Identification of Na+/K+-ATPase inhibition-independent proarrhythmic ionic mechanisms of cardiac glycosides

The current study explored the Na+/K+-ATPase (NKA) inhibition-independent proarrhythmic mechanisms of cardiac glycosides (CGs) which are well-known NKA inhibitors. With the cytosolic Ca2+ chelated by EGTA and BAPTA or extracellular Ca2+ replaced by Ba2+, effects of bufadienolides (bufalin (BF) and cinobufagin (CBG)) and cardenolides (ouabain (Oua) and pecilocerin A (PEA)) on the L-type calcium current (I Ca,L) were recorded in heterologous expression Cav1.2-CHO cells and human embryonic stem cell-derived cardiomyocytes (hESC-CMs). BF and CBG demonstrated a concentration-dependent (0.1 to 100 µM) I Ca,L inhibition (maximal ≥50%) without and with the NKA activity blocked by 10 µM Oua. BF significantly shortened the action potential duration at 1.0 µM and shortened the extracellular field potential duration at 0.01~1.0 µM. On the other hand, BF and CBG at 100 µM demonstrated a strong inhibition (≥40%) of the rapidly activating component of the delayed rectifier K+ current (I Kr) in heterologous expression HEK293 cells and prolonged the APD of the heart of day-3 Zebrafish larva with disrupted rhythmic contractions. Moreover, hESC-CMs treated with BF (10 nM) for 24 hours showed moderate yet significant prolongation in APD90. In conclusion, our data indicate that CGs particularly bufadienolides possess cytosolic [Ca2+]i- and NKA inhibition- independent proarrhythmic potential through I Ca,L and I Kr inhibitions.

and enhanced automaticity 1,5 . In cardiac myocytes, CG-induced accumulation of cytosolic [Ca 2+ ] i is known to accelerate [Ca 2+ ] i -dependent inactivation (CDI) of the L-type Calcium channels (LTCC) and lead to the suppression of the LTCC currents (I Ca,L ) 11,12 , which could be responsible for the shortening of cardiac action potential (AP) duration (APD) in cardiac myocytes of various animal species 11,[13][14][15][16] and the shortening of the cardiac field potential (FP) duration of human induced pluripotent stem cell-derived cardiomyocytes 17 . Moreover, increased [Ca 2+ ] i. and SR Ca 2+ load by NKA inhibition could activate Calcium-Calmodulin Kinase II (CaMKII)-induce phosphorylation of the ryanodine receptor (RyR) on the sarcoplasmic reticulum (SR) of cardiac myocytes and enhanced SR Ca 2+ leak and exacerbate spontaneous cardiac activity 18,19 . However, questions remain as whether NKA inhibition and [Ca 2+ ] i accumulation by CGs is fully responsible for the pro-arrhythmic electrophysiological changes. The dissociation between inotropic and proarrhythmic actions of CGs indicates that the proarrhythmic effects of CGs may not result solely from NKA inhibition [18][19][20][21] . For example, Chan Su (toad venom) and its key ingredient BF have demonstrated more potent cardiotoxicity over Oua as additional proarrhythmic effects were achieved by Chan Su or BF after the Na + /K + -ATPase activity was fully blocked by Oua 20,22 .
CG-induced cardiotoxicity remains to be a threat to human health and it may hamper their clinical potentials as hypertonic and anti-cancer drugs. Hence, it is important to fully understand the proarrhythmic effect and mechanisms of CGs. Here, we explored the potential [Ca 2+ ] i -and NKA inhibition-independent proarrhythmic ionic mechanism of CGs. Bufadienolides (BF and CBG) and cardenolides (Oua and Pecilocerin A (PEA)) ( Fig. 1) with different NKA inhibition potencies 13,[21][22][23][24] , were assayed with the cytosolic Ca 2+ chelated, extracellular Ca 2+ or K + depleted and NKA activity blocked. Effects of CGs on major cardiac ion channel currents and corresponding repolarization durations were assessed in heterologous expression systems, human embryonic stem cell (hESC)-derived cardiomyocytes (hESC-CMs) and zebrafish heart, respectively. Our data confirmed that, under physiological/toxicological concentrations, CGs particularly Bufadienolides like BF could either inhibit LTCC independently from intracellular [Ca 2+ ] i and NKA-inhibition, or, inhibited hERG at both high and low (prolonged effect) concentrations.

