Ca2+ cycling properties are conserved despite bradycardic effects of heart failure in sinoatrial node cells

Background: In animal models of heart failure (HF), heart rate decreases due to an increase in intrinsic cycle length (CL) of the sinoatrial node (SAN). Pacemaker activity of SAN cells is complex and modulated by the membrane clock, i.e., the ensemble of voltage gated ion channels and electrogenic pumps and exchangers, and the Ca2+ clock, i.e., the ensemble of intracellular Ca2+ ([Ca2+]i) dependent processes. HF in SAN cells results in remodeling of the membrane clock, but few studies have examined its effects on [Ca2+]i homeostasis. Methods: SAN cells were isolated from control rabbits and rabbits with volume and pressure overload-induced HF. [Ca2+]i concentrations, and action potentials (APs) and Na+–Ca2+ exchange current (INCX) were measured using indo-1 and patch-clamp methodology, respectively. Results: The frequency of spontaneous [Ca2+]i transients was significantly lower in HF SAN cells (3.0 ± 0.1 (n = 40) vs. 3.4 ± 0.1 Hz (n = 45); mean ± SEM), indicating that intrinsic CL was prolonged. HF slowed the [Ca2+]i transient decay, which could be explained by the slower frequency and reduced sarcoplasmic reticulum (SR) dependent rate of Ca2+ uptake. Other [Ca2+]i transient parameters, SR Ca2+ content, INCX density, and INCX-[Ca2+]i relationship were all unaffected by HF. Combined AP and [Ca2+]i recordings demonstrated that the slower [Ca2+]i transient decay in HF SAN cells may result in increased INCX during the diastolic depolarization, but that this effect is likely counteracted by the HF-induced increase in intracellular Na+. β-adrenergic and muscarinic stimulation were not changed in HF SAN cells, except that late diastolic [Ca2+]i rise, a prominent feature of the Ca2+ clock, is lower during β-adrenergic stimulation. Conclusions: HF SAN cells have a slower [Ca2+]i transient decay with limited effects on pacemaker activity. Reduced late diastolic [Ca2+]i rise during β-adrenergic stimulation may contribute to an impaired increase in intrinsic frequency in HF SAN cells.

Heart failure (HF) reduces pacemaker activity of SAN cells through an increase in intrinsic cycle length (Opthof et al., 2000;Witte et al., 2000;Verkerk et al., 2003;Du et al., 2007). Previously, it has been shown that this was al least due to remodeling of components of the membrane clock. HF impairs the membrane clock by downregulation of I f in the SAN of rabbit (Verkerk et al., 2003) and downregulation of the corresponding hyperpolarization-activated (HCN) channel subunits, HCN4 and HCN2, in the SAN of dogs (Zicha et al., 2005). In addition, TTX-sensitive neuronal Na + channels, Na v 1.1 and Na v 1.6, are downregulated in SAN tissue of HF rats (Du et al., 2007). Finally, HF upregulates I Ks in HF SAN cells of rabbit (Verkerk et al., 2003), but since I Ks plays a limited role in pacemaker activity without adrenergic stimulation (Lei et al., 2000), this change in membrane clock is hardly involved in the increase in intrinsic cycle length during HF (Verkerk et al., 2003).
To date, the effect of HF on [Ca 2+ ] i homeostasis in atrial and ventricular myocytes has been studied in detail (for reviews, see Bers, 2006b;Bers et al., 2006;Kho et al., 2012;Eisner et al., 2013;Luo and Anderson, 2013;Neef and Maier, 2013), but the effect of HF on [Ca 2+ ] i in SAN cells is hardly known. Shinohara et al. (2010) found that HF, induced by rapid ventricular pacing, results in Ca 2+ clock malfunction in SAN of dogs, characterized by a reduction of the slope of late diastolic Ca 2+ elevation (LDCAE) as well as unresponsiveness to isoproterenol and caffeine in intact SAN. Especially the reduced slope of the LDCAE, with associated decrease in localized Ca 2+ releases (LCRs) or Ca 2+ sparks (Bogdanov et al., 2006;Maltsev et al., 2006;Joung et al., 2009Joung et al., , 2010van Borren et al., 2010), may have implications for the Ca 2+ clock (Stern et al., 2014).
In the present study, we evaluated the [Ca 2+ ] i homeostasis in isolated control (CTRL) and HF SAN cells using a well-established rabbit model of volume and pressure overloadinduced HF. We found a slower spontaneous [

