Screening system for drug-induced

RESEARCH ARTICLE
HEALTH AND MEDICINE
Screening system for drug-induced arrhythmogenic
risk combining a patch clamp and heart simulator
2015 © The Authors, some rights reserved;
exclusive licensee American Association for
the Advancement of Science. Distributed
under a Creative Commons Attribution
NonCommercial License 4.0 (CC BY-NC).
10.1126/sciadv.1400142
Jun‐ichi Okada,1,2* Takashi Yoshinaga,3 Junko Kurokawa,4 Takumi Washio,1,2
Tetsushi Furukawa,4 Kohei Sawada,3 Seiryo Sugiura,1,2 Toshiaki Hisada1,2
INTRODUCTION
Cardiac arrhythmias are serious side effects of commonly used drugs.
Torsade de Pointes (TdP) is a rare but potentially lethal type of ventricular arrhythmia often caused by inhibition of the repolarizing potassium current of cardiac myocytes (1). The historical incidence of
drug-related sudden cardiac death led to the establishment of regulatory requirements for cardiotoxicity testing, which include the measurement of repolarizing potassium current through hERG potassium
current, an in vivo QT assay in an animal model, and the examination
of QT interval in healthy volunteers (a thorough QT study) in 2005 (2).
The practice of these guidelines has been successful in preventing
arrhythmia-related accidents. However, pharmaceutical companies,
regulatory agencies, and academic institutions all recognize the limitations of these costly tests in accurately predicting the arrhythmogenic
risk. Accordingly, in 2013, the U.S. Food and Drug Administration
and related organizations presented a drastic paradigm shift in cardiotoxicity testing (3–5). This proposal is based on incorporation of novel
technologies involving computational integration of multiple ion channel assays using cell models of cardiac electrophysiology and electrophysiological tests of stem cell–derived cardiomyocytes. Indeed, in
silico assessment of the arrhythmogenic risk has been widely attempted
using various types of cell or tissue models of cardiac electrophysiology.
However, most of these studies only provide surrogate markers of arrhythmia such as action potential duration, QT prolongation of
pseudo-electrocardiogram (ECG), and QT dispersion, rather than the
occurrence of clinically observed TdP (6–8). Although development
of a standardized method is scheduled for 2016, herein we present
the results of a novel assay system using our multiscale, multiphysics
heart simulation technology coupled with multi-ion current assays developed on expressed human ion channels (Fig. 1).
Our heart simulator, UT-Heart, is a finite element method (FEM)–
based human heart model that can reproduce all fundamental activities of the working heart, including propagation of excitation,
1
Graduate School of Frontier Sciences, The University of Tokyo, 5-1-5 Kashiwanoha,
Kashiwa, Chiba 277-8563, Japan. 2UT-Heart Inc., 3-25-8 Nozawa, Setagaya-ku, Tokyo
154-0003, Japan. 3Global CV Assessment, Eisai Co. Ltd., Tokodai 5-1-3, Tsukua-shi,
Ibaraki 300-2635, Japan. 4Department of Bio-informational Pharmacology, Medical
Research Institute, Tokyo Medical and Dental University, Tokyo 113-8510, Japan.
*Corresponding author. E-mail: [email protected]
Okada et al. Sci. Adv. 2015;1:e1400142
1 May 2015
contraction and relaxation, and generation of blood pressure and flow,
based on the molecular mechanisms of excitation-contraction coupling (9–13). For the detailed analysis of ECG, we coupled the heart
model with the FEM model of the human torso and solved the bidomain equations with a novel solving technique (14, 15). This feature of
the system may potentially replace the thorough QT assay beyond the
limits of cell models of electrophysiology.
As benchmark test chemicals, we selected the following 12 drugs
recognized to have varying degrees of TdP risk: quinidine, D,L-sotalol,
amiodarone, E-4031, cisapride, astemizole, terfenadine, bepridil,
moxifloxacin, verapamil, dofetilide, and ranolazine (7, 16, 17). Pharmacological characterization of these drugs was made by measuring
their inhibitory effects on the following six ion currents/channels
expressed in Chinese hamster ovary (CHO) cells over a wide range
of concentrations: rapid delayed rectifier potassium current (IKr/
hERG), sodium current (INa/Nav1.5), L-type calcium current (ICa/
Cav1.2), slow delayed rectifier current (IKs/KCNQ1 + KCNE1), inward
rectifier potassium current (IK1/Kir2.1), and transient outward current
(Ito/Kv4.3 + KChIP2). The effect of the drugs under various plasma
concentrations was reproduced by changing the parameters of each
channel model in the heart simulator according to the obtained
dose-inhibition relationship. The list of abbreviation is shown in
table S2.
RESULTS
Characterization of drugs
The results of the in vitro evaluation of the inhibitory effects of 12 drugs
on six ion currents tested are summarized in Fig. 2 (see also table S1
for detail). Median inhibitory concentration (IC50) values normalized
by the free effective therapeutic plasma concentration (ETPCunbound)
varied considerably among the drugs; for hERG channels, the hERG
IC50/ETPCunbound ranged from 0.197 (quinidine) to 975 (amiodarone).
