Pacemaker implantation in horses has previously been described (12). In this study dual-chamber pacemakers¹ were implanted in nine horses for AF induction (AF group) and electrophysiological studies (AF and control group). The pacemakers were implanted in sedated (0.01 mg/kg detomidine, 0.01 mg/kg butorphanol and constant rate infusion of 1.0 mg/ml xylazine) standing horses. The pectoral region was surgically prepared and locally anesthetised before a 6–8 cm incision was made. Through the cephalic vein two leads² were advanced to the heart and both were fixated in the right atrium. Pacing ability, threshold, lead impedance and absence of phrenic nerve stimulation were tested before and after fixation. The leads were secured in the cephalic vein and connected to the pacemaker which was placed in a subcutaneous pocket distal to the incision. The incision was sutured in three layers and the horses were treated with antibiotics and non-steroidal anti-inflammatory drugs. The surgery-time was between 60 and 150 min.

The study design is depicted in Fig. 1. AF was induced by high-rate pacing from the pacemaker, pacing from both leads at a rate of 170 min^− 1 (with a delay between the leads of 175 ms and 2.5 V @ 0.4 ms output) which resulted in an atrial pacing rate of 340 min^− 1. AF was initiated by burst-pacing operated manually through the pacemaker, which was continued daily (1–8 h per horse) until sustained AF was achieved. During AF (in the period of AF induction), the pacemaker was in an inhibitory mode to automatically start pacing if the atrial rate fell below 170 min^− 1, thus ensuring a minimum atrial rate of 170 min^− 1 at all times. The horses were constantly monitored with ECG during AF induction, except for a few hours per day, when they were allowed to walk freely in the paddock. The pacemaker was turned off after the horses had been in self-sustained AF for more than 24 h.

Total AF durations were calculated from the surface ECG, and all AF episodes of more than 10 min were included. Episodes shorter than 10 min, burst-pacing and pacemaker activity were not included. If the horses spontaneously cardioverted to SR in the paddock (when not equipped with an ECG) or at night (after dislodging the electrodes), these periods were registered as uncertain AF and were not included in the total AF duration.

To investigate the effect of both short- and long-term AF on atrial remodelling, the procedures described below were performed 3, 9, 27 and 55 days (procedure days) after pacing was initiated. Baseline measurements were obtained on day 0 before pacing was initiated (SR measurements) and on day 1 (AF measurements). If the horses were not in AF on the procedure days, the pacemaker was turned off and the procedures were performed in SR. To study the effect of remodelling on pharmacological cardioversion, intravenous flecainide³ (2 mg/kg, rate 0.2 mg/kg/min) was administered on each of the procedure days. Flecainide was chosen for cardioversion as it previously has been studied in horses, it is used in the clinic to treat paroxysmal AF and has a short half-life (7). The pacemaker was turned off at least 30 min before flecainide infusion began to avoid that spontaneous cardioversion was attributed to the effect of flecainide. A detailed description of the effects of flecainide can be found elsewhere (13). At the end of each procedure day, AF was re-induced following the same induction protocol as described above.

During AF induction and throughout each procedure day, the horses were equipped with a three-lead Holter ECG⁴ with the electrodes placed as a modified base-apex ECG (7). The ECGs from each procedure day were manually analysed offline by a single observer; QRS duration and QT interval were measured and averaged over five consecutive beats and HR was averaged over 10 beats. QT measurements were adjusted beat-to-beat according to species specific HR (QTc), as previously described (14).

Atrial fibrillatory rate (AFR) is a non-invasive measurement of atrial remodelling, and is inversely proportional to atrial fibrillation cycle length (AFCL) (15). AFR was calculated from a 15-min surface ECG recording of AF. ECG analysis including QRST cancellation and AFR calculation was performed using the Cardiolund AFR Tracker software⁵ (16).

