Inhibitory Effects of Cenobamate on Multiple Human Cardiac Ion Channels and Possible Arrhythmogenic Consequences Authors: Andreea Larisa Mateias, Florian Armasescu, Bogdan Amuzescu, Alexandru Dan Corlan, Beatrice Mihaela Radu Published: 2024 ABSTRACT Cenobamate is a novel third-generation antiepileptic drug used for the treatment of focal onset seizures and particularly for multi-drug-resistant epilepsy; it acts on multiple targets: GABAA receptors (EC50 42–194 µM) and persistent neuronal Na+ currents (IC50 59 µM). Side effects include QTc interval shortening with >20 ms, but not <300 ms. Our in vitro cardiac safety pharmacology study was performed via whole-cell patch-clamp on HEK293T cells with persistent/inducible expression of human cardiac ion channel isoforms hNav1.5 (INa), hCav1.2 (α1c + β2 + α2δ1) (ICaL), hKv7.1 + minK (IKs), and hKv11.1 (hERG) (IKr). We found IC50 of 87.6 µM (peak INa), 46.5 µM (late INa), and 509.75 µM (ICaL). In experiments on Ncyte® ventricular cardiomyocytes, APD90 was reduced with 28.6 ± 13.5% (mean ± SD) by cenobamate 200 µM. Cenobamate’s marked inhibition of INa raises the theoretical possibility of cardiac arrhythmia induction at therapeutic concentrations in the context of preexisting myocardial pathology, in the presence of action potential conduction and repolarization heterogeneity. This hypothetical mechanism is consistent with the known effects of class Ib antiarrhythmics. In simulations with a linear strand of 50 cardiomyocytes with variable inter-myocyte conductance based on a modified O’Hara–Rudy model, we found a negligible cenobamate-induced conduction delay in normal tissue, but a marked delay and also a block when gap junction conduction was already depressed. 1. Introduction Cenobamate (formerly known as YKP3089) is a novel antiepileptic drug that was approved by the US Food and Drug Administration in November 2019 and by the European Medicines Agency in March 2021. It is commercialized in the US by SK Life Science Inc. and in Europe by Arvelle Therapeutics Netherlands BV under the commercial names Xcopri® and Ontozry®, respectively [1]. The drug has been successfully used for treating adults with focal onset seizures and particularly for epilepsy resistant to other antiepileptic drugs, achieving a reduction in monthly seizure frequency of up to 55% and unprecedented seizure-free rates of up to 28% [2]. Cenobamate is a tetrazole alkyl carbamate derivative featuring some structural similarity to earlier compounds such as meprobamate, felbamate, carisbamate, and retigabine. Interestingly, cenobamate was discovered initially via a purely phenotypic screening; its main pharmacodynamic effects were identified later and they are still incompletely understood [3]. Its impressive effectiveness is due to action on multiple targets, including positive allosteric modulatory effects on GABAA receptors by binding to a site different from the benzodiazepine binding site [4]. GABAA receptor binding exerts limited agonistic effects and occurs with different affinities for the six α subunit isoforms tested, α1β2γ2 and α2-6β3γ2, with reported EC50 values of 42 to 194 µM ([5] pp. 8–9). The effects on GABAA receptors are somewhat similar to those of barbituric compounds, leading to both tonic and phasic GABAA-mediated inhibition [1]. Another essential contribution to the antiseizure potency of cenobamate is provided by its inhibitory effects on neuronal Na+ channels, particularly on the late (persistent) Na+ current component (reported IC50 of 59 µM) [6]. This selective inhibition of the persistent neuronal Na+ current component may explain the relatively specific inhibition of excitatory pyramidal neuron activity without significant effects on inhibitory interneurons [7]. However, the complex pharmacokinetics of the compound pose some difficulties, requiring careful titration and maintenance, with starting doses of 12.5 mg/day increased gradually to a recommended dose of 200 mg/day and maximally 400 mg/day, where effective peak plasma therapeutic concentrations (Cmax) reach 45.5 µg/mL (169.98 µM), and the average binding to plasma proteins is ~60% (study YKP3089C018) ([5] p. 13). Beyond common side effects such as dizziness, drowsiness, headache [8], and an increased tendency to suicide, cenobamate reduces the QTc interval by more than 20 ms, although not below 300 ms; therefore, it is forbidden for patients with hereditary short QT syndrome. These QT shortening effects were assumed to result from the inhibition of cardiac voltage-dependent Na+ channels and further post-marketing studies were recommended ([9] pp. 11–12). Therefore, the aim of our study was to provide an in-depth evaluation of cenobamate’s effects on different human cardiac ion channels via in vitro patch-clamp experiments according to the principles of the Comprehensive in vitro Proarrhythmia Assay (CiPA) paradigm [10] and to perform a preliminary exploration, via computational approaches, of the possible proarrhythmogenic consequences of clinical relevance. 2. Materials and Methods 2.1. Cell Cultures Experiments were performed on HEK293T cell lines with persistent/inducible expression of different human cardiac ion channel isoforms: hNav1.5 and hKv7.1 + minK (Anaxon AG, Bern, Switzerland), hCav1.2 (α1c + β2 + α2δ1) (B’SYS GmbH, Witterswil, Switzerland), and hKv11.1 (hERG) (Cytocentrics Bioscience GmbH, Rostock, Germany). The cell lines were grown in standard conditions (37 °C, 5% CO2, 90% humidity) in DMEM:F12 medium (D8900, Sigma-Aldrich, Burlington, MA, USA) buffered with 15 mM HEPES, containing 3.15 g/L D-glucose and 0.055 g/L sodium pyruvate, supplemented with 1.2 g/L sodium bicarbonate, 10% fetal bovine serum (F2442, Sigma-Aldrich, Burlington, MA, USA), 2 mM L-glutamine (G7513, Sigma-Aldrich, Burlington, MA, USA), 1% penicillin–streptomycin (P4333, Sigma-Aldrich, Burlington, MA, USA), and with specific antibiotics for selection of transfected cells or induction of gene expression according to manufacturers’ instructions. The medium was changed every 2–3 days and the cells were passed weekly upon detachment with trypsin–EDTA solution. All cell culture media and solutions were sterile filtered with 0.22 μm polyvinylidene fluoride filters. Ncyte® vCardiomyocytes (Ncardia, Leiden, The Netherlands) were cultured in 24-well plates precoated with bovine fibronectin, following manufacturer’s instructions. For the experiments, cardiomyocytes were detached as described previously [11] and plated on ϕ 10 mm glass coverslips, like the other cell lines. 