Effects of proarrhythmic drugs on relaxation time and beating pattern in rat engineered heart tissue

The assessment of proarrhythmic risks of drugs remains challenging. To evaluate the suitability of rat engineered heart tissue (EHT) for detecting proarrhythmic effects. We monitored drug effects on spontaneous contractile activity and, in selected cases, on action potentials (sharp microelectrode) and Ca2+ transients (Fura-2) and contraction under electrical pacing. The Ito-blocker inhibitor 4-aminopyridine increased action potential duration and T2 and caused aftercontractions, which were abolished by inhibitors of ryanodine receptors (RyR2; JTV-519) or sodium calcium exchanger (NCX; SEA0400). 77 Drugs were then tested at 1-10-100× free therapeutic plasma concentrations (FTPC): Inhibitors of IKr, IKs, Ito, antiarrhythmics (8), drugs withdrawn from market for torsades des pointes arrhythmias (TdP, 5), drugs with measurable (7) or isolated TdP incidence (13), drugs considered safe (14), 28 new chemical entities (NCE). Inhibitors of IKr or IKs had no effect alone, but substantially prolonged relaxation time (T2) when combined at high concentration. 15/33 drugs associated with TdP and 6/14 drugs considered non-torsadogenic (cibenzoline, diltiazem, ebastine, ketoconazole, moxifloxacin, and phenytoin) induced concentration-dependent T2 prolongations (10-100× FTPC). Bepridil, desipramine, imipramine, thioridazine, and erythromycin induced irregular beating. Three NCE prolonged T2, one reduced force. Drugs inhibiting repolarization prolong relaxation in rat EHTs and cause aftercontractions involving RyR2 and NCX. Insensitivity to IKr inhibitors makes rat EHTs unsuitable as general proarrhythmia screen, but favors detection of effects on Ito, IKs + Ito or IKs + IKr. Screening a large panel of drugs suggests that effects on these currents, in addition to IKr, are more common than anticipated. Electronic supplementary material The online version of this article (doi:10.1007/s00395-014-0436-7) contains supplementary material, which is available to authorized users.


Introduction
Proarrhythmic side effects of drugs can be life-threatening and have led to a number of drug withdrawals from the market [17]. Predicting the arrhythmogenic potential in preclinical drug development is difficult for several reasons [25,37]. (1) Mechanisms of arrhythmias are complex. This is exemplified by the fact that not only loss-of-function, but also gain-of-function mutations of K ? -and Na ? -channels can cause arrhythmias by promoting ectopic activity, dispersion, and/or re-entry circuits [1]. This corresponds with clinical experiences that both class I (Na ? -channel blockers) and class III (K ? -channel blockers) antiarrhythmic drugs have significant proarrhythmic effects [10]. In addition, dysfunction of the sarcoplasmic reticulum Ca 2?release channel (RyR2) or the SR Ca 2? -storage protein calsequestrin underlie catecholaminergic polymorphic ventricular tachycardia, characterized by increased ectopic automaticity in situations of stress. (2) The heart is equipped with several safety mechanisms, explaining why one hit is rarely sufficient to cause symptomatic arrhythmias. Even patients with inherited rhythm disorders experience clinically relevant arrhythmias relatively late in life and/or only under certain trigger situations such as increased sympathetic drive, hypokalemia, drugs, ischemia, or myocardial scars. (3) The existing preclinical test systems have shortcomings [14,25,37]. In current routine, new chemical entities (NCE) are tested on cells overexpressing the human eag-related gene (hERG), and those with significant inhibitory activity are excluded from further development. Whereas hERG-tests have documented high sensitivity and specificity for this single ion current and the principal relevance for hERG-inhibition in causing torsades des pointes (TdP) is undisputed, the predictive value of this test is limited [21,31]. Several drugs have hERG-inhibitory activity without being associated with TdP-arrhythmias (e.g., verapamil). Others have no relevant hERG-activity at clinically used concentrations, but increase the risk of arrhythmias (e.g., mefloquine and phenytoin; [31]).
The Food and Drug Administration and the European Medicines Agency currently recommend an integrated risk assessment, which includes the results of several experimental tests (e.g., hERG, rabbit Purkinje fibers, and dog telemetry) as well as in silico and clinical data [11,12,38]. The content of current tests may be improved by characterizing NCEs in a whole panel of cell lines, each expressing a different cloned ion channel. This approach provides a more comprehensive picture of a drugs channelaffecting activity [3,42], but needs modeling to predict the integrated effect. An alternative strategy is to test drugs directly on cardiac myocytes or more complex cardiac tissues as the ''real substrate'' for arrhythmias, assuming that any reproducible effect is relevant, independent of its exact mechanism. Unfortunately, isolated adult cardiac myocytes do not beat and, similar to Purkinje fibers or Langendorff-perfused hearts, cannot be examined in large series. Instrumented dogs or rabbits are the most valid models, but cannot be used for screening large number of NCEs, both for ethical and financial reasons.
We have recently developed an automated miniaturized drug screening assay based on our EHT technology and neonatal rat cardiac myocytes, which appears to combine some of the advantages of a relatively intact 3D cardiac tissue, availability at large numbers, robustness and highcontent readout, particularly analysis of contractile force and kinetics [7,16,20]. Experiments with a limited number of model compounds indicated assay sensitivity to detect proarrhythmic effects of drugs [16]. The aim of the present study was to systematically determine the predictive value of the assay by testing a larger number of clinically used compounds with characterized hERG-inhibitory activity and proarrhythmic potential in humans [31] and randomly picked NCEs as well as underlying mechanisms.

