CO2 directly opens hemichannels of connexin26 (Cx26) by carbamylating K125, thereby allowing salt bridge formation with R104 of the neighbouring subunit in the connexin hexamer. The formation of the inter-subunit carbamate bridges within the hexameric hemichannel traps it in the open state. Here, we use insights derived from this model to test whether the range of agonists capable of opening Cx26 can be extended by promoting the formation of analogous inter-subunit bridges via different mechanisms. The mutation K125C gives potential for nitrosylation on Cys125 and formation of an SNO bridge to R104 of the neighbouring subunit. Unlike wild-type Cx26 hemichannels, which are insensitive to NO and NO2−, hemichannels comprising Cx26K125C can be opened by NO2− and NO donors. However, NO2− was unable to modulate the doubly mutated (K125C, R104A) hemichannels, indicating that an inter-subunit bridge between C125 and R104 is required for the opening action of NO2−. In a further test, we introduced two mutations into Cx26, K125C and R104C, to allow disulfide bridge formation across the inter-subunit boundary. These doubly mutated hemichannels open in response to changes in intracellular redox potential.
The human genome contains 21 connexin genes . Connexins are one of the main gene families that can form gap junctions. Each gap junction is made of two hemichannels, comprising six connexin subunits, one in each plasma membrane of two adjacent cells. The hemichannels dock to form a continuous passageway linking the cytoplasm of the coupled cells that permits transfer of ions and small molecules. In addition to this canonical function of connexins as gap junctions, undocked hemichannels can also have physiological functions [2,3]. We have shown that hemichannels of three closely related β connexins, Cx26, 30 and 32, can be opened by CO2 . In the case of Cx26, we have established that this is a direct action in which CO2 binds covalently to K125 via a carbamylation reaction . The carbamylated residue can then form a salt bridge to the neighbouring subunit in the hexamer . This opens the hemichannel, and allows release of ATP, which is an important biological signalling molecule . The CO2-dependent opening of Cx26, and consequent release of ATP, is a key mechanism in the regulation of breathing by CO2 .
In this paper, we demonstrate that it is possible to increase the range of molecules that can alter the gating of Cx26 via a rational mutational strategy. Given that formation of a bridge between residues 125 and 104 in adjacent subunits seems vital for CO2-mediated hemichannel opening , we have explored whether alternative mechanisms of bridge formation might also be effective. NO is an important gaseous signalling molecule that is produced by cells and can act via activation of a guanylate cyclase. NO can also signal via a nitrosylation reaction, whereby cysteine residues become converted to SNO groups (figure 1) . We have tested whether the mutation K125C has the ability to support NO/NO2−-mediated channel opening. If C125 were to be nitrosylated, the resulting SNO group would have the potential to form an inter-subunit salt bridge with Arg104  (figure 1) and, if our model is correct, Cx26K125C should be opened by NO/NO2−. We have additionally demonstrated, in hemichannels with the mutations K125C and R104C, that disulfide bridge formation between subunits will allow hemichannel opening in response to changes in intracellular redox.
3.1. Mutant connexin production
Cx26K125C and Cx26K125C,R104C mutants were produced from wild-type Cx26 in a Puc19 plasmid via the Quikchange protocol. The primers used were as follows: Cx26K125C forward, 5′ CGA AGA GAT CAA AAC CCA GTG CGT CCG TAT CGA AGG GTC CC 3′; reverse, 5′ GGG ACC CTT CGA TAC GGA CGC ACT GGG TTT TGA TCT CTT CG 3′; for Cx26K125C,R104A, a second set of primers was used to introduce the R104A mutation into Cx26K125C—forward, 5′ GGC CTA CCG GAG ACA CGA AAA GAA AGC GAA GTT CAT GAA GG 3′; reverse, 5′ CCT TCA TGA ACT TCG CTT TCT TTT CGT GTC TCC GGT AGG CC 3′; and for Cx26K125C,R104C a second primer set to introduce the R104C mutation into Cx26K125C—forward, 5′ CCG GAG ACA CGA AAA GAA ATG CAA GTT CAT GAA GGG AGA G 3′; reverse, 5′ CTC TCC CTT CAT GAA CTT GCA TTT CTT TTC GTG TCT CCG G 3′. The DNA produced was processed using DPN1 digest and bacterial transformation. DNA mutation was confirmed by genomic sequencing. The Cx26 mutants were subcloned into a pCAG-GS vector for mammalian cell transfection.
