Autonomic Modulation of Atrial Fibrillation

Central Illustration

8][9] The skin is well innervated by sympathetic nerves. 10In dogs, the postganglionic sympathetic nerve fibers of neck and thorax originate primarily from the SG. 11Our canine studies confirmed that the SKNA can be used to estimate the SGNA. 7,12,13 then performed a human study by inserting a temporary pacing wire into the epicardial fat pad to record the intrinsic cardiac nerve activity after open heart surgery in 11 patients. 14We found that lowamplitude SKNA and intrinsic cardiac nerve activity were present at all times, but the burst discharges were observed much less frequently.Figures 2 and 3 of Shen et al 14 show SKNA and intrinsic nerve activity bursts might occur simultaneously.Both SKNA and burst intrinsic cardiac nerve activity were associated with the onset of premature atrial contraction and premature ventricular contraction and AF.To further investigate the relationship between SKNA bursts and cardiac arrhythmias (including AF), we 15 recorded neuECG in 10 patients with AF and 6 patients with ventricular tachycardia (VT) or ventricular fibrillation (VF) episodes.Clustering was defined by an arrhythmic episode followed within 1 minute by spontaneous recurrences of the same arrhythmia.
To better understand the relationship between SKNA bursts and arrhythmia clustering, we 9,16 developed a new method to objectively determine the SKNA bursts.Figure 1 illustrates the method using a 24.7-hour long recording.We first analyzed the average voltage of skin sympathetic nerve activity per sample (aSKNA) in each 60-second window.We then plotted the histogram of aSKNA that shows the proportion of windows with a given amplitude (Figure 1A).There are 2 groups of data; each can be fitted with a Gaussian distribution.The first (left) Gaussian distribution represents the baseline nerve discharges, and the second (right) Gaussian distribution represents the burst activity.We used the mean plus 3Â SD of the first Gaussian distribution (1.0434   mV) as the threshold to separate these 2 groups of activities.Figure 1B shows actual recordings at points a, b, and c in Figure 1A. Figure 1C shows continuous recording over the same period as shown in Figure 1A.
The onsets of AF were indicated by red dots for immediate recurrence of atrial fibrillation (IRAF) and orange dots for non-IRAF.The IRAF was defined as the recurrence of AF within 1 minute from the termination of a previous episode.The horizontal dotted red line indicates the threshold for burst determination (1.0434 mV).The leading edge and trailing edge of each burst was automatically determined.The binary time series graph at the bottom of Figure 1C shows the SKNA burst (black) versus nonbursting period (white).There was a total of 68 SKNA bursts in the entire recording.
Using that method, we showed that the total duration of SKNA bursts associated with AF was longer than that associated with sinus rhythm, indicating elevated sympathetic tone during AF.We found that there was a latency between the onset of SKNA bursts and the onset of AF, with a median of 9.0 (IQR: 5.0-15.5)minutes.In the same study, we also observed a 3-minute latency between SKNA bursts and the onset of VT.The presence of this latency implies that there needs to be a sufficient accumulation of neurotransmitters locally to induce an episode of AF.To achieve that effect, large and sustained SKNA bursts are needed for AF induction.

SKNA BURSTS AND THE TERMINATION OF AF OR AT
While studying the association between SKNA bursts and the onset of AF, we unexpectedly discovered that there is also an association between SKNA bursts with the termination of AT.That study 17 included 11 patients, including 3 with AT and 8 with AF.
One of the patients with AT had a total of 32 episodes of AT.We observed that whereas active SKNA bursts are present prior to the onset of AT, there are also large SKNA bursts and transient heart rate acceleration immediately preceding the termination of AT.
Because this phenomenon was observed in only 1 of 3 patients with AT studied, its clinical importance remains unclear.Similarly, in all 8 patients with paroxysmal AF, large SKNA bursts were frequently, albeit not always, present before the AF termination.Because SKNA bursts often occurred throughout the period of AF perpetuation, it was difficult to conclude that these SKNA bursts were responsible for termination of AF.A potential clinical implication of these findings was that suppression of the SKNA bursts might prevent AF initiation, but suppression might also prevent AT/AF termination in some patients.