2.2. Whole-Cell Patch-Clamp Pharmacology Assays Manual whole-cell patch-clamp experiments were performed at room temperature, with equipment as described in previous studies [11,12]. Pipettes were pulled from Clark borosilicate glass capillaries with inner filament (GC150F-10, Harvard Apparatus, Holliston, MA, USA) and had typical resistances of 2–3 MΩ. Recordings were performed with a WPC-100 resistive feedback amplifier (ESF electronic, Göttingen, Germany) or an Axopatch 200B amplifier (Axon Instruments—Molecular Devices, San Jose, CA, USA) connected to a Digidata 1322A interface driven by the pClamp 8.2 Clampex module (Axon Instruments, part of Molecular Devices, San Jose, CA, USA). For different types of ion channel, we applied specific voltage-clamp protocols as follows:-For hERG, a standard two-step activation protocol was used: holding at −70 mV, a 100 ms step at −50 mV used as current reference for measurements, a 2 s depolarizing step at +40 mV, followed by a 2 s step at −50 mV to elicit the peak hERG current by rapid recovery from inactivation, repeated sweeps at 10 s intervals, plus specific protocols to assess the voltage dependence of activation and inactivation, as described in Thomet et al., 2021 [11].-For hKv7.1 + minK, a general two-step activation protocol was used: holding at −110 mV, a 250 ms step at −80 mV with a brief (50 ms) step at −90 mV at the middle to monitor seal quality, a 3000 ms depolarizing step at +60 mV, followed by a 1000 ms step at −50 mV; repeated sweeps at 20 s intervals.-For hCav1.2, a pharmacology protocol with holding at −80 mV, a first 100 ms step at −50 mV and a second 200 ms step at 0 mV, repeated sweeps at 20 s intervals with drug application after 5 sweeps if current levels were steady, plus an activation–inactivation protocol with holding at −80 mV, 1000 ms depolarizing steps from −80 to +50 mV in 10 mV increments, followed by 250 ms steps at +10 mV and a return to −80 mV; 14 sweeps at 20 s intervals.-For hNav1.5, multiple voltage protocols were used, including the following:-A standard pharmacology protocol with holding at −90 mV, a 20 ms step at −110 mV for seal monitoring, a 40 ms activation step at −10 mV, and repeated sweeps at 1 s intervals. Initially, 60 sweeps were recorded in control conditions, and 30 sweeps at 3 min after drug or vehicle solution application.-An activation protocol with holding at −90 mV, a 20 ms step at −110 mV for seal monitoring, depolarizing 40 ms steps from −80 mV to +60 mV in 5 mV increments, followed by a return to −90 mV; repeated sweeps at 5 s intervals.-A steady-state inactivation protocol with holding at −120 mV, 1000 ms activation steps from −120 mV to +40 mV, followed by a 25 ms step at −30 mV; repeated sweeps at 5 s intervals.-A use-dependent block protocol with holding at −120 mV and 5 depolarizing pulses at −10 mV lasting 10 ms each, with interpulse intervals increasing from sweep to sweep from 5 ms to 50 ms in 5 ms increments.-A CiPA-like protocol for experiments with ATX-II to measure effects on late Na+ current, similar to that described in [11], except for a holding potential of −90 mV, a preconditioning step at −110 mV, and a repeat interval of 8 s. -For hERG, a standard two-step activation protocol was used: holding at −70 mV, a 100 ms step at −50 mV used as current reference for measurements, a 2 s depolarizing step at +40 mV, followed by a 2 s step at −50 mV to elicit the peak hERG current by rapid recovery from inactivation, repeated sweeps at 10 s intervals, plus specific protocols to assess the voltage dependence of activation and inactivation, as described in Thomet et al., 2021 [11]. For hERG, a standard two-step activation protocol was used: holding at −70 mV, a 100 ms step at −50 mV used as current reference for measurements, a 2 s depolarizing step at +40 mV, followed by a 2 s step at −50 mV to elicit the peak hERG current by rapid recovery from inactivation, repeated sweeps at 10 s intervals, plus specific protocols to assess the voltage dependence of activation and inactivation, as described in Thomet et al., 2021 [11]. -For hKv7.1 + minK, a general two-step activation protocol was used: holding at −110 mV, a 250 ms step at −80 mV with a brief (50 ms) step at −90 mV at the middle to monitor seal quality, a 3000 ms depolarizing step at +60 mV, followed by a 1000 ms step at −50 mV; repeated sweeps at 20 s intervals. For hKv7.1 + minK, a general two-step activation protocol was used: holding at −110 mV, a 250 ms step at −80 mV with a brief (50 ms) step at −90 mV at the middle to monitor seal quality, a 3000 ms depolarizing step at +60 mV, followed by a 1000 ms step at −50 mV; repeated sweeps at 20 s intervals. -For hCav1.2, a pharmacology protocol with holding at −80 mV, a first 100 ms step at −50 mV and a second 200 ms step at 0 mV, repeated sweeps at 20 s intervals with drug application after 5 sweeps if current levels were steady, plus an activation–inactivation protocol with holding at −80 mV, 1000 ms depolarizing steps from −80 to +50 mV in 10 mV increments, followed by 250 ms steps at +10 mV and a return to −80 mV; 14 sweeps at 20 s intervals. For hCav1.2, a pharmacology protocol with holding at −80 mV, a first 100 ms step at −50 mV and a second 200 ms step at 0 mV, repeated sweeps at 20 s intervals with drug application after 5 sweeps if current levels were steady, plus an activation–inactivation protocol with holding at −80 mV, 1000 ms depolarizing steps from −80 to +50 mV in 10 mV increments, followed by 250 ms steps at +10 mV and a return to −80 mV; 14 sweeps at 20 s intervals. -For hNav1.5, multiple voltage protocols were used, including the following: For hNav1.5, multiple voltage protocols were used, including the following: -A standard pharmacology protocol with holding at −90 mV, a 20 ms step at −110 mV for seal monitoring, a 40 ms activation step at −10 mV, and repeated sweeps at 1 s intervals. Initially, 60 sweeps were recorded in control conditions, and 30 sweeps at 3 min after drug or vehicle solution application. A standard pharmacology protocol with holding at −90 mV, a 20 ms step at −110 mV for seal monitoring, a 40 ms activation step at −10 mV, and repeated sweeps at 1 s intervals. Initially, 60 sweeps were recorded in control conditions, and 30 sweeps at 3 min after drug or vehicle solution application. -An activation protocol with holding at −90 mV, a 20 ms step at −110 mV for seal