Structural basis for the selective permeability of channels made of communicating junction proteins

Abstract

The open state(s) of gap junction channels is evident from their permeation by small ions in response to an applied intercellular (transjunctional/transchannel) voltage gradient. That an open channel allows variable amounts of current to transit from cell-to-cell in the face of a constant intercellular voltage difference indicates channel open/closing can be complete or partial. The physiological significance of such open state options is, arguably, the main concern of junctional regulation. Because gap junctions are permeable to many substances, it is sensible to inquire whether and how each open state influences the intercellular diffusion of molecules as valuable as, but less readily detected than current-carrying ions. Presumably, structural changes perceived as shifts in channel conductivity would significantly alter the transjunctional diffusion of molecules whose limiting diameter approximates the pore’s limiting diameter. Moreover, changes in junctional permeability to some molecules might occur without evident changes in conductivity, either at macroscopic or single channel level. Open gap junction channels allow the exchange of cytoplasmic permeants between contacting cells by simple diffusion. The identity of such permeants, and the functional circumstances and consequences of their junctional exchange presently constitute the most urgent (and demanding) themes of the field. Here, we consider the necessity for regulating this exchange, the possible mechanism(s) and structural elements likely involved in such regulation, and how regulatory phenomena could be perceived as changes in chemical vs. electrical coupling; an overall reflection on our collective knowledge of junctional communication is then applied to suggest new avenues of research.

1. Introduction: Basic concepts

1.1. Permeability and selectivity

Permeability is the property of a material (or structure) to allow molecules through, as porous rock, soil or paper may let water pass. Substances that move through permeable structures are called permeants. The very nature (porosity) of permeable structures determines that some, but not all molecules go through; in other words, permeability is selective. This selectivity (or selective permeability) is well illustrated by ionic membrane channels that exhibit the property at a high degree. The selectivity of ionic channels is often described in terms of relative permeability to various permeants (e.g. the permeability ratios, P_X/P_K, of an inward rectifier potassium channel to Tl⁺, K⁺, Rb⁺ and Cs⁺ are 1.29, 1.0, 0.74 and 0.18, respectively [79]). That potassium channels are more permeable to K⁺ than practically any other biologically relevant ion is the reason why potassium channels are recognized as such, meaning that despite being permeable to several ions, they are selective for K⁺. A similar characterization has not been conducted for all gap junction channels.

In a population of membrane channels, permeability changes do not necessarily result from altered selectivity; hypothetically, permeability could change proportionally for all permeants, for instance, by simply reducing or increasing the number of available channels. At the level of individual channels, examples of changing permeability without effect on selectivity [13,39] and changing selectivity with variable consequences on permeability [1,39,91,157,163,167,192,196], are readily found. Suppose that channel permeability decreases by simple constriction of the pore to the point [1,91] where the biggest permeants are no longer able to traverse the channel; or conversely, a widening of the pore occurs that allows previously impermeant ions to become permeants. In neither case can the altered profile of permeability ratios be correctly described as “increased selectivity,” but rather should be described as changed channel selectivity. “Increased selectivity” for a particular molecule must refer to its ability to permeate the channel in comparison to other permeants; in the case of our exemplary potassium channel (whose relative Tl⁺/K⁺ permeability is 1.29), acquiring a Tl⁺/K⁺ permeability ratio of 0.5 would mean increased K⁺ selectivity, whether this is achieved by increasing K⁺ or decreasing Tl⁺ (absolute) permeability. If a Na⁺ channel, which is so much less permeable to K⁺ that it is practically “not permeable” to this ion, were to become a K⁺ channel, or vice versa ([92,163]; cf. [157]), this drastic change would be described as a vast increase in selectivity for the specific ion. Harder to imagine is how such profound selectivity changes could occur without at least some modification of the channel’s absolute permeability to each ion (for perspectives, see [1,92,93,97,174,192]).

How do these considerations affect our notion of gap junctions? For a start, it is, by now, obvious that gap junctions must be selective. The practical and less easy to answer questions are: What are they selective for, is selectivity regulated, and, if so, what are the mechanisms of regulation. Data addressing these fundamental questions are, although not exhaustively, reviewed next.

1.2. Gap junction channels are ionic conduits

Gap junctions are permeable structures comprising hydrophilic channels. Thus, their likely permeants are charged or polar molecules, the most conspicuous of which are current-carrying ions. The relative permeability of some gap junction channels for monovalent cations and anions has been investigated [12,208,209,213]. From these studies we can conclude that the selectivity of gap junctions for small ions of either charge is minor compared to the selectivity of Na⁺ and K⁺ channels. Whether or not the diffusion of, and the charge selectivity for, current-carrying ions is a relevant feature of junctional channels is not completely clear. In tissues whose function is not dependent on the transmission of electrical signals, it is difficult to envision that ionic communication through gap junction channels would be directly related to tissue function or failure. Nonetheless, it was recently postulated [113] that during embryonic development, as well as in tissue regeneration and tumor progression, the topical presence/absence of gap junctions is fundamental in the establishment of electrophoretic fields that could guide the movement of morphogens; more interesting, the morphogens themselves could be moving through gap junctions. The establishment of this kind of electrical field would not only require a continual source of current, but would benefit from a steepening of the field’s gradient across the developing tissue derived from low conductivity and strong ion selectivity of the gap junctions involved. Work on this mostly unexplored area might prove intellectually rewarding.

In the heart, where the consequences of decreased gap junction electrical communication can be readily detected, a decrease of open channels decreases intercellular current flow and the propagation of action potentials. Accordingly, in heterozygous Cx43^+/− mice, with significantly reduced ventricular expression of Cx43, electrophysiological signs of slow ventricular conduction, despite normal action potential amplitude, are seen [65]. The presence of conduction defects in Cx43^+/− mice was not confirmed in experiments using voltage sensitive dyes and high resolution epicardial video mapping recordings [130], but in mice with cardiac-restricted, conditional inactivation of Cx43, slower conduction velocity and spontaneous ventricular arrhythmias leading to sudden death at an early age [67], were documented. In the latter work extensive areas of myocardium devoid of Cx43 were seen, suggesting that discontinuity of cell-to-cell coupling, rather than overall low junctional permeability per se, was responsible for the observed symptoms. The consequences of abnormal organization of gap junctions were reviewed recently [58].

1.3. Gap junction channels are molecular sieves

While all gap junction channels are permeable to current-carrying ions, most of them showing channel conductance (γ_j) values roughly within an order of magnitude irrespective of their connexin composition, their permeability to other substances depends on their connexin composition and varies more widely. In an early study [51], the relative permeability of differently composed (Cx26, Cx31, Cx32, Cx37, Cx40, Cx43, Cx45 expressed in HeLa cells) gap junctions to LY (Lucifer Yellow: 443 (MW); 2⁻ (charge)), PI (Propidium iodide: 414, 2⁺), EB (Ethidium bromide: 314, 1⁺), DAPI (4,6 diamidino-2-phenyl-indol: 279, 1⁺) and Neurobiotin (N-2 [aminoethyl]-biotinamide: 287, 1⁺) was explored. The investigators counted neighboring cells receiving dye within 5 minutes of iontophoretic injection into a central cell. With the exception of Cx31 and Cx32 comprised junctions, which showed lower permeability to PI and EB, all junctions were well permeated by all probes. Assuming comparable numbers of open channels among groups, the data suggest the relative permeability of Cx43 junctions is LY>DAPI>EB>PI, whereas that of Cx40 junctions is DAPI≈LY≈PI>EB. Interestingly, Cx40 junctions seemed to display slightly higher permeability to positively charged neurobiotin than to negatively charged LY, while Cx43 junctions did not discriminate between these dyes. Junctions formed by Cx36, perhaps reflecting their small γ_j (~14pS), were permeable to neurobiotin but not LY [190]. Very recently, channel permeability to probes of different sizes and net charge was explored with simultaneous assessment of electrical coupling and cellto- cell fluorescent flux [87]. In this work, negatively charged LY, CF (carboxyfluorescein: 376, 2⁻) and Alx350 (Alexa350: 326, 1⁻) permeated gap junctions in the order Cx43 > Cx45 > Cx26 > Cx40; positively charged NBD (N,N,N-trimethyl-2-[methyl-(7-nitro-2,1,3-benzoxadiol-4-yl)amino]ethanaminium: 280, 1⁺) permeated all connexin channels explored at similar rates (Cx26 ≈ Cx43 ≈ Cx40 ≈ Cx45) whereas EB displayed the order Cx26 ≥ Cx43 ≈ Cx40 > Cx45. Compared with their permeability to K⁺, Cx43 and Cx45 showed similar permeability to NBD and Alx350, while Cx45 showed lower permeability to CF and LY than Cx43; more interesting, Cx40 and Cx26 permeability for NBD was ~3 times that for Alx350, despite both dyes having similar minor diameters. Clearly, regardless of the quantitative rigor of the approach, together these data support the recurring contention [162] that gap junction channels display connexin-specific permeabilities to different substances. If fluorescent probes stand as model permeants for naturally occurring cytoplasmic constituents, then gap junctions act as selective interfaces wherein, depending on the comprising connexin(s), the intercellular exchange of some molecules occurs faster than that of others.

Some words of caution. While binding of dye to cytoplasmic components or sequestration within organelles may occur for some probes and particularly in some preparations, which could obviously affect permeability measurements, legitimate efforts were made by the authors of the works cited here to minimize the impact of these artifacts. For the purpose of this review, all results are presented as interpreted by the authors. Also, it may go unperceived that electrical coupling is often evaluated in cell pairs with voltage clamp techniques, usually involving the invasion of the intracellular space and, plausibly, the dialysis of cytoplasmic (potentially regulatory) components; in comparison, many dye coupling studies were done on cell groups using microinjection techniques, which, while performed with very fine electrodes, may include electrical “ringing” of the cell membrane. It is not clear that junctional coupling is exactly the same between cell groups and pairs, or which method is more accurate or deleterious in the assessment of junctional permeability/conductance. Some of these problems are overcome by the simultaneous measurement of both parameters, perhaps the most attractive way, as done in the latter referenced study [87] and further expounded upon in section 4.1.

2. Determinants of junctional and channel permeability: gating, electrostatic interactions, phosphorylation

Clearly, the permeability of a junction and its comprising channels depends on whether those channels are gated open or close, and on the electrostatic environment of the permeation pathway, both of which could be regulated independently by phosphorylation. These possibilities are discussed, in turn, next.

2.1. Gating

The permeability of membrane ion channels is “mainly” determined by whether they are gated open or closed. When open, their permeation is a function of the electrochemical driving force for their preferred (selected) ion(s); their permeability is a function of pore shape and surface charge, which can be modified by rearrangements of surface charges [1,13,39,91–93,97,157,163,167,174,192,196]. Here we consider similar attributes of gap junction channels.

Gap junction channels display at least one closed configuration and, typically, more than one open configuration ([14,22,32,156,181,185,194,200,208,209]; cf. [194] and Table S1); when these open configurations differ in conductivity, the most conductive is referred to as the fully open state and less conductive configurations as sub-states. The transit between any two of these states is referred to as gating. Experimentally, gating is commonly induced by imposing a voltage gradient across the junction (V_j), and this is by far the fastest and clearest method to evince the dynamics of channel behavior. However, gap junction channels also gate in the absence of an applied voltage (see [30,116,125,173,229] and section 3). When all the channels in a gap junction gate closed, diffusion of permeants from cell to cell is halted. Thus, increasing the relative number of fully closed channels in a junction translates into progressively less junctional permeability to all permeants (uncoupling). Importantly, the impact on junctional permeability of gating to different sub-states is less predictable and likely depends on the nature of the permeants.

Not unexpectedly, the selective permeability of a channel in a sub-state configuration can differ from the “fully open” configuration, with some sub-states becoming impermeable to some permeants. Indeed, in an elegant study by Bukauskas et al. [21] the conductive residual state (a V_j-induced sub-state) of Cx43 channels was demonstrated to be impermeable to dyes that readily permeate the fully open channel. In their experiments, dye (Alexa 350 or EB) injected into one cell of a pair failed to diffuse across a highly conductive junction while the junction was subjected to a large V_j (90 mV). Recovery of intercellular dye diffusion occurred, essentially instantaneously, upon cessation of the V_j step as the channels recovered their full open configurations. This study remains the most elegant demonstration of a gating-induced change in the selective permeability of a gap junction. Clearly the residual state configuration of the channel is permeated by current-carrying ions; the question remains, however, as to whether other small molecules can also diffuse through [8]. Sustained V_j gradients of the magnitude required to gate junctional channels to their residual states are unlikely to be (broadly) present between cells in vivo, although smaller voltage gradients might occur at heterocellular junctions with very low coupling levels. Disappointingly, reliable voltage-independent strategies for stably inducing sub-state configurations have not been identified to date. But favorably, some site mutants can display significantly different frequencies for some vs. other open states when compared to wt junctions.