3.2. HeLa cell culture
HeLa cells were cultured by standard methods in DMEM and 10% FCS with addition of 3 mM CaCl2. For experimentation, cells were plated onto coverslips at a density of 5 × 104 cells per well. Transient transfections were performed using the Genejuice protocol.
3.3. Dye loading protocols
3.3.1. Solutions used
Control saline: 124 mM NaCl, 26 mM NaHCO3, 1.25 mM NaH2PO4, 3 mM KCl, 10 mM d-glucose, 1 mM MgSO4, 1 mM CaCl2.
Zero Ca2+ saline: 124 mM NaCl, 26 mM NaHCO3, 1.25 mM NaH2PO4, 3 mM KCl, 10 mM d-glucose, 1 mM MgSO4, 1 mM MgCl2, 1 mM EGTA.
All solutions were saturated with 95% O2/5% CO2.
Figure 2 summarizes in diagrammatic form the dye loading protocols used to assess NO2−/NO and redox sensitivity. The protocols are based on those that we have previously validated for determining the CO2 sensitivity of Cx26WT hemichannels [4,5,8].
3.3.2. Sensitivity to NO2− and NO donors
Coverslips plated with HeLa cells transiently transfected with Cx26K125C, Cx26K125C,R104A or Cx26WT were exposed to control saline containing 200 µM carboxyfluorescein (CBF) and 100 µM NaNO2 or the NO donors 2-(N,N-diethylamino)-diazenolate 2-oxide (DEA-NONOate, 100 or 10 µM) or sodium nitroprusside (SNP, 100 µM), for 10 min. This was followed by saline with 200 µM CBF for 5 min to allow the hemichannels to close in the absence of a stimulus and thus reduce loss of dye from the cells through any open hemichannels, and a 30 min wash with saline (in the absence of CBF) to ensure that all dye was removed from the outside of the cells. The cells were then immediately imaged to quantify dye loading.
To establish whether the guanylate cyclase blocker H-[1,2,4]oxadiazolo[4,3-a]quinoxalin-1-one (ODQ) could affect the modulation of Cx26K125C by DEA-NONOate, ODQ at 3 µM was pre-applied in control saline for 15 min, and present during the application of the DEA-NONOate in the presence of CBF. This was then followed by the washing procedure as described above.
The control comparison used to establish any baseline loading in the absence of a stimulus followed exactly the same protocol, but no NaNO2 or NO donors were present. Thus, the HeLa cells were exposed to 200 µM CBF in control saline for 15 min, followed by 30 min of washing.
The zero Ca2+-positive control involved exposure of Cx26 (mutant or wild-type) expressing cells to saline containing 200 µM CBF in zero Ca2+ saline for 10 min. This was followed by control saline with 200 µM CBF for 5 min and 30 min of washing with saline. We have previously shown that non-transfected HeLa cells, or HeLa cells transfected with an empty vector have no responsiveness to the zero Ca2+ stimulus [5,8]. Only batches of transfected cells that successfully demonstrated dye loading in response to zero Ca2+ were included in the analysis.
3.3.3. Redox sensitivity
We used two solutions to manipulate intracellular redox: ‘reduced saline’ contained 4 mM N-acetyl-l-cysteine (NAC) and ‘oxidized saline’ contained 1 mM l-buthionine sulfoximine (BSO). The HeLa cells were exposed for 5 min to reduced saline, and then transferred to oxidized saline plus 200 µM CBF for 10 min. They were then washed with reduced saline plus CBF for 5 min before being washed for 30 min in reduced saline.
The no-stimulus control to assess baseline loading was exposure of the cells to 200 µM CBF in reduced saline for 15 min, followed by 30 min of washing in reduced saline.
Each experiment was independently repeated (i.e. separate transfections on different batches of cells on different days) at lease five times as noted in the relevant figure legends.
