Correlation between human ether‐a‐go‐go‐related gene channel inhibition and action potential prolongation

Background and Purpose Human ether‐a‐go‐go‐related gene (hERG; Kv11.1) channel inhibition is a widely accepted predictor of cardiac arrhythmia. hERG channel inhibition alone is often insufficient to predict pro‐arrhythmic drug effects. This study used a library of dofetilide derivatives to investigate the relationship between standard measures of hERG current block in an expression system and changes in action potential duration (APD) in human‐induced pluripotent stem cell‐derived cardiomyocytes (hiPSC‐CMs). The interference from accompanying block of Cav1.2 and Nav1.5 channels was investigated along with an in silico AP model. Experimental Approach Drug‐induced changes in APD were assessed in hiPSC‐CMs using voltage‐sensitive dyes. The IC50 values for dofetilide and 13 derivatives on hERG current were estimated in an HEK293 expression system. The relative potency of each drug on APD was estimated by calculating the dose (D150) required to prolong the APD at 90% (APD90) repolarization by 50%. Key Results The D150 in hiPSC‐CMs was linearly correlated with IC50 of hERG current. In silico simulations supported this finding. Three derivatives inhibited hERG without prolonging APD, and these compounds also inhibited Cav1.2 and/or Nav1.5 in a channel state‐dependent manner. Adding Cav1.2 and Nav1.2 block to the in silico model recapitulated the direction but not the extent of the APD change. Conclusions and Implications Potency of hERG current inhibition correlates linearly with an index of APD in hiPSC‐CMs. The compounds that do not correlate have additional effects including concomitant block of Cav1.2 and/or Nav1.5 channels. In silico simulations of hiPSC‐CMs APs confirm the principle of the multiple ion channel effects.


Introduction
The current paradigm of assessing drug-induced pro-arrhythmic risk is based on a link between drug-induced human ether-a-gogo-related gene (hERG also known as K v 11.1) channel blockade and QT-interval prolongation; for review, see Sanguinetti and Tristani-Firouzi (2006). The abnormal activity of cardiac myocytes such as early after-depolarizations (EADs) is more likely to occur when the cardiac action potential (AP) is prolonged. EADs manifest as a single spike or oscillations of the membrane potential at the repolarising phase of the AP (Keating and Sanguinetti, 2001;Morita et al., 2008;Liu et al., 2012) and are commonly seen in patients with an acquired long-QT syndrome (Veldkamp et al., 2001;Pogwizd and Bers, 2004). EADs are proarrhythmic because of their potential to induce dispersed refractory periods in cardiac tissue, which is a vital condition for the precipitation of arrhythmias. A link between EADs and torsade de pointes (TdP) has been previously studied (Volders et al., 2000), and it is widely accepted that the prolongation of the QT interval is the precursor of EADs and TdP caused by many drugs (Hancox et al., 2008;Sager et al., 2014).
Many experimental and theoretical studies have been performed to investigate the ionic mechanisms of EADs in isolated cardiomyocytes (Zeng and Rudy, 1995;Sato et al., 2010;Liu et al., 2012). The repolarization phase of cardiac AP results from a complex interplay between several ionic currents such as inward sodium current (I Na ), inward calcium current (I CaL ) and several potassium currents mainly rapid delayed rectifier potassium current (I kr ). EADs can be produced either by increasing the inward currents, mainly L-type calcium current (I CaL ), or reducing the outward currents (I Kr ), or both. So, for example, a cell can be made susceptible to EADs by inhibiting I Kr through hERG with dofetilide, activating the late sodium current (I NaLate ) with veratridine or by increasing the conductance of I CaL through Ca v 1.2 channels with BAY K8644 (Horváth et al., 2015). Drugs with unidirectional inhibition of inward and outward currents are generally unable to prolong AP duration (APD) and thus unlikely to induce EADs. Verapamil is one example that simultaneously inhibit I Cal and hERG current without prolonging the QT interval (Zhang et al., 1999). Kramer J et al. (2013) have found that prediction of pro-arrhythmogenity may be improved by considering the effect of drugs on currents from three key ion channels: hERG potassium channels (K v 11.1), sodium channels (Na v 1.5) and calcium channels (Ca v 1.2). The development of multiple ion channel effect models leads to a significant reduction in false-positive and false-negative predictions when compared with hERG assays alone. Recently, the Cardiac Safety Research Consortium and the Food and Drug Administration proposed a new cardiac safety paradigm labelled as 'comprehensive in vitro pro-arrhythmia assay' (CiPA). The new CiPA guidelines advocate studying the pharmacological effects of drugs on multiple ion channels that play an important role in shaping the ventricular AP (hERG, Na v 1.5, Ca v 1.2) instead of only hERG screening, and confirmation of electrophysiological effects using myocyte assays such as human-induced pluripotent stem cell-derived cardiomyocytes (hiPSC-CMs).
Previous studies of pro-arrhythmic effects of hERG inhibitors used a variety of chemical classes with different potencies and different selectivity. In this study, minor changes in the chemical structure of the highly potent and selective hERG inhibitor dofetilide generate compounds with a wide range of IC 50 values. A remarkable linear relationship was observed between the IC 50 value and the degree of AP duration change observed in hiPSC-CM a relationship confirmed using an in silico model. The few derivatives not adhering to this linear relationship showed significant effects on Na v 1.5 and Ca v 1.2 ion channels.