Less predictably, changes in the selective permeability of a junction or channel may occur without alteration of its conductivity or, perhaps more surprisingly, may increase despite a reduction of conductivity (Figure 1 and [49,75]). In an intriguing study [49] of junctional selectivity, transjunctional diffusion of cationic NBD through Cx43 junctions formed by rat insulinoma (Rin) and neonatal rat kidney (NRK) cell pairs was evaluated; the intercellular dye diffusion data were fit with a two compartment model and the resulting diffusion rate constant used as index of the junction’s dye permeability (P_j-NBD). The conductance (g_j) of the same junction was measured, and the ratio P_j-NBD/g_j (or NBD permselectivity) calculated for each junction. Unexpectedly, g_j was not a reliable predictor of P_j-NBD and, more surprisingly, the incidence of sub-state behavior (55–70 pS) appeared to be greatest in the junctions most permeable to NBD. Moreover, treatments designed to increase the frequency of the “fully open” state failed to induce highly permeable junctions. Although the Cx43 junctions were less well permeated by LY than NBD, the P_j-LY/g_j correlation was also non-linear. Concurrently, Eckert [45] reported very similar results for Cx43 and Cx46 junctions (formed by HeLa cells) and Cx43 junctions in BICR/M1Rk cells relative to calcein (Bis[N,N−bis(carboxymethyl)aminomethyl]fluorescein; 622, 4⁻) and LY diffusion. After exhaustive analysis similar to that done by Ek Vitorin [49], but also including corrections for cell volume and calculations of single (unit-conductance) channel permeance, Eckert also concluded that the “strictly linear correlation of ionic versus dye permeance … expected for a simple pore” was absent from these junctions. Accordingly, the author surmised that differential regulation of single channel permeability for larger molecules and for small inorganic ions could explain his results [45]. In a third unrelated study, a significant increase in permeability to LY without a concomitant change in macroscopic g_j or in the biophysical properties of the channels was reported for Cx43 in vascular smooth muscle cells from diabetic rats [17]. Taken together, these data suggest that as the share of channels in different states might vary between cell pairs, and because changes in permeability may occur without detectable alteration of channel conductance, junctional dye permeability and electrical conductance may not be linearly correlated, either at the macroscopic or individual channel level. Large channel conductance does not correlate well with high dye permeability across connexins [206–208] – indeed, the highly conductive Cx37 and Cx40 channels (~300 and ~200 pS, respectively) are poorly permeated by dyes, especially negatively charged dyes, that readily permeate many other gap junction channel types. To be sure, not even channels that could be expected to have similar structural characteristics necessarily display identical conductivity or permeability [2,15,201,227]. Although not well predicted by channel conductance, channel permeability may nonetheless be linked to electrical alterations beyond our present detection capabilities. For instance, changes in the channel’s surface charges may or may not have easily detected conductive consequences in these wide diameter channels, while nevertheless having profound effects on permeation by our experimental probes and other permeants.

Figure 1.

Two scenarios for regulated permselectivity. A, Decreasing g_j without changes in P_j results in increased P_j/g_j. B, Decreasing P_j without changes in g_j results in decreased P_j/g_j. Each columnar arrangement in the upper panels illustrate the proportional composition of junctional plaques, where channels can be ion and dye permeable (hexagonal pore), ion permeable but dye impermeable (here drawn as a pentagonal pore representing changes in diameter/electrostatic environment) or ion and dye-impermeable (grey cross-closed pore). Note that the represented dye (inset) will only fit one pore configuration, and that current-carrying ions will be excluded only from the fully closed pores. Lower panels illustrate the results of redistribution of the channels among these configurations, on the permselectivity of the junctions. Dye (charge) selectivity, which depends on one specific channel configuration, will not be affected by changes in permselectivity. Modified from [75].

A more linear relation between P_j and g_j has been reported for junctions (formed by HeLa cells) composed either of Cx43 or Cx40 [199], or the chicken orthologue of Cx50 [41]. Besides low P_j-LY values for both Cx43 (compared to that from Ek-Vitorin et al. [49] or Eckert [45]) and Cx40 junctions, Valiunas et al. [199] reported that heterotypic Cx40-Cx43 junctions displayed a P_j-LY value intermediate to those of the homomeric counterparts and their single channel records showed channels opening to either the smallest (residual) or the highest conductive state, but noticeably, not to intermediate values. Similarly, Dong et al. [41] found that the P_j-LY of Cx43 is low, yet higher than that of Cx45.6 junctions; to complicate matters still further, when using a series of negatively charged dyes with similar structure but increasing molecular weights (350, 570, 760 Da for Alexa350, Alexa488 and Alexa594, respectively), Cx45.6 junctions were only permeable to the two smaller dyes, yet more permeable to Alexa350 than Cx43; however, when the amino terminal of Cx43 substituted that of Cx45.6 (Cx45.6 -43N), these mutant junctions were more permeable to the two largest, and less permeable to the smallest, than Cx45.6 junctions. These confounding results are similar to those of another study [221] addressed next (section 2.2.) A remarkable finding of Dong et al. [41] pertains to channel gating: in their figure 3, junctional current from either connexin displayed an initial fast relaxation, but when the NT domain of Cx43 substituted that of Cx45.6 (mutant Cx45.6–43N), the initial fast phase was lost. Also remarkable was the γ_j of the mutant Cx45.6–43N, which resembles Cx43 more than Cx45.6 (their figure 4), suggesting that the NT modifies or forms part of the pore (cf. [120]), thereby establishing channel permeability and voltage gating. These studies will be re-addressed jointly with our own data below (section 2.2), where a possible explanation for the disparate observations will be advanced.

2.1 Summary

By directly allowing or denying pore access to possible permeants, channel gating regulates junctional permeability, reasonably, through the regulation of the number of open channels. However, channel gating may also yield alternate open configurations whose selectivities are likely different from that of the fully open channel. Excluding extreme manipulations [21], reliable (pharmacological) methods to induce shifts in channel behavior from multiple to a single sub-state configuration, in order to study the unique characteristics of the latter, are not currently available.

2.2. Electrostatic interactions

Weak electrostatic potentials associated with the pore of connexins were proposed to account for diverging ionic selectivities, in such a manner that the selectivity of any given connexin might not correlate with the measured channel conductance [207]; explicitly, amino acids within or near the pore were postulated to alter the selectivity and conductance of connexin-specific channels by providing precise electrostatic field effects that interact with candidate permeants. Consequently, changes in channel permeability would not derive from a direct (detectable) gating mechanism, but from “variations in primary amino acid sequence and the resulting structure of the pore-forming region of the connexins”. Support for this idea comes from several other studies [41,63,140,195,230]. Of great interest is the work of Weber et al. [221] who used the Alexa probe series to assess the permeability of several connexins in paired frog oocytes. Because most of their calculated absolute permeabilities were ~2 orders of magnitude higher than would be predicted by pure hindered pore diffusion, the authors proposed (electrostatic) affinity between the probes and the pore, which would create a favorable environment within the pore and consequently elevate the concentration and enhance the flux of these permeants. An implication of these studies is that different connexins display different electrostatic potentials to select for specific molecular structure in permeants (cf. [41]). This possibility was further explored using labeled metabolites and Cx43 and Cx32 comprised junctions [60]; whereas Cx32 junctions were more permeable to calcein than Cx43 junctions, the latter were more permeable to ADP and ATP. In addition, in a clever patch-clamp study where junctional flux of cAMP was indirectly measured by its effect on the current of a cyclic nucleotide-modulated membrane channel expressed in the recipient cell of a pair, Kanaporis et al. [88] found that Cx43 is more permeable to cAMP and LY than either Cx26 or Cx40; the authors adhere to the idea that charges near the pore mouth could change selectivity without changing channel diameter, and propose that channel length could modify unitary conductance without parallel changes in permeability to big molecules. These data complement previous work [61] showing that adenosine permeates Cx32 channels better (~12-fold) than Cx43 channels, but that Cx43 channels are much better permeated by AMP, ADP (~8-fold each) and ATP (>300-fold!): seemingly, increasing permeability matching increasing phosphate (negative charge?) content. Does this indicate that Cx43 channels are selective for negatively charged molecules? In our laboratory, we have found no charge selectivity in Cx43 junctions, as will be discussed next.

The studies of Weber et al. [221] stand in contrast to those of Heyman and Burt [75], wherein it was concluded that the dye permeability of Cx43 channels can be described with a hindered diffusion model. In this work, ionic and dye permeability of Cx43 junctions (formed in mammalian cell lines rather than frog oocytes) were explored using patch clamp recordings of single channels and junctions with pipette salt substitutions, dual dye injection and simultaneous measurement of transjunctional dye diffusion; dye data were analyzed with a two-compartment model of intercellular diffusion. According to these studies Cx43 gap junctions are size-selective, but not charge-selective, and their relative permeance sequence of dyes (NBD~Alexa350>LY>Alexa488>>Alexa594) and conductance sequence of ions (KCl>TEACl~K-glutamate) could be explained by diffusion through an aqueous pore “devoid of significant electrostatic influences and reasonably uniform in geometry along its length.” Accordingly, previously reported discrepancies between dye permeability and conductance of Cx43 junctions [49] were explained by the presence of a variable number of channels that were conductive, but either dyepermeable or dye-impermeable (Figure 1) [75].

It is important to consider that gating, as with other ionic channels, is bound to modify the interaction of pore and permeant, irrespective of the role that electrostatic interactions play in channel permeability. If electrostatic interactions do take place [221], gating might change the environment near the pore mouth; if permeability depends exclusively on pore size [75], gating might decrease pore access by a simpler (constriction/obstruction) mechanism.

2.2 Summary

Closed channels are not permeable. The permeability of open channels possibly depends on interactions between charges in/near the pore and of the candidate permeants, as well as on the relative size of the latter. At least for some connexin channels, electrostatic interactions between pore and permeant may improve diffusion of permeants [135] whose limiting diameter approximates the pore’s. Reasonably, the adoption of partially open (partially closed) channel configurations could alter this porepermeant interplay, leading to robust modifications of junctional permeability for a given permeant. Current-carrying ions, by reason of their sizes may be less affected by changes in the pore’s surface charges or limiting diameter, such that discrepancies between junctional conductance and permeability could be expected.

2.3. Phosphorylation

The role of phosphorylation in the regulation of gap junction communication has been often reviewed (for instance [74,85,95,101,103,104,108,122,126,144,178,179,220]); our comments here are germane to permeability changes, if any, produced by kinase or phosphatase activity. We focus exclusively on Cx43 in this discussion as it is the most widely studied. For limited information on other connexin isoforms, we refer the reader to [24,26,71,72,76,100,103,115,134,161,193,204,205,212,226].

Cx43 has sixteen serine, one threonine and five tyrosine residues identified by phosphorylation consensus site prediction programs as probable targets for phosphyorylation. At least 8 of the serines and two of the tyrosines have been confirmed as targets. The kinases mediating these events include PKC, MAPK, cdc, CK1, Src and PKA. Some of these kinases/phosphorylation events affect junctional permeability [104], although not in a consistent fashion. Thus, decreased P_j-LY was reported following activation of protein kinase C (PKC) [81,90,99,184] and PKC depletion or inhibition resulted in increased or restored P_j-LY [19,99]; however, in some settings PKC inhibition can result in decreased P_j-LY [54]. Similarly, activation of mitogen-activated protein kinase (MAPK) decreases P_j-LY [81,90,99,184], but activation of MAPK alone is not always sufficient to increase Cx43 phosphorylation or decrease P_j-LY [82]. Phosphorylation of Cx43 decreases during Ca-induced uncoupling in chicken lens [34]. Cx43 is also phosphorylated by p34cdc2/cyclin B, a kinase active in mitosis when junctional communication decreases [89]. At least some of these changes in junctional communication might not directly or exclusively derive from changes in channel permeability.

Solid evidence for phosphorylation altering channel permeability comes from mutational analyses. Thus, alanine substitution of serines 325, 328 and 330 of Cx43 (Cx43TM) delayed the development of detectable electrical coupling, and reduced average macroscopic g_j, incidence of fully open channels and P_j-LY [102]; these serine residues are substrate for caseine kinase 1 (CK1), which also regulates gap junction assembly. Although the question arises as to how these mutations induce subconductive states, these data indicate that phosphorylation at these sites is needed for the channel to open fully. Interestingly, wild type Cx43 phosphorylated on these sites remains, though at reduced levels, at the intercalated disks of ischemic myocardium when protein dephosphorylated at these sites has migrated to lateral borders of the myocytes.

