Mitochondrial connexin43 and mitochondrial KATP channels modulate triggered arrhythmias in mouse ventricular muscle

Connexin43 (Cx43) exits as hemichannels in the inner mitochondrial membrane. We examined how mitochondrial Cx43 and mitochondrial KATP channels affect the occurrence of triggered arrhythmias. To generate cardiac-specific Cx43-deficient (cCx43−/−) mice, Cx43flox/flox mice were crossed with α-MHC (Myh6)-cre+/− mice. The resulting offspring, Cx43flox/flox/Myh6-cre+/− mice (cCx43−/− mice) and their littermates (cCx43+/+ mice), were used. Trabeculae were dissected from the right ventricles of mouse hearts. Cardiomyocytes were enzymatically isolated from the ventricles of mouse hearts. Force was measured with a strain gauge in trabeculae (22°C). To assess arrhythmia susceptibility, the minimal extracellular Ca2+ concentration ([Ca2+]o,min), at which arrhythmias were induced by electrical stimulation, was determined in trabeculae. ROS production was estimated with 2′,7′-dichlorofluorescein (DCF), mitochondrial membrane potential with tetramethylrhodamine methyl ester (TMRM), and Ca2+ spark frequency with fluo-4 and confocal microscopy in cardiomyocytes. ROS production within the mitochondria was estimated with MitoSoxRed and mitochondrial Ca2+ with rhod-2 in trabeculae. Diazoxide was used to activate mitochondrial KATP. Most of cCx43−/− mice died suddenly within 8 weeks. Cx43 was present in the inner mitochondrial membrane in cCx43+/+ mice but not in cCx43−/− mice. In cCx43−/− mice, the [Ca2+]o,min was lower, and Ca2+ spark frequency, the slope of DCF fluorescence intensity, MitoSoxRed fluorescence, and rhod-2 fluorescence were higher. TMRM fluorescence was more decreased in cCx43−/− mice. Most of these changes were suppressed by diazoxide. In addition, in cCx43−/− mice, antioxidant peptide SS-31 and N-acetyl-L-cysteine increased the [Ca2+]o,min. These results suggest that Cx43 deficiency activates Ca2+ leak from the SR, probably due to depolarization of mitochondrial membrane potential, an increase in mitochondrial Ca2+, and an increase in ROS production, thereby causing triggered arrhythmias, and that Cx43 hemichannel deficiency may be compensated by activation of mitochondrial KATP channels in mouse hearts.


Introduction
Connexin43 (Cx43) is a major connexin that forms gap junction (GJ) channels in cardiac ventricular muscle [44] and is present as unopposed/nonjunctional hemichannels in the sarcolemmal membrane [12,35] and as vesicles within the muscle [27]. Recently, Cx43 is also found in the inner mitochondrial membrane as hemichannels [4,32], and this mitochondrial Cx43 is associated with ADP-stimulated complex 1 respiration, ATP generation [3], mitochondrial K + influx, and the formation of reactive oxygen species (ROS) [39]. In addition, the presence of Cx43 is necessary for ischemic and pharmacological preconditioning [14,34,38] and is also necessary for the activation of mitochondrial K ATP channels [21,29] to reduce ventricular arrhythmias induced by ischemia [20]. It has not yet been established, however, how Cx43 is involved in arrhythmia susceptibility.
In past studies, a 50% reduction of sarcolemmal Cx43 expressed as GJs slowed ventricular conduction of excitation [13], and a reduction below 20% of control levels prominently slowed its ventricular conduction and caused reentrant arrhythmias [11]. Regarding Ca 2+ wave propagation, the expression of GJs determines the propagation velocity because Ca 2+ waves propagate along ventricular muscle by Ca 2+ -induced Ca 2+ release from the SR mediated by Ca 2+ diffusion through GJs [22]. We have reported that in rat trabeculae, the reduction of GJ communication using octanol and heptanol decreased the velocity of Ca 2+ waves [47]. In contrast, we have previously reported that blocking of Cx43 using carbenoxolone increased the velocity of Ca 2+ waves, causing triggered arrhythmias in rat trabeculae [23]. In addition, we have further reported that the increase in arrhythmias by carbenoxolone was suppressed after the addition of diazoxide, suggesting that Cx43 plays important roles in the occurrence of trigged arrhythmias under modulation of mitochondrial K ATP channels [23]. Nevertheless, it has not yet been established whether the block of mitochondrial Cx43 is actually involved in the increase in arrhythmias because we cannot eliminate the possibility that carbenoxolone may block channels other than Cx43 in the previous study [23].
Therefore, using cardiac-specific Cx43-deficient (cCx43 −/− ) mice, we investigated whether Cx43 deficiency can cause triggered arrhythmias due to changes in mitochondrial function, in addition to reentrant arrhythmias due to the reduction in GJ communication observed in Cx43 +/− mice [11,13]. In addition, we investigated whether mitochondrial K ATP channels are involved in the occurrence of triggered arrhythmias.