Group sizes
Numbers (n) for all experiments are provided and refer to independent single measurements. Data subjected to statistical analysis have n of at least five per group. In the case of the APD, experiments on hiPSC-CMs have a minimum of n = 4 in some cases. The n = 4 can discriminate a 15% change in APD 90 (APD at 90% repolarization) with α = 0.95 and β = 0.2, from power calculations. The variability in APD values on a well-to-well basis (in 96-well plate) was measured and can be expressed in terms of a coefficient of variation for CDI cells [commercially available from Cellular Dynamics International (CDI), Madison, WI, USA] after rate correction (1 Hz) is 0.08.

Randomization
Randomization was not applicable, hence not performed.

Blinding
Blinding of experiments is not applicable.
Optical measurement of transmembrane potential signals using voltage-sensitive dyes Two hours before the experiments, the cells were transiently loaded with the voltage-sensitive dye (VSD) di-4-ANEPPS (6 μM, 1 min at room temperature) in serum-free media (DMEM, Gibco 11 966, supplemented with galactose 10 mM and sodium pyruvate 1 mM). Afterwards, the medium containing VSD was replaced by fresh serum-free medium, and the cells were returned to the incubator. The multi-well plate was placed in an environmentally controlled stage incubator (37°C, 5% CO2, water-saturated air atmosphere) (Okolab Inc, Burlingame, CA, USA) of the CelIOPTIQ platform (Clyde Biosciences Ltd, Glasgow, Scotland). The di-4-ANEPPS fluorescence signal was recorded from a 0.2 × 0.2 mm area using a 40× (NA 0.6) objective lens. Excitation wavelength was 470 ± 10 nm using a light-emitting diode (LED), and emitted light was collected by two photomultipliers (PMTs) at 510-560 and 590-650 nm respectively. LED, PMT, associated power supplies and amplifiers were supplied by Cairn Research Ltd (Kent, UK). The two channels of fluorescence signals were digitized at 10 kHz, and the ratio of florescence (short wavelength/long wavelength) was used to assess the time course of the transmembrane potential independent of cell movement (Knisley et al., 2001). Baseline spontaneous electrical activity was recorded by capturing a 20 s segment of fluorescent signal prior to compound (drug) addition. Acute effects of dofetilide and derivatives were assessed by exposure to increasing drug concentration with matched vehicle controls for each concentration. A 20 s recording was then taken 30 min after exposure to the drug or vehicle with only one concentration applied per well. The records were subsequently analysed offline using proprietary software (CellOPTIQ). The procedure was repeated from four to five times, and parallel matched control (vehicle) measurements were taken on cardiomyocytes with equivalent concentrations of vehicle (DMSO). AP parameters were measured, including APD at 50, 75 and 90% repolarization (APD 50 , APD 75 and APD 90 respectively). Data are given as % change from control for the treated groups (vehicle, control and drug). This allowed a single comparison to be made at each concentration, and every experiment was performed with its own set of controls (vehicle). No data were used more than once.
Patch-clamp studies on hERG, Na v 1.5 and Ca v 1.2 channels Currents through hERG channels (Anaxon GmbH) and Na v 1.5 channels stably expressed in HEK293 cells were studied within 8 h of harvest in the whole-cell configuration of the planar patch clamp technique (NPC-16 Patchliner, Nanion Technologies GmbH, Munich, Germany), using an EPC 10 patch-clamp amplifier (HEKA Elektronik Dr. Schulze GmbH, Lambrecht/Pfalz, Germany) (Milligan et al., 2009). Currents were low-pass filtered at 10 kHz using the internal Bessel filter and sampled at 25 kHz. The extracellular solution for hERG current recording contained 140 mM NaCl, 4 mM KCl, 2 mM CaCl 2 , 1 mM MgCl 2 ,5 mM D-glucose and 10 mM HEPES (pH 7.4) (Sigma-Aldrich). The intracellular solution for hERG current recording contained 50 mM KCl, 10 mM NaCl, 60 mM KF, 20 mM EGTA and 10 mM HEPES (pH 7.2). The extracellular solution for measuring sodium currents in HEK cells stably expressing the human clone of Na v 1.5 (GenBank M77235) contained 4 mM KCl, 20 mM NaCl, 1.8 mM CaCl 2 , 0.75 mM MgCl 2 , 5 mM HEPES, 120 mM choline chloride and pH 7.4 using NaOH. The intracellular solution for sodium current recording contained 120 mM CsF, 20 mM CsCl, 5 mM EGTA, 5 mM HEPES and pH 7.4 using CsOH. All chemicals were obtained from Sigma-Aldrich Chemie GmbH (Taufkirchen, Germany). The compound solutions were applied by means of the automated NPC-16 Patchliner planar patch-clamp platform. Data acquisition was done using the PatchMaster software version 2.65 (HEKA Elektronik Dr. Schulze GmbH).
For barium current (I Ba ) measurements through voltagegated Ca 2+ channels, HEK tsA-201 cells were co-transfected with cDNAs encoding the rabbit Ca V 1.2 α 1 -subunit (GenBank X15539) with auxiliary β 2a (Perez-Reyes et al., 1992) as well as α 2 -δ 1 (Ellis et al., 1988) subunits and GFP to identify transfected cells (see Beyl et al., 2012, for details). The transfection of tsA-201 cells was performed using the FUGENE6 Transfection Reagent (Roche Diagonstics GmbH, Mannheim, Germany) following standard protocols. The tsA-201 cells were used until passage number 15. No variation in channel gating related to different cell passage numbers was observed. I Ba were studied by manual patch-clamping (Hamill et al., 1981) using an Axopatch 200A patch clamp amplifier (Axon Instruments, Foster City) 36-48 h after transfection. The extracellular bath solution (in mM: BaCl 2 20, MgCl 2 1, HEPES 10, choline-Cl 90) was titrated to pH 7.4 with methanesulfonic acid. Patch pipettes with resistances of 1 to 4 MΩ were made from borosilicate glass (Clark Electromedical Instruments, UK) and filled with pipette solution (in mM: CsCl 145, MgCl 2 3, HEPES 10, EGTA 10), titrated to pH 7.25 with CsOH. The drugs were applied to cells under voltage clamp using a microminifold perfusion system. Ca 2+ channel block was estimated as peak I Ba inhibition during a train of short (50 ms) test pulses from À80 mV at a frequency of 0.2 Hz. Patch clamp experiments to study hERG, Na v 1.5 and Ca v 1.2 currents were performed at room temperature (22-25°C). All data were digitized and saved to disc. Current traces were filtered at 5 kHz and sampled at 10 kHz. The pClamp software package (Version 7.0 Axon Instruments, Inc.) was used for data acquisition and preliminary analysis. Microcal Origin 7.0 was used for analysis, and sigmoidal curves were fitted using the Hill equation.

In silico studies of hiPSC-CMs' action potentials
The cellular AP model of Paci et al. (2012) for ventricular hiPSC-CMs was used for comparative computational studies of APD 90 prolongation caused by dofetilide and its derivatives. These effects were described by the common pore block model in which the currents through the channels potentially sensitive to drugs were calculated with a coefficient equal to a fraction of channels free of drug: All computations were performed in MATLAB R2015b. AP simulations were performed for a temperature of 310 K (37°C).

Data processing and normalization
Origin 7.0 (Origin Lab Corp., Northampton, MA, USA) was employed for data analysis and curve fitting. The cumulative concentration-inhibition curves were fitted using the Hill equation: where IC 50 is the concentration at which hERG inhibition is half-maximal; C is the applied drug concentration; A is the fraction of hERG current that is not blocked; and n H is the Hill coefficient (Windisch et al., 2011). Data are presented as mean ± SEM for at least five cells from two different batches or for three independent measurements with HEK293 cells.

Statistical comparison
Statistically significant differences were calculated using Student's t-tests and one-way ANOVA and data from independent recordings. Only P-values <0.05 were accepted as statistically significant. Linear correlation was used to confirm a linear relationship between hERG IC 50 and APD data. The data and statistical analysis comply with the recommendations on experimental design and analysis in pharmacology (Curtis et al., 2015).

