Missouri), Dofetilide was synthesized at American Custom Chemicals Corporation (San Diego, California), Tolterodine ( ) from Toronto Research Chemicals (Toronto, Canada), AMG 1 was synthesized at Amgen Medicinal Chemistry (Thousand Oaks, California). Moxifloxacin, Flecainide, Mexiletine, DL-Sotalol, Terodiline, Alfuzosin, and Ranolazine were dissolved in distilled H 2 O to make stock solutions, and the rest of the compounds were dissolved in DMSO to make stock solutions. RESULTS Extracellular FP were recorded from hipsc-cm following treatment with vehicle and 15 ion channel blockers with various selectivity profiles (Table 1 and Supplementary Table 2). Ten compounds were typical herg blockers with a range of potencies (IC50: 7 nm to 330 mm). Some of the compounds had similar potencies for herg and NaV1.5 channels (3-fold), and would be classified as mixed channel blockers. All the drugs had negligible effects on L-type Ca 2þ channel function (IC 50 > 30 mm) (Supplementary Table 1). Correlation of FPD Prolongation With herg Potency The sensitivity of hipsc-cm to detect herg-related FPD prolongation was recorded following treatment with all 15 ion channel inhibitors. For example, the effect of sertindole is shown in Figure 1 and Table 2. Sertindole decreased the amplitude of the repolarization waveform at 30 nm, and reduced it to a flat line at 100 and 300 nm (Fig. 1). This agent also reduced Na þ slope, Na þ amplitude, and beat rate in a concentration-dependent manner. At concentrations of 100 and 300 nm, EADs were detected. The MEC for FPD prolongation (MEC-FPD) was derived for 13 compounds; lamotrigine and L did not alter FPD. A correlation plot of MEC-FPD versus herg-ic 50 for ten typical herg blockers demonstrated significance (Fig. 2A; R 2 ¼ 0.89, P <.0001). However, when all 13 agents (Ten herg blockers plus flecainide, mexiletine, and AMG1) were included in the correlation analysis, the correlation was not significant (Fig. 2B; R 2 ¼ 0.03, P ¼.56). This indicates that FPD detection in hipsc-cm correlates well with herg potency for selective herg blockers, but does not correlate well for multichannel blockers. Assessment of Proarrhythmic Risk-Based on Clinical Exposure Multiples The value of FPD prolongation and EAD in hipsc-cm for predicting proarrhythmic risk was evaluated by comparing MEC values to clinical exposures. The ratio of MEC-FPD/human C max was calculated for 12 drugs (Fig. 3A; Supplementary Table 1); some agents were not included because of lack of human exposure data (AMG1, L768673) or no effect on FPD was observed (lamotrigine). The same approach was used for the ratio of MEC-EAD/ C max (Fig. 3B; Supplementary Table 1). Some drugs (terodiline, moxifloxacin, and mexiletine) were excluded because EADs were not observed at their highest testing concentrations. Compounds were categorized into 3 proarrhythmia categories based on the ratio: low risk (30); intermediate risk (>10, but <30); high risk (<10). Based upon the risk categorization ratio, assay performance of FPD and EAD for predicting TdP risk was calculated (Table 3). As shown in Table 3, at a ratio of < 10, FPD incorrectly identified terodiline as a TdP negative, but mexiletine and ranolazine as TdP positives. At a ratio of > 30, FPD prolongation correctly identified all TdP positive compounds, but ranolazine was labeled TdP positive incorrectly. For EAD, 3 TdP positive compounds, sertindole, cisapride and terfenadine, were all categorized as being TdP negative at 10- and 30-fold ratio; ranolazine was mis-called as TdP positive at a 30-fold ratio. At > 100 ratio, false negatives were eliminated, but ranolazine was still a false positive. Consistent with initial observations, FPD had high sensitivity, but specificity was poor (Table 3). EAD had low sensitivity at 10- and 30-fold ratio, but high sensitivity at 100-fold ratio. Specificity of EAD also depended upon the ratio, which was high at 10-fold ratio, but low at 30- and 100-fold ratios (Table 3). Both FPD and EAD had high positive predictive values at all ratios, however, FPD had a low negative predictive value at TABLE 1. Ion Channel Profiles and TdP Risk of Test Articles IC 50 (nm) Selectivity TdP Risk Compound herg NaV1.5 ICa 2þ IKs herg/nav1.5 herg/ica 2þ Sertindole > ND < þ Dofetilide 70 > > ND < < þ Tolterodine > ND < Cisapride > ND < þ Terfenadine > ND 0.04 < þ Terodiline > ND 0.1 < þ Alfuzosin > ND 0.11 < 0.23 Ranolazine > ND 0.62 < 0.5 Sotalol (D,L) > > ND < 1.7 < 1.7 þ Moxifloxacin > > < 0.74 < 2.4 þ Lamotrigine > ND 2.74 < 0.36 Flecainide > < þ Mexiletine > < 0.71 AMG ND ND 5.5 ND ND L > > ND 27 ND ND ND þ, TdP positive;, TdP negative; ND, not determined.

