Direct ionic communication between electrically excitable cells was discovered in the early 20th century ^[1][2][1,2], but for many decades this form of cell–cell communication was thought to be a property of excitable cells only. Therefore, everyone was surprised when in the 1950s direct cell-to-cell communication was found to exist in the cells of virtually all tissue ^[3][4][5][6][3,4,5,6]. In the mid-1960s, convincing evidence that this form of cell coupling could be regulated down to complete electrical and metabolic cell–cell uncoupling was reported ^[3][7][8][9][3,7,8,9]. However, some hints of the existence of cell–cell uncoupling mechanisms originated almost a century earlier. Indeed, in 1877 Engelmann reported that damaged cardiac cells became independent from their neighboring cells as they died [10]. This phenomenon, called “healing over”, is now known as “cell-to-cell uncoupling”, a property of all coupled cells, mediated by the chemical gating mechanism of gap junction channels; rev. in ^{[11][12][13][14][15]}[11,12,13,14,15].

Nearly a century after Engelmann’s discovery [10], Jean Délèze reported that in cut cardiac fibers “healing over” only occurred in the presence of external Ca²⁺ [16], suggesting that the rise in intracellular calcium concentration ([Ca²⁺_i]) caused by Ca²⁺ influx played a role in the regulation of gap-junctional communication. Soon after, the Ca²⁺-role in cell uncoupling was confirmed by evidence that cell–cell communication was lost in cells subjected to treatments that increase the [Ca²⁺]_i ^{[9][15][17][18][19][20][21][22][23][24][25]}[9,15,17,18,19,20,21,22,23,24,25].

The [Ca²⁺]_i effect on gating has been questioned for more than four decades. Early studies suggested that a [Ca²⁺]_i as high as 40–400 µM is needed ^[26][27][26,27]. In contrast, numerous more recent reports indicate that a much lower [Ca²⁺]_i, ranging from ~100 nM to low µM is effective. Data for the effectiveness of nanomolar [Ca²⁺]_i were first reported in a study ^[28][29][28,29] in which [Ca²⁺]_i of 251 nM (at pH 7.4), ~400 nM (at pH 7.0) and 2.5 µM (at pH = 6.5) were proven effective in uncoupling cardiac cell pairs, one of which was mechanically perforated to allow the influx of extracellular solutions well buffered for Ca²⁺ and H⁺. More recently, a treatment with ionomycin and gramicidin of arterially perfused rabbit papillary muscle uncoupled the cells at ~685 nM or greater [Ca²⁺]_i [30]; these data were confirmed in cells subjected to ischemia/reperfusion [30]. Low [Ca²⁺]_i proved effective on gating in many other cells, such as crayfish giant axons ^[31][32][31,32], rat lacrimal cells [33], Novikoff hepatoma cells ^[34][35][34,35], astrocytes ^[36][37][38][36,37,38], lens cultured cells [39], pancreatic β-cells [40], human fibroblasts [41], and Cx43-expressing cultured cells [42]. Mammalian pancreatic and lacrimal gland cells briefly uncoupled when secretion was stimulated by the application of acetylcholine or other secretagogues at concentrations below those required for maximal secretion ^[43][44][45][43,44,45], as well as by depolarization or cyclic nucleotide load ^[46][47][46,47]. In pancreatic acinar cells, even the treatment with secretagogues at concentrations capable of stimulating maximal secretory activity increases [Ca²⁺]_i only from 180 nM to 860 nM [48], further supporting the effectiveness on channel gating of nanomolar [Ca²⁺]_i.

WResearchers have studied, with Ca²⁺- and pH-sensitive microelectrodes, the relationship between junctional electrical resistance (Rj), [Ca²⁺]_i and [H⁺]_i in crayfish septate axons uncoupled by low pH_i [32]. Cytosolic acidification (pH_i = 6.3), caused by the application of Na⁺-acetate, increased [Ca²⁺]_i by approximately one order of magnitude, from resting values of 100–300 nM, and greatly increased Rj, indicating that that crayfish gap junctions are sensitive to low µM [Ca²⁺]_i [32]. The time course of Rj and [Ca²⁺]_i matched well, while that of Rj and [H⁺]_i did not [32].