monitoring, depolarizing 40 ms steps from −80 mV to +60 mV in 5 mV increments, followed by a return to −90 mV; repeated sweeps at 5 s intervals. An activation protocol with holding at −90 mV, a 20 ms step at −110 mV for seal monitoring, depolarizing 40 ms steps from −80 mV to +60 mV in 5 mV increments, followed by a return to −90 mV; repeated sweeps at 5 s intervals. -A steady-state inactivation protocol with holding at −120 mV, 1000 ms activation steps from −120 mV to +40 mV, followed by a 25 ms step at −30 mV; repeated sweeps at 5 s intervals. A steady-state inactivation protocol with holding at −120 mV, 1000 ms activation steps from −120 mV to +40 mV, followed by a 25 ms step at −30 mV; repeated sweeps at 5 s intervals. -A use-dependent block protocol with holding at −120 mV and 5 depolarizing pulses at −10 mV lasting 10 ms each, with interpulse intervals increasing from sweep to sweep from 5 ms to 50 ms in 5 ms increments. A use-dependent block protocol with holding at −120 mV and 5 depolarizing pulses at −10 mV lasting 10 ms each, with interpulse intervals increasing from sweep to sweep from 5 ms to 50 ms in 5 ms increments. -A CiPA-like protocol for experiments with ATX-II to measure effects on late Na+ current, similar to that described in [11], except for a holding potential of −90 mV, a preconditioning step at −110 mV, and a repeat interval of 8 s. A CiPA-like protocol for experiments with ATX-II to measure effects on late Na+ current, similar to that described in [11], except for a holding potential of −90 mV, a preconditioning step at −110 mV, and a repeat interval of 8 s. Experiments on Ncyte® vCardiomyocytes were performed via an amphotericin-perforated whole-cell approach, filling the pipettes with solution containing amphotericin B 300 µg/mL except for a small region near the tip that contained amphotericin-free solution. The voltage-clamp and current-clamp protocols were similar to those used in our previous studies [11,12,13]; the following protocols were recorded in sequence before and at 3 min after application of cenobamate in the bath to a final concentration of 200 µM:-Nav standard protocol: holding −90 mV, depolarizing step at −10 mV lasting 40 ms;-Cav-Nav separation protocol: holding −80 mV, first step at −20 mV 200 ms, intermediate step at −40 mV 100 ms, second step at −50 mV to +50 mV 200 ms, in 10 mV increments in consecutive sweeps;-hERG standard protocol: holding −70 mV, a brief preliminary step at −50 mV 100 ms, first step at +40 mV 2000 ms, second step at −120 mV/−50 mV 2000 ms (in consecutive sweeps);-A double voltage ramp protocol: holding −80 mV, first ramp from −120 mV to +80 mV in 2000 ms, second ramp from +80 mV back to −120 mV in 2000 ms, sweeps repeated at 10 s intervals;-A current-clamp protocol with external pacing: 3 current stimuli of 2 nA amplitude, duration 0.5 ms (up to 2.5 ms in some experiments) applied at 3 s intervals, first stimulus at 500 ms after start of recording; in cells with spontaneous pace-making, this phenomenon was suppressed by applying a steady hyperpolarizing current of −10 pA or −20 pA. -Nav standard protocol: holding −90 mV, depolarizing step at −10 mV lasting 40 ms; Nav standard protocol: holding −90 mV, depolarizing step at −10 mV lasting 40 ms; -Cav-Nav separation protocol: holding −80 mV, first step at −20 mV 200 ms, intermediate step at −40 mV 100 ms, second step at −50 mV to +50 mV 200 ms, in 10 mV increments in consecutive sweeps; Cav-Nav separation protocol: holding −80 mV, first step at −20 mV 200 ms, intermediate step at −40 mV 100 ms, second step at −50 mV to +50 mV 200 ms, in 10 mV increments in consecutive sweeps; -hERG standard protocol: holding −70 mV, a brief preliminary step at −50 mV 100 ms, first step at +40 mV 2000 ms, second step at −120 mV/−50 mV 2000 ms (in consecutive sweeps); hERG standard protocol: holding −70 mV, a brief preliminary step at −50 mV 100 ms, first step at +40 mV 2000 ms, second step at −120 mV/−50 mV 2000 ms (in consecutive sweeps); -A double voltage ramp protocol: holding −80 mV, first ramp from −120 mV to +80 mV in 2000 ms, second ramp from +80 mV back to −120 mV in 2000 ms, sweeps repeated at 10 s intervals; A double voltage ramp protocol: holding −80 mV, first ramp from −120 mV to +80 mV in 2000 ms, second ramp from +80 mV back to −120 mV in 2000 ms, sweeps repeated at 10 s intervals; -A current-clamp protocol with external pacing: 3 current stimuli of 2 nA amplitude, duration 0.5 ms (up to 2.5 ms in some experiments) applied at 3 s intervals, first stimulus at 500 ms after start of recording; in cells with spontaneous pace-making, this phenomenon was suppressed by applying a steady hyperpolarizing current of −10 pA or −20 pA. A current-clamp protocol with external pacing: 3 current stimuli of 2 nA amplitude, duration 0.5 ms (up to 2.5 ms in some experiments) applied at 3 s intervals, first stimulus at 500 ms after start of recording; in cells with spontaneous pace-making, this phenomenon was suppressed by applying a steady hyperpolarizing current of −10 pA or −20 pA. For experiments, cells were detached with trypsin-EDTA, centrifuged, resuspended in standard extracellular solution, and placed in 24-well culture plates containing ϕ 10 mm glass coverslips sterilized by brief immersion in pure ethanol and soaking. After 30 min incubation at 37 °C to allow cell attachment, each coverslip was removed from the well and placed at the center of a ϕ 35 mm Petri dish, surrounded by a Teflon™ ring with a central hole of the same diameter as the coverslip, with the inner edge covered with silicone grease (Merck KGaA 1.07746.0100, Rahway, NJ, USA). After pressing the Teflon™ ring to create a water-tight seal, 100 μL of standard extracellular solution was pipetted over the coverslip. During patch-clamp experiments, 10 μL of cenobamate intermediate solution in extracellular solution at 10× the desired final concentration was added, approximately compensating for the solution loss by evaporation during the time since mounting the coverslip until drug application. 