Statistics
All measurements were expressed as mean ± SEM. Statistical analysis was performed using ANOVA and a posthoc test (Tukey-Kramer) for multiple comparisons and a paired t-test for two-group comparisons when the data were normally distributed. The Wilcoxon signed-ranks test was used for two-group comparisons when the data were not normally distributed, unless otherwise mentioned. These analyses were performed using software for statistical analysis (Ekuseru-Toukei, Social Survey Research Information Co., Ltd, Tokyo, Japan). Values of P < 0.05 were considered to be significant.

Physical characteristics
cCx43 −/− mice died suddenly within 2 months after birth, and their percent survival was lower than that of cCx43 +/− and cCx43 +/+ mice (Fig. 1A). The heart and body weights of cCx43 −/− mice were not different from those of cCx43 +/+ mice (Fig. 1B).

Triggered arrhythmias in trabeculae
In the trabecula obtained from a cCx43 −/− mouse, electrical stimulation induced arrhythmia expressed as repeated contractions (red arrowheads in Fig. 1C) at 3 mmol/L [Ca 2+ ] o and higher than 3 mmol/L [Ca 2+ ] o , indicating that the [Ca 2+ ] o,min of the trabecula was 3 mmol/L. In contrast, in the trabecula obtained from a cCx43 +/+ mouse, electrical stimulation did not induce contractions even at 7 mmol/L [Ca 2+ ] o (Fig. 1C), indicating that the [Ca 2+ ] o,min of the trabecula was higher than 7 mmol/L and was defined as 8 mmol/L in the present study. The summary data in Fig. 1D show that the [Ca 2+ ] o,min in cCx43 −/− mice was lower than that in cCx43 +/+ mice, suggesting that the susceptibility to arrhythmia in cCx43 −/− mice is higher than that in cCx43 +/+ mice.
We have previously reported that in rat trabeculae, blocking of Cx43 hemichannels by carbenoxolone can increase the occurrence of triggered arrhythmias and that diazoxide, a K ATP channel opener, can suppress the occurrence [23]. Thus, the effect of diazoxide on the [Ca 2+ ] o,min in cCx43 −/− mice was examined. Diazoxide suppressed the arrhythmia induced by electrical stimulation at 3 mmol/L [Ca 2+ ] o in the cCx43 −/− trabecula ( Fig. 2A) and increased the [Ca 2+ ] o,min in cCx43 −/− mice (Fig. 2B), while it did not increase the [Ca 2+ ] o,min in cCx43 +/+ mice ( Figure S3AB), suggesting that activation of mitochondrial K ATP channels can decrease the occurrence of triggered arrhythmias in cCx43 −/− mice.  To determine how ROS production changes during electrical stimulation in cCx43 −/− mice, the slope of DCF fluorescence intensity during 1-and 0.5-Hz stimulation was then observed. The slope of DCF fluorescence during stimulation was higher in cCx43 −/− cardiomyocytes than in cCx43 +/+ cardiomyocytes (Figs. 3A, B, and C and S3CD) and was decreased after the addition of diazoxide especially in cCx43 −/− cardiomyocytes (Figs. 3D, E and S3CD). To further determine the amount of ROS production in cCx43 −/− mice, MitoSoxRed was used. As shown in Fig. 3F, the spatial loading pattern of MitoSoxRed was similar to that of MitoTracker Green in the trabecula, meaning that Mito-SoxRed was mainly loaded within the mitochondria. After the addition of H 2 O 2 , MitoSoxRed fluorescence increased and reached a plateau during 0.5-Hz electrical stimulation (Fig. 3G, H). The ratio of MitoSoxRed fluorescence before the addition of H 2 O 2 (Fl) to that after its addition (Fl H2O2 ) was calculated. The ratio in cCx43 −/− mice was higher than that in cCx43 +/+ mice (Fig. 3I), suggesting that the level of ROS production within the mitochondria was higher in cCx43 −/− mice. Taken together, these results suggest that in cCx43 −/− mice, ROS are produced at higher levels during electrical stimulation and play important roles in arrhythmia susceptibility and that activation of mitochondrial K ATP channels suppresses the increase in ROS production.

