Tetrodotoxin‐Sensitive Neuronal‐Type Na⁺ Channels: A Novel and Druggable Target for Prevention of Atrial Fibrillation

Abstract

Background

Atrial fibrillation (AF) is a comorbidity associated with heart failure and catecholaminergic polymorphic ventricular tachycardia. Despite the Ca²⁺‐dependent nature of both of these pathologies, AF often responds to Na⁺ channel blockers. We investigated how targeting interdependent Na⁺/Ca²⁺ dysregulation might prevent focal activity and control AF.

Methods and Results

We studied AF in 2 models of Ca²⁺‐dependent disorders, a murine model of catecholaminergic polymorphic ventricular tachycardia and a canine model of chronic tachypacing‐induced heart failure. Imaging studies revealed close association of neuronal‐type Na⁺ channels (nNa_v) with ryanodine receptors and Na⁺/Ca²⁺ exchanger. Catecholamine stimulation induced cellular and in vivo atrial arrhythmias in wild‐type mice only during pharmacological augmentation of nNa_v activity. In contrast, catecholamine stimulation alone was sufficient to elicit atrial arrhythmias in catecholaminergic polymorphic ventricular tachycardia mice and failing canine atria. Importantly, these were abolished by acute nNa_v inhibition (tetrodotoxin or riluzole) implicating Na⁺/Ca²⁺ dysregulation in AF. These findings were then tested in 2 nonrandomized retrospective cohorts: an amyotrophic lateral sclerosis clinic and an academic medical center. Riluzole‐treated patients adjusted for baseline characteristics evidenced significantly lower incidence of arrhythmias including new‐onset AF, supporting the preclinical results.

Conclusions

These data suggest that nNa_Vs mediate Na⁺‐Ca²⁺ crosstalk within nanodomains containing Ca²⁺ release machinery and, thereby, contribute to AF triggers. Disruption of this mechanism by nNa_v inhibition can effectively prevent AF arising from diverse causes.

Nonstandard Abbreviations and Acronyms

AF

atrial fibrillation

ALS

amyotrophic lateral sclerosis

CPVT

catecholaminergic polymorphic ventricular tachycardia

I_Na

Na⁺ current

ISO

isoproterenol

Na_v

Na⁺ channel

nNa_v

neuronal Na⁺ channel

NCX

Na⁺/Ca²⁺ exchange

RyR2

ryanodine receptor

SR

sarcoplasmic reticulum

WT

wild‐type

Clinical Perspective

What Is New?

• Riluzole, a neuronal type Na⁺ channel blocker, was investigated in 2 retrospective amyotrophic lateral sclerosis cohorts.

• Patients treated with riluzole had significantly lowered incidence of arrhythmias, including new‐onset atrial fibrillation.

What Are the Clinical Implications?

• The pharmacological class of neuronal‐type Na⁺ channel blockade may offer a unique mechanism to prevent arrhythmias including atrial fibrillation.

Atrial arrhythmia, such as atrial fibrillation (AF), is a leading cause of morbidity and mortality in the United States.1 It is a common comorbidity associated with heart failure and its risk has been associated with “leaky” ryanodine receptor 2 (RyR2) Ca²⁺ release channels.2, 3, 4 The importance of leaky RyR2 to atrial arrhythmogenesis is particularly evident in patients with catecholaminergic polymorphic ventricular tachycardia (CPVT).5 In this pathology, mutations in RyR2 or in the sarcoplasmic reticulum (SR) Ca²⁺‐binding protein calsequestrin increase propensity for AF.6 Emerging evidence also suggests that the link between Ca²⁺ handling and atrial arrhythmias is in part modulated by Na⁺ influx.7

The predominant Na⁺ channel (Na_v) isoform found in the heart (Na_v1.5) is decreased in cardiomyopathy, as well as in AF.7, 8 However, tetrodotoxin‐sensitive neuronal‐type Na⁺ channels (nNa_vs) are upregulated in these pathologies resulting in enhanced persistent Na⁺ current (I_Na).7, 9 The subcellular localization of different Na_V isoforms thus determines the location of late Na⁺ entry relative to Ca²⁺‐handling machinery. This, in turn, may determine whether nNa_v‐mediated Na⁺/Ca²⁺ exchange (NCX) merely contributes to global cytosolic Ca²⁺ overload or acts directly to trigger abnormal Ca²⁺ release.10, 11, 12

Ventricular proarrhythmia is an important limitation of current drug therapies used in AF.13, 14, 15 Thus, the identification of agents that safely and effectively prevent arrhythmogenic trigger in the atria is imperative. Riluzole, an nNa_v inhibitor used to manage amyotrophic lateral sclerosis (ALS),16 effectively suppresses triggered ventricular arrhythmias in multiple animal models.10, 17, 18, 19 Given its extensive safety profile,20 riluzole could potentially safely prevent AF. Here, we provide evidence from preclinical models and patients suggesting riluzole as a safe and effective treatment for atrial arrhythmias.

Methods

Expanded methods are available in Data S1.

The preclinical and retrospective cohort data that support the findings of this study are available from the corresponding author and Dr Mark Munger (mark.munger@hsc.utah.edu), respectively, upon reasonable request.