4 QU AND VARGAS 289 FIG. 1. Field potential traces from hipsc-cm treated with sertindole. Cells were treated with 0, 30, 100, and 300 nm sertindole (as shown). Each waveform has a fast spike (Na þ spike), a repolarization waveform (blue arrows), and EAD waveforms (red arrows). Calibration: 1 mv; 2 s. Sertindole prolonged FPD, decreased Na þ amplitude, decreased beat rate, and triggered EAD in a concentration-dependent manner. TABLE 2. Effects of Sertindole on FP Endpoints in hipsc-cm Testing Conc. (nm) Beat Period Na þ Slope Na þ Amplitude FPD EAD amplitude at 3 and 10 nm, which is times lower than NaV1.5 potency (Table 1) * * * * þ * * * * þ MEC (nm) Note: Data are mean 6 SEM and expressed as % of vehicle. *Significant effect. FPD, field potential duration; EAD, early after-depolarization. 10-fold ratio, and EAD had low negative predictive values at 10- and 30-fold ratios (Table 3). Correlation of Na 1 Slope With Nav1.5 Potency Changes of Na þ slope and Na þ amplitude related to sodium channel blockade was recorded by MEA for all 15 agents. For correlation analysis, compounds with no Nav1.5 IC 50 or MEC for decreasing Na þ slope were excluded. The relationship of Naþ slope reduction in hipsc-cm with NaV1.5 potencies for 4 sodium channel blockers is depicted in Figure 4A; no significant correlation was observed (R 2 ¼ 0.13; P ¼.64). Inclusion of 4 mixed herg/nav1.5 inhibitors (sertindole, terodiline, alfuzosin, ranolazine) with a defined MEC for decreasing Na þ slope and a defined IC 50 for Nav1.5 did not improve the correlation between Na þ slope reduction and Nav1.5 inhibitory potency Figure 4B (R 2 ¼ 0.14, P ¼.37). Analysis of Na þ slope and amplitude demonstrated that dofetilide reduced these parameters, which was unexpected (Fig. 5; Table 4). Dofetilide significantly decreased Na þ slope and Effects of IKs Blockers on FPD in hipsc-cm The potent and selective IKs channel blocker, L was tested for ability to prolong FPD. The concentration response curves for percent inhibition of IKs current and percent changes in FPD measured in hipsc-cm were compared (Fig. 6). L inhibited IKs current potently (IC 50 : 26.9 nm), but did not prolong FPD up to 0.3 mm, a concentration that would be expected to cause maximal inhibition of IKs channels. Beat Rate of hipsc-cm The ability of all 15 ion channel blockers to alter beat rate of hipsc-cm was evaluated, but no consistent pattern was observed. For example, 7 of 10 herg blockers decreased beat rate, but 3 caused an increase (moxifloxacin, ranolazine, terodiline). Of the sodium channel blockers, 3 of 4 decreased beat rate, but lamotrigine caused an increase. Treatment with the IKs inhibitor L had no effect on beat rate. Reproducibility of Responses To understand the reproducibility of responses detected by MEA in hipsc-cm, FPD and beat period changes induced by quinidine were evaluated. Quinidine was included on each MEA plate and used as a positive reference or internal control to assess plate-toplate responses (Fig. 7). The magnitude of changes and concentration response relationships in FPD and beat period recorded in 5 different plates exhibited wide variation in response to quinidine treatment; however, MEC values were reproducible with maximal 2-fold differences among 5 plates (Table 5).