CELL PREPARATION
The investigation was approved by the local ethics committee and conformed to the guiding principles of the Declaration of Helsinki. HF was induced in 5-month-old male New-Zealand White rabbits by combined volume-and pressure-overload in 2 sequential surgical procedures as described previously (Vermeulen et al., 1994;Verkerk et al., 2011). In short, volume overload was produced by rupture of the aortic valve until pulse pressure was increased by about 60-90%. Three weeks later, pressure overload was created by suprarenal abdominal aorta constriction of approximately 50%. Four months after the last operation, the rabbits were anaesthetized [ketamine (50 mg im; Eurovet, Bladel, The Netherlands) and xylazine (10 mg im; Eurovet, Bladel, The Netherlands)], heparinized (5000 IU, LEO Pharma, Breda, The Netherlands), and killed by intravenous injection of pentobarbital (240 mg; Ceva Sante Animale B.V., Maassluis, The Netherlands). Relative heart weight (i.e., heart weight to body weight ratio), relative lung weight (i.e., lung weight to body weight ratio), and presence of ascites were analyzed as described previously (Vermeulen et al., 1994). Sham-operated rabbits undergoing thoracotomy and agematched non-operated rabbits do not differ in various heart failure parameters (Vermeulen et al., 1994) and important cellular parameters for hypertrophy and ionic remodeling (Verkerk et al., 2011). In the present study, therefore, non-operated age-matched healthy animals served as control (CTRL). A total of 12 HF and 13 CTRL rabbits were used.
Single SAN cells were enzymatically isolated from the entire SAN region as described previously (Verkerk et al., 2009). Small aliquots of cell suspension were put in a recording chamber on the stage of an inverted microscope. Cells were allowed to adhere for 5 min after which superfusion with Tyrode's solution was started. Tyrode's solution (36 ± 0.2 • C) contained (in mM): NaCl 140, KCl 5.4, CaCl 2 1.8, MgCl 2 1.0, glucose 5.5, HEPES 5.0, pH 7.4 (NaOH). Spindle and elongated spindle-like cells displaying regular contractions were selected for measurements.

CALCIUM MEASUREMENTS
[Ca 2+ ] i was measured in spontaneously active indo-1 (Molecular Probes, Eugene, OR, USA) loaded myocytes as described previously in detail (van Borren et al., 2010). Such signals are a measure of global [Ca 2+ ] i transients, which are triggered by Ca 2+ influx through sarcolemmal Ca 2+ channels activated during spontaneous action potentials. As indicated in the Introduction section, these spontaneous action potentials in SAN cells likely involve a tight collaboration of both membrane and Ca 2+ clock systems. Of note, in this study we used the term "spontaneous [Ca 2+ ] i transient" analog to "spontaneous action potentials," without assumptions regarding the primary cause of the spontaneous activity. As before, we discerned and analyzed five distinct phases in [Ca 2+ ] i transients of SAN cells (van Borren et al., 2010): (1) (Varro et al., 1993), which also allows specific pre-pacing protocols. In SAN cells, however, we were unable to measure reliably the I NCX during caffeine-induced Ca 2+ transients, due to cell membrane rupture resulting in large "leak" currents in response to the strong and fast caffeine-induced contractions. Therefore, we assessed SR Ca 2+ content as the fractional release of SR Ca 2+ (Bers, 1987;Bassani et al., 1992)

