Induction of cell death and gain-of-function properties of connexin26 mutants predict severity of skin disorders and hearing loss

Connexin26 (Cx26) is a gap junction protein that oligomerizes in the cell to form hexameric transmembrane channels called connexons. Cell surface connexons dock between adjacent cells to allow for gap junctional intercellular communication. Numerous autosomal dominant mutations in the Cx26-encoding GJB2 gene lead to many skin disorders and sensorineural hearing loss. Although some insights have been gained into the pathogenesis of these diseases, it is not fully understood how distinct GJB2 mutations result in hearing loss alone or in skin pathologies with comorbid hearing loss. Here we investigated five autosomal dominant Cx26 mutants (N14K, D50N, N54K, M163V, and S183F) linked to various syndromic or nonsyndromic diseases to uncover the molecular mechanisms underpinning these disease links. We demonstrated that when gap junction-deficient HeLa cells expressed the N14K and D50N mutants, they undergo cell death. The N54K mutant was retained primarily within intracellular compartments and displayed dominant or transdominant properties on wild-type Cx26 and coexpressed Cx30 and Cx43. The S183F mutant formed some gap junction plaques but was largely retained within the cell and exhibited only a mild transdominant reduction in gap junction communication when co-expressed with Cx30. The M163V mutant, which causes only hearing loss, exhibited impaired gap junction function and showed no transdominant interactions. These findings suggest that Cx26 mutants that promote cell death or exert transdominant effects on other connexins in keratinocytes will lead to skin diseases and hearing loss, whereas mutants having reduced channel function but exhibiting no aberrant effects on coexpressed connexins cause only hearing loss. Moreover, cell death-inducing GJB2 mutations lead to more severe syndromic disease.

Connexin26 (Cx26) is a gap junction protein that oligomerizes in the cell to form hexameric transmembrane channels called connexons. Cell surface connexons dock between adjacent cells to allow for gap junctional intercellular communication. Numerous autosomal dominant mutations in the Cx26encoding GJB2 gene lead to many skin disorders and sensorineural hearing loss. Although some insights have been gained into the pathogenesis of these diseases, it is not fully understood how distinct GJB2 mutations result in hearing loss alone or in skin pathologies with comorbid hearing loss. Here we investigated five autosomal dominant Cx26 mutants (N14K, D50N, N54K, M163V, and S183F) linked to various syndromic or nonsyndromic diseases to uncover the molecular mechanisms underpinning these disease links. We demonstrated that when gap junction-deficient HeLa cells expressed the N14K and D50N mutants, they undergo cell death. The N54K mutant was retained primarily within intracellular compartments and displayed dominant or transdominant properties on wild-type Cx26 and coexpressed Cx30 and Cx43. The S183F mutant formed some gap junction plaques but was largely retained within the cell and exhibited only a mild transdominant reduction in gap junction communication when co-expressed with Cx30. The M163V mutant, which causes only hearing loss, exhibited impaired gap junction function and showed no transdominant interactions. These findings suggest that Cx26 mutants that promote cell death or exert transdominant effects on other connexins in keratinocytes will lead to skin diseases and hearing loss, whereas mutants having reduced channel function but exhibiting no aberrant effects on coexpressed connexins cause only hearing loss. Moreover, cell death-inducing GJB2 mutations lead to more severe syndromic disease.
The GJB2 gene encoding connexin26 (Cx26) 2 has an estimated mutation prevalence of 3% in the general population (1).
Globally, an estimated 17.3% of hearing loss cases are linked to bi-allelic GJB2 mutations, highlighting the importance of Cx26 in hearing (1). In addition, numerous syndromic diseases exhibiting hearing deficits and a variety of skin abnormalities are linked to GJB2 missense mutations with autosomal dominant inheritance (2). Interestingly, some speculate that the pervasiveness of GJB2 mutations may result from a selective heterozygote advantage (1) conferred by subclinical epidermal thickening and a stronger cutaneous barrier (3). In humans, Cx26 is expressed in a variety of tissues and, not surprisingly, in several cell types in the cochlea (4) and in keratinocytes of the epidermis (5). Within these tissues, several other members of the connexin family are expressed, most notably Cx30 and Cx43, wherein mutations in their respective genes have also been implicated in syndromic diseases sharing some similar features (2,5,6).
Cx26 is a gap junction protein that oligomerizes in the cell to form hexameric transmembrane channels called connexons (7). Connexons that span the plasma membrane are called hemichannels and may allow a cell to pass small signaling molecules between the cytosol and the extracellular environment (7). However, when hemichannels from adjacent cells dock together, they form a single conduit called a gap junction channel, which connects the cytosol of these cells and facilitates gap junctional intercellular communication (GJIC) (7). ATP, inositol trisphosphate, and cations frequently pass through Cx26 gap junction channels and have been shown to play important roles in regulating cell proliferation and differentiation as well as maintaining ionic homeostasis within tissues (8,9).
The Cx26 polypeptide has four transmembrane domains, two extracellular loops, an intracellular loop, and cytosolic N and C termini. The N-terminal domain (amino acid residues 1-20) is suggested to play a major role in voltage sensing and channel gating (10). The extracellular loops (E1 and E2) (amino acid residues 41-75 and 155-192, respectively) are thought to be key domains for oligomerization and interchannel docking (10). Disease-causing point mutations have been documented in nearly every domain of the Cx26 polypeptide, and depending on the mutation and the motif that harbors the altered residue, variations can occur in connexin folding and trafficking, channel assembly, channel gating, half-life, degradation, and/or interactions between other co-expressed connexins (11). Some mutations have been shown to disrupt several connexin lifecycle characteristics (12), increasing the complexity of delineating how GJB2 point mutations cause diseases that affect one or more organs with varying severity.
In this study, we selected five autosomal dominant GJB2 missense mutations that result in single amino acid substitutions in various domains of the Cx26 polypeptide and are linked to an array of auditory and skin pathologies. The N14K mutation causes a disease that shares symptoms with Clouston syndrome and keratitis-ichthyosis-deafness syndrome (KIDS) (13), the D50N mutation leads to KIDS (14), the N54K mutation results in Bart-Pumphrey syndrome (15), and the S183F mutation causes palmoplantar keratoderma (PPK) and hearing loss (16). Finally, the M163V mutation is linked to moderate hearing loss only (17). Considering the pleiotropic nature of GJB2 mutations, we proposed that Cx26 mutants that give rise to similar clinical presentations would share common mechanisms of action.
Here we found that the N14K and D50N mutants leading to widespread erythrokeratoderma and severe hearing loss caused cell death, the N54K and S183F mutants leading to PPK and hearing loss had trafficking defects and reduced channel function, and the M183V mutant leading to hearing loss alone had reduced channel function. Last, all mutants linked to syndromic disease had transdominant effects on co-expressed connexins.

