Prokaryotic voltage-gated sodium channels are more effective than endogenous Na_v1.5 channels in rescuing cardiac action potential conduction: an in silico study

Abstract

Methods to augment Na⁺ current in cardiomyocytes hold potential for the treatment of various cardiac arrhythmias involving conduction slowing. Because the gene coding cardiac Na⁺ channel (Na_v1.5) is too large to fit in a single adeno-associated virus (AAV) vector, new gene therapies are being developed to enhance endogenous Na_v1.5 current (by overexpression of chaperon molecules or use of multiple AAV vectors) or to exogenously introduce prokaryotic voltage-gated Na⁺ channels (BacNa_v) whose gene size is significantly smaller than that of the Na_v1.5. In this study, based on experimental measurements in heterologous expression systems, we developed an improved computational model of the BacNa_v channel, Na_vSheP D60A. We then compared in silico how Na_vSheP D60A expression vs. Na_v1.5 augmentation affects the electrophysiology of cardiac tissue. We found that the incorporation of BacNa_v channels in both adult guinea pig and human cardiomyocyte models increased their excitability and reduced action potential duration. When compared with equivalent augmentation of Na_v1.5 current in simulated settings of reduced tissue excitability, the addition of the BacNa_v current was superior in improving the safety of conduction under conditions of current source-load mismatch, reducing the vulnerability to unidirectional conduction block during premature pacing, preventing the instability and breakup of spiral waves, and normalizing the conduction and ECG in Brugada syndrome tissues with mutated Na_v1.5. Overall, our studies show that compared with a potential enhancement of the endogenous Na_v1.5 current, expression of the BacNa_v channels with their slower inactivation kinetics can provide greater anti-arrhythmic benefits in hearts with compromised action potential conduction.

NEW & NOTEWORTHY Slow action potential conduction is a common cause of various cardiac arrhythmias; yet, current pharmacotherapies cannot augment cardiac conduction. This in silico study compared the efficacy of recently proposed antiarrhythmic gene therapy approaches that increase peak sodium current in cardiomyocytes. When compared with the augmentation of endogenous sodium current, expression of slower-inactivating bacterial sodium channels was superior in preventing conduction block and arrhythmia induction. These results further the promise of antiarrhythmic gene therapies targeting sodium channels.

INTRODUCTION

Arrhythmia-related sudden cardiac death remains one of the leading causes of mortality in the United States (1). In poorly excitable or fibrotic myocardium, arrhythmias are frequently initiated because of slow action potential (AP) conduction, unidirectional conduction block, and the formation of rotating spiral waves. One potential strategy to improve compromised AP conduction in diseased myocardium is to enhance the amplitude of endogenous sodium channel (Na_v1.5) current, which is yet to be accomplished by pharmacological means. Gene therapies, on the other hand, can specifically target Na_v1.5 channel transcription through intron deletion (2) or CRISPR-Cas9 editing (3) or rescue channel trafficking via overexpression of chaperon protein MOG1 (4, 5) or the common polymorphism H558R (6, 7). However, such methods are either inefficient, suboptimal for patients carrying dominant negative variants of SCN5A (8), or inadequate when Na_v1.5 channels are inactivated by membrane depolarization (e.g., in myocardial ischemia and infarction). An alternative strategy to enhance cardiomyocyte excitability and achieve antiarrhythmic effects is the expression of exogenous voltage-gated sodium channels (VGSCs) with tailored biophysical properties. Pioneering studies from the Rosen laboratory have shown that the forced cardiomyocyte expression of Na_v1.4, the VGSC from skeletal muscle that fires at more depolarized potentials than Na_v1.5, can restore the excitability of ischemic cardiac tissue and decrease the incidence of reentrant arrhythmias (9–14). However, both Na_v1.4 and Na_v1.5 genes are too large to be packaged in single adeno-associated virus (AAV) vectors used for stable human heart expression, and while Na_v1.5 delivery via two AAV vectors is possible (15), its efficacy as an antiarrhythmic therapy remains to be shown.

Recently, Nguyen et al. explored the use of bacterial VGSCs (BacNa_vs) to improve cardiac AP conduction in various physiological and pathological settings in vitro (16–18). BacNa_vs can be engineered to operate in a selectable range of membrane voltages with various kinetics and have small gene sizes (<1 kb) amenable to efficient AAV delivery (19). An engineered BacNa_v channel derived from Shewanella putrefaciens (Na_vSheP D60A), in particular, was shown to successfully augment cardiomyocyte excitability in arrhythmogenic conduction settings in vitro, such as hyperkalemia, in which KCl was added to depolarize resting membrane potential and inactivate the Na_v1.5 channel. The stable expression of BacNa_v in cardiomyocytes was also shown to improve AP conduction in the in vitro models of regional ischemia or myocardial fibrosis, further demonstrating BacNa_v’s potential for use in anti-arrhythmic cardiac gene therapies (16, 17).

In the current study, we constructed an updated computational model of the Na_vSheP D60A channel using voltage- and current-clamp data recorded from a HEK293 cell line stably expressing K_ir2.1, Cx43, and Na_vSheP D60A (KirCxSheP293 cell line) (16). The new model more faithfully replicates the I-V relationship, the AP trace, and restitution properties observed experimentally. Using this model, we sought to compare therapeutic mechanisms of BacNa_v expression vs. equivalent Na_v1.5 augmentation by incorporating the BacNa_v and Na_v1.5 channel descriptions at multiple conductance levels into two adult cardiomyocyte models: the Luo–Rudy guinea pig model and the O’Hara–Rudy human model. Specifically, we first determined the short-term and long-term changes in endogenous ionic currents, excitability, conduction, AP shape, and restitution as a function of added BacNa_v or Na_v1.5 conductance level. We further compared how the BacNa_v expression or Na_v1.5 augmentation in cardiac tissue influences: 1) vulnerability to conduction block due to source-load mismatch or premature pacing, 2) AP conduction stability at high excitation rates and during spiral-wave reentry, and 3) transmural conduction and ECG in cardiac tissues with mutated Na_v1.5 channel (Brugada syndrome). The results of these studies show that compared with augmentation of endogenous Na_v1.5 current, exogenously expressed, slower-inactivating BacNa_v channels more effectively support impulse propagation in critical regimes of depressed or heterogeneous action potential conduction and prevent spiral wave instabilities leading to ventricular fibrillation.

METHODS

Whole Cell Patch Clamp

Dissociated KirCxSheP293 cells were plated on an Aclar coverslip and left to attach for 12–24 h at 37°C. Patch pipettes were fabricated with tip resistances of 1–2 MΩ when filled with pipette solution. The whole cell patch-clamp recordings were performed at 37°C using the Multiclamp 700B amplifier (Axon Instruments), filtered with a 10-kHz Bessel filter, digitized at 40 kHz, and analyzed using WinWCP software (John Dempster, University of Strathclyde). For sodium current recordings, the bath solution consisted of (in mM) 135 NaCl, 1.8 CaCl₂, 1.2 MgCl₂, 2 NiCl₂, 10 HEPES, and 10 glucose; the pipette solution consisted of (in mM) 115 CsCl, 10 NaCl, 0.5 MgCl₂, 10 TEA-Cl, 10 EGTA, 10 HEPES, and 5 MgATP. To measure the activation properties of voltage-gated sodium channels, membrane voltage was stepped from a holding potential of −80 mV to varying 500-ms test potentials (60 to −50 mV, increments of −10 mV). Inactivation of voltage-gated sodium channels was derived from peak currents measured at 0 mV after varying 10-s prepulse potentials (−30 to −60 mV, increments of −10 mV). For the AP recordings, the bath solution consisted of (in mM) 135 NaCl, 5.4 KCl, 1.8 CaCl₂, 1 MgCl₂, 0.33 NaH₂PO₄, 5 HEPES, and 5 glucose. The pipette solution contained (in mM) 140 KCl, 10 NaCl, 1 CaCl₂, 2 MgCl₂, 10 EGTA, 10 HEPES, and 5 MgATP. Action potentials were elicited using a 1 ms pulse at 1.1× the threshold amplitude. Recordings were converted to .mat files for MATLAB using the WinWCP software (John Dempster, University of Strathclyde) and analyzed as previously described (16, 20).