Study Approval

All animal procedures were approved by The Ohio State University and University of Michigan Institutional Animal Care and Use Committees and conformed to the Guide for the Care and Use of Laboratory Animals published by the US National Institutes of Health (publication No. 85‐23, revised 2011). The historical cohort was reviewed and approved as exempt following applicable guidelines involving the ethical treatment of human subjects by the University of Utah's institutional review board before the initiation of data collection. The University of Utah's institutional review board is fully accredited by the Human Research Protection Program.

Preclinical Evaluation of Antiarrhythmic Targeting of nNa_vs

Atrial cardiomyocytes, obtained from cardiac calsequestrin mutant (R33Q) wild‐type (WT) mice or failing canine hearts were enzymatically dissociated for patch‐clamp recordings, confocal immunolabeling, or Ca²⁺ imaging. A subset of mice was used to assess the role of nNa_vs in an in vivo arrhythmia induction.

Clinical Evaluation of Antiarrhythmic Targeting of nNa_vs

We conducted a retrospective cohort study of patients with ALS who were treated with riluzole (study group) versus no riluzole (control group) from the ALS Centre of the Azienda Ospedaliero, Universitaria di Modena, Modena, Italy, and the University of Utah, Salt Lake City, Utah. Data were collected through queries of structured data fields. Variables extracted from structured fields included demographic information (ie, age on the index date, sex, race, and body mass index), laboratory results, diagnostic tests and results, prescription records, and hospitalizations with cardiovascular encounter(s) including acute coronary syndrome, acute myocardial infarction, heart failure, and any arrhythmia including atrial flutter or fibrillation. Race and ethnicity were grouped into white, black, Hispanic, Asian, and other/unknown. Cardiovascular risk factors and specific medications for cardiovascular risk prevention and treatment were collected and identified. Distribution of the clinical and demographic characteristics of the study cohort were described among the overall cohort from both the Italian and US databases, and among riluzole versus no‐riluzole users. Baseline characteristics between the riluzole and no‐riluzole groups were compared using chi‐square test, and Fisher exact test if the number of patients having a specific clinical character was <5. Cardiac pacemaker and implantable cardioverter‐defibrillator placement were collected, if applicable, and cardiac monitoring for AF was conducted based on patient presentation of symptoms.

The primary clinical end point was a composite of arrhythmia and AF. The time‐to‐end point from the first exposure to riluzole or earliest available date on or after the diagnosis of ALS, if no riluzole cohort, was projected using Kaplan–Meier curves where patients were censored when they encountered the end point or at the last follow‐up from each institution. The measure of treatment effect was hazard ratio (HR) and 95% CI estimate from Cox proportional hazard regression model under intention‐to‐treat principles (ie, post‐index date variables were not incorporated into the analysis such as medication adherence) where the effect measure was adjusted for the baseline characteristics that were marginally different (P<0.1) between the riluzole and no‐riluzole groups. The analysis was performed for the overall cohort, then limited to the Utah cohort to qualitatively examine the influence of the heterogeneity across the Italy and US cohorts on the measure of treatment effect. All tests were 2‐tailed with an α of 0.05 for statistical significance. Data analysis was conducted with SAS version 9.4 (SAS Institute Inc).

Results

nNa_v Blockade Prevents Leaky RyR2‐Induced Aberrant Ca²⁺ Oscillations

We first determined whether nNa_v exerts a unique action on intracellular Ca²⁺ handling, which may contribute to atrial arrhythmia initiation. Exposure of calsequestrin‐associated CPVT atrial myocytes to isoproterenol (100 nmol/L) induced self‐sustaining, repetitive Ca²⁺ oscillations, consistent with previous findings.6 These oscillations were abolished by tetrodotoxin (100 nmol/L; Figure 1A). Further, tetrodotoxin prevented induction of repetitive Ca²⁺ oscillations by field stimulation in the presence of isoproterenol in nearly all cells tested (Figure 1A), despite increasing SR Ca²⁺ load (Figure S1). However, field stimulation elicited Ca²⁺ oscillations in all cells upon washout of tetrodotoxin (Figure 1A).

Figure 1. Inhibition of tetrodotoxin (TTX)‐sensitive neuronal Na_Vs (nNa_vs) is sufficient to prevent induction of aberrant, repetitive Ca²⁺ oscillations in R33Q atrial myocytes.

A, Representative examples of the line scan images and corresponding Ca²⁺ transients recorded in field‐stimulated R33Q atrial cardiomyocytes paced at 0.5 Hz and loaded with the Ca²⁺ indicator, Fluo‐3 AM. Cells were treated with isoproterenol (ISO, 100 nmol/L) and subsequently TTX (100 nmol/L; green arrow indicates time when TTX was added) was rapidly applied. β‐Adrenergic stimulation with ISO promoted aberrant, repetitive Ca²⁺ oscillations (median pacing frequency was 2 Hz with the range of 0.5 to 4 Hz), while TTX (100 nmol/L) significantly decreased their incidence. Ca²⁺ oscillations were induced after drug washout (red arrow; number of cells tested depicted under the corresponding bars, N=8 animals for ISO and ISO‐TTX, N=6 animals for ISO‐washout, respectively; ***P<0.0001 McNemar test for ISO vs ISO‐TTX, P<0.0001 Fisher exact test for ISO‐TTX vs ISO‐washout). B, Treatment of ISO (100 nmol/L)‐exposed R33Q atrial myocytes with 10 μmol/L riluzole (Ril) significantly reduced the incidence of aberrant, repetitive Ca²⁺ oscillations, an effect that was washable (median pacing frequency was 2 Hz with the range of 0.5 to 7 Hz; number of cells tested depicted under the corresponding bars, N=10 animals for ISO and ISO‐Ril, N=8 animals for ISO‐washout, respectively; ***P<0.0001 McNemar test for ISO vs ISO‐Ril, P<0.0001 Fisher exact test for ISO‐Ril vs ISO‐washout).

