Neural Mechanisms of Paroxysmal Atrial Fibrillation and Paroxysmal Atrial Tachycardia in Ambulatory Canines

Abstract

Background

The relationship between autonomic activation and the mechanisms of paroxysmal atrial fibrillation (PAF) remains unclear.

Methods and Results

We implanted a pacemaker and a radiotransmitter in 7 dogs (Group 1). After baseline recording, we paced the left atrium (LA) at 20 Hz for one week and then monitored left stellate ganglion nerve activity, left vagal nerve activity and LA electrogram off pacing for 24 hrs. This protocol repeated itself until sustained AF (>48 hrs) was induced in 3±1 weeks. In another 6 dogs (Group 2), we cryoablated left and right stellate ganglia and the cardiac branch of left vagal nerve during the first surgery and then repeated the same pacing protocol until sustained AF was induced in 7±4 weeks (p=0.01). There were PAF (4 ± 2 episodes/d) and paroxysmal atrial tachycardia (PAT, 10 ± 3 episodes/d) in Group 1. Simultaneous sympathovagal discharges were observed to immediately precede the onset of atrial arrhythmias in 73% of episodes. In comparison, Group 2 dogs had no PAF (p=0.046) or PAT episodes (p<0.001). There was nerve sprouting, sympathetic hyperinnervation and a massive elevation of transcardiac norepinephrine levels in both groups.

Conclusions

Intermittent rapid LA pacing results in sympathetic hyperinnervation, PAF and PAT. Simultaneous sympathovagal discharges are common triggers of these arrhythmias. Cryoablation of extrinsic sympathovagal nerves eliminated PAF and PAT, suggesting that simultaneous sympathovagal discharges and these arrhythmias are causally related. Because cryoablation only delayed but did not prevent sustained AF, autonomic nerve activity is not the only factor determining AF maintenance.

Introduction

Atrial fibrillation (AF) is a complex arrhythmia that requires a trigger for initiation and a favorable substrate for maintenance. However, the nature of the trigger remains elusive. Coumel¹ proposed that both sympathetic and parasympathetic nervous systems are important in triggering AF. Heart rate variability analyses suggest that sympathovagal activation occurs prior to the onset of PAF episodes.² In vitro studies showed that combined sympathetic activation and acetylcholine infusion can facilitate the induction of AF, probably through the mechanism of late phase 3 early afterdepolarizations.^3-5 These results suggest that combined sympathovagal discharges may be needed to trigger paroxysmal AF (PAF). We have successfully developed methods for continuous autonomic nerve recordings in ambulatory dogs.^{6, 7} Wijffels et al⁸ showed that chronic rapid atrial pacing leads to sustained AF. We hypothesize that chronic rapid pacing may also lead to PAF. The first goal of the present study, therefore, was to develop a canine model of PAF by chronic intermittent rapid pacing and to use this model to test the hypothesis that specific patterns of autonomic nerve activity (ANA) precede the onset of PAF. The second objective of the study is to perform cryoablation of the stellate ganglia and superior cardiac branch of the vagal nerve to test the hypothesis that cryoablation of these nerve structures can prevent or reduce PAF episodes. Stellate ganglion ablation has been used in patients to control ventricular arrhythmias.^{9, 10} A possible mechanism of its antiarrhythmic effect is reduced cardiac sympathetic innervation. Therefore, a third objective of the study was to test the hypothesis that bilateral stellate ganglion ablation can lead to cardiac sympathetic denervation.

Methods

The study protocol was approved by the Institutional Animal Care and Use Committees. The authors had full access to and take full responsibility for the integrity of the data. We studied 13 Mongrel dogs (22-27kg) consisting of 2 groups. Group 1 consists of 7 dogs that did not undergo cryoablation. Group 2 comprises 6 dogs that underwent cryoablation of sympathovagal nerves during first surgery. The first 7 dogs were assigned to Group 1 and the second 6 dogs to Group 2. Hearts from an additional 6 normal dogs were studied de novo for histological control.