2.3. Chemicals and Solutions In all experiments, the standard extracellular solution had the following composition (in mM): NaCl 140, KCl 2.5, MgCl2 2, CaCl2 2, HEPES 10, D-glucose 10, sucrose 15, pH 7.40 at 25 °C titrated with NaOH 1 M. In experiments performed on hKv11.1 (hERG) channels, the pipette solution contained (in mM): KCl 140, CaCl2 1, MgCl2 2, EGTA 11, HEPES 10, pH 7.20 at 25 °C titrated with KOH 1 M. In experiments on hKv7.1 + minK, this solution was supplemented with MgATP 5 mM, and in later experiments also with creatine phosphokinase (238395, EMD Millipore, Burlington, MA, USA) at 50 U/mL final concentration and phorbol 12-myristate 13-acetate (PMA, 524400, EMD Millipore, Burlington, MA, USA) freshly added from a 10 mM stock solution in DMSO to a final concentration of 10 μM. For experiments on hNav1.5 and hCav1.2, the pipette solution contained (in mM): D-gluconic acid 70, CsCl 60, MgCl2 5, CaCl2 1, Na2ATP 5, EGTA 11, HEPES 10, pH 7.20 by titration with CsOH 1 M. For experiments on Ncyte® vCardiomyocytes, the pipette solution contained (in mM): K+ gluconate 100, KCl 20, CaCl2 1, MgCl2 1, EGTA 11, HEPES 10, Na2phosphocreatine 3, MgATP 4, pH 7.20 at 25 °C titrated with KOH. Amphotericin B (E3789, Sigma, Burlington, MA, USA) was used to prepare a stock solution with a concentration of 30 mg/dL, kept at −20 °C and protected from light. ATX-II (STA-700, Alomone labs, Jerusalem, Israel) was dissolved in pure water to obtain a 20 µM stock solution; aliquots were stored at −20 °C and used to prepare the working solution at 20 nM or 100 nM in extracellular solution. A primary 200 mM stock solution of cenobamate (purchased from Selleckchem, Houston, TX, USA or DC Chemicals, Shanghai, China) in DMSO was stored at −80 °C for several months, with small samples stored at −20 °C for experiments for less than 1 month. In a large number of experiments, including those on hiPSC-CM, the final estimated concentration in the recording chamber was 200 μM, which is slightly higher that the reported maximal plasma therapeutic effective concentration in human trials for a dose of 400 mg/day (Cmax = 169.98 μM, [5] p. 10). In each whole-cell patch-clamp experiment, the same sequence of voltage protocols was applied initially and upon drug exposure, and the effect on peak or plateau current levels was assessed when a steady state was reached. 2.4. Data Analysis and Computer Simulations Throughout the paper, data are reported as mean ± SD or mean ± SEM, as appropriate, with n indicating the number of cells recorded in each experimental condition. Statistical significance was tested via Student’s t test or non-parametric variants if the normality condition was not met, using a critical level of α = 0.05. Computer simulations at single-cardiomyocyte level or on a 1D linear string model with n = 50 cells were developed using C++ scripts and R scripts for postprocessing and visualization (GCC 11.4.0 and R 4.1.2 from the Linux Mint v21.3 distribution), starting from a previously published version of a modified O’Hara–Rudy 2011 human ventricular cardiomyocyte electrophysiology model [14] optimized for pharmacology assays [11], by including cardiac ion channel inhibition data and blocking/unblocking kinetics specific for cenobamate, and considering different gap junction conductance values that are specific for normal or pathological intercellular coupling in the working ventricular myocardium. 3. Results 3.1. Assessment of Inhibitory Effects of Cenobamate on Different Cardiac Ion Channels and Action Potentials Recorded in hiPSC Cardiomyocytes Using the experimental settings and protocols described in Section 2, we performed in vitro assays to test the inhibitory effects of cenobamate on multiple human cardiac ion channels expressed in HEK293T cells. Within an initial set of experiments, the drug was applied at a final concentration of 200 μM. Thus, we measured an average inhibition of peak INa elicited with the standard protocol of 69.53 ± 14.63% (mean ± SD, n = 6 experiments) vs. 0.78 ± 1.82% in control experiments (n = 5) where only the solvent (DMSO) was applied in similar conditions at an equivalent final concentration. Individual results are listed in Table S1 of the Supplementary Materials, and the inhibition in a typical experiment is illustrated in Figure 1a. With these experimental data, we computed an apparent IC50 value for peak INa inhibition by cenobamate of 87.64 μM. Furthermore, we evidenced voltage shifts larger than 10 mV for both the voltage dependence of activation and steady-state inactivation, as shown in Figure 2c, where data points were fitted with Boltzmann charge-voltage functions; a typical experiment with recordings of INa activated at multiple voltage steps initially and in the presence of cenobamate 200 µM is shown in Figure 2a. We also found that the inhibition by cenobamate of peak INa is virtually voltage-independent, by dividing peak INa amplitude for multiple depolarizing voltage steps in the presence of cenobamate by the initial amplitude for equivalent voltage steps in a recording performed before applying the blocker, as illustrated in Figure S1 of the Supplementary Materials. Such a behavior is to be expected for a drug with a neutral molecule at physiological pH. In a subsequent set of experiments, we tested the effects of multiple cenobamate concentrations on the peak Na+ current, assessed with a standard voltage protocol with a single depolarizing step at −10 mV, and on the late Na+ current, assessed with a CiPA-like voltage protocol upon adding ATX-II 20 nM or 100 nM in the bath solution. Typical recordings in this setting are illustrated in Figure S2 of the Supplementary Materials, and experimental measurements are listed in Tables S2 and S3 of the Supplementary Materials. The dose–response curves and apparent IC50 values for cenobamate inhibition of peak and late INa (87.6 µM and 46.5 µM, respectively) are illustrated in Figure 3a. Within an initial set of experiments on L-type Ca2+ channels, we obtained an average inhibition of peak ICaL amplitude elicited with the standard protocol during the depolarizing step at 0 mV of 39.26 ± 13.78% (n = 4) vs. 3.10 ± 2.89% in control experiments (n = 3), as shown in Figure 1b and Table S4 of the Supplementary Materials. We also demonstrated a cenobamate-induced voltage shift, particularly of the relative steady-state inactivation, as shown in Figure 2d. A typical experiment with recordings of ICaL activated at multiple voltage steps between −80 mV and +50 mV, followed by a step at +10 mV, initially and in the presence of cenobamate 200 µM, is shown in Figure 2b. We repeated the experiments at multiple cenobamate concentrations (results listed in Table S5 of the Supplementary Materials), obtaining an apparent IC50 value for hCav1.2 inhibition by cenobamate of 509.75 µM by using a dose–response function with unitary Hill coefficient, as shown