Vagal Nerve Stimulation Reduces Ventricular Arrhythmias and Mitigates Adverse Neural Cardiac Remodeling Post–Myocardial Infarction

Study overview and approval

Animal experiments were approved by the UCLA Institutional Animal Care and Use Committee and performed in accordance with the National Institutes of Health’s Guide for the Care and Use of Laboratory Animals. A total of 41 Yucatan minipigs (Sus scrofa, S&S Farms) of both sexes, weighing 50 to 60 kg at terminal experiment, were used for the present study. Six animals died within 48 hours of MI from SCD, and were not included in final analysis. Surviving animals comprised 3 experimental groups: control (n = 10), MI (n = 12 without sham VNS; n = 4 with sham VNS), and MI + cVNS (n = 9). At the end of terminal experiments, animals were euthanized in concordance with Institutional Animal Care and Use Committee guidelines. Figure 1 presents overall study design and methods used in this project.

Implantation of vagal nerve stimulators

After sedation with midazolam (1 mg/kg, intramuscular) and ketamine (10 mg/kg, intramuscular), animals were endotracheally intubated and mechanically ventilated (tidal volume 300 to 450 mL, rate 10 to 12 breaths/min). Anesthesia was maintained with isoflurane (1%-4%, inhaled), and carprofen (4.4 mg/kg, intramuscular) and buprenorphine (0.02 mg/kg, intramuscular) were administered for analgesia. Ceftiofur (4 mg/kg, intravenous) and cefazolin (1 mg) were administered for perioperative antibiotic prophylaxis. Ear veins and the right or left femoral artery or tail artery were cannulated for medication administration and invasive blood pressure monitoring. Three-lead electrocardiogram, oxygen saturation, and end-tidal carbon dioxide were monitored. Body temperature was maintained using a forced-air normothermia system, and 0.9% NaCl (5-6 mL/kg/h) was administered to replenish insensible fluid loses. The intended incision site over the right ventral neck was anesthetized with a 1:1 mixture of lidocaine (1 mg/kg) and bupivacaine (1 mg/kg). The right external jugular vein and right vagus nerve were exposed. A wireless telemetry device (M01-F1, Data Sciences International) was implanted with 1 lead fluoroscopically guided to the cavo-atrial junction through the right external jugular vein, and the second lead tunneled to a pocket in the lateral neck. A bipolar VNS cuff (PerenniaFLEX Model 304, LivaNova) was placed around the right cervical vagus nerve and attached to a VNS pulse generator (Vitaria Model 7103, LivaNova). VNS devices were tested intraoperatively for evoked bradycardia to confirm device placement. Devices and leads were secured with 4-0 polypropylene sutures to the adjacent fascia, and the fascia overlying the vagus nerve and external jugular vein were each closed separately with absorbable suture. The device pockets and incision were then closed with 4 layers of absorbable suture, and the animals were extubated and recovered. Carprofen (4.4 mg/kg, once daily) was administered for 3 days for postoperative analgesia. Pigs were evaluated twice daily for 10 days, and daily thereafter.

Titration of VNS therapy

To achieve therapeutic levels of VNS, stimulation settings were systematically titrated to those that produced a null heart rate change at 5 Hz frequency and 250 μs pulse width during the on phase of VNS (14 seconds on, 66 seconds off) as shown in Supplemental Figure 1A. To accomplish this, titration was began at 5 Hz, 130 μs, and 0.25 mA. At each titration session, current was increased in 0.25-mA increments until side effects (cough, gastrointestinal upset) were elicited; current was then decreased by 0.25 mA and the pig rested for >1 day at these settings. Once a current of 3 mA was achieved, pulse width was stepped up to 250 μs and a similar process was repeated. Low-intensity stimulation (>1.5 mA) usually results in tachycardia (Supplemental Figure 1B); at higher-intensity stimulation bradycardia is evoked. Neural fulcrum was defined as a null change in heart rate during on-phase VNS, and titration was discontinued once fulcrum was achieved. In the MI + cVNS group, 2 days after MI, VNS was restarted and maintained at 5 Hz, 250 μs, and 17.5% duty cycle (14 seconds on, 66 seconds off), with intensity as previously determined individually to achieve neural fulcrum and maintained as such until termination.