Continuous recordings of autonomic nerve activity

We implanted in all 13 dogs a Data Sciences Inc (DSI) D70-EEE radiotransmitters to record ANA activity and simultaneous ECG according to methods described in detail elsewhere.⁷ Briefly, a DSI transmitter was implanted to subcutaneous tissues. Left stellate ganglion nerve activity (SGNA) was registered by suturing one pair of bipolar electrodes onto the caudal half of the left stellate ganglion (LSG) beneath its fascia. To record cardiac vagal nerve activity (VNA), another pair of bipolar electrodes was sutured onto the superior cardiac branch of the left vagal nerve. The final bipolar pair was sutured onto the LA epicardium or subcutaneously as surface ECG. Telemetered signals from the transmitter were acquired continuously, 24 hrs a day 7 days a week, while the dogs are ambulatory.

Cryoablation of autonomic nerves

Schwartz et al reported that ablation of the caudal half of the left stellate ganglion and T2-T4 thoracic sympathetic ganglia was effective at preventing ventricular arrhythmias.¹⁰ However, atrial sympathetic innervation derives significantly from both left as well as right stellate ganglion (RSG).¹¹ Therefore we ablated both LSG and RSG according to the Schwartz protocolin Group 2. The cranial halves of the SG were spared to limit the potential for Horner’s syndrome,¹⁰ which includes ptosis and miosis ipsilateral to the site of SG damage. For vagal denervation, we ablated the superior cardiac branch of the left vagal nerve. Figure 1A-1D shows the cryoablation procedure. The nerves were cryoablated using SurgiFrost system (CryoCath Technologies, Inc., Montreal, Quebec, Canada) to achieve a tissue temperature of -146°C for 4 min. Subsequently, ANA was recorded from the nerve upstream to the site of ablation.

Figure 1.

LSG and superior cardiac branch of left thoracic vagal nerve. Left edge is cranial and right edge is caudal for all panels. (A): LSG before ablation. (B): Cryoablation of LSG. (C): The left thoracic vagal nerve runs parallel to left phrenic nerve. The superior cardiac branch of the left thoracic vagal nerve is located between these two large nerves (arrow in insert). (D): Superior cardiac branch (arrow) of left vagal nerve was lifted up by the cryoprobe and cryoablated. (E-H): Histology of cryoablated SG. (E) and (F): trichrome stains of RSG, showing fibrosis at the site of ablation (arrows). (G and H): Surviving ganglion cells within the LSG staining positive (brown) for TH.

Surgical procedures

A left thoracotomy was performed under isoflurane anesthesia in all 13 dogs. Blood was sampled simultaneously from the coronary sinus (CS) and aorta (AO) before and immediately after LSG stimulation (20s, 10mA, 20Hz, 2ms pulse width). Cryoablation was then performed in Group 2 dogs. To confirm adequacy of ablation, we electrically stimulated the nerves (20Hz, 2ms pulse width, 20s, 15mA) before and 5 minutes after each ablation. Cryoablation was considered complete when stimulation of each ablated nerve no longer produced any changes of heart rate or blood pressure. Subsequently, DSI recording wires were implanted to the unablated (upper) half of the LSG, the superior cardiac branch of the left vagal nerve cranial to the ablated portion and to LA epicardium. The wires were connected to a DSI transmitter, which was implanted into a subcutaneous pocket. A pacing lead was implanted onto the LA appendage (LAA) and connected to a subcutaneously positioned Medtronic Itrel neurostimulator for chronic atrial pacing (20Hz, twice diastolic threshold, 2ms pulse width). A small right thoracotomy was then performed. The RSG was identified, and the caudal half of the RSG cryoablated. The chest was closed and the animal recovered.