Materials and methods
The investigation conforms to the guide for the care and use of laboratory animals published by the NIH (Publication No. 85 -23, revised 1985). A detailed description of methods can be found in the supplemental file.

Cell isolation and EHT-generation
Heart cells of postnatal d0-d3 Wistar rats and EHTs were prepared as previously described [16,44]. Fresh neonatal heart cells were mixed with medium, fibrinogen and thrombin, and casted into strip-format (12 9 3 9 3 mm) molds in agarose, in which pairs of elastic silicone posts were placed from above. EHTs maintained for up to 4 weeks.

Measurement of contractile parameters
Contractile parameters were evaluated as previously described [16]. In principle, the 24-well-plates with EHTs (14-21 days) were put in a gas-, temperature-, and humidity-controlled incubator with glass roof and customized software-controlled video camera placed on top. Contractile parameters of spontaneously beating EHTs were evaluated using an automated figure recognition algorithm. Deflection of the silicone posts was recorded over time and, based on post geometry and elastic modulus of the silicone, used to calculate force, frequency, fractional shortening, contraction and relaxation time (bpm, T1 [from 20 % to peak] and T2 [from peak to 20 %], respectively).

Drug screening-video optical analysis
All measurements were performed with 14-21 day old EHTs in fresh serum-free DMEM (Biochrom F04115), supplemented with 10 mM HEPES for pH-steadiness, preincubated at 37°C, 40 % O 2 , 7 % CO 2 , 90 % humidity for 2 h. Measurements were done routinely 1 day after feeding with standard EHT medium. The drugs were analyzed in three different concentrations (45 min each, cumulative, 1-1009 free therapeutic plasma concentration [FTPC]). For drug details including solvents see Supplement Table 1. Prior to each measurement, 50 nM epinephrine (Sigma E4643) was added to each well to simulate ''physiological'' conditions and enhance the likelihood of contractile activity within the 60 s recording time. Drugs were added under sterile conditions. After 45 min incubation in a standard incubator, EHTs were transferred to the video optical system (Fig. 1). Due to a sequential mode of measurement (total time *30 min), incubation time varied from the first to the last EHT (45-75 min).
Measurements under perfusion and electrical stimulation, Ca 2? -transients, action potentials Intracellular Ca 2? -transients were analyzed in parallel with force under electrical stimulation and continuous perfusion using a novel setup as described previously [35]. The setup consisted of an inverted microscope, a temperature-and O 2 /CO 2 -controlled chamber for the 24-well EHT plate, a flow rate-controlled perfusion system, platinum-iridium wire electrodes for field stimulation, a fluorescence light source (IonOptix Hyperswitch), a photomultiplier, video cameras and software (both IonOptix) for the evaluation of contractile activity (edge detection mode). Experiments were done at 4 ml flow/min (per well) and 2-4 Hz stimulation in modified Tyrode's solution.