Drugs
Dofetilide was obtained from Sigma, and its derivatives were prepared as previously described (Shagufta et al., 2009). All derivatives were dissolved in DMSO to prepare a 10 mM stock and stored at À20°C. Drug stocks were diluted to the required concentration in extracellular solution on the day of each experiment. The maximal DMSO concentration in the bath (1%) did not affect Ca v 1.2 or Na v 1.5 currents in any of the preparations. (Supporting Information Fig. S1).

Nomenclature of targets and ligands
Key protein targets and ligands in this article are hyperlinked to corresponding entries in http://www. guidetopharmacology.org, the common portal for data from the IUPHAR/BPS Guide to PHARMACOLOGY (Southan et al., 2016), and are permanently archived in the Concise Guide to PHARMACOLOGY 2015/16 .

Dofetilide derivatives library
The small library of derivatives used in this study was previously described by Shagufta et al. (2009). The chemical structures of dofetilide and its 13 derivatives are shown in Figure 1. The structural modifications conserved the phenyl rings on both sides of the molecules and comprised the following: (i) attaching different substituents to the rings (all excluding Dofe30); (ii) changing the substituents on the protonated nitrogen (Dofe54, Dofe60); and (iii) varying chain length (Dofe78, Dofe81).

Figure 1
Chemical structures of dofetilide and its derivatives.
BJP P Saxena et al.

Drug-induced prolongation of APs in hiPSC-CMs
Effects of different concentrations of dofetilide and 13 derivatives on AP parameters were studied in hiPSC-CMs.
The changes in APD (as % of control) are given in Table 1. Figure 2 shows representative effects of dofetilide and two of its derivatives on spontaneous APs in cardiomyocytes.
The derivative Dofe54 represents a potent pro-arrhythmic compound and Dofe33 is an example with weak (if any) pro-arrhythmic activity. Dofetilide-induced concentrationdependent lengthening of the AP was accompanied by incidence of EADs at concentrations of 10, 30 and 100 nM. The highest concentration used (100 nM) dramatically increased the spontaneous rate of myocyte contraction ( Figure 2A). The potent derivative Dofe54 produced a slightly different pattern of AP distortion: smaller amplitude of oscillation during EADs and prolongation of APs at relatively low concentrations was observed. The 10 nM concentration induced approximately 700% prolongation of the AP ( Figure 2B). The Dofe33 exhibited a negligible effect on APD 90 prolongation until 100 nM. At a concentration of 300 nM, the APD 90 was increased by approximately 170% of control. The maximal AP prolongation of 250% was observed at micromolar concentrations ( Figure 2C). The concentration dependence of APD 90 (in % to control) for dofetilide, Dofe54 and Dofe33 are shown in Figure 2D-F. The sigmoidal curves ( Figure 2D, E) were fitted to the Hill equation.
Derivatives Dofe54, Dofe81, Dofe35, Dofe60 and Dofe78 had the most potent effects on APD (maximal level up to approximately 1000%), with incidence of EADs at the higher concentrations. In contrast, derivatives Dofe30, Dofe31, Dofe33, Dofe43, Dofe41 and Dofe45 exhibited relatively less effect on the APD 90 without (if any) incidence of EADs. Derivatives Dofe42 ( Figure 3A, D) and Dofe44 ( Figure 3B, E) did not affect APD 90 even at 1 μM while Dofe45 ( Figure 3C, F) only slightly prolonged the AP. hERG channel inhibition by dofetilide and its derivatives hERG channel inhibition by dofetilide and derivatives was studied in HEK293 cell lines stably expressing hERG channels using an automated planar patch system (see Methods). After application of a given drug concentration, 0.3 Hz pulse trains were applied until a steady-state of hERG current inhibition occurred. hERG current inhibition by Dofe54 is illustrated in Figure 4A. The concentration-inhibition relationships were analysed by plotting the normalized values of peak tail current versus peak tail steady current in the presence of the respective cumulatively applied compound concentrations ( Figure 4B, C). Data points were fitted using Hill equation. Figure 4 illustrates that dofetilide derivatives can be subdivided into the following: (i) high affinity derivatives hERG current inhibition with IC 50 values ranging between 3 and 40 nM ( Figure 4B); and (ii) low affinity derivatives with an IC 50 s of >100 nM ( Figure 4C). The concentration-inhibition curves of group 1 derivatives were close to the dofetilide curve while curves of group 2 derivatives indicated reduced potency (approximately 10-fold) of channel inhibition.