Pacing Protocol and Second Surgery

Table 1 illustrates the pacing protocol. After a week of post-operative recovery, the DSI was turned on to record baseline rhythm for a week. Atrial pacing was then commenced. Pacing was performed for a week at a time, alternating with non-paced monitoring periods. When pacing was switched off, the immediate rhythm was non-sustained pacing-induced AF. When pacing-induced AF terminated, the ensuing rhythm was monitored for 24 hours to determine if there were paroxysmal atrial arrhythmias. After this, pacing was resumed. The alternating pacing-monitoring sequence was repeated until sustained (>48 hours) AF developed. Forty eight hours of monitoring was performed while the dog was in sustained AF. When sustained pacing-induced AF terminated, a further 24 hr of rhythm was monitored. The dog then underwent second surgery. During second surgery, blood was sampled using a protocol similar to first surgery. The dog was then euthanized and the hearts fixed in 4% formalin for 1 hr and stored in 70% alcohol for histology.

Table 1. Pacing protocol.

1st surgery	Stage 1	Stage 2	Stage 3		Repeat stage 3	Stage 4		2nd surgery
	Post-op	Baseline Rhythm	Pacing	No pacing*		Sus. AF	No AF
	1 wk	1 wk	1 wk	24hr		24hr	24hr

Shaded boxes indicate periods with DSI monitoring whereas unshaded boxes indicate unmonitored periods.

^*Once pacing-induced sustained AF terminates, monitoring was performed for 24-hr. The pacing-monitoring cycle is then repeated.

Abbreviations: Sus., sustained

Data analyses

Manual analyses were done for all raw data to correlate ANA with the occurrence of atrial arrhythmia. We identified premature atrial contractions (PACs), PAF and paroxysmal atrial tachycardia (PAT) episodes. PAT was diagnosed when there was an abrupt (>50 bpm/s) increase in the atrial rate to > 200 bpm that persisted for at least 10 s.¹² Isolated PAC was defined as an isolated premature atrial beat followed by a pause before resumption of a regular baseline atrial rhythm. A second method of analyses was to use custom-designed software to automatically import, filter and analyze both the ANA and ECG recordings. Hilbert Transforms were used to convert the filtered ECG signal into its instantaneous amplitudes and frequencies.¹³ The activation cycle lengths (RR intervals) and standard deviation of RR intervals (SDRR) were determined. Data from SGNA and VNA were high-pass (100 Hz) filtered, rectified, integrated with a 100-ms time constant and then summed to represent averaged nerve activity over: (a) 5-s segments for 30-s before and after onset of atrial arrhythmias, (b) 10-s segments over 24-hr for averages of ANA.

Immunohistochemistry

Five micrometer sections were cut from paraffin blocks of the LA and RA appendages and free walls, left sided PVs, LSG, RSG and superior cardiac branch of the left vagal nerve. The sections were stained with tyrosine hydroxylase (TH) to label sympathetic nerves, growth associated protein 43 (GAP43) for growing nerve cones, and Choline Acetyltransferase (ChAT) for cholinergic nerves.

Catecholamine and Nerve Growth Factor ELISA Assays

Plasma Norepinephrine (NE) and nerve growth factor (NGF) were assayed using competitive enzyme immunoassay kits from the Alpco Diagnostics and Promega, respectively. Transcardiac level (ng/ml) was defined by CS minus AO concentrations.

Statistical Analyses

Data were expressed as mean ± SD. Student’s t-tests with unequal variance were used to compare the means of two groups. ANOVA with Neuman-Keuls tests were used to compare means of multiple groups. Repeated measures ANOVA was performed to compare transcardiac NE. Pearson’s Chi-squared tests were performed to assess association, in contingency-table analyses. A p value of ≤ 0.05 was considered statistically significant.

Results

All dogs successfully completed the study protocol. The total duration of monitoring was 6±2 weeks in Group 1 dogs and 9±4 weeks in Group 2 dogs. All Group 2 dogs had transient Horner’s syndrome after 1^st surgery that resolved completely after 1 week.