3.4. Imaging and analysis
The cells were imaged in their living state by epifluorescence (Scientifica Slice Scope, Cairn Research OptoLED 470 nm illumination, 60× water Olympus immersion objective, NA 1.0, Hamamatsu ImageEM EMCCD camera, Metafluor software). Using ImageJ, the extent of dye loading was measured by drawing a region of interest (ROI) around individual cells and calculating the mean pixel intensity for the ROI. The mean pixel intensity of the background fluorescence was also measured in a representative ROI, and this value was subtracted from the measures obtained from the cells. All of the images displayed in the figures reflect this procedure in that the mean intensity of the pixels in a representative background ROI has been subtracted from every pixel of the image. At least 40 cells were measured in each condition per independent repetition. The mean pixel intensities from all independent repetitions were plotted in one cumulative probability distribution. Statistical comparisons of dye loading under different conditions were performed by comparing the individual median values from each independent repetition via the Mann–Whitney U-test. As we have used non-parametric statistical comparisons, we show the median values together with the upper and lower quartiles in the figures.
3.5. Patch clamp recordings
Coverslips containing non-confluent cells were placed into a perfusion chamber at 28°C in sterile filtered standard saline. Standard patch clamp techniques were used to make whole cell recordings and outside out or inside out membrane patch recordings. For whole cell patch clamp, the intracellular fluid in the patch pipette contained: K-gluconate 130 mM, KCl 10 mM, EGTA 10 mM, CaCl2 2 mM, HEPES 10 mM, pH 7.3, and adjusted to 295 mOsm with pure H2O (dilution no more than 2%). All whole-cell recordings were performed at a holding potential of −40 mV with regular steps of 5 s, every 10 s, to −50 mV to assess whole cell conductance.
4.1. The mutation K125C creates an NO2− and NO-sensitive hemichannel
We examined the gating of Cx26 hemichannels via an established assay, whereby loading of CBF into Cx26-expressing HeLa cells is a measure of hemichannel opening . We found that 100 µM nitrite caused dye loading in Cx26K125C-expressing HeLa cells (p = 0.008 compared with background control, figure 3a,b,d). The extent of dye loading in 100 µM nitrite was very similar to that caused by the zero Ca2+ positive control, suggesting that the nitrite stimulus was very effective at opening the hemichannels (figure 3d).
Importantly, we found that HeLa cells expressing wild-type Cx26 had no endogenous sensitivity to nitrite, as no increase in dye loading was seen (figure 3c,e). As a positive control to test for functional hemichannels, we used a zero Ca2+ challenge, which we have previously shown has no effect on dye loading in parental HeLa cells which have no connexin expression [5,8]. The zero Ca2+ stimulus was very effective at causing dye loading in HeLa cells expressing wild-type Cx26 (figure 3c,d,e), demonstrating that functional wild-type hemichannels are present. The lack of effect of NO2− must therefore be due to its inability to open the wild-type Cx26 hemichannels. Taken together, these results demonstrate that wild-type Cx26 does not respond to NO2− with channel opening, but that this new functionality can be conferred on Cx26 by the mutation K125C.
We next tested whether R104 was required for the modulatory action of NO2−, by creating the doubly mutated variant Cx26K125C,R104A. We have previously demonstrated that the Cx26R104A is CO2 insensitive as the loss of the arginine residue at this location prevents formation of the inter-subunit salt bridge with the carbamylated K125. In this doubly mutated variant, if C125 were to be nitrosylated it would be unable to form a salt bridge to A104. We found that HeLa cells expressing Cx26K125C,R104A did not exhibit dye loading in response to NO2− (figure 4a,b). However, the positive control of the zero Ca2+ stimulus evoked dye loading indicating that functional hemichannels were present (figure 4a,b). Overall, while Cx26K125C hemichannels exhibited robust sensitivity to NO2−, neither Cx26WT nor Cx26K125C,R104A hemichannels displayed any sensitivity to NO2− (figure 4c). Thus, we conclude that both C125 and R104 are required for nitrite sensitivity.
NO donors are more usually used to demonstrate activation of the NO signalling pathway and nitrosylation. We therefore additionally tested two NO donors, DEA-NONOate and SNP. Both of these donors at 100 µM caused dye loading into HeLa cells expressing Cx26K125C (figure 5a). However, HeLa cells expressing wild-type Cx26 showed no such NO-evoked dye loading (figure 5b), indicating that this is a specific effect of the NO donors on HeLa cells expressing Cx26K125C. We also tested DEA-NONOate at a lower concentration (10 µM) and found that this too caused dye loading into cells expressing Cx26K125C (p = 0.016, n = 5, figure 5c). However, the extent of the dye loading was less compared with both the higher dose of DEA-NONOate and the zero Ca2+ control (figure 5c).