In order to more accurately determine the [Ca²⁺]_i effective on gating, wresearchers studied Novikoff hepatoma cell pairs by double whole-cell clamp ^[34][35][34,35]. In these Cx43-expressing cells, Ca²⁺_i-gating sensitivity was tested by monitoring the decay of junctional conductance (Gj) at different [Ca²⁺]_i (buffered with BAPA), at pH_i = 7.2 or 6.1 (buffered with HEPES and MES, respectively). Channel gating was activated by [Ca²⁺]_i ranging from 500 nM to 1 µM, irrespective of pH_i [34]. With [Ca²⁺]_i = 0.5–1.0 µM, the Gj dropped to ~25% of the initial values with mean τ’s of 5.9 and 6.2 min, at pH_i = 6.1 and 7.2, respectively. With [Ca²⁺]_i = 3 µM, the cells uncoupled in <1 min (τ = ~20 s) [34]. The effectiveness of high nanomolar [Ca²⁺]_i on gating was confirmed in the same cells with brief (20 s) exposures to 20 µM arachidonic acid [35].

Similarly, a [Ca²⁺]_I < 0.5–1 µM blocked the cell–cell diffusion of Lucifer Yellow in chicken-lens-cultured cells [39], and nanomolar [Ca²⁺]_i drastically reduced the Gj in pancreatic β-cells, with a temperature drop from 37° to 30° and an external [Ca²⁺]_o rise from 2.56 mM to 7.56 mM [40]. Gating sensitivity to nM [Ca²⁺]_i was also reported in astrocytes injected with Lucifer Yellow and Ca²⁺ [36]. In these cells, nanomolar [Ca²⁺]_i prevented cell–cell dye transfer independently of pH_i; the dye transfer was blocked by [Ca²⁺]_i ranging from 150–600 nM [36]. Consistent with these findings is a report that the addition of 20 mM of BAPTA to the patch pipette solutions substantially improves coupling between astrocytes [37], which indicates that gating may even be sensitive to basal [Ca²⁺]_i. Dye coupling was also blocked in cultured astrocytes treated with ionomycin, which increased the [Ca²⁺]_i to 500 nM [38], and similar values were reported in lens-cultured cells [49]. In murine Neuro-2a cells (N2a) expressing Cx43, ionomycin treatment increased the Ca²⁺-influx and reduced the Gj by 95% [42], as the [Ca²⁺]_i increased from ~80 to ~250 nM. All of these data confirm the idea that Ca²⁺_i is a fine modulator of gap-junctional coupling.

As gap junction proteins do not have highly sensitive intracellular Ca²⁺-binding sites, the data described in the previous chapter strongly suggest that Ca²⁺_i affects gating via an intermediate component. Indeed, since the early eighties wresearchers have proposed calmodulin (CaM) as the intermediate of the Ca²⁺ gating effect; rev. in: ^[50][51][52][50,51,52].

In 1981, Johnston and Ramón reported that crayfish giant axons lose their cell–cell gating sensitivity to increased Ca²⁺_i and/or decreased pH_i when they are internally perfused [53]. Their data, confirmed by Arellano and coworkers [54], induced them to suggest that a soluble intermediate mediates the Ca²⁺/H⁺-induced cell uncoupling [53]. In the same year, wresearchers first suggested CaM as the soluble intermediate of Ca²⁺_i-induced gating ^[55][56][55,56]. Our idea was also supported by evidence for CaM binding to the gap junction protein connexin32 (Cx32) and to gap junction fragments from crayfish hepatopancreas ^[57][58][57,58].

In 1988, Arellano and coworkers provided convincing evidence that CaM is in fact the soluble intermediate that had been washed out by the internal perfusion of crayfish axons [54] because when the lateral giant axons were internally perfused with Ca-CaM (pCa 5.5; CaM + SIS-B), the Rj increased from the control values of ~60 kΩ to 500–600 kΩ in ~60 min (Figure 1). In contrast, the axons perfused either with CaM in low Ca²⁺ solutions (pCa > 7; CaM + SIS-A), with CaM-free high Ca²⁺ solutions (pCa 5.5; SIS-B) or with Ca-free solutions (SIS), maintained the Rj at control levels during the 60 min perfusion time (Figure 1). Figure 2 schematically summarizes the results of Arellano and coworkers ^[54][59][54,59]. The same results were reported with either only one axonal segment perfused (Figure 2) or with both segments perfused. Significantly, while 20 min of the internal perfusion of one axon segment with 1 mM Ca²⁺ in the absence of CaM did not change the Rj, the subsequent addition of Na⁺-acetate to the external solution, while maintaining the same internal solution, increased the Rj to ~400 kΩ [54]. This is very significant for two reasons: first, it proves that the septum was not damaged by the SIS perfusion; second, it proves that the uncoupling effect of acetate on the intact axon segment is not just due to an increase in [H⁺]_i [59], but rather to an acetate-induced rise in [Ca²⁺]_i resulting from the drop in pH_i (Figure 2), as also reported by us with different methods [32] (see in the previous).