Paroxysmal Atrial Arrhythmias

Compared to baseline, Group 1 dogs had significantly increased incidence of paroxysmal atrial arrhythmias after chronic intermittent atrial pacing. After 3 ± 1 cumulative weeks of pacing, 3/7 dogs developed PAF (4 ± 2 episodes/d vs. 0 at baseline, p=0.046) and all dogs had PAT (10 ± 3 episodes/d vs. 2 ± 1 episodes/d at baseline, p<0.001) and PACs (4 ± 1 episodes/d vs. 0 at baseline, p=0.002). Figure 2A illustrates the circadian variation of arrhythmic episodes. Out of a total of 283 episodes (31 PAFs, 194 PATs, and 58 PACs), 36% occurred between 4 am to 8 am, 36% between 8 am to 12pm, 16% between 12 pm to 4 pm, 3% between 4 pm to 8 pm, 6% between 8 pm to 12 pm, and 3% between 12 am to 4 am. Figures 2B and 2C illustrate examples of PAC (arrow) and PAT, respectively, in Group 1 (non-cryoablated) dogs. Note the increased sympathovagal discharge preceding both PAC and PAT. Figure 3A is a typical example of PAF in Group 1 dogs. In the initial 20-s, there was relatively quiescent SGNA and VNA associated with a sinus arrhythmia. An abrupt increase in SGNA and VNA immediately (<5-s) preceded the onset of PAF. We also observed multiple episodes of PAT to PAF conversion. Figure 3B is an example. Figure 3C is a 6-s close up of the same episode shown in Figure 3B, straddling the initiation of PAF. An initial increase in VNA (1) followed by increased SGNA (2) resulted in an acceleration of PAT from 521 bpm to 562 bpm, and a paradoxical reduction of ventricular rate (increased RR interval). This phenomenon was probably due to increased repetitive anterograde concealment associated with faster atrial rate and vagal effects on atrioventricular node. A second increase in VNA (3) followed closely by a massive burst of SGNA (4) preceded the onset of PAF by approximately 3-sec.

Figure 2.

Paroxysmal atrial arrhythmias in Group 1. (A) Circadian incidence of paroxysmal arrhythmias (PAC, PAT and PAF combined) over a 24-hr period. (B): Arrow points to the PAC. (C): PAT induced by simultaneous sympathovagal discharge.

Figure 3.

Two examples of PAF. (A) Sinus rhythm to AF conversion. (B) Atrial tachycardia to AF conversion. (C) Magnified from the center of Panel B (line segment above ECG), showing that the elevated VNA accelerated atrial rate, leading to paroxysmal reduction of ventricular rate (prolonged RR interval) before conversion to AF.

There were specific patterns of sympathovagal discharge unique to each arrhythmia type. Figure 4 is a more detailed quantitative analyses of SGNA and VNA 30 seconds before and after the onset of PAT, PAF (panel A), PACs (panel B) and sinus tachycardia (panel C). There were no significant differences in the patterns of ANA preceding PAT and PAF. Hence, the data were combined (panel A). There were distinctive patterns of ANA preceding the onset of different arrhythmias. Sinus tachycardia (C) can be distinguished from atrial arrhythmias (PACs, PAT and PAF) by antecedent sympathetic activation accompanied by withdrawal of VNA. On the other hand, sympathovagal activation rather than sympathetic activation alone, precedes the onset of paroxysmal atrial arrhythmias (A and B). A greater duration and extent of SGNA rise led to tachyarrhythmia (PAT/PAF) whereas a smaller and less sustained increase of SGNA led to a PAC. To further demonstrate a positive association between sympathovagal discharge and arrhythmias, we analyzed the frequency with which sympathovagal discharge occurred during periods with and without arrhythmia. Nerve discharge was defined as the presence of a three-fold increase in averaged integrated nerve activity within a 5-s period over baseline nerve activity. Sympathovagal discharge was observed to immediately (10-15s) precede the onset of PAC, PAF and PAT in 73% of episodes. On the other hand, when analyzing periods without arrhythmia, sympathovagal discharge occurred in only 13.3 ± 3.9% (p<0.001) of the time.

