Connexins (Cxs) are transmembrane proteins pivotal for cell communication as they assemble into hemichannels (HCs) and gap junction channels (GJCs). HCs connect the cell with its external milieu and are relevant pathways for paracrine or autocrine signaling, whereas GJCs connect the cytoplasm of cells in contact, allowing direct cell-cell interactions. In addition, both HCs and GJCs allow the passage of molecules and ions that are essential for maintaining cell homeostasis (Harris, 2007) (e.g., ATP, glucose, NAD⁺, and second messengers, among others) (Kang et al., 2008). Under normal conditions, extracellular Ca²⁺ and resting membrane potential keep HCs with low open probability (Bennett et al., 2016; Lopez et al., 2016), and depending on the Cx isoform, the open probability of HCs can be increased by diverse metabolic and signaling mechanism including, phosphorylation, nitrosylation, interaction with calcium-calmodulin and CO₂ (Sáez et al., 2003; Nijjar et al., 2020).

The human genome presents 21 Cx genes with age and tissue-specific distribution (Willecke et al., 2002; Söhl and Willecke, 2003). For example, Cx26 and Cx30 are co-expressed in glia, skin, and cochlear supporting cells, among others (Ahmad et al., 2003; Forge et al., 2003; Zhao and Yu, 2006). In the membrane, Cxs form different oligomeric channels: homomeric, when HCs are composed of the same Cx type, and heteromeric, when different Cx isoforms conform the channel. The configuration adopted by HCs and a GJCs is extremely important because each Cx confers specific regulatory and intrinsic biophysical properties to the channels, including conductance, permeability, and voltage dependence (Koval et al., 2014).

Mutations in the Cx26 gene (GJB2) can produce genetic sensorineural hearing loss due to cochlear malfunction (Wingard and Zhao, 2015; Verselis, 2017). The exact role of Cxs in the functioning of the cochlea is under debate, but it has been suggested that it involves the regulation and recycling of K⁺, pH maintenance, and the passage of molecules like ATP and IP₃ between supporting cells (Gossman and Zhao, 2008; Verselis, 2017), maintaining the sensitivity and viability of hair cells, and the general homeostasis of the auditory sensory epithelium (Wan et al., 2013). The most common non-syndromic deafness (DFNB1A) has been linked to more than 300 mutations in GJB2 (see http://deafnessvariationdatabase.org/letter/g, then select “GJB2″); these mutations can be missense, non-sense, frameshift, insertions, and deletions; of which non-sense/truncating mutations, such as 35delG, are the prevalent forms in some populations (Mammano, 2018). However, characterization of these mutants in exogenous expression systems have shown that they generally do not form GJCs or that GJCs formed present altered gap junctional permeability (Martínez et al., 2009; García et al., 2016b). Besides, a small group of mutations with a dominant inheritance pattern produces severe skin problems and correspond to syndromic deafness (Avshalumova et al., 2014; García et al., 2016a). An example of syndromic disease is the KID characterized by severe deafness and skin abnormalities such as palmoplantar keratoderma, erythrokeratodermia, ichthyosis, and in some cases, corneal keratitis that leads to blindness (Skinner et al., 1981; García et al., 2016a).

Most syndromic mutations in Cx26 are clustered in the amino-terminal domain (NT) and in the para-helix domain facing the internal and external vestibule of the channel pore, respectively (Sanchez and Verselis, 2014; García et al., 2016b) and in most cases, they form hyperactive HCs (Lee and White, 2009; Levit et al., 2012; Shuja et al., 2016) with altered gating properties (García et al., 2018b). Moreover, some KID mutants, like Cx26S17F, form hyperactive hemichannels only when co-expressed with other wild-type (WT) Cxs, like Cx43 (García et al., 2015). However, the presence of hyperactive HCs in the cochlea of a KID mouse model has not been determined yet.