Crayfish express innexins rather that connexins, but innexins are very similar to connexins and contain CaM-binding sites. In crayfish giant axons both innexin-1 and innexin-2 are expressed [60]. Innexin-1 and -2 contain CaM-binding sites at the CT and CL2 domains (Figure 3). The CaM-binding prediction to the CT and CL2 domains of these innexins were identified by means of a computer program developed at the University of Toronto (http://calcium.uhnres.utoronto.ca/ctdb/ctdb/sequence.html, accessed on 28 November 2021. Copyright © 2021 Ikura Lab, Ontario Cancer Institute. All Rights Reserved).

In 1986, Arellano and coworkers also confirmed their earlier data of gating insensitivity to H⁺_i [59]. In this study, a glass capillary was inserted into one of the axons and one side of the junction was perfused with solutions of pH 7 or 6 (Figure 2A), while monitoring the Rj. Significantly, the Rj remained unchanged when the pH of the perfusate was lowered from 7.1 to 6.0 [59]. WResearchers have confirmed the absence of a direct effect of low pH_i on gating in crayfish axons [32], Xenopus oocyte pairs [61], and Novikoff hepatoma cell pairs [34]. In the Novikoff cells, wresearchers monitored the Gj at different pCa (9, 6.9, 6.3, 6, and 5.5; buffered with BAPTA) and pH_i (7.2 or 6.1; buffered with HEPES an MES, respectively). No significant difference in the Gj was observed between pH_i 7.2 and 6.1 as long as the [Ca²⁺]_i was carefully buffered with BAPTA [34].

In the four decades that followed our reports of the early eighties ^[55][56][62][55,56,62], CaM participation in channel gating has been confirmed by multiple data generated by a variety of experimental procedures that include: treatment with CaM blockers ^{[42][49][55][56][62][63][64][65][66][67]}[42,49,55,56,62,63,64,65,66,67], inhibition of CaM expression ^[68][69][70][68,69,70], overexpression of a CaM mutant (CaMCC) with higher Ca²⁺ sensitivity ^[71][72][71,72], colocalization of CaM and gap junctions by immune-fluorescence microscopy ^[71][72][71,72] (Figure 4), intracellular perfusion of crayfish axons with CaM-containing solutions [54], and in vitro testing of CaM binding to connexins ^{[57][58][71][72]}[57,58,71,72] and synthetic connexin peptides mimicking CaM-binding sites of various connexins ^{[42][52][73][74][75][76][77][78][79][80]}[42,52,73,74,75,76,77,78,79,80]; rev. in: ^[50][51][52][50,51,52].

After our early evidence for a CaM role in gap junction channel function ^{[55][62][63][81]}[55,62,63,81], numerous other membrane channels have been found to directly involve CaM in their gating mechanisms. Indeed, there are an increasing number of channels regulated by CaM. In addition to connexins, they include: voltage-gated calcium (VGCC, CaV) channels, sodium (VGSC, NaV) channels, potassium channels (VGPC, KV), small conductance calcium-activated K⁺ channels (SK), inwardly rectifying potassium channels (Kir, IRK), cyclic nucleotide-gated channels (CNG), ryanodine receptors (RyR), and transient receptor potential channels (TRP), rev. in: ^[82][83][82,83], as well as the water channel aquaporin-0 AQP0), also known as the eye lens protein MIP26 ^{[84][85][86][87][88][89][90]}[84,85,86,87,88,89,90].

CaM is an acidic protein of 148 amino acids, whose sequence is very well preserved from plants to mammals. It is a dumbbell shaped protein, ~65 Å long, made of two fairly spherical lobes of ~35 × 25 Å in size, called the N-lobe and the C-lobe [91]. A short NH₂-terminus is followed by the N-lobe, which is linked to the C-lobe by a flexible amino acid chain. Each of the two lobes has two domains, known as EF-hands [92], which bind Ca²⁺ with nanomolar affinity. The Ca²⁺ affinity of the C-lobe is greater than that of the N-lobe by approximately one order of magnitude. Ca²⁺-binding to Ca²⁺-free CaM (apo-CaM) induces conformational changes that unmask a hydrophobic pocket in each lobe. Ca²⁺-CaM (holo-CaM) interacts with a receptor domain, usually made of a basic amphiphilic alpha-helix, by binding to it hydrophobically and electrostatically.