Figure 4.

Distinctive patterns of left SGNA (black point) and VNA (blue point) 30-sec before and after the onset of PAT and PAF (A), PAC (B) and sinus tachycardia (C) in group 1 dogs, with the time of onset as time zero. There was no significant difference between PAT versus PAF, hence the data were combined. *, p<0.05 vs -30s; **, p<0.01 vs -30s; †, p<0.05 vs -5s

Effects of Cryoablation

Figure 1E shows histological sections of the RSG in group 1 dogs. The area enclosed by a square is enlarged in Panel F. Both were stained with trichrome and the black arrows point to fibrous tissue (blue) at the sites of cryoablation. Adjacent to these sites are surviving nerves (marked as “S”). Panel G is a similar section to panel F, showing TH-positive staining in the surviving portion of the RSG. Panel H is an enlargement of the left portion of panel G. Brown colored structures (arrows) indicate positively stained nerve structures. There are intact (unablated) ganglion cells in this slide.

Effect of Cryoablation on Upstream Autonomic Nerve Activity

Figure 5 shows 24-hour averages of ANA, mean RR and standard deviation of consecutive RR intervals (SDRR) during baseline sinus rhythm (before atrial pacing) on postoperative days (POD) 1 and 10. Compared with group 1 dogs, group 2 dogs had significantly lower POD1 levels of SGNA (2.91 ± 1.46 mV vs. 5.23 ± 1.35 mV, p=0.018), VNA (0.75 ± 0.18 mV vs. 1.12 ± 0.52 mV, p=0.05), mean RR (621 ± 71 vs. 750 ± 43 ms, p=0.006) and SDRR (141 ± 63 ms vs. 201 ± 53 mV, p=0.049). However, the levels of SGNA (p=0.041), VNA (p=0.052), mean RR (p=0.002) and SDRR (p=0.003) recovered by POD10 compared to POD1, to the point where there was no significant difference between POD10 levels of SGNA, VNA and mean RR and SDRR between groups 1 and 2 dogs.

Figure 5.

Effects of cryoablation on upstream ANA, mean RR and SDRR on postoperative days (POD) 1 and 10.

Figure 6 illustrates the effects of ANA on heart rate in group 1 and group 2 dogs during baseline sinus rhythm (before atrial pacing). Both examples were obtained on POD10 and beyond. In group 1 dogs (A), an initial period of increased VNA (a) and relatively quiescent SGNA was associated with a sinus arrhythmia. This was followed by a burst of SGNA (b) and VNA withdrawal (c) that led to the onset of a sinus tachycardia. In group 2 dogs (B), increased VNA (d) and SGNA (e) appear to have only small effects on heart rate, indicating that the left stellate ganglion and vagal nerve were effectively disconnected from the heart. Infrequently, however, large bursts of SGNA (Panel C, f) were associated with sinus acceleration in group 2 dogs. However, the change was not as dramatic or as abrupt as in group 1 dogs. The dissociation between ANA and HR in Group 2 dogs was persistent throughout the entire study period.

Figure 6.

Effects of ANA on heart rate in noncryoablated dogs (A) and cryoablated (B, C) dogs. (A) VNA activity (a) is associated with sinus arrhythmia. SGNA burst (b) and VNA withdrawal (c) were associated with sinus tachycardia in a noncryoablated dog. (B) No or little change of atrial rate with either VNA (d) or SGNA bursts (e) in a cryoablated dog. (C) Large burst of SGNA (f) was associated with atrial rate acceleration in a cryoablated dog.

Effect of Cryoablation on Inducibility of Sustained AF

Group 1 dogs developed sustained AF after 3±1 cumulative weeks (2 to 4 weeks) while Group 2 dogs developed sustained AF after 7±4 weeks (range 3-12 weeks) of atrial pacing (p=0.01). However, there was significant individual variability and overlapping durations between these two groups.