THE POTENTIAL EFFECTS OF AF ON SNA
The rapid ventricular rate and irregular cycle length during AF can lead to intermittently reduced blood pressure, which is known to increase the muscle sympathetic nerve activity (SNA). 18In canine models, the SGNA was significantly elevated 15 seconds before Autonomic Modulation of AF and remained elevated 30 seconds after the onset of the atrial arrhythmia. 3In humans, large and sustained SNA were associated with the temporal clustering of AF and VT/VF. 16These findings suggest that AF itself may increase the SNA, which in turn triggers the next episode of AF and leads to AF cluster.
Because the relationship of AF with the SNA may be bidirectional, reducing SNA by neuromodulation may reduce both the frequency and the total burden of AF.

CARDIAC AFFERENT INNERVATION AND AF
Cardiac afferent reflexes are abnormal in patients with AF during sinus rhythm and dysfunctional during AF. 19In canine models, apnea increases GP activity, followed by vagal bursts and tonic SG firing. 20siniferatoxin decreases SNA and GP nerve activity and abolishes apnea's electrophysiologic response and AF inducibility.These data imply that sensory neurons and afferent nerves play a role in apnea-induced AF.However, because it is difficult to selectively record afferent nerves in ambulatory animals or humans, the direct relationship between afferent nerve firing and onset of AF remains unclear.

SKNA ELEVATION AND MI
A vast majority of patients with AF (82%) have either obstructive or nonobstructive coronary artery diseases. 21A higher burden of coronary artery disease within all arteries supplying blood flow to the atrial myocardium are associated with higher odds of newonset AF at 1 year. 224][25] The elevated sympathetic tone can contribute to the development of AF and other types of cardiac arrhythmias.We recently performed a study to prospectively measure the SKNA in 128

Tsai et al
patients with acute coronary syndrome (ACS) and compare the magnitude with that of control subjects. 26The neuECG was recorded with ECG lead I configuration at baseline, during mental math stress, and during recovery.Approximately 55% of patients with ACS have ST-segment elevation myocardial infarction (STEMI) and 45% have non-ST-segment elevation ACS.Ventricular arrhythmias were seen in 18% of the patients with ACS.The primary findings of the study, as shown in Table 1, were that the baseline, stress, and recovery average SKNA were higher in ACS than in control subjects.The age of the patients was positively associated with aSKNA.The SKNA in the ACS subject showed more nerve bursts in each phase than in the control subjects.Among patients with ventricular arrhythmias, 15 (83%) had nonsustained VT, 1 (6%) had sustained VT, and 2 (11%) had VF.
Figure 2 shows a recording from a patient with STEMI.
Large SKNA bursts correlated with the occurrences of VTs.After adjusting for the potential confounders, the ventricular arrhythmias occurrence elevated by 23% when aSKNA level elevates by 0.1 mV and by 8.33fold if aSKNA level elevated by 1 mV.
There were sex differences of SKNA among control subjects but not among patients with ACS.In control subjects, aSKNA was higher in women than in men at baseline and recovery phases but not during mental stress.These findings suggest that sympathetic reserve (stress/baseline) is higher in men than in women.In comparison, the aSKNA was elevated in both men and women during ACS but there were no with PVI þ RDN in patients with hypertension also suggested a significant beneficial effect of RDN in controlling AF. 28 The role of RDN in AF is currently strongest in those with concomitant hypertension.
These data do not rule out the possibility that RDN achieved AF control by modifying hypertension, a known risk factor for AF. 29 EFFECTS OF RDN IN NORMAL DOGS.We hypothesized that the mechanisms of RDN might result from the remote neuromodulation that involves both the left SG and the brainstem. 30In that study, 9 normal dogs underwent a radio transmitter implantation to record SGNA and were monitored for 2 weeks before and 2 months after radiofrequency RDN procedure.
We found that bilateral RDN caused significant central and peripheral sympathetic nerve remodeling and reduced SGNA.The SG and the brainstem remodeling were documented by histologic examinations.Figure 4 shows the histologic changes in the brainstem, with neurons stained positive for TUNEL.In addition, we performed 18 F-2-fluoro-2-deoxyglucose uptake study in 2 dogs.The results show that there were À16% and À13% changes in 18 F-2-fluoro-2deoxyglucose uptake, respectively, in pons and medulla.These findings indicate that there are both SG and brainstem remodeling after RDN.95% CI: 1.20-1.43)(Figure 5C).These dogs also have a significant number of spontaneous AT episodes, characterized by abrupt onset and termination of the atrial arrhythmias.Figure 6A shows an example of SGNA followed by AT onset (arrow).Figure 6B shows the termination of AT after SGNA activity cessation.