ELECTROPHYSIOLOGY
Action potentials and I NCX were recorded by the amphotericinperforated patch-clamp technique using an Axopatch 200B amplifier (Molecular Devices, USA). Patch pipettes (borosilicate glass, 2-5 M ) were filled with solution containing (in mM): K-gluc 120, KCl 20, NaCl 5, amphotericin-B 0.22, NMDG-Cl (Nmethyl-D-glucammonium chloride) 10, HEPES 10, pH 7.2 (KOH 5.5). Series resistance was compensated by ≥80%, and potentials were corrected for the calculated liquid junction potential. Action potentials were low-pass filtered (cut-off frequency: 1 kHz) and digitized at 1 kHz; I NCX was filtered and digitized at 1 and 0.2 kHz, respectively. Voltage control and data acquisition were accomplished using pCLAMP 8 software (Molecular Devices, USA), while custom software was used for data analysis. Cell membrane capacitance was estimated by dividing the time constant of the decay of the capacitive transient in response to 5 mV hyperpolarizing voltage clamp steps from 0 mV by the series resistance, and amounted to 50.0 ± 6.9 pF in CTRL (n = 5) and 56.1 ± 8.1 pF in HF (n = 5) SAN cells (mean ± SEM, P > 0.05), in line with the values of 50 ± 4 (n = 23) and 52 ± 3 pF (n = 24) that we observed in a previous study (Verkerk et al., 2003). I NCX density and the I NCX -[Ca 2+ ] i relationship were measured as described previously (Weber et al., 2003). In short, a slow depolarizing voltage ramp from −85 to 120 mV results in an increase in [Ca 2+ ] i due to Ca 2+ entry through NCX operating in reverse mode (net outward I NCX ). Upon fast repolarization, Ca 2+ removal occurs resulting in a [Ca 2+ ] i decline due to Ca 2+ efflux through NCX operating in forward mode (net inward I NCX ) (Weber et al., 2003). The relationship between [Ca 2+ ] i and I NCX is membrane potential dependent, with a steeper relationship at more negative membrane potentials (Hove-Madsen and Tort, 2001). We used a repolarizing potential of −60 mV, which is close to the maximal diastolic potential of both CTRL and HF SAN cells (Verkerk et al., 2003). Thapsigargin (2.5 μM) was added to the Tyrode's solution to block SERCA. In addition, chromanol (5 μM) and E4031 (5 μM) were present to block the slow and rapid components of the delayed rectifier K + current, while CsCl (5 mM) was present to block I f . All drugs were obtained from Sigma (Zwijndrecht, The Netherlands) except for E4031, which was a gift from Eisai (Tokyo, Japan), and noradrenaline (Centrafarm, Etten-Leur, The Netherlands). I NCX densities were calculated by dividing current amplitudes by the cell membrane capacitance.

NUMERICAL RECONSTRUCTION OF SODIUM-CALCIUM EXCHANGE CURRENT
For the numerical reconstruction of I NCX on the basis of the simultaneously recorded experimental data on the membrane potential (V m ) and [Ca 2+ ] i (combined voltage and calcium clamp, van Borren et al., 2007), we adopted the I NCX formulation of the Lindblad et al. model for a rabbit atrial myocyte (Lindblad et al., 1996), with I NCX scaled down to 78% of its control value, based on mRNA data from Allah et al. (2011). We thus followed the approach of our previous study in which we reconstructed rabbit SAN I NCX in order to compare it with human SAN I NCX (Verkerk et al., 2013).
The Lindblad et al. (1996) equations not only require values for V m and [Ca 2+ ] i , but also for [Ca 2+ ] e , [Na + ] e and [Na + ] i , which denote the extracellular Ca 2+ concentration, the extracellular Na + concentration, and the intracellular Na + concentration, respectively. [Ca 2+ ] e and [Na + ] e were set to 1.8 and 140 mM, respectively, in accordance with the aforementioned Tyrode's solution, whereas [Na + ] i was set to 5 mM, based on the pipette solution. In some reconstructions, [Na + ] i was raised to 7.5 mM, as indicated.