Effect of Cryoablation on Paroxysmal Atrial Arrhythmias

Compared to Group 1, Group 2 (cryoablated) dogs had significantly fewer episodes of PACs (2±1 episodes/d vs. 4±1/d, p=0.011), and no episodes of PAT (p<0.001) or PAF (p=0.046) after the termination of pacing-induced AF. The PACs in cryoablated dogs were also preceded by sympathovagal discharge.

ELISA Studies

Figure 7 shows the results of plasma NGF and NE in group 1 and 2 dogs. There was an increase in transcardiac NGF in both groups (Group 1, p=0.047; Group 2, p=0.049) after chronic atrial pacing, compared with baseline sinus rhythm. Transcardiac NE sampled during resting state, was significantly increased after chronic atrial pacing compared to baseline sinus rhythm in Group 2 dogs (-1.1 ± 1.4 vs 5.0 ± 2.9 ng/ml, p=0.006) but not significantly increased in Group 1 dogs (0.1 ± 0.4 vs 1.8 ± 0.6 ng/ml, p=0.42). Transcardiac NE was also sampled immediately after LSG stimulation. Transcardiac NE (LSG stimulation) levels were significantly higher after chronic pacing compared to baseline in Group 1 dogs (12.1 ± 4.6 vs 2.2 ± 1.3 ng/ml, p=0.007) but significantly lower in Group 2 dogs (4.0 ± 7.0 vs 9.0 ± 9.0 ng/ml, p=0.035). These results indicate that sprouted sympathetic cardiac nerve endings were functionally active but permanently disconnected from the LSG.

Figure 7.

Upregulation of transcardiac NGF (A) and NE (B) in cryoablated (Group 2) and noncryoablated (Group 1) dogs. Repeated measures analysis of variance (rmANOVA) was performed to examine the effects of group, baseline/pacing, and rest/LSG stimulation on transcardiac NE. The rmANOVA included two-way and three-way interactions between the three factors and accounted for the four measurements from each dog, and allowed different variances for the measurements.

Immunohistochemistry Studies

Figure 8 compares the results of immunostaining in the atria between normal control and groups 1 and 2 dogs. The atrial nerve density for each group was expressed as a mean of nerve densities in the LAA, RAA, and left pulmonary veins (PVs). The densities of TH-positive nerves within the atria were 4249±1198 μm²/mm² in Group 1 and 7724±1476 μm²/mm² for Group 2. Both were significantly higher than 1521±488 μm²/mm² in normal control dogs (p=0.007 and p<0.001 respectively). Group 2 dogs had higher TH-positive nerves than Group 1 dogs (p=0.005). Similarly, the density of GAP43 positive nerves were higher in Group 1 (9218 ± 2214 μm²/mm², p=0.007) and Group 2 (12647 ± 3293 μm²/mm², p=0.002) compared to normal control (2141 ± 405 μm²/mm²). GAP43 nerve density was higher in Group 2 compared to Group 1 dogs (p=0.044). There was a trend towards increased ChAT-positive nerve densities in Group 1 and 2 dogs, compared with normal control, however, this was not statistically significant (Group 1: 7430±2110 μm²/mm², p=0.1; Group 2: 8205±1823 μm²/mm², p=0.08; Normal Control: 4611±4322 μm²/mm²). There was no significant difference between ChAT-positive nerve densities between group 1 and 2 dogs (p=0.89).

Figure 8.

Histological sections of TH, GAP43 and ChAT atrial nerves in normal control, noncryoablated (Group 1) and cryoablated (Group 2) dogs. Both Group 1 and Group 2 dogs have significant nerve sprouting and sympathetic hyperinnervation but parasympathetic nerve densities were not significantly different from normal control.

