Altered calcium handling produces reentry-promoting action potential alternans in atrial fibrillation–remodeled hearts

Atrial fibrillation (AF) alters atrial cardiomyocyte (ACM) Ca2+ handling, promoting ectopic beat formation. We examined the effects of AF-associated remodeling on Ca2+-related action potential dynamics and consequences for AF susceptibility. AF was maintained electrically in dogs by right atrial (RA) tachypacing. ACMs isolated from AF dogs showed increased Ca2+ release refractoriness, spontaneous Ca2+ spark frequency, and cycle length (CL) threshold for Ca2+ and action potential duration (APD) alternans versus controls. AF increased the in situ CL threshold for Ca2+/APD alternans and spatial dispersion in Ca2+ release recovery kinetics, leading to spatially discordant alternans associated with reentrant rotor formation and susceptibility to AF induction/maintenance. The clinically available agent dantrolene reduced Ca2+ leak and CL threshold for Ca2+/APD alternans in ACMs and AF dog right atrium, while suppressing AF susceptibility; caffeine increased Ca2+ leak and CL threshold for Ca2+/APD alternans in control dog ACMs and RA tissues. In vivo, the atrial repolarization alternans CL threshold was increased in AF versus control, as was AF vulnerability. Intravenous dantrolene restored repolarization alternans threshold and reduced AF vulnerability. Immunoblots showed reduced expression of total and phosphorylated ryanodine receptors and calsequestrin in AF and unchanged phospholamban/SERCA expression. Thus, along with promoting spontaneous ectopy, AF-induced Ca2+ handling abnormalities favor AF by enhancing vulnerability to repolarization alternans, promoting initiation and maintenance of reentrant activity; dantrolene provides a lead [...] Research Article Cardiology


Introduction
Atrial fibrillation (AF), the most common sustained cardiac arrhythmia in clinical practice, is associated with significant morbidity and mortality (1). AF prevalence is age dependent and will continue to rise as the population ages (1). Current therapeutic approaches for AF face limitations related to suboptimal efficacy and adverse effect potential; new insights into underlying mechanisms might help identify novel therapeutic strategies (2).
AF-associated remodeling has long been known to alter atrial Ca 2+ handling (3). The rapid atrial rates associated with AF produce early cellular Ca 2+ loading (4), followed by changes in cell Ca 2+ handling that reduce the systolic Ca 2+ transient while causing marked hypocontractility (3). One curious finding in such remodeled cardiomyocytes is that although the steady-state Ca 2+ transient during sustained activity is very small, the initial transient after a period of inactivity is nearly normal (3). One possible explanation would be an alteration in refractoriness of the Ca 2+ transient. Prolonged Ca 2+ transient refractoriness (CTR) decreases Ca 2+ transients with repeated activation, especially at rapid rates, and increases susceptibility to alternans behavior (5).
Abnormalities in cardiac ryanodine receptor (RyR2) function are well known to occur in AF and are associated with aberrant Ca 2+ releases that cause delayed afterdepolarizations, triggered activity, and spontaneous ectopic firing (6)(7)(8)(9). In a mouse knockin model harboring a "leaky" RyR2 (RyR2 R2474S), Xie et al. found increased susceptibility to atrial cardiomyocyte action potential (AP) alternans and atrial arrhythmias (10). A computational study suggested that disrupted Ca 2+ release induced by impaired RyR2 inac-Atrial fibrillation (AF) alters atrial cardiomyocyte (ACM) Ca 2+ handling, promoting ectopic beat formation. We examined the effects of AF-associated remodeling on Ca 2+ -related action potential dynamics and consequences for AF susceptibility. AF was maintained electrically in dogs by right atrial (RA) tachypacing. ACMs isolated from AF dogs showed increased Ca 2+ release refractoriness, spontaneous Ca 2+ spark frequency, and cycle length (CL) threshold for Ca 2+ and action potential duration (APD) alternans versus controls. AF increased the in situ CL threshold for Ca 2+ /APD alternans and spatial dispersion in Ca 2+ release recovery kinetics, leading to spatially discordant alternans associated with reentrant rotor formation and susceptibility to AF induction/ maintenance. The clinically available agent dantrolene reduced Ca 2+ leak and CL threshold for Ca 2+ / APD alternans in ACMs and AF dog right atrium, while suppressing AF susceptibility; caffeine increased Ca 2+ leak and CL threshold for Ca 2+ /APD alternans in control dog ACMs and RA tissues. In vivo, the atrial repolarization alternans CL threshold was increased in AF versus control, as was AF vulnerability. Intravenous dantrolene restored repolarization alternans threshold and reduced AF vulnerability. Immunoblots showed reduced expression of total and phosphorylated ryanodine receptors and calsequestrin in AF and unchanged phospholamban/SERCA expression. Thus, along with promoting spontaneous ectopy, AF-induced Ca 2+ handling abnormalities favor AF by enhancing vulnerability to repolarization alternans, promoting initiation and maintenance of reentrant activity; dantrolene provides a lead molecule to target this mechanism.
tivation might promote atrial alternans and arrhythmia susceptibility in patients with AF (11); however, there has been no direct demonstration.