Next, we compared tetrodotoxin with riluzole, an agent currently employed in the management of ALS.21 At the resting potential, riluzole has been shown to preferentially block tetrodotoxin‐sensitive nNa_vs in dorsal root ganglion neurons.16 First, we examined the impact of riluzole on the tetrodotoxin‐sensitive component of I_Na. Both 100 nmol/L tetrodotoxin and 10 μmol/L riluzole reduced peak I_Na density to similar degrees (peak I_Na at −35 mV of −36.4±3.7 pA/pF versus −25.8±4.4 and −28.1±3.0 pA/pF for control, riluzole and tetrodotoxin, respectively; P=0.0002 Kruskal–Wallis rank sum test; Figure S2A). Both also shifted steady‐state inactivation of I_Na to more hyperpolarized potentials (V_1/2 of −87.2±3.7 mV versus −98.1±3.0 and −95.9±3.2 mV for R33Q, riluzole and tetrodotoxin, respectively; P=0.0345 Kruskal–Wallis rank sum test; Figure S2B). Importantly, riluzole, like tetrodotoxin, prevented aberrant Ca²⁺ oscillations (Figure 1B) and increased SR Ca²⁺ load (Figure S1). These data point to the antiarrhythmic efficacy in CPVT of nNa_V inhibition by riluzole, likely via suppression of late I_Na induced by isoproterenol. Under control conditions, we did not observe significant differences in late I_Na between WT and CPVT (Figures 2B and 5A). Upon addition of isoproterenol (100 nmol/L), CPVT atrial myocytes evidenced an increase in late I_Na (Figure 2B). Both peak as well as late I_Na were reduced by riluzole (Figure 2). Taken together, these data support a role for nNa_v in arrhythmogenic Ca²⁺ release in CPVT atria.

Figure 2. Effect of neuronal Nav (nNav) blockade with riluzole (Ril) on isoproterenol (ISO)‐promoted inward Na⁺ currents (I_Na) in R33Q atrial myocytes.

A, Representative I_Na (top) were elicited by a protocol presented in the inset before (black) or after addition of ISO (100 nmol/L; red) or ISO (100 nmol/L)+Ril (10 μmol/L; purple). (Bottom) Corresponding current‐voltage relationship from control, ISO, and ISO+Ril (n=10, 9, and 7 cells from 5, 6, and 5 mice, respectively). Addition of ISO+Ril reduced I_Na density relative to ISO alone (peak I_Na at −40 mV of −37.1±3.2 vs −22.4±3.9 pA/pF for ISO and ISO+Ril, respectively; P=0.0102 ANOVA, P=0.0418 for ISO vs ISO+Ril). B, Representative traces of persistent I_Na elicited using the protocol shown in the inset were recorded before (black) or after addition of ISO (100 nmol/L; red) or ISO+Ril (10 μmol/L; purple). ISO enhanced persistent I_Na in R33Q cardiomyocytes, while Ril suppressed β‐adrenergic–mediated increase in persistent I_Na (n=9, 8, and 8 cells from 3, 3, and 4 mice for R33Q, ISO, and ISO‐Ril, respectively; P<0.0001 ANOVA, ***P=0.0004 for persistent I_Na integral and ***P<0.0001 for normalized late I_Na). Summary data are presented as persistent I_Na integral Amp‐ms/F (AmsF⁻¹; left) or normalized late I_Na (%; right), which were measured by either integrating I_Na between 50 and 450 ms (left) or normalizing mean persistent INa recorded between 250 and 450 ms by peak current generated by a step to ‐40 mV.

Figure 5. Augmentation of tetrodotoxin (TTX)‐sensitive neuronal Nav (nNav) is sufficient to initiate aberrant, repetitive Ca²⁺ oscillations in wild‐type (WT) atrial myocytes.

A, Representative traces of persistent inward Na⁺ currents (I_Na) recorded in WT atrial cardiomyocytes before (black) and after (red) exposure to β‐pompilidotoxin (β‐PMTX; 40 μmol/L). β‐PMTX increased persistent I_Na relative to control (n=11 and 10 cells from 3 mice, respectively; ***P<0.0001 Wilcoxon rank sum test). B, (Left) Representative line scan images obtained from WT atrial cardiomyocytes exposed to isoproterenol (ISO; 100 nmol/L) and paced at 1 Hz. (Right) Rapid application of β‐PMTX (40 μmol/L) induced aberrant, repetitive Ca²⁺ oscillations. (Median pacing frequency was 0.5 Hz with the range of 0.5 to 2 Hz; number of cells tested depicted under the corresponding bars, from N=3 mice; ***P<0.0001 Fisher exact test for ISO vs ISO–β‐PMTX, ***P<0.0001 Fisher exact test for ISO–β‐PMTX vs ISO‐washout).