N14K and D50N mutants induce cell death in vitro
To assess the impact of Cx26 mutants on cellular health, specific Cx26 mutants were expressed in GJIC-deficient HeLa cells. HeLa cells expressing the GFP-tagged N14K and D50N mutants appeared unhealthy as early as 24 h post-transfection; therefore, we immunolabeled cleaved caspase-3 (CC3) to determine whether cells expressing these mutants were undergoing apoptosis (Fig. 1A). HeLa cells expressing the GFP-tagged N14K and D50N mutants were more extensively labeled for CC3 (Fig. 1B), similar to cells treated with staurosporine, suggesting that the expression of these mutants in HeLa cells triggers apoptosis in vitro. Expression of untagged N14K and D50N mutants also triggered apoptosis in HeLa cells (data not shown), eliminating the possibility that the GFP tag was responsible for the induction of cell death. Consequently, cells expressing the N14K or D50N mutants were deemed not sufficiently healthy for further mutant localization or functional studies.

N54K and S183F mutants have trafficking defects and impaired dye-transfer ability
Cx26 mutants associated with hearing loss and various skin diseases were expressed in HeLa cells to examine how a point mutation in Cx26 may affect its trafficking, cellular localization, and channel function. Cx26 and the M163V mutant formed abundant gap junction plaques at cell-cell interfaces (Fig. 2, A  and B). The N54K and S183F mutants were retained primarily in intracellular compartments (Fig. 2C), partially colocalized with protein-disulfide isomerase (PDI) labeling of the endo-plasmic reticulum ( Fig. 2A), and formed fewer gap junction plaques (Fig. 2B). Likewise, HeLa cells engineered to express untagged Cx26 mutants displayed cellular localizations similar to HeLa cells expressing GFP-tagged mutants (data not shown). The ability of the disease-linked mutants to form functional gap junction channels was assessed by quantifying intercellular transfer of microinjected Alexa Fluor 350 in HeLa cells expressing Cx26 or the various mutants. Cx26 had nearly 100% incidence of dye transfer (Fig. 2D), whereas cells expressing the M163V mutant passed dye ϳ40% of the time. Cells expressing either the N54K or S183F mutants were essentially unable to establish gap junction channels capable of dye transfer (Fig. 2, D and E).