Statistical analysis
Data were expressed as mean ± SEM. Statistical differences were analyzed using the one-way analysis of variance (ANOVA) followed by the Dunett's (all compared to baseline) or Tukey's (all compared to all) adjustment for post hoc multiple comparison, or by paired or unpaired Student's t test, as indicated in the legend of each figure. Results were considered statistically significant if a paired Student's t test revealed a p value of less than 0.05 and the deviation from baseline was at least 15 %. This limit was defined after initial series of experiments had shown that formally significant (t test), but not concentration-dependent effects of drugs often amounted to ±11 %. Further support for the 15 % threshold came from quantifying the mean ± SD of all baseline measurements (n = 221 independent EHTs), which amounted to 99.6 ± 11.4 % (SEM ± 0.77 %).

Mechanisms of twitch prolongation and irregular beating pattern
In a previous study with EHTs, we found that the experimental I Ks -inhibitor chromanol 293b and two drugs known a b e c d Fig. 1 Schematic illustration of the standard operation procedure for evaluating drug effects. EHTs were subjected to measurements at day 14-21. One day before measurement culture medium was changed. Before evaluation, EHTs were transferred to fresh, preincubated (37°C, 7 % CO 2 , 40 % O 2 ), serum-free DMEM supplemented with 10 mM HEPES for pH steadiness, and incubated for 45 min.
Epinephrine (50 nM) was added and contractile parameters were analyzed by the video optical system. Thereafter, EHTs were transferred to fresh preincubated DMEM, including HEPES plus the first concentration of a drug, and incubated for 45 min. Epinephrine was added, measurements were done and this circuit started again with the second concentration of the drug to inhibit I Kr and cause TdP in humans, quinidine and erythromycin, caused concentration-dependent prolongations of relaxation time T2 [16]. These observations suggested that T2 is a useful surrogate for drug-induced prolongations of repolarization and proarrhythmic effects. To test this hypothesis, we evaluated the effects of the I toinhibitor 4-aminopyridine (4AP; IC 50 on I to in adult rat ventricular myocytes: 980 lM [41]) on EHTs. 4AP prolonged T2 concentration dependently reaching significance at 3 mM (Fig. 2). At 30 mM, EHTs showed extremely prolonged relaxation (?502 %), comparable to what have been seen previously with chromanol (?710 %; [16]). Prolonged contractions can either be caused by altered intracellular Ca 2? transients or myofilament response to Ca 2? . To differentiate between these mechanisms, EHTs were subjected to sharp electrode measurements of action potentials and calcium transients (Fura-2; Fig. 3). Action potential characteristics under basal conditions and electrical stimulation were similar as described previously in this model [16]. 4AP increased action potential duration (APD 90 ) from a mean of 113 to 175 ms ( Fig. 3a, b). 4AP also prolonged intracellular Ca 2? transients, whereas the badrenergic agonist epinephrine shortened it, both in the absence and presence of 4AP ( Fig. 3c-f). Quinidine, a still prescribed drug for the treatment of atrial fibrillation, also increased APD 90 from 179 to 232 ms at 100 lM (Supplemental Fig. 5d). These data supported the interpretation that alterations in T2 reflect similar changes in APD and the kinetics of intracellular Ca 2? transients. Accordingly, a high concentration of caffeine, known to open ryanodine receptors (RyR2) in the sarcoplasmic reticulum (SR), reduced beating rate and induced a large and wide twitch (Supplemental Fig. 1).
4AP not only prolonged T2, but also induced beat-to-beat irregularities, variations of twitch amplitude and aftercontractions falling into the relaxation phase of prolonged twitches (Fig. 4). To investigate the role of different cellular effector systems in the T2-prolonging effect of repolarization-prolonging compounds, we determined the effect of 4AP in the absence and presence of tetrodotoxin (I Na ), tetracaine (I Na ), verapamil (I Ca ), thapsigargin (SERCA), SEA0400 (sodium calcium exchanger, NCX) or JTV519 For APD 90 measurements two independent groups of EHTs (n = 3, each) were directly perfused with vehicle control (VC) or 4-aminopyridine (4AP), electrically stimulated and contractions recorded. Preparations were stimulated for at least 1 h before drug exposure and data acquisition. For Fura-2 F340/380 ratio and T2 the two independent groups (n = 8, each) were preincubated with VC or 4AP. Thereafter, EHTs were perfused and electrically stimulated (2 Hz; baseline) for 10 min, before epinephrine (50 nM) was added to the perfusion system.

