Connexin43 mutations linked to skin disease have augmented hemichannel activity

Mutations in the gene (GJA1) encoding connexin43 (Cx43) are responsible for several rare genetic disorders, including non-syndromic skin-limited diseases. Here we used two different functional expression systems to characterize three Cx43 mutations linked to palmoplantar keratoderma and congenital alopecia-1, erythrokeratodermia variabilis et progressiva, or inflammatory linear verrucous epidermal nevus. In HeLa cells and Xenopus oocytes, we show that Cx43-G8V, Cx43-A44V and Cx43-E227D all formed functional gap junction channels with the same efficiency as wild-type Cx43, with normal voltage gating and a unitary conductance of ~110 pS. In HeLa cells, all three mutations also localized to regions of cell-cell contact and displayed a punctate staining pattern. In addition, we show that Cx43-G8V, Cx43-A44V and Cx43-E227D significantly increase membrane current flow through formation of active hemichannels, a novel activity that was not displayed by wild-type Cx43. The increased membrane current was inhibited by either 2 mM calcium, or 5 µM gadolinium, mediated by hemichannels with a unitary conductance of ~250 pS, and was not due to elevated mutant protein expression. The three Cx43 mutations all showed the same gain of function activity, suggesting that augmented hemichannel activity could play a role in skin-limited diseases caused by human Cx43 mutations.

. ILVEN is characterized by pruritic, erythematous, hyperkeratotic papules linearly distributed along Blaschko's lines, and is caused by somatic, rather than germline mutations 28 . The three Cx43 mutations linked so far to human genetic skin disease are all single amino acid substitutions whose functional consequences have not been fully characterized. Studies of several mutations in other connexins linked to skin disease have suggested a generalized role for altered hemichannel activity in the connexin skin disorders 20,[29][30][31][32] . Increases in hemichannel activity have been described for Cx26 mutations causing both Keratitis-Ichthyosis-Deafness (KID) syndrome and palmoplantar keratoderma (PPK) with deafness [33][34][35][36] . Cx30 mutations associated with hidrotic ectodermal displaysia (HED) elicited large currents and increased ATP leakage in cells, consistent with altered hemichannel function 37 . In a similar fashion, Cx31 mutations causing EKVP were shown to result in increased hemichannel activity, ATP leakage, and necrotic cell death when expressed in transfected cells 38 . Even in the case of Cx43, the initial characterization of the Cx43-G8V mutation linked to PPKCA1 in transfected cells showed increased whole cell membrane current, Ca 2+ influx, and cell death when compared to wild-type Cx43, which could have been mediated by an increase in the activity of hemichannels 22 .
Here we report the functional characterization of all three of the Cx43 mutations that have been linked to non-syndromic human skin disease thus far, Cx43-G8V, Cx43-A44V and Cx43-E227D. Using cRNA injected Xenopus oocytes or transiently transfected HeLa cells, we demonstrate that all three mutations form functional gap junctions as efficiently as wild-type-Cx43, with no obvious differences in voltage gating, unitary conductance, protein expression, or cellular localization. We further observed that Cx43-G8V, Cx43-A44V and Cx43-E227D all formed functional hemichannels with greatly increased membrane current flow, a feature not shared by wild-type Cx43. These results suggest that the augmentation of hemichannel function shared by all three mutations may play a role in the pathophysiology of human Cx43 mutations linked to skin disease.

Results
Wild-type and mutant forms of Cx43 show equivalent levels of protein expression in Xenopus oocytes. Equivalent protein expression of wild-type and mutant Cx43 was verified by western blotting of cRNA injected Xenopus oocytes. Immunoblotting for Cx43 revealed ~43 kDa bands of similar intensity in lanes corresponding to oocytes injected with either wild-type Cx43, Cx43-G8V, Cx43-A44V, or Cx43-E227D cRNAs. No Cx43 band was detected in H 2 O injected control cells (Fig. 1a). When the blot was re-probed for β-tubulin, it was detected at similar intensity in all samples, confirming equal protein loading (Fig. 1b). These data showed that levels of Cx43 protein synthesis following cRNA injection were similar for mutant and wild-type Cx43.