Optical Mapping of AP Propagation

Confluent monolayers of K_irCxSheP293 cells (HEK293 cell line stably expressing K_ir2.1, Cx43, and Na_vSheP D60A) were optically mapped using a hexagonal array of 504 optical fibers (Redshirt Imaging), as previously described (16, 18, 20, 21). Briefly, cells were stained with 10 μM Di-4-ANEPPS for 5 min at room temperature and transferred to a temperature-controlled (37°C) recording chamber filled with Tyrode’s solution. Action potential propagation was initiated by 10 ms, 1.2× threshold stimuli using a bipolar point electrode. Fluorescence signals were converted to voltage by photodiodes and acquired at a 2.4-kHz sampling rate with 750-μm spatial resolution. The pacing rate was increased every 10 s in steps of 1 Hz until the maximum capture rate and optically measured transmembrane potentials were recorded at each frequency to construct restitution curves. Conduction velocity (CV) and action potential duration (APD) were determined using custom MATLAB software, as previously described (22, 23).

Model Development

Membrane voltage for single excitable cells (K_irCxSheP293, guinea pig ventricular myocyte, human ventricular myocyte) was modeled using the following differential equation:

$$C_{\mathrm{m}}\frac{\mathrm{d}{V}_{\mathrm{m}}}{\mathrm{d}t}\mathrm{\hspace{0.17em}}=\mathrm{\hspace{0.17em}}-\left(I_{\mathrm{s}\mathrm{t}\mathrm{i}\mathrm{m}}+I_{\mathrm{i}\mathrm{o}\mathrm{n}\mathrm{i}\mathrm{c}}\right),$$ (1)

Where C_m is the membrane capacitance per unit area, V_m is the transmembrane potential, I_stim is the externally applied stimulus current per unit area, and I_ionic is the sum of the individual ionic currents per unit area that contribute to the AP. The K_irCxSheP293 cell model used the same ion channels and parameters described by Gokhale et al. (24), while the guinea pig and human currents were implemented as originally described (25–28). In the non-K_irCxSheP293 models, potassium was the charge carrier for the stimulus current, and the stimulus current was included in the total ion movement for potassium only.

The BacNa_v channel current (I_Na,BacNav) was modeled with activation (m) and inactivation (h) gate as follows:

$$I_{\mathrm{N}\mathrm{a},\mathrm{B}\mathrm{a}\mathrm{c}\mathrm{N}\mathrm{a}\mathrm{v}}\mathrm{\hspace{0.17em}}=\mathrm{\hspace{0.17em}}{G}_{\mathrm{N}\mathrm{a}}{m}^{3}h(V_{\mathrm{m}}-E_{\mathrm{N}\mathrm{a}}),$$ (2)

where G_Na is the maximum conductance through the channel, V_m is the transmembrane potential of the cell, and E_Na is the Nernst potential for sodium ion. The procedure leading to this choice of model can be found in the Supplemental Data (all Supplemental material, figures, tables, and equations may be found at https://doi.org/10.6084/m9.figshare.24135039). The steady-state values (m_∞, h_∞) and time constants (τ_m, τ_h) for both gating variables m and h were estimated from the patch clamp data using standard protocols (24, 29). Following the initial estimation for these four parameters, the normalized conductance time courses were fit using Eq. 3.

$$g\left(v,t\right)\mathrm{\hspace{0.17em}}=\mathrm{\hspace{0.17em}}{(m}_{\mathrm{\infty }}\left(v_{\mathrm{t}\mathrm{e}\mathrm{s}\mathrm{t}}\right)\mathrm{\hspace{0.17em}}{\left(1-\mathrm{\text{exp}}\left(-\frac{t}{{\tau }_{\text{m}}\left(v_{\mathrm{t}\mathrm{e}\mathrm{s}\mathrm{t}}\right)}\right)\right)}^3\mathrm{*}({(h}_{\mathrm{\infty }}\left(v_{\mathrm{h}\mathrm{o}\mathrm{l}\mathrm{d}}\right)-\left(h_{\mathrm{\infty }}\left(v_{\mathrm{h}\mathrm{o}\mathrm{l}\mathrm{d}}\right)-h_{\mathrm{\infty }}\left(v_{\mathrm{t}\mathrm{e}\mathrm{s}\mathrm{t}}\right)\right)\left(1-\mathrm{\text{exp}}\left(-\frac{t}{{\tau }_{\mathrm{h}}\left(v_{\mathrm{t}\mathrm{e}\mathrm{s}\mathrm{t}}\right)}\right)\right)).$$ (3)

Each parameter was allowed to vary using the initial estimates described above and with bounds from the 95% confidence intervals. Fitting was performed using MATLAB’s Curve Fitting toolbox and the trust-region algorithm for nonlinear least squares optimization (30–32). Upon obtaining the final estimates for each parameter value at each test voltage, the estimates were weighted according to our Fit Quality metric (Supplemental Eq. S7, Supplemental Table S2) and fit with continuous functions of V_m (Supplemental Table S3) to facilitate later simulations. The fits with the best scores (Supplemental Eq. S8) were used for all further simulations. Namely, m and h steady-state curves were fit using a Boltzmann sigmoid function, the activation time constant τ_m was fit using the generalized hyperbolic secant function (16), and the inactivation time constant τ_h was fit using a sum of two sigmoid functions to allow for the tail values in the highly positive and negative membrane voltage domains to be different.

Action Potential Fitting

AP traces recorded in KirCxSheP293 cells expressing K_ir2.1, Cx43, and Na_vSheP D60A, were fit using parameters from Gokhale et al. (24) by substituting Na_v1.5 with the BacNa_v current. Conductance values for each channel were determined using a single-objective particle swarm search algorithm. The objective function was a weighted sum of the root mean squared errors (RMSEs) between the experimental and the simulated AP in the time range from the peak of the AP to the end of the time course, the absolute difference in action potential durations at 80% repolarization (APD₈₀), and upstroke velocities (maximum dV_m/dt). The RMSE was only calculated for the specified time range to avoid artifacts from the stimulus. The swarm size was maintained at 70 (10 times the number of variables), and the initial population was selected based on uniform distributions spanning physiologically feasible values (Supplemental Table S6). The algorithm was configured using MATLAB’s Global Optimization Toolbox and run for a total of one hundred times using the Duke Compute Cluster to ensure better global convergence. The algorithm was run with self-adjustment and social-adjustment weights fixed at 1.49. The final set of equations (Supplemental Eqs. S9–S15) and parameters (Supplemental Table S8) describing the Na_vSheP D60A model are presented in Supplemental Information.