Myocardial infarction

For induction of MI, the same anesthesia, analgesia, and monitoring protocol described already in this article was performed. The right or left femoral artery was accessed using the modified Seldinger technique and a vascular access sheath was placed. VA prophylaxis consisted of amiodarone (1.5 mg/kg, intramuscular) and lidocaine (2 mg/kg, intravenous), whereas periprocedural anticoagulation was maintained with heparin (5,000 U, intravenous). Under fluoroscopic guidance, an AL1.0 guide catheter (Medtronic) and 0.014-inch guidewire (Balance Middleweight Universal, Abbott) were used to cannulate the left coronary artery. An angioplasty balloon catheter was advanced over the guidewire to the mid–left anterior descending (LAD) coronary artery to the level of the second diagonal branch and inflated. A suspension (2.25 mL diluted to 6 mL with 0.9% NaCl) of polysterene microspheres (Polysciences Inc.) was used to embolize the distal LAD coronary artery. Hyperacute T waves, T-wave inversion, or ST-segment elevation were evident on surface electrocardiogram. External cardioversion and standard cardiopulmonary resuscitation were performed if the animal developed unstable ventricular tachycardia (VT) or ventricular fibrillation (VF). For animals requiring cardiopulmonary resuscitation, epinephrine (0.01 mg/kg, intravenous) was administered every 3 to 5 minutes, and amiodarone (1.5 mg/kg, intravenous) and lidocaine (2 mg/kg, intravenous) re-administered one additional time. Animals were extubated and recovered as already described. All pigs received carprofen (4.4 mg/kg) daily for 2 days and, if defibrillation was required, buprenorphine (0.02 mg/kg, intravenous) was administered twice daily for 2 days.

Surgical preparation at terminal study

Pigs were sedated, intubated, mechanically ventilated, maintained on isoflurane, and monitored as already described. Tiletamine-zolazepam (4-6 mg/kg, intramuscular) was used for induction and fentanyl citrate (20 mg/kg, intravenous) was administered for analgesia. Vascular access sheaths were placed in bilateral femoral arteries and veins for administration of fluids, monitoring, and placement of intracardiac catheters. Arterial blood gas contents were also evaluated hourly and ventilation parameters adjusted or sodium bicarbonate administered as needed to ensure physiological parameters for experimental studies. The thorax and heart were exposed through a median sternotomy. A pericardial cradle was created to support the heart. The parietal pleural was divided posteriorly and the sympathetic chain isolated at the level of the stellate ganglia. The left vagus was dissected as already described, and the right vagus lead was disconnected from the pulse generator. After completion of surgical procedures, anesthesia was transitioned from isoflurane to alpha-chloralose (25-50 mg/kg/h, intravenous), an anesthetic that does not depress cardiac and autonomic reflexes.

Nerve stimulation

Bipolar platinum nerve cuffs (Microprobes) were placed around the right and left sympathetic chain at the level of the stellate ganglia. A bipolar VNS cuff was placed around the left vagus as already described. Nerve stimulation was performed using a Grass S88 stimulator, interfaced through a PSIU6 photoelectric constant-current isolation unit (Grass Instruments). For sympathetic chain stimulation, threshold was defined as the current required to elicit a 10% increase in heart rate or LV end-systolic pressure at a fixed frequency of 4 Hz and fixed pulse width of 4 ms. For VNS, stimulus threshold was defined as the current required to elicit a 10% change in heart rate at a fixed frequency of 10 Hz and fixed pulse width of 1 ms. Bilateral sympathetic chain stimulation (BSS) was performed at 1.5 times threshold at moderate (4 Hz) and high (10 Hz) intensity to evaluate evoked sympathetic responses. To test whether acute VNS influenced induction of VT or VF, acute bilateral VNS was performed in conjunction with VA inducibility at 1.5 times fulcrum (right) or threshold (left) intensity.

Echocardiography

Echocardiography was performed as previously described.²³ LV ejection fraction was calculated using the modified Simpson method. LV wall thicknesses were measured in diastole.

