Diverse Molecular Mechanisms of Gap Junction Channel Gating

Bruce J. NICHOLSON, Lan ZHOU, FengLi CAO, Hui ZHU and Ye CHEN

Abstract: Gap junction channels are subject to acute regulation by a number of factors, both electrical and chemical. Gating kinetics of the channels reveals two classes of responses: rapid closure in response to voltage; and slower responses associated with the original formation of channels and their closure induced by chemical signals such as lypophillic agents and reduced cytoplasmic pH. Using the paired Xenopus oocyte expression system, we have found that modifications to several domains, including M2 and both extracellular loops, can induce a reversal in the response of gap junctions to transjunctional voltage (Vj). This suggests that voltage gating involves an inherent conformational change of the connexin that is "global" in that many parts of the molecule are involved. In contrast, the kinetically slower closure of Cx43 channels in response to cytoplasmic acidification has been associated with a discrete "ball and chain " mechanism involving the C-terminal domain. Application of a novel oocyte perfusion chamber has revealed that this gating mechanism also requires additional heat labile component(s) of the cytoplasm. Analogous studies of the gating of Cx43 channels by v-src indicate that this response also utilizes a "ball and chain" mechanism. Contrary to previous findings, however, this appears not to depend on phosphorylation of Cx43, but may also require interaction with cytoplasmic factors, possibly by way of SH3 domains in the C-terminal tail. These initial results suggest that kinetically slower chemical gating of gap junctions may occur via interactions between discrete domains of the connexin, mediated in at least some cases by additional soluble factors, while faster, electrically mediated gating events represent global conformational changes of the connexin.

The mechanisms that underlie the switch between open and closed states is appropriately portrayed as a holy grail in the structural and functional analysis of ion channels. Despite many innovative studies, this is understood in surprisingly few cases. The inactivation "ball and chain" gate of K⁺ channels is perhaps the only case where most of the molecular components have been identified (see[1] for review). In other cases certain components have been defined, such as the voltage sensor of Na⁺ and K⁺ channels [2], the ligand binding site, or the gating movement within the pore of ACh Receptors [3]. However, the conformational changes that link these steps are ill defined. Gap junction channels are likely to provide a novel perspective to this field as the intercellular nature of these channels results in certain unique properties. Firstly, the default state of the channels appears to be open, so the gating events involve closure of the channel. Secondly, in cases where the gating process involves general conformational changes of the protein, part of the activation energy barrier of this process must include contributions from the docking interface between the two hemichannels. Finally, the high resolution structures of the gap junction channel that are now becoming available (reviewed in [4]) greatly enhance the prospects for definitive conclusions from these molecular studies.

Closure of gap junction channels is induced by either response to cell trauma (e.g. drop in pH, increase in Ca⁺⁺ or voltage differences between the cells - see [5] for review) or physiological regulators such as phosphorylation (e.g. v-src or MAP kinase - reviewed in [6]). Kinetic studies at the single channel level indicate that gap junction gating can be divided into two general classes: rapid (~2 msec) responses to electrical stimuli (changes in Vj), and; slower (10-40 msec) responses to chemical stimuli (e.g. pH) that are also associated with initial channel formation [7]. We have applied molecular approaches to address the question of whether these kinetic differences reflect fundamentally different gating mechanisms.

We have noted similar changes in the response of other connexins to applied Vj (i.e. increases in conductance compared to the decreases seen in wt channels) in very different mutant constructs involving chimera between the extracellular loops of Cx43 and the transmembrane and cytoplasmic domains of Cx40 (described in [11]). These mutants were originally constructed for the purposes of defining determinants in the specificity of hemichannel docking. In this regard, the results indicated that specificity does not lie exclusively at the level of primary sequence, but also depends on the tertiary structure of the extracellular loops that is in part determined by the transmembrane and cytoplasmic domains (see abstract by Zhu et al., also [11]). However, most of these constructs also displayed responses to Vj that were in part reminiscent of the proline mutant discussed above (Fig 1 D-F). Although analysis of all of these mutants is needed at the single channel level, this result does implicate the docking domains of connexins in the conformational response leading to voltage gating of the channel. This leads to a model of voltage gating, consistent with the earlier studies [9 & 10], as a "global" conformational change of the channel that involves many domains of the connexin sequence other than the sensor and channel lining themselves. As noted above, the energetics of any such "global" rearrangement is likely to be affected by the docking interaction between connexons, as seen here in the responses of these chimeras of Cx40 and 43.