Roles of ROS in triggered arrhythmias
Mitochondrial membrane potential, mitochondrial Ca 2+ , and Ca 2+ sparks TMRM fluorescence was increased during 1-and 0.5-Hz electrical stimulation in cCx43 +/+ cardiomyocytes, while it was slightly decreased in cCx43 −/− cardiomyocytes (Fig. 4A, B, and C), meaning that Cx43 deficiency depolarizes mitochondrial membrane potential during electrical stimulation. Diazoxide increased TMRM fluorescence only in cCx43 −/− cardiomyocytes (Fig. 4D, E). Figure 5A shows that the loading pattern of rhod-2 was similar to that of MitoTracker Green in the trabecula, meaning that rhod-2 was mainly loaded within the mitochondria.
As the next step, Ca 2+ spark frequency was measured in isolated cardiomyocytes because Ca 2+ leak from the SR is deeply involved in the occurrence of delayed afterdepolarizations and triggered arrhythmias [43]. As shown in Fig. 6A, B, Ca 2+ spark frequency in cCx43 −/− cardiomyocytes was higher than that in cCx43 +/− cardiomyocytes, and the frequency in cCx43 +/− cardiomyocytes was higher than that in cCx43 +/+ cardiomyocytes. In addition, Ca 2+ spark frequency in cCx43 −/− cardiomyocytes was decreased after the addition of diazoxide and recovered after the addition of 5-HD (Fig. 6C). These results suggest that Cx43 activates Ca 2+ leak from the SR depending on the degree of its deficiency and that activation of mitochondrial K ATP channels compensates for Cx43 hemichannel deficiency.

Discussion
The present study investigated whether Cx43 hemichannels play some roles in the occurrence of triggered arrhythmias. To the best of our knowledge, this study shows for the first time that Cx43 deficiency activates Ca 2+ leak from the SR, probably due to depolarization of mitochondrial membrane potential, an increase in mitochondrial Ca 2+ , and an increase in ROS production, thereby causing triggered arrhythmias, and that Cx43 hemichannel deficiency may be compensated by activation of mitochondrial K ATP channels in mouse hearts, as discussed below.