PKC activation increases the frequency of subconductive Cx43 channels [127], most likely by direct phosphorylation the connexin protein on serine 368 [105]. Solan et al. [180] showed that phosphorylation/dephosphorylation at S365 regulates availability of S368 for phosphorylation: in normal conditions, either or none of these residues may be phosphorylated, but not both simultaneously. Thus, Cx43 is normally phosphorylated at S365 in the heart; it is dephosphorylated at this site during hypoxia, as the protein moves from the intercalated disks to the lateral membranes; phosphorylation at S368 increases in hypoxia and depends on S365 being dephosphorylated. While the γ_j values of substitution mutants of S365 have not been reported, S365A and S365D junctions are well permeated by NBD and Alexa350 without apparent change in charge selectivity (our unpublished experiments). Interestingly, as in the case of S325, S328 and S330 phosphorylation [102], Cx43 phosphorylated at S368 by PKC during ischemia remains at the intercalated disks, at increased levels compared with the non-ischemic tissue [49].

MAPK phosphorylates residues S255, S279 and S282 [219] of Cx43 and decreases P_j-LY and P_j-NBD, as well as macroscopic g_j after EGF and lysophosphatidic acid stimulation [33,109,218]. Similar decreases of dye and electrical coupling followed activation of the protein tyrosine kinase pp60^v-src, which phosphorylates residues Y247 and Y265 of Cx43 [114,186]. Interestingly, the reduction in dye permeability caused by both MAPK and v-SRC is more extensive than the reduction in electrical coupling, and occurs without detectable changes in channel amplitude, suggesting alterations in channel open probability (P_o) or in channel selectivity [33].

In Cx43 hemichannel reconstitution studies, MAPK [94] and PKC [9] were shown to directly and reversibly reduce permeability to LY and sucrose (M_r ~ 342). With this reconstitution system, it was also shown that all connexins of the hexamer had to be phosphorylated by PKC to abolish permeability to sucrose; however, despite complete phosphorylation (at S368), permeability to the small hydrophilic ethyleneglycol (M_r ~ 62) remained [8]. It is unknown whether this “sizable diameter” represents the residual state of Cx43 channels, but as decreased γ_j has long been associated with PKC activation [125,127], these results suggest that PKC can strongly change dye coupling without a consistently parallel effect on electrical communication, as our published [49] and unpublished observations indicate.

Another phosphorylation site (S306) involved in the regulation of channel permeability was recently reported [158]; in this work, mutant S306A showed slightly smaller channels (95 vs. 119pS) than wt Cx43, as well as reduced g_j in the presence of increased plaque size.

2.3 Summary

From a hypothetical viewpoint, and in face of the above data, phosphorylation can affect junctional permeability by i) changing channel number, ii) triggering or modulating channel gating or iii) modifying electrostatic fields at the orifice of or within the channel pore. Aside from the factual mechanism, and because phosphorylation occurs mainly at the carboxyl terminal domain (“tail”), it is reasonable to imagine that regulation of channel permeability by phosphorylation requires a step whereby the tail or part of it comes to reside near the pore opening or, alternatively, that the tail allosterically transduces signals to, and causes conformational changes in, the pore forming regions. In summary, direct phosphorylation of connexin proteins is a major regulator of junctional permeability, although mechanistic models for how phosphorylation alters channel structure remain speculative. In addition, and beyond the scope of the present review, phosphorylation-induced regulation of junctional permeability may also be indirect [38].

3. Gating elements

3.1. Multiple gating parts?

As integral proteins, connexins have four transmembrane domains (TM1–TM4), two extracellular loops (E1–E2), one intracellular loop (CL) and the intracellular amino and carboxyl terminal domains (NT and CT respectively). All regions of the connexin molecule seem involved in the generation of functional channels (Table S2), whether by taking part in the sorting of the protein to the cell membrane, in the formation of the hydrophilic pore, or in the gating behavior of the junctions. While functional irregularities can be detected on channels with mutations on several regions of the connexin protein, no functional effects have been found for some naturally occurring mutations (e.g. see Cx32 and Erythrokeratoderma variabilis on Table S2); however, the strong link of those “silent” mutations to evident disease asserts that some function of the connexin, if not channel function, is compromised by the mutations. Here, we focus on reports that link domains of the connexin proteins to demonstrable electrical or chemical permeability changes in working gap junctions.

The earliest gating model of gap junction channels derived from electron microscopy studies [197] and postulated that all connexin units in a connexon slide against each other (iris-like) to decrease their normal inclination; the concerted movement results in a clockwise rotation at the cytoplasmic face that closes the pore mouth like a camera shutter. This mechanism implied two channel states (open and closed) corresponding to the two alternate subunit configurations, and was thought to explain the Ca²⁺-induced decrease in permeability [198]. More recently, it was hypothesized that each connexin subunit autonomously responds to environmental alterations, and that channel sub-states arise from conformational changes of single or multiple subunits [139]. This model, developed to explain channel gating triggered by V_j differences, allowed for multiple open states. Perspective on some anatomical aspects of Ca²⁺-triggered gating can be gained by consulting recent papers [3,62,131,189,191], some of which support, at least in part, the shutter model of channel closure. What follows is a more detailed, roughly progressive exposition of relevant evidence for the involvement of multiple connexin regions in the regulation of channel permeability; we expect that a chronological perspective will be useful to understand the development of the current status.

Junctional currents of Cx32 and Cx26 expressed in oocytes show inactivation (relaxation) during an applied V_j pulse [11]; in these early reports two different inactivation time constants were recognized, and two types of gating proposed: fast and slow. To determine which regions of these connexins conferred fast vs. slow gating behavior, connexin chimeras were constructed that swapped regions between Cx32 and Cx26, and gating behavior of these chimeras characterized in oocytes. Based on shifts along the voltage-axis of the conductance-voltage relationship of steady state g_j resulting from swaps of connexin loops, specifically the E1 and CL domains but also the amino terminal half of TM3 [165], it was proposed that E1 contributed to both fast and slow gating mechanisms and to the junctional rectification observed when Cx32 and Cx26 were paired in heterotypic fashion [166].

Verselis et al. [210] proposed that two residues at the M1/E1 border, as well as the 2^nd residue of the NT formed part of a voltage sensor. Similar results (residues between NT and E1 strongly influencing V_J-gating) came from chimeras of multiple connexins, particularly Cx32 and Cx46 [223]. Furthermore, single amino acid substitutions on E1 (C64) or on the neighboring TM1 (M34) and TM2 (R75) regions [141] eliminated Cx26 channels permeability to sulforhodamine (MW 559, 1⁺). Shibayama et al. [176] more recently determined that single point mutations found in patients with oculodentodigital dysplasia (ODDD) affect junctional function: those in the NT (Y17S) and CL (I130T, K134E) domains showed no electrical coupling, while those on TM1 (G21R, A40V) and TM2 (L90V) displayed decreased g_j.

In Cx40-transfected HeLa cells, Bukauskas et al. [83] found multiple γ_j including several sub-states; the γ_j of the main (198pS) and residual (36pS) states, and thus their ratio (γ_jmain/γ_jres = 0.18), were constant and independent from the applied V_j in the 25–100mV range. Short-lived sub-states were often seen, particularly during the early part of a large V_j (−75mV), even with a single active channel; moreover, transitions between the residual and any open state were fast, while those between closed and open states were slow and comparable to the first opening of newly formed channels. In addition, the probability of a channel residing in the main state (P_o) decreased with increasing V_j. Based on all these observations, the authors proposed that the decline of g_j at high V_j reflected partial channel closure, to the residual state, which would alter the junctional molecular selectivity. Thus, with large V_j differences fast gating corresponded to fully open channels transiting to the residual state. In Cx43-transfected HeLa cells and Cx43-expressing fibroblasts from sciatic nerve, CO₂ caused slow transitions from open to closed channels at small V_j (<40mV) values; at larger V_j (>40mV) values, channels displayed fast flickering that was unmodified by CO₂, but CO₂ also caused slow transitions between the residual and closed state [23]. A correspondence between V_j ≈ fast gating and chemical ≈ slow gating was thus drawn.

In neonatal rat heart cells, Cx43 also displays multiple open states; channel records performed with different internal solutions (i.e., mainly KCl, Cs-aspartate or TEA-aspartate) yielded increasing ratios between the main and the residual channel conductance (γ_jmain/γ_jres: 4.2, 5.1 and 6.3, respectively), suggesting that the residual state was more restrictive than the fully open state; again, transitions between open states were fast, whereas transitions involving (to or from) the closed state were slow, leading to the conclusion that the residual state was “the ground state of electrical gating, and the closed state, the ground state of chemical gating” [200]. Similar findings were reported for Cx30 junctions [201].

Cx30, Cx46 and Cx50 hemichannels in HeLa cells also seem to have two gates, one acting on positive membrane voltage (V_m) and reflecting V_j gating, and another “inherent” to hemichannels, acting on negative V_m, and deemed “loop” gating: a slow, multistate mechanism thought to close the channel on normally hyperpolarized cell membranes [203]. Indeed, it was put forward [194] that channel docking and consequent gap junction channel opening can be seen as chemical gating: a “ligand-induced” gating mediated by the “loops” E1 and E2, but a gating that is also voltage sensitive. Using SCAM (substituted cysteine accessibility method) on mutant Cx50 hemichannels, Verselis et al., [211] recently suggested that loop gating results from the movement of residues in the TM1/E1 region (particularly pore lining residues from F43 to D51) that causes pore narrowing.

The fusion of green fluorescent protein to the CT domain of Cx43 (Cx43-GFP) impairs the fast V_j-gating; heterotypic pairing of Cx43-GFP with wt Cx43 facilitates the exposure of the wt hemichannel to an applied V_j, thus allowing a closer examination of its gating behavior. From studies using this junctional system, Bukauskas et al. [20] recently postulated that both fast and slow gating mechanisms share common structural elements and that V_j regulates both of them. According to these authors, the gates operate “in series and in a contingent manner in which the state of one gate affects gating of the other”. This proposition was fully developed on a stochastic model of channel gating in which the fast gates are assumed to be on the cytoplasmic side of the channels, and the slow gates at the region of hemichannel contact [147,160]. The model is eminently biophysical and accurately reproduces experimental results and yields adequate parameters describing the V_j-dependent behavior of junctions. Notably, the model makes no direct assumptions about the connexin parts involved in each gate, but it could be useful, in conjunction with mutational analysis, in determining the role of each connexin region in channel gating.

3.1 Summary

The studies presented above suggest that all regions of the connexin protein likely contribute to gating of the channel. Follows are discussions of data more purposefully implicating the CT, CL and NT in the gating of gap junction channels.

3.2. The ball-and-chain model of channel gating

An early study of Cx43 truncation mutants, Cx43-tr245 and Cx43-tr303 (truncated at residues 245 and 303, respectively) showed changes in channel conductance but not voltage dependent gating. Whereas wt Cx43 channels showed two open states with conductances of 60 and 100pS, the truncation mutants showed single open states with increased conductance (160pS) for the shorter Cx43-tr245 and decreased conductance (50pS) for the longer Cx43-tr303 [57]. These results implicate the CT domain in setting channel conductance(s) and could suggest that a very short carboxyl “tail” is unable to occlude the pore, while the preservation of residues 245–303 in the longer truncation mutant actually increases, in comparison with the still longer wild type protein, the tail’s ability to shape (close) the pore. These authors also showed that the V_j-dependent gating profiles for the truncation mutants did not differ from wild type, which could suggest that the CT domain is not necessary for V_j-dependent gating of Cx43. However, both pore conductance and V_j-dependent gating behavior data from this study are also consistent with a mechanism implicating specific (and different) sequences in the CT domain being necessary for determining these features of channel function. In another study, a modest truncation of Cx45 that eliminated only 26 residues (including a serine-rich region [72]) rendered the Cx45 channel non-functional [84]; however, channel function was rescued by heterotypic pairing with wt Cx45, but fast inactivation was slowed for these heterotypic channels without altering their steady state current, unitary conductance (25pS) or permeability to calcein [84]. Revilla et al [164] also provided evidence supporting the role of the CT in voltage-dependent gating; studying wt and “tail-less” Cx32 and Cx43 expressed in oocytes they observed a decrease in fast inactivation and proposed that the CT was an integral part of the fast V_j-gating mechanism.