Action Potential Propagation

To model AP propagation, the tissue intracellular domain was represented as a network of resistors (Supplemental Fig. S1). This approach allowed the investigation of AP conduction and conduction velocity restitution in one-dimensional (1-D) and two-dimensional (2-D) continuous domains (Supplemental Fig. S1, A and C), as well as 1-D inhomogeneous domains and 1-D models of discrete cells coupled by gap junctions (Supplemental Fig. S1B). The Supplemental Eqs. S2 were applied to conserve current in the intracellular space.

Restitution Relationship Validation

Simulated APD and CV restitution curves were determined from a 2-D continuous monodomain model (50 × 50 nodes; dx = 24 µm). One edge of the domain was stimulated at the test frequencies ranging from 1 to 6 Hz for 10 s (as in the experimental protocol). The CVs and APDs were recorded at the middle of the domain and the RMSEs were calculated between the simulated and experimentally measured curves.

Incorporation of BacNa_v Into Adult Cardiomyocyte Membrane Models

To study the effects of BacNa_v on the time course of an adult cardiac AP, the channel was incorporated into two established mathematical membrane models: the 2007 version of the Luo–Rudy dynamic guinea pig ventricular model (LRd) (25, 27) and the O’Hara–Rudy human ventricular model (ORd) (28). Based on neonatal rat cardiomyocyte experiments performed by Nguyen et al. (16), the endogenous sodium channel (Na_v1.5) conductance was assumed to be unaffected by the incorporation of BacNa_v. The action potential time courses were evaluated as a function of the magnitude of BacNa_v conductance to mimic different gene expression levels. The baseline BacNa_v conductance level (1×) was such that its peak of the I-V curve was equal to that of the endogenous Na_v1.5 current. The additional BacNa_v levels studied were 0.2×, 0.5×, 1.5×, and 2× relative to this baseline value. For comparison, we studied nominal ventricular myocytes with correspondingly added levels of Na_v1.5 conductance.

To ensure the correct implementation of these detailed models, simulations were performed by adding the new BacNa_v current to the MATLAB codes provided by the authors. The effects of the added channels on upstroke velocity, AP amplitude (APA), APD, and resting membrane potential (RMP) were evaluated during continuous pacing at 1 Hz for the five conductance levels of BacNa_v or Na_v1.5 until the maximum change in the state variables was less than 0.01% or until 500 paces had occurred. The resulting steady-state values were stored and used as initial conditions for subsequent simulations. These initial studies were performed using a continuous monodomain 1-D fiber that was 0.5 cm in length (50 nodes; dx = 100 µm) to minimize boundary effects while maintaining computational feasibility.

Current Source-Load Mismatch

To evaluate the impact of the expressed BacNa_v or augmented Na_v1.5 current on the safety factor of conduction, we constructed a fiber of 200 discrete cells coupled by gap junctions (33). A central fiber region of 100 cells was poorly coupled (46.7-fold higher gap junction resistivity) compared with normally coupled surrounding cell regions (50 cells at each end of the fiber). The intracellular space of each cell consisted of 10 nodes (dx = 0.001 cm). As previously shown (33), conduction failure in this setting occurred as the action potential exited the poorly coupled region (site of source-load mismatch). In all cells, the BacNa_v and Na_v1.5 channel were incorporated at seven conductance levels: 0× (base), 0.1×, 0.2×, 0.5×, 1×, 1.5×, or 2× for a total of 7 × 7 = 49 combinations. For these 49 cases, we determined the success of conduction over the mismatch, the safety factor of conduction in the first post-mismatch myocyte, and the conduction delay between the last pre- and the first post-mismatch myocyte. Specifically, the safety factor of conduction was calculated using the method described by Boyle and Vigmond (34), but with the upper time limit of the integration shifted to the peak of the second time derivative of V_m. This increased integration time limit better captured the dynamics of BacNa_v which was slower than that of Na_v1.5 and for the normal time limits led to <1 safety factor even when propagation through the fiber was successful. Furthermore, to assess how different sodium channel kinetics influence AP conduction over the source-load mismatch, we evaluated the success of conduction, safety factor, and conduction delay for the 0.5× Na_v1.5 + 0.5× BacNa_v case with varying activation (τ_m) and inactivation (τ_h) gate time constants of BacNa_v or Na_v1.5 at 0.1×, 0.2×, 0.5×, 1×, 2×, 5×, and 10× of the nominal values (the total of 49 cases for each of Na_v1.5 or BacNa_v).

Vulnerable Window for Unidirectional Conduction Block

The vulnerable window for unidirectional conduction block was determined using the method described by Shaw and Rudy (35). A computational fiber of 300 serially arranged LRd ventricular myocytes was paced from one end (S1; basic cycle length, BCL = 1,000 ms) followed by a premature stimulus (S2, twice the diastolic threshold) applied at the fiber center with varied S1–S2 coupling intervals (CI = 180–450 ms). For each premature stimulus, the fiber was evaluated for the occurrence of bidirectional conduction, unidirectional conduction block, or bidirectional conduction block. Specifically, AP traces from nodes 0.25 cm before and after the central node were analyzed to determine whether an AP fired after the S2 stimulus, evidenced by a depolarization (dV_m/dt > 0) lasting > 0.5 ms that led to a peak V_m > 0 mV. Responses to S2 stimuli at either end of the vulnerable window were further evaluated manually to ensure accurate classification. This protocol was performed for every combination of incorporated 0× (control), 0.1×, 0.2×, 0.5×, 1×, 1.5×, and 2× BacNa_v levels and the equivalent Na_v1.5 levels, for a total of 49 combinations. In addition, the maximum steepness of the spatial repolarization gradient (σ) was calculated for the S1-induced propagation wave for these 49 cases as previously described (36). We also calculated how replacing different fractions of Na_v1.5 current by equivalent levels of BacNa_v current affects vulnerable windows.

Electrical Stability at High Excitation Rates

The stability of LRd model at high excitation rates was tested for 0.5× Na_v1.5 (reduced excitability case), 1× Na_v1.5, and 0.5× Na_v1.5 + 0.5× BacNa_v current levels during both rapid pacing of a 1-D fiber and spiral activity in a 2-D domain. Specifically, a 6-cm-long 1-D LRd fiber (dx = 0.02 cm, 1,200 nodes) was paced from one end at 10 Hz for 25 s. Poincare plots of CL_n + 1 vs. CL_n, APD_n + 1 vs. APD_n, and APA_n + 1 vs. APA_n were constructed from the central node in the fiber during a time window with observed instabilities. For 2-D simulations, we used an isotropic 8.0 cm × 8.0 cm domain (dx = dy = 0.02 cm, 160,000 nodes) with intracellular resistivity set at 450 Ω·cm and fiber radius set at 10 µm. Rotating spiral waves were generated using an S1–S2 protocol whereby an S1 line stimulus was applied at the left edge of the domain followed by a suprathreshold S2 stimulus applied to the top, left quadrant of the domain (37). Circular obstacles of 0.9- and 0.5-cm radius were placed into the center of the domain to evaluate the stability of 2-D electrical activity at different rotation rates of an anchored spiral. No-obstacle domains with free-rotating spirals were used to evaluate the potential for spontaneous spiral breakup by tracking the number of phase singularities with simulation time (averaged in blocks of 500-ms activity). In all 2-D simulations, Poincare plots of CL, APD, and APA were constructed at a peripheral point in the domain (low left corner, x = 0 cm, y = 0.4 cm), except if noted otherwise, to further assess electrical stability during spiral activity.