Prolongation of AP correlates with potency to block hERG
The potency of dofetilide derivatives to prolong AP was related to their apparent affinity for hERG potassium channels. The drugs inhibiting hERG at lower concentrations prolonged the AP and induced EADs at lower concentrations (Table 2). In a first attempt, we failed, however, to observe a quantitative correlation between IC 50 of hERG inhibition Table 1 Changes in APD 90 in hiPSC-CMs after application of dofetilide and derivatives hERG inhibition and AP prolongation BJP and the concentration that increased APD (D 150 ) by 50%. Dofe42 and Dofe44 did not induce prolongation of AP and Dofe45 slightly prolonged the AP at high concentrations. A plot of D 150 versus drug affinities (IC 50 ) is shown in Figure 5 (see also Table 2). Data points corresponding to derivatives that were not efficient at prolonging the AP (Dofe42, Dore44 and Dofe45; Figure 3) are illustrated as red circles in Figure 5. Excluding these data points from analysis led to a strong correlation (r = 0.94, P < 0.05) ( Figure 5, black line) while taking them into account made the correlation non-significant. The predicted relationship between IC 50 and D 150 by mathematical AP model is

Figure 2
Effects of dofetilide and its derivatives Dofe54 and Dofe33 on AP characteristics in hiPSC-CMs. Representative AP recordings of hiPSC cardiomyocytes after incubating with dofetilide, n = 4-5 (A), the high affinity derivative Dofe54, n = 4 (B) and the low affinity derivative Dofe33, n = 4 (C) and plots of APD 90 as % of control versus concentrations of dofetilide, n = 4-5 (D), Dofe54, n = 4 (E) and Dofe33, n = 4 (F). The data points represent the mean ± SEM (see Table 1) and were fitted by a Hill equation for dofetilide and Dofe54. The data points for Dofe33 were connected by lines.   Table 1).

Table 2
Dofetilide and its derivatives: affinity for hERG potassium channels and concentration (D 150   Correlation between D 150 (concentration that prolongs AP in hiPSC-CM by 50%) and IC 50 (half-maximal concentration inhibiting hERG channels in HEK293 cells). A significant linear correlation (r = 0.94, P < 0.05) was observed for 12 data points (black circles) including dofetilide and 11 derivatives. Derivatives Dofe45, Dofe44 and Dofe42 (red circles) were not included in the correlation analysis. Dofe45 prolonged the AP in hiPSC-CM by 50% only at 538 nM and Dofe42 and 44 at >600 nM. The red line represents a prediction of the mathematical simulation of the hiPSC-CM's AP (see Figure 8).
hERG inhibition and AP prolongation BJP indicated by the red line. This model will be discussed in more detail later. In order to examine the possibility that additional block of inward currents may have counterbalanced hERG inhibition, we investigated effects of Dofe42, Dofe44 and Dofe45 on calcium (Ca v 1.2) and sodium (Na v 1.5) channels.

Inhibition of Ca v 1.2 by dofetilide and derivatives
Dofetilide itself does not inhibit Ca v 1.2 even at a high concentration of 100 μM (Supporting Information Fig. S1a). Figure 6A illustrates the effects of Dofe42, Dofe44 and Dofe45 on Ca v 1.2 at the indicated concentrations, and Figure 6B shows the corresponding concentration-inhibition curves obtained during continuous pulsing at 0.2 Hz. Dofe45 was identified as a potent Ca v 1.2 blocker (IC 50 = 190 ± 3 nM, Figure 6B, right panel) while Dofe42 and Dofe44 inhibited Ca v 1.2 with comparably low potencies [IC 50 of 38 ± 9.3 μM (Dofe42) and >100 μM (Dofe44)]. Use-dependent channel inhibition was studied during trains of 1 Hz and 50 ms test pulses (from À80 to +10 mV). After the application of 20 test pulses in control (absence of drug), the cells were incubated for 3 min with drug at rest. Peak current inhibition during the first pulse (1st, Figure 6C) in the presence of the drug reflects 'resting state' block. Additional current inhibition  during a subsequently applied pulse train illustrates usedependent block. Peak current inhibition in control and drug are compared in Figure 6D. Dofe42 and Dofe45 induced pronounced resting state block and additional use-dependent block (compare last currents of the train in control and in drug). No use-dependent block was observed for Dofe44 ( Figure 6C, D middle panel).