Cx26 mutants exhibit dominant-negative effects on Cx26 function
To assess the distribution and function of Cx26 mutants in conditions where both wild-type and mutant Cx26 are A, HeLa cells expressing GFP-tagged Cx26, N14K, and D50N mutants (green) were immunolabeled to highlight CC3 (red) expression in cells undergoing apoptosis. Nuclei were stained with Hoechst (blue). Cells treated with staurosporine were used as a positive control for apoptotic cells. B, a high percentage of apoptotic cells expressing N14K and D50N mutants were quantified and compared with Cx26-expressing cells. One-way ANOVA; ****, p Ͻ 0.0001; n ϭ 3. Scale bar, 40 m. Error bars, S.D. expressed in the same cell, we engineered HeLa cells to express RFP-tagged Cx26 and GFP-tagged Cx26 mutants at approximately a 1:1 ratio. Cx26 and the M163V mutant formed numerous gap junction plaques at the cell surface and were highly colocalized with Cx26-RFP (Fig. 3, A and B). The N54K mutant was mostly distributed in intracellular compartments ( Fig. 3C) but, surprisingly, was also able to traffic to the cell surface and form intermixed gap junction plaques when co-expressed with Cx26-RFP (Fig. 3, A and B). The S183F mutant remained distributed mostly in intracellular compartments ( Fig. 3C) but was also found in several plaques when co-expressed with Cx26 (Fig. 3, A and B). These findings suggest that all mutants were more readily assembled into a gap junction plaque when co-expressed with Cx26. Interestingly, all mutants exhibited a more pronounced intracellular distribution compared with Cx26, but only the N54K mutant significantly increased the intracellular localization of co-expressed Cx26 (Fig. 3C). To further examine the interaction between the Cx26 mutants and Cx26, we quantified Alexa Fluor 350 dye transfer between HeLa cells that co-expressed the mutants and Cx26 (Fig. 3D). Cells expressing M163V and S183F mutants reduced the overall functional gap junctional status of Cx26. However, the N54K mutant abolished the functional gap junctional status of coexpressed Cx26, as assessed by the ability to pass a fluorescent dye (Fig. 3D).

N54K and S183F mutants display transdominant effects on Cx30
Because Cx26 and Cx30 are co-expressed in the same keratinocytes as well as several cell types of the inner ear (4), we engineered HeLa cells to express RFP-tagged Cx30 and GFPtagged Cx26 mutants to assess potential transdominant inter- A, HeLa cells expressing GFP-tagged Cx26, N54K, M163V, or S183F mutants (green) were immunolabeled for PDI (red), a resident protein of the endoplasmic reticulum, and stained with Hoechst (blue) to denote nuclei. Cx26 and the M163V mutant formed numerous gap junction plaques at cell-cell interfaces (arrows). N54K and S183F mutants were mostly located within the cell, whereas the S183F mutant formed few gap junction plaques. B, the N54K and S183F mutants, but not the M163V mutant, formed fewer gap junction plaques compared with Cx26. One-way ANOVA; **, p Ͻ 0.01; n Ն 13. C, the N54K and S183F mutants, but not the M163V mutants, also displayed increased intracellular fluorescence. One-way ANOVA; **, p Ͻ 0.01; n ϭ 6. D, Alexa Fluor 350 dye transfer in pairs or clusters of HeLa cells expressing N54K and S183F mutants was negligible, whereas ϳ40% of the cells expressing the M163V mutant transferred dye. E, example images of dye-transfer experiments showing successful (Cx26) and unsuccessful dye transfer (N54K). One-way ANOVA; ****, p Ͻ 0.0001; **, p Ͻ 0.01; n ϭ 3. The number of cells that were microinjected to test for dye transfer in each case is noted in D. Ctrl, untransfected HeLa cells. Scale bar (A and E), 20 m. Error bars, S.D.
actions. Cx26 as well as the M163V and S183F mutants were able to form numerous gap junction plaques at the cell surface and often co-localized with Cx30 (Fig. 4, A and B). The N54K mutant formed very few plaques and remained in an endoplasmic reticulum-like distribution pattern but also appeared to impair the ability of Cx30 to form abundant gap junction plaques (Fig. 4, A and B). Whereas the N54K and S183F mutants were typically found in intracellular compartments, only the N54K mutant was able to significantly retain co-expressed Cx30 in intracellular compartments (Fig. 4C). Together these findings suggest the N54K mutant exhibits a transdomi-nant effect on Cx30 trafficking. Next we determined whether the Cx26 mutants exhibited transdominant effects on Cx30 channel function using Alexa Fluor 350 dye-transfer experiments (Fig. 4D). Microinjected HeLa cells expressing Cx30 alone or co-expressing Cx30 and Cx26 had a nearly 100% incidence of dye transfer. However, cells co-expressing Cx30 and either the N54K or S183F mutants had a significantly reduced ability to pass dye, indicating that these mutants had a transdominant effect on Cx30 channel function. Interestingly, the M163V mutant did not significantly reduce the ability of Cx30positive cells to pass dye. A, HeLa cells expressing GFP-tagged Cx26 or N54K, M163V, and S183F mutants (green) together with Cx26-RFP (red) were stained with Hoechst (blue). All mutants formed gap junction plaques at cell-cell interfaces (arrows) and colocalized with Cx26-RFP. B, only the N54K mutant formed fewer gap junction plaques than Cx26, and none of the mutants significantly impaired the formation of wild-type Cx26 gap junction plaques. One-way ANOVA; *, p Ͻ 0.05; n Ն 13. C, all mutants displayed increased intracellular fluorescence; however, only the N54K mutant was able to significantly increase the intracellular fluorescence of wild-type Cx26. One-way ANOVA; ****, p Ͻ 0.0001; *, p Ͻ 0.05; n ϭ 6. D, untransfected control cells (Ctrl) or cells expressing the N54K mutant together with Cx26 failed to pass microinjected Alexa Fluor 350 dye. Cells expressing S183F and M163F mutants together with Cx26-RFP had reduced incidences of dye transfer compared with cells expressing Cx26-GFP. One-way ANOVA; ****, p Ͻ 0.0001; n ϭ 3. The number of cells that were microinjected in each case is noted in D. Scale bar, 20 m. Error bars, S.D.