Results
Based on that used in previous in vitro experiments [5][6][7][8][9] and in patients 10 , the concentrations of BF, CBG, Oua and PEA adopted in the current study ranged from 0.1 to 100 µM, whereas 0.01~1.0 µM falls in the therapeutic range and 1.0~10 µM overlapped with the cardiotoxicity range.

Cytosolic [Ca 2+ ] i -independent effects of CGs on major cardiac ion currents. Automated patch
clamping was adopted to screen the impacts of CGs on major cardiac ion currents in heterologous expression cell lines. With the intracellular solutions contained 20 mM of EGTA and 5 mM BAPTA (to chelate the intracellular free Ca 2+ , BF and CBG demonstrated concentration-dependent (1~100 µM) inhibition of I Ca,L over baseline (with maximized inhibition of ~70% and ~60%, and IC 50s of 12.5 µM and 15.3 µM, respectively) which were more remarkable than that of Oua and PEA (with maximized inhibition ~15% and ~10%, and estimated IC 50s of way above 100 µM). The concentration-dependent effects of BF and CBG on I Ca,L inhibition were confirmed by One-Way ANOVA test showing significant intra-group (among different concentrations) differences (p < 0.001). In addition, BF and CBG produced a marked (>40%) inhibition of the rapidly activating component of the delayed rectifier K + current (I kr ) at 100 µM with estimated IC 50s at slightly above 100 µM compared to that of Oua and PEA which were way above 100 µM. On the other hand, all CGs demonstrated limited inhibitions (all <20%) on I Na (Fig. 2) with the estimated IC 50s way above 100 µM. Outcomes from Two-Way repeated ANOVA testing indicated that the effects of BF and CBG on I Ca,L , I Kr and I Na , respectively did not differ from each other whereas the effects of BF and CBG on I Ca,L at 1.0 µM, 10 µM and 100 µM were significantly stronger than Oua and PEA at the corresponding concentrations (p < 0.001). At 100 µM the inhibitory effects of BF and CBG on I kr were much stronger than Oua and PEA (p < 0.001).
To evaluate the accuracy of our automated patch-clamping recording, the effects of BF and CBG on I Kr were independently verified by Shanghai Institute of Materia Medica, Chinese Academy of Science, where hERG-CHO cells (different from hERG-HEK293 we used) and a Sophion QPatch Automated Patch Clamp Systems (Biolin Scientific, Stockholm, Sweden) were used. At 40 µM (the highest concentration used), BF and CBG suppressed I kr by ~20% and ~40%, respectively, while Cisapride (a positive control) achieved a full blocking of hERG. Therefore, these results confirmed our findings on I Kr .
The cytosolic Ca 2+ -independent effects of BF on I Ca,L were further validated in Cav1.2-CHO cells under Ca 2+ -free extracellular conditions (Fig. 2D). With the extracellular Ca 2+ replaced by Ba 2+ which carried the current through Ca 2+ channels, application of 10 μM BF rapidly reduced I Ca,L density by ~30%, suggesting that BF may possess a direct inhibitory effect on LTCC.
NKA inhibition-independent effect of BF and CBG on I Ca,L . As it is well-received that NKA-inhibition by CGs is the cause of the increase cytosolic [Ca 2+ ] i and CDI 1, 3-5 , we further validated the role of NKA inhibition in the inhibitory effect of BF and CBG on I Ca,L . As Oua has been commonly adopted as a standard NKA inhibitor, we treated Cav1.2-CHO and hESC-CMs with 10 µM Oua which is known to completely block the NKA activity 20,22 . As shown in Fig. 3A, BF and CBG concentration-dependently inhibited I Ca,L in Cav1.2-CHO (with extracellular KCl replaced by CsCl to further suppress the NKA activity) with similar magnitudes as shown in Fig. 2A. The concentration-dependent effects of BF and CBG on I Ca,L inhibition were confirmed by One-Way ANOVA test showing significant intra-group (among different concentrations) differences (p < 0.001). Moreover, BF (1.0 and 10 µM, CBG not tested) significantly inhibited I Ca,L in hESC-CMs and the effect was reversed by Bay K8644 (1 µM), an opener of LTCC (Fig. 3Ba, Bb). Noticed that the magnitude of peak I Ca,L inhibition was greater than that during the current decay (Fig. 3Bc).