We speculated that abnormal Ca 2+ handling associated with AF might increase CTR and that Ca 2+ handling disturbances might not only predispose to abnormal Ca 2+ releases, causing spontaneous atrial ectopic activity, but also promote alternans behavior and susceptibility to reentry. We assessed this possibility in a dog model of electrically maintained AF, with a combination of cellular Ca 2+ imaging, simultaneous tissue optical Ca 2+ /transmembrane potential (Vm) optical mapping, and in vivo electrophysiology. Our results suggest that AF-associated Ca 2+ handling disturbances may be central not only to the trigger for reentry in terms of ectopic beat formation but also to the substrate for reentry through enhanced susceptibility to AP alternans. 2+ release restitution. Figure 1A shows representative recordings of cytosolic Ca 2+ transients (CaTs) at various S1-S2 coupling intervals from each group. AF prolonged CTR, which averaged 197 ± 13 ms in AF atrial cardiomyocytes (ACMs) versus 144 ± 7 ms in CTLs (P < 0.01; Figure 1B). Figure 1C shows typical CaT restitution curves, fitted by monoexponential relationships. AF increased the steepness of CaT restitution, decreasing the restitution time constant (τ) from 591 ± 18 ms in CTL ACMs to 315 ± 13 ms (P < 0.001; Figure 1D).

AF-related changes in CTR and Ca
Changes in the threshold for cellular CaT alternans. It has long been recognized that changes in [Ca 2+ ] i are central to the development of repolarization alternans (12). Figure 2A shows cellular CaT recordings at progressively higher frequencies. In each case, beat-to-beat CaT alternans eventually appeared. The threshold frequency for alternans in each ACM in each group is shown in Figure 2B. The CaT alternans frequency threshold was significantly reduced, by 40%, in AF ACMs.
Changes in CaT and sarcoplasmic reticulum Ca 2+ content. The results in Figure 1 and Figure 2 show that Ca 2+ release dynamics are altered in AF ACMs. To relate these changes in dynamics to basic Ca 2+ handling properties, we determined the effects of AF on CaT properties and sarcoplasmic reticulum (SR) Ca 2+ content under fixed-rate pacing. Supplemental Figure 1A (supplemental material available online with this article; https://doi.org/10.1172/jci.insight.133754DS1) shows representative CaT recordings. Diastolic [Ca 2+ ] i was increased, while CaT amplitude was significantly decreased, in AF ACMs (Supplemental Figure 1B). Changes in SR Ca 2+ content were assessed by simultaneously measuring caffeine-induced CaTs and sodium-calcium exchange (NCX) currents. Supplemental Figure 2A shows representative recordings of caffeine-induced CaTs and NCX currents from CTL and AF ACMs. AF significantly increased caffeine-induced NCX current amplitude and SR Ca 2+ content (Supplemental Figure 2, B and C) without affecting [Ca 2+ ] i decay kinetics (Supplemental Figure 2D).
In situ alterations in atrial Ca 2+ handling dynamics and heterogeneity. To relate AF-induced single-ACM changes to cellular alterations in intact tissue, we performed optical [Ca 2+ ] i mapping in coronary-perfused canine atrial preparations. Figure 4A shows typical maps of S2/S1 CaT ratio at 4 S1-S2 coupling intervals. The S2/S1 CaT ratio was significantly decreased at all coupling intervals below 220 ms in AF atria compared with CTLs (all P < 0.001; Figure 4B), corresponding to the single-ACM results ( Figure 1). However, in situ mapping revealed an additional effect of AF: a marked increase in regional variability with premature activation ( Figure 4A). Spatial CaT inhomogeneity was quantified as the coefficient of variation (COV) of S2/S1 CaT ratio (COV ratio). The COV ratio was strikingly increased in AF atria, with significant changes at all coupling intervals below 240 ms compared with CTLs (P < 0.001; Figure 4C).
We further quantified the spatial variability in Ca 2+ release dynamics by comparing the restitution τs at multiple locations. Figure 5A shows representative CaT restitution curves at 4 atrial locations. The τs were smaller overall in AF atria than in CTLs ( Figure 5B), corresponding to AF effects on single-ACM restitution ( Figure 1C). In addition, AF substantially increased the spatial variability in restitution kinetics ( Figure 5C).
Alterations in atrial CaT and action potential duration alternans in atrial tissues. We evaluated the effects of AF on ACM alternans in situ with simultaneous optical mapping of CaT and action potential duration (APD) alternans. Figure 6, A and B, illustrate, respectively, CaT and APD alternans thresholds in a CTL atrium. Recordings from single pixels are shown, along with maps at the left of the amplitude of beat-tobeat changes at each pixel in the recording field. Consistent with the single-ACM results, the median BCL thresholds for CaT and APD alternans were significantly increased in AF versus CTL atria (median CaT alternans threshold in CTL 200 ms vs. 255 ms for AF; APD threshold: 180 ms in CTL vs. 225 ms for AF, P < 0.001 for each; Figure 6, C and D). Of note, in all cases and conditions, Ca 2+ alternans occurred at a longer cycle length threshold (by ~20 ms) than APD alternans (Supplemental Figure 4).