Cx43 mutations form functional gap junction channels in Xenopus oocytes. To examine the ability of Cx43 skin disease mutations to make gap junction channels, wild-type and mutant Cx43 were expressed in Xenopus oocyte pairs and gap junctional conductance was measured (Fig. 1c). Control oocyte pairs injected with water showed negligible conductance (0.28 µS), while cells with wild-type Cx43 channels had an average conductance of 29.6 µS. Cell pairs expressing Cx43-G8V, Cx43-A44V, or Cx43-E227D had mean conductance levels of 29.3, 33.4 or 30.3 µS respectively. Conductance levels of cell pairs expressing either mutant, or wild-type, forms of Cx43 were all significantly greater than the H 2 O injected negative controls (one way ANOVA, p < 0.05), but not significantly different from each other, consistent with the equivalent levels of Cx43 protein expression (Fig. 1a). Wild-type Cx43, Cx43-G8V, Cx43-A44V, and Cx43-E227D were readily detected in lanes corresponding to each injection condition with similar band intensities. (b) To confirm equal sample loading, blots were stripped and reprobed with an antibody against β-tubulin, which was present at comparable levels in all lanes. (c) Gap junctional conductance levels for cell pairs expressing either wild-type, or mutant, forms of Cx43 were all significantly greater than the H 2 O injected negative controls (p < 0.05, one way ANOVA), but not significantly different from each other. Values are the mean ± SEM. Thus, all three of the skin disease associated Cx43 mutations formed gap junction channels with macroscopic conductance levels equal to wild-type Cx43.
Skin disease mutations do not affect Cx43 voltage gating. To examine whether skin disease causing mutations altered the voltage gating properties of Cx43 gap junction channels, oocyte pairs were subjected to hyperpolarizing and depolarizing transjunctional potentials (V j ) while recording junctional currents (I j ). As previously described [39][40][41] , I j s of wild-type Cx43 gap junction channels decreased symmetrically in a voltage-dependent manner (Fig. 2a). Channels in cell pairs expressing Cx43-G8V, Cx43-A44V, or Cx43-E227D ( Fig. 2b-d) behaved in a similar fashion. Steady-state voltage gating was compared by plotting V j (normalized to the value at ±20 mV) against G j (Fig. 2e). Analysis of wild-type Cx43 showed an approximately symmetric decline in steady state conductance at increasing values of V j , as has been reported previously 33,40,42 . Cell pairs expressing Cx43-G8V, Cx43-A44V, or Cx43-E227D displayed steady state gating that strongly resembled that of wild-type Cx43.
Cx43-G8V, Cx43-A44V, and Cx43-E227D all form active hemichannels in unpaired Xenopus oocytes. To test for changes in hemichannel activity, cRNAs encoding wild-type and mutant forms of Cx43 were injected into single Xenopus oocytes that had also received antisense oligonucleotides directed against the endogenous Xenopus Cx38 43,44 . Membrane currents were recorded while the cells were clamped to different membrane voltages. Oocytes injected with H 2 O instead of connexin cRNA showed insignificant current flow at membrane voltages between −30 and +60 mV (Fig. 3a). We found that oocytes expressing wild-type Cx43 also showed negligible membrane current flow between −30 and +60 mV (Fig. 3b). In sharp contrast, Xenopus oocytes expressing Cx43-G8V, Cx43-A44V, or Cx43-E227D all displayed large outward hemichannel currents upon depolarization (Fig. 3c-e). To compare differences in hemichannel activity, mean currents were plotted against the membrane potential (Fig. 3f). H 2 O injected control cells, or wild-type Cx43 expressing cells showed minimal membrane currents at all tested voltages. The Cx43-G8V injected cells had much larger hemichannel currents than wild-type Cx43 injected cells at all tested membrane potentials. Cx43-A44V injected oocytes displayed membrane currents that were greater than wild-type Cx43 at membrane potentials ≥+20 mV. Cx43-E227D injected cells had larger membrane currents than wild-type Cx43 at membrane potentials ≥+10 mV. This increased membrane current implied the presence of augmented hemichannel activity by the Cx43 mutants associated with genetic skin disease.