Brugada Syndrome Model

For modeling a Brugada syndrome (loss-of-function T1620M mutation in SCN5A), we simulated transmural conduction in a 1-D fiber of 165 guinea pig ventricular myocytes as previously described (16, 38). Specifically, the fiber was divided into the endocardial (cells 1–60), midmyocardial (cells 61–105), and epicardial (cells 106–165) region and stimulated at the endocardial end. The three regions were differentiated by the density of the transient outward potassium current (I_to) and the ratio of current density between the slow and rapid rectifying potassium currents (I_Ks:I_Kr). Endocardial cells had zero I_to and an 11:1 ratio of I_Ks:I_Kr; midmyocardial cells had a max I_to of 0.2125 pA/pF and a 4:1 ratio of I_Ks:I_Kr; and epicardial cells had an I_to of 0.25 pA/pF and a 35:1 ratio of I_Ks:I_Kr. Brugada severity was simulated at two levels by increasing both I_to maximum conductance and the speed of fast inactivation for the endogenous Na_v1.5 current as previously described (39). Specifically, mild Brugada case was modeled with 1.5× faster I_Na inactivation and 3× maximum I_to conductance and severe Brugada case was modeled with 3.5× faster I_Na inactivation and 7× maximum I_to conductance. Simulated treatments consisted of incorporation of BacNa_v current, additional mutant Na_v1.5 current, and healthy Na_v1.5 current, all at the 0.5× level. The virtual ECG electrode was placed 2 cm away from the epicardium along the fiber axis (38). Pseudo-ECG signals and deviation from the healthy control ECG were calculated as previously described (16).

Numerical Methods

Simulations of single-cell dynamics were performed using MATLAB and the scripts provided by the Yoram Rudy laboratory on their website. Simulations of propagation in 1-D and 2-D were solved using a custom script that calculated the ionic currents from provided models and applied a finite difference approximation of currents in the spatial domain combined with either forward Euler’s method or MATLAB’s ode15s to solve AP evolution in space and time. If forward Euler was used, the time step was fixed at 5 µs. For the 1-D source-load mismatch, vulnerability window, and 2-D spiral wave simulations, the Cardiowave software package was used (40).

Statistical Analysis

All data are presented as means ± SD unless otherwise specified. Comparisons between different possible sodium channel model formulations are discussed in the Supplemental Information.

RESULTS

Improved BacNa_v Model

Nguyen et al. have recently described a model of the Na_vSheP D60A channel based on whole-cell voltage and current clamp recordings (16). In the current work, we performed patch-clamp experiments and incorporated results from six additional KirCxSheP293 cells to better estimate the dependences of time constants (τ_m, τ_h) and steady-state activation (m_∞) and inactivation (h_∞) curves on the membrane voltage (V_m) in a range from −60 to −20 mV using a new quality metric of the fit (see Supplemental Data). This metric assigns greater weight to patch-clamp fits with higher adjusted correlation coefficients and higher signal-to-noise ratios. Among the different models of BacNa_v conductance, we found that m³h formulation gave the best overall fit of AP (Supplemental Tables S7 and S8) and was adopted for all the subsequent simulations using BacNa_v current. Importantly, compared with Nguyen et al. (16), this newly developed model showed improved fits of the I-V curve, AP trace, and APD and CV restitution relationships derived from optical mapping studies in cell monolayers (Fig. 1). Additional details on optimization and validation of the improved BacNa_v model are provided in Supplemental Information (Supplemental Fig. S2 and Supplemental Tables S1–S8).

Figure 1.

Improved (current) model of BacNa_v (Na_vSheP D60A) outperforms original (Nguyen) model in fitting experimental data. A–D: comparison of current and Nguyen model fits of normalized membrane current-voltage (I-V) curve (A), representative action potential (AP) trace (B), and normalized AP duration (APD; C) and conduction velocity (CV; D) restitution curves recorded in HEK293 cells expressing BacNa_v, Cx43, and K_ir2.1 channels. Normalizations were performed relative to the maximum I, APD, or CV values. E: comparisons for the current and Nguyen model fits in A–D. RMSE, root mean square error.

Effects of BacNa_v Expression vs. Na_v1.5 Augmentation in Ventricular Tissue Models

Normal AP conduction.

Using the newly developed model of BacNa_v channel, we sought to computationally compare the electrophysiological effects of introducing de novo BacNa_v current vs. augmenting endogenous Na_v1.5 current in ventricular myocytes. The expression of BacNa_v and equivalent augmentation of Na_v1.5 were simulated at five different levels (conductance values) using two membrane models, LRd and ORd. One-dimensional computational fibers constructed from these models were paced 500 times (i.e., 8.33 min of 1-Hz pacing) to acquire steady state (Supplemental Fig. S3). At the steady state (Fig. 2), we found that BacNa_v expression yielded a more pronounced increase in action potential amplitude (APA) in both LRd (Fig. 2A) and ORd (Fig. 2B) models compared with equivalent Na_v1.5 augmentation, while resting membrane potential (RMP) was not appreciably changed (up to 0.03%) in any of the cases, consistent with modest changes in intracellular K⁺ concentration ([K⁺]_i). As expected, both BacNa_v and Na_v1.5 yielded an increase in the maximum AP upstroke and conduction velocity (CV) with effects being more pronounced for Na_v1.5 augmentation (Fig. 2), predominantly because of a faster activation kinetics of Na_v1.5_, as well as a higher BacNa_v-induced increase in intracellular Na⁺ concentration ([Na⁺]_i) in the LRd fiber, which resulted in a decrease in E_Na and I_Na driving force (not shown). On the other hand, APD was decreased at low BacNa_v expression levels and then moderately increased for higher BacNa_v expression, while the effects of increased Na_v1.5 on APD were minimal (Fig. 2). In the LRd (but not ORd) fiber, intracellular Ca²⁺ concentration ([Ca²⁺]_i) was also increased as a result of the increased Na⁺ influx and activity of the Na⁺-Ca²⁺ exchanger, more so with the BacNa_v than Na_v1.5 addition but no early or delayed afterdepolarizations were observed in any of the simulations. We also assessed steady-state AP parameters and ionic concentrations at a higher pacing rate (2 Hz, Supplemental Fig. S4) and found that most of the trends in studied parameters found for 1-Hz pacing were maintained during 2-Hz pacing, while baseline (0×) parameters changed as expected {i.e., APD, APA, upstroke, CV, and ([K⁺]_i) were decreased, RMP was depolarized, and [Na⁺]_i and [Ca²⁺]_i were increased during 2-Hz vs. 1-Hz pacing}.

Figure 2.

Effects of BacNa_v expression and Na_v1.5 augmentation on action potential (AP) shape, conduction, and ion concentrations in Luo–Rudy dynamic guinea pig ventricular myocyte model (LRd) and O’Hara–Rudy human ventricular myocyte model (ORd). A and B: steady-state action potential duration (APD), AP amplitude (APA), maximum AP upstroke velocity, conduction velocity (CV), resting membrane potential (RMP), intracellular Na⁺ ([Na⁺]_i), K⁺ ([K⁺]_i), and Ca²⁺ ([Ca²⁺]_i) concentrations in a 1-dimensional (1-D) LRd (A) or ORd (B) fiber (50 nodes; dx = 0.01 cm) as a function of increasing conductance of BacNa_v (blue) or Na_v1.5 (orange) channels. One end of the 1-D fiber was paced 500 times at 1 Hz until a steady state was reached and results for the last (500th) paced AP were shown for the central myocyte.