Effects of pharmacological modulation of cellular Ca 2+ release behavior. To explore the relationship between RyR2 leakiness, increased CTR, and CaT alternans, we used dantrolene (10 μM) to suppress RyR2 leak (13). Dantrolene substantially reduced SCaS frequency, Ca 2+ spark-mediated leak, and SCaT events in AF cardiomyocytes (Figure 3 and Supplemental Figure 3). The addition of dantrolene decreased CTR in ACMs from AF dogs toward CTL values ( Figure 1, A and B: from 197 ± 13 ms in AF ACMs to 150 ± 6 ms in dantrolene-treated [AF + DTL] ACMs, P < 0.01). Dantrolene also moved the time course of CaT restitution in AF cells toward CTL values (Figure 1, C and D). Similarly, dantrolene increased the CaT alternans threshold in ACMs from AF dogs ( Figure 2B), from a median of 3 Hz in AF to 4 Hz in AF + DTL (P < 0.001). Correspondingly, dantrolene decreased the BCL threshold for CaT and APD alternans in perfused atria from AF dogs ( Figure 6, C and D). Dantrolene also returned the kinetic properties of AF-remodeled atrial Ca 2+ restitution toward control values, increasing the S2/S1 CaT ratio ( Figure 4B) and the τ of calcium release restitution ( Figure 5B), while decreasing the COV ratio ( Figure 4C) and COV τ ( Figure 5C) in AF atria. Thus, reducing RyR2 leak with dantrolene reversed many of the abnormalities in CaT and APD kinetic properties associated with AF. To determine whether we could reproduce AF-associated abnormalities in isolated CTL dog cardiomyocytes by enhancing RyR2 leak, we applied low-dose caffeine (100 μM), known to increase RyR2 leak and SCaS frequency (14). Consistent with expectations, caffeine significantly increased SCaS frequency, Ca 2+ spark-mediated leak, and SCaT events in CTL cardiomyocytes ( Figure 3 and Supplemental Figure 3). The addition of caffeine increased CTR in ACMs from CTL dogs toward AF values ( Figure 1, A and B: from 144 ± 7 ms in CTL cardiomyocytes to 193 ± 10 ms in caffeine-treated cardiomyocytes, P < 0.01). Caffeine also shifted the time course of [Ca 2+ ] i restitution toward AF values (Figure 1, C and D) and reproduced AF effects on cellular CaT alternans threshold ( Figure 2). Correspondingly, caffeine increased the BCL threshold for CaT and APD alternans in perfused atria from CTL dogs ( Figure 6, C and D). Caffeine also mimicked AF effects on restitution, decreasing the S2/S1 CaT ratio ( Figure 4B) and τ of Ca 2+ release restitution curves ( Figure 5B), while increasing the COV ratio ( Figure 4C) and COV τ ( Figure 5C) in CTL atria.
Alternans behavior and AF. The results presented above show that AF-induced remodeling increases the susceptibility to CaT and APD alternans by altering CaT dynamics and increasing their heterogeneity. We then examined the relationship between alternans occurrence/pattern and AF induction. Representative recordings of the transition from spatially concordant alternans (SCA) to spatially discordant alternans (SDA) from an AF atrium are shown in Figure 7A. During dynamic pacing at threshold BCL, CaT and APD alternans were initially spatially concordant across the atrial surface. With further decreases in BCL, SCA transitioned to SDA. Whenever SDA CaT occurred, SDA APD alternans also appeared ( Figure 7, B and C). The inducibility of SDA CaT/APD was significantly increased in AF atria (occurring in 6/8, or 75.0%, of AF atria vs. 0/6, or 0%, of CTL atria, P < 0.01); dantrolene reduced SDA CaT/APD incidence in AF atria from 75% to 1/8 (12.5%); caffeine increased SDA CaT/APD incidence in CTL atria from 0% to 4/6 (66.7%, P < 0.01, Figure 7C). Upon the onset of SDA, the beat-to-beat behavior of discordant APD alternans paralleled that of Ca 2+ alternans, with a somewhat smaller amplitude (Supplemental Figure 5).
The induction of AF was closely linked to that of SDA, appearing after the development of SDA in all cases and in none in which SDA failed to develop. AF was induced in 75% (6/8) of AF-remodeled atria, versus 0% of 6 CTL atria (P < 0.01). AF inducibility of AF atria was suppressed by dantrolene (to 1/8, 12.5%, of AF atria, P < 0.01); in contrast, AF inducibility of CTL atria was promoted by caffeine, occurring in 4/6, 66.7%, of CTL atria (P < 0.05). AF initiation was related to the development of functional block-associated reentry, as illustrated in Supplemental Figure 6. During discordant alternans associated with rapid pacing, the impulse blocked in a long APD region, inducing reentry around the line of block and reentrant rotor formation. Figure 8 illustrates the relationship between zones of SDA and reentrant activity. Figure 8A shows a right atrial (RA) pseudo ECG from 1 optically mapped preparation. The ΔAPD map for complexes a and b (the penultimate paced complexes prior to AF initiation) are shown in the upper-left of Figure 8B. There are substantial and discordant interbeat APD differences between regions in the mapped area. The subsequent panels show phase maps corresponding to the time points indicated on the pseudo ECG. Dynamic pacing induced AF with consistent counterclockwise rotor activity in the field of view, as shown by the PSs at the core tip of the rotor at 8 snapshots of phase maps within a rotational cycle. The location of the core tip PS coincided closely with the location of nodal lines during SDA.
In vivo correlates. We then sought to determine whether the phenomena we studied at the cellular and isolated atrial level also occur in vivo. We recorded activation recovery intervals (ARIs) in anes- thetized CTL or AF dogs, before and after dantrolene (2.5 mg/kg, IV). Figure 9A shows representative recordings of atrial ARI alternans from a CTL dog, as well as in an AF dog, before and after dantrolene administration. Compared with CTL dogs, the median BCL threshold for ARI alternans was significantly increased in AF dogs ( Figure 9B: AF 210 ms vs. CTL 150 ms, P < 0.001). Dantrolene administration significantly decreased the BCL threshold for ARI alternans in AF dogs (to 170 ms, P < 0.01). These results indicate that in vivo vulnerability to atrial APD alternans is increased in AF dogs and that dantrolene attenuated this effect.
Supplemental Figure 7 shows recordings of AF induction by atrial burst pacing in a CTL dog, along with an AF dog before and after dantrolene administration. Compared with CTL dogs, the AF vulnerability and AF duration were markedly increased in AF dogs (AF vulnerability: 61% ± 6% in AF vs. 5% ± 3% in CTL, P < 0.001; AF duration: 379 ± 96 s in AF vs. 14 ± 8 s in CTL, P < 0.01). Dantrolene administration markedly decreased the AF inducibility and AF duration in AF dogs (AF inducibility: 28% ± 7% in AF + DTL vs. 61% ± 6% in the same AF dogs before DTL; AF duration: 61 ± 23 s in AF + DTL vs. 379 ± 96 s in the same AF dogs before DTL; all P < 0.01).