Figures 6C and 6D
show that RDN decreased the number of AT episodes and duration.Both the present study and the study by Zhang et al 32 showed that RDN reduced sympathetic tone.These effects led to a reduction of ventricular arrhythmias in the Zhang et al 32 study and a control of AT episodes in our study.

POSSIBLE MECHANISMS OF NEURAL REMODELING
AFTER RDN.Dying back neurodegeneration is introduced by either genetic lesions or toxic conditioninduced axonal loss. 33The "dying back" hypothesis proposes that degeneration of each axon starts at the distal end and moves retrogradely. 34Alzheimer disease, Parkinson disease, and amyotrophic lateral sclerosis are among the diseases associated with dying back degeneration. 35Injury-induced axon degeneration has been used extensively to study axon degeneration. 35Those studies have identified molecular mechanisms that act either to protect or to promote degeneration.We reported that following RDN, an average of 40% of the ganglion cells in the right and left SG are TUNEL-positive, indicating cell death. 30We also observed brainstem remodeling in dogs that underwent RDN.These changes are best explained by injury-induced dying back degeneration.The process of transneuronal degeneration then induced brainstem remodeling. 36However, we do not have definitive proof that these processes are responsible for all the histologic changes observed in the SG and the brainstem.GP AND LIGAMENT OF MARSHALL ABLATION.Sympathetic nerves from the middle cervical and SG pass along the ligament of Marshall (LOM) to innervate the left ventricle. 37GP modulate the autonomic interactions between the extrinsic and intrinsic cardiac ANS. 38Both the GP and the LOM are targets during catheter ablation of AF. 39 Zhao et al 40   the SG can inhibit left SG function and improve cardiac remodeling in a large animal model of chronic MI. 44 Lee et al. 45 used botulinum injection into the celiac plexus for celiac plexus block and found it is an effective alternative to RDN in managing refractory hypertension.Whereas botulinum has significant and prolonged functional effects, direct botulinum injection into the superior cervical ganglion in rabbits did not cause significant histologic changes. 46Because it is thought that neurotoxin application to the nervous structures can modulate them, randomized clinical trials have been conducted to study the effects of botulinum toxin on postoperative AF.Although some studies show significant beneficial effects during the postoperative period and during long-term followup, [47][48][49] a subsequent larger study did not reproduce these positive results. 50A common limitation of these studies is that the mechanisms of postoperative AF is complex. 51Because inflammation, ischemia, and other factors may contribute to the onset of postoperative AF, neuromodulation alone may not be effective in preventing its occurrence.The negative results may not be applicable to ordinary AF, which is more dependent on ANS activity as its triggers.
However, because botulinum toxin injection does not cause histologic changes, its affects may be shortlasting and may not be ideal for permanent neuromodulation.
RAPID ELECTRICAL STIMULATION.The excessive use of axons may lead to intracellular Ca 2þ overload, a common mechanism of axonal degeneration and cell death. 52,53That is probably why sustained electrical stimulation of the perforant path causes epileptic brain damage in rats. 54Both the dendritic and somal degenerative changes in that study 54 closely resemble the "excitotoxic" type of damage that the putative transmitters glutamate and aspartate are known to cause.Based on those findings, we hypothesized that rapid electrical stimulation of the peripheral sympathetic axons can modulate/damage the sympathetic neurons in the stellate and other ganglia, leading to reduced sympathetic outflow and better arrhythmia control.Our laboratory has tested this hypothesis using several different models.