Cx43 and arrhythmias
In past studies, a reduction of GJs in the sarcolemma slowed ventricular conduction of excitation and caused reentrant arrhythmias [11,13]. On the other hand, in our previous study, carbenoxolone reduced GJ permeability but nevertheless increased the velocity of Ca 2+ waves along  (Fl) to that after its addition (Fl H2O2 ). *P < 0.05 vs. Cx43 +/+ multicellular ventricular myocardium with an enhancement of delayed afterdepolarizations, causing triggered arrhythmias in trabeculae [23]. Furthermore, in the previous study, carbenoxolone increased Ca 2+ spark frequency even in isolated single cardiomyocytes that contained no GJ communication. It is possible, however, that block of the other channels except for Cx43 by carbenoxolone could be involved in arrhythmia susceptibility in the previous study. Thus, in the present study, we used cardiac-specific Cx43deficient mice to eliminate the effect of the channels other than Cx43 hemichannels.
In the present study, cCx43 −/− mice showed mitochondrial depolarization (Fig. 4A, B, and C) and an increase in mitochondrial Ca 2+ (Fig. 5B, C). In past studies, it was reported that mitochondrial depolarization played important roles in postischemic arrhythmias [1] and glutathione oxidation-induced arrhythmias [6] and that an increase in mitochondrial Ca 2+ increased ROS production [30]. Actually in the present study, ROS production was increased with an increase in mitochondrial Ca 2+ in cCx43 −/− mice (Figs. 3A, B, C, G, H, and I and S3CD), and antioxidant peptide SS-31 and NAC decreased arrhythmia susceptibility (Fig. 2C-F), consistent with the past study that mitochondria-targeted antioxidant decreases arrhythmias [40]. In addition, it is well known that an increase in ROS production activates Ca 2+ leak from the SR [16,46,48] and that a higher Ca 2+ spark frequency is deeply involved in the occurrence of triggered arrhythmias [43], suggesting that a higher Ca 2+ spark frequency in cCx43 −/− mice (Fig. 6A, B)  arrhythmia susceptibility (Fig. 1C, D) and higher mortality (Fig. 1A). Taken together, it is reasonable to assume that in cCx43 −/− mice, Ca 2+ leak from the SR is activated by an increase in ROS production, probably due to mitochondrial depolarization and an increase in mitochondrial Ca 2+ , thereby causing triggered arrhythmias.

Roles of K ATP channels
It has been reported that mitochondrial Cx43 is needed for diazoxide-induced cardioprotection [33] in ischemia-reperfusion injury [19,34,37], suggesting that mitochondrial Cx43 plays some important roles in the cardioprotection. In the present study, however, despite the deficiency of Cx43 hemichannels, diazoxide suppressed mitochondrial depolarization (Fig. 4D) and decreased mitochondrial Ca 2+ (Fig. 5E, F), ROS production (Figs. 3D and S3CD), Ca 2+ leak from the SR (Fig. 6C), and arrhythmia susceptibility ( Fig. 2A, B) in cCx43 −/− mice. It has been reported that diazoxide prevents mitochondrial depolarization and decreases ROS production induced by anoxia/reoxygenation [9], hypoxia/reoxygenation [28], and ischemia/reoxygenation [15] and that it reduces ROS production and arrhythmias during prolonged ischemia [20]. These results suggest that different from diazoxide-induced cardioprotection in ischemia-reperfusion injury, Cx43 deficiency may be compensated by activation of mitochondrial K ATP channels in cCx43 −/− mice, although we cannot determine whether the other channels except for mitochondrial K ATP channels work for the compensation.

Limitations
Transgenic mouse models deficient in some kind of protein would compensate for the deficiency. Also in the present study, it is possible that the effect of diazoxide may be enhanced by this kind of compensation. We assume, however, that the amount of compensation is limited because the results in the present study are almost consistent with those in the previous study using carbenoxolone [23]. Cx43 exists as hemichannels in the sarcolemma [12,35] and as vesicles within the muscle [27] in addition to its expression as hemichannels in the inner mitochondrial membrane. It has been reported that in plakophilin-2-deficient mice, Cx43 hemichannels in the sarcolemma play some roles in Ca 2+ spark frequency even in isolated single cardiomyocytes [17]. It is unlikely, however, that Cx43 hemichannels in the sarcolemma worked for the occurrence of arrhythmias   [45], which is lower than the [Ca 2+ ] o,min of cCx43 +/− and cCx43 +/+ mice (Fig. 1D). On the other hand, we assume that mitochondrial Cx43 hemichannels are open under physiological conditions because if they were closed, blocking of Cx43 by carbenoxolone [23] and Cx43 deficiency in the present study would not increase Ca 2+ spark frequency (Fig. 6B). The measurements were performed under unphysiological conditions, that is, at room temperature, with low frequency of electrical stimulation and [Ca 2+ ] o ranging from 0.2 to 7 mmol/L. Thus, caution is needed to interpret the results in the present study.