Discussion

In the present study we demonstrated that intermittent rapid pacing may lead to PAF and PAT episodes. We also found that simultaneous sympathovagal discharges preceded the onset of these paroxysmal atrial arrhythmias, and that specific patterns of autonomic nerve discharge differentiated between PAC, PAF/PAT and sinus tachycardia. Cryoablation of the stellate ganglia and the superior cardiac branches of vagal nerve eliminated all episodes of PAF and PAT. Surprisingly, these antiarrhythmic effects were associated with cardiac nerve sprouting and sympathetic hyperinnervation in the atria of cryoablated dogs. The data suggests that decentralization rather than denervation of sympathovagal nerves is the antiarrhythmic mechanism of stellate ganglion and vagal nerve ablation.

Sympathovagal Discharge and Paroxysmal Atrial Arrhythmias

In the present study, we found that sympathovagal discharge characterizes the onset of paroxysmal atrial tachyarrhythmia. By contrast, sinus tachycardia is preceded by sympathetic discharge alone. These data highlight the importance of concurrent vagal activation as an arrhythmic trigger, even in the presence of sympathetic hyperinnervation. In addition, atrial tachyarrhythmias (PAF and PAT) could be distinguished from ectopic beats by a greater and more sustained antecedent elevation of SGNA. Because of such specific ANA patterns associated with different arrhythmias, it is reasonable to conclude that ANA is a trigger for these arrhythmias. The hypothesis is strengthened by (1) consistent findings across a large number of arrhythmic episodes, (2) the close temporal relationship of ANA discharge and arrhythmia onset, (3) ANA activation preceded, rather than followed, the onset of arrhythmia and (4) elimination of arrhythmia with sympathovagal cryoablation. The cellular mechanism of these observations can be explained by late-phase 3 early afterdepolarization³ due to increased intracellular calcium and shortened action potential during sympathovagal discharge.¹⁴

Effects of Cryoablation on Autonomic Nerve Activity

We demonstrated that cryoablation was an effective method to achieve permanent damage to the SG, as evidenced by fibrosis at the site of ablation. The lesion caused by cryoablation was discrete, and there was evidence of surviving neurons upstream to the lesion site. The presence of immunopositivity to TH at the upstream site confirmed that the neurons were functionally viable, and enabled successful recording of ANA from the surviving neurons. The fact that SGNA and VNA both recovered to levels similar to noncryoablated dogs further supports this conclusion. However, these neurons were effectively disconnected from the heart, as discharges of SGNA or VNA were frequently not associated with any significant changes in heart rate, and LSG stimulation did not trigger cardiac NE release. Immediately after cryoablation, on POD 1, there was a dramatic reduction of SGNA and VNA, probably due to reversible damage induced by the low temperature. This was associated with a reduction of mean RR and SDRR compared with noncryoablated dogs. Nerve recordings confirmed significant disconnection of both sympathetic and vagal nerves from the heart, as the heart rate failed to follow the changes of ANA. On POD 10, there was a recovery of SGNA, VNA, mean RR and SDRR. At the same time, there was increased association of SGNA and VNA with appropriate changes in heart rate. These findings suggest re-establishment at least partially, of neural connections by nerve sprouting between the surviving neurons and the non-SG neural pathways that traffic to the heart. Alternatively, these changes could be due to increased circulating catecholamines induced by sympathetic discharges elsewhere in the body. Histological analyses confirmed increased innervation of the heart by sympathetic nerves both in cryoablated and in noncryoablated dogs as compared with normal controls. We conclude that cryoablation of extrinsic cardiac nerves effectively disrupted the connection between the central nervous system and the heart.