NKA inhibition-independent effect of BF on [Ca 2+ ] i transients in hESC-CMs. [Ca 2+ ] i transients
in hESC-CMs is facilitated by Ca 2+ -induced Ca 2+ release (CICR) from SR which depends on the Ca 2+ influx through LTCC. A previously report that Chan Su and BF were capable of suppressing [Ca 2+ ] i in neonatal rat cardiac myocytes with NKA activity fully inhibited by 10 µM Oua 20 . Similarly, we observed that on hESC-CMs pre-incubated with 10 µM Oua for 3 minutes, BF (10 µM) reduced the [Ca 2+ ] i transients amplitude (from 6.47 ± 0.90 to 4.18 ± 0.77, p < 0.001, n = 7) and duration at 50% and 80% recovery (CaD50 and CaD80), and increased time to peak (Fig. 4). Such findings support the inhibitory effect of BF on LTCC.

Cytosolic [Ca 2+ ] i -independent effects of CGs on APs of hESC-CMs.
To correlate the CG-induced changes in cardiac ion currents with transmembrane activities, APs were recorded in single hESC-CMs under cytosolic [Ca 2+ ] i -independent condition (with 5 mM EGTA added in the pipette solution). At 1.0 µM, all CGs tended to shorten APDc (APD90, PAD50 and APD30) and it was more evidenced with BF and CBG. Yet only the effects of BF were statistically significant (Fig. 5A). Moreover, a moderate reductions of AP amplitude (APA) and a more depolarized shift of the maximal diastolic potential (MDP) were observed with BF and CBG, yet only BF-induced APA reduction and CBG-induced positive shift of MDP were significant (Fig. 5A). No changes in the maximal upstroke velocity (dV/dt Max ) and beating frequency were noted. BF (10 µM) treatment for 1-2 minutes achieved more remarkable changes in APs, followed by polymorphic arrhythmia-like AP waveform at 5 minutes and AP firing ceased at 5~8 minutes (data not shown).

Effects of CGs on cardiac field potentials of hESC-CMs.
Effects of CGs on the transmembrane activity of hESC-CMs were further tested by recording the extracellular FP of hESC-CM clusters without altering the cytosolic Ca 2+ . A trend of decreasing FP duration (FPD) by BF, Oua and PEA was noticed while only BF exerted a significant (p < 0.05) concentration-dependent (0.01~1 µM) effect (Fig. 6A,C). Moreover, BF and Oua tended to increase the beating frequencies yet the changes were insignificant (Fig. 6B). At 10 µM, all CGs stopped the beating of hESC-CMs in 3~5 minutes (data not shown) and the beating resumed after washout (data not shown).

Effects of CGs on APs of zebrafish.
As all CGs at 10 µM abolished the transmembrane electrical activity in vitro in hESC-CMs, the effects of higher doses of CGs were tested in day-3 zebrafish larva which demonstrated insensitivities to CGs indicated by retained normal heart rhythm and APs after exposing to up to 10 µM of all CGs. However, zebrafish larvae responded to 100 µM BF and CBG demonstrated decreased heart rates (by ~27% and ~38%, respectively), prolonged APDc (by ~52% and ~63%, respectively) and early-after depolarization and polymorphic arrhythmia-like changes (Fig. 7). Effects of BF and CBG are significantly distinguishable from that of Oua, PEA and digoxin (used as a control) which showed little effect on heart rate and APDc (Fig. 7).

