Expression and Impact of Adenosine A3 Receptors on Calcium Homeostasis in Human Right Atrium

Increased adenosine A2A receptor (A2AR) expression and activation underlies a higher incidence of spontaneous calcium release in atrial fibrillation (AF). Adenosine A3 receptors (A3R) could counteract excessive A2AR activation, but their functional role in the atrium remains elusive, and we therefore aimed to address the impact of A3Rs on intracellular calcium homeostasis. For this purpose, we analyzed right atrial samples or myocytes from 53 patients without AF, using quantitative PCR, patch-clamp technique, immunofluorescent labeling or confocal calcium imaging. A3R mRNA accounted for 9% and A2AR mRNA for 32%. At baseline, A3R inhibition increased the transient inward current (ITI) frequency from 0.28 to 0.81 events/min (p < 0.05). Simultaneous stimulation of A2ARs and A3Rs increased the calcium spark frequency seven-fold (p < 0.001) and the ITI frequency from 0.14 to 0.64 events/min (p < 0.05). Subsequent A3R inhibition caused a strong additional increase in the ITI frequency (to 2.04 events/min; p < 0.01) and increased phosphorylation at s2808 1.7-fold (p < 0.001). These pharmacological treatments had no significant effects on L-type calcium current density or sarcoplasmic reticulum calcium load. In conclusion, A3Rs are expressed and blunt spontaneous calcium release at baseline and upon A2AR-stimulation in human atrial myocytes, pointing to A3R activation as a means to attenuate physiological and pathological elevations of spontaneous calcium release events.


Introduction
Cyclic AMP (cAMP) signaling plays a crucial role in modulating calcium regulatory proteins involved in cardiac excitation-contraction coupling, such as L-type calcium channels, the sarcoplasmic reticulum (SR) calcium channel, also named the ryanodine receptor (RyR2), and phospholamban that regulates the activity of the SR calcium pump [1,2]. Physiological and pathological modulation of cAMP signaling, in turn, involves a large number of G proteincoupled receptors (i.e., GPCRs) and phosphodiesterases [3,4]. Within GPCRs, adenosine receptors play a key role in the regulation of myocardial function and rhythm [5]. In this context, the Gi-protein coupled adenosine A 1 receptor (A 1 R) is expected to reduce cAMP production and attenuate the sympathetic tone, and the A 1 R is a pharmacological target for 2 of 12 the regulation of supraventricular arrhythmias [6]. However, excessive A 1 R activation can also accelerate atrial fibrillation (AF) [7] or favor its induction by shortening the refractory period via activation of the G-protein coupled inwardly rectifying potassium channel [8]. Furthermore, both the A 1 R and the Gi-protein coupled adenosine A 3 receptor (A 3 R) has been attributed important roles in ischemic preconditioning and cardio protection [9][10][11]. Moreover, the adenosine A 2A receptor (A 2A R) and A 2B receptor (A 2B R) are Gs-protein coupled receptors that are expected to stimulate cAMP synthesis and favor cAMP-dependent phosphorylation of key calcium regulatory proteins. Indeed, the A 2A R displays an overlapping distribution with the RyR2 and has previously been shown to selectively modulate spontaneous calcium release from the SR [12]. Moreover, A 2A R expression is upregulated in patients with AF and prevention of A 2A R activation normalizes spontaneous calcium release in patients with AF to levels observed in patients without AF [13], and diminishes the induction of arrhythmic responses in electrically paced myocytes from patients with AF [14]. Because the A 3 R is expected to inhibit adenylate cyclase, activation of this receptor would be expected to dampen spontaneous A 2A R-induced calcium release and contribute to maintaining a low incidence of spontaneous calcium release at baseline. Currently, the functional role of A 3 Rs in atrial myocytes remains elusive, and there are notable differences in A 3 R expression or binding of agonists to A 3 Rs in atria from humans and small rodents [15]. However, since there are speciesdependent differences in the expression of G-protein coupled receptors and their binding constants for A 3 R agonists [15,16], this study aims to determine the expression of A 3 Rs in human right atrial samples and the functional impact on intracellular calcium homeostasis in human right atrial myocytes.