5 290 TOXICOLOGICAL SCIENCES, 2015, Vol. 147, No. 1 FIG. 2. Correlation of FPD prolongation in hipsc-cm with herg potency. MEC-FPD was plotted against the herg-ic 50 determined with PatchXpress. (A) Ten herg blockers (black square). (B) Ten herg blockers plus 3 multichannel blockers (red circle) included. Reference lines: 1:1 (green) and 1:3 (blue). FIG. 3. Clinical exposure multiples of FPD prolongation and EAD induction in hipsc-cm. (A) Ratios of FPD prolongation and (B) ratios of EAD induction. Colored lines indicate risk categories: green is low risk (ratio 30), red is high risk (ratio < 10). TABLE 3. Assay Performance of FPD and EAD in Predicting TdP Risk FPD EAD TdP Sertindole Dofetilide Cisapride Terfenadine Terodiline Sotalol (D,L) Moxifloxacin Flecainide Mexiletine Tolterodine Alfuzosin Ranolazine Sensitivity Specificity Positive predictive value Negative predictive value ,30, 100 refer to the MEC/Cmax ratio. Red: positive. Green: negative. White: not tested and not included in performance calculations. DISCUSSION FPD Prolongation Is Mediated Primarily by herg Channels in hipsc-cm Prior reports in hipsc-cm have demonstrated that herg channel blockade can cause FPD prolongation (Asai et al., 2010; Braam et al., 2010; Caspi et al., 2009; Clements and Thomas, 2014; Harris et al., 2013). Consistent with those reports, all ten herg blockers tested in this current study prolonged FPD in hipsc-cm at a pharmacological concentration that correlated with herg channel inhibition. The analysis clearly supports the use of hipsc-cm as a human tissue model to detect QT interval prolongation risk induced by herg channel inhibition. However, when 3 multichannel blockers (flecainide, mexiletine, AMG1) were included, the correlation became nonsignificant, because the MEC for delayed repolarization occurred at a much lower concentration than herg channel potency. For example, the MEC for FPD prolongation of flecainide was 0.1 um, which was 16-fold lower than its herg potency value (IC50: 1.6 mm). Therefore, FPD measurement in hipsc-cm could be used to assess repolarization delay elicited by specific herg inhibitors, but this endpoint could overestimate the repolarization risk for agent with a complex ion channel profile, eg, mixed herg/ NaV1.5 blockers. All of the agents had negligible inhibitory

6 QU AND VARGAS 291 FIG. 4. Correlation of Na þ slope reduction in hipsc-cm with Nav1.5 potency. Concentrations that caused significant reduction of Na þ slope were plotted against Nav1.5-IC 50 determined by PatchXpress. (A) Four sodium channel blockers. (B) Four sodium channel blockers plsu 4 herg blockers. Compounds with no Nav1.5 IC 50 or MEC for decreasing Na þ slope were excluded. FIG. 5. Dofetilide decreased Na þ amplitude of FP traces in hipsc-cm. Traces from cells treated with 0.3, 1, 3, and 10 nm dofetilide are shown. Each waveform has a fast spike (Naþ spike), a repolarization waveform (blue arrows), and EAD waveforms (red arrows). Calibration: 2 mv; 2 s. Dofetilide decreased Naþ amplitude, beat rate, and triggered EAD in a concentration-dependent manner. effects on CaV1.2 channels; therefore calcium channel blockade did not contribute to FPD findings in this study. Proarrhythmia Risk Assessment in hipsc-cm: Not Ready for Primetime Drug-induced proarrhythmia was assessed by determining the MEC for FPD and EAD induction in hipsc-cm. Risk assessment was performed by comparing the MEC to known clinical exposure, eg, MEC-FPD/C max or MEC-EAD/C max ratio. This approach was similar to prior investigations of TdP risk that used herg-ic 50 /C max ratio as a metric (Gintant et al., 2011; Redfern et al., 2003). Based on MEC-FPD potency relative to the clinical exposure, drugs with a ratio > 30 were considered low risk, and drugs with a ratio < 10 were classified as high risk for TdP. In the MEA test system, 7 of 8 torsadogens were classified as high risk, and