Connexin hemichannels are known to be blocked by the presence of extracellular cations such as calcium and gadolinium 6,45,46 . To confirm that currents observed in single cells expressing the Cx43 mutations were mediated by connexin hemichannels, oocytes were successively stepped from −70 mV to +20 mV while being perfused with extracellular medium containing either 2 mM Ca 2+ , 0.2 mM Ca 2+ , or 5 µM Gd 3+ . In an example shown for a Cx43-E227D expressing oocyte (Fig. 4), hemichannel currents were initially suppressed by the presence of 2 mM Ca 2+ , and increased markedly when the extracellular calcium in the perfusate was reduced tenfold to 0.2 mM. Cx43-E227D hemichannel currents were also efficiently suppressed by perfusion with 5 µM Gd 3+ , and recovered rapidly when the Gd 3+ was washed out with medium containing 0.2 mM Ca 2+ . Finally, Cx43-E227D hemichannel currents were suppressed back to their initial levels by increasing the extracellular calcium back to 2.0 mM. On average, Cx43-E227D hemichannel currents were inhibited 90 ± 4% (mean ± SD, n = 3) following perfusion with 5 µM Gd 3+ . Similar results were obtained for Cx43-G8V (86 ± 9%, n = 3) and Cx43-A44V (87 ± 3%, n = 3). These data suggest that the increased membrane current seen in oocytes expressing mutant forms of Cx43 is due connexin hemichannel activity. Single hemichannels formed by skin disease mutations. The conductance and gating of hemichannels formed by mutant Cx43 subunits was explored using cell-attached or excised patch recordings. Xenopus oocytes expressing mutant Cx43 subunits showed large-conductance channels in symmetric 140 mM KCl. Figure 5 shows currents of single Cx43-G8V, Cx43-A44V and Cx43-E227D hemichannels obtained by applying ±70 mV voltage ramps. Open channel currents of Cx43-A44V and Cx43-E227D were largely linear over the voltage range whereas Cx43-G8V hemichannel currents exhibited slight inward rectification. Mean slope conductances, measured at V m = 0 mV, were 265 ± 8.6 for Cx43-G8V (n = 5), 232 ± 5.5 for Cx43-A44V (n = 6), and

Cx43 mutations also form gap junctions in transfected HeLa cells.
To determine mutant protein localization and function in mammalian cells, gap junctional communication-deficient HeLa cells were transiently transfected with the mutant forms of Cx43. Immunofluorescent staining verified protein expression and localization for Cx43-G8V (Fig. 7a,b), Cx43-A44V (Fig. 7c,d), and Cx43-E227D (Fig. 7e,f). All three mutations showed proper trafficking to the cell membrane, especially at the regions of cell-to-cell contact, as shown by punctate staining (white arrowheads). Thus, Cx43 mutant proteins were properly expressed, and junctionally targeted in mammalian cells.
The ability of Cx43 mutants to form functional gap junction channels was also analyzed by dual whole cell patch clamp in the transiently transfected HeLa cells (Fig. 7g). As expected 10,47 , untransfected HeLa cells failed to form gap junctions, with a mean conductance of 0.06 ± 0.03 nS. Consistent with our results obtained in paired Xenopus oocytes above (Fig. 1), the mean junctional conductance of Cx43-G8V, Cx43-A44V, and Cx43-E227D cell pairs were 11.3 ± 1.9, 9.5 ± 1.2, and 11.1 ± 1.0 nS, respectively, values that were more than two orders of magnitude greater than the untransfected cell pairs, but not statistically different from each other (p > 0.05, one way ANOVA). Previous biophysical studies have documented that wild-type Cx43 forms gap junction channels with a unitary conductance of ~110 pS 39,[48][49][50] . To determine if the Cx43-G8V, Cx43-A44V, and Cx43-E227D mutations altered single channel conductance, unitary currents were measured in poorly coupled cell pairs. An example of a single-channel current recorded from a cell pair transfected with Cx43-A44V is shown in Fig. 7h. At a transjunctional voltage of −60 mV, the single-channel was primarily in the open state, with a few transitions to the closed state. An amplitude histogram of the recording, illustrated on the right of the trace, showed a major peak at 6.8 pA, corresponding to the fully open state, and yielding a unitary conductance of 113 pS. The Cx43-G8V, Cx43-A44V, and Cx43-E227D mutations all displayed unitary conductance of gap junction channels similar to wild-type Cx43, with mean values (±SEM) of 110 ± 9 (n = 2), 113 ± 8 (n = 3), and 109 ± 8 pS (n = 3) respectively.

Discussion
We have functionally characterized three human Cx43 mutations linked to non-syndromic skin-limited genetic disease. Cx43-G8V, Cx43-A44V and Cx43-E227D all formed functional gap junctions as efficiently as wild-type Cx43, with similar voltage gating, unitary conductance, protein expression, and cellular localization. In addition, Cx43-G8V, Cx43-A44V and Cx43-E227D formed functional hemichannels that mediated greatly increased membrane current flow. We found that wild-type Cx43 failed to form ion-conducting hemichannels under physiological conditions, as has been previously reported 46,[51][52][53] . The shared gain of hemichannel function by Cx43-G8V, Cx43-A44V and Cx43-E227D suggests that augmented hemichannel activity could be a common feature of Cx43 mutations linked to skin disease.