In our previous studies, we showed that some KID mutations in the NT of Cx26 (e.g., S17F, G12R, and N14Y) produce aberrant Cx-Cx interactions that could lead to the formation of heteromeric HCs and GJCs with Cx43 (García et al., 2015). Since Cx26 and Cx43 are co-expressed in skin cells, we have proposed that the channels resulting from the oligomerization of mutant Cx26 and Cx43 are responsible for the skin phenotype observed in KID (García et al., 2016b). Moreover, keratinocytes and supporting cells of the organ of Corti also express Cx30 together with Cx26 (Zhao and Santos-Sacchi, 2000; Ahmad et al., 2003; Forge et al., 2003). In fact, Cx26 and Cx30 co-assemble into heteromeric channels in most cochlear supporting cells (Ahmad et al., 2003; Sun et al., 2005) with some heterogeneity due to differences in the expression ratio between Cx26 and Cx30, for example, Deiters’ cells express more Cx30 than Cx26 (Sun et al., 2005; Ortolano et al., 2008; Jagger and Forge, 2015; Liu et al., 2015). However, it remains unknown whether Cx26-KID mutants interact with Cx30 producing aberrant HCs and GJCs and whether such interaction alters the inner ear homeostasis.

In this study, we found that cochlear tissue expressing Cx26S17F presented increased membrane permeability in the supporting cells of the organ of Corti through the formation of hyperactive HCs, producing damage of hair cells in the organ of Corti. Moreover, we show that mutant Cx26S17F can associate with Cx30, producing hyperactive HCs with reduced sensitivity to extracellular Ca²⁺ and standard HC blockers. Molecular dynamic models of mutant heteromeric show alteration in residues involved in Ca²⁺ coordination. These results provide an explanation for the deafness phenotype in KID syndrome.

Generation of constitutive knock-in (KI) Cx26S17F. Cx26S17F carrier mice (Cx26^+/floxS17F) kindly donated by Dr. Klaus Willecke (Schütz et al., 2011) to Emma Infra Frontiers repository (strain EM:05215) was acquired to reproduce KID syndrome in mice. For this, Cx26^+/floxS17F mice, which has the Cx26 WT gene flanked by loxP sites and Cx26S17F cloned after the WT gene, allowed the expression of Cx26S17F under the WT promoter in every tissue that expressed Cx26, after crossbreeding with CMV-Cre mice (strain #6054, Jackson Laboratory), giving rise to KI Cx26S17F (Cx26^+/S17F; KI) mice that presented skin problems as shown by skin disease (Supplementary Figure S1). Genotyping was done by PCR as described (Schütz et al., 2011) using the following primers loxP-F: CTA TCA GCA GCC TAG AGG AGG, and loxP-R: CAT GAT GCG GAA GAT GAA GAG, with an expected amplicon of 250 bp for Cx26 WT allele, 330 bp for floxed allele Cx26^+/floxS17F and 380 bp after deletion of Cx26 WT gene for KI Cx26S17F (Supplementary Figure S1B); and Cre F: GCG GTC TGG CAG TAA AAA CTA TC and Cre R: GTG AAA CAG CAT TGC TGT CAC TT, with an expected amplicon of 100 bp (not shown), according to the manufacturer’s instructions.

Cochlear explant cultures. Cochleae were obtained according to the protocols approved by the Ethical and Animal Care Committee of the Universidad de Valparaíso, considering all efforts to avoid mice suffering and the use of in vitro approaches when available. Briefly, P1 to P7 animals (7 days postnatal) were sacrificed by decapitation. The explants were obtained by skull dissection from the temporal bone; both inner ears located in the temporal bone were isolated, and cochleae were extracted as described (Parker et al., 2010). Then, once separated from the spiral ligament, the sensory epithelium was transferred on glass coverslips pre-coated with Geltrex (Gibco) in a plating medium (Dulbecco’s Modified Eagle Medium, supplemented with 5% FBS and 5% HS), and maintained for 24 h at 37°C in a humified incubator under 5% CO₂ atmosphere and 95% humidity (Supplementary Figure S2).

Ex-vivo Cx26S17F conditional KI (cKI Cx26S17F). Cx26^+/floxS17F cochlear explants were treated with Cre-recombinase recombinant protein fused to a TAT sequence (TAT-Cre) enzyme, allowing the recombination to express cKI Cx26S17F ex-vivo. TAT-Cre recombinase was used at 10 U/ml in culture media and incubated overnight at 37°C, according to the manufacturer’s instructions (Sigma-Aldrich) (Supplementary Figure S3). All experiments were done after ∼16 h incubation.