Echocardiographic examinations were performed to quantify left atrial and ventricular size and function. All echocardiographic examinations were performed using a portable Vivid I ultrasound system with a 3.1 MHz phased array transducer⁶ at rest and without sedation, as previously described (17). Echocardiographic examinations were performed before pacemaker implantation, at baseline and on all procedures days (before, days 0, 1, 3, 9, 27 and 55) in both SR and AF when possible. Horses in the AF group were examined in AF before administration of flecainide, and an additional examination was performed if they cardioverted to SR after flecainide treatment. Assessment of ventricular size and function was performed from the examinations obtained in AF while assessment of atrial size and function was performed from the SR examinations as the left atrial fractional area change measurements only can be obtained in SR. Horses in the control group were examined before flecainide administration. Colour flow Doppler was used to assess valvular competences. Analysis was performed offline⁷ by a single observer and two-dimensional echocardiography (2D) and anatomical M-mode (AMM) were used to assess the left atrial (LA) and left ventricular (LV) size and function. Values are reported as the average of three non-consecutive cardiac cycles. Ventricular measurements are described in the Additional file 1. Two-dimensional echocardiographic variables for the assessment of LA size and function were measured as follows: left atrial area at mitral valve closure (LAA_min), left atrial area at onset of the P wave (LAA_a), left atrial area at mitral valve opening (LAA_max), left atrial diameter at mitral valve opening (LAD), passive left atrial fractional area change (LA-Fac_passive = (LAA_max – LAA_a) / LAA_max), active left atrial fractional area change (LA-Fac_active = (LAA_a – LAA_min) / LAA_a) and total left atrial fractional area change (LA-Fac_total = LAA_max – LAA_min) / LAA_max) (18). Valvular insufficiency was classified as: none, clinically insignificant, mild, moderate or severe (19).

The horses were euthanized between 58 and 63 days after AF initiation by bolt-stunning, and cardiac tissue was harvested for microscopy and quantitative polymerase chain reaction (qPCR, described in Additional file 1: Table S1) from the left atrial appendage (LAA), right atrial appendage (RAA) and right ventricle (RV). Biopsies for microscopy were sliced (4 μm) and stained with Picro-Sirius Red. Detailed information about the preparation of the tissue can be found in the Additional file 1. Full biopsy sections were scanned, using a ZEISS Axioscan slide scanner equipped with a 20x, 0.8 NA objective.⁸ Scanned images were analysed with the software ZEN 2.3 Blue edition⁹ in combination with the extension software ZEN Intellesis,¹⁰ which was used for image segmentation to distinguish between background, muscle and collagen. We applied two approaches for collagen assessment. First, the full sections including large vessels, epi- and endocardial tissue were analysed. Second, quantification of interstitial fibrosis, where three areas from each section containing either high amounts (interstitial_max) or low amounts (interstitial_min) of collagen between the cardiomyocytes were chosen, by eye and blindfolded to the observer, and analysed as the average of interstitial_min and interstitial_max.

Investigators were blinded to the group allocation and procedure day throughout all analyses. Data analyses were performed using Microsoft Excel¹¹ and GraphPad Prism.¹² Changes in AFR and other ECG parameters between baseline and day 55, were assessed using student’s t-tests where paired analysis were performed when appropriate. Difference in fibrotic area and ion channels expression (measured with qPCR) between AF and control were also assessed using student’s t-test. The qPCR data were subsequently Bonferroni corrected. Echocardiographic data were compared to baseline measurement with a one-way ANOVA followed by Dunnett’s post hoc test for multiple comparisons. Data are presented as mean ± SD and p ≤ 0.05 was considered significant.

All horses were continuously paced into self-sustained AF until they no longer spontaneously cardioverted to SR. The pacing time required to achieve self-sustained AF varied among horses (18 ± 9.9 days, range 8–36 days). AF stability increased over time for all horses; at the beginning of AF induction, burst-pacing only resulted in short AF episodes, but as pacing continued, longer AF episodes occurred and eventually only a small number of burst-pacing sessions were needed for AF to become sustained (Fig. 2, panel C). The AF became less organized over time as it developed into more stable AF (Fig. 2, panel E and F). In contrast, no ECG changes were observed in the control group between day 0 and day 55 (Fig. 2, panel A + B). Total AF duration ranged from 15 to 46 days (37 ± 11 days), and the AF burden for each horse is illustrated in Fig. 3. The total duration of unknown AF was on average 2.4 days per horse and was not included.

Offline assessment of echocardiographic image and data quality led to the exclusion of ventricular measurement obtained in AMM for two horses from the AF group (on day 1 and 55), and ventricular measurement obtained in 2DE for 1 horses in the AF group at day 55 and for atrial images in one horse during AF at day 1.