STATISTICS
Data are mean ± SEM. Groups of HF SAN cells were compared with groups of CTRL SAN cells using Fisher's exact test, Two-Way Repeated Measures ANOVA followed by pairwise comparison using the Student-Newman-Keuls test, and unpaired t-test. Paired t-tests were used to compare drug effects within a group. An F-test was used to compare frequency dependency between groups. P < 0.05 is considered statistically significant. Table 1 summarizes the general characteristics of the HF model. Consistent with previous reports of the volume-and pressureoverload rabbit model of HF (Vermeulen et al., 1994;Baartscheer et al., 2003a,b;Verkerk et al., 2003Verkerk et al., , 2011Wiegerinck et al., 2006) body weight was similar, but heart weight was significantly higher in HF compared to CTRL rabbits. Consequently relative heart weight was significantly higher in HF rabbits. In addition, HF rabbits had an increased lung weight and relative lung weight. Moreover, six out of the 12 HF rabbits used in the present study exhibited ascites as assessed during autopsy, while none of the 13 CTRL rabbits did. Finally, we measured the weight of the right atrium to determine whether HF affects also the right atrium including the SAN. We found that both the absolute weight of the right atrium and the relative right atrial weight were significantly higher in HF rabbits. Figure 1A shows typical recordings of [Ca 2+ ] i transients from SAN cells of a CTRL and a HF rabbit; Figure 1B summarizes the  (Weber et al., 2003) and frequencydependent (Hussain et al., 1997;Dibb et al., 2007), while in the SAN an increase in late diastolic [Ca 2+ ] i rise was observed at higher beating rates (Joung et al., 2009). The role of the action potential duration in the observed slower [Ca 2+ ] i transient decay is likely limited, because the action potential duration in CTRL and HF SAN cells is similar (Verkerk et al., 2003). However, the decay of the [Ca 2+ ] i transient has a strong frequency dependency with a slower decay at slower frequencies (Hussain et al., 1997;Dibb et al., 2007). Thus, the slower [Ca 2+ ] i transient decay that we observed in HF SAN cells may be related to its slower intrinsic frequency. To test this hypothesis, we determined the relationship between frequency and [Ca 2+ ] i properties by plotting each of the five [Ca 2+ ] i transient parameters of every cell against its own frequency. The data have been fitted with linear regression lines (Figure 2). The linear fits to the CTRL as well as HF SAN cell data have all slopes significantly different from zero (see Table 2). However, neither the slopes nor the intercepts are significantly different between CTRL and HF for each of the five parameters (  Figure 1B) is larger than expected from the linear regression of the [Ca 2+ ] i transient time constant of decay vs. frequency relationship. Using the slope and intercept of the regression line for all cells, which amount to 151 (intercept) and −26 (slope), respectively, the difference in frequency between HF and CTRL cells explains only a ≈17% difference.

HF REDUCES THE FREQUENCY AND Ca 2+ TRANSIENT DECAY OF SAN CELLS
From these data we conclude that HF reduces the frequency and slows the Ca 2+ transient decay of SAN cells, the latter partially explained by the slower frequency in HF SAN cells.

HF DOES NOT AFFECT THE SR Ca 2+ CONTENT OF SAN CELLS
Our experiments demonstrate a reduced [Ca 2+ ] i transient decay rate, partially explained by a slower intrinsic frequency. The decline of the [Ca 2+ ] i transient is mainly due to uptake into the SR through SERCA and extrusion of Ca 2+ via the sarcolemmal Na + -Ca 2+ exchange (NCX), while mitochondrial Ca 2+ uptake and sarcolemmal Ca-ATPase may contribute for a very small fraction (Bers, 2006b;Dibb et al., 2007). In ventricular myocytes, a slower decline of the [Ca 2+ ] i transient by downregulation of SERCA results in a reduced SR Ca 2+ content (Bers, 2006b;Bers et al., 2006), while a slower decline of the [Ca 2+ ] i transient due to downregulation of the NCX may result in increased SR Ca 2+ content (Bers, 2006b;Eisner et al., 2013). To gain insight into the relationship in SAN cells, we analyzed the amplitude of the normal spontaneous [Ca 2+ ] i transient as a fraction of the caffeine-evoked transient in the presence of 5 mM NiCl 2 , termed fractional release, which is an estimate of the proportion of the SR Ca 2+ content released during each spontaneous transient.  Regression lines as in Figure 2. Slope and intercept data are ± standard error.
All slopes are significantly different from zero.
and a HF rabbit. The averaged fractional release of CTRL and HF SAN cells did not differ significantly [0.23 ± 0.04 (n = 23) vs. 0.22 ± 0.03 (n = 18)] ( Figure 3B). The caffeine-evoked transient in the presence of NiCl 2 can also be used to determine the aforementioned contribution of mitochondrial Ca 2+ uptake and sarcolemmal Ca-ATPase, because SERCA and NCX function is effectively by-passed by caffeine and Ni 2+ , respectively (Díaz et al., 2004