7 292 TOXICOLOGICAL SCIENCES, 2015, Vol. 147, No. 1 TABLE 4. Effects of Dofetilide on FP Endpoints in hipsc-cm Defetilide (nm) FPD Na þ Slope Na þ Amplitude Beat Period EAD None * None * * * * þ * * * * þ 30 NB NB NB NB NB MEC (nm) Note: Data are mean 6 SEM and expressed as % of vehicle. *Significant effect. FPD, field potential duration; EAD, early after-depolarization; NB, no beating. TABLE 5. MEC of Quinidine Determined in 5 Different Plates MEC FPD Beat Period Plate Plate Plate Plate Plate FIG. 6. L had no effect on FPD in hipsc-cm (blue). Concentration-response curve for KvLQT1/minK current inhibition (black) was fitted by Michaelis Menten equation (E ¼ E max * [L768673]/ (IC 50 þ [L768673])) with E max ¼ 100, and IC 50 ¼ 26.9 nm. Red dotted line represents 10% effects. FIG. 7. Reproducibility of responses: variability of FPD and beat period with positive control, quinidine. Left panel: FPD (% of vehicle) data from 5 different plates; right panel: beat period (% of vehicle) from 5 different plates. For plate 3, there was no beating at and 2.5 mm. 100% were identified at 30-fold. In contrast, 2 of 4 drugs known to have low risk for TdP were misclassified as having risk, ie, ranolazine and mexiletine. These latter drugs are widely accepted as having no TdP risk in clinical practice (Redfern et al., 2003; Sossalla and Maier, 2012), so the high risk categorization is problematic. Ranolazine can block herg function, and prolong QT interval, but does not induce TdP; in fact, it suppresses long QT related ventricular arrhythmias in experimental models

8 QU AND VARGAS 293 (Sossalla and Maier, 2012). Extensive electrophysiological studies demonstrated that ranolazine inhibits late Na þ currents, which opposes the APD- and QT-prolonging effects of herg block (Antzelevitch et al., 2011). Mexiletine is an antiarrhythmic drug used to treat abnormal heart rhythms by blocking cardiac sodium channels (Brunton et al, 2011), and is not associated with TdP in humans (Redfern et al., 2003). Collectively, the present findings indicate FPD prolongation in hipsc-cm has high sensitivity, but low specificity for TdP prediction; thus, this metric (FPD/C max ratio) has the potential to misclassify safe compounds, ie, high false-positive rate. An advantage of MEA recording in hipsc-cm is the ability to detect repolarization waveform abnormalities that resemble EAD. Early depolarization of ventricular myocytes represents a trigger event that has been implicated as the primary mechanism for ventricular arrhythmia induction in acquired and congenital long QT syndromes, including TdP (Weiss et al., 2010). Interestingly, at the concentration ranges used in our study, 9 compounds elicited EAD in hipsc-cm. However, when the proarrhythmia ratio based on EAD was considered, sertindole, tolterodine, cisapride, terfenadine, and alfuzosin had values >30. At this cutoff level, 60% of the TdP-positive drugs were mislabelled as false negatives (Kramer et al., 2013; Redfern et al., 2003). These compounds were correctly identified at supratherapeutic concentrations (>100-fold), which suggests that EAD induction is less sensitive than FPD for proarrhythmia detection. Calculation of sensitivity and specificity based upon FPD and EAD changes confirmed these observations (Table 3). FPD generally exhibited high sensitivity, and low specificity. In regard to performance, EAD had similar performance characteristics as the FPD endpoint, that is low sensitivity at ratios <100 coupled with low specificity at high multiples. Overall, FPD and EAD exhibited high positive predictive values, but low negative predictive values; thus hipsc-cm will have the propensity to generate false positives in predicting TdP risk, which will lead to undeserved attrition of new drug candidates. Discordance of Na 1 Spike and NaV1.5 Channel Inhibition The initial upward potential detected by MEA recording (Na þ spike) is believed to represent the depolarization phase of the ventricular action potential, which is initiated and sustained by cardiac Na þ channels (Zipes and Jalife, 2013). Inhibition of sodium channels in hipsc-cm would be predicted to decrease Na þ slope and Na þ amplitude. However, the present study indicated that there was no significant correlation between hnav1.5 potency and reduction of the Na þ slope