Shortly after the observations detailed above were published, the CT domain was implicated in the chemically induced uncoupling of several homomeric junctions formed by distinct connexins. In the initial studies, truncation of gradually longer CT segments progressively decreased and eventually abolished functional expression of Cx43 channels, while also lessening the electrical uncoupling induced by cytoplasmic acidification [116]. The pH_i sensitivity of Cx43 channels was found to depend on specific CT regions, including residues 261–300 and 374–382 [48], as well as an intact CL domain [46]. Coexpression of the CT as an independent particle with Cx43 truncated at residue 257 (M257) partially restored pH_i sensitivity to M257 and also imparted to Cx32 junctions higher pH_i sensitivity [129]. These observations lead to the ideas that: i) the CT closes the channel by interacting with another part of the molecule (channel) near the pore (a ball-and-chain mechanism similar to that of shaker potassium channels [128]) and ii) the “receptor”, of which the CL could be part [46], was preserved among connexins as several, but not all, gap junctional channels seemed to share a similar closing mechanism [183]. It is important to note that in these original studies a direct effect of protons on γ_j or V_j-gating was not shown. Indeed, it was submitted that in lens connexins expressed in HeLa cells, the pH- and V_j-dependent gating mechanisms were functionally and structurally separated [44]. Furthermore, in reconstituted hemichannels, regions of the CT that are likely participants of pH-induced uncoupling (the proline rich 253–270 segment and the 288–359 segment) were not involved in the PKC-induced decreased permeability to sucrose and LY [7], suggesting that pH- and PKC-mediated regulation of Cx43 junctions do not share a common mechanism. However, other studies do support the involvement of a CT-CL interaction in the overall function of the channels, including V_j-gating. For example, in HeLa cells, Cx26 channels allow the transfer of neurobiotin and LY [51] whereas Cx30 channels are only permeable to neurobiotin, not LY [121]. To determine whether any domain(s) of Cx26 could confer on Cx32 a Cx26-like behavior, chimeras were constructed in which either the CT, CL or both of Cx26 replaced these domains of Cx32. The P_j-LY and V_j-sensitivity, as well as channel conductance, typical of Cx26 was reproduced only in the chimeras with both CT and CL from Cx26.

More observations supporting the CT’s role in fast V_j-gating were collected in N2A cells expressing the Cx43-M257 mutant, which displayed no fast V_j-gating kinetics [50]; interestingly, this study also showed that heterotypic docking of wt Cx43 and wt Cx45 also eliminated fast gating, leaving behind a stronger slow gating. Moreover, in accordance with the CT generating the residual channel configuration during fast inactivation, the presence of the CT domain as an independent protein was sufficient to establish multiple channel sub-states missing from the M257 channels [124]. However, recovery of fast V_j-gating upon coexpression of the CT was not clearly shown, and a mechanism by which multiple channel conductance values were generated was not proposed. The elimination of the fast gating by heterotypic pairing [50] is suggestive of connexon to connexon modulation. We are unaware of similar observations in other systems.

The central element of the ball-and-chain hypothesis is the interaction of the CT with a receptor-like region [43,80,98,228]. To demonstrate that such interaction occurs and is enhanced by low pH, Duffy et al. [42] used biophysical methods (mirror resonance, enzyme-linked sorbent assay, nuclear magnetic resonance spectroscopy) to detect protein binding and structural changes, and showed that the Cx43CT binds to the region 119–144 on the carboxyl half of Cx43CL (L2) in vitro; additionally, they showed that low pH induced a higher α-helical order of the Cx43L2 peptide structure, but only minor modifications in Cx43CT. Significantly, in a manner consistent with the L2 peptide specifically binding to CT [172], the delivery of this “receptor” particle through a patch pipette caused loss of the residual state of Cx43 junctions, and this effect was not reproduced with a mutated (H126K-I130N) version of L2. In addition, L2 modified the V_j-dependence of Cx43, making it similar to that of the truncated Cx43-M257. More recently, the mutation of histidine 142, which forms part of the L2 peptide, was shown to affect fast gating of Cx43 junctions in frog oocyte pairs, further supporting the role of the CL in a particle-receptor mechanism of channel closure [175]. Mutation of H142 to glutamate caused an apparent decrease of transitions from fully open to the residual state (as well as the appearance of small, ~50pS, low-frequency events) and prolonged the mean open time of the channel ([175], their figure 3).

Summary 3.2

In agreement with a proposed stochastic model of channel gating [20], the ball-and-chain model of gating suggests that both the chemical and V_j-induced gating of Cx43, and likely other connexins, involves the interaction of CT and CL domains of the channel proteins.

3.3. The cork hypothesis

In the oocyte system, Cx38 is more sensitive to CO₂ and V_j than Cx32. From studies on protein chimeras created by swapping domains between these connexins, Wang et al. [214,215] proposed a major role for the CL in both gating types, although the participation of other domains was not discarded. They initially postulated differential roles for the 1^st and 2^nd segments of the intracellular loop (CL₁ and CL₂) in V_j-gating and CO₂-induced (pH) gating, respectively. As deletion of most of the CT of Cx32 did not affect its CO₂-sensitivity, but replacement of the arginines at the beginning of the CT (dubbed CT₁) with asparagines greatly increased CO₂-gating, Peracchia and Wang [152] suggested that interaction occurs between negative residues in CL₁ with positive residues either in CT₁ or CL₂, with the open channel resulting from CL₁-CT₁ interaction and the chemically (pH, Ca) gated closed channel resulting from CL₁-CL₂ interaction. Evidence that positive charges at CT₁ were key to the CO₂-sensitivity, but not V_j-sensitivity was presented [216] and incorporated into a model in which CT₁ acts as “a latch that keeps the channel open by immobilizing the CL, a potential gating element.” Breaking the CL₁-CT₁ interaction would be part of the gating process that enables the CL (not CT) to move toward the channel pore and close the channel. It remained unclear whether the CL₁-CL₂ interaction occurred within a single connexin unit or between neighboring units. From careful consideration of the above data, it would seem that CT₁ would actually be involved in both V_j and chemical gating.

Subsequently, atypical V_j and CO₂ sensitivities were reported for junctions made of concatenated connexins or connexins with single point mutations designed to weaken the association of CT with the CL; it was proposed that the chemical gate is voltage sensitive, and that the chemical gate and slow gate are one and the same. The slow (chemical) gate, distinct from the conventional V_j-sensitive gate and responsive to changes in Ca²⁺ and H⁺, could possibly involve calmodulin (CaM), or other small acidic protein bound to the CT (cf. [59,148,149]) or the NT [40,150,154] to act as a negatively-charged channel-plugging molecule (“cork” gating model, [40,153]). Notably, inhibition of CaM expression with oligonucleotide antisense to CaM mRNA did reduce the V_j and/or CO₂ sensitivities of otherwise very sensitive mutants of Cx32, as well as of wt Cx38 and Cx45 ([151,154,155]) and the Ca²⁺-dependent g_j decrease of wt Cx50 junctions [29]. Very recently, it was reported that in Cx26 channels, a pH-induced interaction between the CT and the carboxyl half of the CL occurs that can be inhibited by taurine binding to the same CL region [117]; taurine, a protonated aminosulfonate, is known to block channels that contain Cx26 [119], perhaps through the postulated plugging mechanism. Ancillary proteins, binding to the CT instead of the CL domain, have also been postulated in the uncoupling of Cx43 junctions by Src tyrosine kinase [173].

Summary 3.3

The cork hypothesis suggests that CT/CL interactions maintain channels in the open state. A V_J-sensitive, slow (chemical) gating occurs by interaction between the two halves of the CL, or by the occlusion of an accessible pore by an associated plugging element (calmodulin).

3.4. The NT domain plug

Wang et al. [214], in their domain-swapping studies of highly pH sensitive Cx38 and less pH sensitive Cx32 channels found no role for the NT in channel sensitivity to pH and V_j. In contrast, working with mutants of Cx32 and Cx26, whose hemichannels display opposite gating polarities, Verselis et al. [210] proposed that residues in the NT in combination with others in the M1/E1 border constituted part of the voltage sensor. In studies by Dong et al. [41], the fast gating of junctional currents from Cx43 and Cx45.6 channels was lost when the NT of Cx45.6 was substituted by that of Cx43.

The latter studies implicate the NT and V_j-dependent gating. The idea that the NT participates in the permeability and/or gating of connexins received a considerable boost from very recent structural work on Cx26. Specifically, crystallographic analysis of hCx26M34A, a site mutant of Cx26 that decreases junctional dye coupling, yielded images of a prominent density in the center of hemichannels at the level of the transmembrane domains, which was interpreted [142] as a physical obstruction, an occluding “plug” blocking the pore. This “plug” could comprise the NT, the CL or the CT, but because of previous data linking channel gating with movements of charges within the NT, this domain was proposed as the most likely candidate. Accordingly, deletion of residues 2–7 from the NT domain reduced plug density [143]. Higher resolution analysis of wt Cx26 channel structure revealed three potentially important components of the permeation pathway: channel entrance, pore funnel and extracellular cavity, corresponding to the intracellular, transmembrane and extracellular regions, respectively. The channel entrance includes eleven positively charged residues from TM2 and TM3 that could create a favorable environment for negatively charged permeants [120]. The funnel surface is lined by the NT (residues A2 to I9), thus forms part of the pore under at least some conditions. Because the funnel-like structure forms a constriction at the cytoplasmic side of the pore (their figure 4), the size and electrical character of the side chains residing in this area were postulated to have a strong effect on the cutoff size and selectivity of the channel. The inner walls of the extracellular cavity of the pore are formed by the twelve N-terminal halves of E1 (6 from each hemichannel). In Cx26, lysine 41 (K41) lies at the TM/E1 boundary narrowing the pore and creating a positively charged environment between the funnel and the negatively charged extracellular cavity; the authors postulated that K41 and the N-terminus, despite their distance, could indeed act together to sense the voltage field, in agreement with early proposals ([210]). More, exquisitely detailed considerations of the solved structure are presented in the rest of the Maeda paper [120]. This high resolution structure further suggests that while the N-terminal helixes (NTH) line the funnel, the loop connecting the NTH and TM1 is very flexible (cf. [86,159]) such that the NTH is drawn to the inner wall of the channel by interaction with a methionine (M34) of the neighbor TM1; this interaction was proposed to keep the channel open. Weakening of the interaction with M34 could thus narrow the funnel, effectively reducing channel permeability. It was hypothesized that “unbound” NTHs assemble to form the observed “plug” [120], physically closing the pore; the release of a single NTH could destabilize the structure and lead to subconductance states. In conclusion, although the roles of other parts of the molecule (CT, CL) are not discarded, the NT domain(s) might both sense the voltage and effect channel closure by coalescing into a central plug.

Summary 3.4

Independent from the possible role of other connexin regions, the NT model of channel closure hypothesizes that V_J-sensitive gating occurs by the concerted action of the NT’s organizing into a plugging element in the center of the pore. Subconductive states may result from the disconcerted destabilization of a single NT unit. Importantly, such structural destabilization carries the potential to deeply affect channel permeability.

Gating and Permeability - Summary

For now then, at least two channel gates seem operative: i) a fast gate, recognized in the fast component of V_j-inactivation, likely yielding a residual channel state, and ii) a slow or chemical gate (sensitive to voltage, pH, Ca²⁺, etc), recognized in the slow V_j-inactivation, yielding fully closed channels. The participation of each region in either gate is not, at this point, perfectly clear or patently autonomous. The fast gate could be formed by the CT, the CL and/or the NT domains, and the slow gate by interaction between regions of the CT and CL and perhaps secondary gating elements (calmodulin, taurine). The chemical gate may also be modulated by voltage. Channel permeability could thus be determined by the interplay of multiple domains that participate, isolated or in conjunction, in the closure or partial closure of the channel. Hence, subconductive states could result from the effects of phosphorylation on a number of CT domains, or from the release of the NT-helix(es) into the pore center. In addition, shifts in the electrostatic field near the pore mouth, which could be caused or triggered by chemical (pH, direct phosphorylation) or electrical modulation, and may not be detectable by changes in channel behavior, are also likely to affect the diffusion of molecules from cell to cell.

Due in part to the fact that some connexins have been more thoroughly studied than others, and in part to the volitive or circumstantial preferences of the researchers doing the work, as the reader must have noted, some of the gating mechanisms here reviewed are better or exclusively described for gap junction channels composed of some, but not all connexins. Unclear at this point is whether all gap junction channels, irrespective of connexin composition, share all, some or none of these gating mechanisms. We think it appropriate to personally consider whether homologous mechanisms with variable strength on each connexin, or a core mechanism(s) with multiple triggers depending on the connexin and cell types, could better explain our cumulated observations; and in either case, to keep in mind that regarding channel gating we are not at closure.

4. Electrical conductivity and chemical permeability: unconnected measures of connexin function or the keys to channel selectivity?

4.1. Defining permselectivity

Very few studies have simultaneously measured dye permeability and electrical conductance of individual channels or junctions; consequently, whether these parameters changed in parallel or differentially in response to an intervention remained unclear until recently. The general impression now is that electrical conductance and permeability to fluorescent dyes or metabolites are two distinct measures of junctional function, neither predictive of the other. Moreover, because low (or absent) junctional dye permeability can coexist with substantial electrical conductance, reports of uncoupling in response to some treatments may have inaccurately reflected (under or overestimating, depending on which parameter was assessed) the physiological effects of those treatments. Regrettably, it is not always possible, and it can be a laborious challenge, to perform electrophysiological recordings in all experimental settings; certainly, a rigorous evaluation of junctions in large groups of cells with microelectrode techniques is fundamentally unattainable, even assuming homogeneity of cell coupling (however, see [4,37,70,188]).