To better understand the effects of the BacNa_v expression and Na_v1.5 augmentation on the two ventricular myocyte models (LRd in Fig. 3 and ORd in Supplemental Fig. S5), we plotted the AP shapes (Fig. 3A, Supplemental Fig. S5A) and corresponding inward and outward currents for the original (control, 0×) model and with incorporation (+) of 1× BacNa_v or 1× Na_v1.5. Relative current contributions during AP were plotted using the “currentscape” visualization method by Alonso and Marder (41) (Fig. 3B, Supplemental Fig. S5B). With incorporated BacNa_v, relative contributions of inward currents to the total inward current in the LRd model remained the same except for the L-type Ca²⁺ current that showed reduced contribution during the initial BacNa_v activity (Fig. 3B). On the other hand, the relative outward current contributions remained virtually unchanged during the entire AP (Fig. 3B). Doubling the Na_v1.5 conductance in the LRd model had negligible effects on both relative inward and outward current contributions (Fig. 3B). The main effect of the Na_v1.5 doubling was the earlier occurrence of peak I_Na (Fig. 3C, left) due to accelerated charging of the membrane at the subthreshold potentials, while the peak I_Na was reduced because of the relative timing difference between the decrease in I_Na driving force (V_m − E_Na) and increase in G_Na caused by faster rising V_m. The BacNa_v incorporation in the LRd model resulted in marginally decreased peak I_Na, likely because of increased [Na⁺]_i, and a minimally changed timing of peak I_Na (Fig. 3C, left). Past AP upstroke, the absolute inward and outward currents were increased with BacNa_v incorporation (Fig. 3C, left and middle), leading to a biphasic change in the total current with increased inward followed by increased outward current (Fig. 3C, right), leading respectively to increase in early AP plateau and a slight decrease in APD (Fig. 2). Unlike the LRd model, the ORd model showed more prominent current changes in response to BacNa_v expression, including relative increases in I_Ks, I_NaCa, I_Kb, and I_K1 and decrease in I_NaK (Supplemental Fig. S5B). Incorporating BacNa_v in the ORd model also resulted in an earlier peak I_Na (Supplemental Fig. S5C), and, similar to LRd, an increase in the total early inward followed by outward current, which augmented AP amplitude and early plateau without the increase in APD. The equivalent Na_v1.5 augmentation in ORd model caused negligible effects on the AP shape and relative current contributions, an earlier peak I_Na, and, unlike in the LRd model, significantly increased peak I_Na. When compared with LRd model, this peak I_Na increase was the result of a slower decrease in I_Na driving force (V_m − E_Na) relative to an increase in G_Na.

Figure 3.

Effects of BacNa_v expression and Na_v1.5 augmentation on ion currents in the Luo–Rudy dynamic guinea pig ventricular myocyte model (LRd). A: action potential (AP) traces for the control (1× Na_v1.5), added (+1×) BacNa_v, and added (+1×) Na_v1.5 current in LRd myocytes. B: “currentscape” plots showing relative contributions of individual currents (denoted by color segment height at each time point) to total inward and outward current during the AP. C: absolute total (top) and early (bottom) inward (left), outward (middle), and net (right) currents during the AP. Early currents are shown during the first 5–25 ms of the AP. Y axis was zoomed in select plots to increase clarity.

Conduction over source-load mismatch.

We previously showed that in an ORd 2-D cardiac tissue model, simulated expression of BacNa_v rescued conduction block at sites of source-load mismatch created by nonconducting obstacles (16). Since incorporating BacNa_v in LRd and ORd models yielded a higher AP amplitude and early plateau compared with Na_v1.5 incorporation (Fig. 2), we hypothesized that these changes would be advantageous for overcoming large conduction delays at sites of source-load mismatch (42). We, therefore, constructed a discrete LRd fiber model [see methods (33)] with a central region comprised of 100 poorly coupled myocytes surrounded by two normally coupled 50-myocyte regions (Fig. 4A). For the chosen coupling resistances and standard LRd membrane model (control, 0×), conduction initiated at one end of the fiber blocked at the entry from poorly into normally coupled cells (Fig. 4B). Interestingly, while doubling Na_v1.5 conductance (+1× Na_v1.5) did not rescue the conduction block (Fig. 4C), adding as little as +0.2× BacNa_v overcame the conduction block because of source-load mismatch (Fig. 4D). Similarly, +0.2× BacNa_v expression, but not +1× Na_v1.5 augmentation, rescued the conduction block in an analogous 1-D fiber made using the ORd model (Supplemental Fig. S6). This result, in conjunction with the fact that the LRd and ORd fibers showed consistent trends in AP parameters in response to Na_v1.5 augmentation or BacNa_v expression (Fig. 2, Supplemental Fig. S4), prompted us to perform remaining simulations using a computationally less expensive LRd model.

Figure 4.

Effects of BacNa_v expression and Na_v1.5 augmentation on overcoming current source-load mismatch. A: schematic of the discrete Luo–Rudy dynamic guinea pig ventricular myocyte model (LRd) fiber (10 nodes/cell, dx = 0.001 cm) consisting of 2 well-coupled peripheral regions (cells 1–50 and 151–200, gap junction resistivity r = 1.5 Ω-cm²) and a poorly coupled central region (cells 51–150, gap junction resistivity R = 70 Ω-cm²) used to study conduction over source-load mismatch at the interface between 150th and 151st cell. B–D: action potential (AP) traces in the pre-mismatch cell 150 (black) and postmismatch cell 151 (red) shown for the control case (1× Na_v1.5, B), Na_v1.5 augmentation (+1× Na_v1.5, C), and BacNa_v expression (+0.2× BacNa_v, D). E–G: heatmaps of the conduction outcome (E) and safety factor of conduction (F) in the post-mismatch cell, and conduction delay across the source-load mismatch (G) for 49 combinations of (Na_v1.5, BacNa_v) conductance levels [from (0×, 0×) to (2×, 2×)]. H–J: analogous heatmaps to those in E–G for 49 combinations of BacNa_v gating time constants [($\tau$_m, $\tau$_h) from (0.1×, 0.1×) to (10×, 10×)]. K–M: AP traces at the source-load mismatch for cases that failed to conduct [i.e., Na_v1.5 with a 10× slower $\tau$_h (K) and BacNa_v with a 5× faster $\tau$_h (L)] and case with successful conduction for BacNa_v with a 2× faster $\tau$_h (M).

Specifically, we first explored how conduction velocity and safety factor of conduction [calculated using a modified calculation of Boyle and Vigmond (34) to account for the slower BacNa_v kinetics] change with loss of myocyte coupling in a 1-D LRd fiber for sodium channel conductances set at 0.5× Na_v1.5 (reduced excitability), 1× Na_v1.5 (control), and 0.5× Na_v1.5 + 0.5× BacNa_v (Supplemental Fig. S7). Expectedly, we found that 1× Na_v1.5 and 0.5× Na_v1.5 fibers had highest and lowest starting CV, respectively, and that the lower safety factor of conduction in 0.5× vs. 1× Na_v1.5 case yielded a conduction block at higher levels of coupling. Importantly, adding 0.5× BacNa_v to 0.5× Na_v1.5, increased the safety factor of conduction and prevented conduction block at levels of decoupling where 1× Na_v1.5 fiber failed to conduct. Overall, these studies demonstrated that the expression of BacNa_v channels rather than equivalent augmentation of endogenous Na_v1.5 current was more beneficial for the ability to sustain AP conduction in severely decoupled cardiomyocytes.