Role of I Kr and I Ks in T2 prolongations
The role of I Kr and I Ks for action potential repolarization in rat heart is still poorly understood [32]. We studied the involvement of these two currents in rat EHTs by applying the reference I Kr blocker E-4031 (IC 50 7.7 nM [43]) and the selective I Ks blocker HMR-1556 (IC 50 10.5 nM [36]) alone or in combination (1-1,000 nM, Fig. 5). Neither E-4031 nor HMR-1556 affected T2 even at high concentrations (1,000 nM). The combined application also did not affect T2 at up to 100 nM, but caused a substantial increase at 1,000 nM ([8-fold). Given the high selectivity of E-4031 and HMR-1556 for I Kr and I Ks (HMR-1556 IC 50 on I to : 33.9 lM; on I Ca : 27.5 lM; on I Kr : 12.6 lM [36]), respectively, the effect of the combination suggests a role of these two currents for determining the repolarization reserve in rat EHTs.

Screening of proarrhythmic compounds under spontaneous beating
To test the usefulness of our screening system for the detection of proarrhythmic compounds, we analyzed a large panel of drugs associated with arrhythmias. The selection was made according to a list of drugs published by Redfern and colleagues [31] which related various levels of proarrhythmic risk with inhibition of I Kr (hERG). We tested the effect of the 46 compounds of this list which were commercially available (details Supplemental Table 1) at 1-, 10-, and 100-fold FTPC, n = 4-8 each. In addition, we tested moxifloxacin, an important inhibitor of bacterial gyrases, associated with prolongation of the QTinterval, but not with arrhythmias [28]. The selection of compounds encompassed 8 clinically used antiarrhythmic drugs (Group I), 5 drugs withdrawn from the market for TdP (Group II), 7 drugs with measurable incidence of TdP in humans (Group III), 13 drugs with isolated reports of TdP (Group IV) and 14 drugs devoid of TdP reports and therefore considered safe (Group V; Figs. 6, 7). Under the experimental conditions (which included a *EC 50 [6] epinephrine concentration, 50 nM) and in the absence of interventional drugs, EHTs showed a typical and reproducible beating pattern, consisting of periods with high frequency Note that both inhibitors prevented the marked 4APinduced T2 prolongation and after contractions, but did not completely normalize T2 or beating frequency (SEM) T1 and T2 of 67 ± 8 (±1) and 92 ± 18 (±2) ms, respectively. Given that the bursts occurred by chance either fully inside the 60 s recording window or only partially, the total number of beats per 60 s was relatively meaningless. In contrast, the frequency of beating in the burst was stable and systematically affected by epinephrine (?15-20 %) and carbachol (reversed epinephrine effect and induced partial stop; Supplemental Fig. 3).
In aggregate, the rat EHT responses categorized the 47 drugs in 4 groups (Fig. 8). Group 1 caused irregular beating as the main effect (n = 5), group 2 induced a concentration-dependent increase in T2 (n = 18). T2effective concentrations were 5 fold (domperidone, disopyramide) to 100 fold FTPC (e.g., ebastine, diltiazem, sotalol, or moxifloxacin). Group 3 (n = 7) had variable effects on EHT contraction including prolongation of contraction time T1 (haloperidol, sertindole, diphenhydramine, and mefloquine), shortening of T2 (terfenadine) and a decrease in force (verapamil and astemizole). was already seen at 300 nM, approximately 4-fold FTPC. Group 4 encompassed 17 compounds without any effect on the parameters studied. It contained specific I Kr -blockers such as ibutilide and dofetilide, which was in contrast to the results obtained with human EHTs [33].

