Testing the nonclinical Comprehensive In Vitro Proarrhythmia Assay (CiPA) paradigm with an established anti‐seizure medication: Levetiracetam case study

Abstract Levetiracetam (LEV), a well‐established anti‐seizure medication (ASM), was launched before the original ICH S7B nonclinical guidance assessing QT prolongation potential and the introduction of the Comprehensive In Vitro Proarrhythmia Assay (CiPA) paradigm. No information was available on its effects on cardiac channels. The goal of this work was to “pressure test” the CiPA approach with LEV and check the concordance of nonclinical core and follow‐up S7B assays with clinical and post‐marketing data. The following experiments were conducted with LEV (0.25–7.5 mM): patch clamp assays on hERG (acute or trafficking effects), NaV1.5, CaV1.2, Kir2.1, KV7.1/mink, KV1.5, KV4.3, and HCN4; in silico electrophysiology modeling (Virtual Assay® software) in control, large‐variability, and high‐risk human ventricular cell populations; electrophysiology measurements in human induced pluripotent stem cell (hiPSC)‐derived cardiomyocytes and dog Purkinje fibers; ECG measurements in conscious telemetered dogs after single oral administration (150, 300, and 600 mg/kg). Except a slight inhibition (<10%) of hERG and KV7.1/mink at 7.5 mM, that is, 30‐fold the free therapeutic plasma concentration (FTPC) at 1500 mg, LEV did not affect any other cardiac channels or hERG trafficking. In both virtual and real human cardiomyocytes, and in dog Purkinje fibers, LEV induced no relevant changes in electrophysiological parameters or arrhythmia. No QTc prolongation was noted up to 2.7 mM unbound plasma levels in conscious dogs, corresponding to 10‐fold the FTPC. Nonclinical assessment integrating CiPA assays shows the absence of QT prolongation and proarrhythmic risk of LEV up to at least 10‐fold the FTPC and the good concordance with clinical and postmarketing data, although this does not exclude very rare occurrence of QT prolongation cases in patients with underlying risk factors.


first published in 2005 and
recently revised through a Q&A process, 2 aims at assessing QT prolongation and proarrhythmic potential of new pharmaceuticals before first-in-human dosing. The two core assays of this guidance are an in vitro hERG assay and an in vivo QT assay, generally conducted in the dog. Follow-up in silico and in vitro proarrhythmia assays, as initially described in the Comprehensive in vitro Proarrhythmia Assay (CiPA) paradigm, 3,4 can also be integrated in the assessment ( Figure 1). Since Levetiracetam (LEV, Keppra®), an anti-seizure medication (ASM) approved in many countries as monotherapy or adjunctive treatment of seizures in pediatric and adult patients, was first launched in 2000, before the introduction of the original ICH S7B guidance and CiPA approach, nonclinical cardiovascular safety data were limited to studies in dog Purkinje fibers and in anesthetized and conscious dogs. A Thorough QT (TQT) study was also conducted in healthy volunteers at therapeutic and supratherapeutic doses, 5 in compliance with the ICH E14 clinical guidance. 6 At the time of the submission, neither available nonclinical data nor the TQT assessment revealed any risk of QT prolongation and Torsade de Pointes (TdP). However, until now, the direct effects of LEV on hERG-mediated I Kr current and other cardiac ion currents remained unknown. Therefore, the goal of this work was to "pressure test" the CiPA paradigm with a "real-life" well-established ASM such as LEV and check the concordance of the whole nonclinical dataset with clinical and postmarketing data. The three nonclinical components of CiPA were used: standardized in vitro cardiac ion channel patch clamp assays, in silico modeling of the human ventricular action potential, and electrophysiology and contractility measurements in human induced pluripotent stem cell-derived cardiomyocytes (hiPSC-CM). In addition to the CiPA and S7B-related assays, LEV was also tested on hERG trafficking, which can, if inhibited, trigger QT prolongation after chronic treatment. 7 As they were never published so far, the nonclinical in vitro and in vivo data previously collected in dogs and postmarketing data, although this does not exclude very rare occurrence of QT prolongation cases in patients with underlying risk factors.

K E Y W O R D S
cardiac safety, CiPA, ICH S7B, levetiracetam, nonclinical, QT prolongation, torsade de pointes F I G U R E 1 Schematic representation of core and follow-up assays of the ICH S7B nonclinical guidance and nonclinical components of the Comprehensive in Vitro Proarrhythmia Assays (CiPA) approach (adapted from ICH E14/S7B Q&A training material and FDA Impact Story (Impact Story: Improved Assessment of Cardiotoxic Risk in Drug Candidates: The Comprehensive in vitro Proarrhythmia Assay | FDA are presented alongside the newly generated data to establish a comprehensive integrated cardiovascular safety risk assessment of LEV.