Hemodynamic and electrophysiological data acquisition and analysis

A solid-state pressure catheter (Mikro-Tip Model SPR-350, Millar Instruments) was placed in the left ventricle through the femoral artery and used to measure rate of change of LV pressure (dP/dt) and heart rate. A lead II electrocardiogram and bipolar electrocardiograms were measured with surface electrodes and quadripolar or duodecapolar electrode catheter (Boston Scientific). These data were acquired, digitized, and analyzed offline using a data acquisition system (Cambridge Electronics Design Power1401 and Spike2). LV contractility or systolic function was evaluated with the maximum rate of change of LV pressure (LV dP/dt_max) while lusitropy was evaluated using the minimum rate of change of LV pressure (LV dP/dt_min).

A custom, 128-electrode thin film array (NeuroNexus) was placed over the left ventricle scar-border zone (or equivalent area in controls) to acquire unipolar epicardial electrograms. Data was sampled at 1,380 Hz using an AlphaLab SnR acquisition system (Alpha Omega). Activation recovery intervals (ARIs) were analyzed from each epicardial electrogram using Scaldyn (University of Utah) as previously described.²⁴ Maps were plotted using Fiji²⁵ with identical scaling and bilinear interpolation for the 3 experimental groups simultaneously. Activation time was identified as the time from the beginning to the minimum dV/dt in the activation wavefront and recovery time by the maximum dV/dt of the repolarization wavefront. Activation delay was defined as the maximum minus minimum repolarization time, whereas activation dispersion was assessed as the variance of activation times. Repolarization time at 90% repolarization (RT₉₀) was defined as the 95th percentile minus the 5th percentile of local repolarization times as previously described.²⁶ Repolarization gradient was defined as the RT₉₀ divided by the geometric distance between corresponding electrodes. ARI was calculated as the difference between recovery time and activation time, and ARI dispersion as the variance in ARI. ARI serves as a strong in vivo surrogate for local action potential duration.²⁷

Arrhythmia inducibility

VA inducibility was performed using programmed electrical stimulation (PES) using a Micropace EPS 320 cardiac stimulator (Micropace EP) interfaced with a Prucka CardioLab System (GE Healthcare). PES was performed at a stimulus train (S1) 20% faster than the resting heart rate followed by an S2 extrastimulus with a 10 ms decrement until 200 ms or until effective refractory period was reached. If the first extrastimulus did not induce sustained VT or VF, a second (S3) extrastimulus was added and the decrement process repeated. This was performed through a third (S4) extrastimulus. For animals receiving cVNS therapy, PES was performed twice: without acute VNS, and with bilateral acute VNS for 30 min before and during PES. Animals that did not experience sustained VT or VF for >30 seconds with 3 extrastimuli were considered noninducible. Nonsustained ventricular tachycardia (NSVT) was defined as at least 3 or more ventricular beats at a rate of >120 beats/min that terminated within 30 seconds. Animals were classified as noninducible, induced for NSVT, or induced for sustained VT/VF. An arrhythmogenicity index was calculated to quantify the ease of VT/VF inducibility based on the stimulus drive train and number of extrastimuli as previously described.²³ Higher values indicate greater ease of inducibility.

Tissue collection and processing

The LV anterior myocardium adjacent to the LAD coronary artery (consisting of scar, border, and remote regions), right stellate ganglia, and first thoracic dorsal root ganglia (DRG) were collected and rinsed in cold phosphate-buffered saline (PBS). Myocardial samples were fixed in 10% formalin (Sigma-Aldrich) at room temperature for 72 hours, washed with PBS 4 times for 1 hour each, and preserved in 70% ethanol. Stellate and DRG were fixed in 4% paraformaldehyde (Electron Microscopy Sciences) at 4°C for 24 hours, washed with PBS 4 times for 1 hour each, and preserved in PBS with 0.02% sodium azide.

Masson’s trichrome staining was performed on LV samples spanning the scar-border zone, just adjacent to the LAD, and evaluated for structural abnormalities. Abnormalities included increased myocytolysis, decreased staining for muscle bands, reactive hypertrophy, and presence of dysmorphic myocytes. The following arbitrary scale was used: 3 = very abundant, 2 = moderate, 1 = low, and 0 = none or rare. Three to four sections were evaluated per animal, and values were recorded for subepicardium, midwall, and subendocardium. These were added on the first pass and an average value for each animal was calculated.