Results
To determine the functional impact of A 3 Rs in the human atrium, we analyzed the expression and functional electrophysiological impact of A 3 Rs in myocytes from 53 patients without a previous history of AF. Table 1 summarizes the clinical features of the study population.

Adenosine A 3 R Expression
First, we aimed to determine the expression levels of A 3 R mRNA in comparison to the other adenosine receptors. The results shown in Figure 1 indicate that A 1 R mRNA is the most abundant, followed by the A 2A R and the A 3 R. Specifically, the expression of A 2A R mRNA constituted the 35 ± 5% of the A 1 Rs, while A 3 R accounted for only 9.2 ± 3.3%.

Adenosine A3R Expression
First, we aimed to determine the expression levels of A3R mRNA in comparison to the other adenosine receptors. The results shown in Figure 1 indicate that A1R mRNA is the most abundant, followed by the A2AR and the A3R. Specifically, the expression of A2AR mRNA constituted the 35 ± 5% of the A1Rs, while A3R accounted for only 9.2 ± 3.3%.

Impact of Adenosine A3Rs on Calcium Homeostasis at Baseline
However, since GPCRs can be located in macromolecular clusters where they exert a strong regulation of specific molecular functions [17], we first determined how selective A3R inhibition affected calcium homeostasis at baseline. As shown in Figure 2A,B, the selective A3R antagonist MRS1191 significantly increased the incidence of ITI (Figure 2A), without affecting the caffeine releasable SR calcium load ( Figure 2B). Furthermore, MRS1191 did not have a significant effect on the ICa amplitude ( Figure 2C), but significantly increased the ICa inactivation ( Figure 2D).

Impact of Adenosine A 3 Rs on Calcium Homeostasis at Baseline
However, since GPCRs can be located in macromolecular clusters where they exert a strong regulation of specific molecular functions [17], we first determined how selective A 3 R inhibition affected calcium homeostasis at baseline. As shown in Figure 2A,B, the selective A 3 R antagonist MRS1191 significantly increased the incidence of I TI (Figure 2A), without affecting the caffeine releasable SR calcium load ( Figure 2B). Furthermore, MRS1191 did not have a significant effect on the I Ca amplitude ( Figure 2C), but significantly increased the I Ca inactivation ( Figure 2D).

Impact of Crosstalk between A 3 R and A 2A R on Spontaneous Calcium Release
Since we have previously shown that A 2A R activation contributes to a higher incidence of I TI in human atrial myocytes [12] and A 3 R activation would counteract this, we tested whether there is a crosstalk between A 3 Rs and A 2A Rs in human atrial myocytes. For this purpose, we first exposed human atrial myocytes to 200 nM CGS21680 to simultaneously stimulate A 2A Rs and A 3 Rs. As shown in Figure 3A, this significantly increased the I TI frequency four-fold. Interestingly, subsequent exposure to MRS1191 induced an additional 3-fold increase in the I TI frequency (p < 0.001), suggesting that the A 3 R blunts the effect of A 2A R activation. Analysis of the caffeine-releasable SR calcium load revealed that the treatment with CGS21680 or CGS21680 + MRS1191 did not affect the SR calcium load significantly ( Figure 3B), suggesting that the increased incidence of I TI is not caused by a higher SR calcium load. Accordingly, there was no significant correlation between SR calcium load and I TI frequency (p = 0.789; Figure 3C). Immunofluorescent labeling of the RyR2 phosphorylated at s2808 revealed that phosphorylation was significantly higher in myocytes incubated with CGS21680 + MRS1191 than in control myocytes from the same patient ( Figure 3D), suggesting that s2808 phosphorylation could contribute to the higher incidence of I TI observed after exposure to CGS21680 + MRS1191.
To determine whether A 2A R activation increases the I TI frequency by increasing the propensity of the RyR2 to open spontaneously, we analyzed the incidence of calcium sparks resulting from the opening of individual RyR2 clusters [18]. Figure 4A,B shows that both CGS21680 and CGS21680 + MRS1191 induced a dramatic increase in the calcium spark density. This concurred with a significant reduction of the calcium spark amplitude, which was most pronounced in the presence of CGS21680 + MRS1191 ( Figure 4C). On the contrary, the treatments had no significant impact on the width ( Figure 4D) or the decay of the calcium sparks ( Figure 4E).