| Cardiac ion currents in patch clamp assays
Chinese hamster ovary (CHO) cells stably transfected with hERG, Na V 1.5, Ca V 1.2 (co-expressed with β2/α2δ1), K V 1.5, K V 4.3 (coexpressed with KChiP2.2), or K ir 2.1 were cultured in Ham's medium with L-glutamine, while HEK293 cells stably expressing HCN4 or K V 7.1 (co-expressed with mink) were cultured in DMEM with L-glutamine. All cell lines were stably transfected with cDNA encoding for the human isoform of the different tested ion channels. to compare LEV effects versus vehicle. In the hERG GLP assay, samples of formulations collected from the reservoir tube after preparation, as well as samples collected from the recording chamber at the end of the experiment were analyzed for LEV concentration determination using a HPLC/UV analytical method.
The hERG trafficking assay was similar to the "acute" hERG assay, except that it was an automated patch clamp conducted in non-GLP conditions, at room temperature, and that cells were incubated with LEV for 24 to 28 h. A preliminary washout experiment was conducted prior to the trafficking experiment to check the reversibility of the direct hERG current block. For this, the top concentration of LEV (7.5 mM) was first applied on the cells, followed by at least 10 min of bath solution (without LEV) as washout. The same voltage protocol as for the acute hERG was applied before and after washout and tail current amplitude was measured, as well as in the trafficking experiment, at the end of the 24-28 h of incubation.

| In silico human adult cardiomyocyte electrophysiology modeling
In silico human adult cardiomyocyte electrophysiology modeling was performed using the software Virtual Assay® (v. 3

.2 © 2018
Oxford University Innovation Ltd.), similar to previous works demonstrating high accuracy of predictions for drug-induced proarrhythmia. [9][10][11][12] Simulations of action potential (AP) and calcium transient (CT) were performed using the ToR-ORd model, 13 which is the most updated human ventricular cardiomyocytes model available in the literature. Simulations were repeated with two older models, that is, ORd 14 and ORd2-CiPA, 15  (iii) High-risk: conductances varied as in. 9 This population has been specifically designed with a low repolarization reserve, to maximize the risk to develop drug-induced early-after-depolarizations (EADs), associated with proarrhythmic risk.
The three populations were paced at 1 Hz for 500 beats to reach a steady state, and then calibrated using experimental AP and CT biomarkers from nondiseased human hearts. 10  To verify that our results are not dependent on the random sampling of the ionic conductances, simulations were repeated twice, using two different sets of cell populations (results shown in the Table S1).
Simulations of LEV were performed at multiple concentrations (0.25, 0.75, 2.5, and 7.5 mM) using the data generated in the patch clamp assays as input of a simple pore-block model 18 for the different ion channels. Since no IC 50 was obtained due to very weak inhibitory activity of LEV on the different currents, we considered the actual % of current block observed at each tested (iii) Slow pacing (50 beats at 0.2 Hz) plus β-adrenergic stimulation (β-AS), which was simulated as a simple twofold increase in the conductances of the slow delayed rectifier potassium current (G Ks ) and the L-type calcium current (G CaL ).
The last beat in each simulation was analyzed, to compute the following biomarkers: AP duration at 40% and 90% of repolarization (APD 40 and APD 90 ); AP triangulation, defined as the difference between APD 90 and APD 40 (Tri 90-40 ); maximum upstroke velocity (dV/dt MAX ); peak voltage (V peak ); resting membrane potential (RMP); CT duration at 90% of repolarization (CTD 90 ); CT peak (CT peak ); electromechanical window (EMw), defined as the difference between APD 90 and CTD 90 . AP traces displaying a positive derivative after 150 ms were classified as EADs.
Based on previous studies, we considered the following markers of proarrhythmia: (i) occurrence of EAD in the in silico population at any tested concentration; 9 (ii) more than 10% shortening of the EMw at 10-fold FTPC. 10

| Human induced pluripotent stem cell-derived cardiomyocytes (hiPSC-CM) function
Experiments on hiPSC-CM were adapted from a protocol previously described. 12  trols. Impedance data, that is, Cell Index (CI), beat rate, and CI amplitude, as well as ECR (Extracellular Recording) data, that is, sodium spike amplitude and field potential duration (FPD) were recorded by RTCA CardioECR at baseline, 30 min, 1, 6, 12, and 24-h posttreatment. These data were normalized to baseline and vehicle for matching time points. FPD was corrected for heart rate using Fridericia's formula. Intergroup comparisons were made using first a two-way (treatment, time) ANOVA followed by an unpaired Student's t test when the ANOVA was significant (p < .05).
No statistical analysis was conducted for the sodium spike amplitude due to the high variability of this parameter.

| Electrocardiography and hemodynamics in conscious telemetered dogs
The experiments were conducted under GLP conditions, were reviewed by the institutional ethics committee, and approved by the ECG electrodes were placed in Lead II position and waveforms were recorded using the ART™ acquisition system (DataSciences Inc.

| Cardiac ion currents in patch clamp assays
In the hERG manual patch clamp assay, a slight inhibition of tail LEV did not significantly inhibit the other cardiac ion currents, except K V 7.1/mink-mediated current which was reduced by 8% versus vehicle control at 7.5 mM (p < .05; Table 1). All positive controls exhibited their expected effects.
In the hERG trafficking assay, all LEV concentrations produced <5% inhibition of tail current after 24-h incubation, indicating the absence of effects on trafficking ( Figure 2B).