Inhibition of Na v 1.5 by dofetilide derivatives
Dofetilide (100 μM) did not inhibit Na v 1.5 (Supporting Information Fig. S1b) while block was observed for Dofe42, Dofe44 and Dofe45. Figure 7A shows representative current traces illustrating the inhibition of Na v 1.5 by derivatives at indicated concentrations. The concentration-inhibition curves for all three derivatives were first estimated at a holding potential of À140 mV where all Na v 1.5 are available ( Figure 7B, Wang et al., 2015). Dofe44 inhibited cardiac sodium channels with an IC 50 of 23.3 ± 1.9 μM (n = 6) compared with statistically less potent Dofe45 (IC 50 s of 69.7 ± 1.0 μM, n = 6, P < 0.05) and Dofe42 (77.9 ± 9.7 μM, n = 6, P < 0.05). However, the reported resting potentials of iPSC-CM range between À75 and À63 mV (Hoekstra et al., 2012) would induce substantial inactivation. In order to evaluate block of inactivated Na v 1.5, we investigated I Na inhibition at a holding potential of À80 mV where more than 60% of Na v 1.5 were in an inactivated state (Wang et al., 2015). Interestingly, the concentration-response curves where significantly shifted towards lower drug concentrations ( Figure 7B, Dofe42: 5.6-fold, Dofe44: fivefold and Dofe45: 7.7-fold), suggesting that inactivated Na v 1.5 are blocked with higher affinity. The application of test pulses at a higher frequency (1 Hz) did not induce additional channel inhibition.

Computational studies support experimental findings
The in silico AP model (Paci et al. 2012) for ventricular hiPSC-CM was run with a pacing of 1 Hz until limit cycling was achieved in order to determine control APD 90 . In the first series of calculations, we have described a prolongation of AP under inhibition of hERG potassium channels. The drug dose D was set as a multiple of IC 50 , that is, D = x × IC 50 , enabling the use of the nonlinear forward mapping F: x → APD 90 (x).
The factor x corresponding to a prolongation of the control APD 90 by 50% was determined by solving (Engl et al., 2009) the nonlinear inverse problem F(x) = 1.5 × (control APD 90 ). The predicted relationship between IC 50 and D 150 is shown in Figure 5 (red line). Figure 8A displays AP simulations at different levels of inhibition of the hERG channel. The APD exhibited a linear correlation with logarithm of concentration of hERG channel blocker ( Figure 8B). In order to test whether inhibition of inward currents compensates for the APD changes seen with hERG channel block, we simulated APs for different concentrations of half-maximal Ca v 1.2 and Na v 1.5 inhibition (for Ca v 1.2, IC 50 = 200 nM and Na v 1.5, IC 50 = 10 μM). Both IC 50 s are characteristic for Dofe 45 (Table 2, Figures 6 and 7). The simulation ( Figure 8C) surprisingly coincides with experimental records ( Figure 3C). The 'selective inhibition' of the hERG channels by Dofe45 would induce a substantial prolongation of the AP. Figure 8D illustrates the sensitivity