Cell culture and transfections. HeLa cell culture and transfections were done as described (García et al., 2015). Cx26S17F was cloned in pcDNA3.1 C T-GFP-TOPO (Invitrogen). A plasmid with the Cx30 cloned in a vector containing msfGFP was acquired from the Addgene reservoir (#69019; donated by Dr. David C. Spray. Albert Einstein College of Medicine, New York, USA) and subsequently cleaved to clone into the pNT-mCherry vector (Clonetech, kindly donated by Dr. María Estela Andrés, Universidad Católica de Chile), which allowed the visualization of Cx30 in two colors, according to the experiment to be carried out. For transient expression of the different Cxs in HeLa cells, transfections were performed using the protocol described for Lipofectamine 2000 (Invitrogen). For double transfections with Cx26S17F/Cx30, 1 µg of Cx26S17F and 0.7 µg of Cx30 plasmids were used. All the experiments were performed 16 h after transfection.

Immunofluorescence and colocalization analysis. Cochlear Explants were fixed in PBS-4% PFA for 1 h at room temperature, washed with PBS, and stored at 4°C until use. The following primary and secondary antibodies were used: mouse αCx26 and rabbit αCx30 (Lifetech); αMouse-Cy2 (Jackson InmunoResearch) and αRabbit-Cy3 (Jackson InmunoResearch). Antibodies were diluted in blocking solution (PBS-Triton X-100 1% + 2% BSA +2% normal goat serum) as follows: mouse αCx26 1:100; rabbit αCx30 1:100; αMouse-Cy2 1:500, αRabbit-Cy3 1:500. Additionally, the stereocilia were detected by phalloidin-Texas Red (Sigma-Aldrich) staining. In all the above experiments, the cells were examined and acquired in an upright confocal laser-scanning Nikon Eclipse C1-Plus microscope (Nikon Instruments) using the EZ-C1 software and were analyzed with the FIJI software (NIH). In addition, optical sections were acquired for every 1 µm in the z-axis, and 3D confocal images were reconstituted from confocal stack images using NIS elements software (Nikon Instruments). In addition, some experiments for stereocilia visualization were analyzed by super-resolution microscopy (Zeiss ELYRA S1, SR-SIM) at the Center for Advance Microscopy at the Universidad de Concepción, Chile (https://cmabiobio.cl/en/).

Immunofluorescence was carried out in transfected HeLa cells as described previously (García et al., 2015); with the following primary and secondary antibodies, respectively: mouse αCx26 1:200 (Lifetech), rabbit αCx30 (Lifetech), and αMouse-Cy2 1:1000 (Jackson InmunoResearch) and αRabbit-Cy3 1:1000 (Jackson InmunoResearch). Cell nuclei were counterstained with DAPI at room temperature for 10 min. Expression of each Cx was quantified in each optical section of 1 µm of confocal stack images using NIS elements software (Nikon Instruments). Plasma membrane was not labeled directly, instead we estimated its localization between cellular appositions (cell-cell contacts) and cell surface, observed using the bright field light from confocal microscope images, similar criteria has been used before (Lauf et al., 2002).

To evaluate the colocalization strength of Cx30 with Cx26WT or Cx26S17F, Pearson and Manders’s coefficients were calculated by cellular area in images showing HeLa cells co-expressed Cxs, using NIS elements software (Nikon Instruments). Ten ROIs of the same size were randomly distributed in a 1 µm optical section in the z-axis for each cellular compartment, membrane, cytoplasm, and the cellular appositional zone (GJ Zone). ROIs that did not show expression of both Cxs were not considered in this quantification. The threshold of every image was set visually. Statistical analyzes were made in Graph Pad Prism software (GraphPad Software Inc., USA).

Proximity ligation assay (PLA). PLA methodology was used according to the manufacturer’s instructions (Duolink, Sigma-Aldrich). The PLA anti-rabbit minus probe binds to the αCx30 antibody, whereas the PLA anti-mouse plus probe binds to the antibody against αCx26 or with the probable interaction partner respectively. Subsequently, the samples were incubated with secondary antibodies conjugated with the PLA probe. After that, amplification and detection protocols were done as manufacturer´s instructions. In all the above experiments, the cells were examined and acquired in an upright confocal laser-scanning Nikon Eclipse C1-Plus microscope (Nikon Instruments) using the EZ-C1 software and were analyzed with the FIJI software (NIH).