| In silico human adult cardiomyocyte electrophysiology modeling
Based on the patch clamp assays, we scaled the following ion channel conductances in the in silico human cardiomyocyte models to repro- for the Na + -Ca 2+ exchanger and L-Type Ca 2+ current (Figure 4), in agreement with previous findings. 9,10 Results were found to be consistent across different populations and several human action potential models (Table S1).

| Action potential recordings in dog Purkinje fibers
At stimulation frequencies of 1 and 0.5 Hz, LEV induced no statistically or biologically (i.e., >10%) significant changes in any of the AP parameters compared to vehicle control (Table 2). When the frequency was increased from 1 to 3 Hz at the top concentration (5.9 mM), there was no significant difference between the vehicle and LEV in MRD. Sotalol induced its expected effects, that is, significant increases in APD 60 (45%) and APD 90 (44%) at 1 Hz, reaching 58-59% increase at 0.5 Hz, without any effect on other AP parameters.  Results are shown for all tested concentrations, from control (no drug, 0x) to 30x the free therapeutic plasma concentration (FTPC). In each boxplot, the central mark is the median of the population, box limits are the 25th and 75th percentiles, and whiskers extend to the most extreme data points not considered outliers, which are plotted individually as separate crosses.

| Integrated QT risk assessment
regression model applied to the individual data (0.017 ms/(μg/mL)) was not statistically different from zero.

| DISCUSS ION
LEV is a long-established ASM approved in many countries as monotherapy or adjunctive treatment of seizures in pediatric and adult patients. It binds synaptic vesicle protein SV2A. 21 Since it was launched in 2000, before the introduction of the ICH S7B guidance defining the preclinical assessment of QT prolongation risk, 1 the nonclinical cardiovascular safety pharmacology package only comprised a dog Purkinje fiber assay and several in vivo studies in dogs. A TQT study was conducted in healthy volunteers at therapeutic and supratherapeutic doses of LEV, 5 in compliance with the ICH E14 clinical guidance. 6 However, the effects of LEV on cardiac ion channels, in particular hERG, were never assessed so far.
In 2013, the CiPA paradigm 3,4 was introduced to further assess TdP risk beyond hERG and QT prolongation, using a set of three nonclinical in vitro and in silico assays. The recently adopted ICH E14/S7B Q&As 2 which partly integrates CiPA assays (Figure 1) now offers to drug makers multiple strategies to evaluate proarrhythmic risk and potentially avoid a mandatory clinical TQT study. Here, we used LEV to retrospectively test the predictivity of the CiPA assays and verify the concordance of the full nonclinical dataset with available clinical and postmarketing data.

| Minimal effects of LEV on cardiac ion currents
Standardized patch clamp assays are the first of the three nonclinical CiPA assays. We tested the direct effects of LEV on a panel of eight ion channels involved in cardiac action potential, including hERG. While some currents such as Na V 1.5 or Ca V 1.

| Absence of proarrhythmic effects in different virtual human cardiomyocyte populations
The

| Absence of proarrhythmic effects in dog and human in vitro action potential assays
Before CiPA emerged, the dog Purkinje fiber assay was a commonly used in vitro model to assess potential proarrhythmic effects of drugs. 26

| Very rare post-marketing QT prolongation cases
Despite the negative TQT study in healthy volunteers, very rare cases of QT prolongation have been reported in the FDA Adverse Event Report System (FAERS) database, with an estimated reporting rate of 0.56 per 100 000 patient-years of treatment with LEV. 28 Using the background reporting in the FAERS database as a reference, QT prolongation for LEV was not considered to have a higher reporting rate than expected. Although the primary mechanism of QTc prolongation/TdP is known to be hERG blockade, many risk factors can contribute, such as KCNH2 gene mutation, electrolyte disturbances, pre-existing heart disease, bradycardia, or gender. 29,30 Of interest, among published LEV cases, two patients, both females, were carrying a KCNH2 gene mutation. 31 hypothyroidism. This is also reflected by the in silico simulations which showed that it is only when several risk factors such as bradycardia (slow pacing) and adrenergic stimulus are combined that some repolarization abnormalities start to be observed in a very low number of cells (<1%). Some authors have also considered epilepsy as a possible interfering factor for drug-induced QT prolongation. 34,35 Three potential mechanisms have been considered by which cardiac repolarization thus QT interval could be altered during seizures: increased catecholamine release, 36

ACK N OWLED G M ENTS
The authors wish to acknowledge Eric Martel and Nigel Gillard for their contribution to the dog studies, as well as Nancy Yuen for her valuable scientific advice.

Blanca Rodriguez holds a Wellcome Trust Fellowship in Basic
Biomedical Sciences (214 290/Z/18/Z), and an NC3Rs Infrastructure for Impact Award (NC/P001076/1). All the studies included in this work were sponsored by UCB.

DATA AVA I L A B I L I T Y S TAT E M E N T
The datasets generated for the present work are available from the corresponding author upon reasonable request.

E TH I C S S TATEM ENT
All animal studies were carried out in compliance with European legislation on laboratory animal use and welfare in force at the time of study completion. All experimental protocols were approved by the institutional animal care and use committee.