in Figure 3b. For IKs, the average inhibition of the plateau current at +60 mV was 36.68 ± 13.02% (n = 8) vs. 23.66 ± 9.86% in control experiments (n = 6), as shown in Figure 1c and Table S6 of the Supplementary Materials. Thus, the estimated net average inhibition (cenobamate 200 μM—control) was 36.68 − 23.66 = 13.02%, resulting in an estimated IC50 for IKs (Kv7.1 + minK) of 1336.1 μM. Finally, in experiments on hERG channels, we obtained an average peak IKr inhibition by cenobamate 20 μM of 5.3 ± 1.8% vs. 4.1% in control experiments, as shown in Figure 1d. Subsequent experiments on hERG channels performed with a cenobamate concentration of 200 µM provided a relative peak IKr inhibition of 10.97 ± 3.11% (Table S7 of Supplementary Materials). With these data, we computed an average peak IKr inhibition at 200 µM cenobamate of 10.97 − 4.1 = 6.87%, resulting in an estimated IC50 of 2711.2 µM. However, we preferred to consider for further computations the apparent cenobamate IC50 for IKr (hERG) of 1869 μM reported in preliminary studies required for drug approval ([5] p. 10). We also performed a series of amphotericin-perforated whole-cell experiments on Ncyte® vCardiomyocytes hiPSC-CM preparations. In this setting, cenobamate 200 µM induced an average APD90 shortening by 28.6 ± 13.5% (mean ± SD, n = 5) and a resting potential (RP) depolarization from −51.3 ± 7.8 mV (n = 8) to −40.1 ± 10.4 mV (n = 7). The results are listed in Table S8 of Supplementary Materials, while a typical current-clamp recording showing external stimulus-triggered action potentials (APs) and the distribution of AP duration (APD90) and RP values are illustrated in Figure 4a,b. 3.2. Estimation of State-Dependent Blocking and Unblocking Rates of Cenobamate for Nav1.5 Channels In this approach, we adopted the modulated receptor hypothesis (MRH) of Na+ channel blocks by local anesthetics [15,16] for our extended Markov state diagram of Nav1.5 channel gating, including conformation-specific blocked states, as illustrated in Figure 5. To estimate state-specific blocking and unblocking kinetics, we applied, initially and during exposure to cenobamate, a special voltage protocol devised for use-dependent block, with repeated brief depolarizing pulses at −10 mV separated by variable intervals at a hyperpolarized potential (−120 mV), with durations between 5 and 50 ms, as described in Section 2. Recordings using this protocol in initial conditions and with cenobamate application are illustrated in Figure S3 of the Supplementary Materials. By noting with td = 10 ms and tr = 5–50 ms the times spent at a depolarized potential and at a resting hyperpolarized potential during each pulse, with on the fraction of available (non-inactivated or blocked) channels that produce the peak Na+ current during each depolarizing pulse of the sequence, and with τi and τr, the apparent macroscopic time constants of inactivation and recovery from inactivation in control conditions (in the absence of a blocker), we obtained using several approximations a general recurrence formula that allows computation for each experiment of the recovery time constant, knowing the inactivation time constant (obtained by the monoexponential fit of the inactivation part of the recording during the depolarization pulse) and the decay in peak INa amplitude during pulse n relative to the peak amplitude during the preceding pulse (1 − on)/on−1 (detailed explanations in Section S1 of the Supplementary Materials):τr=tr/ln1−e−td/τi−ln1−onon−1 Furthermore, for application of the same voltage protocol during exposure to cenobamate, by using a similar reasoning and considering a composed “inactivated-or-blocked” state, we obtained a formula for computing an estimate of the blocking/unblocking time constant specific for the inactivated conformation of the channels (τib) starting from τi and an apparent time constant for open channel blocks (τob):τib=1/1/τr−ln1−e−td(1/τi+1/τob)−ln1−onon−1/tr Using these computation methods, we obtained by analyzing data obtained in four experiments performed on hNav1.5-expressing HEK293 cells, with application of a voltage-clamp protocol including five depolarizing pulses (−10 mV, td = 10 ms) with interpulse intervals at −120 mV increasing from 5 to 50 ms in 5 ms increments (tr = 5–50 ms), an estimated τib of 4.03 ± 0.70 ms (mean ± SD, see Table 1). Starting from this estimated value, we computed kib−1 and kib, assuming that their ratio (the inactivated-specific Kd) should follow the same proportion relative to the open-specific Kd, as in the case of lidocaine-Nav1.5, estimated from the experimental data of Cardona K et al., 2010 [17]:kob = 5173 M−1 ms−1, kib = 4998.4 M−1 ms−1, kob−1 = 0.0128 ms−1, kib−1 = 0.0384 ms−1 Kd (I) = kib−1/kib = 0.0384 ms−1/0.0049984 µM−1 ms−1 = 7.6825 µM Kd (O) = kob−1/kob = 0.0128 ms−1/0.005173 µM−1 ms−1 = 2.4744 µM Kd (I)/Kd (O) = 3.1048 For cenobamate, using the same state-specific dissociation constant ratio, and noting with [D] the drug (cenobamate) concentration used in experiments (estimated to 98.44 µM based on apparent inhibition of peak INa amplitude) Kd (O) = 87.77 µM; Kd (I) = 87.77 µM × 3.1048 = 272.5083 µM kib × [D] + kib−1 = 1/τib = 0.2484 ms−1 => 0.2484 = (kib−1/272.5083) × 98.44 + kib−1 kib−1 = 0.2484/(1 + 98.44/272.5083) = 0.18252 ms−1 kib = (0.2484 − 0.18252)/98.44 = 0.06593 ms−1/98.44 µM = 0.0006698 µM−1 ms−1 To estimate the blocking and unblocking rates specific for the open channel conformation (kob and kob−1), we assessed the apparent macroscopic time constant of current decay via the monoexponential fitting of the inactivation phase of INa elicited with the standard Nav voltage protocol in the same cell before and during the addition of cenobamate 200 µM. Thus, we obtained, in drug exposure and control experiments, the time constants presented in Table 2. With these kinetic data, we directly estimated the two rates, by making a supplementary assumption: the dissociation constant (Kd) is equivalent to the IC50 of cenobamate on peak INa amplitude (87.8 µM). Consequently, we obtained the system of two algebraic equations:Kd = kob−1/kob = 87.77 µM 200 µM × kob + kob−1 = 0.62 ms−1 leading to:kob = 0.62/287.77 = 0.00215 µM−1 ms−1 kob−1 = 0.189 ms−1 3.3. Inclusion of Pharmacological Inhibition Experimental Data in a Ventricular Cardiomyocyte Electrophysiology Model: Testing Effects on Action Potential and Computing Proarrhythmogenic Risk Predictors We applied all the experimental inhibition data