We also examined whether blockade of other Na_V isoforms and/or RyR2 in addition to tetrodotoxin‐sensitive Na_V isoforms confers additional benefit beyond that achieved with 100 nmol/L tetrodotoxin. To that end, we used R‐propafenone22, 23, 24, 25 (300 nmol/L) and ranolazine26, 27, 28, 29 (10 μmol/L) to assess their effect on cellular arrhythmia potential. At these concentrations, ranolazine and R‐propafenone achieved similar levels of peak I_Na reduction relative to riluzole (peak I_Na at −40 mV of −22.4±3.9 pA/pF versus −23.9±2.7 pA/pF and −23.3±1.5 pA/pF for isoproterenol+riluzole versus isoproterenol+ranolazine and isoproterenol+R‐propafenone, respectively; Figures 2A and 3A). Furthermore, both of these agents reduced induction of aberrant Ca²⁺ oscillations (Figure 3C and 3D). Notably, the extent of reduction in cellular arrhythmia propensity was proportional to the extent of late I_Na reduction achieved by these agents (Figure 3B), suggesting that nNav blockade is sufficient for the antiarrhythmic effect of Na_V blockers. Furthermore, the effects of these 2 compounds were, in part, washable (Figure 3C and 3D).

Figure 3. The extent of late inward Na⁺ currents (I_Na) inhibition corresponds to prevention of aberrant, repetitive Ca²⁺ oscillations in R33Q atrial myocytes.

A, (Left) I_Na obtained by step protocol illustrated in 1‐second intervals. (Right) In isoproterenol (ISO; 100 nmol/L)‐treated R33Q atrial myocytes, addition of ranolazine (Ran; 10 μmol/L, blue trace and bar; n=5 from N=4 animals) or R‐propafenone (R‐prop; 300 nmol/L, orange trace and bar; n=4 from N=3 animals) significantly reduced peak I_Na (P=0.0016 ANOVA, *P=0.0418 for ISO vs ISO+Ran and *P=0.0246 for ISO vs ISO+R‐prop) relative to ISO alone (n=7 from N=5 animals). Notably, there was no difference in peak I_Na reduction between the groups. B, (Left) Representative persistent I_Na elicited using the protocol shown in the inset. (Right) ISO (100 nmol/L) increased persistent I_Na, while Ran and R‐prop reduced it (n=13, 13, 7, and 7 cells from 6, 6, 6 and 4 mice for R33Q, ISO, ISO+Ran, and ISO+R‐prop, respectively; P<0.0001 ANOVA, ***P<0.0001, **P=0.0082, and *P=0.0480). Notably, R‐prop reduced ISO‐induced persistent I_Na to a greater extent than Ran (P=0.0480). C, Treatment of ISO (100 nmol/L)‐exposed R33Q atrial myocytes with Ran (10 μmol/L) significantly reduced the incidence of aberrant, repetitive Ca²⁺ oscillations, which was washable (number of cells tested depicted under the corresponding bars, N=5 animals; ***P=0.0009 McNemar test for ISO vs ISO‐Ran and for ISO‐Ran vs ISO‐washout). D, R‐prop abolished aberrant, repetitive Ca²⁺ oscillations, an effect that was only partially washable only after 10 minutes. (Number of cells tested depicted under the corresponding bars, N=4 animals; ***P=0.0005 McNemar test for ISO vs ISO–R‐prop, *P=0.0455 for ISO–R‐propafenone vs ISO‐washout).

Next, we confirmed the importance of NCX to the proarrhythmic process in this model.6, 11 Acute NCX inhibition (5 mmol/L NiCl₂) abolished repetitive Ca²⁺ oscillations in all cells tested (Figure S3A). Since we were unable to electrically stimulate cardiomyocytes in the presence of NiCl₂, we repeated these experiments during NCX inhibition with SEA0400 (1 μmol/L). SEA0400 prevented induction of arrhythmogenic Ca²⁺ oscillations in over two thirds of cells tested (Figure S3B).

nNa_vs Closely Associate With Ca²⁺‐Handling Machinery

In order to examine close proximity of nNa_v with SR Ca²⁺‐release machinery, which may contribute to aberrant Ca²⁺ release through NCX in atrial myocytes, we performed confocal microscopy. This approach identified multiple nNa_v isoforms (Na_v1.1, Na_v1.3, and Na_v1.6) localized near RyR2 and NCX in atrial myocytes from R33Q hearts (Figure 4A and 4B, top). Proximity ligation assays10 confirmed close association (within 40 nm)30 of all 3 nNa_v isoforms with RyR2 and NCX (Figure 4 and 4B, bottom). These results place nNa_vs near enough to leaky RyR2 to promote arrhythmias via aberrant NCX.

Figure 4. Neuronal Na⁺ channel (nNa_V) and ryanodine receptor 2 (RyR2) colocalize to the same discrete subcellular regions.

Representative confocal micrographs of myocytes isolated from R33Q mice labeled for (A) RyR2 (red) and (B) Na⁺/Ca²⁺ exchange (NCX; red) with various Na⁺ channel (Na_v) isoforms (Na_v1.x, green). These often resulted in an overlap between the immunofluorescent signals (yellow) when overlaid. (Right) Close‐up views of regions highlighted by dashed white boxes. (Bottom) Representative fluorescent proximity ligation assay signal for RyR2 (A) and NCX (B) with different nNa_v isoforms (Na_V1.x).