Stellate ganglia sections were stained using a 2-day protocol with the following combinations: 1) protein gene product 9.5 (PGP9.5), tyrosine hydroxylase (TH), vesicular acetylcholine transporter (VAChT); 2) PGP9.5, TH, neuropeptide Y (NPY); and 3) PGP9.5, glial fibrillary acidic protein (GFAP), S100 Calcium Binding Protein B (S100B). The right stellate ganglia were paraffin embedded to acquire 5-μm sections along the long-axis of the ganglia. Slides with paraffin-embedded sections underwent 2 consecutive, 10-minute baths in xylene. Slides were then placed in ethanol baths of varying concentrations (100% for 10 min, followed by 95% for 5 minutes, followed by 70% for 5 minutes). Finally, slides were placed in dH₂O for 5 minutes. Slides were placed in an antigen unmasking solution, which was made by diluting 2.5 mL of antigen retrieval buffer in 250 mL of dH₂O and boiling to 90°C, for 25 minutes, then cooled for 1 hour. Slides were then placed in a soaking buffer solution (PBS with 0.4% TX-100) for 30 minutes at 4°C. Tissue wells were created on slides using a hydrophobic barrier pen and 50 to 100 μL PBS was added to each well depending on tissue size. The slides were placed in a humidified chamber and agitated for 5 minutes. PBS was then replaced with blocking buffer solution (PBS + 10% donkey serum with 0.1% TX-100) for 1 hour. Primary antibody solution was then made by adding primary antibodies (Supplemental Table 1) to the blocking buffer solution, which replaced the blocking solution in the wells. Slides were agitated overnight at 4°C. Slides were then warmed to room temperature and washed with PBS with 0.2% TX-100 3 times. Secondary antibody solution was then made by adding secondary antibodies (Supplemental Table 2) to the blocking buffer solution, which then replaced the primary antibody solution in each well. Slides were agitated gently for 1 hour at room temperature then washed 3 times with PBS. PBS was then replaced with refractive index matching solution that covered each section entirely and cover slips were placed directly after to secure the tissue. Slides were imaged using confocal microscopy.

DRG were stained using a 7-day protocol with the following combinations: 1) PGP9.5, calcitonin gene-related peptide (CGRP), neuronal nitric oxide synthase (nNOS); and 2) PGP9.5, GFAP, S100B. The right first thoracic DRG from each subject was embedded in agarose to acquire 100-μm sections along the long-axis of the ganglia. Each section was placed in 1 mL of blocking solution containing 10% donkey serum, 0.2% Triton X-100, and PBS for overnight incubation with slight agitation. A primary antibody solution was made using blocking solution and primary antibodies (Supplemental Table 1). After 3 nights of incubation with slight agitation, the sections were washed with PBS every hour for at least 6 washes. A secondary antibody solution was made with blocking solution and secondary antibodies (Supplemental Table 2). Sections were incubated for 3 nights with slight agitation. Sections were then washed with PBS every hour for at least 6 washes. Sections were mounted on slides and refractive index matching solution was added in drops until the sections were covered entirely, and a coverslip was placed to secure the tissue. Slides were imaged using confocal microscopy.

Confocal imaging was performed using a Zeiss LSM-880 confocal laser microscope (Zeiss). The following lasers were used: 488 (Green), 562/Cy3 (Red), and 647 (Blue). Imaging was done at 10x and stitched together using the tile scan feature. Maximum intensity projections for a representative Z-stack were subsequently used for analysis.