In the original clinical reports of these mutations, Cx43-G8V was reported to target to cellular interfaces and support the intercellular passage of fluorescent dye as a GFP tagged construct 22 , consistent with our experimental data. In contrast, Cx43-A44V and Cx43-E227D failed to localize at cell-cell junctions, and aggregated in the Golgi apparatus when transfected into HeLa cells as hemagglutinin-tagged constructs, with no test of channel function in this study 24 . In contrast to this report, we found that untagged versions of both Cx43-A44V and Cx43-E227D targeted to cellular interfaces and supported gap junctional coupling in two different expression systems. In addition, the gap junction channels that we recorded had unitary conductance and voltage gating properties that would be expected for Cx43. We suspect that the hemagglutinin-tag, or possible high levels of overexpression, may have altered the improper localization reported by Boyden et al. 24 .
Our single channel studies further revealed that the unitary conductance of hemichannels formed by all three mutants is ~250 pS, close to twice that of their corresponding gap junction channels, as would expected from the series docking of two identical hemichannels in a cell-cell channel 8 . We did not observe single hemichannels in wild-type Cx43 injected oocytes, but our values of unitary conductance for the three different mutants are similar to the ∼220 pS unitary conductance previously reported for wild-type Cx43 hemichannels in HeLa cells 54,55 . In contrast, the gating characteristics of hemichannels formed by Cx43 mutants exhibit significant differences from previous reports of wild-type Cx43 hemichannels 55,56 . In solutions containing Ca 2+ /EGTA (free Ca 2+ concentrations <10 −7 M), we found that all three mutant hemichannels primarily resided in the fully open state at low to moderate inside negative voltages. All of the mutant hemichannel currents also exhibited gating to subconductance states at negative voltages, whereas depolarizing voltages produced only brief transitions to intermediate states. These hemichannel properties are markedly different from those of wild-type Cx43, which required depolarization exceeding +40 mV for their activation, and exhibited long-lived transitions to subconductance states at high positive voltages. Complete characterization the gating, ion permeability, and pharmacology of Cx43-G8V, Cx43-A44V, and Cx43-E227D hemichannels will require additional studies.
Cx43 is a widely expressed protein, and is present in the skin across the epidermis 21,57,58 . Most mutations in the GJA1 gene, encoding Cx43, result in oculodentodigital dysplasia (ODDD) 59 , a disorder that manifests with neuropathies, facial, dental, and digit abnormalities and very rarely skin disease 60 . Recently, GJA1 mutations were identified in patients with three distinct, non-syndromic, skin-limited diseases, who lacked any of the diagnostic features of ODDD. Cx43-G8V was found in three patients from two unrelated families with PPKCA1 22 , Cx43-A44V was detected in two unrelated patients, one with EKVP and the other with ILVEN 24,27 , and Cx43-E227D was identified in two unrelated patients with EKVP 24 . There is overlap in the clinical features presented in these disorders, suggesting that a common functional consequence, such as augmented hemichannel activity, could underlie the pathology resulting from the three distinct mutations. As described in the introduction, analysis of mutations in other connexins associated with epidermal disorders has suggested a general role for augmented hemichannel function in the pathology of skin disease 19,20 . Other recent studies have suggested that wild-type Cx43 hemichannel activity was promoted by Cx26 mutations through the formation of heteromeric hemichannels. The first study examined the KID syndrome mutation Cx26-S17F. Although Cx26-S17F was unable to form hemichannels or gap junction channels when expressed alone 61 , it showed significantly increased hemichannel activity when co-expressed with wild-type Cx43 35 . The second showed that two PPK mutations, Cx26-H73R and Cx26-S183F, both failed to form hemichannels when expressed alone in Xenopus oocytes. Like Cx26-S17F, co-expression of either Cx26 PPK mutant with Cx43 showed significantly increased hemichannel activity, compared to Cx43 alone. Co-immunoprecipitation showed that Cx43 was efficiently pulled down with either Cx26-H73R and Cx26-S183F, confirming the formation of heteromeric hemichannels 33 . In the present work, we found that expression of skin disease causing Cx43 mutations resulted in augmented membrane currents mediated by active hemichannels. The addition of the data for Cx43, to the cumulative reports on Cx26, Cx30 and Cx31 mutations, makes a strong case that increased hemichannel activity linked to connexin mutations associated with epidermal disorders may contribute to disease pathology.