parameter. Furthermore, the discordance between NaV1.5 potency and Na þ slope effects was independent of the ion channel profile of the compounds. For example, the very selective herg blockers, dofetilide, and sertindole, also depressed Na þ slope and Na þ amplitude, functional effects unrelated to their Na þ channel potency. The apparent effects of dofetilide on cardiac conduction in hipsc-cm do not translate to human clinical evidence, because this drug has no effect on PR or QRS intervals at therapeutic drug levels (Falk and DeCara, 2000). In our previous study of human embryonic SC-CM using patch clamp (Qu et al., 2013), the inhibitory effect of Na þ channel blockers was much less potent on decreasing action potential upstroke velocity as compared with inhibitory potency in the NaV1.5 current assay. Consistent with the literature (He et al., 2003; Jonsson et al., 2012; Peng et al., 2010; Satin et al., 2004), the diminished Na þ currents and the lower upstroke velocity in SC- CM would contribute to the lower sensitivity of this cell-type to Na þ channel inhibition. Most of the drugs decreased Na þ slope more potently, at concentrations below the NaV1.5-IC50 value, and this was especially true for dofetilide. Does this mean that the Na þ spike parameter is more sensitive to hnav1.5 inhibition, or is it related to unique electrophysiological characteristics of hipsc-cm? It is likely the latter, because the action potential of hsc-cm is different from adult (mature) cardiomyocytes (He et al., 2003; Hoekstra et al, 2012; Jonsson et al., 2012; Peng et al., 2010; Satin et al., 2004). For example, SC-CM displays spontaneous phase 4 depolarization, which is not seen in adult normal ventricular myocytes, The action potential of SC-CM is different from adult mature cardiomyocyte (He et al., 2003; Hoekstra et al, 2012; Jonsson et al., 2012; Peng et al., 2010; Satin et al., 2004). Consistent with these features, SC-CM are reported to have reduced expression of transient outward K current (Ito), slow component of delayed rectifier K current (IKs), and inward rectifier K currents (IK1) (Liang et al., 2013). IK1 is the most important current responsible for the cardiac myocyte resting membrane potential, so decreased expression could explain the depolarized membrane potential of hsc-cm (Doss et al., 2012; Roden and Hong, 2013). Therefore in contrast to normal adult ventricular myocytes, hsc-cm have a repolarization and maximum diastolic potential that depend critically on IKr due to the absence of IK1 and decreased contribution of Ito and IKs (Doss et al., 2012). When IKr is inhibited by herg blockers, there are minimal potassium currents reserved to repolarize the membrane potential in hsc-cm, thus a more depolarized diastolic potential would inactivate Na channels and impede depolarization. Therefore agents that block herg would decrease Na þ spike recorded with MEA in hipsc-cm. The unique physiology and pharmacological response of hsc-cm would lead to the misidentification of a highly selective herg inhibitor as a mixed ion channel blocker with a liability for both delayed ventricular conduction and depolarization. No Effect on Repolarization Following IKs Inhibition A prior study indicated that APD recordings in human embryonic SC-CM were not sensitive to IKs inhibition following treatment with L (Qu et al., 2013). In this study, the FPD from hipsc-cm was also insensitive to L Liang et al. (2013) have presented convincing data that hipsc-cm and hesc-cms have low expression of KCNQ1 and KCNE1 by quantitative PCR compared with adult human ventricular tissue. KCNQ1 and KCNE1 encode the slow component of delayed rectifier potassium channel responsible for IKs (Sanguinetti et al., 1996). Our pharmacological studies indicate that hipsc-cm and hesc-cm lack sensitivity for detecting action potential prolongation related to IKs block due to lower channel expression. LIMITATIONS There were several limitations in this study: (1) Our evaluation and conclusions are limited by the number of agents tested, especially the assessment of only 4 Na þ channel blockers and 1 IK blocker. A larger set of test compounds, including both TdPpositive and negative controls, would expand the qualification dataset, and assist in understanding the performance of MEA recordings in hipsc-cm as a new model for assessing proarrhythmic risk; (2) Some of the agents tested, including moxifloxacin, mexiletine, and terodiline, were not tested at high enough concentrations, which led to FPD and EAD data gaps; however this did not interfere with the overall conclusions of the study, in our opinion; (3) The definition of MEC was based upon an effect size of 10% that was applied to all study endpoints;