Three representative 500- × 500-μm regions of interest (ROI) were selected for each stellate section on Zen Black (Zeiss). Sixteen-bit full-resolution images were exported and analyzed on ImageJ Software (National Institutes of Health). The PGP9.5, TH, and VAChT combination was analyzed for cholinergic transdifferentiation by using the cell counter feature in ImageJ to obtain a total cell count using the PGP9.5 channel and a count of cells double-labeled for TH and VAChT. The PGP 9.5, TH, and NPY combination was similarly analyzed for NPY expression. Glial analysis was performed by selecting 15 satellite glial cells (5 cells per ROI) from each animal for individual analysis. The length and area of each satellite glial cell was measured on ImageJ using the GFAP stain. The immunoreactive area (μm²) per cell for each ROI was calculated by converting the GFAP channels to binary images. The thresholding tool in ImageJ was then used to calculate the total GFAP stain area in each ROI. A ratio of total GFAP area to total cell count based on the PGP9.5 channel was then calculated for each ROI and averaged for each animal.

One representative 500- × 500-μm ROI was selected for each DRG section. Sixteen-bit full-resolution images were exported and analyzed on ImageJ Software. The PGP9.5, CGRP, and nNOS staining combination was analyzed for cells labeled for nNOS or CGRP and PGP9.5. To evaluate glial activation, the immunoreactive area (μm²) per cell for each ROI was similarly calculated by converting the GFAP channels to binary images. The thresholding tool was then used to calculate the total GFAP stain area in each ROI. A ratio of total GFAP area to total cell count based on the PGP9.5 channel was then calculated for each animal.

Statistical analysis

Data are presented as mean ± SEM. Normality was assessed using the Shapiro-Wilk test. For normally distributed variables, differences in SD were tested using the Brown-Forsythe test. For normally distributed variables, comparisons were made using standard analysis of variance if SDs were not significantly different, and Welch’s analysis of variance if SDs differed. For non-normally distributed variables, the Kruskal-Wallis test was used. Pairwise comparisons were subsequently performed using the t test or Wilcoxon rank-sum test, and the 2-stage step-up method of Benjamini, Krieger, and Yekuteili was used to control for multiple comparisons. For comparisons between solely 2 groups, the t test or Wilcoxon rank-sum test was used. Categorical variables were studied using the chi-square test. A 2-sided P value of <0.05 was deemed statistically significant. Statistical analysis was conducted using Prism 9 (GraphPad Software).

Implementation of cVNS in Yucatan minipigs

To study the effects of cVNS on MI (Figure 1A), we developed a strategy for cVNS delivery at the neural fulcrum.²¹ Our preliminary studies suggested that titration to therapeutic VNS parameters required approximately 30 days to prevent side effects in pigs. Because the majority of myocardial remodeling occurs within the first month after MI, we implanted right vagus nerve cuffs, pulse generators, and telemetry units and titrated to therapeutic settings before MI (Supplemental Figure 1). This strategy allowed us to begin therapeutic VNS on-demand after MI. Both MI + cVNS pigs and MI + sham cVNS pigs underwent VNS pulse generator and lead implantation, followed by titration to VNS fulcrum (∼4 weeks). Once VNS fulcrum was achieved (5 Hz, 250 μs, mean current 1.8 ± 0.2 mA), therapy was discontinued and pigs underwent MI at least 1 day after cessation of therapy to ensure no memory effect from VNS titration.²⁸ MI was induced in MI, MI + sham cVNS, and MI + cVNS pigs using percutaneous embolization of the LAD coronary artery, and infarcts were allowed to mature for 6 to 8 weeks before terminal study. VNS therapy was restarted 2 days after MI in the MI + cVNS group. Among all animals undergoing MI induction, 6 experienced SCD within 2 days of MI, confirming the adequacy of this model. A sensitivity analysis, reported later in this article, found no significant difference in sympathetic function, electrophysiological parameters, cardiac structural remodeling, or arrhythmia inducibility in MI compared with MI + sham cVNS groups (Supplemental Figure 2). As such, these were considered as one group for the primary analysis reported in this study. Age- and weight-matched controls that had not undergone any interventions were used for comparison.