HF AFFECTS NEITHER THE I NCX NOR ITS [Ca 2+ ] i DEPENDENCY
Previously, the I NCX density was found to be similar in CTRL and HF SAN cells of rabbit (Verkerk et al., 2003). These results were obtained by measurements of Ni + -sensitive currents during a descending ramp, while the free [Ca 2+ ] i was buffered at 150 nM (Verkerk et al., 2003). HF may affect the relationship between [Ca 2+ ] i and I NCX (Díaz et al., 2004). Therefore, we next determined the relationship between [Ca 2+ ] i and I NCX as described previously by Weber et al. (2003). Figures 4A,B, show the protocol used and typical examples, respectively. A slowly depolarizing voltage ramp from −85 to 120 mV results in an increase in [Ca 2+ ] i due to Ca 2+ entry through reverse mode NCX ( Figure 4B). Upon repolarization to −60 mV, Ca 2+ removal occurs (forward mode NCX) resulting in a [Ca 2+ ] i decline ( Figure 4B, gray lines), which correlates well with the measured I NCX (Figure 4B, black lines) (cf. Weber et al., 2003). Figure 4C shows  Figure 4C). In addition, the time constants of both the [Ca 2+ ] i and I NCX decline during the repolarizing step to −60 mV, another measure of NCX function (Weber et al., 2003), do not differ significantly between CTRL and HF SAN cells ( Figure 4D). From these data we conclude that HF does not affect the I NCX density nor the I NCX -[Ca 2+ ] i dependency. In addition, given the unaltered NCX and mitochondrial Ca 2+ uptake and sarcolemmal Ca-ATPase activity of SAN cells, we conclude that the slower decay of the [Ca 2+ ] i transient is due to reduced SERCA activity.

HF HARDLY AFFECTS [Ca 2+ ] i MODULATION BY ACETHYLCHOLINE AND NORADRENALINE
[Ca 2+ ] i transients in SAN cells are importantly modulated by β-adrenergic (Rigg et al., 2000;Vinogradova et al., 2002a;Joung et al., 2009;van Borren et al., 2010)  now also the late diastolic [Ca 2+ ] i rise is significantly lower in HF SAN cells (Figure 5B, bottom right). Muscarinic receptor stimulation with 50 nM acetylcholine significantly decreased the frequency and [Ca 2+ ] i transients parameters in both CTRL and HF SAN cells, with exception of the time constant of decay which was significantly increased in both groups of cells (Figure 6). In presence of acetylcholine, only the frequency differs significantly between HF and CTRL SAN cells.
From these data we conclude that HF does not affect importantly the [Ca 2+ ] i modulation by acethylcholine and noradrenaline. However, in presence of noradrenaline, HF SAN cells have a reduced late diastolic [Ca 2+ ] i rise compared to CTRL SAN cells and this might have implications for pacemaker activity formation during β-adrenergic stimulation.

EFFECTS OF HF ON THE I NCX DURING THE DIASTOLIC DEPOLARIZATION PHASE
The [Ca 2+ ] i transients are longer in HF SAN cells, likely due to reduced SERCA activity. In general, it is assumed that this will promote NCX mediated Ca 2+ extrusion, and thus will increase the I NCX -mediated net inward current during the diastolic depolarization (Ju and Allen, 1998;Sanders et al., 2006;Lau et al., 2011). Such an effect may, at least partially, counteract the previously observed decrease in I f -mediated inward current in our used model of HF (Verkerk et al., 2003). However, NCX function during the cardiac cycle is not only determined by [Ca 2+ ] i , but also by the intracellular Na + ([Na + ] i ) concentration, and the membrane potential (V m ) (Blaustein and Lederer, 1999 (Figure 1) and of our previous study (Verkerk et al., 2003 ; Figure 1), although the differences in frequency and diastolic depolarization rate between CTRL and HF SAN cells did not reach the level of statistical significance due to the small number of cells (data not shown). Figures 7A,B, show a selected AP and [Ca 2+ ] i transient of a CTRL and HF SAN cell, which closely represent the mean differences in [Ca 2+ ] i transient and AP properties. Figure 7C shows the reconstructed I NCX , based on the simultaneously recorded Ca 2+ transients and APs of Figures 7A,B. In the CTRL SAN cell, I NCX is inward for the entire AP, with a small amplitude early in the AP and a maximal amplitude late during repolarization. The I NCX declined during the diastolic depolarization phase, consistent with the decrease in [Ca 2+ ] i . In HF, I NCX during the AP is close to that of the CTRL SAN cells, except that the amplitude is slightly larger during the diastolic depolarization. This is likely due to the slightly higher [Ca 2+ ] i in the diastolic depolarization phase, since the APs were almost identical and an identical [Na + ] i concentration of 5 mM was used for CTRL and HF SAN cells in both the pipette solution and calculations. Figure 7D extends our analysis to an increased [Na + ] i , as frequently observed in ventricular myocytes of our used HF model (Baartscheer et al., 2003b). Elevation of the [Na + ] i by 50% to 7.5 mM, in line with the observations of Baartscheer et al. (2003b), decreased the diastolic I NCX in both CTRL and HF SAN cells. Interestingly, I NCX may become even shortly outward early during the AP. Figure 7E shows the I NCX of the CTRL SAN cell with 5 mM [Na + ] i and the I NCX of the HF SAN cell with 7.5 mM [Na + ] i . It is evident that under these conditions I NCX in HF SAN cell is close to that of the CTRL SAN, except for the outward I NCX in the HF SAN cell. Similar results are obtained with the NCX model equations of Faber and Rudy (2000) (data not shown).
From these data we conclude that the decreased [Ca 2+ ] i transient decay in HF may result in slightly increased I NCX during the diastolic depolarization phase, but that this effect is counteracted by a HF-induced increase in [Na + ] i .