From the echocardiographic examinations performed in SR after cardioversion with flecainide (AF group) we found that the left atrial area (LAA_min) significantly increased after day 9 (p < 0.01, Fig. 5) compared to baseline while no significant changes were observed for LAA_max or LAD. The LA-Fac_active significantly decreased after day 3 (p < 0.01, Fig. 5) while no significant changes were observed for LA-Fac_passive but a tendency for LA-Fac_total was found at several time points and showed significant difference at day 27 (p < 0.01). No differences were observed between examinations performed before pacemaker implantation and at baseline (after pacemaker implantation, before induction of AF). Before cardioversion with flecainide the horses in the AF group had an echocardiographic examination while in AF. Data from these examinations revealed a significant enlargement of the LAA_min and LAA_max between baseline (AF baseline = day 1) and day 55 (p < 0.01). No changes in LAD or LA-Fac_total were observed. Measurement of LAA_a, LA-Fac_active and LA-Fac_passive were not possible as no P wave was visible during AF. We observed no changes in either atrial size or function in the control group (Fig. 5 and Additional file 1: Table S2). Measurements of atrial size and function before the pacemaker was implanted, at baseline, during the AF episodes and after cardioversion are illustrated in Additional file 1: Table S2.

Colour flow Doppler revealed a moderate tricuspid valve insufficiency in one horse at inclusion that did not progress during the study. Several horses had mild and clinically insignificant valve insufficiencies. None of the horses developed valvular insufficiencies beyond the “mild” category throughout the study.

Results on cardioversion rates with flecainide have previously been published (13). Briefly, one horse was in stable AF on procedure day 3, and cardioverted 3 min after flecainide infusion was initiated. On day 9, 5/5 horses cardioverted (6–12 min after flecainide infusion began), on day 27, 5/6 horses cardioverted (4–62 min after infusion began) and on day 55, 2/6 horses cardioverted (96 and 185 min after infusion began, Fig. 3). There was a correlation between the time from infusion start until cardioversion to SR and the cumulative duration of AF (p < 0.001, R² = 0.80) (13).

With the Picro-Sirius Red staining of connective tissue we found a significantly increased collagen deposition in the left atria in the AF group (compared to the control group) for both the full LAA sections (p < 0.05, Fig. 6A, C) and the analysis of interstitial fibrosis (p < 0.05, Fig. 6D, F). There were no significant differences found for the full RAA section (p = 0.14, Fig. 6B, C), although a tendency was seen in the analyses of interstitial fibrosis (p = 0.06, Fig. 6E, F). Quantification of collagen deposition of the RV sections (Fig. 6C) showed no differences between the two groups (p = 0.97).

The pacemaker ensured a constant rate between 170 and 340 min-1 in the atria, however this did not lead to sufficient capture nor the development of self-sustained AF. Burst-pacing was therefore introduced and combined with pacing from the pacemaker it proved feasible for inducing self-sustained AF in horses. Other animal studies (8, 20) have used neurostimulators to automatically burst-pace the atria which is desirable as it is less time-consuming for the operators and ensures a uniform and a most likely faster AF induction. The pacing-time required to achieve self-sustained AF varied among horses, suggesting a variation in AF vulnerability. Atrial pacing led to increased AF vulnerability and increased AF duration in all horses; this is the main feature of atrial remodelling where “AF begets AF” (2). One horse (#6) required a particularly long pacing-time, which could be attributed to its small size (408 kg), as AF is difficult to induce in small horses (11). Despite the prolonged period before persistent AF was induced in this horse, intensive pacing did seem to remodel the heart, as this horse could not be pharmacologically cardioverted by flecainide on day 55.

A number of studies on electrical properties in horses with induced AF have been published (5, 7‐9) and in one of them, spontaneous cardioversion to SR allowed refractory period measurements without drug interferences (8). The equine atrial refractoriness is rate-dependent and decreases with AF (7‐9) in a similar way to humans (21), dogs (22), pigs (23) and goats (2). Shortening of the refractory period allows more fibrillatory waves to co-exist, which further promotes AF. In this study, we measured the rate of the fibrillatory waves while the horses were in AF. Several studies have validated AFR as a measurement of electrical remodelling (24‐26), and this method allowed us to study the progression of electrical remodelling non-invasively and without drug interference. We found a significant increase in AFR between day 1 and day 55, which is representative of electrical remodelling. We have previously studied AFR in horses with both short-term induced AF and spontaneous persistent AF (26). Similar to the findings in this study, the AFR in horses with induced short-term AF was lower (269 ± 36 fpm) compared to horses with spontaneous persistent AF (364 ± 26 fpm) (26). These values correspond well with AFR values in patients with paroxysmal and persistent AF, respectively (25, 27).