9 294 TOXICOLOGICAL SCIENCES, 2015, Vol. 147, No. 1 determination of effect sizes based on the variability characteristics of each parameter, eg, FPD versus beat rate, was not considered; (4) The magnitude of effects detected vary greatly in different plates based on responses to quinidine, however, the MEC determined is reproducible across plates, which would not impact ratio estimates from study to study; (5) The use of a single estimate of human C max data is a potential limitation, because there are multiple sources for clinical exposure data, which could introduce selection bias into the ratio calculation; (6) The findings in this study depict response for a particular type of hipsc-cm (icell), and may be different for other cell sources. CONCLUSIONS Although recent publications (Clements and Thomas, 2014; Harris et al, 2013) indicate a great potential for MEA recordings in hsc-cm to assess nonclinical proarrhythmic risk, the current evaluation demonstrates that this model has a high false-positive rate which will confound risk assessment, and could lead to premature termination of drug candidates. Our conclusions are supported by recent publications. Du et al. (2015) provided evidence that ipsc-cms possess unique action potential morphology and they do not predict atrial, ventricle, or nodal cells. In regard to the spontaneous beat rate phenomena, Kim et al. (2015) demonstrated that hips-cms have negligible I f (funny) current, which is the primary pacemaker current responsible for sino-atrial automaticity in the normal heart; in contrast, automaticity of hips-cm is initiated by intracellular Ca 2þ release from the sarcoplasmic reticulum linked to the Na-Ca exchanger. These observations are consistent with our findings that MEA recording in hipsc-cm was unable to differentiate Na þ channel blockade from herg blockade due to reduced repolarization reserve in hsc-cm, and that beat rate was a highly variable parameter in response to ion channel inhibition. Therefore, there continues to be an on-going need to perform extensive pharmacological evaluation of hsc-cm (eg, hesc-cm and hipsc-cm) to understand fully the translation of stem-cell proarrhythmia data to experimental findings in conventional nonclinical test systems, and human clinical investigations. SUPPLEMENTARY DATA Supplementary data are available online at oxfordjournals.org/. ACKNOWLEDGMENTS We thank Cyprotex US, LLC for executing this study, BaoXi Gao and Mei Fang at Amgen Inc. for assisting in test article preparation and handling, and Thomas Monticello and Michael Engwall at Amgen Inc. for reviewing and editing the manuscript. FUNDING This study is funded by Amgen Inc. REFERENCES Abassi, Y. A., Xi, B., Li, N., Ouyang, W., Seiler, A., Watzele, M., Kettenhofen,R., Bohlen, H., Ehlich, A., and Kolossov, E., et al. (2012).Dynamicmonitoring of beating periodicity of stem cell-derived cardiomyocytes as a predictive tool for preclinical safety assessment. Br. J. Pharmacol. 