Alterations in Ca 2+ handling proteins. To gain insights into the potential molecular basis of abnormalities of Ca 2+ handling in AF atria, we performed Western blots on a range of relevant proteins in isolated cardiomyocytes (for original images, see Supplemental Figure Figure 8B), Ser2808-phosphorylated RyR2 (Supplemental Figure 8C), and the Ser2814-phosphorylated form (Supplemental Figure 8D). Phosphorylation ratios (Ser2808-and Ser2814-phosphorylated RyR2 to total RyR2) were not significantly affected (Supplemental Figure 8, E and F). Phospholamban (PLB) was not significantly affected by AF, including total PLB (Supplemental Figure 9C), the Ser16-phosphorylated form (Supplemental Figure 9D), and phosphorylation ratios (Ser16-and Thr17-phosphorylated PLB divided by total PLB; Supplemental Figure 9, F and G); nor was sarco/endoplasmic reticulum Ca 2+ -AT- Maps of S2/S1 CaT amplitude ratio (color scale at right) during progressive decreases in S1-S2 coupling interval for each condition (left columns). Original recordings of the basic CaT (the first CaT) and the premature stimulation-induced CaT (the second CaT) from sites 1 and 2 for each condition are shown in the right columns. (B) Average S2/S1 CaT amplitude ratio for different S1-S2 coupling intervals in each group. (C) Average coefficient of variation (COV) of S2/S1 (COV S2/S1) CaT amplitude ratio for different S1-S2 coupling intervals in each group. Horizontal lines show mean ± SD. N = 6 for CTL and CTL + Caff; N = 8 for AF and AF + DTL. 2-way repeated-measures ANOVA followed by Bonferroni-corrected post hoc tests. CTL + Caff, CTL atria after caffeine treatment; AF + DTL, AF atria after dantrolene treatment. insight.jci.org https://doi.org/10.1172/jci.insight.133754 Pase 2a (SERCA2a) (Supplemental Figure 9H). AF significantly increased the expression of Thr17-phosphorylated PLB (P < 0.05; Supplemental Figure 9E). FKBP12.6 expression was nonsignificantly reduced in AF (Supplemental Figure 10A), whereas calsequestrin (CSQ) showed a statistically significant, about 30%, decrease (Supplemental Figure 10B).

Discussion
In this study, we evaluated for the first time to our knowledge the relationships among Ca 2+ mishandling induced by AF, the susceptibility to different forms of atrial alternans behavior in atrial tissues and intact hearts, and AF susceptibility. We relate findings in single cells to those in intact atria with the use of Ca 2+ imaging and V m analysis and to in vivo consequences with studies in intact, anesthetized dogs. Furthermore, the potential molecular changes underlying Ca 2+ mishandling were studied with the use of immunoblots. Our results indicate that AF remodeling enhances spontaneous single-cell Ca 2+ release, prolongs the refractory period of the CaT, increases the slope of the CaT restitution curve, and predisposes to CaT alternans. The AF-induced changes were mimicked in control cells by increasing spontaneous Ca 2+ release with low-dose caffeine and suppressed in AF cells by dantrolene. Correspondingly, in coronary-perfused preparations, AF atria displayed increased Ca 2+ handling heterogeneity and susceptibility to CaT and APD alternans, particularly SDA, which was associated with enhanced AF inducibility accompanied by rotor formation at nodal lines. Dantrolene decreased Ca 2+ handling heterogeneity and suppressed alternans behavior and AF vulnerability in AF atria. Finally, in vivo studies in dogs with AF-induced remodeling confirmed increased susceptibility to rate-dependent atrial repolarization alternans, along with enhanced AF vulnerability and sustainability, which were attenuated by intravenous dantrolene administration. Ca 2+ handling changes in AF and potential arrhythmogenic role. A variety of Ca 2+ handling abnormalities have been associated with AF, based on studies of both ACMs isolated from AF patients (6)(7)(8)(9)15) and experimental animals (16)(17)(18). Spontaneous diastolic Ca 2+ release and RyR2 dysfunction are almost ubiquitous findings (6,9,(16)(17)(18)(19). Several studies show enhanced delayed afterdepolarizations resulting from diastolic Ca 2+ leak, associated with spontaneous ectopic activation generation (8,9,16,17,19), although one report indicates Ca 2+ silencing that protects against spontaneous Ca 2+ release and activity (15).
Here, we confirmed an increase in spontaneous cellular Ca 2+ release events in ACMs from AF dogs; moreover, this finding was associated with increased CTR, altered CaT restitution, and greater susceptibility to CaT alternans. A causative role of spontaneous Ca 2+ leak in altered CaT kinetics was supported by the observation that low-dose caffeine reproduced the alterations AF caused and that dantrolene, which suppresses Ca 2+ leak (13,19), reversed the AF-induced changes. Further support for a role of Ca 2+ mishandling in arrhythmogenesis was obtained by noting the relationship between CaT and APD alternans behavior in AF dog atria, which caused SDA and associated rotor formation along with greatly increased AF susceptibility. Again, these changes were suppressed by dantrolene. Finally, we obtained evidence for similar phenomena in vivo. These results argue strongly that Ca 2+ mishandling due to AF remodeling plays a role in AF promotion not only by favoring ectopic impulse formation but also by increasing the susceptibility to reentry.