In the following in silico drug testing, we simulated cases in which the
heart is exposed to each drug at concentrations 0 (control), 1, 3, 5, 10,
30, 50, 100, 300, 500, and 1000 times of ETPCunbound. If the arrhythmia
event occurred, we also tested the concentration between these levels
to narrow down the arrhythmia threshold.
1 of 7
Downloaded from http://advances.sciencemag.org/ on October 24, 2016
To save time and cost for drug discovery, a paradigm shift in cardiotoxicity testing is required. We introduce a novel
screening system for drug-induced arrhythmogenic risk that combines in vitro pharmacological assays and a multiscale heart simulator. For 12 drugs reported to have varying cardiotoxicity risks, dose-inhibition curves were
determined for six ion channels using automated patch clamp systems. By manipulating the channel models implemented in a heart simulator consisting of more than 20 million myocyte models, we simulated a standard
electrocardiogram (ECG) under various doses of drugs. When the drug concentrations were increased from therapeutic levels, each drug induced a concentration-dependent characteristic type of ventricular arrhythmia, whereas
no arrhythmias were observed at any dose with drugs known to be safe. We have shown that our system
combining in vitro and in silico technologies can predict drug-induced arrhythmogenic risk reliably and efficiently.
RESEARCH ARTICLE
cells modeled) included in the heart simulator. Simulations were performed by coupling the heart model with the torso at either 1 Hz
[60 beats per minute (bpm)] or 2/3 Hz (40 bpm) to reflect the standard
12-lead ECG used for the thorough QT test (Fig. 1, movie S1, and fig.
S1). The simulation results at 1.2 Hz (72 bpm) are also available in fig.
S2. Cisapride is a drug used for the treatment of dyspepsia and was
withdrawn from the market because of the risk of TdP (16). When this
drug was applied to the model, the QT interval was prolonged in a dosedependent manner, and TdP was observed at 30 times of ETPCunbound
[Fig. 3A, heart rate (HR) = 60 bpm (left) and 40 bpm (right)]. By contrast, verapamil, which is recognized as a safe drug with no published
reports of TdP in humans (16), exhibited a similar, dose-dependent
QT interval prolongation, but never induced TdP even at 50 times
ETPCunbound [Fig. 3B, HR = 60 bpm (left) and 40 bpm (right)]. Above
this dose (100 times ETPCunbound), the membrane excitation did not
propagate owing to the inhibition of Na current. The arrhythmogenic
thresholds of the 12 drugs at both HRs are shown in Fig. 3C (the ECG
responses of the 12 drugs are shown in fig. S2). We did not observe TdP
even at higher doses for amiodarone, moxifloxacin, or verapamil.
We evaluated the predictability of our simulation by comparing the
obtained TdP threshold values with the actual risk reported in the literature. For this purpose, we used the risk categories by Redfern and
Fig. 1. Diagram of the assay system. Dose-inhibition curves of drugs
were determined for six ion channels from in vitro experiments using Lawrence (7, 16, 17) with three modifications. First, E-4031, which is
the automated patch clamp system. On the basis of these data, conditions not included in these studies, was categorized as 1 because it is a pure
of the in silico heart model under specific drug concentrations were simu- hERG blocker developed as an anti-arrhythmic drug, although a prolated by modulating the channel parameters in cell models of cardiac arrhythmic effect has also been reported (movie S2 and fig. S1B) (18).
electrophysiology to yield the 12-lead ECG.
Second, ranolazine, which was also not included in these studies, was
categorized as 5 according to the literature
(19). Third, we recategorized amiodarone
as 4 (drugs for which there have been
isolated reports of TdP in humans) because
the incidence of TdP is quite low for this
drug in clinical trials (20). Further, a
meta-analysis on 6500 patients revealed that
the prophylactic use of amiodarone reduces
the risk of arrhythmia/sudden cardiac death
in high-risk patients with recent myocardial infarction or congestive heart failure (21).
Amiodarone was excluded from the list of
category 1 drugs by Lawrence et al. (17),
whereas in the analysis by Mirams et al.
(7), amiodarone was the only drug that
resulted in a large prediction error (false
negative). Finally, although we are also aware
of recent reports of arrhythmogenicity of
moxifloxacin, we used the original categorization by Redfern and Lawrence in this study
(16, 17). The threshold values relative to
ETPCunbound are shown for each drug in
each category in Fig. 3C. The threshold
values did not differ appreciably between
the three HR, although TdP was induced
at lower threshold values by some drugs, reFig. 2. Dose-inhibition relation of drugs on ion currents. Left column (top to bottom): amiodarone, producing the reported arrhythmogenic
astemizole, bepridil, cisapride, dofetilide, and D-sotalol. Right column (top to bottom): E-4031, moxifloxacin, risk of bradycardia by prolongation of venquinidine, ranolazine, terfenadine, and verapamil. In each graph, relative activities of ion currents are tricular repolarization (1). Although the
plotted as a function of drug concentration in logarithmic scale. Black line, INa; black dotted line, INaL; number of drugs tested was small, all drugs
red line, ICa; dark blue line, IKr; green line, IKs; light blue line, IK1; orange line, Ito; green circle, ETPCunbound. in categories 1 and 2 have relatively low
In silico preclinical analysis of cardiotoxicity
Drug effects at each specific dose were simulated by introducing the
current inhibition of six channels to all the cell models (22,750,008
1 May 2015
2 of 7
Downloaded from http://advances.sciencemag.org/ on October 24, 2016
Okada et al. Sci. Adv. 2015;1:e1400142
RESEARCH ARTICLE
Okada et al. Sci. Adv. 2015;1:e1400142
1 May 2015
3 of 7
Downloaded from http://advances.sciencemag.org/ on October 24, 2016
and expensive. Accordingly, we examined
how the reduction in the number of analyzed channels affects the predictability for
TdP risk. We repeated the simulations
using the results for four channels (IKr/
hERG, INa/Nav1.5, ICa/Cav1.2, and IKs/
KCNQ1 + KCNE1) and three channels
(IKr/hERG, INa/Nav1.5, and ICa/Cav1.2),
and compared the results with those using
six channels. In Fig. 5B, the obtained
threshold values were compared among
the simulations using a different number
of channels. We found that a reduction in
the number of channels from six to four
did not alter the TdP threshold for any
of the drugs tested. However, a further reduction to three channels disabled the detection of TdP for quinidine (Fig. 5A).
