Due to the crucial role of Cx43 in the body and the heart in particular, mutants have a high risk of being eliminated at an early stage of fetal development (110–112), and thus, only a few reports link Cx43 mutations to human sudden cardiac death. Van Norstrand et al. identified two missense mutations in GJA1 resulting in E42K and S272P mutations in Cx43 in sudden infant death syndrome (113). Additionally, Cx43 mutations have been identified as the cause of oculodentodigital dysplasia, an autosomal dominant syndrome characterized by craniofacial and limb abnormalities (114). Arrhythmias are strikingly absent in this disease except for the I130T mutant (located in the Gap19 sequence; Figure 1) that forms leaky HCs and is linked to ventricular tachycardia (17). More recently, several studies demonstrated that increased HC opening is linked to ventricular arrhythmogenesis at the cellular level, with evidence coming from mouse models of Duchenne muscular dystrophy (115–118) and of arrhythmogenic right ventricular cardiomyopathy (119), as well from human data on Cx43 HC-linked excitability disturbances leading to triggered APs (tAPs) in heart failure (77). These experimental studies will be further discussed below.

Atrial cardiomyocytes express both Cx40 and Cx43, with Cx40 showing the largest expression. Several mutations as well as a rare polymorphism have been reported in the Cx40 GJA5 gene in human atrial fibrillation (AF) patients (120, 121), but little or nothing is known about their functional impact on Cx40-based HCs. As this Review focuses on Cx43 HCs, we include in the discussion below a recent report of Cx43 HC involvement in AF associated with a rare human mutation in the MYL4 gene (c.234delC) as a heritable cause of AF (122).

Duchenne muscular dystrophy. Duchenne muscular dystrophy (DMD) is an X-linked genetic disease resulting in muscle degeneration owing to the absence of the protein dystrophin in affected males; it also leads to dystrophic cardiomyopathy, which is a major cause of patient mortality (123). In mdx mice, a genetic DMD model expressing non-functional dystrophin (further referred to as Dmd-mdx), Gonzalez et al. demonstrated that arrhythmias induced by β-adrenergic stimulation with isoproterenol, as well as the ensuing animal mortality, were significantly decreased after HC inhibition with Gap26 or Gap19 (115). ECG recordings demonstrated increased premature ventricular contractions (PVCs) in Dmd-mdx relative to wild-type animals (Figure 3A). The increased PVC frequency resulted from increased HC opening, based on ethidium bromide (EtBr) dye uptake studies. Increased HC opening resulted in depolarization of ventricular cardiomyocytes, leading to “triggered activity,” i.e., above-threshold depolarization events manifested as tAPs (117), a cellular arrhythmic manifestation. Increased HC opening resulted from isoproterenol-induced nitric oxide formation, leading to nitrosylation of Cx43 at Cys271 (Figure 1) and subsequent lateralization of Cx43 (Figure 3A). Accordingly, biotinylation experiments showed an increased density of lateralized HCs (117).

Contribution of Cx43 HCs to ventricular and atrial arrhythmogenesis. Overview of signaling cascades leading to HC-related arrhythmic responses in ventricular (A–D) and atrial cardiomyocytes (E). Arrhythmogenic consequences of HC opening are given for the ventricular cascades A–E and are further explained in the main text. APD, action potential duration; cKO, conditional knockout; CM, cardiomyocyte; NAD+, nicotinamide adenine dinucleotide; ventr., ventricular.