and state-dependent blocking kinetics for Nav1.5 to a modified O’Hara–Rudy 2011 model adapted for pharmacology simulations and specifically for computing the proarrhythmogenic risk predictor Qnet, as described in [18,19]. The starting C++ script, described in [11] (available on GitHub at https://github.com/bamuzesc/Qnet-for-CiPA.git and on Zenodo at https://zenodo.org/record/5615548#.YXvbQ7hyGSp, accessed on 10 October 2024), has been completed by adding pharmacological inhibition data for the newly defined compound cenobamate, and the state-specific Nav1.5 blocking–unblocking rates have been inserted in the Hodgkin–Huxley-type gating model of INa defined in the subroutine RGC (rates, gates, and currents), as described in Section S2 of the Supplementary Materials. These changes were performed according to the method used by Cardona et al., 2010 [17], by defining a blocked state probability b and expressing its rate of change by a first order differential equation:db/dt=m3(1−fINaP)hj+fINaPhpjpkob+m31−(1−fINaP)hj−fINaPhpjpkib[D](1−b)−−m3(1−fINaP)hj+fINaPhpjpkob−1+m31−(1−fINaP)hj−fINaPhpjpkib−1b and inserting its value, computed by numeric integration, in the current equation:INa=(1−b)GNa(V−ENa)m3(1−fINaP)hj+fINaPhpjp With this edited script, we computed Qnet values for cenobamate concentrations ranging between 1 and 25 × Cmax (170 µM). These values are plotted in Figure 6a, along with those for a 12-compound drug panel used as the Comprehensive in vitro Proarrhythmia Assay (CiPA) [10,20] training set [21]. These large Qnet values, which are the result of the predominant inhibition by cenobamate of depolarizing ion currents (INa and ICaL) and negligible effects on repolarizing currents (IKr and IKs), demonstrate convincingly the lack of torsadogenic risk for cenobamate. The action potential (AP) shapes generated by the script, using the default parameter set for subendocardial ventricular cardiomyocytes (cell type = 0) at different cenobamate concentrations and in control conditions, are illustrated in Figure 6b. At 1 × Cmax, a clinically relevant concentration, there is already a significant shortening of AP duration (computed APD90 of 270.5 ms vs. 304.0 ms in control conditions), consistent with the QTc shortening recorded in experiments on animal models and clinical trials and with our experimental data on APD90 shortening by cenobamate 200 µM in hiPSC-CM. In Figure 6c,d,e, we illustrate the effects of cenobamate concentrations in the range 1–5 × Cmax on AP shape and duration for the three cardiomyocyte variants of the O’Hara–Rudy 2011 model: subendocardial (cell type = 0), subepicardial (cell type = 1), and midmyocardial (cell type = 2). Interestingly, for the midmyocardial type, deemed to form the major part of ventricular myocardium, APD is shortened but remains above 300 ms, which is similar to experimental findings on QTc interval shortening during cenobamate obtained in clinical study YKP3089C020 performed on healthy volunteers ([9], p. 11). 3.4. Computational Study of Cenobamate Effects on AP Propagation in Ventricular Tissue Although cenobamate does not present any significant torsadogenic risk, its local anesthetic-like properties and relatively strong inhibition of the human cardiac Na+ channel isoform hNav1.5 imply that it may alter the rapid depolarization (phase 0) of the AP, particularly in ventricular cardiomyocytes, subsequently slowing down conduction velocity and possibly creating favorable conditions for the activation of reentry circuits in pathological myocardium. To explore in more detail these potentially hazardous effects, we performed some theoretical computations to assess conditions prone to triggering ventricular reentry arrhythmias (Section S3 of Supplementary Materials) as well as simulations with a 1D model of the human ventricular myocardium using the modified ventricular cardiomyocyte electrophysiology model described above. These preliminary computations prove that it is conceivable to have, in the pathological myocardium, short (less than 4.5 cm) potential (latent) reentry circuits resulting from the combination of a fast (normal) and slow (pathologic) pathway, united at the ends, which would not give reentry in basal conditions but would lead to reentry in the presence of cenobamate. For example, at a cenobamate concentration of 200 µM, the minimum length of a two-pathway circuit that would allow slow-to-fast reentry will decrease to only 6 mm for an estimated refractory period on the fast pathway of 300 ms. For the 1D ventricular tissue propagation model, we simulated a linear string of 50 cardiomyocytes with the default geometry parameters of the O’Hara–Rudy 2011 model united end-to-end by intercalated discs with gap junction conductances Gj varying between highly altered pathological values (lower limit 300 pS/pF) and normal physiological values (upper limit 6000 pS/pF). The modified cell electrophysiology O’Hara–Rudy model with included cenobamate pharmacology data developed by us was used for each cell in the string, and the total current density equation was supplemented with a term describing longitudinal intermyocyte current flow obtained by multiplying Gj with the actual potential difference between adjacent cardiomyocytes. Simulation results with different parameter sets (Gj between 300 and 6000 pS/pF and cenobamate concentration between 0 and 6 × Cmax) are presented in Tables S9 and S10 of the Supplementary Materials, and for some selected relevant values in Figure 7. 4. Discussion By using a combination of experimental in vitro assays, theoretical computations, and computer simulations, we succeeded in evidencing some novel pharmacological properties of cenobamate and to explore their potential arrhythmogenic consequences, particularly in the pathological myocardium with altered gap junction connectivity and slow conduction velocity. The specific and relatively strong inhibitory effects of the drug on cardiac voltage-dependent Na+ channels (hNav1.5) and L-type Ca2+ channels were a bit surprising, given that previous studies performed either on neuronal TTX-sensitive hNav1.7 channels stably expressed in HEK293 cells or on newborn rat CA3 hippocampal neurons found a relatively weak inhibitory potency on the peak (transient) Na+ current (estimated IC50 > 500 µM), and a much stronger effect on the late (persistent) INa, with IC50 values of 41.8 µM ([5] p. 8) and 53.1 µM [22], respectively. This strong inhibition by cenobamate of