Augmentation of nNa_v Activity is Proarrhythmic

Next, we examined whether augmented nNa_v–mediated Na⁺ influx is sufficient to induce arrhythmogenic Ca²⁺ oscillations. Exposing WT myocytes to isoproterenol alone was insufficient to elicit Ca²⁺ oscillations. However, persistent I_Na induced by nNa_v augmentation (β‐pompilidotoxin, 40 μmol/L; Figure 5A)10, 31 promoted arrhythmogenic Ca²⁺ oscillations in WT cardiomyocytes exposed to isoproterenol (Figure 5B). This aberrant Ca²⁺ release resulted in a reduced SR Ca²⁺ load (Figure S4). Taken together, these data are consistent with the notion that enhanced nNa_v‐mediated Na⁺ influx is necessary as well as sufficient to produce proarrhythmic Ca²⁺ oscillations in WT atrial myocytes.

nNa_vs Modulate Atrial Arrhythmias in Mice

To determine the effects of nNa_v inhibition on atrial arrhythmias in mice with leaky RyR2,32 we treated them with riluzole (15 mg/kg IP).33 Atrial arrhythmia inducibility by atrial‐burst pacing was reduced by half following riluzole treatment (Figure 6A). In contrast, augmentation of nNa_v‐mediated Na⁺ influx by β‐pompilidotoxin (40 mg/kg IP) promoted atrial arrhythmias in WT mice: caffeine and epinephrine challenge induced atrial arrhythmias in 58% (7 of 12) of β‐pompilidotoxin–treated mice, compared with 15% (2 of 13) of untreated controls (Figure 6B). Taken together, these data suggest that perturbing local Na⁺‐Ca²⁺ crosstalk via modulation of nNa_v potently regulates atrial arrhythmia risk in vivo.

Figure 6. Modulation of tetrodotoxin (TTX)‐sensitive neuronal Nav (nNav) channel correspondingly modulates atrial arrhythmias in mice.

A, Simultaneous surface ECG (lead II) and intracardiac atrial electrograms with frequent, rapid P waves and irregular RR intervals suggestive of atrial arrhythmia such as atrial flutter and atrial fibrillation in R33Q mice after burst pacing. Pretreatment with riluzole (Ril; 15 mg/kg IP), targeting plasma concentrations of ~10 μmol/L,33 reduced the atrial arrhythmia inducibility (n=7 mice; **P=0.0160 Wilcoxon signed rank test). B, Representative surface ECG recordings of wild‐type (WT) mice treated (top, red ECG) or untreated (bottom, black ECG) with β‐pompilidotoxin (β‐PMTX; 40 mg/kg IP) and exposed to catecholamine challenge with epinephrine (Epi, 1.5 mg/kg) and caffeine (Caff, 120 mg/kg). Since increased heart rate has been linked to reduced arrhythmia inducibility in calsequestrin null mice, and WT mice show higher heart rate relative to calsequestrin null mice,32 all WT animals were pretreated with ivabradine (3 mg/kg) for 10 minutes before any intervention. Epi+Caff challenge during β‐PMTX exposure precipitated repetitive P waves and irregular RR intervals suggestive of atrial arrhythmia in over 50% of WT mice, which is a 3‐fold increase relative to β‐PMTX–untreated mice (number of mice tested and those positive for atrial arrhythmias depicted under the corresponding bars; *P=0.0410 Fisher exact test).

Targeting nNa_vs With Riluzole Prevents Arrhythmias in a Canine Cardiomyopathy Model

Next, we examined the contribution of nNa_vs to atrial arrhythmogenesis in a clinically relevant chronic tachypacing‐induced canine cardiomyopathy model (4 months of tachypacing).4 This model allowed us to test the relevance of nNa_v blockade with riluzole in a much more complex pathology that goes beyond leaky RyR2. Riluzole (10 μmol/L) reduced the integral of persistent I_Na in the presence of isoproterenol (100 nmol/L) in nonfailing atria by 66±4%. This is consistent with the previous observation that about half of late I_Na in nonfailing canine ventricles is carried by tetrodotoxin‐sensitive nNa_vs.34 Atrial myocytes isolated from cardiomyopathic canines evidenced enhanced persistent I_Na at baseline, relative to controls (Figure 7A and 7B). In contrast to controls, isoproterenol (100 nmol/L) did not further enhance persistent I_Na in failing atrial cells (Figure 7A and 7B). In the context of increased post‐translational modification of Ca²⁺ cycling proteins in this model,4 these results may point to increased post‐translational modification of Na_vs at baseline in failing hearts. Concurrently, proximity ligation assay revealed reduced incidence of Na_V1.1 and Na_V1.3 localizing in proximity to RyR2, while Na_V1.6 localization in proximity to both RyR2 and NCX was significantly increased (Figure S5). Importantly, riluzole (10 μmol/L) reduced the integral of persistent I_Na in failing myocytes to the same level as in nonfailing atria (−67±9 versus −74±13  Amp.ms/F for failing and nonfailing atria, respectively; Figure 7A and 7B). Notably, enhanced persistent I_Na integral in failing atrial myocytes translated into aberrant Ca²⁺ cycling: all isoproterenol‐treated (100 nmol/L) failing atrial myocytes studied evidenced frequent, self‐sustaining Ca²⁺ oscillations (Figure 7C and 7D). Riluzole (10 μmol/L) significantly reduced these events, an effect that was reversed upon washout (Figure 7C and 7D). Taken together, these data suggest the translatability of targeting nNa_vs with riluzole in complex pathologies.