Effects of drugs under electrical pacing
Drugs which caused prolongations of T2 also reduced beating rate in the bursts in many cases (Supplemental Fig. 4), raising the question which effect is the cause and which the consequence. We therefore measured EHT contractility under continuous electrical stimulation and perfusion and carefully evaluated the time course of the effects of quinidine (100 lM) and erythromycin (1 mM; Supplemental Fig. 5). Quinidine first induced a marked, time-dependent prolongation of T2 and a reduction in force, which was then followed by the development of a slower rhythm independent of pacing. Erythromycin also first prolonged T2, reduced force and then induced an autonomous chaotic contraction pattern. Effects of both drugs were fully reversible after washout. The data suggest that the drugs exert primary effects on processes underlying relaxation and that slowing of rate is a consequence.

Screening of new chemical entities (NCEs) under spontaneous beating
To get a rough estimate of the frequency of rat EHT effects in a non-selected group of drugs a chemical library was analyzed. NCEs were purchased from Maybridge and randomly chosen by respecting the rule of five [26]. 28 NCEs were tested at 0.1, 1, and 10 lM. Most NCEs had no effect on EHT contractility. Three NCEs prolonged T2 and one reduced force. In two additional cases EHTs stopped beating and didn't react to electrical pacing (Supplemental Table 2). Overall, an effect on EHT contractility was observed in 21 % of the investigated NCEs.

Discussion
Assessing the proarrhythmic risks of drugs remains a challenge in preclinical drug development. Current experimental models in preclinical toxicology determine the effect of NCEs on ion channels commonly involved in arrhythmogenic drug effects, particularly hERG channels [39], on electrophysiological surrogates of arrhythmias such as triangulation in Langendorff-perfused rabbit hearts [23,24], on action potential duration in rabbit Purkinje fibers [15] or the QTc intervall in telemetrically surveyed dogs [13]. Newer models with higher throughput and/or a human cardiomyocyte context include measurements of heart rate in zebrafish [27], of Ca 2? transients in isolated guinea-pig [30] or human pluripotent stem cell (hPSC)derived cardiomyocytes [5] or of electric field potentials in hPSC-cardiomyocytes [2]. The present study in rat EHTs is, to the best of our knowledge, the largest head-to-head comparison of proarrhythmic drugs performed so far. It showed that time of relaxation (T2) in rat EHTs is essentially insensitive to selective blockers of I Kr or I Ks , but prolonged by inhibition of I to , combined full inhibition of I Kr and I Ks or combined inhibition of I Ks and I to . This suggests that rat EHTs monitor mainly I to and I Ks effects of drugs. The I to -blocker 4AP also prolonged action potential duration and Ca 2? transients at T2-effective concentrations, indicating that, in this experimental model, T2 prolongation follow similar prolongations of repolarization and Ca 2? transients. The data suggest that, for drug screening purposes, T2 is a useful surrogate for time of repolarization in rat EHTs. The high percentage of known proarrhythmic drugs that induced concentration-dependent T2 prolongations, aftercontractions and/or irregular beating (group 1 and 2, 23/38 = 61 %) indicates that many clinically used drugs possess effects on I to and/or I Ks in addition to their well-characterized inhibitory action of I Kr (hERG).