In addition to nitrosylation, NO can signal via the activation of guanylate cyclase and the cGMP-dependent protein kinase. We therefore tested whether ODQ, a blocker of guanylate cyclase, affected the modulatory action of DEA-NONOate on HeLa cells expressing Cx26K125C. We found that 3 µM ODQ did not affect the degree of dye loading caused by 10 µM DEA-NONOate (figure 5c,d). We therefore conclude that guanylate cyclase does not contribute to the modulatory action of the NO donors on Cx26K125C.
4.2. The mutation K125C, R104C creates a redox-sensitive hemichannel
Disulfide bridges represent a further way in which a bridge between subunits at residues 125 and 104 could form. A disulfide bridge would be sensitive to intracellular redox as this would potentially reduce the disulfide bridge to break it apart or oxidize the cysteine SH groups to promote bridge formation. We therefore made a further variant of Cx26, Cx26K125C,R104C, to test whether we could create a redox-sensitive hemichannel.
We used l-BSO and NAC to manipulate the redox state of HeLa cells via altered intracellular concentrations of glutathione . BSO inhibits γ-glutamylcysteine synthase, an enzyme essential for glutathione synthesis . BSO will therefore cause a decrease in the concentration of intracellular glutathione and will make the cell relatively oxidized and promote disulfide bridge formation . Conversely, NAC, by acting as a precursor, can increase the intracellular concentration of cysteine, and thus enhance intracellular glutathione concentrations . NAC will therefore make the cell more reduced and tend to break disulfide bridges .
We used dye loading in the presence of NAC as a reduced control, and transferred the cells to solutions containing BSO as the oxidized stimulus. Cells exposed to the reduced saline did not exhibit enhanced dye-loading (figure 6). However, when cells were exposed to the oxidized saline, there was a big increase in dye loading compared with control experiments (p = 0.021, figure 6). This redox sensitivity is not a property of wild-type Cx26, as HeLa cells stably expressing Cx26 showed no redox-sensitive dye loading (figure 6).
4.3. Patch clamp recordings demonstrate the kinetics of agonist sensitivity
To confirm the dye loading experiments, we performed whole cell patch clamp recordings from HeLa cells expressing wild-type Cx26, Cx26K125C and Cx26K125C,R104C. HeLa cells expressing wild-type Cx26 reliably responded to a CO2 challenge (figure 7). This response was rapid, having a time course that could be fitted with an exponential function with a time constant of around 5 s. HeLa cells expressing wild-type Cx26 showed no conductance changes in response to NO2−, BSO or NAC (figure 7). Thus, as we saw with dye loading, the wild-type Cx26 hemichannel has no sensitivity to these compounds.
HeLa cells expressing Cx26K125C responded to 100 µM NO2− (figure 7a,b). Interestingly, this response was some 20 times slower in onset than the response of wild-type Cx26 to CO2 (figure 7e). HeLa cells expressing Cx26K125C,R104C exhibited a slow conductance increase when exposed to BSO (figure 7c,d). When transferred to NAC, the conductance increase initially accelerated before diminishing back towards the previous baseline (figure 7c). This initial acceleration of the conductance change may be due to the continued presence of BSO inside the cell, and the time required to wash it out, before the depleted intracellular glutathione levels can be replenished sufficiently to reduce intracellular redox and thus close the hemichannel.
Interestingly, removal of NAC revealed a slow and reversible conductance increase, suggesting that the native redox state of the HeLa cells is slightly oxidized compared with the highly reduced state induced by inclusion of NAC (figure 7c,d). The increase in conductance following removal of NAC also had a time constant some 20 times slower than the response of wild-type Cx26 to CO2 (figure 7e).