Nevertheless, there remains the feasibility of performing, under well-controlled circumstances, the simultaneous/sequential measurement of both parameters (permeability and conductance) for the same junction ([45,47,136,199,207,209,221]). Because junctional permeability to current-carrying ions is a common denominator of all gap junctions, irrespective of connexin composition, and the electrical conductance of single channels a straightforward measure of their ionic permeability, it is reasonable to measure (and compare) the selectivity of differently composed gap junctions and channels (as we would for other ionic channels) by their permeability ratio P_X/P_I, where _X is any given permeant (dye, metabolite, amino acid, peptide, nucleotide), and P_I represents the permeability to current carrying ions (i.e., conductivity). As already described, junctions may undergo changes in permeability that may not be reflected by parallel changes in their electrical conductivity, but should be detected as changes in the ratio of those parameters, P_j-X/g_j, which we have dubbed junctional permselectivity. A related approach is to simultaneously assess the transjunctional diffusion of two permeants dissimilar in size, charge, or other significant feature [75]; their permeability ratio, P_X/P_Y - where _X and _Y are the selected permeants, can be determined under changing experimental conditions, and the junctional size or charge selectivity thus explored. Given that the selectivities and the unitary conductances of homomeric-homotypic junctions are likely distinctive, it should be possible with either or both methods to portray each connexin by their normal behavior and response to specific (chemical or otherwise) triggers. The attributes, both in control conditions and under selected stimuli, of junctions containing more than one connexin can then be characterized. We surmise that these studies can dissect, in a manner approaching the circumstances of living tissues, the normal function and regulation of single- and multi-connexin junctions.

4.2. Selectivity of multi-connexin junctions: long and short-term regulation

The junctional permeability of different homomeric gap junctions to a given molecule can vary by orders of magnitude and, as well, can display different patterns of permeability to various molecules [60,61]. How do junctions containing multiple connexins whose homomeric counterparts display different channel permeabilities, conductances, and regulatory properties work in physiological settings? A study by Heyman et al. [76] illustrates one possible scenario relevant to cells, such as those of the atria and endothelium, that co-express Cx40 and Cx43.

Heyman et al [76] showed that Cx40 junctions are cation-selective and Cx43 junctions are not charge selective (P_j-NBD/P_j-Alexa350: ~ 11 and ~1.22, respectively). Junctions formed by cells that co-express these connexins displayed variable charge selectivity, but always equal to or between the selectivities of the homomeric junctions. PKC activation had no effect on the dye permeability or selectivity of Cx40 junctions, but was shown to reduce or abolish the permeability of homomeric Cx43 junctions without affecting their charge selectivity. In co-expressing cells, PKC activation increased charge selectivity to levels comparable to that of homomeric Cx40 junctions. These data indicate i) that PKC regulation of the multi-connexin junction relied exclusively on the Cx43 content of the junction (whether in homomeric, heteromeric or heterotypic channels), and ii) that the potential consequences of such regulation on junctional selectivity and cell-to-cell molecular exchange are vast. Moreover, such connexin coexpression presents the opportunity for junctional regulation on flexible time frames: regulation of gating, permeability or selectivity could be effected in terms of minutes by kinases or other fast triggers acting upon the molecular domains of specific connexins (in homomeric or heteromeric partnerships) [76], [225], while longer term (hours, days) regulation might be exercised through the expression level of each protein and come about from processes (transcription, translation, protein sorting) that determine channel composition (homomeric, heteromeric, heterotypic) of junctional plaques [119].

4.3. Permeability of chimera Cx43-CT37: non-charge selective pores with selective tails?

The regulation of junctional permeability/selectivity without (detectable) electrical changes has been suggested here and elsewhere [17,45,49,61]. Here, we provide further support for this possibility.

Measurements of the main γ_j of Cx43 vary from 90 to 125pS, whereas that of Cx37 is well over 300pS (Table 1). In our hands, Cx43 and Cx37 show highly variable γ_j in Rin cells [25,49], but the conductances of the fully open channels are, nonetheless, ~100 and >300pS, respectively. Despite the 3-fold lower conductance, the pore of Cx43 channels is not charge selective [76] and is readily permeated by dyes that do not permeate Cx37 channels well; in turn, Cx37 channels are poorly permeated by dyes and display significant cation selectivity (ratio P_j-NBD/P_j-Alx350 ~ 8; our unpublished observations). Because non-channel related functions of connexins have been suggested [35,36] and growth-arrest of Rin cells by Cx37 expression [25] launches in circumstances where the presence of gap junctions is improbable (low cell density), our laboratory is presently dissecting the roles of the pore and CT domains of Cx37 in growth control. One logical approach is to define the performance of the pore and the CT of Cx37 when associated with the complementary part of a different connexin, preferably one with no growth effect of its own in the cell type under study [25]. With this in mind, we created a chimera with the pore domain (residues 1–236, which includes the NT, TM1-4, E1-2 and CL) of Cx43 and the CT domain of Cx37 (residues 236–333), Cx43-CT37, and examined its growth suppressive and biophysical properties (specifically: dye charge selectivity (P_j-NBD/P_j-Alexa350), permselectivity (P_j-NBD/g_j) and conductance). As expected from our previous data suggesting that channel selectivity is principally determined by the pore domain of the channel [76], the charge selectivity of junctions formed by Cx43-CT37 is indistinguishable from that of wt Cx43 junctions (Figure 2A). However, intriguingly, the P_j-NBD/g_j of Cx43-CT37 junctions more closely resembles that of Cx37 junctions than Cx43 junctions (Figure 2B). These observations suggest that while the pore domain, alone, determines the charge selectivity of Cx43 junctions, the CT domain strongly modifies the overall permeability. Does this decreased permeability of Cx43-CT37 result from interaction of the CT of Cx37 with the pore domain of Cx43?

Table 1.

Multiplicity of connexin channel conductance (γ_j) values. For this and Table 2, cell types used: Human: SKHep1(liver cancer endothelial), HeLa (cervical cancer), HEK293 (embryonic kidney), NHEK (neonatal epithelial); Rodent: MEF (mouse embryonic fibroblasts), N2A (mouse neuroblastoma), MNCM (mouse neonatal cardiac myocytes), Rin (rat insulinoma), NRK (normal rat kidney), C6 (rat glioma fibroblasts); Canine: MDCK (Madin Darby canine kidney); Avian: CEF (chicken lens embryonic fibroblast); Amphibian: Frog (Xenopus laevis oocytes). Other abbreviations and signs: Δ, deletion; N.A., not applicable. LY, Lucifer Yellow. EB, Ethidium bromide. CMTX, Charcot-Marie-Toot X-linked disease. EKV: Erythrokeratoderma variabilis. DFNB1, Nonsyndromic Hearing Loss and Deafness. KID, Keratitis- Ichthyosis-Deafness. MOT, mean open time. PKC, protein kinase C.

Connexin	Cell type	γj (pS)	Findings or effects of mutations	Refs.
Cx43	MEF	94, 47, and transitions in the 30–70 range	N.A.	[14]
Cx43	SKHep1	Cx43 = 60; 90	Cx43CT truncation changed γj	[83]
Cx43	N2A	102; 85; 74	S50V: 75; 60pS; Q49H-S50V: 53pS	[81]
Cx43	N2A	99–114; ~22	L2 (aa 119–144) suppress residual	[77]
Cx43	N2A; HeLa	98–123 (~105) on K aspartate	T8M: ~106; N206S: 107; unaffected γj; unchanged LY, but decreased EB permeability	[48]
Cx43	HeLa	58; 17; on K aspartate	V5/6 His tag: 28and 8 pS	[19]
Cx45	HeLa	27; 11; on K aspartate	V5/6 His tag: 21and 5 pS	[19]
Cx40	HeLa	130; 26; on K aspartate	V5/6 His tag: 126 and 27 pS	[19]
Cx32	N2A	Cx32 = 70; 25, 20, 10	S26L same γj	[53]
Cx32	N2A	70–75	P87A: reduced substate = 5–20pS	[71]
Cx30	HeLa	130–218(~163); 20–35 (~26)	K aspartate, at 34–35°C; γj increased with temperature	[91]
Cx30	HeLa	160; 27	N.A.	[10]
Cx26	Lipid bilayer	165, 70, 35	M34A: 70, 45 pS	[25]
Cx26; Cx32	N2A	135; 53	Cx26 cation select 2.6/1	[86]
Cx26, Cx30	HeLa	Cx26 = 80–110 Cx30 = 120–150	Cx30 bigger channels are cation selective & less permeable; similar channel number by gj value	[97]
Cx31	N2A, HeLa	85	Permeable to LY, Alx350, EB, DAPI	[1]
Cx43, Cx32, Cx26	SKHep1	Cx43 = 60; 90. Cx32 = 120; 150	N.A.	[84]
Cx43, Cx45	Heart	Cx43+/+ = levels from 7 to 66 Cx43−/− = 33; 13 (Cx45?)	Recorded in embryonic mouse ventricular myocytes	[7]
Cx45.6 & Cx46	N2A	Cx45.6 = 184 (~202)	Cx45.6–46NT: 116 (~130); 35; 18pS	[90]
zfCx43.4	HeLa	33; 12	Zebra fish Cx43.4 is an orthologue of mammalian & avian Cx45	[20]
Cx46	Frog	−30mV : 270 (Mean ± SD: 258±37); +30mV: 120 & ~40	Frog oocyte membrane patches; γj segregates with 2nd half of M1	[34]
Cx50	HeLa	220, 43	Cation selectivity: γj KCl/Kglutamate:1:2	[85]
Cx50	Rin	250; 200; 65; 35	S395A: unchanged γj	[43]
Cx37	N2A, Rin	340–375; 60–80	V156D = 47, 250; K162E = 53, 342	[38]

Figure 2.

Characterization of junctional permselectivity properties of chimera Cx43-CT37 (43C37). A, Dye charge selectivity (P_j-NBD⁺/P_j-Alexa350⁻) of Cx43-CT37 resembles that of non-selective Cx43 channels/junctions rather than that of cation-selective Cx37 channels/junctions (inset). B, NBD permselectivity of 43C37 is lower than that of Cx43 and closer to that of Cx37 (inset). C, Channel recordings (left) and their all-points histograms (right) of 43C37 display mainly ~100pS events (upper trace), although smaller transitions (central and lower traces) and evidence that both channel amplitudes coexist (zoomed in trace) can be seen. From baseline (dashed lines: current in the absence of V_j), cumulative current levels (dotted lines) are labeled on the histogram peaks; the distance between peaks, labeled to the left of traces, correspond to the events amplitudes. When present, downward arrows mark start (black) and end (gray) of voltage pulses; a slow channel closure is shown (asterisk) in the central trace. Calibration applies to all traces except the amplified section. D, Histogram of 43C37 events amplitude, showing a single peak between 95 and 110 pS. n=experiments; N=events.

Despite displaying lower permselectivity than Cx43wt junctions, which may suggest interaction (or conceivably lack of interaction [152,216]) of the Cx37CT with the Cx43 pore domain, the unchanged charge selectivity of Cx43-CT37 junctions suggests that if such interaction occurs (fails to occur), it does not affect the hypothetical electrostatic fields that would determine charge selectivity. So, does that interaction take place? In general terms it seems that the interaction of CT domain(s) with conserved receptor-like region(s) in the pore domain supports sub-state behavior, at least for Cx43 junctions [124]. In our observations to date, the channels of the Cx43-CT37 chimera junctions show a high frequency of transitions between closed and fully open states with no clear residual state evident (Figure 2C, upper trace); subconductive states are seen in some record segments (Figure 2C, lower traces), but these are not frequent. Thus, the histogram of events from Cx43-CT37 junctions shows a single peak in the 90–110pS range (Figure 2D). These results could indicate that a pore-CT interaction typical of wtCx43 does not occur in the Cx43-CT37 chimera. Certainly, this conclusion seems to be at odds with the behavior of Cx43-Cx37 heteromeric channels, where intermediate channel conductances and altered V_j-gating suggest such interactions [16]. A possible explanation may reside in the different channel structures: thus, in the Cx43-CT37 chimera the CT “sees” a pore formed exclusively by Cx43 domains, while in the Cx43,Cx37 heteromers a finite chance for each CT to interact with its kin part of the pore can be expected. Because relatively modest changes in channel behavior could have robust consequences on the overall operation of the junctions formed by those channels [4436], further study of Cx43-CT37 and similarly designed chimeras is warranted. The V_j-gating of Cx43-CT37 has not yet been defined, and the complementary chimera, Cx37-CT43 awaits exploration.