In a further scrutiny of Dmd-mdx mice, Himelman et al. reported that dystrophin loss is associated with microtubule reorganization as hyperdense structures (118), which were found to enhance ROS production and oxidation of calmodulin kinase II (CaMKII) (Figure 3B). Oxidized CaMKII consequently provoked a decrease in Cx43 phosphorylation of Cx43 residues S325/S328/S330 (Figure 1) that set the stage for Cx43 remodeling, lateralization, and increased HC opening. Accordingly, the isoproterenol-triggered PVC frequency correlated well with EtBr HC dye uptake. Moreover, PVCs were absent in phosphomimetic Cx43 S325E/S328E/S330E mice, which also displayed decreased cardiomyopathy and increased survival. The QT time, reflecting the AP plateau phase and repolarization, was prolonged in Dmd-mdx mice while normal in S325E/S328E/S330E animals. The observed beneficial effect of phosphomimetic S325E/S328E/S330E is in line with the observation that phosphorylation of the Ser triplet is associated with proper localization of Cx43 to the IDs, while hypophosphorylation leads to lateralization (124–126) and increased probability of HC opening. In contrast, transgenic animals carrying a slightly different non-phosphorylatable mutant S325A/S328T/S330A showed decreased, not increased, HC activity as investigated by patch clamp approaches in mouse ventricular cardiomyocytes (127), which requires further scrutiny.

Arrhythmogenic right ventricular cardiomyopathy. Strict spatiotemporal control of [Ca2+]i dynamics is crucial for normal cardiomyocyte function, and recent evidence demonstrated that Cx43 HC opening triggers disturbed [Ca2+]i dynamics in a plakophilin-2–knockout (PKP2-KO) mouse model of arrhythmogenic right ventricular cardiomyopathy (ARVC) (119). PKP2 is a desmosomal ID protein involved in regulating intercellular junction assembly (128); it also controls gene transcription related to Ca2+ cycling and heart rhythm (129). Mutations in PKP2 are associated with the majority of genetic causes leading to ARVC (130). Ventricular cardiomyocytes isolated from inducible-PKP2-KO (PKP2-cKO) mice before the appearance of overt cardiomyopathy (i.e., still-intact tissue) have a widened intercellular space and display Cx43 remodeling and HC lateralization in the right but not the left ventricle (119) (Figure 3C). Interestingly, the phenotype is associated with ultrastructural appearance of linear plaque-like HC arrays (119, 129, 131). At the functional level, the phenotype demonstrated abnormal [Ca2+]i dynamics that were diminished in PKP2-cKO/Cx43+/– mice and suppressed by Gap19. The disturbed [Ca2+]i dynamics entailed increased sparking frequency and Ca2+ transient amplitudes, as well as increased sarcoplasmic reticulum (SR) and mitochondrial Ca2+ loading, culminating in early and delayed Ca2+ alterations as cellular arrhythmogenic signs. In addition to Gap19 inhibition of HCs, the increased Ca2+ spark frequency was also reduced by PKC inhibitors. PKP2-cKO cardiomyocytes further displayed increased phosphorylation of RyR2 at amino acid residue T2809 located in a domain involved in modulating Ca2+ gating. These cellular data were complemented by ECG studies in Langendorff-perfused hearts of PKP2-cKO mice, which, in contrast to wild-type animals, demonstrated long runs of ventricular tachycardia upon challenge with isoproterenol/rapid pacing (Figure 3C).

HC-linked arrhythmogenic mechanisms in ventricular cardiomyocytes. De Smet et al. and Lissoni et al. provided a detailed account of the mechanisms and consequences of Cx43 HC opening in response to challenge of mouse and pig left ventricular cardiomyocytes with caffeine or β-adrenergic agonists (70, 77). A typical caffeine response is represented in Figure 2E, demonstrating a slow transient inward current on which, superimposed, appear brief spike-like current events. The transient current is mediated by the electrogenic action of the Na+/Ca2+ exchange (NCX) transporter, which for every Ca2+ extruded brings three Na+ inside the cell (Figure 2G). In contrast, the spiking currents result from brief Cx43 HC opening events (first linked to Cx43 in 1990 by Pott and Mechmann, ref. 132). Other proposed channels, like sarcolemmal RyRs (133, 134), Panx1 channels (135), and the transient receptor potential channels TRPP2 (PKD2) and TRPP5 (PKD2L2) (136), were excluded (77). Atrial cardiomyocytes display similar spiking opening activity of Cx43 HCs (137). In ventricular cells, HC activity was found to vary between species, with mouse and human cardiomyocytes displaying the highest activity while activity in pig cells was lower (77). It is often assumed that the high HC opening activity in single dissociated cardiomyocytes is a by-product of cell dissociation or linked to the absence of GJs. However, macro-patch approaches mapping HC activity at cell ends showed that HC opening activity was larger in GJ-connected cardiomyocyte cell pairs than in single cardiomyocytes (77).