To further explore conditions for successful conduction over the 1-D source-load mismatch (Fig. 4A), we assessed how different conductance levels (from 0 to 2×) of Na_v1.5 or BacNa_v affected 1) the ability of myocytes to conduct over the site of mismatch (Fig. 4E), 2) the safety factor of conduction in the first post-mismatch cell (Fig. 4F), and 3) the resulting conduction delay over the mismatch (Fig. 4G). These studies revealed that without BacNa_v, no endogenous Na_v1.5 conductance level was able to support successful conduction over the mismatch (Fig. 4E). In contrast, incorporation of BacNa_v (at 0.1× to 1.5× level) was able to rescue failed conduction at each of the studied Na_v1.5 levels (including 0×), with safety factor (Fig. 4F) and conduction delay (Fig. 4G) that were respectively increased and decreased with increased BacNa_v expression. We then focused on the 0.5× Na_v1.5 case (reduced excitability) where we added either 0.5× Na_v1.5 current or 0.5× BacNa_v current and varied their on-off kinetics (0.1× to 10× t_h and 0.1× to 10× t_m). We found that varying kinetics of 0.5× Na_v1.5 current did not rescue conduction over the mismatch for any of the studied t_h and t_m values (including 10× t_h, Fig. 4K). On the other hand, successful conduction in the 0.5× Na_v1.5 + 0.5× BacNa_v case was blocked when t_h of BacNa_v was decreased by more than 0.5×, independent of the t_m value (Fig. 4H). While the safety factor of conduction was variable (Fig. 4I), conduction delay was decreased for larger t_h values (Fig. 4J), demonstrating the importance of relatively slow inactivation of BacNa_v in its improved ability to support conduction over the mismatch region compared with Na_v1.5. In fact, while augmenting Na_v1.5 conductance brought AP amplitude to a level induced by BacNa_v expression, a relatively fast inactivation of Na_v1.5 compared with BacNa_v did not support prolonged early AP plateau in premismatch cells that was necessary to sufficiently charge postmismatch cells over their activation threshold. This importance of early AP plateau was further evident when comparing faster inactivation (shorter t_h values) of BacNa_v that prevented (Fig. 4L) and slower inactivation (longer t_h values) that supported (Fig. 4M) AP conduction over the mismatch.

Vulnerability to Unidirectional Conduction Block

Increased vulnerability to unidirectional conduction block in cardiac tissue is the direct contributor to higher incidence of reentrant arrhythmias (35). To further compare how BacNa_v expression and Na_v1.5 augmentation influence critical regimes of cardiac conduction, we studied their effects on vulnerability to unidirectional conduction block during premature pacing based on the protocol of Shaw and Rudy (35). Specifically, a fiber of 150 LRd myocytes was paced with a train of S1 stimuli at a 1,000-ms cycle length from one end followed by a premature S2 stimulus that was delivered in the middle of the fiber at different S1–S2 coupling intervals (180–450 ms, Fig. 5A). Using this protocol, we determined a window of time (i.e., “vulnerability window”) over which the delivered S2 stimulus would cause a unidirectional (arrhythmogenic) conduction block rather than a failed or bidirectional (nonarrhythmogenic) conduction (Fig. 5B). We then varied the amplitudes of coincorporated BacNa_v and Na_v1.5 from 0–2× of their respective nominal values and examined the effects on 1) vulnerability window to unidirectional conduction block (Fig. 5, C and D) and 2) maximum slope of spatial repolarization gradient during S1 conduction (Fig. 5, E and F) shown to directly determine vulnerability to unidirectional block (43). We found that increases in BacNa_v and Na_v1.5 currents synergistically shortened the vulnerability window (Fig. 5C), which was a result of a decreased steepness of S1 repolarization gradient (Fig. 5E). Based on these simulations, we then assessed how the size of vulnerability window resulting from a specific amount of Na_v1.5 current changes when a fraction of this current is replaced by an equivalent amount of BacNa_v current (Fig. 5D). Interestingly, for any starting levels of Na_v1.5 current equal to or less than the healthy physiological level (1×), replacement of any fraction of this current by an equivalent amount of BacNa_v current shortened the vulnerable window (Fig. 5D). Thus, in cardiac tissues with reduced excitability, BacNa_v expression rather than Na_v1.5 augmentation would more efficiently decrease likelihood of unidirectional conduction block because of a premature excitation. For supraphysiological Na_v1.5 levels (1.5× and 2×), replacement of up to 1× Na_v1.5 with BacNa_v still shortened the vulnerability window, and only when completely replacing 1.5× or 2× Na_v1.5 with equivalent levels of BacNa_v resulted in vulnerable window prolongation (Fig. 5D). Similarly, slope of S1 repolarization gradient was reduced in the presence of BacNa_v, increasingly more as the fraction of Na_v1.5 current replaced by BacNa_v current was increased (Fig. 5E). Together, these simulations demonstrated beneficial effects of BacNa_v expression vs. Na_v1.5 augmentation in reducing the vulnerability to unidirectional conduction block in cardiac tissues with reduced excitability.

Figure 5.

Effects of BacNa_v expression and Na_v1.5 augmentation on vulnerability to unidirectional conduction block by premature excitation. A: schematic of a 3-cm-long fiber of Luo–Rudy dynamic guinea pig ventricular model (LRd) myocytes where a 30-s long S1 stimulus train [basic cycle length (BCL) = 1,000 ms] was followed by a premature S2 stimulus applied at various coupling intervals (CI = 180–450 ms). B: action potential (AP) traces at nodes 125, 150 (S2 site), and 175 demonstrating the three possible outcomes of S1–S2 stimulation at given CIs, namely failed conduction (left), unidirectional conduction (middle), and bidirectional conduction (right). C: heatmap of vulnerability window widths for 49 combinations of (Na_v1.5, BacNa_v) conductance levels [from (0×, 0×) to (2×, 2×)]. D: relative changes in the vulnerability window width derived from heatmap in C for cases when a specified BacNa_v conductance level (from 0.1× to 2×) is used to replace equivalent Na_v1.5 conductance level, with total BacNa_v+Na_v1.5 conductances shown color-coded. E and F: equivalent heatmaps to those in C and D, showing the maximum spatial repolarization gradients and gradient changes induced by S1 stimulus as a function of Na_v1.5 and BacNa_v conductance levels.

Conduction Stability at High Excitation Rates

We then explored the effects of BacNa_v expression and Na_v1.5 augmentation on the stability of AP conduction at fast excitation rates in 1-D fiber and 2-D tissue domains made of LRd myocytes. Specifically, continuous pacing at a constant rate of 10 Hz in a 1-D fiber with reduced excitability (0.5× Na_v1.5) resulted in an unstable AP conduction characterized by pace-to-pace variability in CL (Fig. 6A), APD, and APA (Supplemental Fig. S8A). The addition of 0.5× Na_v1.5 resulted in a modestly improved stability, which was further improved when 0.5× BacNa_v was added instead of 0.5× Na_v1.5. We then assessed the stability of spiral wave activity in a 2-D tissue domain using central obstacles of different sizes to generate anchored spiral waves (akin to anatomical reentry) with different rotation rates (Fig. 6B, Supplemental Fig. S8B). Consistent with the results from the 1-D fiber, and as evidenced from Poincare plots, an increase in spiral rotation rate due to decreased obstacle size (from 0.9 cm to 0.5 cm) in 2-D tissue with reduced excitability (0.5× Na_v1.5) was associated with moderate electrical instability, which, along with the spiral rotation rate, was significantly increased by the addition of 0.5× Na_v1.5. In contrast, the addition of 0.5× BacNa_v increased the stability of AP conduction despite an increase in rotation rate (Fig. 6, B and C, Supplemental Fig. S8B). Notably, in the no-obstacle case (Fig. 6C), free-rotating spirals with the added 0.5× Na_v1.5 underwent spontaneous breakup after 3 s of activity, which was also evident from an increase in the number of phase singularities (Fig. 6C). The 0.5× Na_v1.5 spiral also underwent spontaneous breakup, but only after 19 s of activity, while the +0.5× BacNa_v spiral remained stable during the entire simulation period (30 s). To further demonstrate the destabilizing effects of Na_v1.5 augmentation vs. BacNa_v expression on spiral wave activity, we replaced the +0.5× BacNa_v current during spiral rotation with the +0.5× Na_v1.5 current, which immediately destabilized the spiral wave and induced breakup (Supplemental Movie S1 at https://doi.org/10.6084/m9.figshare.22818413, Supplemental Fig. S8C). Collectively, these simulations showed that at high excitation rates, AP propagation in cardiac tissues with reduced excitability may be stabilized by the expression of BacNa_v channels but not by equivalent augmentation of endogenous Na_v1.5 current.