DISCUSSION
In the present study, we characterized the [Ca 2+ ] i cycling of single SAN cells isolated from CTRL rabbits and rabbits with volume and pressure overload-induced HF. The mechanisms behind [Ca 2+ ] i cycling in ventricular myocytes are diverse and complex (for reviews, see Guatimosim et al., 2002;Eisner et al., 2005;Bers, 2006b;Neef and Maier, 2013). In short, the [Ca 2+ ] i transient is triggered by Ca 2+ influx through sarcolemmal Ca 2+ channels, which results in release of Ca 2+ from the SR via ryanodine-2 (RyR2) channels. This so-called Ca 2+ -induced Ca 2+ release (CICR) is importantly modulated by T-tubuli organization, with co-localization of I Ca,L and RyR2 channels (Øyehaug et al., 2013),

www.frontiersin.org
February 2015 | Volume 6 | Article 18 | 7 and the open probability of RyR2 channels (Guatimosim et al., 2002). The amplitude of the [Ca 2+ ] i transient depends importantly on the SR Ca 2+ content (Bassani et al., 1992;Trafford et al., 2000Trafford et al., , 2001Díaz et al., 2004;Dibb et al., 2007;Briston et al., 2011). [Ca 2+ ] i transient amplitudes (Eisner et al., 2005;Bers, 2006a;Neef and Maier, 2013). The diastolic Ca 2+ concentration is regulated by the [Ca 2+ ] i transient decline, especially during rapid pacing (Laurita et al., 2003), and leak of RyR2 channels (Neef and Maier, 2013  HF in ventricular myocytes may affect many [Ca 2+ ] i transient characteristics. It is reported that HF decreases the upstroke velocity of the [Ca 2+ ] i transient by T-tubuli disorganization (Øyehaug et al., 2013) and increased open probability of RyR2 channels (Guatimosim et al., 2002). An increase of RyR2 open probability, as may occur during HF in ventricular myocytes, also leads to an increased amount of Ca 2+ sparks, resulting in an increased diastolic [Ca 2+ ] i concentration and reduced SR Ca 2+ content since more Ca 2+ is pumped out of the cell by the NCX (Neef and Maier, 2013). Furthermore, HF in ventricular mycoytes typically slows the [Ca 2+ ] i transient decay due to reduced SERCA function. Such reduced Ca 2+ uptake by SERCA, in combination with a frequently observed upregulation of the NCX and consequently increased Ca 2+ efflux, will reduce the SR Ca 2+ content and systolic Ca 2+ transient amplitude (for reviews, see Bers, 2006b;Eisner et al., 2013;Neef and Maier, 2013