Prolonged effects of CGs on action potentials of hESC-CMs. A previous study has demonstrated
that CGs such as digoxin are able to delay cardiac repolarization at nanomolar concentrations associated with inhibitions of the expression and trafficking of hERG 25 . We tested the prolonged effects of 10 nM BF on APs in hESC-CMs. Cells treated with BF ≤ 10 nM for 24 hours maintained rhythmic contractions. The APs showed moderate yet significant prolongation in APD90 (Fig. 5B).  27 were adopted to validate the impacts of CG-induced ion current changes on APs. The ORd model showed that increased [Na + ] i was able to reduce APD (Fig. 8A). Yet reduced I NaK (70%, 50%, 25% and 10% of control) alone failed to positively shift MDP and shorten APD (Fig. 8B). Next, reduced I Ca,L (70%, 50%, 25% and 10% of control) alone (Fig. 8C) or with 0%  I NaK (Fig. 8D) led to marginal/moderate decrease in APDs. On the other hand, the simulated effects of reduced I Kr (70%, 50%, 25% and 10% of control) on APs confirmed the experimental findings (Fig. 8E). Interestingly, the LRd model revealed marked APD shortening caused by reduced (70%, 50%, 25% and 10% of control) slow inactivation inward current (I CI ), an potential equivalent of I Ca,L (Fig. 8F). (Aa) Representative AP waveforms recorded in hESC-CMs exposed to 1 µM of BF, CBG, Oua and PEA for 1-2 minutes. (Ab) Representative AP waveforms recorded in hESC-CMs treated with BF (10 µM) for different time points. (Ac) Bar-graphs show APD90, APD30, APA and MDP measured in hESC-CMs 2 minutes post application of BF, CBG, Oua and PEA. * p < 0.05, † p < 0.01, vs. baseline (two-tailed paired Student t-tests). Data are presented as mean ± SEM. (B) Representative superimposed AP waveforms recorded in hESC-CMs exposed to 0.1% DMSO (as vehicle control) and BF (10 nM) for 24 hours. n = 7.

Discussion
The current study identified the NKA inhibition-independent proarrhythmic mechanisms of CGs which involve key cardiac ion channel currents and transmembrane potentials. With the cytosolic Ca 2+ largely chelated or extracellular Ca 2+ depleted, bufadienolides demonstrated a cytosolic [Ca 2+ ] i -independent I Ca,L inhibition, accompanied by APD shortenings. With the NKA activities completely inhibited by Oua, effects of bufadienolides on I Ca,L persisted. Further, a strong I kr blockage in hERG-HEK cells and a prolonged APD in zebrafish larva were observed by BF and CBG at 100 µM, whereas prolonged APD was also observed in hESC-CMs treated with 10 nM BF for 24 hours. Lastly, results from computer simulation largely validated the experimental findings on the relationship between ion currents and APs.
Data from the current study indicate that NKA inhibition-independent I Ca,L inhibition could contribute to the greater proarrhythmic effects of BF and CBG compared with Oua and PEA. While it is known that 10 μM of Oua could totally blocked the NKA activities 20,22 and BF and CBG are 4~7 folds more potent NKA blockers than Oua 23 , the more potent proarrhythmic effects of BF in APD shortening 13 28,29 . In addition, BAPTA, the other calcium chelator we added to the intracellular solution (5 mM), could, by itself, fully block CDI 29 . Furthermore, our data showed that the I Ca,L inhibitory effect of BF retained in Cav1.2-CHO cells with extracellular Ca 2+ replaced by Ba 2+ , while depletion of extracellular Ca 2+ is known to fully blocked CDI 29 . Secondly, we gathered evident to support that BF and CBG possess NKA inhibition-independent effects. The concentration-dependent I Ca,L inhibition by BF and CBG peaked at 100 µM and it further suggests that BF and CBG are capable of inducing NKA inhibition-independent proarrhythmic effects as <2.5 µM of BF or CBG could be sufficient to fully block the NKA activities 18,[22][23][24] . Next, we noticed that the effects of BF and CBG on I Ca,L and [Ca 2+ ] i transient persisted in cells incubated with 10 µM Oua. The rapid inhibition of I Ca,L and shortening of APs by BF and CBG support direct effects on LTCC since the effects of NKA inhibition and CDI on APs could take longer time 13 . While CDI of I Ca,L is known to accelerate the decay of currents without supressing the peak currents, our finding that BF, without/with NKA activity blocked by Oua, markedly inhibit peak I Ca,L with magnitudes greater than that for the decay current (Fig. 3Bc), further support the notion that CGs are capable of NKA inhibition-and CDI-independent I Ca,L inhibition.
I kr blocking by CF and CBG offers an additional support to the NKA inhibition-independent proarrhythmic mechanisms of CGs. Our data show that BF and CBG at 100 µM demonstrated a strong and unprecedented I kr blocking (>40%) which could be behind the APD prolongation in zebrafish lava. However, the impact of a potential I Ca,L inhibition, in completion with I Kr inhibition, on APD in Zebra fish remains unclear. Result from computer simulation may suggest that APD is more sensitive to I Kr inhibition. It appears that zebrafish is less sensitive to CG-induced cardiotoxicity probably due to inter-species variations and the arrhythmogenic concentration of CGs is less physiologically relevant. With more clinical relevant concentrations (the recommended digoxin concentrations for the treatment of chronic heart failure are around 0.8 ng/ml (1.2 nM) with a half-life of 36 hours and the toxic level is more than 2 ng/ml 5,30 ), we demonstrated that CGs could prolongation APD and such an effect could be attributed to the chronic effects of low concentration CGs on hERG inhibition 25 .
The well-documented effect of CGs on depolarizing the MDP and shortening APD in isolated cardiac myocytes 13, 14 appeared to be less obvious in hESC-CMs (Fig. 5). This discrepancy could be due to the shorter exposure time (1-2 minutes) designed to favours the observation of the direct effects of CGs on the ion channels; a more depolarize (~60 mV) MDP 31 in hESC-CMs compared with the −80 mV ~ −90 mV MDP in isolated human ventricular myocytes, and the EGTA (10 mM) added in the pipette solution that could suppress CDI and minimize the APD shortening effect of CGs in hESC-CMs.
Effects of CGs on APD are supported by the simulated results in principle although only moderate APD shortening was simulated by the ORd human ventricular myocyte model (2011) in responds to a marked I Ca,L reduction. Such a discrepancy may reflect the limitations with the ORd model, which has been shown failing to simulate considerable changes in APs with marked inhibitions of I NaK and I Ca, L 32 . Compared with the moderate negative correlation between [Na + ] i and APDs simulated by the ORd model, Grandi et al. 33 , using a mathematical model for Ca 2+ handling and ionic currents in the human ventricular myocyte, demonstrated that Na + loading could be a major determinant of ventricular APD shortening, and therefore that model could be a better choice for modeling effects of NKA block and I Ca,L block.
While the arrhythmogenic potential associated with LTCC inhibition could be less understood compared with hERG inhibition, the association between reduced I Ca,L and cardiac arrhythmias is supported by the clinical findings. For example, loss-of-function mutations in the LTCC have been associated with QT shortening and severe arrhythmia in patients with the Brugada syndrome 34 .
Overall, our data may explain the dissociation between the positive inotropic effect and the proarrhythmic effects of CGs. For example, the additional LTCC inhibition by BF over Oua may undermine the hypertonic effect of BF by attenuating the [Ca 2+ ]i and accelerate the proarrhythmic effects in the meantime.