Figure 7. Riluzole (Ril) reduces enhanced, persistent inward Na⁺ currents (I_Na) and prevents induction of aberrant, repetitive Ca²⁺ oscillations in canine heart failure (HF) atrial myocytes.

A, Representative traces of persistent I_Na integral elicited using the protocol shown in the inset. Recordings were made in control (top) and failing (bottom) atrial myocytes before (black) and after exposure to isoproterenol (ISO; 100 nmol/L, red) and after treatment with Ril (10 μmol/L, purple). At baseline, atrial cardiomyocytes from failing hearts showed a larger persistent I_Na integral relative to control. ISO (100 nmol/L) enhanced persistent I_Na only in control cardiomyocytes. Ril reduced persistent I_Na integral in both control and failing atrial myocytes. B, Summary data are presented as persistent I_Na integral Amp‐ms/F (AmsF⁻¹), which was measured by integrating I_Na between 50 and 450 ms (n=7 and 9 cells from 5 control and 3 failing dogs; P<0.0001 Kruskal–Wallis test; *P=0.0126 for control vs ISO, *P=0.0209 for control vs ISO+Ril, *P=0.0268 for control‐ISO vs HF‐ISO, ***P<0.0001). C, Representative examples of the line‐scan images and corresponding Ca²⁺ transients recorded in canine HF atrial cardiomyocytes loaded with Ca²⁺ indicator, Fluo‐3 AM, and paced at 0.5 Hz with field stimulation. (Top) Cells were treated with ISO (100 nmol/L) and subsequently Ril (10 μmol/L; purple bar indicates time when Ril was added) was rapidly applied. (Bottom) Resumption of field stimulation failed to induce Ca²⁺ oscillations during concomitant exposure to ISO and Ril; however, washout of Ril resulted in their reinitiation. D, Ril significantly reduced the incidence of aberrant, repetitive Ca²⁺ oscillations, an effect that was washable (median pacing frequency was 0.5 Hz with the range of 0.5 to 1 Hz; number of cells tested depicted under the corresponding bars, N=3 animals for ISO and ISO‐Ril, N=2 animals for ISO‐washout, respectively; **P=0.0009 McNemar test for ISO vs ISO‐Ril, P<0.0001 Fisher exact test for ISO‐Ril vs ISO‐washout incidence; ***P=0.0005 Friedman rank sum test for Ca²⁺ oscillations frequency).

Targeting nNa_vs With Riluzole Controls New‐Onset AF in Patients With ALS

Based on the results from the canine model, we examined the effect of riluzole on atrial arrhythmias in human patients via a retrospective cohort study. The research cohort consisted of 184 Italian patients prescribed riluzole, 314 US patients prescribed riluzole, and 735 riluzole‐free patients from the United States (Table). Compared with the no‐riluzole patients, patients taking riluzole were age equivalent and had significantly more cardiovascular risk factors, more active disease, and more recorded cardiovascular events. In addition, the riluzole cohort were prescribed significantly more cardiovascular risk prevention and treatment medications, including β‐adrenergic blockers, Ca²⁺ channel blockers, aspirin, angiotensin system antagonists, and digitalis. The trends were consistent from both overall and US cohort analyses.

Table 1.

Baseline Characteristics

	Riluzole, %		No Riluzole, %	Riluzole vs No Riluzole P Value
	Italy and United States (n=501)	United States (n=314)	United States (n=735)	Overall Cohort	United States
Age ≥65 y	44.6	40.4	43.7	0.7475	0.3331
Concurrent conditions
Hypertension	25.5	14.0	7.2	<0.0001	0.0005
Hyperlipidemia	12.1	6.4	2.9	<0.0001	0.0072
Diabetes mellitus	7.8	7.0	5.0	0.0481	0.2042
AMI or ACS	3.0	0.0	0.8	0.0036	0.1084
Active smoking	15.1	2.2	0.0	<0.0001	<0.0001
CAD	9.2	1.6	1.2	<0.0001	0.6344
HF	7.6	0.3	0.8	<0.0001	0.3644
Stroke	3.4	0.0	0.0	<0.0001	n/a
Current medication
β‐Blocker	6.6	0.3	0.0	<0.0001	0.1258
CCB	5.6	0.0	0.0	<0.0001	n/a
ASA	20.0	15.6	5.7	<0.0001	<0.0001
Statins	12.8	10.8	3.0	<0.0001	<0.0001
ACEI or ARB	20.8	11.1	2.3	<0.0001	<0.0001
Digitalis	0.8	0.0	0.0	0.0146	n/a

ACEI indicates angiotensin‐converting enzyme inhibitor; ARB, angiotensin receptor blocker; ASA, acetyl salicylic acid (aspirin); CAD, coronary artery disease (other than acute myocardial infarction [AMI] or acute coronary syndrome [ACS]); CCB, calcium channel blocker; HF, heart failure; and n/a, not applicable.