Impact of Crosstalk between A3R and A2AR on Spontaneous Calcium Release
Since we have previously shown that A2AR activation contributes to a higher incidence of ITI in human atrial myocytes [12] and A3R activation would counteract this, we tested whether there is a crosstalk between A3Rs and A2ARs in human atrial myocytes. For this purpose, we first exposed human atrial myocytes to 200 nM CGS21680 to simultaneously stimulate A2ARs and A3Rs. As shown in Figure 3A, this significantly increased the ITI frequency four-fold. Interestingly, subsequent exposure to MRS1191 induced an additional 3-fold increase in the ITI frequency (p < 0.001), suggesting that the A3R blunts the effect of A2AR activation. Analysis of the caffeine-releasable SR calcium load revealed that the treatment with CGS21680 or CGS21680 + MRS1191 did not affect the SR calcium load significantly ( Figure 3B), suggesting that the increased incidence of ITI is not caused by a higher SR calcium load. Accordingly, there was no significant correlation between SR calcium load and ITI frequency (p = 0.789; Figure 3C). Immunofluorescent labeling of the RyR2 phosphorylated at s2808 revealed that phosphorylation was significantly higher in myocytes incubated with CGS21680 + MRS1191 than in control myocytes from the same patient ( Figure 3D), suggesting that s2808 phosphorylation could contribute to the higher incidence of ITI observed after exposure to CGS21680 + MRS1191. Data were recorded in nine myocytes from eight patients before (blue) and after perfusion of myocytes with MRS1191 (purple). Significant differences are indicated with * p < 0.05 (paired t-test). Table 2 summarizes the impact of the two treatments on all calcium spark features analyzed. Table 2. Impact of the treatment with 200 nM CGS21680 (CGS) and 1 µM MRS1191 (MRS) on the incidence of calcium sparks and their properties in human atrial myocytes. Significant differences between treatment and control are indicated with * p < 0.05, ** p < 0.01, *** p < 0.001.  (B) SR calcium load; (C) relationship between ITI frequency and SR calcium load. Data were recorded in 21 myocytes from 19 patients before (blue) and after perfusion of myocytes with CGS21680 (yellow) and CGS21680 and MRS1191 (orange); (D) RyR2 s2808 phosphorylation in 106 myocytes from 19 patients. Significant differences between the control and treatments are indicated with * p < 0.05; *** p < 0.001, ‡ p < 0.01 for CGS21680 + MRS1191 vs. CGS21680. Data in (A,B) were analyzed using ANOVA with Welch's correction (p-values are given above the bars) and Games-Howell post-test. Unpaired t-test was used in panel (D).

Frequencies
To determine whether A2AR activation increases the ITI frequency by increasing the propensity of the RyR2 to open spontaneously, we analyzed the incidence of calcium sparks resulting from the opening of individual RyR2 clusters [18]. Figure 4A,B shows that both CGS21680 and CGS21680 + MRS1191 induced a dramatic increase in the calcium spark density. This concurred with a significant reduction of the calcium spark amplitude, which was most pronounced in the presence of CGS21680 + MRS1191 ( Figure  4C). On the contrary, the treatments had no significant impact on the width ( Figure 4D) or the decay of the calcium sparks ( Figure 4E). (B) SR calcium load; (C) relationship between I TI frequency and SR calcium load. Data were recorded in 21 myocytes from 19 patients before (blue) and after perfusion of myocytes with CGS21680 (yellow) and CGS21680 and MRS1191 (orange); (D) RyR2 s2808 phosphorylation in 106 myocytes from 19 patients. Significant differences between the control and treatments are indicated with * p < 0.05; *** p < 0.001, ‡ p < 0.01 for CGS21680 + MRS1191 vs. CGS21680. Data in (A,B) were analyzed using ANOVA with Welch's correction (p-values are given above the bars) and Games-Howell post-test. Unpaired t-test was used in panel (D).