Thus, at least for the drugs tested in this
study, assays on four channels are necessary.
Next, we assessed the efficacy of the six
channels assayed in this study for accurate
risk assessment. We tested the TdP risk of
hypothetical multichannel blockers created by adding 50% inhibition of the five
Fig. 3. In silico risk assessment of drug-induced arrhythmogenesis. (A and B) Two-lead ECGs with ion currents not investigated in this analvarying doses of drugs (control, 1×, 10×, 30×, 50× of ETPCunbound) are shown for cisapride (A) and verapysis one by one to the pharmacological
amil (B) at HR = 40 bpm (left column) and HR = 60 bpm (right column). For both drugs, a dose-dependent
characteristics of verapamil under 50 times
prolongation of QT intervals was observed, although TdP was only observed for cisapride. (C) Threshold
values relative to ETPCunbound are shown for each drug categorized as high risk (1, 2, and 3) or safe (4 and ETPCunbound. We also examined the effect
of IK1 inhibition because none of the drugs
5). TdP was not observed for drugs in categories 4 and 5. ND, not detected.
tested in this study displayed an inhibitory
action on this channel. We found that the
threshold values. Most importantly, TdP was not observed with drugs addition of IK1 inhibition introduced a minor rhythm disturbance
in categories 4 and 5. Thus, our analysis produced no false positives. (fig. S3), confirming the significance of assaying this channel. Together,
for high-throughput screening of TdP risk, combinations of four to six
Comparison with other biomarkers for TdP risk
of the channels selected in this study are recommended, depending on
Because the multiscale simulation provides us with various parameters the affordability and time. Of course, inclusion of other ion currents
from the molecular to the organ level simultaneously, we compared the mediated by sodium/potassium pumps, calcium pumps, and sodiumpredictability of the present method (Fig. 4A) with other biomarkers in- calcium exchangers would potentiate the power of predictability.
cluding hERG IC50 (Fig. 4B), corresponding to the hERG test, and QT
interval prolongation (Fig. 4C), corresponding to the thorough QT test, Importance of realistic, three-dimensional simulation
both recommended in the current guidelines. We also examined the In the analysis presented above, inclusion of IKs inhibition (switch
MICE model that compares the inhibitory effects of a drug on the major from three-channel effects to four-channel effects, Fig. 5) caused
depolarizing current ICa/Cav1.2 and repolarizing current IKr/hERG dur- TdP by quinidine at as low as 7.5 times ETPCunbound, although the
ing the action potential plateau (22) [Fig. 4D, evaluated as −log(Cav1.2 inhibitory effect at this concentration (about 30% of IKs IC50) was at
IC50) + log(hERG IC50)], as well as the transmural dispersion of re- most 20%. Of course, the interplay with other ion currents plays a sigpolarization (TDR) (Fig. 4E). TDR was measured by the difference be- nificant role. However, we hypothesized that the inhibition of the IKs
tween the maximum and minimum action potential durations under channel could amplify the TDR originating from the heterogeneous
maximum tolerable concentration without arrhythmia. Simulations were expression of this channel species to increase the TdP propensity
performed at HR = 60 bpm. In clear contrast to the present method, the (23). To test this hypothesis, we repeated the simulation with the
hERG IC50, QT interval prolongation, the MICE model (22), and TDR heart model in which the whole ventricles consisted of either epicarcould not discriminate the low-risk drugs (categories 4 and 5: blue rect- dial, midmyocardial (M), or endocardial cells, as well as with the
angle) from the high-risk drugs (categories 1 to 3: red rectangle).
two-layered model consisting of epicardial and endocardial cells.
As shown in Fig. 6, TdP was not observed in those models with nonSensitivity analysis
physiological cell distributions, whereas with homogeneous distributions,
Although the automated patch clamp technique coupled with the in especially in the model with M cells, sustained depolarization was obvitro ion channel expression system has enabled the high-throughput served. Sustained depolarization is caused by early after depolarization,
screening of drug toxicity, multiple channel assays are still time-consuming which is a sign of arrhythmia. However, its appearance at ETPCunbound
RESEARCH ARTICLE
Okada et al. Sci. Adv. 2015;1:e1400142
1 May 2015
4 of 7
Downloaded from http://advances.sciencemag.org/ on October 24, 2016
(8, 25, 26). To our knowledge, however, this is the first
report of in silico TdP risk assessment based on a realistic
heart model producing a realistic standard ECG.