In 2000, we prthe coposed a CaM-based “cork-type” mechanism of gap junction chemical gating ^[112]. This “cork” model envisions a physical obstruction of the cytoplasmic mouth of the channel by a CaM lobe ^{[19][50][112][113]}, probably combined with conformational changes in the connenexins, caused by Ca²⁺-CaM binding to the gating site. The model is based on numerous findings that suggest a direct CaM role in gating; rev. in ^{[50][51][52][113]}[50,51,52,113]. Experimental evidence indicates that the chemical/slow gate is a sizable, negatively-charged particle, likely to be a CaM lobe ^[69][114][69,114]. There are many reasons why wresearchers think that CaM is the most likely gating candidate. [Ca²⁺]_i in the high nM to low μM values activates chemical gating; rev. in ^[19][52][19,52]. In view of the fact that the cytoplasmic domains of connexins do not contain sequences able to bind Ca²⁺ at such low concentrations, the effect of Ca²⁺_i on the channel gating is most likely mediated by a CaM-like protein, most likely CaM itself. Indeed, CaM binds to connexins ^{[52][57][58][71][72][77][115]}[52,57,58,71,72,77,115] which in fact have CaM-binding sites; most of the sites are at the second half of the cytoplasmic loop (CL2), but some are also at the NH₂-terminus (NT-site) and at the NH₂-end of the COOH-terminus (CT1); rev. in ^{[19][52][75][113]}[19,52,75,113]. Most relevant for gating are likely to be the CL2 and NT sites ^{[52][73][74][75][77][108][109]}[52,73,74,75,77,108,109]. Peptides mimicking the CaM-binding site sequences located at CL2, NT, and CT1 of several connexins bind Ca²⁺-CaM with high affinity ^{[42][52][73][74][75][76][78][79][80][99][103]}[42,52,73,74,75,76,78,79,80,99,103]. Most important is the binding of CaM to the CL2 domain, which has been experimentally confirmed by Jenny Yang’s team for Cx43 [80], Cx44 [79] Cx50 [78], and Cx45 [77] and by Katalin Török’s team for Cx32, Cx35, Cx45, and Cx57 ^[73][74][73,74]. CaM and connexins co-localize at gap junctions (Figure 4) and intracellular sites ^{[71][72][77][104][116]}[71,72,77,104,116]. Recently, the direct binding of CaM to Cx45 has been visualized in living cells by Bioluminescence Resonance Energy Transfer (BRET) [77]; the interaction of CaM and Cx45 was Ca²⁺-dependent and prevented by W7; the CL2’s CaM binding site (res. 164–186) was confirmed by a study reporting its high-affinity interaction (K_D = ~5 nM) with a peptide matching the CL2 domain of Cx45’s CL2, tested with a fluorescence-labeled CaM [77]. On the other hand, however, another study provided evidence for both Ca²⁺-dependent and Ca²⁺-independent CaM-binding to the CL2 domains of Cx45, Cx32, Cx35, and Cx57 ^[73][74][73,74]. The Ca²⁺-independent binding of CaM to the CL2 domain ^[73][74][73,74] confirms earlier data suggesting that the CaM is anchored to the Cxs at normal [Ca²⁺]_i (~50 nM) ^{[69][71][72][77]}[69,71,72,77]. Each of the two negatively-charged CaM lobes is ~25 × 35 Å in size [91], which is the approximate size of the positively-charged cytoplasmic mouth (vestibule) of the channel ^{[117][118][119]}[117,118,119] (Figure 95). So, a CaM lobe would fit nicely in the mouth (vestibule) of the connexon (Figure 95B). Evidence from a three-dimensional electron density map of isolated gap junctions, which display crystalline (hexagonal) channel arrays (see in the following), studied by X-ray diffraction, proves that the channels are in a closed state as they are inaccessible to sucrose due to a blocking particle at both channel ends similar in size to a CaM lobe ^{[120][121][122]}[120,121,122] (see in the following). Significantly, in a double-whole-cell-clamp (single-channel) study the chemical/slow gate opens and closes fully and very slowly (transition time = ~10 ms) [123] and the open-to-closed channel transitions, and vice versa, often displayed fluctuations [123]. This further supports the idea that a large particle may transiently flicker in and out of the mouth of the channel before closing the channel completely [123].

W Researchers are proposing two types of CaM-mediated cork-gating mechanism: “Ca-CaM-cork” and “CaM-cork”. In the former, the gating involves Ca²⁺-induced CaM activation. In the latter, gating would occur without a [Ca²⁺]_i rise above the resting values and in most cases would require either a connexin mutation ^[69][114][69,114] or the application of large Vj gradients [124].