HC opening in ventricular cardiomyocytes at negative diastolic potential requires both caffeine stimulation and [Ca2+]i elevation; importantly, [Ca2+]i elevation by itself is not sufficient and neither is caffeine stimulation under conditions of strong [Ca2+]i buffering (70). It was found that Cx43 interacts with RyR2, and preventing this interaction using RyRHCIp peptide blocked HC activation. These findings suggest three prerequisites to open Cx43 HCs in ventricular cardiomyocytes: (a) activation of RyR2 by caffeine, adrenergic stress, or rapid pacing; (b) [Ca2+]i elevation; and (c) molecular interaction between RyR2 and Cx43 (Figure 2F).

Cx43 HC opening in ventricular cardiomyocytes can also be triggered by isoproterenol, and combined isoproterenol/rapid pacing stimulation results in very strong HC opening activity (77). These stimulation protocols are well known to increase SR Ca2+ loading, leading to increased Ca2+ sparking activity (138–140). However, Ca2+ release combined with RyR2-Cx43 interaction fulfills only two, not all three, of the conditions necessary for HC activation, but it seems obvious that the third condition of RyR2 activation is mediated by Ca2+ itself, as occurs in Ca2+-induced Ca2+ release (CICR).

How does HC opening lead to arrhythmogenic responses in ventricular cardiomyocytes? De Smet et al. demonstrated that HC opening frequently occurs in association with Ca2+ waves triggered by β-adrenergic activation combined with rapid pacing (77). Such Ca2+ waves typically lead to delayed afterdepolarizations (DADs), caused by the electrogenic consequences of NCX-mediated Ca2+ extrusion during the wave (entry of three Na+ for every Ca2+ extruded). HC opening results in an approximately 1.6-mV depolarization per open HC and another of about 1.3 mV resulting from electrogenic effects of NCX extrusion of HC-linked elevated microdomain [Ca2+]i (77) (Figure 2G and Figure 3D). As such, HC spiking activity occurring during a Ca2+ wave may thus increase DAD amplitude and tAP frequency. This was confirmed in human ventricular cardiomyocytes from failing hearts, where HC inhibition with TAT-Gap19 significantly decreased DAD amplitude and tAP frequency.

In addition to HC activity occurring during Ca2+ waves, HC opening was also found to precede the Ca2+ waves. In this case, there was a brief approximately 50-millisecond delay between HC opening activity and the start of the Ca2+ wave, suggesting a causal linkage (77). The majority of such waves started at cell ends where HC density is highest. Opening of HCs at these sites not only triggers depolarization but also leads to Ca2+ entry and [Ca2+]i elevation at HC-dyad microdomains estimated to attain 0.8 μM per HC for 1.0 mM extracellular Ca2+ (3.4 μM for 1.8 mM external Ca2+) (77). Such HC-initiated Ca2+ waves invariably led to DADs and tAPs in ventricular cardiomyocytes from failing human hearts (see De novo generation of DADs in Figure 3D), which were effectively suppressed by TAT-Gap19. These cellular data were confirmed in arterially perfused ventricular tissue wedges from failing human hearts demonstrating significantly increased DAD rates (~1 per 3 seconds) and tAPs (~1 per 7 seconds) following 2 Hz pacing/isoproterenol, compared with wedges from non-failing hearts (Figure 3D). DAD amplitudes and DAD/tAP frequencies were all significantly reduced by TAT-Gap19 and increased by TAT-CT9.