INTRINSIC CYCLE LENGTH IS PROLONGED IN HF SAN CELLS
We found a lower frequency of spontaneous [Ca 2+ ] i transients in HF SAN cells, indicating that intrinsic cycle length was prolonged in HF SAN cells (Figure 1). This finding agrees with previously findings in intact SAN (Opthof et al., 2000) and single SAN cells (Verkerk et al., 2003) of rabbit. The frequency of the intrinsic cycle length based on the [Ca 2+ ] i transients was 14% slower in HF SAN cells (Figure 1) and this percentage closely matches the increased intrinsic cycle length (15%) of our previous study with the same rabbit HF model and measured with perforated patch clamp methodology (Verkerk et al., 2003).
We  (Figure 2, Table 2). All parameters, except the [Ca 2+ ] i transient time constant of decay, increased upon increased frequencies. The frequency dependency was not significantly different in HF compared to CTRL SAN cells (Figure 2, Table 2). Nevertheless, the frequency dependency relationships might have influenced some principal findings because spontaneous activity is lower in HF SAN cells. The relative small difference in intrinsic cycle length of CTRL and HF SAN cells, the rather large variation between cells, and the modest steepness of the frequency-dependencies (Table 2, Figure 2) may all have contributed to the absence of significant differences in frequency dependency between HF and CTRL SAN cells.
Recently, Herrmann et al. elegantly solved the problem of intrinsic rate differences in SAN cells by using electrical field stimulation (Herrmann et al., 2013). Using this approach, they investigated the contribution of the murine sodium-calcium exchanger  (Figure 4) nor the mitochondrial Ca 2+ uptake and sarcolemmal Ca-ATPase (Figure 3), indicating that downregulation of SERCA activity also contributes to the slower [Ca 2+ ] i transient decay. Any contribution of (changes in) T-tubular organization, sarcolemmal calcium currents, and action potential duration can be ruled out because T-tubuli are absent in rabbit SAN cells (Masson-Pévet et al., 1979), and I Ca,T and I Ca,L densities as well as action potential duration are unaffected in HF rabbit SAN cells (Verkerk et al., 2003). HF in SAN cells did not affect the diastolic and systolic [Ca 2+ ] i concentrations, systolic [Ca 2+ ] i rise, late diastolic [Ca 2+ ] i rise (Figure 1), and SR Ca 2+ content (Figure 3). The unaffected diastolic Ca 2+ concentration, at first sight, might appear to be inconsistent with the decrease of the [Ca 2+ ] i transient decline. It should be noted, however, that HF SAN cells have a lower frequency due to their longer diastolic depolarization phase (Verkerk et al., 2003), which leaves more time for Ca 2+ reuptake and/or removal. The unchanged slope of the LDCAE, associated with LCRs and Ca 2+ sparks, suggests that RyR2 open probability is not affected in HF SAN cells. Our finding contrasts findings in a canine model of rapid pacing-induced heart failure, where it was found, using isolated right atrial preparations, that LDCAE was reduced (Shinohara et al., 2010). Differences in HF model, species and preparations might contribute to the contrasting findings.
The unaltered SR Ca 2+ content agrees with the unaffected [Ca 2+ ] i transient amplitude in HF SAN cells, but is somewhat surprising given that reduced SERCA activity is supposed to result in a lower SR Ca 2+ content. The latter is due because NCX and SERCA activity compete for Ca 2+ during the [Ca 2+ ] i transient decline, and reduced SERCA function thus indirectly favors greater Ca 2+ efflux via the NCX (for reviews, see Guatimosim et al., 2002;Eisner et al., 2005;Bers, 2006b;Neef and Maier, 2013). We observed a similar SR Ca 2+ content in HF and CTRL SAN cells despite the slower [Ca 2+ ] i transient decline in HF SAN cells. This suggests a compensatory increase in Ca 2+ influx. I Ca,T and I Ca,L densities were not affected in HF SA node cells (Verkerk et al., 2003), but we cannot exclude that the longer diastolic depolarization phase in HF SAN cells (Verkerk et al., 2003) may result in a larger background Ca 2+ influx via L-type Ca 2+ channels (Verheijck et al., 1999). Alternatively, increased [Na + ] i as occurs during HF may promote Ca 2+ influx via reversed NCX activity (Despa et al., 2002). Indeed, by reconstruction of the I NCX , based on action potentials and simultaneously measured [Ca 2+ ] i transients, we found that the I NCX became outward, thus resulting in Ca 2+ influx, early during the AP (Figures 7D,E) under conditions of elevated [Na + ] i as occur during HF.