Characterization of rat EHTs as a test system for testing proarrhythmic drugs
The underlying hypothesis of this study was that repolarization-inhibiting drugs prolong action potential duration a b Fig. 7   double asterisk other types of arrhythmias in humans known. The type of depiction and most information independent of EHTs were adopted from Redfern et al. (2003). Case reports (indicated by asterisk), estimated plasma concentration and IC 50 values on I to /I Ks were obtained by a literature search and Ca 2? influx and thereby time of relaxation, a parameter easily assessable in EHT. EHTs are engineered threedimensional cardiac tissue constructs in which cardiac cells are embedded in a fibrin matrix and, over 10-14 days, form a highly organized longitudinally oriented network [16]. The 24-well mini-EHTs system is designed for a robust, automated measurement of spontaneous or electrically stimulated contractile activity at a medium throughput scale. It is stable for weeks and has the advantage that measurements can be done at near-physiological conditions: 37°C, steady state beating (not dying preparations such as isolated myocytes, papillary muscles, Purkinje fibers, and Langendorff hearts), auxotonically work-performing heart muscle constructs, and absence of invasive experimental interventions (e.g., patch clamp or microelectrodes). On the other hand, it is not well suited for measuring effects of drugs on action potentials, field potentials, or ion channels directly. We therefore performed several experiments to evaluate whether the relaxation time T2 is indeed a suitable surrogate of time of repolarization.
The following observations favor this assumption. (1) T2 is easily determined and highly reproducible (11 % SD, 0.77 % SEM, n = 221). (2) The duration of contraction and relaxation corresponds well with that of action potentials in rat EHTs and both were prolonged in the presence of 4AP as well as quinidine. Action potential duration (APD 90 ) measured with sharp microelectrodes at 36°C amounted to a mean of 113 ms, time to 90 % Ca 2? return to a mean of 124 ms at 37°C, and T1 ? T2 (at 80 % relaxation) amounted to a mean of 159 ms at 37°C (this study and [16]). Thus, it is reasonable to assume that prolongations of repolarization affect T2. (3) Drugs with well-characterized actions on I to [41], I Kr [43], and I Ks [36] had reproducible, concentration-dependent and reversible effects on T2. 4AP started to prolong T2 at 3 mM, which corresponds to 3-fold IC 50 in adult rat cardiac myocytes [41] and confirms the prominent role of I to for rodent heart repolarization [19]. The lack of effect of E-4031, dofetilide, ibutilide, and HMR-1556 even at 100-fold IC 50 concentrations confirms the minor role of I Kr and I Ks for repolarization in rat cardiac myocytes [29]. Interestingly, however, the combined application of high concentrations of E-4031 and HMR-1556 markedly prolonged T2. Given the high selectivity of both drugs for I Kr and I Ks [36,41], respectively, the data suggest that I Kr and I Ks can substitute for each other, but together play a role for repolarization in rat EHTs. The strong effect of HMR-1556 in the presence of 3 mM 4AP (and the lack of an E-4031 effect under this condition) suggests a greater role for I Ks than I Kr . The necessary channel subunits are expressed in rat EHTs (KCNH2a, KCNQ1a, and KCNE1/E2 mRNAs are present in rat EHTs [16]).
Whereas all this argues for repolarization as an important parameter of relaxation time T2 in our system, it is obvious that other mechanisms have to be considered and that T2 prolongations alone do not prove effects on repolarization. Any effect on beating rate will affect T2 because of the well-known frequency-dependent acceleration of relaxation. Such effect could be excluded in case of quinidine and erythromycin (Supplemental Fig. 5), but may play a role in other cases. Drugs that directly affect intracellular Ca 2? handling could alter T2 independently of repolarization. Indeed, caffeine at a high concentration (5 mM) which enhances the open probability of RyR2 reduced beating rate and caused widening of contraction twitches (Supplemental Fig. 1). Thapsigargin, a selective inhibitor of SERCA, did not affect time of relaxation as such (3-30 nM), but aggravated the relaxation-slowing effect of 4AP (Supplemental Fig. 2). Another possibility to affect T2 independently of repolarization is an effect on myofilament Ca 2? sensitivity. Ca 2? sensitizers such as EMD 57033 [34] shift the force-pCa curve to the left and thereby prolong relaxation as recently shown in mouse EHTs [35]. The unexpected T2-prolongation under diltiazem may represent this mechanism. A study in isolated adult rat ventricular myocytes reported a mild myofilament Ca 2? sensitizing effect of diltiazem [9]. Such an effect may also explain the unexpected lack of significant negative inotropic effects of this compound at the highest tested concentration (10 lM) which was in contrast to verapamil. Conversely, epinephrine (Supplemental Fig. 3) and all other cAMP-dependent drugs increase the PKA-dependent phosphorylation of troponin I, myosin binding protein C and others and shift the curve to the right. This favors relaxation and abbreviates T2. Taken together, T2-prolongation of rat EHTs can be caused not only by blockers of repolarization (mainly I to , I Ks , very little contribution of I Kr ), but also rate-slowing drugs, Ca 2? sensitizers or blockers of cAMP-dependent pathways such as carbachol (Supplemental Fig. 3), which needs to be considered in drug screening efforts.
The pronounced prolongation of relaxation at high concentrations of 4AP and aftercontractions were sensitive to both SEA0400 and JTV519 (Fig. 4), but relaxation remained significantly prolonged in their presence (*200 %). This finding is interesting as it suggests that the primary effect of 4AP, the inhibition of repolarization, has time-dependent (see time course in Fig. 4) secondary effects that affect time of relaxation. Very likely, this secondary effect corresponds to increased filling of the SR with Ca 2? with the final consequence of RyR2-mediated spontaneous Ca 2? release and NCX-mediated Ca 2? extrusion. The latter transport is electrogenic and contributes to prolonged depolarization and after depolarizations, visible in our system as marked prolongations of T2 and aftercontractions. An interesting speculation is that group 1 (Fig. 8) drugs directly interfere with RyR2 and/or NCX or SR function and therefore induce irregular beating, whereas group 2 drugs primarily affect repolarization and therefore prolong T2 before causing extra beats. On the other side, group 1 and 2 effects cannot be firmly separated because three group 1 drugs also caused T2 prolongations. Interestingly, tetracaine (also flecainide (0.5 lM); data not shown) did not mimic the effect of JTV519 (Supplemental Fig. 2), although both have inhibitory effects on RyR2mediated Ca 2? release [18,40]. Possibly, their main effect on I Na which is not shared by JTV519 could explain the difference.