N54K displays mild transdominant effect on endogenous Cx43
It has been demonstrated that some Cx26 mutants have the abnormal ability to interact with Cx43 (12). Therefore, we engineered Cx43-positive rat epidermal keratinocytes (REKs) to express Cx26 mutants to determine their potential for transdominant effects on endogenous Cx43 in a tissue-relevant cell. Only the N54K mutant formed fewer gap junction plaques than Cx26, but none of the mutants impaired the formation of Cx43 plaques (Fig. 5, A and B). The S183F mutant extensively colo-calized with Cx43, consistent with its reported interaction with Cx43 (26). The N54K and S183F mutants were frequently found in intracellular compartments, and interestingly, these mutants also increased the intracellular localization of co-expressed Cx43 (Fig. 5C), suggesting a possible transdominant interaction with endogenous Cx43. REKs engineered to express Cx26 mutants were microinjected with Alexa Fluor 350, which readily passes through both Cx26 and Cx43 gap junction channels. Only cells expressing the N54K mutant exhibited a signif- , and S183F mutants (green) and RFP-tagged Cx30 (red) were stained with Hoechst (blue). S183F and M163V mutants formed abundant gap junction plaques at cell-cell interfaces (arrows) and colocalized with Cx30, whereas S183F mutants were also located within the cell. N54K mutants remained mostly within the cell, although they formed a few gap junction plaques (arrows). Only the N54K mutant assembled fewer gap junction plaques at the cell surface and also demonstrated an ability to reduce Cx30 gap junction plaque formation. One-way ANOVA; ***, p Ͻ 0.001; n Ն 22. C, the N54K and S183F mutants exhibited increased intracellular fluorescence, although only the N54K mutant increased the intracellular fluorescence from wild-type Cx30. One-way ANOVA; **, p Ͻ 0.01; *, p Ͻ 0.05; n ϭ 6. D, cells expressing the M163V mutant together with Cx30 had dye transfer abilities similar to cells expressing Cx26 and Cx30, or Cx30 alone, whereas the incidences of dye transfer in cells expressing the N54K and S183F mutants together with Cx30 were markedly reduced. One-way ANOVA; ****, p Ͻ 0.0001; n Ն 3. The number of cells that were microinjected in each case is noted in D. Scale bar, 20 m. Error bars, S.D.

Cx26 mutants predict disease severity
icant decrease in detectable dye transfer, suggesting that the N54K mutant had a modest transdominant-negative effect on endogenous Cx43 (Fig. 5D).

N54K and S183F mutants display reduced hemichannel function
Aside from GJIC and trafficking defects, mutant connexins may further disrupt hemichannel properties (18). We therefore used a propidium iodide (PI) dye uptake assay to assess the function of mutant hemichannels in normal, calcium-contain-ing extracellular solution (ECS ϩ Ca 2ϩ ) and divalent-cationfree extracellular solution (ECS Ϫ Ca 2ϩ ), which stimulates hemichannels to open (19). All mutant-expressing HeLa cells displayed minimal PI uptake in ECS ϩ Ca 2ϩ conditions, and similar to Cx26, the M163V mutant exhibited a nearly 100% incidence of PI uptake in ECS Ϫ Ca 2ϩ , indicating fully functional hemichannel activity (Fig. 6B). Interestingly, in ECS Ϫ Ca 2ϩ , cells expressing the N54K mutant demonstrated a highly variable incidence of PI uptake, whereas cells expressing the S183F mutant displayed no increase in PI uptake, Figure 5. The N54K mutant exhibits transdominant inhibition of endogenous Cx43 in REKs. A, REKs expressing GFP-tagged Cx26, N54K, M163V, and S183F mutants (green) with endogenous Cx43 immunolabeled in red were stained with Hoechst (blue). The N54K and S183F mutants were often found intracellularly but also formed gap junction plaques at the cell surface, whereas the M163V mutant exhibited a localization pattern similar to Cx26. B, the N54K mutant formed fewer gap junction plaques compared with Cx26; however, endogenous Cx43 gap junction formation was not impaired by any of the mutants. One-way ANOVA; **, p Ͻ 0.01; n Ն 24. C, the N54K and S183F mutants exhibited increased intracellular fluorescence and also increased the amount of intracellular fluorescence from wild-type Cx43. One-way ANOVA; ****, p Ͻ 0.0001; ***, p Ͻ 0.001; *, p Ͻ 0.05; n Ն 6. D, REKs expressing the N54K mutant had reduced dye transfer capabilities compared with untransfected REKs (Ctrl) and REKs expressing Cx26. One-way ANOVA; **, p Ͻ 0.01; n ϭ 3. The number of cells microinjected is noted in D. Scale bar, 20 m. Error bars, S.D.
indicating non-functional hemichannels (Fig. 6B). It is important to note that divalent-cation-free solution was used only to stimulate hemichannel opening. As a result, some cells exhibited a rounded morphology, which may indicate that the cells are beginning to become unhealthy (Fig. 6A).