Discussion and conclusion
Potential pro-arrhythmic effects in early stages of drug development have often been assessed solely by examining hERG channel block. The principle role of hERG channel block for AP repolarization and its consequences have been extensively discussed (Sanguinetti and Tristani-Firouzi, 2006). The new CiPA paradigm proposes that drugs should be tested by screening multiple ion channels including I Kr , I Ks and I K1 as well as I NaLate and I CaL and predicting their effect on the human APD using in silico models to integrate the effects of a number of ion channels (Sager et al., 2014). The CiPA scheme also suggests analysing pro-arrhythmic effects using human iPSC-derived cardiac muscle as a surrogate for human myocardium. However, not all ion channels expressed in human myocardium are equally well represented in hiPSC-CM. In particular, studies suggest that I K1 , I Ks and I NaLate currents have minimal contributions to the electrophysiology of the iPS cells (Paci et al., 2012).
The Na v 1.5 channel that generates the upstroke phase and the Ca v 1.2 responsible for maintaining the plateau phase of AP are known to be active. At the end of the plateau phase and beginning of repolarization, inward currents are small (largely inactivated) and countered by the activation of outward K + currents, predominately hERG, which is responsible for initiating the repolarization phase, is well represented in iPS cell. We have previously reported the absence of I NaLate effect in the presence of ranolazine in hiPSC-CM (Hortigon-Vinagre et al., 2016), which shows that presence of I NaLate in hiPSC is unlikely. Yang et al. (2014) reported an enhancement of I Na-L by dofetilide after chronic (5 h) drug exposure. Drug effects in our experiments were, however, studied a after short-time (several minutes) of application and no increase in I NaLate was observed. It is under discussion whether commercial hiPSC cell lines contain a range of cell types or simply broadspectrum features. The majority of hiPSC cells appear to have ventricular phenotype, and they are likely to operate as a functional syncytium via gap junction links (Bett et al., 2013;Kane et al., 2016).
Combining in silico studies with hERG (and other ion channels) inhibition and effects on APD should enable a more profound understanding of pro-arrhythmic potential. To test this concept, we compared the prolongation of APs  2) and 1 μM (Na v 1.5), orange 100 nM (Ca v 1.2) and 10 μM (Na v 1.5), magenta 500 nM (Ca v 1.2) and 1 μM (Na v 1.5), green 500 nM (Ca v 1.2) and 10 μM (Na v 1.5). See also Supporting Information Table S1 comparing the values used in silico AP models of adult ventricular myocytes and hiPSC-CM. of hiPSC-CM by the selective hERG inhibitor dofetilide and 13 derivatives with respect to their potencies to inhibit hERG ( Figure 5). Derivatives retained the common scaffold of dofetilide while changing the functional group on both the ends or modifying the central nitrogen or altering the length of the molecule. The 13 derivatives inhibited hERG potassium channels with IC 50 s ranging from 3 to 300 nM ( Figure 4B, C). Examining the effects of these derivatives on hiPSC-derived cardiac myocyte APD revealed a correlation between the concentration (D 150 ) inducing a 50% increase of APD 90 of the cardiac AP with half-maximal concentrations (IC 50 s) of hERG channel inhibition ( Figure 5 and Supporting Information Fig. S2).
There was no correlation between the K i values (affinity of derivatives to hERG estimated in radioligand studies; Shagufta et al., 2009), and IC 50 s measured in electrophysiological experiments (see Supporting Information Table S2) was observed. All derivatives (except Dofe30) were similarly active in the binding study while IC 50 s measured in patch clamp experiments varied over two orders of magnitude (from 2.6 to 296 nM, Table 2). The lack of correlation between K i and IC 50 indicates that the interaction of these derivatives with their binding pocket is not the only determinant of hERG channel inhibition (Saxena et al., 2016). The K i value reflects the affinity of a derivative for the binding pocket putatively located in the channel pore while the IC 50 s estimated in patch-clamp studies are affected by the following: (i) channel state-dependent drug effects (Fernandez et al., 2004;Sanguinetti and Tristani-Firouzi, 2006;Stork et al., 2007;Perry et al., 2010;Windisch et al., 2011); (ii) their ability to pass the entrance barrier or leave the channel cavity; and (iii) their affinity to the binding pocket within the channel. In this regard, it is interesting to note that the IC 50 s estimated from hERG inhibition in functional studies are in a reciprocal relation to the molecular weight of the tested derivatives (Supporting Information- Fig. S3). It is tempting to speculate that the dependence of IC 50 on the molecular size is caused by an energetic barrier at the channel pore entrance. In such a scenario, bulkier molecules with higher molecular weight leave the channel with lower probability, resulting in lower off rates and correspondingly in lower IC 50 values.