C e r v i c a l v a g a l n e r v e s t i m u l a t i o n .6][57] Our canine experiments showed that lowlevel vagal nerve stimulation (VNS) could suppress SGNA and the incidences of paroxysmal AT in ambulatory canines. 58This was followed by another study of VNS during sustained AF in a canine model. 59We found that VNS reduced SGNA and ventricular rate.A histologic examination of the left SG showed extensive damage.In the damaged region, the cells showed pyknotic nuclei, reduced tyrosine hydroxylase staining, increased percentage of tyrosine hydroxylase-negative ganglion cells, and positive TUNEL staining.Because VNS is clinically used to control drug-resistant epilepsy, we recorded SKNA in 26 patients with drug-resistant epilepsy who were admitted for video electroencephalographic monitoring.Among them, 6 were treated with VNS and 20 were not.We found that patients with VNS had significantly lower SKNA than those without VNS.
These findings are consistent with the effects of VNS in canine models.Because reduced ventricular rate was observed while the stimulator was off, the rate control was not achieved by the direct activation of the parasympathetic nerves.
S u b c u t a n e o u s n e r v e s t i m u l a t i o n .In addition to innervating the heart, the SG also innervates the cutaneous nerves of the neck and the upper thorax. 11 hypothesized that stimulating the subcutaneous sympathetic nerves in those areas would have the same effects as stimulating the sympathetic components inside the vagal nerve.To test that hypothesis, we performed a series of studies in canine models by using implanted VNS devices to stimulate the subcutaneous sympathetic nerves.In the first study, 60 we performed subcutaneous nerve stimulation (ScNS) at the Xinshu acupoint and at the left lateral thoracic nerve for 2 weeks.The results show significant SG remodeling and suppression of paroxysmal AT in ambulatory dogs.We then performed a study to test the hypothesis that ScNS can control the ventricular rate in persistent AF.We found that thoracic high-output (2.5 mA and 3.5 mA) stimulation had the opposite effects. 63A limitation of the abovementioned studies is that we performed an incision to look for subcutaneous nerve structures for stimulation.The need for a several centimeter-long incision may prevent clinical translation.We hypothesize that because skin is well innervated by the sympathetic nerves, a blindly inserted electrode might be able to stimulate the sympathetic nerves and cause SG remodeling.We performed a canine study to test that hypothesis. 64 T r a u s s t i m u l a t i o n .Because the cutaneous layer of external auditory canal (Ramsay Hunt zone) was innervated by a sensory branch of the X (10th) cranial (vagus) nerve, transcutaneous electrical stimulation of the nervous system on the Ramsay Hunt zone might also stimulate the vagal nerve and therefore is beneficial for seizure control. 65Yu et al 66 performed low-level tragus stimulation in a canine model and showed that it reverses AF inducibility, suggesting a potential noninvasive treatment of AF.The term "low level" is used to describe a voltage level that does not slow the sinus rate or atrioventricular conduction.In that first canine study, the average threshold was 9.8 V. Assuming the impedance is w1,000 U, the strength of stimulation is high enough to cause neural remodeling.Stavrakis et al 67,68 attempted to translate these findings to humans.
They showed that noninvasive tragus stimulation was associated with 85% lower AF burden than the ear lobe stimulation. 69However, the positive results were more likely caused by the increased AF burden in the control group than the reduction of AF burden in the experimental group.Another randomized clinical trial showed that Tragus stimulation reduces myocardial ischemia-reperfusion injury in patients with STEMI. 70ese encouraging preliminary results suggest that Tragus stimulation might be useful in AF control, but more studies are needed to determine the mechanisms of action and to support that conclusion.