Impact of Crosstalk between A 3 R and A 2A R on L-type Calcium Current
Finally, the treatment with CSG21680 and CGS21680 + MRS1191 was used to assess the impact of A 2A Rs and A 3 Rs on I Ca . Figure 5A shows that neither A 2A R nor A 3 R activation had any impact on I Ca amplitude. However, concurrent activation of A 2A Rs and inhibition of A 3 Rs with CGS21680 + MRS1191 significantly increased time-dependent inactivation of I Ca ( Figure 5B). Moreover, Figure 5C shows that the time constant for fast I Ca inactivation (tau) was inversely correlated with the I TI frequency recorded in the same cell (p < 0.001). In contrast, Figure 5D showed only a weak correlation between tau and I Ca density, suggesting that I Ca inactivation by calcium influx through the proper L-type calcium channel is modest. However, Figure 5E shows that a brief transient exposure to caffeine, to eliminate SR calcium release-induced I Ca inactivation, unmasks a steeper correlation between tau and the I Ca density (p < 0.01). Even so, Figure 5F shows that the tau for the I Ca inactivation elicited after caffeine exposure is not modified by the treatments with CGS21680 and MRS1191, suggesting that A 2A Rs and A 3 Rs have a minor impact on I Ca amplitude or inactivation.  Table 2 summarizes the impact of the two treatments on all calcium spark features analyzed. Table 2. Impact of the treatment with 200 nM CGS21680 (CGS) and 1 μM MRS1191 (MRS) on the incidence of calcium sparks and their properties in human atrial myocytes. Significant differences between treatment and control are indicated with * p < 0.05, ** p < 0.01, *** p < 0.001.

Impact of Crosstalk between A3R and A2AR on L-type Calcium Current
Finally, the treatment with CSG21680 and CGS21680 + MRS1191 was used to assess the impact of A2ARs and A3Rs on ICa. Figure 5A shows that neither A2AR nor A3R activation had any impact on ICa amplitude. However, concurrent activation of A2ARs and inhibition of A3Rs with CGS21680 + MRS1191 significantly increased time-dependent inactivation of ICa ( Figure 5B). Moreover, Figure 5C shows that the time constant for fast ICa inactivation (tau) was inversely correlated with the ITI frequency recorded in the same cell (p < 0.001). In contrast, Figure 5D showed only a weak correlation between tau and ICa density, suggesting that ICa inactivation by calcium influx through the proper L-type calcium channel is modest. However, Figure 5E shows that a brief transient exposure to caffeine, to eliminate SR calcium release-induced ICa inactivation, unmasks a steeper correlation between tau and the ICa density (p < 0.01). Even so, Figure 5F shows that the tau for the ICa inactivation elicited after caffeine exposure is not modified by the treatments with CGS21680 and MRS1191, suggesting that A2ARs and A3Rs have a minor impact on ICa amplitude or inactivation.  (D,E) correlation between tau and I Ca density before and after transient exposure to 10 mM caffeine. Currents were recorded in 19 patients before (blue) and after perfusion of myocytes with CGS21680 (yellow) and CGS21680 + MRS1191 (orange). Statistical significance was evaluated using Pearson's product-moment correlation; (F) Time constant, tau, for the decay of the first I Ca after transient exposure to caffeine. Significant differences between control and treatments are indicated with; *** p < 0.001, ‡ p < 0.01 for CGS21680 + MRS1191 vs. CGS21680. Data were analyzed using ANOVA (p-values are given above the bars) and Tukey's post-test in (A,B,F). Number of cells are given below bars.