Patients with ALS treated with riluzole had significantly fewer overall cardiac arrhythmias. Five of 498 patients taking riluzole recorded tachyarrhythmia events versus 31 end points of 735 from the no‐riluzole cohort over the maximum follow‐up of 12 years, which resulted in the crude and adjusted HR of 0.28 (95% CI, 0.11–0.73; P=0.0088) and 0.25 (95% CI, 0.07–0.85; P=0.0272), respectively. The respective estimates when the analysis was limited to the US cohort were 0.25 (95% CI, 0.08–0.81; P=0.0210) and 0.21 (95% CI, 0.06–0.72; P=0.0129). The majority of arrhythmic events were distributed equivalently between supraventricular and ventricular tachycardia. The Kaplan–Meier curves of the arrhythmia events are presented in Figure 8, which reveals that arrhythmia events were reduced early and primarily throughout the first 2 years after riluzole prescribing. The majority of patients taking riluzole who recorded a tachyarrhythmia did not have cardiovascular disease. The rate of AF was lower in the rilozule group relative to the no‐riluzole group (P=0.0492). Specifically, AF occurred in 22 patients with ALS: 13 from Italy and 9 from the United States. One AF event was recorded after the riluzole prescription versus 9 events from the no‐riluzole cohort. Most patients who encountered AF had underlying cardiovascular disease (77%), with the majority treated with an angiotensin system antagonist; however, only 3 were treated with a β‐adrenergic blocker.

Figure 8. Riluzole prevents cardiac arrhythmias in patients with amyotrophic lateral sclerosis (ALS).

Two retrospective cohorts of ALS one exposed to riluzole vs no riluzole (controls), were compared by Cox proportional hazard models. The time‐to‐first composite arrhythmic events was analyzed using Kaplan–Meier production limit estimator. A, Overall cohort, (B) US cohort. HR indicates hazard ratio.

Riluzole safety was determined during the study period by recording any abnormal neutrophil count, or transaminase level (ie, alanine aminotransferase/serum glutamic pyruvic transaminase, aspartate aminotransferase/serum glutamic oxaloacetic transaminase, bilirubin, or gamma‐glutamyl transferase levels). There was no record of any abnormal laboratory value in the study cohort.

Discussion

AF is the most common sustained cardiac arrhythmia.35 It is postulated that early use of rhythm control strategies may prevent arrhythmia progression.36 Current antiarrhythmic drugs such as flecainide or amiodarone are moderately beneficial in restoration and maintenance of sinus rhythm but produce serious adverse effects such as ventricular tachycardia, negative inotropy, and extracardiac toxicity.37, 38 Therefore, there is a need for an effective and safe alternative to current antiarrhythmic therapy for AF. In this study, we have examined a nNa_v‐mediated mechanism for atrial arrhythmias in various preclinical models. These studies identify nNa_vs as a druggable target for safe and effective atrial arrhythmia prevention. To translate these results into a real‐world setting, we undertook a retrospective cohort study of patients with ALS. Riluzole‐treated patients with ALS had fewer AF and ventricular arrhythmias than those undergoing conventional ALS therapy without riluzole. Taken together, these findings suggest that riluzole merits serious consideration as treatment for atrial arrhythmias.

Role of nNa_vs in Atrial Arrhythmias

Here, we demonstrate for the first time that nNa_Vs play a key role in the development of atrial arrhythmias in the presence of genetic (murine‐CPVT; Figure 1) or acquired (canine‐heart failure; Figure 7) Ca²⁺ handling dysfunction.6, 39 nNa_v blockade effectively suppressed atrial arrhythmias on the cellular level as well as in vivo (Figure 6). Contrariwise, acute experimental augmentation of nNa_v function was sufficient to precipitate cellular arrhythmias in healthy atrial myocytes (Figure 5). Further, augmentation of nNa_v activity reduced SR Ca²⁺ load (Figure S4), while inhibition of these channels had the opposite effect (Figure S1). This argues against global SR Ca²⁺ overload being the mechanism underlying the observed arrhythmias. Of note, blockade of other Na_V isoforms, including Na_V1.5, with R‐propafenone22, 23, 24, 25 or ranolazine26, 27, 28, 29 in addition to nNa_Vs did not confer any apparent additional benefit beyond that achieved with 100 nmol/L of tetrodotoxin. This suggests that blockade of nNa_Vs is an important component of the antiarrhythmic mechanism of Na_V blockers.

Our structural results point to the close proximity of nNa_vs to Ca²⁺‐handling machinery (RyR2 and NCX) as a potential factor underlying their privileged role in arrhythmogenesis (Figure 4 and Figure S5). In light of these findings, recent observations in patients with AF, which have revealed a reduction in Na_v1.5 and upregulation of nNa_Vs,7 take on interesting implications. Mainly, remodeling within Na⁺/Ca²⁺ nanodomain, composed of nNa_Vs and Ca²⁺‐handling machinery (Figure S5), may potentially compensate for failing excitation‐contraction coupling. Inversely, the late I_Na carried by these channels (Figure 7A and 7B) and the consequent NCX, can facilitate aberrant Ca²⁺ release through sensitized leaky RyR2 (Figure 7C and 7D).40, 41 This is in line with other reports of increased nNa_V and enhanced late I_Na in failing rat, canine, and human hearts.7, 9, 42 Hence, as proposed by our study, nNa_v blockade can be an effective antiarrhythmic strategy independent of remodeling within Na⁺/Ca²⁺ nanodomain or the pathogenesis of leaky RyR2.7, 10, 17, 18, 43 Furthermore, since ranolazine has been previously demonstrated to substantially affect multiple tetrodotoxin‐sensitive and ‐resistant Na_V isoforms (Na_V1.1, Na_V1.4, Na_V1.5, Na_V1.7, and Na_V1.8)26, 27, 28, 29 at concentrations that are achieved therapeutically (<10 μmol/L), tetrodotoxin‐sensitive nNa_V blockade may, in part, be an explanation for the effectiveness of ranolazine in reducing late I_Na in failing human atria.7 However, despite our study pointing to nNa_vs as antiarrhythmia targets, future research will need to determine the specific Na_v isoform, or the combination thereof, necessary for the antiarrhythmic effect of riluzole and other Na_V blockers, such as ranolazine. Taken together, these results indicate a direct role for nNa_V‐mediated Na⁺ influx in arrhythmia initiation, rather than it acting simply as a compounding factor.