Methods
Molecular cloning. Human Cx43 was cloned into pCS2 + 62 for functional studies as previously described 33 .
Hemichannel current recording. Whole cell hemichannel currents were recorded 24 hours after cRNA injection into Xenopus oocytes using a GeneClamp 500 amplifier operated by a PC-compatible computer using a Digidata 1440A interface and pClamp 10 software (Axon Instruments, Foster City, CA). Electrodes (1.5 mm diameter glass, World Precision Instruments, Sarasota, FL) were pulled to a resistance of 1-2 MΩ (Narishige, Tokyo, Japan) and filled with 3 M KCl, 10 mM EGTA, and 10 mM HEPES, pH 7.4. In most cases, cells were recorded in MB medium without added calcium 34 . Hemichannel current-voltage (I-V) curves were obtained by clamping cells at −40 mV and imposing voltage steps in 10 mV increments ranging from −30 to +60 mV. For perfusion experiments testing hemichannel block by cations, extracellular solutions were exchanged using a six channel perfusion valve control system and a slotted bath oocyte recording chamber (VC-6 and RC-1Z, Warner instruments, Hamden, CT).
For patch-clamp recordings of single-hemichannel currents, Xenopus oocytes were manually devitellinized in a hypertonic solution consisting of (in mM) 220 Na-aspartate, 10 KCl, 2 MgCl 2 , and 10 HEPES, and then placed in ND96 medium for recovery. Individual oocytes were moved to a recording chamber (RC-28; Warner Instruments) containing the patch pipette solution, which consisted of (in mM): 140 KCl, 1 MgCl 2 , 5 HEPES, 1 CaCl 2 , and 3 EGTA, pH adjusted to 8.0 with KOH. The bath compartment was connected via a 3-M agar bridge to a ground compartment containing the same IPS solution. Single-hemichannel records from voltage steps and ramps were leak subtracted by measuring the leak current during full-closing events and extrapolating linearly with voltage.
Recording of gap junctional conductance. In Xenopus oocyte pairs, junctional conductance (G j ) was measured by initially clamping both cells in a pair at −40 mV (a transjunctional potential (V j ) of zero). One cell was subjected to alternating pulses of ±20 mV and the current produced by the change in voltage was recorded in the second cell, which was equal in magnitude to the junctional current (I j ). Conductance was calculated by dividing I j by the voltage difference, G j = I j /(V1 − V2) 64 . Gating properties were determined by recording the junctional current in response to hyperpolarizing or depolarizing V j s in 20-mV steps. Steady-state currents (I jss ) were measured at the end of the voltage pulse. Steady-state conductance (G jss ) was calculated by dividing I jss by V j , normalized to ± 20 mV, and plotted against V j .
For recordings of single channel junctional currents, HeLa, or N2A cells were transfected with cDNA corresponding to the mutant connexins. Junctional currents were measured using the dual whole cell patch-clamp technique as described previously 65 . Single gap junction channel currents were visualized during washout of 100% CO 2 saturated media, which uncouples cells completely. All electrophysiological measurements were obtained using Axopatch 1D patch clamp amplifiers (Molecular Devices, San Jose, CA). Data were acquired by using pClamp 9.2 software; analysis was performed with pClamp 9.2. Currents were filtered at 0.5-1 kHz and sampled at 2-5 kHz.
Cell transfection. HeLa cells were grown to 50% confluence on 22 mm 2 coverslips and transfected with wild-type or mutant Cx43 using Lipofectamine 2000 (Invitrogen, Carlsbad, CA) as described 10,47,67 . To facilitate cell survival, the calcium concentration in the culture media was elevated to 4 mM by the addition of supplemental CaCl 2 24 hours after transfection.

Immunofluorescent staining of transfected cells. HeLa cells were fixed in 1% paraformaldehyde in
PBS 24-48 hours after transfection and blocked with 5% BSA dissolved in PBS with 0.02% NaN 3 and 0.1% Tx-100 added. Cells were immunostained with a polyclonal Cx43 antibody followed by a Cy3 conjugated goat anti-rabbit secondary antibody (Jackson ImmunoResearch, West Grove, PA). Cells were photographed on a BX51 microscope using a DP72 digital camera (Olympus America, Waltham, MA).
Statistical analysis. Differences in data sets were analyzed for statistical significance with Origin 6.1 software (Microcal Software, Northampton, MA). Multiple comparisons were done with one-way ANOVA. The data are presented as the mean ± SEM of the indicated number of experiments. Statistical significance was designated for analyses with p < 0.05.