Discussion
Due to a high carrier frequency of GJB2 mutations in the population and Ͼ100 disease-causing Cx26 mutants now identified in humans (20 -22), understanding how distinct GJB2 mutations cause disease is critical when considering any prospective therapies. The potential scope of Cx26 dysfunction is vast, considering the possibility of gain or loss of connexin functions that may present anywhere throughout the short Cx26 life cycle. Furthermore, the complexity of Cx26 dysfunction is compounded by its potential to interact with several co-expressed connexin binding partners in numerous different tissues, including the skin and inner ear. The pleiotropic nature of Cx26-linked diseases raises obvious questions as to how specific GJB2 gene mutations lead to hearing loss or to hearing loss combined with skin disorders of varying severity. Here we studied five autosomal dominant Cx26 mutants: four that cause hearing loss in addition to various skin disorders (N14K, D50N, N54K, and S183F) and one that causes hearing loss alone (M163V). We found that the N14K and D50N mutants, which cause widespread and severe skin lesions, overtly induced cell death, whereas the N54K and S183F mutants, which cause regional and moderate skin lesions, had defects in connexin trafficking as well as impaired gap junction and hemichannel function (Fig. 7). In addition, the M163V mutant, which causes moderate hearing loss alone, displayed only impaired gap junction function (Fig. 7) with no transdominant properties on other co-expressed members of the connexin family. We also showed that whereas all mutants displayed dominant negative properties on wild-type Cx26, the syndromic N54K and S183F mutants further exhibited transdominant effects on co-expressed Cx30 and Cx43 (Fig. 7). Last, despite these transdominant properties, co-expressed connexins can partially rescue Cx26 mutant delivery to the cell surface.
Several extensive reviews have discussed the functional characteristics of many distinct disease-causing Cx26 mutants that were investigated using heterologous expression models (12,22,23). These reports have elucidated some major characteristics of disease-causing mutants: a reduction or ablation of gap junction formation/function, altered selectivity of permissible molecules, and aberrant hemichannel activity. Whereas syndromic and non-syndromic mutations affect amino acid residues in cytoplasmic, extracellular, and transmembrane domains, the majority of syndromic mutations cluster to the N-terminal and first extracellular domains (20 -22, 24). Additionally, it seems that mutants that alter molecular selectivity lead only to non-syndromic hearing loss, whereas aberrant hemichannels are almost exclusively associated with severe syndromic disease (22,24). Recently, the S17F mutant, which leads to a relatively severe form of KIDS, has been shown to form essentially non-functional homomeric gap junctions or hemichannels yet oddly produces hyperactive heteromeric hemichannels with Cx43 (12). This finding further established that aberrant hemichannels are a common characteristic of KIDS mutants, but importantly, it demonstrated the need to carefully examine mutants for transdominant effects on co-expressed connexins, including non-traditional binding partners. It is important to note that a transdominant interaction with a Figure 6. Cx26 mutants linked to syndromic disease display reduced hemichannel function in HeLa cells. A, isolated HeLa cells that were engineered to express GFP-tagged Cx26, N54K, S183F, or M163V mutants were incubated in PI-containing ECS ϩ Ca 2ϩ or ECS Ϫ Ca 2ϩ , and the incidence of PI uptake was quantified from random image fields. Image fields captured GFPtagged connexins (green) and intracellular PI uptake (red). B, in ECS Ϫ Ca 2ϩ , cells expressing the N54K mutant displayed highly variable incidence of PI uptake, and cells expressing the S183F mutant displayed minimal PI uptake. Cells expressing the M183V mutant, much like Cx26, exhibited nearly 100% incidence of PI uptake. All mutants, except for S183F, demonstrated increased PI uptake in ECS Ϫ Ca 2ϩ compared with control ECS ϩ Ca 2ϩ conditions. Some cells exhibited a rounded morphology in ECS Ϫ Ca 2ϩ , which may indicate that they were becoming unhealthy. Two-way ANOVA; ****, p Ͻ 0.0001; n ϭ 3. The number of image fields is noted below the scatter plot in B. Scale bar, 50 m. Error bars, S.D.