4.4. The Cx43-M257 mutant: tail-less but regulated?

The CT domain of Cx43 possesses many elements potentially involved in regulating channel functions, including gating, permeability and selectivity; consequently, many published loss-of-function studies have compared the function of truncation mutants, lacking more or less of the CT domain, to function of wt Cx43 channels (see discussion of ball-and-chain and phosphorylation data above and [73,222] for reviews of Cx43 binding partners). In this context, we have examined the permselective behavior of junctions composed of Cx43-M257 formed by Rin or N2A cells. Because junctional permselectivity of wtCx43 expressed in Rin cells is highly variable [49], possibly due to the changeable phosphorylation state of the CT domain(s), it was not entirely surprising to find that Rin Cx43-M257 junctions displayed a less variable P_j-NBD/g_j relationship (Figure 3A). Although average P_j-NBD/g_j for Cx43-M257 was only slightly smaller than the average for wtCx43 junctions (Figure 3B), when compared to those wt Cx43 junctions with the lowest or highest permselectivities (a fair comparison if the tail is necessary for achieving those states), Cx43-M257 junctions were significantly different from both components of wt Cx43 behavior (Figure 3A–B). The simplest explanation for these observations is that the CT not only participates in (partial) closure and/or modulation of channel selectivity, but can also serve to increase junctional permselectivity, possibly by stabilizing a highly permeable channel configuration by binding to other cytoplasmic regions of the channel (CT-CT, CT-CL or even CT-NT). The possibility that intrachannel interactions could stabilize the open channel configuration has been previously suggested to occur as a consequence of interaction between i) NT and TM1 domains [120] and ii) CT (region contiguous to TM4) and CL (“latch” function, [216]) domains. The data in Figure 3A/B were quite surprising to us in respect to the NBD permselective characteristics of Cx43-M257 junctions as formed by N2A cells, which was consistently lower than observed in Rin cells across the explored g_j range (Figure 3A–B). In our studies of Cx43-M257 function in these cells (Figure 3C & D), as well as in previously published research [50], the Cx43-M257 channels were observed to transit between fully open and closed states almost exclusively (Figure 3D) resulting in a histogram of events with a peak at 100–120pS. Smaller channel transitions are only occasionally observed (Figure 3c, second and third traces). Thus, the low permselectivity of the M257 mutant in N2A cells cannot be explained by a decrease in channel conductivity. The different permselective profile of Cx43-M257 junctions in N2A vs. Rin cells can only be satisfyingly explained by assuming that i) Cx43-M257 channels still possess regulatory regions (in the 20–30 amino acids of the CT domain still present in the M257 mutant [216] or in another region of the protein [41]) and ii) these regions are differentially regulated in Rin vs. N2A cells, or alternatively, that in N2A cells another connexin is expressed that heteromizes with Cx43-M257 and confers on the heteromeric channels and junction a different permselectivity profile.

Figure 3.

Junctional permselectivity properties of truncated mutant Cx43-M257 (M257). A, P_j-NBD/g_j of wt Cx43 (white circles) and mutant M257 in Rin (gray circles) and N2A (black circles) cells. Notice that M257 displays a more linear P_j-NBD/g_j correlation, although with a much lower slope in N2A cells. B, Box plots of P_j-NBD/g_j data from panel A (median and means indicated by narrow and thick lines inside boxes, respectively; percentiles indicated to the left of Cx43 box; black dots depict outliers); inset: individual P_j-NBD/g_j values displayed against g_j (symbols as in A) suggest that permselectivity is not a function of the absolute level of electrical coupling. Cell groups (Mean±SD): Cx43 (all wt Cx43 pairs; 0.072±0.055; n=18), high (high permeability wt Cx43 pairs; 0.113±0.048; n=9), low (low permeability wt Cx43 pairs; 0.030±0.017; n=9), (M257 (Cx43-M257 in Rin cells; 0.057±0.034; n=10) and M257* (Cx43-M257 in N2A cells; 0.003±0.002; n=12) cells. P_j-NBD/g_j of M257 is only slightly lower than that of all wt Cx43 cell pairs (p<0.5). However, P_j-NBD/g_j of M257 differed significantly from both the highest and the lowest permeability groups, whether or not outliers were included in the comparison: M257 (n=9) vs. Cx43 high (n=8), p<0.005; and vs. Cx43 low (n=9), p<0.05. Behavior of mutant Cx43-M257 differed between cell types (p<0.001). C, Single channel recordings of M257 in N2A cells display mainly amplitudes comparable to fully open configuration of wt Cx43 channel (upper two traces), although smaller events (lower trace) and evidence of visits to the residual state (middle trace) can be found. Voltage gradients expressed at the top of each trace. D, Histogram of events amplitude of Cx43-M257 channels in N2A cells shows single peak between 100 and 115 pS. Labels as in Figure 2.

In summary, these data argue that drastic P_j changes can occur in the absence of most regulatory regions of the CT domain, and such P_j regulation occurs without remarkable changes in the electrical properties or behavior of the channels; the data also strongly suggest that regulation of P_j is cell type dependent, possibly due to co-expression of an additional connexin that significantly modifies the regulation of the mutant junctions.

5. Selectivity in context

5.1. Why should gap junctions be selective?

If the sole purpose of junctional communication were to allow the unrestricted diffusion of substances from cell to cell, it would be better served by cell fusion as in mammalian skeletal muscle (which, incidentally, is devoid of connexins in its adult state). The existence of selectivity argues that gap junctions not only facilitate (by providing pathways for) the intercellular exchange of molecules, but they also constrain selectively the diffusion of other such molecules and cytoplasmic components. Such an arrangement is consistent with cells maintaining a high degree of individualization while still cooperating in the function(s) of the tissue(s) they comprise, a likely advantage when one considers the existence of heterocellular junctions. But what are the purposes (or consequences) of selectivity in homocellular junctions? To attempt an answer to this question, let us first consider that selectivity provides for controlled diffusion of some, but not all, substances between cells that form gap junctions. Second, the very occurrence of transjunctional diffusion implies that cell-to-cell differences in the content (concentration) of such substances must exist, or develop transiently when triggered by physiological stimuli. Therefore, cell populations might not be completely homogenous at all times, but their heterogeneities have so far not been recognized by our limited means of study.

If cells within a tissue are, by force of being subject to minor and major variations in milieu and their individual responses to those variations, eventually moved from an initial similitude, and yet they are part of a functionally defined group within an organized whole, then their ability to spatially and timely share common chemicals (to recoup homogeneity) may appear somewhat less remarkable; but their need for a well regulated junctional selectivity becomes positively more intriguing. Furthermore, the simple closure of gap junctions could result in tissue macro or micro-compartments fulfilling important functions both in physiological and pathological settings. Some recent studies may help to place into physiological context the need for regulated (and changing) junctional permeability and selectivity. A few samples may be sufficiently illustrative.

In ovarian follicles, luteinizing hormone (LH) induces MAPK-dependent phosphorylation at residues 255, 262 and 279/282, which is followed by a transient (<5hs) decrease in dye coupling between the cumulus and mural granulosa cell compartments, and between granulosa cells, but not between cumulus cells and oocyte [137]. This reduction of junctional permeability precedes the resumption of meiosis and can be generally reproduced by the junctional uncoupler carbenoxolone (CBX) and by Cx37 antibody injected into the oocyte. These data suggest that signals from the surrounding somatic cells keep oocytes arrested in prophase, and that interruption of such communication is the stimulus that reinitiates meiosis. Recently, it has been suggested that cyclic GMP might constitute the critically interrupted message in this system [138].

In an aesthetically pleasing work, Locke et al. [119] demonstrated that developmental changes in the connexin proteins expressed in the mouse mammary gland are closely related to the progress of milk secretion capabilities. Thus, Cx26 is the main connexin expressed by alveolar epithelial cells during pregnancy, but Cx32 is rapidly induced after delivery. As expected from this change in expression, the channels in the epithelial cell junctions transition from homomeric-homotypic Cx26 channels to heteromeric Cx26-Cx32 channels to homomeric-homotypic Cx32 channels. When these native channels were reconstituted (as connexon hemichannels) into liposomes, differences in their permeability to several metabolites were documented that could be linked to changes in their Cx26-Cx32 stoichiometry. Cx32 channels are wider than Cx26 channels, as measured by their permeability to cyclodextrins [118]. However, while heteromeric Cx26-Cx32 channels were selectively permeable to cAMP, cGMP, and IP₃, their homomeric counterparts were not [119]. Furthermore, permeability of homomeric Cx26 channels and heteromeric channels with high Cx26/Cx32 stoichiometry was inhibited by taurine, an amino acid derivative that plays an important role in milk protein synthesis. Finally, inhibition by taurine decreased as the content of Cx32 increased, and did not occur at all in Cx32 homomeric channels. The authors (coolly) concluded that “channel stoichiometry, permeability, and chemical gating character change in precisely the desired fashion after parturition to maximize molecular and electrical coupling to support coordinated milk secretion.” Modulation of junctional selectivity occurred, in this and the previous case, on a rather long time frame; faster accommodation of physiological needs is also possible, as suggested by Heyman et al [76] and by the next study.

In the mammalian retina, AII amacrine cells (AII) relay input from rod bipolar cells, which collect dim signals, to cone bipolar cells, which collect bright signals and in turn contact ganglion cells. AII cells also make gap junctions between themselves. Thus, AII cells participate in homocellular and heterocellular junctions. Although several connexins are expressed in the retina [66,69,177] whose permeability to cAMP could vary (cf. [88]), it seems that AII cells express preferentially Cx36, while Cx45 may be found in the other cell types [55,68,69,77,170], thus making it plausible that homocellular junctions are homotypic, and heterocellular junctions heterotypic. Xia and Mills [225] showed that upon exposure to the neuromodulator dopamine, dye coupling decreased more rapidly across the homocellular than across heterocellular junctions; however, a membrane-permeant cAMP agonist modulated both types of channel equally, and nitric oxide (NO) preferentially reduced AII-bipolar cell dye coupling. These observations imply that both homocellular and heterocellular junctions are sensitive to cAMP, and therefore, their differential response to dopamine is due to changes in the local concentration of cAMP within each cell; the authors propose that Cx36, characterized by low conductive channels amenable to closure by cAMP, would restrict cAMP elevation to within AII cells, which in turn would tend to close AII-AII junctions more swiftly; alternatively, heterocellular junctions have a higher threshold for cAMP modulation that is readily surpassed by the cAMP agonist. In either case, it seems that coupling between the elements of the neural retinal circuit can be altered, through the modulation of its gap junctions, to favor different pathways in response to dopamine and NO. These observations are very significant for the modulation and neural conduction of bright and dim signals from the retina.

The data reviewed thus far suggest that: i) gap junctions are selectively permeable; ii) gap junction permeability and selectivity vary in a connexin-composition dependent manner; iii) junctional permselectivity depends on limiting diameter of the pore and the surface charges near its mouth and along the pore’s length; and iv) gating, a strong regulator of junctional permselectivity, results from multiple mechanisms involving various regions of the connexin protein. This last proposal carries vast possibilities, for it implies, on one side, that gating can be altered by targeted modifications of single or multiple connexin regions, as repeatedly shown; and on the other side, that the elimination of one gating element, as in the CT truncation mutants, may not completely abolish the gating ability of channels by the remaining mechanism(s).

Plausibly, elimination/perturbation of a gating element cannot be without consequences. Yet, a coupling defect is not always found in some cases where a mutation has the potential to alter channel function. This is particularly disconcerting when such mutations are strongly associated with disease (Table S2). Perhaps, unfortunately, our present knowledge on the function and regulation of junctional coupling is incomplete and inadequate. We and others [45,47,87,199] have performed simultaneous measurements of electrical conductance and transjunctional dye diffusion, in good part to assess whether the ionic and chemical permeability/selectivity of gap junctions are regulated parameters. Our contrasting results only emphasize the need for further exploration if we are to achieve a better understanding of this fascinating facet of cell communication.

Accepting the fundamental premise that gap junctions are selective, it becomes apparent that the importance of this selectivity is related not to their allowing the diffusion of our experimental probes, but the diffusion of their partner cells’ molecular messages. Although a number of endogenous substances have been found to permeate gap junction channels, the goals, conditions or consequences of their transjunctional diffusion are not easily identified. We highlight here the complexity of this issue. Consider that if all connexins are permeable to a given substance (classically – current carrying ions), the junctional diffusion of that substance does not justify the existence of many connexin isoforms or the expression of multiple connexins within tissues, unless the exchange of that particular substance is so essential that it must be preserved while halting the diffusion of other molecules. On the other hand, if our hypothetical substance is of such extreme need (classically - ATP), then cells may be able to produce it by themselves, obviating the need for intercellular exchange. Given these and similar considerations, a question recurs as one of the most relevant to the physiological meaning of junctional coupling: What substances are essential permeants of communicating junctions? Some possibilities are offered next.