Drug-induced relaxation slowing and beating irregularities in rat EHTs
We selected drugs according to a list published by Redfern and colleagues [31] that grouped drugs according to their proarrhythmic potential. 15/33 drugs with various degree of risk for TdP (Class I-IV) showed T2 prolongation or beating irregularities in our system (group 1 and 2). The percentage did not clearly differ between those that were withdrawn from the market for TdP (1/5) and those with only isolated TdP reports (6/13). In addition, T2-prolongation was observed in 6/14 drugs considered safe. This clearly indicates that the rat EHTs test does not sufficiently discriminate between high and low risk drugs, very likely due to the lack of I Kr sensitivity.
Nevertheless, screening the large number of drugs revealed a number of interesting novel informations. (1) It is apparent that T2 prolongations and beating irregularities were seen mainly in those drugs which reach high FTPC in clinical use and were therefore tested at high absolute concentrations (compare concentration range of Group 1 and 2 drugs with group 3 and 4; Fig. 8). This confirms the general rule in pharmacology that drugs with low potency at their target have a higher chance of off-target toxicity than high affinity drugs. (2) Most effects were seen at 30-100-fold FTPC and higher. This is at the upper limit generally considered a critical safety margin [31] and point to a relatively low sensitivity of our assay. Rather than indicating simple non-specific effects (drugs such as sematilide and ciprofloxacin were tested at up to 300 lM) the data with E-4031, HMR-1556 and 4AP indicate that almost complete block of one or more currents is required to overcome safety mechanisms of repolarization in a relatively intact system such as the EHT. As such, the low sensitivity probably reflects the situation in vivo better than isolated cells [28]. Pharmacokinetic peculiarities and timedependent effects likely add to high concentration requirements. For example, amiodarone is very lipophilic and accumulates in cells over time. With an apparent volume of distribution of 20-200 l/kg, the low FTPC of amiodarone (0.3 nM) is orders of magnitude lower than cellular concentrations in the steady state. (3) Concentrations in which group 1 and 2 drugs exerted their effects on EHTs were generally well above their IC 50 for hERG (gray bars in Fig. 8). Notable exceptions are domperidone and phenytoin, where relaxation slowing occurred at concentrations below hERG IC 50 . Since E-4031 at high concentration markedly potentiated the effect of HMR-1556 without having an effect alone, the data suggest that inhibition of I Kr may participate in the effect of group 1 and 2 drugs, but is not sufficient to fully explain them. Inhibition of other currents must come into play. We found reports of I to -inhibiting activity for 14 drugs (blue bars in Fig. 8). In four of these (imipramine, propafenone, quinidine, and clarithromycin) published IC 50 values for hERG and I to were at or below the threshold for T2 prolongation or irregularity, providing a likely mechanism of action. Moreover, we found reports of I Ks -inhibiting effects for 20 drugs (green bars in Fig. 8). In six of these cases (bepridil, thioridazine, propafenone, quinidine, cibenzoline, and sotalol), the IC 50 was at or below the threshold for group 1 or 2 effects. It is interesting that quinidine at the T2threshold concentration (100 lM) inhibits all three currents I Kr , I to , and I Ks . Other currents involved in rat cardiac repolarization such as I K1 , I ss , and I Kx [19] or I KATP [4] and I KNa are also potential targets, but not much is known about effects of the drugs investigated. Taken together, the data indicate that many clinically used drugs, some of them associated with TdP, others not, inhibit cardiac repolarizing currents in addition to I Kr . This supports recent data suggesting that combined channel block underlies clinical proarrhythmia [28].