Alternans behavior in clinical AF. A wide range of observations point to the importance of repolarization alternans in clinical AF (20). Since the first report of atrial monophasic APD alternans in a patient with atrial tachycardia (21), a number of studies have confirmed the occurrence and potential pathophysiological role of alternans in AF. Kim et al. showed increased APD restitution slope in patients with AF versus controls (22). Narayan et al. demonstrated that APD alternans occurred at much lower rates in patients with AF than controls and disorganized to complex oscillations leading to AF as rate increased further (23). In the latter 2 studies, altered repolarization kinetics occurred in patients with both persistent AF and paroxysmal AF. Patients with persistent AF were cardioverted just before electrophysiological study and were therefore quite analogous to our AF dogs, which had electrically maintained AF until the terminal experiment. Although patients with paroxysmal AF do not have AF-induced remodeling per se, they show spontaneous cellular Ca 2+ release (9), which our findings suggest might explain their susceptibility to APD alternans.
Recent computational modeling studies further support the role of Ca 2+ handling abnormalities in atrial alternans and clinical AF. Chang et al. used a human atrial tissue model incorporating AF-associated electrophysiological remodeling and performed a sensitivity analysis of ion channel/transporter properties that might be involved in alternans behavior (11). They concluded that decreased RyR2 inactivation is the only parameter variation that reproduces AF-associated APD alternans behavior at relatively slow rates and that RyR2 kinetics may play a critical role in governing proarrhythmic APD alternans. In a subsequent study with a 3-dimensional human atrial model, they noted a critical role of Ca 2+ -driven alternans, not only in the initiation of AF, but also in its persistence by causing unstable scroll waves that meander and break up to produce multiple daughter wavelets (24).
Spatially concordant and discordant alternans. Alternans can be spatially concordant or discordant (25,26). SCA occurs when all the cells are alternating in phase, can be induced by rapid pacing in most normal hearts, and is not necessarily arrhythmogenic. By contrast, SDA is more malignant and frequently observed in pathological states. The large refractory gradients over short distances resulting from SDA can produce conduction block and reentrant excitation that results in tachyarrhythmias and fibrillation (25)(26)(27). In our optical mapping study, SCA was induced in all atria, both CTL and AF, but SDA occurred in 0% of CTL atria versus 75% of AF atria. Furthermore, SDA immediately preceded AF onset in all 6/6 AF atria.
To gain insights into the much greater susceptibility of AF atria to SDA CaT, we studied the spatial heterogeneity of Ca 2+ handling. We found a clear increase in Ca 2+ handling heterogeneity in AF atria, evidenced by increased COV S1/S2 ratios and COV τ of the Ca 2+ release restitution curve. The increased Ca 2+ handling heterogeneity was associated with a transformation from SCA CaT to SDA CaT upon abrupt BCL reduction in AF atria. Dantrolene decreased Ca 2+ handling heterogeneity and abrogated SDA Ca 2+ occurrence in AF atria. By contrast, low-dose caffeine increased Ca 2+ handling heterogeneity and increased the inducibility of SDA CaT in CTL atria. These findings implicate spatial heterogeneity in Ca 2+ handling remodeling in the occurrence of SDA CaT. The mechanism for SDA APD is still poorly understood. In this study, we observed that SDA CaT and SDA APD always occurred concurrently in atrial tissue. Dantrolene and low-dose caffeine administration, which affected the occurrence of SDA CaT, abrogated and promoted the occurrence of SDA APD, respectively. These results suggest that SDA CaT is central to the development of SDA APD.
Potential mechanisms underlying Ca 2+ handling abnormalities. Cellular Ca 2+ handling is a complex process involving a wide range of proteins involved in Ca 2+ transport, subcellular organelles and architecture, and regulatory processes. A number of molecular mechanisms have been implicated in CaT alternans. SERCA2a is responsible for cytosolic Ca 2+ uptake. SERCA2a downregulation is a classical feature of cellular remodeling in heart failure, conspiring with increased RyR2 leak and PLB hypophosporylation to reduce ventricular cardiomyocyte Ca 2+ stores and cause hypocontractility (28). SERCA2a downregulation is a principal correlate of repolarization alternans in heart failure (29), and alternans and arrhythmogenesis are suppressed by SERCA2a gene transfer (30). However, SERCA2a expression was not changed in our AF ACMs (Supplemental Figure  9H). Consistent with our observation, SERCA2a overexpression and inhibition had little effect on alternans behavior in rat ACMs (31). Similarly, enhancing SERCA2a function with a gain-of-function PLB mutation or mildly suppressing SERCA2a function had minimal effects on Ca 2+ alternans in mouse hearts (32).
A number of recent studies suggest that enhanced RyR2 Ca 2+ leak can cause cellular CaT alternans (10,33). Local Ca 2+ sparks promote CaT alternans by a spark-induced spark mechanism in a computational model (34). Post-myocardial infarction dogs with enhanced RyR2 leak due to redox modulation of RyR2 similarly show susceptibility to CaT alternans (35). These observations are consistent with our findings indicating a central role of AF-induced RyR2 Ca 2+ leak in CaT and APD alternans. Mathematical modeling of the complex determinants of Ca 2+ alternans does predict, at least under certain conditions, increased Ca 2+ spark rate increasing the likelihood of alternans (33). On the other hand, extensive observations suggest that, on the contrary, enhancing RyR2 function (e.g., with caffeine) reduces RyR2 refrac- toriness (thereby alleviating Ca 2+ alternans) whereas suppressing RyR2 Ca 2+ release promotes alternans. For example, Sun et al. found that a gain-of-function RyR2 mutation shortened Ca 2+ release refractoriness and suppressed alternans in the mouse ventricle (32). Interestingly, the mutation did not cause spontaneous Ca 2+ release events, unlike the effect of AF in our model. Similarly, Wang et al. noted that sensitizing RyR2 with 200 μM caffeine suppressed alternans in the ventricles of isolated rabbit hearts (5). Alternans is clearly a complex and multifactorial phenomenon, but the refractoriness of the cellular Ca 2+ transient is clearly a key determinant of atrial Ca 2+ alternans (36), as we observed in the present study.