Noteworthy, riluzole reduced persistent I_Na integral in nonfailing atria by 66±4%. This is consistent with previous observations that about half of late I_Na in canine ventricles is comprised of tetrodotoxin‐sensitive nNa_vs.34 Furthermore, riluzole reduced persistent I_Na by similar absolute extents in both failing and nonfailing atrial myocytes. This corresponded to a greater proportional impact on persistent I_Na integral in failing myocytes (78±3%) versus nonfailing myocytes (66±4%) as a result of enhanced persistent I_Na in failing atria. Together, these data indicate a role for nNa_V remodeling within Ca²⁺‐handling nanodomains and increased post‐translational modification of Na_Vs in failing atria. This notion is consistent with findings in atria from patients with chronic AF, where an increase in nNa_V isoforms accounted for increased late I_Na in AF.7 However, despite the aforementioned studies providing a parallel to our results, we cannot rule out the potential involvement of Na_V1.5 in AF nor potential Na_V1.5 inhibition by riluzole to the drug's antiarrhythmic mechanism.

Riluzole‐Treated Human Patients are Protected From AF

Whereas both tetrodotoxin and riluzole effectively suppressed atrial arrhythmias in animal models, concerns over toxicity render tetrodotoxin an untenable clinical option. In contrast, riluzole has a proven safety record as a treatment for ALS. In line with this, our data also suggest that riluzole was well tolerated. Hence, in our retrospective cohort study of arrhythmia risk among patients with ALS, riluzole‐treated patients with ALS had both fewer AF and ventricular tachycardia diagnoses compared with nontreated patients (Figure 8). To our knowledge, this is the first evidence to show that riluzole may have antiarrhythmic properties in humans. However, because of the nonrandomized nature of this study, further controlled studies are necessary to confirm and extend this finding. Given that ventricular proarrhythmia is a critical limitation of current AF drug therapies,13, 14, 15 riluzole and potentially other nNa_v inhibitors may provide an urgently needed safe alternative.

Limitations

A major limitation of small animal models is the translatability of findings into clinically relevant models of human disease. This factor is mitigated by our data demonstrating the antiarrhythmic efficacy of riluzole in a chronic tachypacing‐induced canine cardiomyopathy model. Riluzole can activate small‐conductance Ca²⁺‐activated K⁺ channels,44 which are upregulated in heart failure.45 While the possible contribution of this effect to riluzole's antiarrhythmic mechanism is not clear, our results with other Na_V inhibitors suggest that it is not necessary. Whether small‐conductance Ca²⁺‐activated K⁺ channel activation contributes to riluzole's antiarrhythmic effect will be an interesting subject for future study. Na⁺ channel blockers such as flecainide, R‐propafenone, or ranolazine can block RyR2, mitigating the impact of leaky RyR2.25, 29, 46 However, riluzole (10 μmol/L) did not affect Ca²⁺ sparks in permeabilized ventricular cardiomyocytes, a surrogate for RyR2 function.17 Even so, this mechanism merits further study. It is important to note that our retrospective cohort study of patients with ALS helps translate findings from animal models to humans. However, the study's observational and nonrandomized nature offers limited insight into the causality of observed effects. This limitation was mitigated, in part, by adjusting outcomes for baseline characteristics. Finally, outcomes were retrospectively obtained and were only available from existing data; therefore, we were not able to include any arrhythmia observations outside of the databases. Since other studies demonstrated similar incidence of arrhythmias,47 and valid results despite a similar limitation, we do not believe there would be differential ascertainment between the 2 cohorts.48 However, because of the unique character of the population (ie, ALS), further controlled studies are necessary to confirm and extend our findings.

Conclusions

We used experimental and preclinical animal models to delineate a mechanistically driven therapeutic strategy for atrial arrhythmias and provide evidence for its translational potential from a community‐based retrospective cohort. Specifically, we identify a nanodomain rich in nNa_Vs and Ca²⁺‐handling machinery (NCX and RyR2) that forms the basis of aberrant, self‐sustained Ca²⁺ release, resulting in atrial arrhythmias in vivo. Importantly, inhibition of nNa_vs with riluzole demonstrates efficacy in preventing atrial arrhythmias in both animal models and human patients. Thus, riluzole has the potential to be repurposed as a therapy for preventing AF.

Sources of Funding

This work was supported by National Institutes of Health grants HL074045, HL063043 (to Györke), and HL127299 (to Radwański), as well as a Saving tiny Hearts grant and American Heart Association 19TPA34910191 (to Radwański).

Disclosures

Kim reports research grants from AstraZeneca, Myriad Genetics, and Pacira Pharmaceuticals. The remaining authors have no disclosures to report.

Supporting information

Data S1 Figures S1–S5 References 49–51

(J Am Heart Assoc. 2020;9:e015119 DOI: 10.1161/JAHA.119.015119.)