Main Findings
While the A 3 R has been attributed an important role in preconditioning and cardio protection [10,19], little is known about its functional role in human atrial myocytes. Here, we analyzed the impact of the A 3 R on intracellular calcium homeostasis and report that even though A 3 R mRNA expression is modest compared to the expression of the A 1 R and the A 2A R, endogenous activation of the A 3 Rs at baseline blunts the incidence of the spontaneous calcium release-induced I TI . Furthermore, crosstalk between A 3 R and A 2A R upon activation of both receptors reduces the incidence of both calcium sparks and I TI , demonstrating that A 3 R activation diminishes spontaneous, A 2A R-mediated, calcium release in human atrial myocytes. The findings also suggest that A 3 R activation could be a means of attenuating arrhythmogenic calcium release events induced by pathological elevations of the adenosine level.

Impact of the A 3 R on Calcium Homeostasis at Baseline
Previous electrophysiological studies in human atrial myocytes have reported a cAMPtonus at baseline [20], which is regulated by phosphodiesterases and modulates I Ca amplitude [21] as well as the incidence of spontaneous calcium release [22]. Consistent with this, we have also shown that the ruptured whole-cell patch configuration dialyses adenosine out of the cell, leading to a reduction of the I TI frequency, presumably because the endogenous adenosine level is sufficient to induce spontaneous, A 2A R-mediated, calcium release at baseline [13]. In accordance with these findings, we here observe that selective inhibition of the A 3 R with MRS1191 increases the basal I TI frequency, suggesting that endogenous adenosine not only activates A 2A Rs, but also A 3 Rs at baseline, and that the latter attenuates A 2A R-mediated activation of adenylate cyclase. Interestingly, we do not observe any significant effect of A 3 R inhibition on I Ca density or SR calcium loading, pointing to compartmentalization of A 3 R-mediated signaling. This finding is similar to previous observations on the impact of pharmacological manipulation of Gs-protein coupled receptors in human atrial myocytes where acute pharmacological manipulation of A 2A R or treatment of patients with β-adrenergic receptor blockers had no impact on I Ca density or SR calcium load [12,13,23].

Impact of Crosstalk between A 3 Rs and A 2A Rs on Calcium Homeostasis
Since AF has previously been associated with increased A 2A R expression and activation that promotes spontaneous calcium release [13], concurrent activation of the A 3 R would be expected to dampen A 2A R-mediated stimulation of calcium release. The present findings demonstrate that when A 2A Rs and A 3 Rs are activated simultaneously, A 3 R activation does indeed attenuate significantly the A 2A R-mediated increase in the incidence of both calcium sparks and I TI . In support of this finding, both A 3 R and A 2A R bind adenosine with an affinity of approximately 300 nM [24]. Similar to observations at baseline, A 3 R inhibition did not modify the I Ca density when the A 2A Rs and A 3 Rs were stimulated simultaneously with CGS21680. However, A 3 R inhibition did speed up I Ca inactivation and this was correlated with the I TI frequency but not with the I Ca amplitude recorded in the same cell. This, combined with the higher incidence of calcium sparks and increased RyR2 phosphorylation at s2808 observed upon concurrent activation of A 2A R and A 3 R suggests that the A 3 R selectively targets cAMP-dependent phosphorylation of the RyR2 and that this not only leads to a higher I TI frequency, but also leads to a faster calcium-release induced I Ca inactivation. Interestingly, the tau for I Ca inactivation was inversely proportional to the I Ca amplitude when cells had previously been exposed to caffeine to clear the SR calcium content and prevent calcium-release induced inactivation. However, even under these conditions, CGS21680 or CGS21680 + MRS1191 did not affect the I Ca amplitude significantly, confirming that A 2A R and A 3 R-dependent signaling targets the RyR2 but not the L-type calcium channel.