Cx26 mutants predict disease severity
non-traditional binding partner, such as Cx43, should be considered a gain-of-function property (11). Therefore, we also explored possible transdominant effects on Cx30, which is coexpressed with Cx26 in the skin and cochlea, as well as Cx43, which is co-expressed with Cx26 in the skin and also found in the cochlea. Furthermore, at least six additional syndromic mutants have been identified in patients that still require functional analysis (21). This report describes the M163V and S183F mutants (25,26) and the N54K mutant, which was reported once in a patient with Bart-Pumphrey syndrome (15) and remains otherwise unexamined.
N14K and D50N mutants are linked to KIDS, which is one of the most severe Cx26-linked skin disorders. Patients present with erythrokeratoderma, PPK, and frequent cutaneous infections, which can lead to fatal septicemia early in life (27). We found that these mutants strongly induced cell death such that meaningful channel function information was unattainable. However, reports suggest they form "leaky" hemichannels when expressed in Xenopus oocytes, where they induce blebbing and substantially reduce cell viability (23,28,29). Cx26 hemichannels are sensitive to hyperpolarization by extracellular Ca 2ϩ , such that they remain closed under physiological conditions (19). Interestingly, some studies have shown that high extracellular Ca 2ϩ conditions can improve the viability of cells expressing such mutants, suggesting that elevated hemichannel activity is tightly linked to cell death in vitro (23, 30, 31). Our Every subsequent square displays the localization (intracellular and/or membrane trafficking), channel formation, and channel function (highly functional (double arrows), reduced function (wavy arrows), non-functional (no-entry symbol)) of each Cx26 mutant (green) and is divided into quadrants corresponding to the context of cells co-expressing the mutant with wild-type Cx26 (blue), Cx30 (purple), or Cx43 (red). Mutants linked to severe syndromic disease are colored red, mutants linked to moderate syndromic disease are colored yellow, and mutants linked to non-syndromic hearing loss are colored green.