Although we could successfully identify safe drugs
from a limited data set, there are multiple problems to
be solved for accurate quantitative risk assessment. First,
in this study, we only used the conduction block model
based on the steady-state inhibitory effect of drugs, and
ignored their state- or voltage-dependent binding and
unbinding. Indeed, as discussed by Mirams et al. (7),
the time scale of binding and unbinding is much longer
than the time scale of channel transition. The impact of
this phenomenon should be examined in future studies,
as well as the effect of changes of the detailed kinetic
parameters in the gating processes. Second, although
this simulation model is based on the detailed mechanisms of cardiac electrophysiology, it cannot reproduce
Fig. 4. Comparison of the predictability
the effect on channels not evaluated experimentally.
of currently proposed biomarkers for
arrhythmogenic risk for drugs in each To address this problem of unknown drug effects, we
category. (A) Threshold (Th) for TdP (current simulated the effect of a hypothetical new drug by addmethod). For drugs not inducing TdP, an ing the inhibitory action of other channels to the phar○ value is assigned at HR = 60. (B) hERG in- macological characteristics of verapamil. These studies
hibition quantified by hERG IC50/ETPCunbound. showed that other channels may have only a minor ef(C) Prolongation of QT interval at the ETPC. fect on the arrhythmogenesis.
(D) MICE model evaluated as −log(Cav1.2 IC50) +
Although the four to six channels assayed in this
log(hERG IC50). D,L-Sotalol was not evalu- study were required for screening purposes, our conated for the MICE model because Cav1.2 IC50 could not be identified. (E) TDR measured
clusions remain speculative given the small number of
by the difference between maximum and minimum action potential durations under
analyzed compounds. Thus, further studies are required,
maximum tolerable concentration without arrhythmia. Concentrations are 100 (amiodarone),
100 (astemizole), 5 (bepridil), 10 (cisapride), 30 (dofetilide), 50 (D,L-sotalol), 5 (E-4031), particularly when considering the growing diversity in
100 (moxifloxacin), 1 (quinidine), 20 (ranolazine), 10 (terfenadine), and 50 (verapamil) times new drug targets. Finally, we simulated only cases of fixed
ETPC. Except for the present method, values of biomarkers were distributed evenly among regular HR in this study. Because TdP is often precipitated by the irregular short-long-short ventricular cycle,
the categories.
we simulated such a sequence under the influence of cisapride, and found that TdP could be induced at a slightly
does not recapitulate the real risk of this drug. Together, these data suggest lower drug concentration compared with the regular rhythm (fig. S6).
that only the three-dimensional heart simulation with physiological cell Such a pacing protocol is also clinically important because it
distribution enables the accurate prediction of TdP risk, whereas risk corresponds to the provocation test in electrophysiology laboratories.
assessment based on only the cell model of electrophysiology may be Nevertheless, comprehensive analyses of the effects of basic cycle length
and pacing patterns needing a large amount of computation are remisleading.
quired in future studies.
Although the incidence of TdP is extremely low after drug marketing, its occurrence is often precipitated by confounding factors
DISCUSSION
including hypokalemia, liver dysfunction, female gender, preexisting
Various methods have been proposed for the accurate assessment of heart diseases, and genetic polymorphism in potassium channels
the TdP risk of drugs (7, 8, 16, 17, 22, 24). In particular, in silico studies modifying functions or sensitivity to drugs (1, 27, 28). As an examare expected to facilitate drug development by saving on the time and ple, we simulated a case of a common polymorphism with slight alcost inherent to animal experiments. However, most in silico studies terations of current (reduced by 25% and increased by 25%), but
examining the drug effect on multiple ion channels only evaluate with retention of the sensitivity to cisapride similar to the wild-type
surrogate markers including the relative inhibitory actions on the major channel (28). This apparently silent change in channel function deion currents dominating the repolarization phase of the myocardium creased the threshold value for TdP and thus increased arrhythmia
(22), the hERG IC50 relative to ETPCunbound (16), or the action propensity (fig. S4). Concomitant use of other drugs may also modpotential duration of the cell model of cardiac electrophysiology (7), ify the arrhythmogenic risk. We tested the combination of amiodarrather than directly assessing the TdP occurrence. The use of whole one and clarithromycin assuming a synergistic (multiplicative)
heart experiments allows direct assessment of TdP risk but requires interaction between them, and found that the concomitant use of
repeated measurement to compensate for biological variability, re- macrolide antibiotics induces TdP with this safe drug (fig. S5), as
quiring considerable time and cost. To overcome these problems, re- previously reported (29). With further improvement, the simulation
searchers have attempted multiscale heart simulations that can directly power of our model will enable us to evaluate the effects of other
assess the TdP risk based on cellular and molecular electrophysiology factors.
RESEARCH ARTICLE
Okada et al. Sci. Adv. 2015;1:e1400142
1 May 2015
5 of 7
Downloaded from http://advances.sciencemag.org/ on October 24, 2016
using the Sophion QPatch HTX system
and software (QPatch Assay Software 5.0,
Biolin Scientific) using the following cell
lines, assay buffers, and voltage protocols.
Cell lines. hERG (KCNH2 gene) channels were stably expressed in CHO-K1 cells.