Figure 1: Current traces from oocyte pairs injected with cRNA for: (A) Cx26/Cx26; (B) Cx26 P87L/Cx26; (C) Cx26 P847L,V84P/Cx26; (D)-(F) Cx43 paired with various chimera of Cx40 (thin line) and Cx43 (thick line). These mutations (B,D,E & F) all cause a reversal in the response to changes in Vj (increase instead of decrease in conductance). The effect of the P87 mutant could be reversed by the reintroduction of P at 84 (C).

In sharp contrast to the case with voltage gating described above, the closure of gap junctional channels in response to cytoplasmic acidification has been associated with discrete domains. Demonstration that the truncated C-terminal domain of Cx43 could still mediate gating of the channel led to the proposal that it serves as a "particle" in a "particle-receptor" model of gating [12] analogous to the "ball and chain" inactivation gate of K⁺ channels [1]. This model, however, may not be universal, as the less sensitive pH responses seen in Cx32, and truncated forms of Cx43, do not appear to require the C-terminal region, but may involve close interactions between its membrane proximal regions and the cytoplasmic loop [13]. In order to further explore the factors that mediate this gating, we have developed a novel, paired oocyte perfusion/voltage clamp chamber. The two oocytes are maintained in electrically isolated chambers, communicating only through a small hole. This allows one of the oocytes to be cut open for perfusion with a variety of pseudo-intracellular solutions, while the other oocyte can remain intact with an independently controlled extracellular environment.

Previous studies using other perfused systems where the connexin composition was unknown had indicated that the gap junctions could respond to changes in pH and Ca⁺⁺ independently [14], but with significantly reduced sensitivity compared to their synergistic action [15]. In the current system, where the connexin composition is controlled and the perfusion conditions can be monitored more readily, we were surprised to find that perfusion resulted in complete loss of responsiveness of Cx43 and Cx32 to pH (Fig. 2A) and Ca⁺⁺ (data not shown) changes. This was demonstrated not to be due to irreversible changes/denaturation of the gap junction upon perfusion, as the responsiveness to reduced pH or elevated Ca⁺⁺ on the intact side was fully retained (Fig 2A). Furthermore, it was found that the response of Cx43 to pH could be restored with the addition of an oocyte cytoplasmic extract (Fig 2B). The somewhat slower kinetics and incompleteness of this response is likely to stem from the dilution of this extract on its reintroduction to the system. The activity appears to be heat labile, and may be different for Cx43 and 32, as the latter appears less stable. Possible candidates for this factor(s) abound, ranging from various Ca⁺⁺ binding proteins to kinases, etc. However, even without their definitive identification, it is clear that pH gating is more complex than previously thought, as the connexin cannot be considered in isolation.

Figure 2: (A) A novel chamber that allows intracellular perfusion of one of a pair of coupled Xenopus oocytes was used to demonstrate the loss of pH induced closure of Cx43 junctions upon removal of the cytoplasmic milieu, although the reversible uncoupling in response to acidification was retained on the intact side. (B) Reversible pH induced uncoupling could be restored to the perfused side by addition of an oocyte cytoplasmic extract.

A logical extension of this model would be that the signal for triggering interactions of the "ball" with the rest of the channel would be addition of a phosphate by the v-src tyrosine kinase. However, contrary to this earlier work, we found that changing tyr 265 to phe had no effect on the inhibition by v-src (Table I). This was consistent with our earlier finding that truncations of the C-terminus at residue 257 only partially eliminated v-src inhibition of coupling. Preliminary studies on deletions of specific domains within the C-terminus (mutants kindly provided by Taffet and Delmar - SUNY, Syracuse) have supported this contention, and implicated the two putative SH3 target domains in the C-terminus (between residues 240 and 280) as playing a role in this process. This suggests that v-src mediated gating of Cx43 may also require interactions with other cytoplasmic factors, presenting even closer analogies to the mechanism deduced for pH gating. Based on initial co-precipitations with the Cx43 tail (Zhou et al, in preparation), pp60^v-src itself represents a possible candidate for such an interacting factor.