Miscellaneous effects on EHT function
Some drugs caused T1 prolongations (haloperidol, sertindole, diphenhydramine, and mefloquine), T2 acceleration (terfenadine) or negative inotropic effects (verapamil). Whereas the latter is the expected main effect of the drug, the mechanism of the other is unclear at present. T1 prolongations could indicate reduced conduction velocity in the EHT, a typical consequence of I Na inhibition. However, pure I Na blockers such as TTX or lidocaine did not prolong T1 (data not shown), arguing against this idea. An alternative may be an effect on gap junction conduction. At least mefloquine is known to inhibit numerous connexins (C9), including C 9 43 [8]. T2 shortening could be due to accelerated repolarization by stimulation of K-currents or accelerated myofilament relaxation, but also due to an inhibition of I Na . In fact, terfenadine, which also shortened action potential duration in the SCREENIT test system (10 lM [23]), has recently been suggested to cause arrhythmias not by a TdP mechanism, but by its strong I Nablocking activity [46].
Taken together, rat EHTs are not suitable as a general screening assay for proarrhythmic drug effects due to the small contribution of I Kr for rat EHT repolarization. On the other hand, the assay is a simple and robust system for the analysis of non-hERG-related drug effects on cardiac function. Relaxation time was shown to be a particularly suited screening parameter, sensitive to drugs affecting cardiac repolarization, but also Ca 2? handling or myofilament function. The high fraction of drugs with known arrhythmogenic effects that prolonged relaxation or induced irregular beating in this hERG-insensitive system suggest that I to and I Ks effects add to the proarrhythmic risk of drugs and require further consideration.