165, Antzelevitch, C., Burashnikov, A., Sicouri, S., and Belardinelli, L. (2011). Electrophysiologic basis for the antiarrhythmic actions of ranolazine. Heart Rhythm 8, Asai, Y., Tada,M., Otsuji, T. G., and Nakatsuji, N. (2010). Combination of functional cardiomyocytes derived from human stem cells and a highly-efficient microelectrode array system: an ideal hybrid model assay for drug development. Curr. Stem Cell Res. Ther. 5, Braam, S. R., Tertoolen, L., van de Stolpe, A., Meyer, T., Passier, R., and Mummery, C. L. (2010). Prediction of drug-induced cardiotoxicity using human embryonic stem cell-derived cardiomyocytes. Stem Cell Res. 4, Brunton, L. L., Blumenthal, D. K., Murri, N., Dandan, R. H., and Knollmann, B. C. (2011). Goodman & Gilman s The Pharmacological Basis of Therapeutics, 12th ed. McGraw-Hill, New York. Caspi, O., Itzhaki, I., Kehat, I., Gepstein, A., Arbel, G., Huber, I., Satin, J., and Gepstein, L. (2009). In vitro electrophysiological drug testing using human embryonic stem cell derived cardiomyocytes. Stem Cells Dev. 18, Cerignoli, F., Charlot, D., Whittaker, R., Ingermanson, R., Gehalot, P., Savchenko, A., Gallacher, D. J., Towart, R., Price, J. H., McDonough, P. M., et al. (2012). High throughput measurement of Ca2þ dynamics for drug risk assessment in human stem cell-derived cardiomyocytes by kinetic image cytometry. J. Pharmacol. Toxicol. Methods, 66, Clements, M., and Thomas, N. (2014). High-throughput multi-parameter profiling of electrophysiological drug effects in human embryonic stem cell derived cardiomyocytes using multi-electrode arrays. Toxicol. Sci. 140, Doss, M. X., Diego, J. M., Goodrow, R. J., Wu, Y., Cordeiro, J. M., Nesterenko, V. V., Barajas-Martinez, H., Hu, D., Urrutia, J., and Desai, M., et al. (2012) Maximum diastolic potential of human induced pluripotent stem cell-derived cardiomyocytes depends critically on IKr. PLoS One 7, e Du, D. T., Hellen, N., Kane, C., and Terracciano, C. M. (2015). Action potential morphology of human induced pluripotent stem cell-derived cardiomyocytes does not predict cardiac chamber specificity and is dependent on cell density. Biophys J. 108, 1 4. Falk, R. H., and DeCara, J. M. (2000). Dofetilide: a new pure class III antiarrhythmic agent. Am. Heart J. 140, Gintant, G. A., Gallacher, D. J., and Pugsley, M. K. (2011). The overly-sensitive heart: sodium channel block and QRS interval prolongation. Br. J. Pharmacol. 164, Guo, L., Abrams, R. M. C., Babiarz, J. E., Cohen, J. D., Kameoka, S., Sanders, M. J., Chiao, E., and Kolaja, K. L. (2011). Estimating the risk of drug-induced proarrhythmia using human induced pluripotent stem cell derived cardiomyocytes. Toxicol. Sci. 123, Harris, K., Aylott, M., Cui, Y., Louttit, J. B., McMahon, N. C., and Sridhar, A. (2013). Comparison of electrophysiological data from human-induced pluripotent stem cell-derived cardiomyocytes to functional preclinical safety assays. Toxicol. Sci. 134, He, J. Q., Ma, Y., Lee, Y., Thomson, J. A., and Kamp, T. J. (2003). Human embryonic stem cells develop into multiple types of cardiac myocytes: action potential characterization. Circ. Res. 93, Hoekstra, M., Mummery, C. L., Wilde, A. A., Bezzina, C. R., and Verkerk, A. O.(2012). Induced pluripotent stem cell derived

10 QU AND VARGAS 295 cardiomyocytes as models for cardiac arrhythmias. Front. Physiol. 3, 346. ICH S7B: International Conference on Harmonisation of Technical Requirements for Registration of Pharmaceuticals for Human Use. ICH Steering Committee. The non clinical evaluation of the potential for delayed ventricular repolarization (QT prolongation) by human pharmaceuticals S7B. Available at: Accessed June 10, ICH E14: Guidance for Industry: E14 Clinical Evaluation of QT/ QTc Interval Prolongation and Proarrhythmic Potential for Non-Antiarrhythmic Drugs RegulatoryInformation/Guidances/ucm pdf Ivashchenko, C. Y., Pipes, G. C., Lozinskaya, I. M., Lin, Z., Xu, X., Needle, S., Grygielko, E., Hu, E., Toomey, J., Lepore, J., et al. (2013). Human-induced pluriopotent Stem cell-derived cardiomyocytes exhibit temporal changes in phenotype. Am. J. Physiol. Heart Circ. Physiol. 305, H913 H922. Jonsson, M. K., Vos, M. A., Mirams, G. R., Duker, G., Sartipy, P., de Boer, T. P., van Veen, T. A. B. (2012). Application of human stem cell-derived cardiomyocytes in safety pharmacology requires caution beyond herg. J. Mol. Cell. Cardiol. 52, Kehat, I., Kenyagin-Karsenti, D., Snir, M., Segev, H., Amit, M., Gepstein, A., Livne, E., Binah, O., Itskovitz-Eldor, J., and Gepstein, L. (2001). Human embryonic stem cells can differentiate into myocytes with structural and functional properties of cardiomyocytes. J. Clin. Invest. 