Frequent Premature Atrial Contractions Lead to Adverse Atrial Remodeling and Atrial Fibrillation in a Swine Model

After the animals were fasted overnight, anesthesia was induced with an intramuscular injection of ketamine and acepromazine and maintained by inhalation of 1% to 5% isoflurane (1.5% to 2% for most animals), with each swine mechanically ventilated with 100% oxygen. Femoral venous access was obtained by a percutaneous puncture using the modified Seldinger technique with ultrasound guidance. A temporary, steerable mapping catheter was placed through a femoral vein sheath and atrial effective refractory periods (ERPs) were measured at each of the following pacing sites: right atrial free wall, right interatrial septum, and coronary sinus (Figure S1A). Single paced atrial extrastimuli were delivered at a 400-ms coupling interval from the previous atrial complex and decremented by 10 ms until loss of capture, with a 5-second pause between extrastimuli (sensed extrastimuli that were delivered to mimic spontaneous PACs). The atrial ERP was defined as the longest coupling interval without atrial capture. After a 10-minute waiting period, AF induction was attempted 3 times from the 3 atrial sites (total of 9 attempts) with burst atrial pacing (cycle length, 50 ms of 15 seconds; Figure S1B). All atrial pacing, including extrastimulus testing (ERP measurements) and burst pacing (AF induction), was performed using the same stimulus strength (4 mA; duration 2 ms) for uniformity.

AF sustainability, defined as the average of the maximum AF duration at each of the 3 sites, and AF inducibility, defined as the percentage of an inducible AF duration of ≥5 seconds among a total of 9 AF induction attempts, were assessed.

The right neck was cut down to expose the external jugular vein. The vein was tied distally, and after an anterior venotomy was performed, 2 pacemaker leads were introduced directly. One bipolar active fixation lead (Medtronic 5076) was placed in the right atrial appendage (RAA) for sensing. A second pacing lead was implanted in the distal coronary sinus aiming to maximize atrial dyssynchrony (Lat-PAC; Figure 1B) or the interatrial septum (Sep-PAC; Figure 1C). In the regular atrial pacing group (Reg-130), only a single lead was introduced into the distal coronary sinus. A biventricular pacemaker (Syncra or Viva CRT-P; Medtronic) was used to create paced PACs. The RAA lead was connected to the atrial port for sensing intrinsic atrial depolarizations, and the second lead was attached to the LV port for sequential atrial pacing and prevention of oversensing. The right ventricular port was plugged.

After 1 week of recovery, the pacemakers were programmed from sensing-only mode (ODO) to DDD mode to create paced atrial bigeminy (50% PAC burden; Figure S2). For the Reg-130 group, the pacemakers were programmed to the SSI mode to allow constant atrial overdrive pacing at 130 beats/min. In a preliminary study, the mean atrial rates during sinus rhythm and bigeminal PACs were 89±10 versus 121±12 beats/min, respectively (Figure S3). Therefore, regular atrial pacing at 130 beats/min would allow higher atrial rates than those during bigeminal PACs.

For the sensing lead in the RAA, the atrial pacing threshold was set to the minimum level to avoid any unnecessary atrial pacing. The PAC coupling interval (sensed atrioventricular delay on the pacemaker) was programmed to a short coupling interval of 250 ms. This aimed to ensure that the PACs were not conducted to the ventricle, thereby negating any potential confounding attributable to ventricular irregularity. The pacing output was set to ensure atrial capture of the bigeminal PACs while avoiding ventricular and phrenic nerve capture. Pacemaker interrogations were performed at monthly intervals to confirm effective sensing, atrial capture, and absence of AF. Ten pigs underwent the same baseline and terminal electrophysiology study and echocardiography protocol but no pacemaker implantation and were included as controls (n=10).

Transthoracic echocardiography (TTE; M5Sc-D probe/Vivid E95 system; GE Healthcare) was performed at baseline and monthly thereafter. All swine were sedated with midazolam (0.3–0.5 mg/kg), ketamine (12.5 mg/kg), and inhaled isoflurane (1%–5%) by mask. Echo images were then acquired using a 3.5-MHz transducer placed in the left parasternal area in the right lateral decubitus position. The LA area was measured in a blinded fashion using images that were obtained from a modified apical 4-chamber view at end-systole, taking care to exclude the LA appendage and pulmonary veins (Figure S4A). LA function was quantified by measuring the peak LA reservoir strain using 2-dimensional speckle-tracking strain analysis software (EchoPAC version 201; GE Healthcare). A longitudinal strain curve was generated during sinus rhythm and gated for the atrial wall motion among 6 LA segments obtained from the modified apical 4-chamber view at a frame rate of ≥70 frames per second. The peak LA strain was assessed by measuring the average of the peak longitudinal strain across 6 LA segments using the QRS onset as a reference point (Figure S4B). LV function was also quantified by measuring the LVEF using M-mode in the parasternal short-axis view (Figure S5).