Cx26 mutants predict disease severity
observations further demonstrate that Ca 2ϩ -free conditions can elevate the activity of Cx26 and mutant hemichannels, which may lead to cytotoxic effects. Nevertheless, reduced Ca 2ϩ sensitivity by N14K, D50N, and several other KIDS mutants promotes hemichannel opening that can diminish transmembrane ion gradients and release molecules, including ATP, that affect cell viability (32). Excessive ATP release is also known to stimulate purinergic signaling capable of mobilizing intracellular Ca 2ϩ and releasing pro-inflammatory cytokines (33). In the epidermis, these mutants may also be able to disrupt the normal epidermal Ca 2ϩ gradient and lipid processing (34), which can result in barrier defects and a compensatory hyperproliferative response that drives hyperkeratosis in KIDS (35). Interestingly, Asn-14 and Asp-50 are pore-lining residues (35), suggesting that changes to the structure of the Cx26 channel pore can produce the aberrant hemichannel properties that stand at the forefront of KIDS skin pathogenesis.
The N54K mutant is linked to moderately severe Bart-Pumphrey syndrome, featuring PPK, knuckle pads, leukonychia, and deafness; the S183F mutant is linked to PPK with deafness. Although no well-defined pathogenic mechanisms for PPK have been established, our findings suggest that the high intracellular retention of N54K and S183F mutants is a common characteristic of Cx26 mutants linked to PPK. Indeed, this trafficking defect has been indicated in previous reports (21, 36 -38). A patient with Bart-Pumphrey syndrome harboring the N54K Cx26 mutant (15) was reported to have a compensatory increase in epidermal Cx30 expression. Asn-54 is an invariably conserved residue in the connexin family among numerous species (15) and forms hydrogen bonds with Leu-56 from the opposing hemichannel (10). Interestingly, we found that the N54K mutant impaired Cx30 trafficking and dye transfer, suggesting that epidermal Cx30 compensation may act to overcome a transdominant hindrance of Cx26:Cx30 heteromeric gap junction formation between keratinocytes. This study and only a handful of others (31, 39 -42) investigated Cx26 mutants in a relevant keratinocyte model in addition to connexin-deficient reference cells. We found that only REKs expressing the N54K mutant had reduced dye transfer. This suggests that N54K Cx26 may also exert transdominant effects on endogenous Cx43, which may be an important etiological factor in Bart-Pumphrey syndrome skin. Recently, Shuja et al. (26) demonstrated that S183F Cx26 does not form gap junctions or hemichannels in Xenopus oocytes, and junctional conductance was reduced in cells co-expressing Cx43. We also found that the S183F mutant inhibited dye transfer in cells co-expressing Cx30 and promoted partial intracellular retention of co-expressed Cx43, pointing toward transdominant interactions of co-expressed connexins as a mechanism of disease. We suspect that we may not have observed the reported S183F mutant functional inhibition of Cx43 seen in Xenopus oocytes (26) in keratinocytes due to the limited sensitivity of the dye transfer assay or to the dosage of the expressed S183F mutant. Because Ser-183 is a highly conserved residue in the connexin family and among many species (16), we suggest that the Ser-183 residue may have an indirect role in interprotomer binding. Additionally, moderately leaky heteromeric hemichannels formed by S183F Cx26 and Cx43 (26), similar to those composed of S17F Cx26 and Cx43, may also contribute to PPK. Nevertheless, our findings provide additional evidence for the impact of transdominant interactions between connexins in skin disease, where disease severity may be linked to the extent of transdominant influence, particularly in skin regions exposed to greater mechanical stress.
In this study, only the M163V mutant is linked to hearing loss without added skin disease. Because Cx26 is proposed to play an important role in potassium buffering and/or recycling in the inner ear (4) and dozens of hearing loss Cx26 mutants display reduced or no gap junction function (22), it is not surprising that we found that the M163V mutant had a dominantnegative effect on Cx26. However, Cx30 is also highly expressed with Cx26 in the inner ear (4), such that compensation might be able to rescue hearing. Interestingly, several studies have shown that Cx30 does not functionally compensate for Cx26 in the inner ear (43)(44)(45). This may explain why the M163V mutant leads to hearing loss despite the finding that HeLa cells co-expressing the M163V mutant and Cx30 passed dye at levels not unlike those expressing Cx30 and/or Cx26. Including our findings from syndromic mutants, we suggest that mutants with trafficking defects and/or transdominant properties, and mutants that induce cell death lead to skin phenotypes, whereas mutants that have reduced gap junction function but do not interfere with other co-expressed connexins produce non-syndromic hearing loss.
Human epidermis expresses seven different connexins in overlapping populations of keratinocytes (6,46), making the epidermis a more robust system from the prospective of connexin expression compared with the cochlea. Fortunately, this high degree of redundant intercellular communication affords the epidermis resiliency in response to GJB2 mutations that have minor effects on protein function. Because fewer connexin types are expressed within the cochlea and compensation may be limited, smaller perturbations in connexin function are capable of disrupting normal hearing. This may speak to the reason why GJB2 mutations almost never produce skin disease alone but Cx26-linked syndromic and non-syndromic deafness is common. Although the majority of autosomal dominant GJB2 mutations produce hearing loss in addition to skin disease, we posit that the strongest predictors of syndromic disease severity actually stem from the transdominant status and gain-of-function properties of Cx26 mutants. Finally, this study provides evidence in support of genetic screening when faced with complex syndromic diseases.

Cx26 mutants predict disease severity cDNA constructs and transfections
cDNA encoding human Cx26 was provided by Dr. C. C. Naus (University of British Columbia, Vancouver, Canada). PCR was used to add XhoI and EcoRI restriction sites to the 5Ј and 3Ј ends of Cx26, and the resulting cDNAs were cloned into the pEGFP-N1 vector (Clontech) and sequenced for verification with a 17-amino acid linker sequence separating the Cx26 and GFP moieties. Constructs encoding human N14K, D50N, N54K, M163V, and S183F Cx26-GFP were further obtained from NorClone by using a QuikChange site-directed mutagenesis kit (Stratagene, La Jolla, CA) in accordance with the manufacturer's instructions. RFP-tagged Cx30 and Cx26 constructs were described previously (18), and all constructs were validated by sequencing. GFP and RFP tags were shown not to dramatically affect connexin trafficking or protein function, as shown previously (38). Cells at ϳ60% confluence were transiently transfected using Lipofectamine 2000. Transfection mixtures contained 200 l of Opti-MEM reduced serum medium (Life Technologies catalog no. 31985-070); 1 l of Lipofectamine 2000 transfection reagent (Invitrogen catalog no. 11668019); and 1 g of GFP-tagged Cx26, N14K, D50N, N54K, M163V, or S183F cDNA constructs. For co-expression experiments, transfection mixtures differed in that 0.5 g of GFP-tagged constructs plus 0.5 g of RFP-tagged Cx26 or Cx30 were added to produce roughly equal expression of mutant to wild-type connexins. The mixture was gently swirled, incubated at room temperature for 10 min, and added dropwise to cells growing in DMEM. All cells were used for experiments between 24 and 48 h following transfection.