Nav1.5 (SCN5A gene, hNav1.5 channel)
channels were stably expressed in CHO-K1 cells.
Cav1.2 (CACNA1C/CACNB2/CACNA2D1
genes) channels were stably expressed
in CHO cells. IKs (KCNQ1/KCNE1 genes)
channels were stably expressed in CHO-K1
cells. IK1 (KCNJ2 gene) channels were stably
expressed in CHO-K1 cells. Ito (KCND3/
KChIP2 genes) channels were stably expressed in CHO-K1 cells. hERG, Nav1.5,
IKs, and IK1 cell lines were established in
Eisai. The Ito cell line was established by
the Tokyo Medical and Dental University.
The Cav1.2 cell line was purchased from
ChanTest Corporation.
Assay buffers. For IKr, IK1, and Ito:
extracellular solution and internal solution containing 145 mM NaCl, 4 mM
Fig. 5. Effect of the number of channels assayed on risk prediction. (A) Drug effect of quinidine on KCl, 2 mM CaCl2, 1 mM MgCl2, 10 mM
ECG simulated using the data on six (left), four (middle), and three (right) channels (ch). When using the Hepes, 10 mM glucose, pH adjusted to 7.4
data on only three channels, arrhythmogenicity was not reproduced. (B) Relative threshold values for TdP with NaOH, and 120 mM KCl, 5 mM
of each drug examined using the data on three (red), four (blue), and six (black) channels. A difference was CaCl , 1.8 mM MgCl , 10 mM KOH/
2
2
detected in quinidine.
EGTA, 4 mM Na2-ATP (adenosine triphosphate), 10 mM Hepes, pH adjusted
to 7.2 with KOH. For INa: extracellular solution and internal solution
containing 145 mM NaCl, 4 mM KCl, 2 mM CaCl2, 1 mM MgCl2,
10 mM Hepes, 10 mM glucose, pH adjusted to 7.4 with NaOH, and
135 mM CsF, 10 mM NaCl, 1 mM EGTA, 10 mM Hepes, pH adjusted
to 7.3 with CsOH. For ICa: extracellular solution and internal solution
containing 137 mM NaCl, 4 mM KCl, 1.8 mM CaCl2, 1 mM MgCl2,
10 mM Hepes, 10 mM glucose, pH adjusted to 7.4 with NaOH, and
80 mM L-aspartic acid, 130 mM CsOH, 5 mM MgCl2, 4 mM Na2ATP, 0.1 mM (tris)guanosine triphosphate, 5 mM EGTA, 10 mM
Hepes, pH adjusted to 7.2 with L-aspartic acid. For IKs: extracellular solution and internal solution containing 145 mM NaCl, 4 mM KCl, 2 mM
CaCl2, 1 mM MgCl2, 10 mM Hepes, 10 mM glucose, pH adjusted to
Fig. 6. Significance of the three-dimensional heart model with realistic tissue structure for the accurate prediction of arrhythmogenic risk. 7.4 with NaOH, and 110 mM K-gluconate, 20 mM KCl, 2 mM MgCl2,
Effects of varying doses of quinidine up to 10 times of ETPCunbound on ECG 5 mM EGTA, 4 mM Na2-ATP, 10 mM Hepes, pH adjusted to 7.2
were evaluated using the model with three types of ventricular myocytes with KOH.
Voltage protocols. IKr was induced by the application of depolar(endocardial, M, and epicardial cells) arranged transmurally (physiological),
endocardial cell only (endo), M cell only (M), epicardial cell only (epi), and izing step pulses to +20 mV with 2000-ms duration, followed by step
two-layered model with epicardial and endocardial cells (epi, endo). Ar- pulses to −50 mV with 2000-ms duration from a holding potential of
rhythmia was observed only in the physiological model, suggesting the −80 mV every 15 s. The inhibition ratio of the tail current at −50 mV
importance of three-dimensional heart simulation with realistic tissue struc- was calculated as hERG inhibition. INa was induced by the application of
ture in which inhibition of IKs augmented the TDR.
30 repetitive depolarizing pulses to −10 mV with 50-ms duration from
a holding potential of −90 mV at 2 Hz. The inhibition ratio of peak
inward current at the 30th pulse was calculated as INa inhibition. For
MATERIALS AND METHODS
ranolazine, the inhibitory effect on the late component of INa was
also included using the previously reported parameters (30, 31).
In vitro current assay
Ion currents were recorded in CHO or Chinese hamster lung (CHL) ICa was induced by the application of depolarizing step pulses to 0 mV
cells expressing either hERG (IKr), Nav1.5 (INa), Cav1.2/b2/a2-d (ICa), with 150-ms duration from a holding potential of −40 mV every 5 s. The
KCNQ1 + KCNE1 (IKs), Kir2.1 (IK1), or Kv4.3 + KChIP2 (Ito) channels inhibition ratio of peak inward current was calculated as ICa inhibition,
RESEARCH ARTICLE
Relative current ¼
1
1 þ 10ðlogIC50 logxÞ∙h
Nonlinear least square fits were solved by GraphPad Prism version
6.02 (GraphPad Software Inc.). Data are expressed as means ± SEM.