At baseline, electrical and mechanical atrial dyssynchrony were evaluated during sinus rhythm or regular pacing from the lateral LA pacing leads or interatrial septal pacing leads at 120 beats/min. The maximum P-wave duration was measured to assess the total atrial conduction time (electrical dyssynchrony) between sinus rhythm and that during regular atrial pacing from the septum and coronary sinus (LA). The maximum P-wave duration was measured from the earliest onset to the latest offset in all ECG leads at a sweep speed of 100 mm/s (Figure 2A). Measurements were made using digital calipers on an electrophysiologic recording system. Furthermore, 2-dimensional speckle tracking strain analysis software (EchoPAC, version 202; GE Vingmed ultrasound) was used to assess mechanical intra-atrial dyssynchrony. The difference in time to peak of the earliest and latest activated segments among the 6 LA segments was measured and the mechanical regional incoordination was assessed during sinus rhythm and at the 2 different pacing sites (Figure 2B).

After 16 weeks, the electrophysiology study and TTE were repeated under general anesthesia using the same protocol as the baseline study. Before the study, the pacemakers were turned off for 24 hours to prevent any acute effects of frequent PACs. If burst pacing–induced AF did not convert to sinus rhythm spontaneously after 7 minutes, electrical cardioversion was performed (this never occurred during the baseline study). In addition, a transseptal puncture was undertaken, guided by intracardiac echocardiography and fluoroscopy using a deflectable sheath (Agilis; Abbott Technologies) and BRK needle. After the deflectable sheath was introduced into the LA, the mean LA pressure was obtained during sinus rhythm. The mean aortic pressure was also measured by the femoral artery sheath during sinus rhythm. A multipolar grid catheter (HD-Grid; Abbott) with 3-3-3–mm interelectrode spacing was then introduced into the right atrium (RA) as well as LA and high-density 3-dimensional electroanatomic maps were created (NavX EnSite Velocity system version 3.0; Abbott) during sinus rhythm (Figure S6). The distribution of the unipolar electrogram voltage among all the RA and LA points was evaluated offline. Conduction velocity was assessed on the LA posterior wall using 5 electrogram pairs from each pacing site for each animal, blinded to group. The local activation time was measured perpendicular to the isochrones in areas of least isochronal crowding, with the conduction velocity measured as the surface distance between each point pair divided by the difference in the local activation time.

An acute study with 5 additional swine was conducted to examine the LA pressure during bigeminal nonconducted PACs versus conducted PACs, delivered from the coronary sinus, to better understand the hemodynamic effect of the PACs against a closed mitral valve. The LA pressure and aortic pressure were recorded during sinus rhythm, bigeminal nonconducted PACs at a coupling interval of 250 ms, and bigeminal conducted PACs at a coupling interval of 350 ms, which allowed conduction to the ventricle, using the same methods as the terminal study.

The swine were euthanized and full-thickness 2-cm² samples were obtained from the LA posterior wall, LA anterior wall, and RA appendages. Sections were preserved in buffered formalin and embedded with paraffin. Sections from each sample were then stained with Masson trichrome and quantification of fibrous tissue was performed on photomicrographs taken using brightfield microscopy. Fibrosis was then quantified by analysis of magnified (×20) images from each section using ImageJ version 1.52 software (https://imagej.nih.gov/ij) by counting the number of blue-stained pixels. Perivascular areas were avoided. Fibrous tissue content was expressed as a percentage of field area and averaged across 7 pictures in each LA site. Analyses were performed on coded specimens blinded to pacing group assignment.