5.2. Small change? The true value of cell-to-cell trade (by gap junctions)

The earliest recognized evidence of cytoplasmic continuity -without cell fusion- within a tissue was as an intercellular “pathway” of low electrical resistance, a “nexus” between cardiac myocytes that supported the essentially undelayed conduction of action potentials from cell to cell. Postulating these nexi, also befittedly named electrical synapses, raised the concept of cardiac muscle as a “functional syncytium” in regards to its electrical properties [10,168]. Histologically evident structures located at sites of plasmalemmal contact and named “gap” junctions for their appearance in TEM images, were confirmed [107] to contain hydrophilic channels, which could, hypothetically at least, support the syncitial function of the nexus. SEM of freeze-fractured membranes showed that gap junctions are plaques or clusters of these channels, often tightly packed in a beehive pattern [123]. Much has been done since the first isolation, almost 40 years ago [53,64], of gap junction plaques and their constituent proteins, the connexins. Among profuse data commented on elsewhere [18] and in this review series, it is now well established that connexins form hexameric, barrel-like structures with a 60° rotational symmetry called connexons, which cross the thickness of the plasmalemma in transverse orientation. In the extracellular space these connexons dock with each other to form dodecameric complexes that bridge the gap between contacting plasmalemmae and, more pertinently, surround hydrophilic pores that connect the neighboring cytoplasms.

Gap junction (or communicating) channels have long been thought of as conduits for the synchronization of cell populations, likely serving to enhance their contractile, secretory or other, responses to ordinary stimuli [18]; they were also very early on presumed to participate in growth control, as suggested by the contact inhibition of cell proliferation and the poor junctional coupling of cancer cells [6,182]. Because of the hydrophilic nature of the pores and the lack of evidence for active transport mechanisms, it is widely accepted that the transfer of molecules across gap junctions occurs by simple diffusion. Hence, the functional synchronization of groups of cells likely follows the diffusion of molecules from cell to cell. Such a model is quite appropriately illustrated by the heart, where the junctional transfer of current-carrying ions underlies one physiological outcome (synchronous beating) for this kind of cell communication. Notably, the gradient provided by the space distribution of the action potential seems an obvious condition for junctional diffusion. The heart could indeed typify gap junctions’ logistic organization: pacemaker cells generate a gradient that drives transjunctional diffusion. While there is little indication that pacemaker cells exist within all tissues, cell populations are not homogeneous, even in culture. In fact, an epithelial system can be recreated where the heterogeneous expression of hormone receptors makes the cells with the higher expression pacemakers for the initiation of Ca²⁺ waves through gap junctions [111]. More compelling, pacemaker activity has been recently proposed to underlie the cyclic secretion of insulin [112,224], and interstitial Cajal or Cajal-like cells, which may express Cx43, could be the source of pacemaker activity in various smooth muscle arrays [106,133,146,171,187,188,217]. But independently from the temporal or structural setting for cell coordination, conviction upholds that where gap junctions are concerned, functional synchrony is achieved through junctional diffusion. Questions arise as to whether all gap junction-based cell coordination depends on the diffusion of current-carrying ions, or whether more complex molecules are involved. If so, which are those molecules?

From a pure diffusion perspective, any molecule with certain physicochemical attributes (i.e., appropriate size, shape, charge) can permeate gap junctions. Thus, all small hydrophilic substances are good candidates for transjunctional diffusion and make a long list of possible junctional permeants. Reducing this list to manageable proportions requires us to reconsider that relevant permeants must not simply be present in the cytoplasm: they must support cell synchronization. And because functional changes are unlikely to be triggered by substances that are always present, two different but not mutually exclusive scenarios are possible: i) relevant substances are present transiently (or severely reduced transiently and in a localized manner), and ii) net movement of substances through the junction, net permeation, requires a transjunctional electrochemical gradient. More plainly, we posit that relevant junctional permeants are produced (or depleted) in some cells of a tissue thereby creating a gradient for their diffusion towards (or from) their neighbors. Moreover, if they are to synchronize the tissue, such molecules should directly or indirectly modify the functional state of the “recipient” or target cells. Under these stipulations, in addition to the importance of current carrying ions and second messenger molecules as permeants (e.g.: cAMP, IP₃, Ca²⁺), two additional attractive candidates are oligopeptides and oligonucleotides, both of which may diffuse relatively well through at least some gap junctions to regulate function in neighboring cells.

Small peptides (6–15 residues) with appropriate sequences have the potential to block enzymatic activity and protein binding sites [56,169]. The deprimerones, an intriguing type of oligopeptide, were described many years ago as peptides associated with DNA and nuclear RNA that are able to control transcription and translation in vitro; interestingly deprimerones variety and content might be decreased during carcinogenesis [78]. Subsequent studies demonstrated that very short synthetic peptides with sequences related to acidic chromatin peptides specifically bind DNA and/or effectively inhibit transcription [5,27,31]. Recently, similarly short peptides were suggested to be the long-sought for chalones, hypothetical substances that would exert cell growth-inhibition within the same organ where they were secreted [52]. Still more recently, intercellular diffusion of peptides up to 10 amino acids in length (relative molecular mass ~ 1800 Da) was suggested as important in antigen presentation [132]. The potential physiological implications of such data make finding whether these classes of peptides are produced and exchanged intercellularly, in vivo, in sufficient amounts to explain physiological synchronization of tissue activity, a beguiling enterprise.

Short interfering RNAs (siRNAs) have been created as synthetic tools to hinder protein expression by mRNA binding. siRNA’s endogenous counterparts, microRNAs (miRNAs), are now known to exist in substantial amount and variety. Because several miRNAs have been confirmed to block mRNA translation, miRNAs are currently being recognized as important regulators of gene expression. Accordingly, both concentration and specificity of miRNAs seem to change with the cell type and functional state and, more interestingly, significant differences exist between the miRNA contents of normal and cancer cells [110]. Dysregulation of miRNA has also been linked to a series of muscle disorders, including some associated with cardiovascular disease [28,145]. Recent data indicate that oligos of the size of miRNAs (~20–22 nucleotides) can traverse gap junctions [96,202]. As with small peptides, the important question from the gap junction field would be whether miRNAs are, under natural conditions, produced in sufficient concentration to effectively diffuse from cell-to-cell and yield physiological effects.

It is noteworthy that, unlike electrical coupling, whose effects depend on the profuse diffusion of ions, the intercellular exchange of even small numbers of these oligopeptides and oligonucleotides could have profound effects on recipient cells due to their potential ability to regulate gene expression at the transcriptional and translational level. Therefore, we speculate that oligopeptides and oligonucleotides, the “small change” of the cell, will prove to play important roles in the cost-effective synchronization of tissues through gap junctions.

Highlights.

• >Multiple connexin regions define permeability and selectivity of gap junctions.
• >Gating, electrostatic interactions and phosphorylation modify channel function.
• >Junctional molecular and ionic permeability may change in non-parallel fashion.
• >Multiple mechanisms of channel gating have been proposed.
• >Naturally existing junctional permeants are not broadly known.

Table 2.

Mutations within all regions of the connexin protein can affect the functionality, conductivity or permeability of gap junctions. Notice that for several mutation, particularly in Cx32 and Cx26, an association with disease but no channel dysfunction, have been identified. Abbreviations and symbols listed on Table 1.