Electrophysiology simulation and ECG analysis
Simulation of drug effects on human cardiac electrophysiology was performed using the healthy adult model of the UT-Heart (10, 14). Cell
models specific for the ventricular myocyte (32) and conduction system
(33) were implemented to each FEM node to reproduce the physiological distribution of different cell species (endocardial, M, and epicardial
cells) in the ventricular wall (10, 23). The propagation of excitation was
solved with the bidomain formulation (34, 35) using the parallel multilevel technique we previously developed (14). To achieve the physiological speed of propagation, a minor modification of the activation gate
parameters was necessary in the ventricular cell model (32). To save
computational time, only the ventricles were used, and the stimulus
was applied to the root of the conduction system at 1.2, 1, and 2/3
Hz to mimic the regular sinus rhythm. In most cases, we calculated
five beats, although the duration was extended when necessary. From
the body surface potential data, we obtained the 12-lead ECG
according to the standard lead locations and measured the QT intervals in limb lead II. Computation was performed with a parallel computer system [Intel Xeon E5-2670 (2.6 GHz), 254 cores] and the
computational time was 3 hours for a single beat. All the program
codes were written in-house.
SUPPLEMENTARY MATERIALS
Supplementary material for this article is available at http://advances.sciencemag.org/cgi/content/
full/1/4/e1400142/DC1
Fig. S1. Simulated 12-lead ECG.
Fig. S2. Drug effects on two-lead ECG.
Fig. S3. Effect of hypothetical multichannel blockers on ECG.
Fig. S4. Effect of cisapride for a subject carrying a common polymorphism of hERG channel.
Fig. S5. Effect of concomitant use of clarithromycin on the arrhythmogenicity of amiodarone.
Fig. S6. Effect of irregular heartbeats.
Table S1. Inhibitory actions of 12 drugs on six ion channels.
Table. S2. Abbreviations.
Okada et al. Sci. Adv. 2015;1:e1400142
1 May 2015
Movie S1. Simulation of regular heartbeat.
Movie S2. Simulation of drug-induced arrhythmia.
Reference (36)
REFERENCES AND NOTES
1. Y. G. Yap, A. J. Camm, Drug induced QT prolongation and torsades de pointes. Heart 89,
1363–1372 (2003).
2. R. R. Shah, Drugs, QTc interval prolongation and final ICH E14 guideline: An important
milestone with challenges ahead. Drug Saf. 28, 1009–1028 (2005).
3. M. A. Hamburg, Advancing regulatory science. Science 331, 987 (2011).
4. K. R. Chi, Revolution dawning in cardiotoxicity testing. Nat. Rev. Drug Discov. 12, 565–567
(2013).
5. P. T. Sager, G. Gintant, J. R. Turner, S. Pettit, N. Stockbridge, Rechanneling the cardiac
proarrhythmia safety paradigm: A meeting report from the Cardiac Safety Research Consortium. Am. Heart J. 167, 292–300 (2014).
6. S. Suzuki, S. Murakami, K. Tsujimae, I. Findlay, Y. Kurachi, In silico risk assessment for druginduction of cardiac arrhythmia. Prog. Biophys. Mol. Biol. 98, 52–60 (2008).
7. G. R. Mirams, Y. Cui, A. Sher, M. Fink, J. Cooper, B. M. Heath, N. C. McMahon, D. J. Gavaghan,
D. Noble, Simulation of multiple ion channel block provides improved early prediction of
compounds’ clinical torsadogenic risk. Cardiovasc. Res. 91, 53–61 (2011).
8. K. A. Beattie, C. Luscombe, G. Williams, J. Munoz-Muriedas, D. J. Gavaghan, Y. Cui, G. R. Mirams,
Evaluation of an in silico cardiac safety assay: Using ion channel screening data to predict QT
interval changes in the rabbit ventricular wedge. J. Pharmacol. Toxicol. Methods 68, 88–96
(2013).
9. H. Watanabe, S. Sugiura, H. Kafuku, T. Hisada, Multiphysics simulation of left ventricular
filling dynamics using fluid-structure interaction finite element method. Biophys. J. 87,
2074–2085 (2004).
10. J. Okada, T. Washio, A. Maehara, S. Momomura, S. Sugiura, T. Hisada, Transmural and apicobasal gradients in repolarization contribute to T-wave genesis in human surface ECG.
Am. J. Physiol. Heart Circ. Physiol. 301, H200–H208 (2011).
11. T. Washio, J. Okada, S. Sugiura, T. Hisada, Approximation for cooperative interactions of a
spatially-detailed cardiac sarcomere model. Cell. Mol. Bioeng. 5, 113–126 (2011).
12. S. Sugiura, T. Washio, A. Hatano, J. Okada, H. Watanabe, T. Hisada, Multi-scale simulations
of cardiac electrophysiology and mechanics using the University of Tokyo heart simulator.
Prog. Biophys. Mol. Biol. 110, 380–389 (2012).
13. T. Washio, J. Okada, A. Takahashi, K. Yoneda, Y. Kadooka, S. Sugiura, T. Hisada, Multiscale
heart simulation with cooperative stochastic cross-bridge dynamics and cellular structures.
SIAM J. Multiscale Model Simul. 11, 965–999 (2013).
14. T. Washio, J. Okada, T. Hisada, A parallel multilevel technique for solving the bidomain
equation on a human heart with Purkinje fibers and a torso model. SIAM Rev. 52, 717–743
(2010).