Study Limitations
In the present study, we have focused on the impact of A 3 Rs on intracellular calcium homeostasis. However, being a Gi-protein coupled receptor, it is conceivable that the A 3 Rs could also modulate the activity of other ion channels that are regulated by Gi-protein coupled receptors [8,9] or influence the activity of other Gs-protein coupled receptors [25]. This, in turn, would potentially influence the net impact of A 3 R activity on the amplitude and frequency of calcium release-induced afterdepolarizations. Similarly, the relative impact of A 3 Rs on A 2A R-mediated signaling will depend on the spatial distribution of the A 1 R, the A 2A R, and the A 3 R with respect to target proteins, such as the RyR2, Ltype calcium channels, phospholamban, etc. While this issue has been addressed in nonmyocardial preparations [25][26][27], such information is currently limited for atrial myocytes. In this regard, we did show that a gradual elevation of intracellular adenosine levels to pathological levels strongly increases the incidence of calcium waves and I TI and that this could be reversed by selective A 2A R inhibition [13], suggesting that A 2A R activation plays a prominent role in pathological elevations of the adenosine level. Moreover, this study uses human atrial myocytes that are well suited for translational studies of receptor-mediated modulation of electrophysiological function. However, we cannot rule out that our findings could potentially be affected or present variability due to variations in concurrent disease, risk factors, or pharmacological treatments among the study population. In this context, age has been shown to affect I Ca density [28] and sex has been shown to have differential effects on I Ca density and I TI frequency [29]. However, because this study does not compare different groups of patients, this issue is primarily expected to increase variability between measurements, rather than the effect of pharmacological treatments. Similarly, the present study was conducted in right atrial myocytes, and we cannot rule out that some of the findings may be specific to the right atrium.

Clinical Implications and Conclusion
We have previously reported that the expression of A 2A Rs is upregulated in AF, and underlies a higher incidence of calcium-release induced I TI and afterdepolarizations in atrial myocytes from patients with AF [13]. The relevance of these findings is further underscored by higher plasmatic adenosine levels and lower adenosine deaminase activity in patients with AF [30]. In this context, the present findings, showing that A 3 R activation regulates the impact of A 2A R activation on RyR2 phosphorylation and spontaneous calcium release, suggest that selective activation of adenosine A 3 Rs might be suitable to dampen pathological elevations of the incidence of spontaneous calcium release-induced electrical activity. In particular, cardioprotective approaches targeting the A 3 R during ischemia, where adenosine levels are known to surge [31], could be a point of departure to explore the potential of A 3 Ra as a novel target to prevent induction of atrial ectopic atrial activity.

Myocyte Isolation
Atrial myocytes were isolated from tissue fragments collected from 53 patients, without a previous history of AF, undergoing cardiac surgery at Hospital de la Santa Creu i Sant Pau in Barcelona. Clinical and echocardiographic data of these patients are summarized in Table 1. Myocytes were isolated from right atrial samples, as previously described [28].
Each patient gave written consent to obtain blood and tissue samples that would otherwise have been discarded during surgery.

Quantitative Real-Time PCR
Total RNA was isolated from human right atrial samples using a commercially available kit. First-strand cDNA was synthesized from 1 mg of total RNA. cDNA was amplified using TaqMan master mix and primers from Thermo Fisher Scientific (Waltham, MA, USA) for the human glyceraldehyde-3-phosphate dehydrogenase (GAPDH): Hs00266705_g1; for human A 1 R (ADORA1): Hs00181231_m1; for human A 2A R (ADORA2A): Hs00169123_m1; for human A 2A B (ADORA2B): Hs00386497_m1; and for human A 3 R (ADORA3): Hs00181232_m1.