cVNS rescues cardiac mechanical function after MI

To determine whether chronic VNS influences basal LV contractile function after MI, we performed echocardiography in controls, MI, and MI + cVNS animals (Figure 2A). MI animals had significant impairments in LV systolic function compared with controls, with LV ejection fraction (EF) of 40.6% ± 1.2% compared with 71.6% ± 0.4% (P < 0.001). Chronic VNS therapy rescued changes in EF to 67.4% ± 1.3%, although this remained significantly different from controls (Figure 2B). Similarly, MI resulted in greater LV end-systolic volumes and LV internal diameter in diastole, which were significantly improved in MI + cVNS animals. Wall thickness in the anterior wall, the LV region primarily perfused by the LAD, significantly decreased in MI vs control animals (0.82 ± 0.02 cm vs 0.57 ± 0.04 cm; P < 0.001), which was partially attenuated with cVNS therapy to 0.65 ± 0.01 cm (Figure 2B). The myocardial regions not perfused by the LAD remained unaffected (data not shown). To evaluate functional sympathetic control of the heart, we performed BSS and tested whether MI or MI + cVNS influenced changes in inotropy or lusitropy. Baseline hemodynamics and thresholds used for sympathetic nerve stimulation are reported in Supplemental Tables 3 and 4. MI resulted in dysfunctional inotropic and lusitropic responses to BSS, which were restored to levels similar to controls in MI pigs receiving cVNS therapy (Figure 2C). This was evident with both moderate- and high-intensity stimulation, suggesting that cVNS ameliorated reduced cardiac sympathetic reserve evident in MI animals.

cVNS reduces the susceptibility to VAs after MI

MI and its associated structural abnormalities in the border zone are known to increase the risk of life-threatening VAs. To evaluate whether cVNS reduces VAs after MI, we performed PES using extrastimulus pacing in MI and MI + cVNS pigs. A representative recording of VT induction is shown in Figure 3A; as noted, the MI animal was induced with 2 extrastimuli, whereas the MI + cVNS subject remained noninducible despite 3 extrastimuli. VT or VF was induced in 92% of MI pigs, whereas the remaining animal experienced NSVT (Figure 3B). In MI + cVNS subjects, 11% were inducible for VT/VF at baseline (P < 0.001 vs MI), whereas the remaining were noninducible (Figure 3B). When active VNS was re-established in MI + cVNS pigs for 30 minutes before and during the extrastimulus protocol, 0% were inducible for VT/VF and 11% experienced NSVT (P < 0.001 vs MI) (Figure 3B). The ease of inducibility, quantified based on the rate of the drive train and number of extrastimuli required to induce VT/VF, was significantly greater in MI compared with MI + cVNS subjects (4.4 vs 0.4; P < 0.001) but similar regardless of whether active VNS therapy was applied (Figure 3C).

VNS exerts antiarrhythmic effects in part via stabilization of cardiac electrophysiological function

To evaluate the substrate-level mechanisms by which cVNS ameliorated VAs, we performed high-resolution electrical mapping of the LV scar-border zone in MI and MI + cVNS animals, and corresponding anatomic region in controls. Heterogeneous areas of myocardial scars contribute to ventricular arrhythmogenesis by facilitating slow conduction and re-entry. These regions, as well as subsets with late tissue activation, are the clinical targets of catheter ablation, the current standard of care for re-entrant VAs. Figure 4A demonstrates representative activation maps for control, MI, and MI + cVNS animals. As quantified in Figure 4B, MI animals displayed substantial conduction block (ie, no increase in proportion of LV tissue activated relative to time) compared with controls, which was qualitatively improved with cVNS therapy. Consistent with prior work, we found that MI quantitatively resulted in greater mean ventricular activation time, activation dispersion (variance of activation time), activation delay, and maximal activation time (Figures 4C to 4F). Chronic VNS therapy normalized the majority of these measures to levels comparable with controls, including mean activation time, delay, and maximal activation time, whereas activation dispersion, a sensitive measure for scar, was improved relative to MI (17.6 ± 1.5 ms² vs 28.5 ± 2.9 ms²; P = 0.008) but remained greater than control (17.6 ± 1.5 ms² vs 10.0 ± 1.4 ms²; P = 0.001; Figure 4D). These findings suggest that cVNS substantially remodels myocardial scars and ameliorates the majority of the conduction abnormalities observed after MI.