108, Kim, J. J., Yang, L., Lin, B., Zhu, X., Sun, B., Kaplan, A. D., Bett, G. C., Rasmusson, R. L., London, B., and Salama, G. (2015). Mechanism of automaticity in cardiomyocytes derived from human induced pluripotent stem cells. J. Mol. Cell. Cardiol. 81, Kramer, J., Obejero-Paz, C. A., Myatt, G., Kuryshev, Y. A., Bruening-Wright, A., Verducci, J. S., and Brown, A. M. (2013). MICE models: superior to the HERG model in predicting Torsade de Pointes. Sci. Rep. 3, Liang, P., Lan, F., Lee, A. S., Gong, T., Sanchez-Freire, V., Wang, Y., Diecke, S., Sallam, K., Knowles, J. W., Wang, P. J., et al. (2013). Drug screening using a library of human induced pluripotent stem cell-derived cardiomyocytes reveals disease-specific patterns of cardiotoxicity. Circulation. 127, Ma, J., Guo, L., Fiene, S. J., Anson, B. D., Thomson, J. A., Kamp, T. J., Kolaja, K. L., Swanson, B. J., and January, C. T. (2011). High purity human-induced pluripotent stem cell-derived cardiomyocytes: electrophysiological properties of action potentials and ionic currents. Am. J. Physiol. Heart Circ. Physiol. 301, H2006 H2017. Peng, S., Lacerda, A. E., Kirsch, G. E., Brown, A. M., and Bruening- Wright, A. (2010). The action potential and comparative pharmacology of stem cell-derived human cardiomyocytes. J. Pharmacol. Toxicol. Methods 61, Puppala, D., Collis, L. P., Sun, S. Z., Bonato, V., Chen, X., Anson, B., Pletcher, M., Fermini, B., and Engle, S. J. (2013). Comparative gene expression profiling in human induced pluripotent stem cell derived cardiocytes and human and cynomolgus heart tissue. Toxicol. Sci. 131, Qu, Y., Gao, B., Fang, M., and Vargas, H. V. (2013). Human embryonic stem cell derived cardiac myocytes detect herg-mediated repolarization effects, but not Nav1.5 induced depolarization delay. J. Pharmacol. Toxicol. Methods 68, Rao, C., Prodromakis, T., Kolker, L., Chaudhry, U. A., Trantidou, T., Sridhar, A., Weekes, C., Camelliti, P., Harding, S. E., and Darzi, A., et al. (2013). The effect of microgrooved culture substrates on calcium cycling of cardiac myocyte derived from human induced pluripotent stem cells. Biomaterials 34, Redfern, W. S., Carlsson, L., Davis, A. S., Lynch, W. G., MacKenzie, I., Palethorpe, S., Siegl, P. K., Strang, I., Sullivan, A. T., Wallis, R., et al. (2003). Relationships between preclinical cardiac electrophysiology, clinical QT interval prolongation and torsade de pointes for a broad range of drugs: evidence for a provisional safety margin in drug development. Cardiovasc. Res. 58, Roden, D. M., and Hong, C. C. (2013). Stem cell-derived cardiomyocytes as a tool for studying proarrhythmia: a better canary in the coal mine? Circulation 127, Sager, P. T., Gintant, G., Turner J. R., Pettit S., and Stockbridge, N. (2014). Rechanneling the cardiac proarrhythmia safety paradigm: a meeting report from the Cardiac Safety Research Consortium. Am. Heart J. 16, Sanguinetti, M. C., Curran, M. E., Zou, A., Shen, J., Spector, P. S., Atkinson, D. L., and Keating, M. T. (1996). Coassembly of K(V)LQT1 and mink (IsK) proteins to form cardiac I(Ks) potassium channel. Nature 384, Satin, J., Kehat, I., Caspi, O., Huber, I., Arbel, G., Itzhaki, I., Magyar, J., Schroder, E. A., Perlman, I., and Gepstein, L. (2004). Mechanism of spontaneous excitability in human embryonic stem cell derived cardiomyocytes. J. Physiol. 559, Sossalla, S., and Maier, L. S. (2012). Role of ranolazine in angina, heart failure, arrhythmias, and diabetes. Pharmacol. Ther. 133, Takahashi, K., and Yamanaka, S. (2006). Induction of pluripotent stem cells from mouse embryonic and adult fibroblast cultures by defined factors. Cell 126, Weiss, J. N., Garfinkel, A., Karagueuzian, H. S., Chen, P., and Qu, Z. (2010). Early afterdepolarizations and cardiac arrhythmias. Heart Rhythm 7, Zhang, D., Shadrin, I. Y., Lam, J., Xian, H. Q., Snodgrass, H. R.,and Bursac, N. (2013). Tissue-engineered cardiac patch for advanced functional maturation of human ESC-derived cardiomyocytes. Biomaterials 34, Zipes, D. P., and Jalife, J. (2013). Cardiac Electrophysiology: from Cell to Bedside Elsevier Health Sciences.