[Ca 2+ ] i transient during β-adrenergic and muscarinic receptor stimulation
We found that the β-adrenergic and muscarinic receptor stimulation was hardly affected by HF. In both CTRL and HF SAN cells, acetylcholine significantly decreased the frequency and [Ca 2+ ] i transients parameters, with exception of the time constant of decay which was significantly increased in both groups of cells (Figure 6). The ACh induced effects were largely similar in CTRL and HF SAN cells. Noradrenaline, on the other hand, significantly increased the frequency and most [Ca 2+ ] i transient parameters in both CTRL and HF SAN cells (Figure 5) (Verkerk et al., 2003). The slower [Ca 2+ ] i transient decay will promote Ca 2+ transport across the sarcolemma by the NCX (Bers et al., 2006), an electrogenic process that delivers inward current at diastolic potentials. The observed slower [Ca 2+ ] i transient decay may thus result in increased I NCX during the diastolic depolarization phase, thereby partially counteracting the previously observed decrease in I f -mediated inward current in HF (Verkerk et al., 2003). Indeed, reconstructed I NCX , based on the simultaneously recorded Ca 2+ transients and APs, demonstrated a slightly larger amplitude during the diastolic depolarization phase ( Figure 7C). However, when we incorporated an increased [Na + ] i , as frequently observed in ventricular myocytes, including those from our rabbit HF model (Baartscheer et al., 2003b), the I NCX in the diastolic depolarization phase is close to that of the CTRL SAN ( Figure 7E). Thus, the slower [Ca 2+ ] i transient decay in HF has likely a limited role in the slower intrinsic cycle length in HF SAN cells. The slope of late diastolic [Ca 2+ ] i elevation, associated with localized Ca 2+ releases (LCRs) or Ca 2+ sparks (Bogdanov et al., 2006;Maltsev et al., 2006;Joung et al., 2009Joung et al., , 2010van Borren et al., 2010), was unaffected under basal conditions, suggesting that the Ca 2+ clock is not the primary cause of the slower intrinsic cycle length in HF SAN cells. Considering the importance of late diastolic [Ca 2+ ] i rise in setting pacemaker activity, however, the lower late diastolic [Ca 2+ ] i rise in HF SAN cells in presence of noradrenaline suggest that an impaired increase in intrinsic frequency under such conditions may be related to reduced Ca 2+ clock function.

LIMITATIONS
We used indo-1 to measure [Ca 2+ ] i , because this ratiometric indicator is less prone to cell contractions and loss of dye. However, indo-1 had an adverse effect on cell viability and rendered many smaller cells quiescent (Lancaster et al., 2004). We cannot rule out the possibility that indo-1 has affected the principal findings, but we assume that its effect is limited because the differences in spontaneous activity of CTRL and HF SAN cells in our present study are highly similar to the differences that we observed with perforated patch clamp measurements (Verkerk et al., 2003).
In our study, we used the complete SAN for cell isolation. However, the SAN is not homogeneous in its composition. This seems valid for membrane currents (Boyett et al., 2000), but also the expression of several calcium handling proteins varies across the node (Musa et al., 2002;Lancaster et al., 2004), all of which are expressed at a lower level in the center of the SAN compared with the periphery, although such findings are debated (Lyashkov et al., 2007). Cells from the center and periphery may differ in cell capacitance. We exclude cell location-dependency as an explanation for our principal findings because cell capacitance did not differ between control and HF.
We reconstructed I NCX based on simultaneous measurements of [Ca 2+ ] i and spontaneous action potentials as well as the [Na + ] i of 5 and 7.5 mM used in our pipette solution and found in our model of HF previously (Baartscheer et al., 2003b). Data on [Na + ] i in SAN cells are extremely sparse and we were not succesful in [Na + ] i measurements with benzofuran isophthalate (SBFI) ourselves. However, the used 5 mM [Na + ] i is close to the concentrations of 4.5 and 4.0 mM in rabbit multicellular SAN preparations and single SAN cells, respectively, found by Choi et al. (1999). In our reconstructions, we used the Lindblad et al. (1996) equations for I NCX . We could not make use of I NCX equations from more recent models, in particular the rabbit SAN cell models by Maltsev and Lakatta (2010b) and Severi et al. (2012), because these require data on the concentration of Ca 2+ in a sarcolemmal subspace rather than global [Ca 2+ ] i .

CONCLUSIONS
In our rabbit model of HF, SAN cells have reduced SERCA activity and reduced intrinsic frequency, both resulting in a slower Ca 2+ decay. The decreased [Ca 2+ ] i transient decay in HF SAN cells may result in slightly increased I NCX during the diastolic depolarization phase, but this effect is likely counteracted by HFinduced increase in [Na + ] i . Reduced late diastolic [Ca 2+ ] i rise during β-adrenergic stimulation may contribute to the impaired increase in frequency under this condition in HF SAN cells.