In addition to the increased SR Ca 2+ loading, we found a significant reduction in CSQ expression in AF-remodeled atria. CSQ is the major Ca 2+ binding protein in the SR and acts as a Ca 2+ buffer regulating SR free Ca 2+ concentration (38). CSQ may also inhibit RyR2 Ca 2+ release independently of its buffering action. In heterozygous CSQ-knockout mice, moderate reductions (about 25%) in CSQ expression can increase SR Ca 2+ leak and arrhythmia susceptibility in the presence of β-adrenoceptor stimulation (39). In CSQ-knockout mice, spontaneous atrial Ca 2+ sparks are increased, and atrial tachypacing induces AF/flutter, which can be suppressed by the RyR2 inhibitor (R)-propafenone (40). These findings suggest that the reduced CSQ expression that we noted in AF ACMs (Supplemental Figure 10) may at least in part explain the spontaneous RyR2 leak that we observed. Finally, we found a reduction in RyR2 expression in AF ACMs (Supplemental Figure 8). Genetic suppression of RyR2 function prolongs Ca 2+ release refractoriness and promotes Ca 2+ release alternans in mouse hearts (41), suggesting that the RyR2 downregulation that we observed may also have contributed to the cellular phenotype.
Potential limitations. In this study, we noted that increased SR Ca 2+ leak as quantified by spontaneous Ca 2+ spark activity was associated with and paralleled prolonged CTR, which in turn was closely associated with Ca 2+ and APD alternans. The refractory period of the CaT has multiple determinants, including L-type Ca 2+ current (I CaL ) properties, SR Ca 2+ load, and the intrinsic refractoriness of RyR2. I CaL is known to be reduced in AF (42), and this change could have contributed to the alterations in CTR that we noted. In addition, caffeine and dantrolene would both be expected to alter SR Ca 2+ stores, which could also affect CTR. Changes in diffusional distance between RyR2 clusters in association with changes in cellular ultrastructure and alterations in intracellular buffering could also have played a role. Much more detailed studies of beat-to-beat changes in the relationships among SR Ca 2+ stores, I CaL , and SR Ca 2+ release will be needed to determine the precise mechanisms underlying the changes in CTR that we observed with AF, caffeine, and dantrolene.
We used a single animal model for this study, electrically maintained AF for 1 week in the dog. We used this model to test a specific hypothesis: that AF-induced Ca 2+ handling remodeling can affect the recovery of the CaT and thereby affect alternans behavior and the susceptibility to reentry. However, we cannot know whether different durations of AF might have different effects. Furthermore, many animal models are available for the study of AF (43), and there is no a priori reason to think that similar changes occur in other models. In addition, the model does not consider the effects of structural modeling, such as atrial fibrosis, which plays a key role in AF pathophysiology (1,44).
In the dual optical mapping studies, the probes for [Ca 2+ ] i and V m imaging share the same excitation band but have distinct emission bands. The fluorescence signals Rhod-2, AM, and RH 237 dyes emitted showed a shallow band of crosstalk in the emission spectra. Fluorescence signals were separated into different detectors using dichroic mirrors and emission filters. The 2 charge-coupled device (CCD) cameras must be precisely aligned to ensure that CaT and V m signals are indeed acquired from the same myocardial sites. We aligned the cameras with a grid reference placed in the field of view, and absolute precision of alignment cannot be ensured.
Novel findings and significance. The principal potentially novel finding of this study is that AF-associated Ca 2+ handling abnormalities predispose not only to spontaneous atrial ectopic firing, as previously noted (8,9), but also to the initiation and maintenance of atrial reentry. This observation places Ca 2+ handling disturbances squarely in the center of AF-promoting mechanisms and supports efforts to design novel compounds that improve Ca 2+ handling to suppress AF (2,45,46).
There have been prior studies that demonstrate beneficial effects of dantrolene on cellular abnormalities in AF models. Avula et al. (19) showed that sheep with myocardial infarction have spontaneous Ca 2+ release-related atrial ectopic activity that initiates atrial reentry and that dantrolene can suppress both ectopic firing and spontaneous AF episodes in the sheep. Their study followed up on prior work demonstrating Ca 2+ -driven atrial ectopy in a canine model of chronic atrial infarction (17). Hartmann et al. (47) similarly showed that dantrolene suppresses Ca 2+ sparks and delayed afterdepolarizations in ACMs from patients with AF. Our work differs from these studies in assessing the relationships between AF-induced changes in Ca 2+ handling, cellular alternans behavior, and reentry susceptibility, along with assessing in detail dantrolene's effects on the substrate for reentry rather than just the trigger, subjects that have not been addressed before in the literature, to our knowledge.
Our study potentially explains a consistent observation that has been somewhat enigmatic. In mouse models with aberrant Ca 2+ handling and spontaneous Ca 2+ releases, AF inducibility and persistence are consistently increased (18,40,48,49). Furthermore, suppression of abnormal Ca 2+ releases reduces susceptibility to AF induction and the persistence of any AF induced. Arrhythmia inducibility and maintenance are generally related to the reentrant substrate, and it has been unclear why they should be altered by spontaneous Ca 2+ leak events. Our observation that spontaneous Ca 2+ release leads to repolarization alternans that facilitates reentrant AF induction and maintenance can potentially explain the prior findings. Furthermore, the observation that a clinically available agent, dantrolene, suppresses the induction of SDA CaT and SDA APD in association with reduced AF vulnerability and AF sustainability points to the potential translational relevance of our findings to the search for new antiarrhythmic drug targets and agents for the treatment of this very common and problematic condition.