FUTURE DIRECTIONS
Increased sympathetic nerve activity plays an important role in initiating AF.Neuromodulation methods may reduce sympathetic outflow and achieve AF control.Small randomized clinical trials suggested that neuromodulation methods might be useful in AF control in highly selected patient populations, but the results have not been widely reproduced.Furthermore, none of the studies were powered to demonstrate a mortality benefit.Because AF is a chronic disease that affects a large number of patients, even a small benefit might have a significant impact on public health.Finding a neuromodulation procedure that has a small benefit to a large number of patients might be as important as finding a procedure that delivers a large benefit to a small group of highly selected patients.
We propose that both paroxysmal and persistent AF may benefit from neuromodulation.In patients with paroxysmal AF, the AF burden may be reduced USA.E-mail: Peng-Sheng.Chen@cshs.org.

A
B B R E V I A T I O N S A N D A C R O N Y M S ACS = acute coronary syndrome AF = atrial fibrillation ANS = autonomic nervous system aSKNA = average skin sympathetic nerve activity AT = atrial tachyarrhythmia ECG = electrocardiography/ electrocardiogram GP = ganglionated plexi IRAF = immediate recurrence of atrial fibrillation LOM = ligament of Marshall MI = myocardial infarction PVI = pulmonary vein isolation RDN = renal denervation ScNS = labeling VF = ventricular fibrillation VNS = vagal nerve stimulation VT = ventricular tachycardia J A C C : B A S I C T O T R A N S L A T I O N A L S C I E N C E V O L .8 , N O . 1

FIGURE 1 A
FIGURE 1 A Method to Objectively Determine the SKNA Bursts sex differences at baseline, during stress, and after recovery.Another finding of the study is that the aSKNA was correlated with systemic norepinephrine concentration.Blood samples collected from ACS (n ¼ 20) and control (n ¼ 20) groups were used to study the association between aSKNA (mV) and systemic plasma norepinephrine level (ng/mL).After adjusting for age, the aSKNA was positively correlated with plasma norepinephrine level (R 2 ¼ 0.20, b ¼ 0.65 AE 0.23 mV per ng/mL; P ¼ 0.007).Heightened sympathetic nerve activity likely contributed to the elevated norepinephrine levels.EFFECTS OF NEUROMODULATION RENAL DENERVATION.Renal denervation (RDN) is an extensively studied neuromodulation procedure both for hypertension and for AF control.Relevant to this review, Steinberg et al 27 performed a multicenter, single-blind, randomized clinical trial that included 302 randomized participants.The investigators found that among patients with paroxysmal AF and hypertension, RDN added to catheter ablation, compared with catheter ablation alone, and significantly increased the likelihood of freedom from AF at 12 months.The HR was 0.57, indicating a highly clinically significant effect.A meta-analysis comparing pulmonary vein isolation (PVI) alone

Figure 3
shows the histologic changes in the SG after bilateral RDN.Figures 3A and 3B show increased fibrosis and cell death (Masson trichrome stain).Figures 3C and 3D show a significantly increased number of tyrosine hydroxylase-negative neurons in the SG.Figures 3E and 3F show positive terminal deoxynucleotidyl transferase dUTP nick end labeling (TUNEL) staining, documenting cell death.