Figure 6.

Effects of BacNa_v expression and Na_v1.5 augmentation on the stability of action potential (AP) conduction during rapid excitation. A: AP traces during 2 s of 10 Hz-pacing shown for the central node of a Luo–Rudy dynamic guinea pig ventricular myocyte model (LRd) fiber and the corresponding cycle length (CL) Poincare plots. B: membrane voltage (V_m) snapshots (at t = 2 s) and CL Poincare plots (from the location marked with a white star) during 2 s of rotor activity in 8 cm × 8 cm LRd domain with a 0.9 cm (top) or 0.5 cm (bottom) circular central obstacle. C, top: V_m snapshots at specified times and phase singularity number (PS No., shown averaged over successive 500-ms windows) plots during 10 s of rotor activity in an obstacle-free 8 cm × 8 cm LRd domain. C, bottom: V_m snapshots during spiral breakup (when present) and corresponding CL Poincare plots (from the location marked by the white star) during 10 s of rotor activity shown (top). The 3 cases are shown: 0.5× Na_v1.5 (control), +0.5× Na_v1.5 augmentation, and +0.5× BacNa_v expression.

Brugada Syndrome Therapy

Finally, we compared various methods to increase I_Na as a potential gene therapy approach for Brugada syndrome (BrS). Specifically, we modeled AP conduction through a transmural LRd fiber consisting of endo-, midmyo-, and epicardial cardiomyocytes (38, 39) having 1) a healthy ion channel makeup, 2) BrS channel makeup (Brugada), 3) Brugada treated with 0.5× BacNa_v expression, 4) Brugada treated with healthy 0.5× Na_v1.5 expression, and 5) Brugada treated by mutated 0.5× BrS Na_v1.5 augmentation (Fig. 7A). The rate of inactivation in BrS Na_v1.5 and I_to level were varied to model 1) a “mild” Brugada case with attenuated APA and increased phase 1 notch in the epicardium and midmyocardium, leading to prominent J wave and the “saddleback” ECG shape (Fig. 7B) and 2) a “severe” Brugada case resulting in prominent APA attenuation and increase in phase 1 notch in the epicardium and midmyocardium and complete loss of AP dome in the epicardium, leading to “triangular” ECG shape with significant ST elevation (Fig. 7C) (38). While neither the expression of healthy Na_v1.5 nor augmentation of BrS Na_v1.5 rescued the pathological features of mild or severe Brugada phenotype, the expression of BacNa_v successfully rescued the changes in the AP amplitude, notch, and dome, leading to normalization of the ECG waveforms (Fig. 7, B and C). Together, the BacNa_v expression in cardiomyocytes rather than equivalent augmentation of Na_v1.5 current was able to improve impaired AP conduction and spatial heterogeneity characteristic of Brugada disease. In particular, slower inactivation of BacNa_v compared with Na_v1.5 channels prolonged the inward Na current to counteract the pathological increase in outward I_to, thereby rescuing changes in early repolarization responsible for increased Brugada phenotype severity.

Figure 7.

Effects of BacNa_v expression and Na_v1.5 augmentation in Brugada syndrome (BrS). A: schematics describing simulated transmural ventricular action potential (AP) conduction [60 endocardial, 45 midmyocardial, and 60 epicardial Luo–Rudy dynamic guinea pig ventricular model (LRd) myocytes; initiated at the endocardial end] and the location of pseudo-ECG measurement 2 cm from the epicardial surface. B and C: simulated AP traces from the centers of endocardial, midmyocardial, and epicardial regions, corresponding pseudo-ECG traces, and deviations from the healthy ECG (Brugada severity) shown for healthy tissue (black dashed), Brugada tissue harboring mutated (BrS) Na_v1.5 (black solid), Brugada treated with BacNa_v expression (purple), Brugada treated with healthy Na_v1.5 expression (yellow dashed), and Brugada treated with BrS Na_v1.5 augmentation (yellow solid). Characteristics of mild (B) and severe (C) Brugada cases are described in methods.

DISCUSSION

Recent advances in gene therapy research may lead to effective strategies to rescue deficits in cardiac tissue excitability and conduction that underlie various cardiac arrhythmias. Specifically, static or dynamic decrease in cardiac Na⁺ channel (Na_v1.5) current induced by loss-of-function mutations in SCN5A gene, environmental insults (e.g., ischemia, hyperkalemia), or premature excitation, in conjunction with an underlying fibrotic substrate with current source-sink mismatches and reduced coupling, can greatly compromise AP conduction, increase dispersion of repolarization and vulnerability to unidirectional block, and trigger life-threatening reentrant arrhythmias (44). Genetic augmentations of endogenous Na⁺ current amplitude in cardiomyocytes via two-virus delivery of wild-type SCN5A gene (15) or overexpression of MOG1 (trafficking chaperon of Na_v1.5) (4), as well as de novo expression of bacterial VGSCs (BacNa_v) (16), have been proposed as strategies to overcome arrhythmogenic conduction abnormalities in the heart. In this study, we directly compared the ability of augmented Na_v1.5 vs. expressed BacNa_v current to rescue conduction deficits in various pathological settings in silico and found that expression of BacNa_v channels in cardiomyocytes holds unique promise as a gene therapy for conduction-related arrhythmogenic disorders.

Our results build on a previously published model for the Na_vShep D60A bacterial sodium channel (16), by using a more extensive set of patch clamp and optical mapping results. The new model (Supplemental Fig. S2) yields improved fits of the experimental I-V curve, AP trace, and APD and CV restitution curves (Fig. 1) and compared with the original Nguyen et al. model, it has higher values of ${\tau }_m$ in the voltages between −40 and −20 mV, but lower otherwise, and shows a +5-mV shift in ${\tau }_h$ curve and the +5-mV shift and gentler slope in m_∞ curve. These effects lead to slightly slower channel activation, inactivation, and de-inactivation compared with the previous model, and, importantly, significantly better fits to the APD and CV restitution curves recorded in an engineered heterologous expression system comprising monolayers of excitable HEK293 cells expressing Kir2.1, Cx43, and BacNa_v channels.