Conclusions
We collected evidence to support that bufadienolides such as BF possess NKA inhibition-independent proarrhythmic effects associated with LTCC and hERG blocking.

Materials and Methods
CG compounds. The structure formulas of BF, CBG, Oua and PEA are shown in Fig. 1. BF, CBG and PEA were prepared in our laboratory at Jinan University, Guangdong, China. BF and CBG were isolated and purified (>99%) from the traditional Chinese medicine Chan Su derived from toad venom 35 while PEA was isolated and purified (>99%) from the plant Asclepias curassavica L. The quality of BF, CBG and PEA were confirmed by 1D NMR spectra assay. BF: 1  Oua and digoxin were purchased from Sigma-Aldrich (St. Louis, MO, USA). BF, CBG, and PEA stocks were made using DMSO whereas Oua and digoxin stocks were prepared in distilled water. The stocks were diluted using external solutions or cultural medium (for MEA assay) to make up the final concentrations for recording. The final concentration of DMSO is ≤0.2%.

Chemicals.
Unless specified, all reagents for electrophysiology assays, including Bay K8644, were obtained from Sigma-Aldrich (St. Louis, USA).

Cells.
Heterologous expressing systems were adopted for evaluating the effects of CGs on major voltage-gated cardiac ion channel currents, including the repolarising rapid component of the outward rectifier potassium current (I Kr ), depolarizing peak sodium current (I Na ) and depolarizing I Ca,L . Stably expressing cell lines included: hERG-HEK293 as HEK293 cells expressing human KCNH2 or Ether-a'-go-go-Related gene (hERG) which encode hKv11.1 channel for I kr (Chan Test, Cleveland, USA); SCN5A-HEK293 as HEK293 cells expressing human SCN5A and SCN1B which encodes hNav1.5/β1 subunit of sodium channels for I Na (Anaxon AG, Berne, Switzerland) and Cav1.2-CHO as CHO cells expressing human CACNA1C/CACNB2/CACNA2D1 genes which encodes hCav1.2/β2/α2δ1 channel or LTCC for I Ca,L (ChanTest, Cleveland, USA).
hESC-CMs were adopted as the most relevant in vitro human model for drug testing. H3 hESCs (WiCell Research Institute, Madison, USA) were differentiated into cardiomyocytes following the published protocols 31,36 . Cells were maintained at 37 °C in a humidified CO 2 (5%) incubator in RPMI 1640 Medium containing 2% of B-27 ® supplement (+Insulin, 1 mL/50 mL) and 1% of Penicillin-Streptomycin-Glutamin (0.5 mL/50 mL), all from Invitrogen (Singapore). hESC-CMs 30~35 days post differentiation were used for MEA and AP recordings. Two-step voltage protocols were adopted for recording I Ca,L , I Kr and I Na while the corresponding ion currents were evoked at the 2 nd step. I Ca,L was recorded by depolarization of the cell membrane potential from a holding potential of −80 mV (50 ms) to +10 mV and held for 100 ms. I kr was recorded by a 1 st pulse (+30 mV for 2000 ms) followed by a 2 nd pulse at voltage −50 mV for 4000 ms to evoke the I Kr peak tail currents. I Na was measured by clamping the cells from a holding potential of −120 mV (5 ms) to 0 mV for 10 ms. The I Ca,L , I Na and I Kr were validated by corresponding blockers (positive controls) which achieved full blocks.

Cardiac ion channel currents measurement by automated patch-clamping. Effects of CGs on