Five different biological replicates (tissue samples) were generated from the specific regions of the heart. Tissue chunks were harvested, snap frozen, and kept at −80 °C until lysis. Tissue pieces (50 mg) were aliquoted in 2-mL reinforced tubes containing 2.8-mm ceramic beads and lysed with Fisherbrand Bead Mill 24 homogenizer using a lysis solution containing radioimmunoprecipitation assay lysis buffer (9806; Cell Signaling Technology), protease inhibitors (P8340; Sigma-Aldrich), and phosphatase inhibitors (PhosSTOP; Sigma-Aldrich). All the lysates were clarified twice with ultracentrifugation, aliquoted, and kept at −20 °C. Rapid Gold bicinchoninic acid assay kit (Thermo Fisher Scientific) was used to quantify protein concentration. Proteins were denatured at 70 °C for 10 minutes in SDS, and 10 μg of samples were loaded on either 4% to 12% NuPAGE Bis-Tris gels or 3% to 8% NuPAGE Tris-Acetate gels (Thermo Fisher Scientific). After gel electrophoresis, proteins were transferred to 0.45-µM polyvinylidene difluoride membranes using a wet-transfer system (XCell II Blot Module). Membranes were subsequently blocked with 5% milk for 1 hour at room temperature and incubated overnight at 4 °C with primary antibodies. The following primary antibodies were used: extracellular proteins: collagen 1 (ab138492; Abcam), collagen 8 (102-11285; RayBiotech), fibronectin (ab6328; Abcam), periostin (ab92460; Abcam), LOX (lysyl oxidase; ab174316; Abcam), transforming growth factor-β (TGF-β; 3711; Cell Signaling Technology), TGF-β-r2 (ab186838; Abcam), p-Smad2/3, and Smad 2/3 (8828 and 8685; Cell Signaling Technology); and internal control: GAPDH (5174; Cell Signaling Technology). Equal protein loading and transfer onto polyvinylidene difluoride membranes was verified with Ponceau S staining solution (59803; Cell Signaling Technology). After the overnight primary antibody incubation, polyvinylidene difluoride membranes were washed several times with TBST and incubated with horseradish peroxidase–linked secondary antibody (NA9340 or NA9310; GE Healthcare) for 1 hour at room temperature. SuperSignal West Femto Maximum-Sensitivity Substrate, an enhanced chemiluminescence reagent from Thermo Fisher Scientific (34095), was used to visualize the bands. Semiquantification (relative abundance) of proteins was conducted by comparing the relative intensity of bands against their respective GAPDH abundance using ImageJ software. Samples were run, transferred to polyvinylidene difluoride membrane, exposed, and developed at the same time to minimize variability in loading and chemiluminescence substrate exposure. Equal protein loading was also validated in the aliquots using Coomassie and UV absorbance in gels. Normality testing was performed using the Shapiro-Wilk test. Statistical significance was calculated using the nonparametric test, Kruskal-Wallis 1-way ANOVA, followed by the Benjamini, Krieger, and Yekutieli procedure. Statistical analysis for molecular data was performed in R statistical software (version 4.3.1).

Continuous variables were summarized using mean and SD (if normally distributed) or median and interquartile range (IQR; if skewed); categorical variables were summarized as proportions. A comparison of means between continuous variables was performed using either an unpaired t test or 1-way ANOVA for normally distributed data or the Kruskal-Wallis test for skewed data. Normality was assessed by using the Shapiro-Wilk test. For ANOVA, a Tukey post hoc test was conducted for multiple comparisons. Data analysis was performed using SPSS for Windows (version 23; IBM). Two-tailed P values <0.05 were considered statistically significant.

Baseline electrical atrial dyssynchrony was greater during lateral LA pacing than atrial septal pacing and sinus rhythm (P-wave duration: sinus rhythm, 58.8±7.6; septal pacing, 74.2±8.2; lateral pacing, 93.0±9.0 ms; P<0.001; Figure 3A). Mechanical atrial dyssynchrony assessed by a 2-dimensional speckle tracking strain analysis was greater during lateral LA pacing than septal pacing and sinus rhythm (sinus rhythm, 24.4±11.1; septal pacing, 35.7±13.0; lateral pacing, 57.5±11.6 ms; P<0.001), confirming greater electrical and mechanical dyssynchrony during lateral LA pacing (Figure 3B).