When the updated BacNa_v current formulation was added to the LRd and ORd models, and long-term pacing was performed to equilibrate intracellular ion concentrations and other AP parameters (Supplemental Fig. S3), the effects on AP shape were similar to previously reported (16). Importantly, compared with the equivalent addition of Na_v1.5 current, BacNa_v expression induced a considerably lower increase in AP upstroke velocity and CV, but greatly increased AP amplitude and membrane potential during early repolarization (Fig. 2). Unlike the augmentation of Na_v1.5 current, which because of its fast kinetics primarily affected the AP depolarization, the slower activating and inactivating BacNa_v current had pronounced effects on the total membrane current, yielding an early increase in inward current followed by an increase in outward current (Fig. 3C, Supplemental Fig. S5C). The increase in the total outward current was mainly contributed by relative increase in the K⁺ driving force (V_m − E_K) caused by a higher early plateau potential, which amplified the relative contribution of I_Ks to late repolarization (Fig. 3B, Supplemental Fig. S5B), resulting in decreased APD (for BacNa_v levels of up to 1.5×, Fig. 2). Notably, this BacNa_v-induced decrease in APD is opposite from the effects of gain of late Na_v1.5 current mutations, which can cause APD prolongation and triggered arrhythmias (45), and is in agreement with our previous experimental study (16).

Given the increase in early plateau potential and AP amplitude with the addition of BacNa_v current, we hypothesized that BacNa_v expression could be particularly effective in supporting AP conduction at sites of long conduction delays caused by severe gap junctional decoupling (Supplemental Fig. S7) or current source-load mismatch (Fig. 4, Supplemental Fig. S6). In a discrete LRd cardiac fiber with a region of poorly coupled cells bounded by two normally coupled regions, we found that even at the lowest studied level (0.1×) of BacNa_v, conduction over a source-load mismatch was successfully rescued (Fig. 4). In contrast, doubling the endogenous Na_v1.5 current could not rescue conduction, suggesting that the slower kinetics of the BacNa_v current yielding increased amplitude and early repolarization of AP was critical to support a sufficient and long-lasting membrane voltage gradient at the site of mismatch that enabled activation of the post-mismatch cell even after a 5- to 10-ms delay. In support, speeding up inactivation of BacNa_v by fivefold disabled this rescuing effect (Fig. 4L).

Along with the improved conduction over a source-load mismatch, the simulated expression of BacNa_v channels was superior to the augmentation of Na_v1.5 current in reducing the vulnerability to unidirectional conduction block during premature stimulation. A shorter vulnerability window in the presence of BacNa_v vs. Na_v1.5 current was observed in all cases of reduced cardiomyocyte excitability (0 < Na_v1.5 < 1×) and was associated with less steep repolarization gradient (Fig. 5). Although CV was decreased when Na_v1.5 current was replaced with an equivalent BacNa_v current, findings from previous studies (43) suggest that a slower recovery from inactivation of BacNa_v is the main reason for the observed shortening of vulnerable window. Since unidirectional conduction block is one of the fundamental conditions for induction of reentrant activity in diseased myocardium (46), these results further supported the promise of BacNa_v expression as an effective gene therapy for cardiac arrhythmias.

In addition to the effective prevention of arrhythmogenic conduction block at sites of source-load mismatch or premature excitation, BacNa_v expression was superior to Na_v1.5 augmentation in counteracting dynamic instabilities in AP conduction resulting from high-frequency excitation induced by rapid pacing or self-sustained rotor activity akin to anatomical or functional reentry (Fig. 6, Supplemental Fig. S8). The stabilization of wavefront-waveback interactions by BacNa_v expression in these simulations could be mainly attributed to its effects on action potential amplitude and duration. Specifically, in BacNa_v-expressing tissues, shorter APD and higher APA (that increased excitatory current ahead of the wavefront) allowed for the formation of a faster and tighter spiral (Fig. 6C) without a loss of stability. In contrast, Na_v1.5 augmentation significantly increased CV without appreciably decreasing APD which intensified the interactions between spiral wavefront and waveback, ultimately leading to spiral breakup, a hallmark of ventricular fibrillation (47, 48).

It is likely that gene therapies targeting an increase in cardiomyocyte excitability would greatly benefit patients with Brugada syndrome (BrS) with loss-of-function mutations in the SCN5A gene (49). These mutations are shown to either impair the trafficking of Na_v1.5 channels and reduce their membrane density or alter channel gating properties to decrease the Na⁺ flow through channels. The mechanisms underlying the ECG abnormalities and arrhythmogenesis in patients with BrS are still debated (50) but can be attributed to 1) impaired AP depolarization, slowed conduction, and presence of structural abnormalities and/or 2) transmural repolarization heterogeneity aided by the presence of strong I_to in epicardial but not endocardial cells. Our simulations (Fig. 7) show that decreased transmural heterogeneity and restoration of the normal ECG shape in BrS hearts could be achieved by expressing BacNa_v channels, but not by alternative approaches expressing wild-type Na_v1.5 channels or increasing trafficking of endogenous mutated Na_v1.5 channels. Importantly, in addition to improving the velocity of AP conduction, slower inactivation kinetics of BacNa_v vs. Na_v1.5 uniquely counteracted epicardial loss of AP dome induced by the presence of strong I_to.

There were several limitations to the study. Although the new NavSheP D60A channel model showed improved fits of multiple experimental measurements (Fig. 1), the impact of parameter uncertainty on BacNa_v’s interactions with the dynamics of used AP models (LRd and ORd) was not studied given the large search space, the complex interactions between currents, and the lack of experimental data with NavSheP D60A expression in adult cardiomyocytes that would constrain and support these studies. Additional in vivo and ex vivo measurements will be needed to guide the development of improved BacNa_v models that would reproduce experimentally observed cell-to-cell variability (24, 51). Furthermore, based on the experimental results in neonatal rat ventricular myocytes (16), we modeled BacNa_v expression without altering currents through other ion channels, pumps, or exchangers, which may differ for adult cardiomyocytes. We also assumed that the therapeutic BacNa_v or Na_v1.5 expression is uniform in all cardiomyocytes, while viral delivery is known to result in variable efficiencies and levels of gene expression (52). Finally, we modeled 2-D tissue domains as isotropic, homogeneous, nonfibrotic media, which are different from pathological substrates encountered in various cardiac diseases. Nevertheless, based on the consistency of outcomes in conditions of reduced excitability, source-load mismatch, premature excitation, and genetic disease, we postulate that therapeutic effects of BacNa_v expression would similarly surpass those of Na_v1.5 augmentation in more realistic in silico models of cardiac gene therapy, and in vivo. Mechanistically, these therapeutic advantages could be primarily attributed to the slower inactivation kinetics of BacNa_v vs. Na_v1.5 channels.

In conclusion, we show that in addition to current amplitude, the kinetic properties of Na⁺ channels should be carefully considered when developing optimal gene therapies for cardiac conduction abnormalities. Our in vitro and in silico methods to customize BacNa_v properties (17, 19, 20) may offer a versatile platform to design gene therapies with disease-specific antiarrhythmic effects beyond those achievable with augmentation of endogenous Na_v1.5 current.

DATA AVAILABILITY

Data will be made available upon reasonable request.

SUPPLEMENTAL DATA

A: https://doi.org/10.6084/m9.figshare.24135039.

B: https://doi.org/10.6084/m9.figshare.22818413.

GRANTS

This work was funded by National Institutes of Health Grants U01HL134764 and R01EB032726 (to N.B.) and American Heart Association Predoctoral Fellowship PRE826228 (to T.W.).

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the authors.

AUTHOR CONTRIBUTIONS

D.N., C.S.H., and N.B. conceived and designed research; D.N., T.W., and H.X.N. performed experiments; D.N. analyzed data; D.N., C.S.H., and N.B. interpreted results of experiments; D.N. prepared figures; D.N. drafted manuscript; D.N., T.W., C.S.H., and N.B. edited and revised manuscript; D.N., T.W., H.X.N., C.S.H., and N.B. approved final version of manuscript.