Electrical remodelling also describe changes in the genes encoding ion channels during AF. In this study, no difference in ion channel expression was observed between the AF and control group. Other studies have found a down-regulation in the L-type Ca²⁺ channel, the α subunit of cardiac Na⁺ channels and Kv4.3 channel (28, 29), however, as these analyses are underpowered similar conclusions cannot be made from the present study.

Atrial fibrillation is also known to cause atrial enlargement and contractile dysfunction that can persist several days after cardioversion (30). In patients, atrial contractile dysfunction, also known as “atrial stunning”, is considered an important risk factor for thromboembolisms (31). In this study, we observed a significant reduction in atrial function (LA-Fac_active) three days after AF induction was initiated. Atrial remodelling has also been documented in dogs after 6 weeks of continuous pacing (22) and a study on instrumented goats tachypaced into AF, showed that atrial function was reduced already after 5 min of AF and that the reduction increased with AF duration (32). Similar results have been described for horses with spontaneous (18) and induced AF (8, 33).

Atrial fibrillation is associated with impaired ventricular function and heart failure is a common comorbidity in AF patients. Although not fully elucidated, this is likely to be the result of a tachycardia-induced cardiomyopathy caused by a persistently increased HR (34). In human patients, the increased HR is often managed with rate-control drugs such as beta-blockers (1). Other animal models also have increased HR, and these are often treated medically with digoxin, or surgically with an AV nodal ablation (20). Horses have a high vagal tone, and if they develop persistent AF they usually have a normal to slightly increased HR (35). In accordance with this, we found no significant difference in HR between the AF and control group. In addition, no changes in ventricular size or function from the echocardiographic examinations were found and no differences in ventricular fibrosis between control and AF group were observed. These results are in accordance with observations from the equine clinic, where horses with spontaneous AF rarely develop heart failure without underlying cardiac disease (18, 36). Tachycardia induced heart failure albeit medical rate control has been shown in pigs and dogs, while goats did not show signs of heart failure. Increased myocardial fibrosis was also evident in the ventricles of the pigs and dogs but not in the goats (20).

Cardioversion to sinus rhythm.

The success rate of medical cardioversion to SR decreases as AF progresses (1). A human clinical study previously reported that the cardioversion success rate decreased from 86 to 22% if the patients treated with flecainide had been in AF for longer than 10 days (37). We found similar results in this study, as the cardioversion success rate fell from 100% on days 3 and 9 to 33% on day 55. This suggests that the horse could be an interesting model for testing novel antiarrhythmic drugs, as it seems that the response to medical treatment is comparable.

In addition to electrical and functional remodelling, AF is also known to cause structural remodelling, which includes hypertrophy of the cardiomyocytes, myolysis, alterations in connexin expression and interstitial fibrosis, all of which increase conduction heterogeneity and support the maintenance of AF (30, 38). To our knowledge, this is the first study to compare structural remodelling in horses with and without AF and we found significantly increased amounts of collagen in the AF group. The difference between groups was more pronounced in the left atrium compared to the right atrium, where only a tendency was present. Several studies in both human patients (39) and animal models (20) show an increase in collagen as a consequence of AF. The literature seems to agree on increased fibrosis in the left atrium in patients with AF, yet there are mixed results regarding right atrial fibrosis (40). Some studies report an equal amount of fibrosis in both the left and right atrium, while other studies report similar findings as this study, where less fibrosis was observed in the right atrium compared to the left (41‐43). Atrial fibrosis increases with age which have been suggested to be a contributing factor to the increased AF burden in the aging population (44). In this study two out of the three control horses were above the age of 15 years which might have caused an increase in fibrosis in the control group.

This study is limited by the relatively low number of animals included. The small sample size is strengthened by the longitudinal design of the study which facilitates repeated measurements, however, both the AF and SR measurements holds several missing values. The missing AF values are attributed to the prolonged AF induction, which could have been avoided by continuous burst-pacing. Sinus rhythm measurements were limited by the attenuated ability of flecainide to cardiovert longstanding AF. Future studies should consider either a more potent drug than flecainide or electrical cardioversion in order to obtain SR measurements. AFR only provides a summary of the atrial activity and local heterogeneity cannot be captured with this method. Technical errors prevented us from assessing local changes in AFCL and atrial refractoriness. Future studies should explore endo- and epicardial atrial recordings and mapping of fibrillation patterns.