Patch-Clamp Technique
Isolated myocytes were subjected to the perforated patch technique using a HEKA EPC-10 amplifier (HEKA Elektronik, Lambrecht/Pfalz, Germany). Series resistance compensation was not applied. The extracellular solution contained (in mM): NaCl 127, TEA 5, HEPES 10, NaHCO 3 4, NaH 2 PO 4 0.33, glucose 10, pyruvic acid 5, CaCl 2 2, and MgCl 2 1.8 (pH = 7.4). The pipette solution contained (in mM): aspartic acid 109, CsCl 47, Mg 2 ATP 3, MgCl 2 1, na 2 phosphocreatine 5, Li 2 GTP 0.42, HEPES 10 (pH = 7.2 with CsOH), and 250 µg/mL amphotericin B. I Ca density and properties were measured using a 200 ms depolarization from a holding potential of −80 mV to 0 mV. A 50 ms prepulse to −45 mV was used to inactivate the Na + current. The I Ca amplitude was normalized to the cell capacitance to obtain the I Ca density. The decay of the I Ca was fit with a double exponential to obtain the time constants tau-1 and tau-2 for fast and slow I Ca inactivation. This study focused on the fast time constant, which is modulated by calcium release from the SR and by calcium entry through the L-type calcium channel. I TI currents were recorded at a holding potential of −80 mV in 4 × 30 s intervals to determine the I TI frequency. Brief exposure (6s) to 10 mM caffeine at a holding potential of −80 mV was used to release calcium from the SR and the time integral of the resulting transient inward NCX-current was used to assess the SR calcium load. Transformation of the charge carried by the NCX-current assumed a stoichiometry of 3 Na + :1 Ca 2+ for the NCX. Working solutions containing 200 nM CGS21680 and/or 1 µM MRS1191 were prepared from 1 mM stock solutions dissolved in DMSO.

Confocal Imaging
Confocal calcium images (512 × 140 pixels) were recorded at 90 Hz with the Leica SP5 AOBS resonance-scanning confocal microscope in fluo-4 loaded myocytes, as described previously [32]. Experiments were carried out at room temperature. Calcium sparks were detected and clustered in 2 × 2 µm 2 regions of interest, termed spark sites, using a custom-made algorithm based on continuous wavelet transform of the temporal profile at every spatial location, as described elsewhere [33]. The calcium spark frequency and the number of spark sites were normalized to the cell area to obtain the calcium spark density (sparks/s/1000 µm 2 ) and the spark site density (spark sites/1000 µm 2 ). In addition, we calculated the number of sparks per site (sparks/site/s). A series of morphological features were measured for each spark signal: Relative amplitude of the peak to the local baseline (F/F0), full duration at half maximum (FDHM), decay constant of an exponential fit (tau), the coefficient of determination of the exponential fit (R2), and full width at half maximum (FWHM).

Data Analysis
Experimental data were collected and analyzed without knowledge about clinical data and clinicians did not know the experimental results. Statistical analysis was carried out using RStudio 4.2.2 statistical software. Unless otherwise stated, data were averaged for each patient and results are given as mean ± s.e.m. with indication of the number of patients in each group. Fisher's exact test was carried out for categorical data. Student's t-test was used for paired or unpaired comparisons, and ANOVA, ANOVA with Welch correction or Kruskal-Wallis were used for comparison of multiple effects, as indicated. Tests used are indicated for each figure and statistically significant effects are indicated with p-values or *: p < 0.05, **: p < 0.01; ***: p < 0.001.

Institutional Review Board Statement:
The study was conducted in accordance with the Declaration of Helsinki, and approved by the ethics committee at Hospital de la Santa Creu i Sant Pau, Barcelona, Spain (protocol code AZAR-AF_2015, approved 30/11/2015).

Informed Consent Statement:
Informed consent was obtained from all subjects involved in the study.

Data Availability Statement:
The data that support the findings of this study are available from the corresponding author upon reasonable request. Some data may not be made available because of privacy or ethical restrictions.