Conventional patch-clamp recordings. Whole cell configuration of the patch-clamp technique was used to measure I Ca,L in Cav1.2-CHO and hESC-CMs and APs in hESC-CMs 31 . The signal was amplified using an Axopatch 700B patch clamp amplifier (Molecular Devices, Foster City, USA) and low-pass filtered at 5 kHz. Patch pipettes were fabricated from glass capillaries using a Sutter P-97 microelectrode puller (Novato, CA, USA) and the tips were heat polished with a microforge (NARISHIGE MF-900, Tokyo, Japan) to gain a resistance of 2-4 MΩ. The electrical signals were sampled at 2.5-10 kHz and filtered at 2 kHz using a low-pass filter. Data acquisition was achieved using the Digidata 1440 A (Axon Instrument). Data analysis and fit were performed using clamp fit 10.2 (Axon Instrument) and Origin 7.0 software (Origin Lab Corporation). A pClamp software (Version8.1; Axon Instrument) was used to generate voltage-pulse protocols, acquire and analyze data. I Ca,L recording in Cav1.2-CHO. To verify the CDI-independent effects of BF, CaCl 2 in the extracellular solution was replaced by BaCl 2 so to let Ba 2+ ions carried the current through Ca 2+ channels 38 . I Ca,L recording in hESC-CMs. I Ca,L was recorded in hESC-CMs as previously described 38 . Patch pipettes solution contained (in mM): CsCl 120, MgCl 2 3, MgATP 5, EGTA 10, and HEPES 5, adjusted to pH 7.2 with CsOH. External solution contained (in mM): NaCl 140, CsCl 10, CaCl 2 1.8, MgCl 2 1, glucose 10, and HEPES 10, adjusted to pH 7.4 with NaOH. To eliminate the 'run-down' effect during I Ca,L recordings, Ba 2+ was also used in the external solution (BaCl 2 1.8 mM) to replace Ca 2+ as charge carrier of calcium channel current 11 . Current-Voltage curve were generated by voltage clamp protocols consisting of V hold = −80 mV followed by a 3 s long pre-pulse at −50 mV to inactivate Na + and T-type Ca 2+ channels, then a family of 300 ms depolarization from −50 mV to 50 mV in 10 mV increments.
Calcium channel current densities were obtained by dividing current amplitudes by membrane capacitances. Steady state inactivation variables of I Ca,L were determined using a two-pulse gapped protocol. Potential was held at −40 mV, then pulsed to a conditioning pre-pulse ranging from −80 mV to +10 mV for 2000 ms, returned to −40 mV for 10 ms, and stepped to 0 mV for 250 ms at 10 s intervals. Voltage-dependence of activation curve and steady state inactivation curve were fitted with Boltzman equation where G is the conductance at various test potentials and was calculated from the peak current according to G = I/(V − Vrev), Vrev is the reversal potential obtained by extrapolating the linear part of the I/V curve to its intersection with the voltage axis. Gmax is maximum conductance; V 1/2 and κ are half-activation voltage and the slope factor. The concentration-response data were fitted with Hill equation: I/Imax = 1/[1+ (D/IC 50 ) n ], where I is the peak current in various concentrations of compound, Imax is the maximal peak current, D is the compound concentration, IC 50 is the drug concentration for 50% inhibition, and n is the Hill coefficient.
The time course of recovery from inactivation of I Ca,L was studied using a two-pulse protocol: a 250-ms pre-pulse (P1) at 0 mV from the holding potential of −50 mV followed by a variable recovery period and a 250-ms test pulse (P2) at 0 mV to assess the amount of current recovered. Each two-pulse sequence was separated by a 30 s interval. The time course of recovery for I Ca,L was determined by fitting the data points to a single exponential function: I/Imax = 1 − exp(−t/τ), where Imax and I were the peak current at pre-pulse (P1) and test pulse (P2), respectively; t was the variable recovery time; τ was the recovery time constant. Multi-electrode Arrays. The extracellular FP produced by hESC-CMs was measured by multi-electrode array (MEA) assay using Multi Channel Systems MCS GmbH (Aspenhaustrasse, Reutlingen, Germany).