Collectively, these observations indicate that the opening of a few Cx43 HCs per cardiomyocyte (sometimes one to four coincident HC openings; Figure 2E) may be sufficient to produce arrhythmogenic responses in ventricular cells and tissues from failing human hearts. It is important to note here that HC opening will only result in depolarization when opening occurs at negative Vm, i.e., during diastole. HCs are poorly selective channels with a reversal potential of around 0 mV. Consequently, HC opening at positive Vm, e.g., during the AP plateau or early repolarization, will result in Vm changes in the direction of the 0-mV axis, i.e., opposite to the responses at negative Vm.

Figure 3D summarizes the signaling cascade starting from stressors (caffeine, isoproterenol, tachycardia) leading to increased HC opening in ventricular cardiomyocytes and its arrhythmic consequences in heart failure. Based on the evidence presented in the cascades of Figure 3, A–D, we conclude that arrhythmic HC contributions may occur through three distinct mechanisms summarized in Table 3 above.

Three levels of HC contributions to arrhythmogenic responses in ventricular cardiomyocytes

Cx43 HCs in AF. While most evidence for arrhythmogenic contributions of HCs comes from work on ventricular cells and tissues, some evidence also points to potential arrhythmogenic involvement of HCs in the atria. Ghazizadeh et al. (141) used an original approach starting from a rare human mutation in the MYL4 gene (Icelandic c.234delC), coding for myosin light chain-4 (MYL4) protein and representing one of the identified heritable causes of AF (122). They explored mutant MYL4 properties in atrial cells derived from human embryonic stem cells (hESCs) and in zebrafish\, in which the cmlc1 gene is a putative ortholog of human MYL4. To include more common AF-generating mechanisms, they applied database information from atrial patient biopsies and tested specific MYL4 mutations at the single-cell (hESC) and organ (zebrafish) levels to explore structural, functional, and transcriptional features predisposing to AF. The work pointed to involvement of retinoic acid signaling in cardiomyocyte cell polarity in mutant MYL4 (Figure 3E) and demonstrated that MYL4 associated with F-actin in atrial biopsies from human subjects in normal sinus rhythm, while MYL4 shifted to the sarcolemma in biopsies from AF patients. MYL4 also coimmunoprecipitated with Cx43, and Cx43 association with actin was decreased in MYL4–/– hESCs, which we suggest removes the high [Ca2+]i brake on HC activity (Figure 2D) and thus leads to increased HC opening. In terms of electrical and Ca2+ consequences, experiments in MYL4–/– zebrafish demonstrated increased AP duration (APD) and Ca2+ transient amplitudes in comparison with MYL4+/+ animals. Interestingly, NAD+, a compound known to be released through Cx43 HCs (142–144), was also elevated. These MYL4–/–-associated alterations in APD, Ca2+ transients, and NAD+ release were all suppressed by Gap19 inhibition of Cx43 HCs. It was furthermore shown that retinoic acid signaling enhanced PKC activity and inhibition of PKC mitigated the changes in APD, Ca2+ transients, and NAD+ release. Taken together, this study brings up three reasons for increased HC function in atrial cells: (a) Cx43 lateralization, (b) PKC activation, and (c) actomyosin-linked disturbances favoring HC opening when [Ca2+]i rises above 500 nM, as suggested here.

The multitude of Cx43 interactions with other proteins makes “location” of Cx43 a crucial determinant of its channel functions. In this regard, cardiomyocytes derived from induced pluripotent stem cells still lack the extreme degree of cardiomyocyte differentiation, in particular at the level of T-tubules, CICR, and the IDs where HCs reside in perinexal areas surrounding the GJ plaques (4, 58, 104–106, 145–149). Thus, conclusions regarding altered HC presence and function always need to be verified in experiments on primary cardiomyocytes.