15. J. Okada, T. Sasaki, T. Washio, H. Yamashita, T. Kariya, Y. Imai, M. Nakagawa, Y. Kadooka,
R. Nagai, T. Hisada, S. Sugiura, Patient specific simulation of body surface ECG using the
finite element method. PACE 36, 309–321 (2013).
16. W. S. Redfern, L. Carlsson, A. S. Davis, W. G. Lynch, I. MacKenzie, S. Palethorpe, P. K. Siegl,
I. Strang, A. T. Sullivan, R. Wallis, A. J. Camm, T. G. Hammond, Relationships between
preclinical cardiac electrophysiology, clinical QT interval prolongation and torsade de
pointes for a broad range of drugs: Evidence for a provisional safety margin in drug
development. Cardiovasc. Res. 58, 32–45 (2003).
17. C. L. Lawrence, M. H. Bridgland-Taylor, C. E. Pollard, T. G. Hammond, J.-P. Valentin, A rabbit
Langendorff heart proarrhythmia model: Predictive value for clinical identification of Torsades
de Pointes. Br. J. Pharmacol. 149, 845–860 (2006).
18. L. Buchanan, G. Kabell, M. N. Brunden, J. K. Gibson, Comparative assessment of ibutilide,
D-sotalol, clofilium, E-4031, and UK-68,798 in a rabbit model of proarrhythmia. J. Cardiovasc.
Pharmacol. 220, 540–549 (1993).
19. P. T. Sager, Key clinical considerations for demonstrating the utility of preclinical models to
predict clinical drug-induced torsades de pointes. Br. J. Pharmacol. 154, 1544–1549 (2008).
20. B. Darpo, Spectrum of drugs prolonging QT interval and the incidence of torsades de
pointes. Eur. Heart J. Suppl. 3, K70–K80 (2001).
21. A. T. M.-A. Investigators, Effect of prophylactic amiodarone on mortality after acute myocardial infarction and in congestive heart failure: Meta-analysis of individual data from
6500 patients in randomised trials. Lancet 350, 1417–1424 (1997).
22. J. Kramer, C. A. Obejero-Paz, G. Myatt, Y. A. Kuryshev, A. Bruening-Wright, J. S. Verducci,
A. M. Brown, MICE models: Superior to the HERG model in predicting Torsade de Pointes.
Sci. Rep. 3, 2100 (2013).
23. C. Antzelevitch, Cellular basis for the repolarization waves of the ECG. Ann. N. Y. Acad. Sci.
1080, 268–281 (2006).
6 of 7
Downloaded from http://advances.sciencemag.org/ on October 24, 2016
except for ranolazine in which the late component was evaluated (30). IKs
was induced by the application of depolarizing step pulses to +40 mV
with 2000-ms duration, followed by step pulses to −50 mV with 500-ms
duration from a holding potential of −80 mV every 5 s. The inhibition ratio
of tail current at −50 mV was calculated as IKs inhibition. IK1 was induced
by the application of hyperpolarizing step pulses to −120 mV with 500-ms
duration, followed by ramp pulses to +40 mV with 2500-ms duration from
a holding potential of 0 mV every 15 s. The inhibition ratio of current amplitude at steady phase at −120 mV was calculated as IK1 inhibition. Ito was
induced by the application of depolarized potential to +20 mV with 500-ms
duration from a holding potential of −80 mV every 15 s. The inhibition
ratio of current amplitude at 100 ms of depolarizing step pulses was
calculated as Ito inhibition.
Drug concentrations were varied up to 200 times of ETPCunbound.
Drug effects were analyzed for each channel by normalizing the current
by its maximum value obtained in the absence of the drug and fit to the
Hill equation, where x and h represent the drug concentration and Hill
constant, respectively.
RESEARCH ARTICLE
Okada et al. Sci. Adv. 2015;1:e1400142
1 May 2015
35. B. Roth, How the anisotropy of the intracellular and extracellular conductivities influences
stimulation of cardiac muscle. J. Math. Biol. 30, 633–646 (1992).
36. P. Gluais, M. Bastide, D. Grandmougin, G. Fayad, M. Adamantidis, Clarithromycin reduces
Isus and Ito currents in human atrial myocytes with minor repercussions on action potential
duration. Fundam. Clin. Pharmacol. 17, 691–701 (2003).
Funding: This research was supported by the Japan Society for the Promotion of Science
(JSPS) through its Funding Program for World-Leading Innovative R&D on Science and Technology (FIRST Program). Author contributions: J.O. wrote program codes, carried out computations, drafted the manuscript, and participated in the design of the study. T.Y. and J.K.
carried out experiments and participated in the design of the study. T.W. wrote
program codes and carried out computations. T.F. and K.S. conceived the study and participated in its design. S.S. participated in the study design and helped to draft the manuscript. T.H. helped to write the program codes and participated in the design and
coordination of the study. Competing interests: The authors declare that they have no
competing interests. Data and materials availability: The following cell lines were obtained
under a material transfer agreement: Ito (CHO-K1) cell line stably overexpressing the KCND3/
KChIP2 genes (J.K.); hERG cell line (CHO-K1) stably expressing the KCNH2 gene, Nav1.5 cell line
(CHL) stably expressing the SCN5A gene, Iks cell line (CHO-K1) stably expressing the KCNQ1/
KCNE1 genes, and Iks cell line (CHO-K1) stably expressing the KCNJ2 gene (K.S. and T.Y.).