Effects of Cryoablation on Inducibility of Arrhythmia

Elvan et al¹⁵ found that radiofrequency ablation of the atria eliminated pacing-induced sustained AF, probably through autonomic tone modulation. In the present study, we found that cryoablated dogs took significantly longer periods of atrial pacing to achieve sustained electrically induced AF, a finding compatible with the results of that previous study. However, all dogs eventually developed sustained AF. One possibility is that sustained AF occurred primarily due to the changes of electrophysiological characteristics (e.g. effective refractory period) of the atria or the thoracic veins rather than due to sporadic ANA. Alternatively, rapid atrial pacing may have induced nerve sprouting of the intrinsic autonomic ganglia. The intrinsic autonomic ganglia are capable of developing spontaneous neural activity independent of the extrinsic control.^{16, 17} These neural activities could be a cause of sustained AF. A recent clinical study¹⁸ showed that recipients with orthotopic cardiac transplantation had a very low incidence (0.3%) of AF, as compared with the 21% incidence of AF in patients who have undergone coronary bypass surgery. Only 3 patients had AF in the former group, and all 3 had bicaval anastomosis. The authors suggested that the complete isolation of thoracic veins in these patients is a reason for a low incidence of AF. However, complete isolation from both extrinsic and intrinsic nervous system could also contribute to the low incidence of AF in these patients.

Autonomic Nerve Modulation as a new Strategy for AF Management

The long-term efficacy of currently available antiarrhythmic drugs for preventing AF recurrence is far from ideal, because of limited efficacy and potential side effects, particularly proarrhythmia.¹⁹ In the past decade, non-pharmacological electrophysiological therapies have become more popular. Targeted destruction of PV foci by radiofrequency catheter ablation suppresses PAF.²⁰ PV denervation enhances long-term results of circumferential ablation for PAF.²¹ However, problems include potential recurrence and a small but nontrivial risk of PV stenosis, systemic thromboembolism, pericardial effusion, cardiac tamponade, esophageal perforation and phrenic nerve paralysis. These limitations stimulate research toward the development of less aggressive and yet effective procedures. We demonstrated in this study that effective autonomic decentralization can be performed by targeting some but not all extra-cardiac autonomic nerves. This limited strategy was effective at preventing paroxysmal atrial arrhythmias while sparing the LA. Our findings support the assertion that ablation of extra-cardiac nerves can be an effective alternative to ablation of intra-cardiac nerves by LA ablation for the treatment of AF.

Absence of Cardiac Denervation After Stellate Ganglia Ablation

Wijffels et al⁸ pioneered the concept of AF begets AF using an sheep model of intermittent pacing. The authors hypothesized that progressive electrophysiological remodeling was the mechanism by which intermittent pacing induces sustained AF. Subsequent studies^{22, 23} suggest a contribution of neural remodeling to AF maintenance. An unexpected finding of the present study is that bilateral stellate ganglia ablation failed to either prevent the induction of sustained AF or produce significant cardiac denervation. Rather, it produced nerve sprouting and sympathetic hyperinnervation above and beyond that induced by rapid atrial pacing. A possible mechanism is the upregulation of cardiac NGF, which might have induced nerve sprouting from the ganglion cells in the intrinsic autonomic nervous system.²⁴ The increased transcardiac NE levels suggest that these TH-positive nerves were functional. These findings suggest the antiarrhythmic effects of stellate ganglion ablation were achieved by disconnection between the nerve discharges upstream of the ablation site and the atria, rather than by inducing cardiac denervation. Because sympathetic nerves are responsible not only for releasing but also for reuptake of the NE, the presence of abundant sympathetic nerve terminals might be beneficial by providing a sink to NE in the heart.

Study Limitations

We did not perform cryoablation of sympathetic and vagal nerves alone. Therefore, we could not determine the relative importance of sympathetic and vagal nerve activation as a trigger for paroxysmal atrial arrhythmias. It is possible that ablation of stellate ganglia alone could prevent PAF. However, previous studies have shown that partial and/or heterogeneous denervation might promote, rather than impede, the development of electrically induced AF.^{25, 26} A more complete ablation of extrinsic autonomic nervous system as done in this study might be needed to effectively prevent PAF. We ablated only the left vagal nerve because paroxysmal AF usually originates from the pulmonary veins. Whether or not the intact right vagal nerve contributed to the anti-arrhythmic effects is unknown.