The change in the LA area after 16 weeks of paced atrial bigeminy is summarized in Figure 4A. The Lat-PAC group had the greatest increase in the LA area, followed by the Sep-PAC and Reg-130 groups, and then the nonpaced controls (terminal – baseline; controls, 0.9±0.4 cm²; Reg-130, 3.0±0.8 cm²; Sep-PAC, 4.1±2.0 cm²; Lat-PAC, 5.9±1.2 cm²; P<0.001; Figure 4A). Changes in the LA area over time among the 3 groups are shown in Figure S7. Gradual LA enlargement was observed upon establishing PACs in the Sep-PAC and Lat-PAC groups. Decrease in the peak reservoir strain was greatest in the Lat-PAC group, followed by the Sep-PAC, Reg-130, and control groups, in that order (terminal – baseline; controls, −0.7±4.2%; Reg-130, −8.6±3.0%; Sep-PAC, −12.7±4.1%; Lat-PAC, −17.3±3.2%; P<0.001; Figure 4B). Echocardiographic measures at the terminal study in each group are summarized in Table S1.

Overall, the change in average atrial ERP after 16 weeks of atrial bigeminy at the 3 sites (terminal − baseline) did not differ in the Lat-PAC and Sep-PAC groups. The atrial ERP only decreased significantly in the Reg-130 group (terminal − baseline: controls, −3.2±23.7; Reg-130, −24.4±13.7; Sep-PAC, 8.3±36.0; Lat-PAC, −4.0±23.6 ms; P=0.05; Figure 5A). On the posterior LA wall, the Lat-PAC swine had significantly lower conduction velocities at 16 weeks versus the Reg-130 and control groups (controls, 1.5±0.2; Reg-130, 1.3±0.1; Sep-PAC, 1.3±0.2; Lat-PAC, 1.1±0.2 m/s; P<0.001; Figure 5B). The distribution of unipolar voltage values in the LA and RA among all points is shown in Figure S8. The acquired points for use of the analysis did not differ among the groups (LA: controls, 2016±542; Reg-130, 2322±520; Sep-PAC, 1786±629; Lat-PAC, 2109±648 [P=0.47]; RA: controls, 2441±673; Reg-130, 2083±283; Sep-PAC, 2399±834; Lat-PAC, 2672±901 [P=0.70]). The Lat-PAC group had a higher proportion of low-voltage zones in both the RA and LA, as defined by a percent area <1.0 mV, than the other groups, with a leftward shift in the voltage distribution (LA: controls, 1.8±1.7%; Reg-130, 4.6±3.2%; Sep-PAC, 7.4±11.2%; Lat-PAC, 17.2±11.8% [P=0.02]; RA: controls, 2.7±0.9%; Reg-130, 7.5±1.9%; Sep-PAC, 9.2±4.2%; Lat-PAC, 15.1±8.6% [P=0.01]).

In the terminal study, both the Lat-PAC and Sep-PAC groups exhibited a significantly higher mean LA pressure than the controls (Figure S9A) despite no significant change in mean aortic pressure (Figure S9B).

The control pigs exhibited a higher mean ventricular heart rate than the PAC pigs, but no difference was found between the Sep-PAC and Lat-PAC groups (controls, 88±11; Sep-PAC, 65±8; Lat-PAC, 66±9 bpm; P<0.001). PACs were nonconducted to the ventricle in all pigs, and the Lat-PAC and Sep-PAC groups had no change in LVEF. In contrast, atrial constant pacing at 130/min for 16 weeks led to a significant small LVEF decline (terminal − baseline: controls, 0.7±3.3; Reg-130, −6.5±4.9; Sep-PAC, 0.5±4.0; Lat-PAC, −2.5±6.2%; P=0.006; Figure S10).

After 16 weeks of PACs, the increase in the average duration of induced AF between the terminal and baseline study was greatest in the Lat-PAC group (controls, −1 [IQR, −16, 7]; Reg-130, 22 [11, 64]; Sep-PAC, 16 [IQR 6, 26]; Lat-PAC, 131 [IQR 30, 192] seconds; P<0.001; Figure 6A). The increase in AF inducibility between the terminal and baseline study was also greatest in the Lat-PAC group, followed by the Sep-PAC and Reg-130 groups, and then controls (controls, −2.2±16.2%; Reg-130, 24.4±18.8%; Sep-PAC, 32.6±23.2%; Lat-PAC, 49.3±13.0%; P<0.001; Figure 6B). All induced AF events were composed of irregular atrial activity and no regular atrial tachyarrhythmias were observed.