Three derivatives (Dofe42, 44 and 45) failed, however, to fit a linear correlation ( Figure 5). We hypothesize that the ineffectiveness of derivatives Dofe45, Dofe44 and Dofe42 to prolong the AP might be due to their interference with Ca v 1.2, Na v 1.5 and potentially other ion channels. Dofe45 was subsequently shown to be a potent inhibitor of Ca v 1.2. In a first series of experiments, performed at a low stimulus frequency (0.2 Hz), this derivative inhibited Ca v 1.2 with an IC 50 of 190 ± 3 nM ( Figure 6A, B). It is well established that open and inactivated channels may have a higher affinity for inhibitors than channels in the resting state (Hondeghem and Katzung, 1977). Therefore, additional measurements were made at a higher frequency of (1 Hz), which is comparable with the beating frequency of iPSC-CM. The shorter (50 ms) pulses (1 Hz) revealed some additional use-dependent channel inhibition by Dofe42 and 45 ( Figure 6C, D). Thus, 1 Hz pulsing can enhance Cav1.2 inhibition due to additional block of open and/or inactivated channels. However, both derivatives inhibited Ca v 1.2 predominantly in the resting state ( Figure 6). A comparison of I Na inhibition at À140 and À80 mV close to the resting potential of iPSC-CM, where more than 60% of Na v 1.5 channels are in an inactivated state (Hoekstra et al., 2012;Wang et al., 2015), revealed that Dofe42, Dofe44 and Dofe45 preferentially inhibit inactivated Na v 1.5. This study is mainly focused on primary targets of these derivatives like I Na and I CaL , and it is very unlikely that these derivatives would have an effect on secondary targets such as I Ks , Na/K pump, NCX and/or SR Ca 2+ release.
As shown in Figures 5 and 8, our in silico studies on the AP model (Paci et al. 2012) at a resting potential of À80 mV reproduced the link between hERG inhibition (IC 50 ) and prolongation of the AP (D 150 ). Furthermore, accounting for inhibition of hERG, Cav1.2 and Nav1.5 by Dofe45 reproduced the principal features of AP changes observed on hiPSC-CM (compare Figures 3C and 8C). The acceleration of early repolarization (phase 1) and inhibition of the AP overshoot are obviously caused by simultaneous inhibition of sodium channels while the prolongation of the AP was predominantly balanced by simultaneous block of Ca v 1.2. Hence, as illustrated in Figure 8D, selective hERG inhibition by 300 nM Dofe45 would induce a more pronounced AP prolongation. The inability of Dofe42 and Dofe44 to prolong the AP is hard to explain exclusively by inhibition of Ca v 1.2 and Na v 1.5 as these channels appear to be blocked only at high concentrations. But the conditions of the ion channel assay are not the same as those of the iPSC-CMs. The oscillatory voltage changes and the temperature will almost certainly alter the level of activation/inactivation of the currents. Both Na v 1.5 and Ca v 1.2 show voltage-and time-dependant effects of drugs, and inhibition of inactivated Na v 1.5 by Dofe42 and Dofe44 was stronger at À80 mV relative to À140 mV ( Figure 7). Furthermore, Ca v 1.2 showed use-dependent block of by Dofe42 ( Figure 6C, D). Therefore, the precise effect of drugs on both of these channels in the context of an AP in IPSC-CMs is difficult to assess. Also, while Ca v 1.2 and Na v 1.5 are the most likely candidates for alternative drug actions, these derivatives may also modulate other ion channels that contribute to the AP shape.
The implications of this work are that potency of hERG current inhibition correlates linearly with an index of APD in hiPSC-CMs. This simple relationship, confirmed in silico, allows data gained in one standard assay to predict the effect on another, that is, the IC 50 of a drug in an ion channel hERG screen predicts the dose required to increased APD 90 in iPSC-CMs or vice versa. Furthermore, compounds that do not correlate will have additional effects including concomitant block of Ca v 1.2 and/or Na v 1.5 channels. Finally, the study shows that while in silico simulations can confirm the principle of the effects of Ca v 1.2 and Na v 1.5 inhibition on APD 90 , more comprehensive voltage clamp data are required to accurately predict the consequences of Ca v 1.2 and Na v 1.5 block on AP shape and duration in silico.