Methods
Canine AF model. Adult male mongrel dogs (20 to 35 kg) were obtained from LAKA Inc (9319.2763) and assigned to CTL and AF groups. After ketamine (5.3 mg/kg, IV)/diazepam (0.25 mg/kg, IV)/1.5% isoflurane anesthesia, AF dogs underwent fluoroscopically guided bipolar pacing lead insertion into the RA appendage and connection to a pacemaker in the neck. After 24 hours' postoperative recovery, dogs were atrial paced at 600 bpm to maintain fibrillatory atrial activity for 7 days. The CTL animals were handled identically to AF dogs, but their pacemaker was not activated.
The left coronary artery was cannulated, and the LA tissue was dissected free and perfused with 1.8 mM Ca 2+ containing Tyrode's solution (37°C, 100% O 2 ). Leaks from arterial branches were ligated, and tissues were perfused with Ca 2+ -free Tyrode's solution for 15 to 20 minutes, followed by Ca 2+ -free Tyrode's solution containing collagenase (~0.45 mg/mL, CLS II, Worthington) plus 0.1% bovine serum albumin (MilliporeSigma) for 1 hour. Digested tissue was carefully minced and agitated, and LA ACMs were harvested. After isolation, cells were kept in 0.2 mM Ca 2+ -containing Tyrode's solution for Ca 2+ imaging experiments (16).
CTR was determined with an S1S2 protocol (51). CaTs were induced with field stimulation at 1 Hz for 10 seconds (S1), followed by a single S2 pulse. A series of S1S2 pulse protocols with progressively decreased S1-S2 coupling intervals (from 1000 to 100 ms) was applied while CaTs were continuously recorded. CTR was defined as the longest S1-S2 coupling intervals that failed to induce a CaT. The Ca2 + release restitution curves were constructed by plotting each S2/S1 CaT ratio against the corresponding S1-S2 coupling interval. The resulting data points were fitted with a single exponential function.
Caffeine-induced CaTs and corresponding NCX currents were simultaneously recorded in ACMs for quantification of SR Ca 2+ content. Membrane potential was held at 80 mV. SR Ca 2+ content was assessed by rapidly applying caffeine (10 mmol/L) after 1 minute of Ca 2+ loading with continuous pacing. Caffeine-induced NCX current was integrated to calculate SR Ca 2+ content (8).
Confocal Ca 2+ imaging. To assess SCaSs, ACMs were incubated with 10 μM Fluo-4 AM (Invitrogen, Thermo Fisher Scientific) for 15 minutes. Subsequently, ACMs were placed on glass coverslips and superfused with 1.8 mM Ca 2+ -containing Tyrode's solution for 10 minutes to allow intracellular de-esterification. Fluorescence images (16 bits, 10-ms intervals, pixel size = 0.2 μm) were acquired with a ZEISS LSM 5 LIVE confocal fluorescence microscope equipped with a 63×/1.4 Plan Apochromat oil objective. Fluo-4 AM was excited with a 488nm argon laser, and emission signals were collected over 505 nm. Background subtracted fluorescence emission signals (F) were normalized to baseline fluorescence (F 0 ) by averaging 20 images. The changes in [Ca 2+ ] i are presented as ΔF/F 0 (where ΔF = F -F 0 ) (52). The SCaS frequency was calculated according to previously described approaches (53). Briefly, a region mask was chosen to define the cell outline and prevent interference from the cell exterior. A nonlinear partial differential equation-based method was combined with anisotropic diffusion and median filtering to improve signal-to-noise ratio. Fourth-order diffusion was applied to smooth small gradient noisy areas, while the areas having larger gradients remained nondiffused (54). Median filtering was applied subsequently to remove impulsive spikes generated by noise (55). The detection step used image segmentation to determine spark regions. Morphology dilation was performed first, and then connected components were labeled, followed by calculation of spark numbers and size. After finishing the preliminary calculation of sparks for each frame, the spark number was verified and counted. For spark centers that were in the region of a prior spark, the newly detected sparks were considered a continuation of the event in the previous frame and assigned the same ID number. The variables analyzed included Ca 2+ spark frequency, spark area, spark amplitude, spark mass, and spark-mediated leak. The spark frequency was calculated as the total detected Ca 2+ events over the recording time in the cell area of field of view. Spark amplitude was measured by the ratio of maximum F/F 0 at a spark event, indicating the highest relative fluorescence ratio. The spark area was measured at the time of maximum spark amplitude. The spark mass was calculated as spark amplitude × spark area × 0.33(change in fluorescence/initial fluorescence [ΔF/F 0 × μm 2 ]) and SR Ca 2+ spark mediated-leak = spark mass × spark frequency(ΔF/F 0 × S −1 ), using a similar method to Benoist et al.'s method of Ca 2+ spark detection (56).