Abnormal myocardial repolarization properties, particularly heterogeneity, have been implicated in arrhythmias after MI. Representative repolarization maps and repolarization times for each study group are shown in Figures 5A and 5B. To study the influence of cVNS on repolarization properties after MI, we first quantified LV ARI, a surrogate measure for action potential duration (APD) that can be readily measured in vivo without causing myocardial injury.²⁷ Interestingly, we found no significant difference in mean ARI across the scar-border zone, or equivalent region, in control compared with MI and MI + cVNS animals (Figure 5C). To evaluate heterogeneity in ARI, we examined ARI dispersion, and found that MI induced substantial increases in ARI dispersion (399 ± 75 ms² vs 63 ± 22 ms²; P < 0.001), which were normalized using cVNS therapy (110 ± 32 ms²) (Figure 5D). Because differences in ARI may be related to differences in activation dynamics, we tested whether differences in repolarization time persisted across the scar-border zone by evaluating differences in the 95^th relative to 5th percentile of repolarization time (RT₉₀). MI resulted in substantially greater differences in RT₉₀ across the mapped region relative to control, which was ameliorated with cVNS therapy (Figure 5E). Steep repolarization gradients have recently been associated with the development or maintenance of re-entrant arrhythmias.^5,29,30 As anticipated, MI pigs had steep repolarization gradients compared with controls (3.7 ± 0.6 ms/mm vs 1.6 ± 0.1 ms/mm; P = 0.007), which was similarly improved with cVNS therapy (2.0 ± 0.4 ms/mm) (Figure 5E). Taken together, these findings strongly suggest that cVNS not only improves conduction across the infarcted myocardium, but improves the repolarization properties of myocytes themselves in vivo, protecting the heart against VAs.

cVNS exerts its antiarrhythmic effect partly by reducing aberrant remodeling of the myocardial substrate

Given the observed improvement in mechanical and electrophysiological properties of cVNS in infarcted hearts, we evaluated structural changes in the scar-border zone. We performed Masson’s trichrome staining on LV sections spanning the scar-border zone, just adjacent to the site of LAD embolization. Compared with MI animals (Figures 6A and 6B), we qualitatively observed less myocytolysis, improved staining of muscle bands, less reactive hypertrophy, and more transmural scar in cVNS + MI pigs (Figures 6C and 6D). These specific findings, particularly a reduction in myocytolysis (Figure 6A, arrows), is consistent with reduced potential for VT/VF. Quantification revealed a marked reduction in border zone abnormality among MI + cVNS compared with MI animals (1.8 ± 0.2 vs 0.6 ± 0.2; P = 0.001; Figure 6E).

MI alters neural and glial phenotypes in the peripheral nervous system, which are partly attenuated with cVNS

In both humans and animal models, myocardial injury leads to extracardiac neural remodeling, which may drive sympatho-excitation after MI.^31,32 Stellate ganglia surgically removed from patients with recurrent VT showed evidence of oxidative stress, glial activation, and changes in neuronal phenotypes.³² To evaluate for changes in neuronal phenotypes in our porcine models, we colabeled stellate ganglia for: 1) PGP9.5 (a pan-neuronal marker), TH, and VAChT; and 2) PGP9.5 and NPY. We found a significant increase in TH+/VAChT+ neurons after MI, which was unchanged with cVNS therapy (Figure 7A). Similarly, the proportion of NPY+ neurons significantly increased after MI, a change that trended toward normalization with cVNS (Figure 7B). To evaluate for glial activation, stellate ganglia were colabeled with S100B, GFAP, and PGP9.5. MI resulted in a substantial increase in glial activation, as evaluated by glial area and glial area relative to neurons per section, which was reversed to control levels with cVNS treatment (Figure 7C). Similar findings were evident on evaluation of length, width, and thickness of glial cells in the stellate, whereas no difference was found in neuronal cell size (Supplemental Figure 3).

We similarly evaluated changes in afferent neuronal phenotypes and glial activation in the first thoracic DRG, the primary source of general visceral fibers that innervate the heart. nNOS expression significantly increased after MI, which was mitigated using cVNS (Figure 8A). The expression of CGRP was similar in the 3 study groups (Figure 8B). Consistent with the stellate, MI led to significantly increased glial activation in the DRG, which was ameliorated using cVNS (Figure 8C). There was no significant difference in overall neuron size, nNOS + neuron size, or CGRP + neuron size by study group (data not shown).