persistent INa is similar to that exerted by other antiseizure drugs: IC50 2.0 µM for riluzole, 79.2 µM for carbamazepine, and 118.8 µM for lamotrigine ([5] p. 8). This may explain the peculiar effectiveness of this medication in genetic epilepsies with gain-of-function mutations in the SCN1A gene and increased persistent INa [23]. In our experiments on hNav1.5 in the presence of anemone sea toxin ATX-II, we also noticed the strong inhibition by cenobamate of the persistent Na+ current, with a similar apparent IC50 of 46.5 µM. Such effects on multiple human cardiac ion channels could represent an argument for the adequacy of the mechanistic in vitro approach of the CiPA paradigm [10,20,21], which is currently only an optional method under the recently adopted S7B/E14 Q&A cardiac safety testing guidelines, and is still often limited to in vitro hERG screening [24]. Given the significant inhibition of the peak Nav1.5 INa current, with an estimated apparent IC50 below 100 µM, which is in a clinically relevant range of concentrations, we considered it important to perform an in-depth study of hNav1.5 channel block by cenobamate, including state-dependent blocking kinetics and affinities, and other relevant effects. Indeed, we noticed significant shifts in the voltage dependence of both the activation and steady-state inactivation of INa in the presence of cenobamate at 200 µM. One of the most important published studies on the effects of cenobamate on neuronal INa, that of Nakamura et al., 2019 [22], found a similar hyperpolarizing shift of the voltage dependence of inactivation, with a −6 mV change in half-inactivation potential at 100 µM cenobamate, with no significant change in the slope factor and voltage dependence of activation. The authors suggested that the main inhibitory effect of cenobamate on Na+ channels may be due to the modulation of the fast voltage-dependent inactivation. Via an elegant method, they also estimated a high affinity for the inactivated conformation of Na+ channels (KI = 48.2 µM) and a low affinity for the resting (closed) conformation (KR = 797.8 µM). This state dependence of drug binding to Na+ channels also led us to assume, in the case of hNav1.5, different blocker affinities and hence blocking–unblocking rates for the three main conformations: open (O), inactivated (I), and closed (C); this hypothesis was adopted in previous state-dependent blocking kinetics studies on Nav1.5, particularly those for the class Ib antiarrhythmic lidocaine [17,25]. Two main mechanistic hypotheses have been proposed for Na+ channel block and the use-dependent inhibition of peak Na+ current by local anesthetics [26]: the modulated receptor hypothesis (MRH) [15,16] and the guarded receptor hypothesis (GRH) [27], as discussed in O’Leary ME and Chahine M, 2017 [28]. The MRH postulates different drug binding affinities for different conformations or states of the channels: high for the open (O) conformation, even higher for the inactivated (I) conformation, and a very low affinity for the closed (C) conformation; drug binding does not affect the gating kinetics itself. The GRH supposes a fast aqueous pathway to the blocking site accessible for charged compounds in the open state only, the block being thus modulated by both the activation and inactivation gates. Given that cenobamate is a neutral molecule at physiological pH (physiological charge 0; strongest acidic pKa 14.28; strongest basic pKa −1.7; logP 1.66; water solubility 0.936 mg/dL—https://go.drugbank.com/drugs/DB06119, accessed on 10 October 2024), we postulated that the blocking site on hNav1.5 channels should also be accessible via a lateral lipophilic pathway and binding/unbinding should also occur in the inactivated and closed conformations; therefore, we considered the MRH model as the most adequate in this case. Mutagenesis studies have shown that the voltage sensors (transmembrane helix S4) of Na+ channel domains I–III contribute the activation gating current [29,30,31,32], while S4 of domain IV is largely involved in fast voltage-dependent inactivation [31,33]. In addition, it has been shown that some highly conserved aromatic residues located on S6 of domain IV, a Phe near the inner end of the selectivity filter [34,35] and a Tyr near the inner activation gate [36], are the main determinants of local anesthetic binding [37]. Therefore, our supplementary working hypothesis in resolving the state-dependent Nav1.5-cenobamate binding affinity by assuming equal ratios of open vs. inactivated-specific dissociation constant for cenobamate and lidocaine, although weak, has some logical grounds. Of course, it will be very important for further studies to more accurately assess this state-dependent affinity and blocking kinetics, if possible via model-independent approaches and avoiding simultaneous multi-parameter optimization methods, since the real mechanism of inhibition is still putative and incompletely validated. Some reports claimed that local anesthetics, like lidocaine and its charged derivatives, also reduce the INa gating current and charge, postulating that the main inhibitory effect occurs via gating charge immobilization [38,39]. Another difficult problem is related to the effects of temperature on Nav1.5 inhibition by cenobamate. Due to the small volume of cell suspension used in our experiments and lack of continuous perfusion, we could not perform recordings at physiological temperature (PT: 35–37 °C). There are also only a few published studies describing the pharmacological effects on Nav1.5 channels at PT. Our own and other studies proved an approximately four-fold reduction in the apparent IC50 of chloroquine binding on Nav1.5, and more than a two-fold reduction for hydroxychloroquine when comparing experimental results on peak Na+ current at PT vs. room temperature, i.e., a stronger binding at higher temperature [11,40], and even higher differences for the effects of these two drugs on late Na+ currents [40]. Another study exploring the inhibition of different Nav isoforms by cannabidiol showed decreased affinity and higher IC50 values at higher temperatures for Nav1.2 [41], but in this case multiple binding sites were present, as indicated by the Hill coefficients of ~3, some of them located at the fenestration between domains I and IV or close to the pocket for the inactivation particle [42]. By similarity with temperature effects on the blocking and unblocking rates of hERG channel pore blockers [43], we can assume that Nav1.5 central cavity blockers increase both their binding and unbinding rates with increasing temperature. On the other hand, Nav channel gating is also temperature-dependent. A recent study on multiple Nav isoforms evidenced the hyperpolarizing shifts of the voltage dependence of activation with increasing temperature, amounting up to −10 mV for hNav1.5 upon changing from 15 °C to 35 °C [44]. Our own unpublished data, including voltage-clamp recordings on hNav1.5 at 22 °C and 35 °C, resulted in estimates of the Nav1.5 activation gating temperature coefficient Q10 of 1.98 and the inactivation Q10 of 1.2; measuring the area under the curve for the Nav1.5 current at these temperatures yielded a ratio of 0.64 (at physiological vs. room temperature), i.e., the activation, but also the inactivation, kinetics are faster. For charged compounds that can only access the central cavity blocking site during the open state, this would imply a lower binding at higher temperatures, whereas for neutral compounds that can also access this site via the lipophilic lateral pathway, binding would be stronger if the affinity for the inactivated conformation is higher compared to that for the open conformation. We may add to these complex effects the temperature dependence of drug protonation/unprotonation, if the charged and neutral forms feature different binding affinities, as in the case of lidocaine [45]. Due to the multiplicity of mechanisms and opposed direction of effects, for local anesthetic-type Nav blockers it is quasi-impossible to extrapolate experimental results on drug inhibition at room temperature to predict changes in affinity at PT [46]. Turning to the cardiac consequences of the inhibitory effects of cenobamate on multiple ion channels that we identified in this study, we were of course pleased to retrieve, in single-cardiomyocyte electrophysiology simulations, an AP shortening similar to the QT interval shortening greater than 20 ms found in 31% of healthy volunteers who received a 200 mg oral dose of cenobamate and 66% of those who received 500 mg vs. 6–17% in negative control subjects given a placebo in study YKP3089C020 ([9] p. 11). In view of our data, this cardiac depolarization shortening is a combined result of the inhibition of persistent INa and ICaL, both directly and indirectly, by the subsequent reduction in myocyte sarcoplasmic calcium concentration and current via Na+/Ca2+ exchangers (NCX), which provide, via the electrogenic exchange of one extruded Ca2+ ion for three internalized Na+ ions, a net inward current during the AP plateau phase [47]. With high positive Qnet values that preclude any proarrhythmogenic risk via the inhibition of delayed rectifier K+ currents and place the compound in the low torsadogenic risk group (Figure 6a), there is still a possibility that cenobamate will trigger reentry arrhythmias via inhibitory effects on INa and, subsequently, on conduction velocity, combined with a reduction in AP duration and hence in refractory period. Both preliminary theoretical computations and more elaborate computer simulations on a 1D string of ventricular cardiomyocytes showed that, in the pathological myocardium with more than ten-fold increased intermyocyte gap junction resistance, INa inhibition by cenobamate should exert a marked slowing down of conduction velocity, leading to a significant difference in velocity compared to an adjacent region of the normal myocardium and the possibility of activation of a reentry circuit either by slow-to-fast pathway conduction or by fast-to-slow conduction if AP propagation failed on the slow pathway or the cardiomyocytes are within a narrow vulnerability window for reentry during the late stage of phase 3 [48]. We identified some upper and lower bounds for space and time intervals, conduction velocity differences, and pacing frequencies that would result in reentry, but of course this exploration has to be continued using more sophisticated and realistic 2D or 3D myocardium models, given the extremely large variability in velocity and refractoriness induced by conditions such as myocardial ischemia [49]. Early simulation studies of cardiac AP propagation identified a minimal 5- to 10-fold gap junction resistance increase as a necessary condition for unidirectional block ([50], cited by [51]), but these conditions can be easily met, particularly during ischemia-related acidosis [52,53,54,55]. We also note that 1D cardiomyocyte tissue models at different levels of complexity have been successfully used to study various problems, including unidirectional block [25,56,57,58,59,60]. Finally, it is worth mentioning that the complex pharmacokinetics and biodisponibility of cenobamate play an important role for the in vivo and clinical relevance of these effects. Complex studies were performed to characterize the single-dose pharmacokinetics and interactions with bupropion, midazolam, warfarin, and omeprazole [61] or to assess the mass balance and metabolic profile using a radioactively labeled compound [62]. Accurate and validated liquid chromatography–mass spectrometry (UHPLC-MS/MS) protocols were developed, using, for example, a water–acetonitrile multistep elution gradient for cenobamate extraction from plasma samples [63]. Distribution and bioavailability assays found different species-specific percentages of protein binding of cenobamate, ranging from 35.4% in rabbit and 43.2% in mouse to 61.0% in human and 60.7–70.7% in non-human primates ([5] p.10). Given the average plasma protein binding of 60%, the free compound levels may be lower than the total concentrations assessed during pharmacokinetic studies, even at maximal effective plasma therapeutic concentrations; therefore, the biological effects may be limited and the risks reduced. In assessing clinical effects, a 0.5 × Cmax concentration may provide the most relevant results. 5. Conclusions Cenobamate, applied at concentrations in the clinically relevant range, resulted in the significant inhibition of Nav1.5 and Cav1.2 channels, which may explain the dose-dependent QTc shortening effects and the interdiction for administering the drug to patients with short QT syndrome. These blocking effects, particularly on Nav1.5, although generally deemed to be antiarrhythmic, may also trigger arrhythmias in pathological conditions such as myocardial ischemia or other structural heart diseases by slowing down conduction and favoring reentry [64].