AP measurement in hESC-
Clusters of contracting hESC-CMs were plated on Matrigel ® coated MEA chips containing 59 titanium electrodes and 1 internal reference electrode. Stocks of CGs were diluted to various concentrations in cell cultural medium (RPMI1640 basal medium). Data was acquired with an interval of 5 minutes at baseline and post drug applications. The FPD was analysed and corrected by the beating frequency (FPDc) with Fridericia's formula (FPDc = FPD/interspike interval 1/3 ) 39 .
Cardiac AP measurement in zebrafish. Wild-type zebrafish (AB, ZIRC) were maintained as described 40 .
All animal experiments were carried according to the regulations of Institutional Animal Care and Use Committee (Biological Resource Center of Biopolis of Singapore, license no. 120787), which approved this study. Developmental stages are in hours post fertilization (hpf) at 28.5 °C 41 .
Micropipettes for AP measurement on whole zebrafish larvae were prepared by pulling fire-polished borosilicate glass capillaries (World Precision Instruments) using the Flaming/brown micropipette puller P-1000 (Sutter Instrument). The micropipette was filled with internal buffer contained (in mM): NaCl 174, KCl 2.1, MgSO 4 .7H 2 0 1.2, Ca(NO 3 ) 2 .4H 2 O 1.8, HEPES 15, adjusted to pH 7.2. The micropipette tip was positioned right above the pericardial region of the zebrafish heart. The electrical signals were recorded by pCLAMP 10 software (Molecular Devices) after amplification via Multiclamp 700B amplifier (Molecular Devices, Foster City, USA) and digitization through Axon Digidata 1440A digitizer (Molecular Devices). Data were analysed with Clampfit 10 software (Molecular Devices). For controls, the zebrafish larvae were mounted (laterally) in 1% low melting agarose in a glass dish and submerged in external buffer: 1X Egg water (0.6 g/L sea salt in reverse osmosis purified water). To assess the effect of drugs on the heart activity of zebrafish, the mounted larvae are pre-treated with the various drugs (diluted in1X Egg water). APDs were determined by measuring the adjacent peaks of the action potentials using the pClamp software. APDs were corrected by heart rate as above-mentioned. Statistical analysis. Numerical data are presented as mean ± standard division (SD) or mean ± standard error of mean (SEM). Comparisons were made with paired and unpaired (two-tailed) Student t-test, One-Way repeated measures ANOVA followed by the Tukey's post hoc testing and Two-Way repeated measures ANOVA followed by the Bonferroni post hoc testing using GraphPad Prism 5.0 (GraphPad Prism 5.0, La Jolla, USA). A p-value of <0.05 was considered statistically significant.