J 3
A C C : B A S I C T O T R A N S L A T I O N A L S C I E N C E V O L .8 , N O . 1 0 , 2 0 2 Tsai et al O C T O B E R 2 0 2 3 : 1 3 9 8 -1 4 1 0 Autonomic Modulation of AF RDN IN DOGS WITH ACUTE MI.Because acute MI significantly increases sympathetic tone and cardiac arrhythmia, it is an excellent model to determine whether RDN is antiarrhythmic.Yamada et al 31 performed RDN in a rabbit model of MI and heart failure by coronary ligation.They found that RDN reversed atrial electrical and structural remodeling and suppressed AF inducibility.Zhang et al 32 reported that RDN in a canine model of MI decreased whole-body and local tissue sympathetic activity and reversed neural remodeling in the heart and SG.We carried out a similar study to investigate the effects of RDN on MI in 11 dogs (Figure 5A) (Unpublished data, 2023).The MI was created by left circumflex artery ligation at week 0. All dogs had SGNA, subcutaneous nerve activity, and superior left GP nerve activity monitored by the implanted D70-EEE radio transmitter (Data Sciences International).The control group (n ¼ 5) had MI only and was followed for 8 weeks.The experimental group (n ¼ 6) underwent RDN 4 weeks after MI and was followed for an additional 8 weeks.We used the ratio of integrated SGNA after MI and at baseline to gauge the changes of sympathetic tone.At the end of follow-up, the ratio was 1.78 (95% CI: 1.34-2.22) in the control group, which was significantly (P ¼ 0.004) higher than that of the RDN group (0.94; 95% CI: 0.86-1.02)(Figure5B).The ratio of RR interval at 2 months and RR interval at baseline was significantly (P ¼ 0.030) shorter in the control group (1.14; 95% CI: 0.98-1.29)than in the RDN group(1.31;

FIGURE 2 SKNA 1 .
FIGURE 2 SKNA Bursts and Nonsustained Ventricular Arrhythmia in a Patient With STEMI

FIGURE 3
FIGURE 3 Histology of the Left SG in a Dog With Bilateral RDN

FIGURE 4
FIGURE 4 Immunofluorescence Microscopy Images of the Brainstem at Level 1 in a Bilateral RD Dog

FIGURE 5
FIGURE 5 RDN After MI

FIGURE 6
FIGURE 6 Effects of RDN on ATs by suppressing triggers and reducing the secondary elevation of the SNAs that help maintain AF.Neuromodulation may also result in better rate control in persistent/permanent AF.CONCLUSIONS ANS plays an important role in the initiation and maintenance of cardiac arrhythmias.Neuromodulation procedures can reduce sympathetic outflow through axonal damage and dying back degeneration or through electrical stimulationinduced intracellular calcium accumulation and cell death (See Central Illustration).Transneuronal degeneration may further magnify the effects of neuromodulation procedures by remodeling the remote nerve structures.Although randomized clinical trials have suggested that neuromodulation could improve the outcomes of catheter ablation, the challenge is to find a noninvasive or minimally invasive neuromodulation method that can be easily applied to all patients with AF.FUNDING SUPPORT AND AUTHOR DISCLOSURES This study was supported in part by the Ministry of Science and Technology, Taiwan (grant MOST 110-2314-B-037-111), the Kaohsiung Medical University Hospital Research Foundation (grants SI11001, SI11101, KMUH105-5M07, KMUH-S10707, and NK111P24), the US National Institutes of Health (grants R01HL139829 and OT2OD028190), and the Burns and Allen Chair in Cardiology Research, Cedars-Sinai Medical Center, Los Angeles, California.Dr Peng-Sheng Chen and Lan S. Chen are co-inventors of U.S. Patents awarded to Indiana University.All other authors have reported that they have no relationships relevant to the contents of this paper to disclose.ADDRESS FOR CORRESPONDENCE: Dr Peng-Sheng Chen, Department of Cardiology, Smidt Heart Institute, Cedars-Sinai Medical Center, 8700 Beverly Boulevard, Davis 1016, Los Angeles, California 90048, Tsai et alJ A C C : B A S I C T O T R A N S L A T I O N A L S C I E N C E V O L .8 , N O . 1

TABLE 1
26e Average SKNA in Control Group and ACS Group Adjusted b were obtained from multiple linear regression models adjusted for age, sex, cigarette smoking, alcohol drinking, and betel-quid chewing.bP< 0.05 for significant difference between ACS and control.From Huang et al26with permission.ACS ¼ acute coronary syndrome; adj.b ¼ adjusted mean; aSKNA ¼ average skin sympathetic nerve activity. a