Optical mapping of V m and [Ca 2+ ] i . The right atrium was dissected free, and the right coronary artery was cannulated and perfused with Krebs solution (in mM: 120 NaCl, 4 KCl, 1.2 MgSO 4 0.7, 1.2 KH 2 PO 4 , 25 NaHCO 3 , 5.5 glucose, 1.25 CaCl 2 ; in 95% O 2 /5% CO 2 ) at 20 mL/min and 37°C. Any leak from arterial branches was ligated with silk thread to maintain adequate perfusion. After 10 minutes of stabilization and electromechanical uncoupling with blebbistatin (15 μM) (57), the preparation was loaded with Rhod-2, AM (Invitrogen, Thermo Fisher Scientific), and RH 237 (MilliporeSigma) for simultaneous optical mapping of CaT and V m . Subsequently, the tissue was excited with a laser (Thorlabs) at a wavelength of 520 ± 45 nm. The bandpass filters for RH 237 and Rhod-2, AM, were 675 ± 25 nm and 590 ± 15 nm, respectively (58,59). Fluorescence images were recorded with 2 CCD cameras (CardioCCD, Redshirt Imaging) focused on a 1.5 × 1.5 cm 2 region. A pair of bipolar electrodes were positioned on the superior RA appendage, and the pacing stimuli were 2 ms pulse widths × 1.5′threshold square wave current.
Ca 2+ /APD alternans was elicited during S1-S1 pacing with 15-second pulse trains separated by 30-second intervals to minimize pacing memory. Starting at 500 ms, the BCL was shortened to 300 ms in 100-ms intervals, to 180 ms in 20-ms intervals, and then in 10-ms steps until Ca 2+ /APD alternans was induced. APD 80 was calculated as the time from maximal upstroke velocity (dF/dt max ) to 80% repolarization. CaT or APD alternans was quantified by subtracting the CaT amplitude or APD 80 for consecutive beats; alternans was defined as interbeat CaT amplitude or APD 80 differences averaging more than 5% over 6 stimuli. The threshold BCL was defined as the maximum BCL that induced Ca 2+ or APD alternans.
To assess the transition from spatial concordant to discordant Ca 2+ /APD alternans, the BCL was decreased from the APD alternans threshold BCL in 10-ms steps until AF was induced or 1:1 capture failed. AF was defined as a rapid (>400 bpm) irregular atrial rhythm lasting more than 1 second. If AF lasted more than 30 seconds, a 5-minute rest period was allowed after AF termination. The phase of alternans was termed positive for long short APD sequences (or high low CaT sequences), and negative for short long APD sequences (or low high CaT sequences). To evaluate spatial characteristics, alternans was classified as SCA or SDA. SCA was defined by at least 10% of the field of view displaying alternans that was entirely in phase. SDA was defined when the field of view simultaneously displayed both positive-and negative-phase alternans, with both positive and negative alternans occupying more than 10% of the surface area.
CTR was determined with an S1-S2 protocol (51). A single extrastimulus (S2) was delivered after 10 consecutive S1 stimuli at a BCL of 300 ms and decremented in 10-ms steps until failure to elicit a CaT. To quantify CaT recovery from refractoriness, we evaluated the ratio of the S2-induced CaT amplitude to the S1-induced CaT amplitude (S2/S1 CaT ratio) at S1-S2 coupling intervals ranging from 300 to 170 ms for each condition (5). Maps of S2/S1 CaT ratio were constructed at different S1-S2 coupling intervals for each condition. The Ca 2+ release restitution curves were constructed by plotting the S2/S1 CaT ratio against the corresponding S1-S2 coupling interval. The resulting data points were fitted with a single exponential function. The Ca 2+ release restitution curves were constructed in 4 locations. The decay τ of each Ca 2+ release restitution curve was determined to calculate the average τ of the restitution curve. The COV (SD/mean) of the S2/S1 CaT ratio (COV ratio) and of the Ca 2+ release restitution τ (COV τ) was determined to evaluate the spatial dispersion of Ca 2+ handling.
Phase analysis was used to detect and follow PSs (60,61). Hilbert transformation was performed first on the filtered signal from each pixel with the mean value during the recording period. The phase angle (θ) was calculated as the inverse tangent of the imaginary versus real part of the transformation result. Singularity detection was performed by using the concept of topological charge, implemented as a series of convolution operations. PSs were identified by a topological charge of ±1 (62). The algorithm was implemented as a line integral of differential phase maps around each pixel on an 8-pixel length enclosed path. The pixels around which the phase progresses through a complete cycle from -π to π are of great interest (63). To avoid noise effects, the double-ring method was used (64).
In vivo electrophysiological study. Dogs were anesthetized with morphine (2 mg/kg, SC)/alpha-chloralose (120 mg/kg, IV) and ventilated, and a median sternotomy was performed. Bipolar electrograms were recorded with bipolar Teflon-coated stainless steel electrodes sutured on the RA appendage (RAA). ARIs were measured as indices of APD (65), from the steepest portion of the initial atrial depolarization deflection to the end of the subsequent atrial repolarization signal. ARI alternans was defined by interbeat differences more than 5% over at least 6 successive beats, induced by a decremental S1-S1 pacing protocol (2 ms, 1.5 × diastolic threshold current stimuli). Starting at 300 ms, the BCL was decremented every 15 seconds in 10-ms steps until alternans occurred. Each pacing sequence was separated by at least 30 seconds to minimize the pacing memory. The BCL threshold for ARI alternans was defined as the maximal BCL that induced alternans. To assess AF vulnerability, an S1-S2 pacing protocol was applied at 8 discrete sites: RAA, left atrial appendage, and right and left atrial posterior walls, inferior walls, and Bachmann's bundles. AF vulnerability was defined as the percentage of atrial sites at which AF more than 1 second was reproducibly elicited by a single premature extrastimulus. To estimate mean AF duration in each dog, AF was induced with 1-second burst pacing applied 10 times (2 ms, 50 Hz, 4 × threshold current stimuli) in the RAA. Sustained AF (>20 minutes) was terminated by direct current cardioversion, and a 20-minute rest period was allowed for cardiac electrical recovery. If sustained AF was induced twice, no further AF induction was performed. After baseline data