Figure S11 summarizes the LA and aortic pressure during nonconducted versus conducted PACs. The magnitude of the additional a-wave generated by the PAC atrial contraction was, paradoxically, much smaller during nonconducted (PAC-a in Figure S11B) than conducted PACs (PAC-a in Figure S11C), presumably because of reduced time for atrial filling. Figure S11D summarizes the LA pressure values of sinus rhythm–a versus nonconducted PAC-a versus conducted PAC-a. Whereas the LA pressure during conducted PACs (at a longer coupling interval) is higher than during nonconducted PACs, the difference did not reach statistical significance because of the small sample size (P=0.10).

LA and RA histological fibrosis in each group is summarized in Figure 7. The Lat-PAC group had the greatest degree of fibrosis, followed by Sep-PAC and Reg-130, and then control pigs, on LA anterior wall (controls, 3.9±1.7%; Reg-130, 5.6±1.2%; Sep-PAC, 5.8±1.6%; Lat-PAC, 7.7±1.6%; P<0.001; Figure 7E) and LA posterior wall (controls, 4.0±1.5%; Reg-130, 6.2±1.7%; Sep-PAC, 6.6±1.4%; Lat-PAC, 8.6±1.0%; P<0.001; Figure 7F).

Molecular analysis of fibrosis-related proteins demonstrated that the chronic PAC model was characterized by upregulation of several extracellular matrix proteins as well as TGF-β1 signaling (Figure 8; Figure S12). Collagen 1 levels were significantly higher in the Lat-PAC group than the Sep-PAC or control groups. Collagen 8 was also more abundant in the Lat-PAC than the control group (P=0.05), whereas no differences were observed in fibronectin or periostin proteins. LOX, an essential protein involved in crosslinking and stabilization of collagen,⁸ was more abundant in the Lat-PAC group than the control group (P=0.05). TGF-β signaling–related proteins were also higher in the Lat-PAC group, with increases in latent and active monomer TGF-β protein, as well as the intracellular second messenger of TGF-β receptor activation, phosphorylated SMAD2/SMAD3. No statistically significant differences were observed in the TGF-β receptor type 2 levels. These results suggested that TGF-β1 mediated activation of SMAD signaling, with a resultant increase in collagen and LOX proteins in the Lat-PAC group.

This study supports the notion that frequent PACs, particularly dyssynchronous ones arising from the lateral LA, are not simply an epiphenomenon, but may directly cause atrial myopathy and contribute to AF pathogenesis. This supports the notion that frequent PACs themselves, in the absence of AF, may contribute to atrial myopathy and predispose to stroke.^9–12 Whereas the appropriate management strategy remains unproven, further studies will determine whether early PAC suppression or anticoagulation may prevent the development of adverse atrial remodeling and AF as well as reduce the risk of stroke. These data also suggest that PAC frequency is not sufficient as a predictor, but that the PAC location also may be critical to determining future AF risk.

The use of a swine model has inherent limitations because of differences in physiology compared with humans. However, this model enables a prospective assessment of the direct effects of high-burden PACs over a 16-week period under controlled conditions, which would be impossible in humans. Whereas a 50% PAC burden is uncommonly seen in clinical practice, this model allowed us to test whether PACs would lead to atrial remodeling over a realistic 16-week time period. The control animals were not instrumented (to contain costs, preserve pacemakers or leads, and minimize the number of invasive procedures). However, it is possible the presence of an atrial lead alone could affect atrial remodeling. We observed significant differences between the 2 PAC groups, suggesting that other factors (other than mitral valve closure) affect the relative effects of PACs on remodeling. Further studies are required to characterize the PAC burden and duration required in humans for the development of atrial myopathy.

In a swine model, high-burden PACs for 16 weeks resulted in an atrial myopathy and remodeling characterized by slow conduction and fibrosis development with a heightened propensity for AF. PACs led to a fibrotic atrial substrate without any change in the atrial ERP, suggesting a process distinct from tachy-pacing–induced atrial remodeling. Lateral LA PACs appear to be associated with a greater degree of atrial myopathy and remodeling and longer duration of induced AF than septal PACs. These data provide evidence that frequent PACs are not an epiphenomenon of AF and can cause an atrial substrate supporting AF. Whether early suppression of specific dyssynchronous PACs can prevent future AF or strokes requires further study.

Figures S1–S12

Table S1

Graphical Abstract