Limitations
In the myocardium and hiPSC-CM hERG, Ca v 1.2 and Na v 1.5 channels function under different conditions than in patch clamp experiments on mammalian cells. In order to relate patch clamp data to the hiPSC-CM assay, it would be desirable to study these ionic currents at the beating hERG inhibition and AP prolongation BJP frequency of hiPSC-CM (~1 Hz) at 37°C. But most patch clamp ion channel assays place constraints on the design of the pulse protocol. As illustrated in Figure 6C, D, continuous 1 Hz pulsing with even short (50 ms) test pulses results in peak current decay of calcium currents caused by channel inactivation. Application of longer test pulses (e.g. 300 ms, corresponding to the length of the ventricular cardiac AP) at 1 Hz leads to inactivation by 30 and 40%, during a train of 20 pulses. Further optimization of experimental (temperature, test pulse length and shape, holding potential, pacing frequency, etc.) and theoretical conditions (analysis of participation of additional ionic currents in AP shaping) is required to achieve a higher level of congruence between the different assay data.

Supporting Information
Additional Supporting Information may be found online in the supporting information tab for this article.
https://doi.org/10.1111/bph.13942 Figure S1 Dofetilide (100 μM) does not inhibit Cav1.2 or Nav1.5. Figure S2 Estimation of dose required to prolong the action potential by 150% (D 150 ). Figure S3 Relationship between potencies of dofetilide derivatives to inhibit hERG (IC 50 ) and their MW (MW). Table S1 Major maximal conductance of ion channels used for AP simulations of human embryonic stem cell-derived myocytes described in Paci et al. () and corresponding values used for adult ventricular cardiomyocyte models. hERG inhibition and AP prolongation BJP