Region	Prep	Mutant	Effects or observations	Refs.
		CX43
NT	Frog	Y265F	Inhibits phosphorylation & gj decrease induced by pp60v-src protein kinase	[88]
NT	HeLa, N2A, HEK293	Cx43D12S,K13G	Reduced Vj-gating & γj (30pS?); high gj; permeability to LY remains	[26]
NT	N2A	Y17S	Afunctional	[79]
TM1	N2A	G21R; A40V	Afunctional	[79]
E1	N2A	F21up (codon duplication)	Afunctional	[79]
E1	MNCM	G60S	No gj development: impaired protein traficking to the membrane	[46]
TM2	N2A	L90V	Lower gj	[79]
CL	N2A	Cx43 + L2 peptide	L2 (but not L2-H126K-I130N mutant) slows gating, abolishes residual γj	[77]
CL	Frog, N2A	H142E	Decreased fast voltage inactivation & frequency of 80pS substate, and increased Po	[78]
CL	N2A	I130T	Lower gj, unchanged γj	[79]
CL	N2A	K134E	Lower gj; lower γj	[79]
TM3	SKHep1	Point mutation (sic)	Changed selectivity (sic)	[83]
TM3	SKHep1	S158F, S158L, S158D	Reduced permeability to LY on well coupled pairs with higher γj	[84]
TM3	Frog	T154A	Closed pore	[6]
E2	N2A	R202H	Afunctional	[79]
CT	SKHep1	Cx32 & Cx26-like truncation	Vj-dependence not changed	[84]
CT	Frog	M257; M247; M243; M241	Truncations: M257 gj same as wt, but gradually decreased to zero as CT got shorter	[21]
CT	Frog	Δ241–257	Unchanged expression level	[21]
CT	Frog	S244G; S255A	Same expression level as M257	[21]
CT	Frog	M257	pH-gating: decreased pKa from 6.6 to 6.1 (as that of Cx32)	[44]
CT	SKHep1	Truncation (sic)	Changed γj	[83]
CT	Frog	Truncation at M257	Decreased pH-sensitivity; Cx43CT alone enhance the pH-gating of M257 & Cx32	[51]
CT	Frog	Δ261–280; Δ281–300; Δ281–290; Δ291–300	These deletions decreased pH-sensitivity	[23]
CT	Frog	M361; M374; M257	These truncations decreased pH-sensitivity	[23]
CT	Frog	D379N; D379Q; D378N-D379N-E381Q; P277A-280A; P375A-377A; R374Q-R376Q; S364P	These single and multiple point mutations decreased pH-sensitivity	[23]
CT	Frog	D378N-E381Q	Increased pH-sensitivity	[23]
CT	Frog	D379K; D378N; E381Q; D378N-D379N; D379N-E381Q; D378N-E381Q	Single and multiple point mutations that caused no clear changes in pH sensitivity	[23]
CT	Frog	Δ241–260; Δ301–330; Δ321–340; Δ341–360; Δ364-363; Δ364–373	These deletions caused no perceptible changes in pH-sensitivity	[23]
CT	Frog	M257	Loss of insulin- and IGF-induced gating	[31]
CT	Frog	Δ261–280	Decreased insulin-gating	[31]
CT	Frog	Δ301–320	Increased insulin gating	[31]
CT	Frog	Δ241–260; Δ281–300	Unchanged insulin-gating	[31]
CT	Frog	RCx32-257stop	Abolished the fast mechanism of Vj-gating	[70]
CT	Frog	S257stop; G242stop	G242stop eliminated fast Vj-gating	[69]
CT	Frog	R243Q; D245Q	R243Q decreased & D245Q increased Vm-dependence without effect on Vj -gating	[69]
CT	N2A	M257	Eliminated fast Vj-gating & residual γj, increased MOT, slowed channel transitions	[50]
CT	C6	M257	Lower hemichannel Po than wtCx43	[37]
CT	NRK, MDCK	S365A; S365D	Normoxic and ischemic adult mouse heart; p-S365 reduces PKC dependent p-S368	[82]
		CX32
NT	Human	G12R; G12R	Mutation in human keratinocytes; EKV; intercellular traficking? Gating polarity?	[74]
NT	Frog	G12S	Afunctional	[53]
NT	Frog	G12V; G12Y; G12S; G12P	V, Y, S mutants afunctional; P mutant forms heterotypic channels with Cx32 & Cx26; G12 might be a hinge in the N-terminal peptide MDWGTLQSILGGVNK	[36,65]
NT	Frog	G12P; G12Y; G12S	G12 forms a “hinge” that allows the NT to move toward the pore	[36]
NT	Frog	N2E; N2D	Rectification	[55]
NT-TM1	Frog	R22P; R22G	Functionally incompetent	[67]
TM1	Frog	S26L	Normal Vj-dependence; γj unchanged in CsCl (70pS) but smaller (43 vs. 54 in wt) in LiCl; lower permeability to Li+; slight anion-selectivity	[53]
TM1	Frog	I30N; P87A	Slightly altered Vj-dependence	[53]
TM1	Frog	M34T	Altered Vj-gating; γj65 (brief) & 15pS (long) openings; lower Po	[53]
TM1	Frog	M34T; V35M; V38M	Most altered Vj-gating: V35M; in heterotypic pairing (Cx32) peak gj at negative Vm; lower Po	[53]
TM1-E1	Human	R42P	Mutation in human keratinocytes; EKV; Vj-gating?	[74,95]
E1	Frog	L56F	More pronounced Vj-dependence as Cx32wt, but unchanged pH-dependence	[67]
E1	Frog	C53S; C60S; C64S	Inhibit gj developing	[16]
TM2	Frog	P87A	Vj -gating similar to wt	[53]
TM2	Frog	L90H	Functionally incompetent	[67]
TM2	Frog	Substitutions: P87G; P87A;	Unchanged, symmetrical homotypic Vj-gating; P87A/Cx32 asymmetrical Vj-gating	[71]
TM2	Frog	Substitutions: P87V	Afunctional	[71]
TM2	Human	C86S	Voltage gating? EKV; mutation in human keratinocytes	[74]
CL	Frog	E102G; Δ111–116	Vj-gating as wt	[53]
CL	Frog	V95M	Functionally incompetent	[67]
CL	Frog	E102G; Δ111–116	More pronounced Vj-dependence than Cx32wt; Δ111–116 paired with wt showed rectification; both showed higher pH-dependence	[67]
TM3	Human	F137L	Heterotypic formation; EKV ; mutation in human keratinocytes	[74]
E2	Human	R180X; E183K	Hearing impairment; mutation in human keratinocytes	[74]
E2	Frog	C168S; C173S	Inhibit gj developing	[15]
E2	Frog	C168S; C173S; C179S; L167T	Inhibit gj developing	[16]
E2	Frog	P172S	Functionally incompetent	[67]
E2	Frog	D169N, D178N, D178Y	Reduce calcium block of hemichannels	[28]
CT	Frog	Truncated 58 last amino acids	No effect on Vj and pH gating	[94]
CT	Frog	S233A; S240A	No effect on Vj and pH gating	[94]
CT	Frog	C217S	No effect on gj developing	[16]
CT	Frog	E208L; Y211stop	Functionally incompetent	[67]
CT	Frog	R220stop	More pronounced Vj-dependence as Cx32wt, but unchanged pH-dependence	[67]
CT	Frog	E208K; R215W; Y211stop	Afunctional	[13]
CT	Frog	R215Q; R215stop	Afunctional	[13]
CT	Frog	R238H; R280G; C217stop; R220stop; R265stop; S281stop	Hemichannel incorporation decreases with CT shortening; minimum length should include Arg215	[13]
CT	Frog	Y211stop; E208K; R215W	CMTX mutations; inhibit hemichannel & gap junction formation	[4]
CT	Human skin	Δ218–221	Phosphorylation & pH-gating; mutation in human keratinocytes	[74]
CT	HeLa, Frog	F235C	Normal localization; lower Vm and Rm; Vj shifted to neg Vm; (high Po of hemichannels?)	[40]
CT	Frog	Cx32-D225	CT truncation eliminates fast Vj gating; slow gate becomes faster	[60]
CT	Frog	HCx32-220stop	Abolished the fast mechanism of Vj gating	[70]
CT	Frog	C217stop; R220stop; R265stop; S281stop; R238H	Normal hemichannel & gj formation	[4]
CL & CT	Frog	NT-to-CT link, “tandem”; Substitution of R 215, 219, 220, 223 & 224: 5R/E & 5R/N	Homotypics: tandem & 5R/E afunctional; 5R/N functional; heterotypics with Cx32 rectify to positive V & are more CO2-sensitive than Cx32wt (5R/E > 5R/N > tandem > Cx32); CO2 enhance Vj-rectification in similar order of degrees.	[61]
All domains	Schwann cells	>150 mutations	CMTX mutations may affect all domains and the biosynthesis, assembly, targeting or function of gap junctions	[66]
		CX40
NT	N2A	E9,13K	Reduced Vj gating sensitivity, γj & spermine block	[52]
NT	N2A	E9K; E13K	Reduced gj min; reduced Vj gating	[52]
NT	Frog	Cx40-TR (CT-truncated)	Unchanged Vj gating & CO2 sensitivity	[58]
NT	N2A	H15Q+K16A	Reduced spermine block & Vj gating, increased gj min and Vo; unchanged gamma j	[41]
NT	HeLa, N2A, HEK293	Cx40E12S,E13G	Reduced Vj gating, block by spermine & γj (<100pS); LY-permeable	[26]
		CX50
NT	Frog	D3N	Inverted gating polarity; increased Vj-gating & CO2 sensitivity of chemical gating	[59]
NT	Frog	G22R	Afunctional; heteromeric w. Cx46 faster Vj -gating	[96]
NT	Rat lens	L7Q	Microphthalmia, pulverulent cataract	[42]
TM1/E1	Frog	F43C	Impaired hemichannel opening on low Ca++	[92]
E1	Frog, HeLa	S50P	Afunctional; alters Vj gating when mixed with wt Cx46 or Cx50	[18]
E1	Frog; CEF	E84K	Decreased gj and dye permeability (LY, calcein); no hemichannel effect	[3]
TM2	Frog	P88S	Afunctional; dominant negative single subunit	[57]
TM2	HeLa, N2A	P88S	Decreased gj; dominant negative	[9]
CT	CEF; Rin	S395A	Increased fully open channels; less responsive to PKA activation	[43]
		CX26
NT	N2A, HeLa	T8M	Decreased EB, but not LY transfer; same γj	[48]
NT	Frog	T8M	Vj gating inverted asymmetry	[47]
NT	NHEK	G11E	Increased intracellular Ca++; cells death; KID syndrome	[89]
NT	Lipid bilayer	M34A	Decreased γj	[25]
NT	Frog	I71N	Recessive hearing loss; reduced membrane incorporation and hemichannel activity	[49]
TM1	Frog	V37I	Afunctional, DFNB1	[12]
TM1	Frog	M34T	Afunctional homotypic; inverted Vj response in heterotypic with wt	[80]
TM1	Human	N54K	Barth-Pumphrey Syndrome (BPS): dermatological disorders & hearing loss	[72]
TM1	Human, HeLa	M34T	Hearing loss; decreased γj, no LY transfer or Ca++ waves; model: pore constriction	[10]
TM1	Frog	M34E	Reduced conductance; stable connexons	[2]
E1	NHEK	D50N	Increased intracellular Ca++; cells death; KID syndrome	[89]
E1	Human	D66H	Vohwinkel's syndrome: mutilating keratoderma w. sensorineural deafness	[45]
E1	Frog, HeLa	W44C	Afunctional, dominant negative (gj, dye transfer.); deafness	[11]
E1	Frog, HeLa	W77R	Afunctional; no dominant negative	[11]
E1	Frog	W44C	Afunctional, DFNB1	[12]
E1	Frog	A40V; G45E	Increased hemichannel Po & Vj gating	[27]
E1	Human	P58A	Hearing loss	[64]
E1	Human	L76P	Hearing loss	[5]
E1-TM2	Human	R75Q, R75W	Afunctional; hearing loss & palmoplantar keratoderma	[17]
E1-TM2	Frog	R75A; R75N; R75D; R75K; R75F; R75Y	Afunctional; altered Vj sensitivity of hemichannels	[17]
TM2	Frog	P87L; P87G	Afunctional; heterotypic pairing with wt yielded low gj & reversed Vj-gating	[87]
TM2	Frog	W77R	Cytoplasmic labeling; DFNB1	[12]
TM2	Frog	F83L polymorphism	Normal function	[12]
TM2	Frog	V84L	Normal function; DFNB1 associated	[12]
TM2	Frog	L90P	Afunctional; DFNB1	[12]
TM2	Frog	P87L	Unstable connexons	[2]
TM2	Frog	V84L	Unchanged gj, LY permeability & γj; low IP3 permeability; stable connexons	[2,8]
CL	Frog	S113R	Afunctional; DFNB1	[12]
TM3	Frog	R143W; V153I	Afunctional	[47]
TM3	Frog	T135A	Closed pore; unstable connexons	[2,6]
E2	Frog	M163V; R184P	Afunctional; DFNB1	[12]
E2	Frog, N2A	rCx26-N159D	D159 in sheep, human & the rat mutant linked to high hemichannel Po	[29]
TM4	Frog	N206S	Vj-gating shows less asymmetry	[47]
TM4	Frog	L214P	Afunctional	[47]
TM4	N2A, HeLa	N206S	Decreased EB, but not LY transfer; same γj	[48]
CT	Human, HeLa	R127H	Much reduced LY transfer	[10]
		Cx30
NT	Frog	T5M	Afunctional, dominant negative (decreases gj)	[30]
NT	Human, Mouse	T5M	Hearing loss; reduced endocochlear potential; reduced dye but not electrical coupling; disturbed Ca2+ waves	[76]
		Multiple connexins and regions
CX32, Cx43; NT	Frog	N2E; N2R/K; G5D; G5R/K; T8D; T8R/K	Mutations on chimera Cx32*Cx43E1 caused change in gating polarity & unitary conductance of hemichannels	[54]
CX32, Cx43; NT	Frog	N2E; N2R/K; G5D; G5R/K; T8D; T8R/K	Mutations on chimera Cx32*Cx43E1. Dispersed pore charges determine hemichannel charge selectivity & conductance	[56]
CX30.3; NT	Human	G12D	CX30.3 related to EKV	[73]
Cx37; NT	HeLa, Frog	Many partial deletions; 2–8 alanine substitution	>9 aa needed for plaque formation; all mutants impair dye permeability (neurobiotin)	[39]
Cx45.6, Cx46; NT	Frog, N2A	Cx45.6*46N (Cx46NT + Cx45.6 other domains)	Slower inactivation kinetics, lower Vj sensitivity & γj	[90]
Cx38NT on Cx32	Frog	Chimera Cx32/38N	Cx32-like pH-dependent gating; more relaxed Vj -dependence	[93]
Cx46, Cx32, Cx43; TM1	Frog	Switching TM1 between Cx46 & chimera cx32E143: cx32M146E143	Single hemichannel features follow TM1	[33]
Cx30.3; TM1	Human	R22H	Cx30.3 mutation in human keratinocytes, linked to EKV	[73]
Cx46; TM1	Frog	L35C	Cx46; calcium-gate located external to L35C	[62]
Cx46, Cx32*; TM1	Frog	M1 swap between Cx46 & chimera Cx32E143	γj segregates with 2nd half of M1	[34]
Cx32, Cx26; E1	Frog	Cx32*26E1; K41E-E42S (Cx26*32ES)	In heterotypic paring with wt Cx32 or Cx26, chimera caused weak rectification & point mutation increased Vj-dependent gating	[75]
Cx46; E1	Frog	N63S	Increased hemichannel Mg blockade	[22]
Cx45; E1	HeLa	Q49H; Q49H-S50V; S50V	Q49H mostly nonfunctional; S50V decreased γj	[81]
Cx32 & Cx43;E1	Frog	Cx32E143: Cx43E1 on Cx32	Yields membrane conductance (hemichannels) that is V, Ca++ & CO2 (pH)-sensitive, and takes up several dyes	[63]
Cx30.3; TM2	Human	T85P	Mutation in human cells; EKV: circinate or targetoid erythema	[73]
Cx32, Cx38; CL	Frog	Cx32/38I: Cx38CL on Cx32	Chimera shows Cx38-like pH-dependent & Vj-dependent gating	[93]
Cx30.3; TM3	Human	F137L	Mutation in human cells; EKV: peeling; hypertrichosis; circinate, gyrate erythema	[73]
hCx37-S319; TM3	N2A, Rin	K162E and V156D on hCx37-S319 variant	K162E: Lower Vj-gating (Vo=37 vs 30mV); γj 58–342; 20–92; (wt: 374; 74). V156D: γj = 47–250 (flickering)	[38]
Cx30.3; TM4	Human	F189Y	Mutation in human cells; EKV: hyperpigmentation, hypertrichosis	[73]
Cx46, Cx43; CT	Frog	C218, 283, 321A; Cx43ΔCT	Cysteines participate in hemichannel close-gating	[68]
Cx45; CT	HeLa	Cx45 CT truncated	Afunctional; rescued by wt proteins	[35]
Cx32, Cx38; NT, TM1, E1, TM2, CL, TM3, E2, TM4, CT	Frog	Cx32/38C; Cx38N-M2/32I-C; Cx38N-I/32M3-C; Cx32N-M381-C; Cx32N-I/38M3-C; Cx38/32I;Cx32-S17A,T18A;Cx32-d33	Chimeras created by domain shifts affecting single or multiple domains; all listed here afunctional	[93]
Cx43, Cx38, Cx46	Frog	Hemichannels; homotypic and heterotypic junctions	pH regulation susceptibility intrinsic to hemichannels, but modified by interactions between connexons; all domains seem involved.	[24]
TM1;TM4; CT	HeLa, Frog	F30V; P191S; F209I	Zebrafish Cx43 short fin mutations: low gj; abnormal Vj gating; F209I less affected	[32]