The residue 125 is near the intracellular loop and thus on the cytoplasmic side of the plasma membrane. The dynamics of hemichannel opening will therefore result from a combination of how quickly the agonist can permeate into the cell to reach the binding site on residue 125 and the propensity for the covalent reaction at residue 125 to occur. For the redox sensitivity, there is the additional complication that NAC and BSO act indirectly to change the intracellular concentration of glutathione, which then alters intracellular redox state to either make or break the disulfide bridge between C125 and C104 of adjacent subunits. It is therefore not surprising that the new agonist sensitivities that we have introduced into the mutated channels have considerably slower kinetics than the native sensitivity to CO2. Nevertheless, our results with BSO and NAC imply that glutathione metabolism is highly dynamic within the HeLa cells and is consistent with other observations of glutathione synthesis in these cells .
To the best of our knowledge, this is the first time that novel agonist sensitivity has been introduced into connexin hemichannels. With our hypothesis-driven mutational strategy, we have converted wild-type Cx26 from a CO2-sensitive hemichannel to an NO/NO2−-sensitive hemichannel in Cx26K125C and a redox-sensitive hemichannel in Cx26K125C,R104C (figure 1). Crucially, wild-type Cx26 hemichannels do not exhibit sensitivity to either NO donors, NO2− or intracellular redox. This contrasts with some other connexin and pannexin hemichannels that can be regulated by NO and intracellular redox [13–15]. Our results are important in two respects. First, they provide further support for our hypothesis that bridges involving residues 125 and 104 of adjacent subunits can gate the opening of Cx26. We initially proposed this hypothesis for the gating of Cx26 by CO2 . By introducing mechanistically different bridges at the same location in the protein, we have shown that alternative agonists can open the hemichannel, providing they promote inter-subunit bridge formation between residues 125 and 104, even if this is via a different mechanism. Second, our new evidence further strengthens the concept of hemichannels as agonist-gated channels somewhat analogous to ligand-gated receptors. In a traditional ligand-gated receptor, the ligand binds and elicits a conformational change, which allows ion flow through the receptor channel. For hemichannels, the ligand or agonist binds to cause conformational change, and the opening of the hemichannel allows the fluxes of small molecules in addition to ions. ATP, an important physiological signalling molecule in many organ systems including the brain, is known to permeate connexin hemichannels [4,16,17]. Cx26 and related β connexins, which are opened by CO2 [4,5,8], and other hemichannels such as Cx37, Cx40 and Cx43, which are opened by NO , therefore provide mechanisms for physiologically important ligand-gated ATP release.
In wild-type Cx26, the available crystal structure  suggests that K125 and R104 of adjacent subunits are only about 6 Å apart. Once carbamylation has occurred this distance will fall to about 3 Å, which is suitable for formation of a salt bridge. In Cx26K125C C125 is 10.5 Å from R104 in the neighbouring subunit (figure 1), and with SNO formation this distance would fall to about 7–8 Å. Ordinarily, this would be too far for interaction between the SNO group and R104. Similarly, in Cx26K125C,R104C the two introduced cysteine residues would be nearly 16 Å apart—a gap far too large for disulfide bridge formation. Nevertheless, our data strongly suggest the formation of an SNO bridge between C125 and R104, and a disulfide bridge between C125 and C104. This implies considerable flexibility in the protein so that these residues can come into sufficient proximity to allow interaction and bond formation. Our elastic network models  based on the crystal structure  suggest that Cx26 is highly flexible with harmonic motions of the major alpha helices that could bring these residues into close apposition and permit bond formation. These data therefore provide additional support for our previous elastic network models. The large distance of these interactions may also contribute to their slow formation, and be an additional reason why NO2− and intracellular redox cause rather slow channel opening.
Finally, we have generated molecules with new functionality that can, respectively, give NO/NO2− and intracellular redox-mediated ATP release. The closest precedent to this change of agonist specificity for a receptor is the creation of the DREADD receptors from the muscarinic receptor . DREADD receptors can be opened by the synthetic agonist clozapine-N-oxide, and have been highly useful analytical tools to test the roles of neuronal populations [21,22]. Our modified Cx26 hemichannels may conceivably be useful in the analysis and perturbation of NO and redox signalling.
The datasets supporting this article have been uploaded as part of the electronic supplementary material.
We thank the MRC for support and the Wellcome Trust for a Vacation Scholarship to D.C.
- Received November 10, 2014.
- Accepted January 22, 2015.
© 2015 The Authors. Published by the Royal Society under the terms of the Creative Commons Attribution License http://creativecommons.org/licenses/by/4.0/, which permits unrestricted use, provided the original author and source are credited.