Submitted 25 November 2014
Accepted 4 April 2015
Published 1 May 2015
10.1126/sciadv.1400142
Citation: J. Okada, T. Yoshinaga, J. Kurokawa, T. Washio, T. Furukawa, K. Sawada, S. Sugiura,
T. Hisada, Screening system for drug-induced arrhythmogenic risk combining a patch clamp
and heart simulator. Sci. Adv. 1, e1400142 (2015).
7 of 7
Downloaded from http://advances.sciencemag.org/ on October 24, 2016
24. D. Bottino, R. C. Penland, A. Stamps, M. Traebert, B. Dumotier, A. Georgiva, G. Helmlinger,
G. S. Lett, Preclinical cardiac safety assessment of pharmacological compounds using an
integrated system-based computer model of the heart. Prog. Biophys. Mol. Biol. 90, 413–443
(2006).
25. G. R. Mirams, M. R. Davies, Y. Cui, P. Kohl, D. Noble, Application of cardiac electrophysiology simulations to pro-arrhythmic safety testing. Br. J. Pharmacol. 167, 932–945
(2012).
26. N. Zemzemi, M. O. Bernabeu, J. Saiz, J. Cooper, P. Pathmanathan, G. R. Mirams, J. Pitt-Francis,
B. Rodriguez, Computational assessment of drug-induced effects on the electrocardiogram:
From ion channel to body surface potentials. Br. J. Pharmacol. 168, 718–733 (2013).
27. F. Z. He, H. L. McLeod, W. Zhang, Current pharmacogenomic studies on hERG potassium
channels. Trends Mol. Med. 19, 227–238 (2013).
28. B. D. Anson, M. J. Ackerman, D. J. Tester, M. L. Will, B. P. Delisle, C. L. Anderson, C. T. January,
Molecular and functional characterization of common polymorphisms in HERG (KCNH2)
potassium channels. Am. J. Physiol. 286, H2434–H2441 (2004).
29. D. Justo, D. Zeltser, Torsades de pointes induced by antibiotics. Eur. J. Intern. Med. 17, 254–259
(2006).
30. C. Antzelevitch, L. Belardinelli, A. C. Zygmunt, A. Burashnikov, J. M. Di Diego, J. M. Fish,
J. M. Cordeiro, G. Thomas, Electrophysiological effects of ranolazine, a novel antianginal agent with antiarrhythmic properties. Circulation 110, 904–910 (2004).
31. M. Chevalier, B. Amuzescu, V. Gawali, H. Todt, T. Knott, O. Scheel, H. Abriel, Late cardiac
sodium current can be assessed using automated patch-clamp. F1000Res. 3, 245 (2014).
32. T. O’Hara, L. Virag, A. Varro, Y. Rudy, Simulation of the undiseased human cardiac ventricular action potential: Model formulation and experimental validation. PLOS Comput. Biol. 7,
e1002061 (2011).
33. P. Stewart, O. V. Aslanidi, D. Noble, P. J. Noble, M. R. Boyett, H. Zhang, Mathematical
models of the electrical action potential of Purkinje fibre cells. Philos. Trans. A Math. Phys.
Eng. Sci. 367, 2225–2255 (2009).
34. C. S. Henriquez, Simulating the electrical behavior of cardiac tissue using the bidomain
model. Crit. Rev. Biomed. Eng. 21, 1–77 (1993).
Screening system for drug-induced arrhythmogenic risk
combining a patch clamp and heart simulator
Jun-ichi Okada, Takashi Yoshinaga, Junko Kurokawa, Takumi
Washio, Tetsushi Furukawa, Kohei Sawada, Seiryo Sugiura and
Toshiaki Hisada (May 1, 2015)
Sci Adv 2015, 1:.
doi: 10.1126/sciadv.1400142
This article is publisher under a Creative Commons license. The specific license under which
this article is published is noted on the first page.
For articles published under CC BY-NC licenses, you may distribute, adapt, or reuse the article
for non-commerical purposes. Commercial use requires prior permission from the American
Association for the Advancement of Science (AAAS). You may request permission by clicking
here.
The following resources related to this article are available online at
http://advances.sciencemag.org. (This information is current as of October 24, 2016):
Updated information and services, including high-resolution figures, can be found in the
online version of this article at:
http://advances.sciencemag.org/content/1/4/e1400142.full
Supporting Online Material can be found at:
http://advances.sciencemag.org/content/suppl/2015/04/27/1.4.e1400142.DC1
This article cites 36 articles, 8 of which you can access for free at:
http://advances.sciencemag.org/content/1/4/e1400142#BIBL
Science Advances (ISSN 2375-2548) publishes new articles weekly. The journal is
published by the American Association for the Advancement of Science (AAAS), 1200 New
York Avenue NW, Washington, DC 20005. Copyright is held by the Authors unless stated
otherwise. AAAS is the exclusive licensee. The title Science Advances is a registered
trademark of AAAS
Downloaded from http://advances.sciencemag.org/ on October 24, 2016
For articles published under CC BY licenses, you may freely distribute, adapt, or reuse the
article, including for commercial purposes, provided you give proper attribution.