Immunofluorescent labeling
HeLa cells or REKs grown to ϳ80% confluence on sterile glass coverslips were washed with phosphate-buffered saline (PBS) and fixed in an ice-cold solution of 80% methanol and 20% acetone for 10 min at 4°C. Coverslips were then washed in PBS, blocked in a 2% bovine serum albumin (BSA) solution (diluted in PBS) for 30 min, and then incubated at room temperature for 1 h with the following primary antibodies diluted in BSA solution: 1:500 mouse anti-PDI (Assay Designs catalog no. SPA-891), 1:500 rabbit anti-Cx43 (Sigma-Aldrich catalog no. C6219), or 1:200 rabbit anti-cleaved caspase-3 (Sigma-Aldrich catalog no. C8487). Secondary antibodies 1:500 Alexa Fluor 488-conjugated anti-mouse (Invitrogen catalog no. A11017) and 1:500 Alexa Fluor 555-conjugated anti-rabbit (Invitrogen catalog no. A21429) were used to detect primary antibodies. Cells were then incubated for 10 min at room temperature with Hoechst 33342 (1:1000 diluted in distilled water) (Molecular Probes catalog no. H3570), mounted, and imaged with a Zeiss LSM 800 confocal Airyscan microscope equipped with ZenWorks software. Images were captured with a ϫ63 oil immersion objective at room temperature. Gap junction plaques between cells were quantified in a blinded fashion by counting the number of green and red punctae at individual cell-cell interfaces. A minimum of 13 separate images were captured for each mutant, and a one-way ANOVA was performed on the means of three biological replicates. For each mutant, six images were used to quantify the fluorescent signal within intracellular compartments. Briefly, monochromatic fluorescence was measured using ImageJ by carefully tracing the cellcell interface and the intracellular regions of each cell. Values indicate the percentage of total cellular fluorescence located within intracellular compartments.

Dye-transfer studies
HeLa cells or REKs grown to ϳ60% confluence and engineered to express GFP-tagged Cx26 or Cx26 mutants (and RFPtagged connexins for co-expression experiments) as described above were microinjected with Alexa Fluor 350 (410 Da; Molecular Probes catalog no. A10439) to assess gap junction dye transfer as described previously (49). Briefly, cells were microinjected using a fine glass needle attached to an Eppendorf Fem-toJet automated microinjector. Cells were imaged using a Leica DM IRE2 epifluorescent microscope to visualize GFP and RFP, and then 1 min following microinjection, they were imaged again to visualize the spread of Alexa Fluor 350. All images were captured with a ϫ20 objective at room temperature. The incidence of dye transfer within each trial was quantified as the percentage of microinjected cells that passed dye to neighboring cells, and a one-way ANOVA was performed on the means of three biological replicates. Tukey's post hoc test compared the means of each condition with the condition in which cells expressed GFP-tagged Cx26. Untransfected HeLa cells were used as negative controls, and HeLa cells expressing Cx26-GFP only or Cx26-GFP in addition to Cx26-RFP or Cx30-RFP served as positive controls for all dye transfer experiments in HeLa cells. Last, for dye transfer experiments in REKs, untransfected REKs or REKs expressing Cx26-GFP or mutants were used.

Hemichannel assay
HeLa cells were seeded at low density to isolate cells from one another and then were engineered to express Cx26 or Cx26 mutants as described above. PI dye uptake assays to evaluate hemichannel activity were performed as noted previously (18). Briefly, cells were washed in ECS (containing 142 mM NaCl, 5.4 mM KCl, 1.4 mM MgCl 2 , 2 mM CaCl 2 , 10 mM HEPES, 25 mM D-glucose, osmolarity 298 mOsM, pH-adjusted to 7.35 using NaOH) and then twice in either ECS ϩ Ca 2ϩ or divalent-cation-free solution (ECS Ϫ Ca 2ϩ ) (same as ECS ϩ Ca 2ϩ but with Ca 2ϩ and Mg 2ϩ substituted with 2 mM EGTA). ECS ϩ Ca 2ϩ or ECS Ϫ Ca 2ϩ containing 1 mg/ml PI (668.4 Da; Invitrogen) was added to the cells and incubated at 37°C for 15 min. Cells were washed three times with ECS ϩ Ca 2ϩ , and then isolated cells were imaged to visualize GFP and PI using the Leica microscope and OpenLab software. For each replicate, the number of cells containing PI was recorded as a percentage of the total number of GFP-positive cells, and a two-way ANOVA was performed on the means of three biological replicates. Sidak's post hoc test compared the means of ECS ϩ Ca 2ϩ and ECS Ϫ Ca 2ϩ conditions.

Statistical analysis
GraphPad Prism version 6 was used for all statistical analysis, and statistical significance was noted when p was Ͻ 0.05. All plots display individual points and the mean Ϯ S.D.