To date, the reasons for the discrepancy between these data and earlier studies that implicated tyr 265 in this process[21] remain to be elucidated. One possibility is that cRNA forv-src and Cx43 were simultaneously injected in the previous work. In contrast, our results reflect only inhibitory effects of v-src after stable expression levels of Cx43 had been reached, and thus should predominantly reflect effects on existing channels. Thus, the greater levels of inhibition that were correlated with tyrosine phosphorylation by Swenson et al.[16] may be related to effects on the biosynthesis of Cx43. Certainly, phosphorylation on serines has already been correlated with regulation of Cx43 assembly [17], so this would not be without precedent.

Table I: Fractional Inhibition of Various Cx43 Constructs upon Expression of pp60^v-src in Xenopus oocytes

Acknowledgments: We would like to express our thanks to LieXian Xu, Steve Taffett, Mario Delmar, Jurgen Schwarz and Klaus Willecke for providing several of the mutant constructs used in these studies. We would also like to thank Heather Kuruvilla and Joe Smith for providing preliminary data and Charles Fourtner for discussions on all of this work. Support was provided by NIH awards from the NCI, HLBI and GMS to BJN, a Max Planck Prize to BJN and Klaus Willecke, and an AHA Investigator award to BJN.

[1] C. Miller, 1990: annus mirabilis of potassium channels. Science 252 (1991) 1092-1096.

[2] E. Perozo et al., S4 mutations alter gating currents of Shaker K⁺ channels. Biophys. J. 66 (1994) 345- 354.

[3] N. Unwin, Acetylcholine receptor channel imaged in the open state. Nature 373 (1995) 37-43.

[5] R. Bruzzone et al., Connections with connexins: the molecular basis of direct intercellular signaling. Eur. J. Biochem. 238 (1996) 1-27.

[6] A.F. Lau et al., Regulation of connexin43 function by activated tyrosine protein kinases. J. Bioenerg. and Biomemb. 28 (1996) 359-368.

[7] F.F. Bukauskas et al., Biophysical properties of gap junction channels formed by mouse Cx40 in induced pairs of transfected human HeLa cells. Biophys. J. 68 (1995) 2289-2298

[9] J.B. Rubin et al., A domain substitution procedure and its use to analyze voltage dependence of homo- typic gap junctions formed by connexins 26 and 32. Proc. Nat. Acad. Sci. USA 89 (1992) 3820-3824.

[10] T.M. Suchyna et al., Identification of a proline residue as a transduction element involved in voltage gating of gap junctions. Nature 365 (1993) 847-849.

[11] S. Haubrich et al., Incompatibility of connexin 40 and 43 hemichannels in gap junctions between mammalian cells is determined by intracellular domains. Mol. Biol. Cell 7 (1995) 1995-2006.

[12] G.E. Morley et al., Intramolecular interactions mediate pH regulation of connexin43 channels. Biophys. J. 70 (1996) 1294-1302.

[13] X.G. Wang and C. Peracchia, Positive charges of the initial C-terminus domain of Cx32 inhibit gap junction gating sensitivity to CO₂ Biophysical J. 73 (1997) 798-806

[14]

[15]

[16] K.I. Swenson et al., Tyrosine phosphorylation of the gap junction protein connexin43 is required for the pp60^v-src-induced inhibition of communication. Cell Regulation 1 (1990) 989-1002.

[17] L.S. Musil and D.A. Goodenough, Biochemical analysis of Connexin43 intracellular transport, phosphorylation, and assembly into gap junctional plaques. J. Cell Biol. 115 (1991) 1357-1374.