Clinical perspective

A mechanistic and rationale basis of VNS therapy is imperative for appropriate clinical application in the setting of MI and heart failure.^9,20 Although low-intensity acute VNS is known to reduce the propensity for arrhythmias during experimentally induced coronary ischemia, the efficacy of cVNS therapy in experimental models is strongly influenced by stimulation parameters.²¹ For example, Radcliffe et al¹⁵ found no effect of cVNS therapy on pacing-induced cardiomyopathy in an ovine model when delivered at settings leading to a 30% bradycardia,¹⁵ whereas Zhang et al⁵⁶ targeted a 10% bradycardia in a similar canine model and found attenuation of heart failure development, although VNS therapy was started earlier in their study. As demonstrated by our VNS titration protocol, increasing intensity of VNS initially evokes tachycardia followed by bradycardia; such an effect is consistent in dogs, pigs, and humans.^19,21,57 As such, dosing and delivery of VNS therapy, analogous to medications, has substantial impact on its efficacy and unintended off-target effects. Considerable differences in neural targets and stimulation parameters were evident in the 3 major trials of cVNS for reduced ejection heart failure.¹⁶ The ANTHEM-HF (Autonomic Regulation Therapy to Enhance Myocardial Function in Heart Failure) trial operated at VNS parameters consistent with the neural fulcrum and led to improvements in both LV function and quality of life.58, 59, 60 On the other hand, the VNS delivery modality and stimulation parameters for the 2 other major trials suffered from an inability to reach target VNS parameters due to side effects, which may partly explain the lack of clinical efficacy observed. Specifically, a post hoc analysis of the NECTAR-HF (Neural Cardiac Therapy For Heart Failure) trial, which found no improvement in LV function with cVNS, demonstrated that only 12% of the patients randomized to VNS therapy showed evidence for therapeutic engagement of the autonomic nervous system inputs to the heart.⁶¹ The present work provides definitive evidence that cVNS exerts significant antiarrhythmic effects and mitigates aberrant cardiac and neural remodeling in large mammals when delivered in a rationale, neuroscience-guided manner. Such factors must be considered in further clinical trials of VNS or other device-based therapy.

Study Limitations

The present study has several limitations inherent to its design and defined objectives. We solely evaluated efficacy of right-sided VNS, which is the more commonly used modality in the clinical setting for treatment of cardiac pathologies. Pigs were titrated to neural fulcrum before MI for this study, which may not be feasible in the clinical setting. Further work will delineate the optimal time to initiate VNS, and fascicular stimulation may obviate the need for titration by avoiding off-target effects of VNS.⁶² To demonstrate that our observed findings were mediated by cVNS after MI, rather than VNS-induced cardioprotection evolving from the titration phase, VNS was turned off before MI induction and not restarted until 2 days after MI. Furthermore, a group of animals, termed sham VNS + MI, were similarly titrated to neural fulcrum and underwent MI, but received no reactive VNS therapy after MI. Our sensitivity analysis clearly demonstrates that these animals were identical to MI alone with regard to myocardial performance, ventricular activation and repolarization properties, arrhythmia inducibility, and scar-border zone histologic characteristics (Supplemental Figure 2). In addition, our evaluation of cardiac mechanical function was performed in the sedated state because echocardiography was not feasible in the conscious state in pigs. Among our cohort, 6 animals died within 48 h after MI due to SCD despite standard cardiopulmonary resuscitation. These animals were likely more susceptible to malignant VAs, and may have derived more substantial benefit from cVNS because prior studies have found greater impairment in parasympathetic function in animals susceptible to ischemia-induced arrhythmias compared with those resistant.^63,64 Further work is necessary to delineate the specific mechanisms by which cVNS imparts its cardioprotective effects, including its impact on neurotransmitter dynamics, scar-border zone properties, and remodeling of neural circuitry. In addition, studies investigating changes in biomarkers and heart rate variability with MI and cVNS may aid in clinical translation of this work.