In the early 1980s, Makowski and coworkers described the structure of crystalline (hexagonal) gap junctions isolated from mouse liver in a high-resistance configuration (closed channels) by analyzing X-ray diffraction data at 18 Å resolution (Figure 184 and Figure 195B) ^[120][121][120,121]. The channels’ gated condition of these isolated gap junctions was proven by the evidence that the channels were impermeable to sucrose [120]; in their words: “Analysis of diffraction patterns from isolated gap junctions in 50% sucrose shows that the sucrose fills the extracellular gap but fails to enter the channel. It is possible that the channel is closed at both cytoplasmic surfaces, excluding sucrose. This suggests that the isolated junctions are in a high resistance state” [120]. Indeed, the three-dimensional map of the electron density demonstrated that the channels were blocked at both cytoplasmic ends by a small particle; in their words: “Its position blocking the channel suggests that it may comprise a gating structure responsible for the control of channel permeability, X-ray diffraction studies of junctions in varying concentrations of sucrose (Makowski et al. 1984a) [122] indicated that in these preparations the channel was closed to the penetration of sucrose and that a solvent region approximately 100 Å long and centered on the six fold axis remained free of sucrose” [121].

Significantly, the blocking particle, spherical in shape, is approximately 30–35 Ǻ in diameter [121] (Figure 184 and Figure 195B), which is remarkably similar in size to a CaM lobe (Figure 195A). Indeed, in their words: “The channel has a diameter of 20–30 Ǻ along most of its length but appears to narrow to a minimum diameter of about 15 Ǻ in the extracellular half of the bilayer. Both the sucrose results and the three-dimensional map are consistent with the idea that a structure located near the cytoplasmic surface of the membraned is blocking the channel in these preparations” [121]. These studies support our evidence that gap junctions with crystalline (hexagonal) channel arrays are in an uncoupled (gated) state (Ca-CaM locked gate, Figure 13, Figure 14, Figure 15 and Figure 16) ^{[11][17][18][127][128][129][130]} and suggest that a CaM lobe is gating the channel.

This article has reviewed four decades of data supporting the direct role of CaM in gap junction channel gating. The Ca-CaM-cork gating mechanism [113], proposed over two decades ago [112], is based on evidence from the effect of CaM inhibitors, the inhibition of CaM expression, the expression of a CaM mutant (CaMCC) with higher Ca²⁺-sensitivity, CaM-connexin co-localization at gap junctions, the presence of high-affinity CaM-binding sites in connexins, the expression of connexin mutants, the gating effect of repeated large Vj pulses, data at the single channel level, the recovery of lost gating competency by addition of Ca-CaM to internally perfused crayfish axons and, finally, X-ray diffraction data on isolated gap junction fragments.

One may ask: why is it important to understand in detail the gating mechanism of gap junction channels? It is important not only because cell–cell uncoupling is more than just a safety mechanism for protecting healthy cells from damaged neighbors (healing over), but because the gating sensitivity to [Ca²⁺]_i in the high nanomolar range indicates that the fine modulation of direct cell–cell communication provides cells with the means for regulating tissue homeostasis. In addition, and more importantly, the field should be encouraged to test the effect of recently discovered disease-causing CaM mutants on gap junction function [135].

Indeed, recent evidence of diseases caused by CaM mutations ^{[136][137][138][139][140][141]}[136,137,138,139,140,141] suggests the potential role of CaM mutants in diseases affecting gap junction function. Almost two dozen CaM mutations have been found to cause cardiac malfunctions, most of which occur in the CaM’s C-lobe, one in the N-lobe, and one in the linker between the C- and N-lobes. In most cases the electrocardiogram (ECG) demonstrates the presence of Long QT Syndrome (LQTS), a change that affects the electrical activity of the heart, which is often associated with Catecholaminergic Polymorphic Ventricular Tachycardia (CPVT) phenotype and Idiopathic Ventricular Fibrillation (IVF). CPVT patients manifest ventricular tachycardia that can lead to death by ventricular fibrillation. In most cases, cardiac malfunctions have been attributed to the effect of CaM mutations on the ryanodine receptor (RyR2) and the cardiac L-type voltage-gated Ca²⁺ channel. However, other membrane channels, potential targets of CaM mutants, have also been suggested. Curiously, however, in spite of strong evidence for the direct CaM role in gap junction channel regulation, the potential consequences of these CaM mutants on direct cell–cell communication—a mechanism fundamental for the function of virtually all vertebrate and invertebrate organs—have not yet been addressed. Therefore, it is clear that future efforts should be aimed at testing the effect of these CaM mutants, and the future discovered CaM mutants, on gap junction mediated communication [135].