Fluorescence recovery after photobleaching (FRAP) assessed the functional state of gap junctions. Cochlear explants or HeLa cells were loaded with 1.6 µM calcein-AM (MW 994.86; net charge -4, Sigma-Aldrich) dissolved in Hanks for 30 min at 37°C protected from light. Then, the excess calcein was removed with two washes of 5 min with Hanks. Using 100% laser power of the confocal microscope, the fluorescence of one of the coupled cells through GJs was bleached. Subsequently, using the 5% power laser, pictures were taken every 1 min for 10 min, obtaining the fluorescence recovery rate in the bleached cell and the concomitant fluorescent reduction in the coupled contacting cells. Instead of bleaching 1 cell for cochlear explants, we bleached a small group of supporting cells (2-4) using 100% laser power. This was because cells in the explants were smaller and more packed between them; in addition, the organ of Corti has two three-layered cells around the tunnel of Corti, making it challenging to bleach just 1 cell. After bleaching, the fluorescent recovery rate was measured in different supporting cells by taking pictures every 1 min for 10 min using 5% laser power. FRAP experiments were performed and acquired in an upright confocal laser-scanning Nikon Eclipse C1-Plus microscope (Nikon Instruments) using the EZ-C1 software and were analyzed with the FIJI software (NIH).

Dye uptake measurements. The dye uptake approach assessed the functional state of HCs expressed in cochlear explants or HeLa cells (García et al., 2015). Dye uptake was determined in cochlear explants maintained for 24 h on coverslips 25 mm in diameter. The culture medium was removed and changed to normal Hank´s solution (mM: 10 HEPES, 140 NaCl, 5.3 KCl, 0.1% glucose, 0.34 Na₂HPO₄, 1.26 CaCl₂, 0.5 MgCl₂; pH 7.4) or in Hank´s divalent cation-free solution (DCFS) containing DAPI (1 µM) as HC activity tracer. Fluorescence microscope pictures were taken in time-lapse mode every 30 s for 10 min in the basal condition (with extracellular Ca²⁺), 5 min in DCFS, and then 5 min in the presence of DCFS with the 100 µM carbenoxolone (CBX). At the beginning and end of the recording, phase contrast photographs were taken to assess cell morphology.

Transiently or stable transfected HeLa cells were grown on 25 mm diameter coverslips. Experiments were done 16 h after transfection in normal Hank´s recording solution and DCFS. Both solutions contained 5 µM ethidium bromide (Etd) as an HCs activity tracer. Pictures were taken every min for 20 min in the basal condition, 10 min in DCFS, and then 10 min in DCFS plus 100 µM La³⁺. To assess cell morphology, contrast photographs were taken at the beginning and end of the recording. All dye uptake experiments were performed in an upright Nikon Eclipse TE 2000U microscope (Nikon Instruments) using the NIS Elements software and were analyzed with the FIJI software (NIH).

Connexin hemichannel currents. Molecular biology, channel expression, and electrophysiology in Xenopus laevis oocytes were performed as described (García et al., 2015; 2018b). cDNA for hCx30, hCx26, or hCx26S17F was subcloned in the pGEM-HA vector (Promega) for in vitro transcription. Macroscopic currents were evaluated at 24 or 48 h post cRNA injection in oocytes bathed in extracellular solution containing 1.8 mM Ca²⁺. All experiments were performed at room temperature (RT, 22°C). Macroscopic Cx HCs currents were recorded using the two-electrode voltage-clamp technique using a Warner oocyte clamp (OC-725C; Warner Instruments). Micropipettes housing the two Ag/AgCl electrodes were pulled to a resistance of 0.5–1 MΩ and filled with 3 M KCl. The bath solutions (mM: 118 NaCl, 1,8 CaCl₂, 2 KCl, and five HEPES, pH 7.4) was connected to two 3 M KCl pools using agar salt bridges. Membrane currents were recorded from oocytes in response to depolarizing voltage steps from a holding potential of −80 mV and stepped in 20 mV increments from–60 mV to 60 mV and returning to −80 mV. Currents were low-pass filtered at 200 Hz and sampled and digitized at 2 kHz. Current amplitude was determined as I _tail (V)/I _max for each voltage tested.

Cell damage/death quantification. Cell damage/death was quantified in HeLa cells with the Annexin V/Propidium iodide (PI) assay (BioLegend^®), according to the manufacturer’s instructions with modifications. Cells were seeded on 25 mm coverslips, washed twice with PBS and incubated with 400 µl Annexin V binding buffer, plus 5 µl of Pacific Blue™ Annexin V and 10 µl of PI solution. Solutions were homogenized, and cells were incubated for 15 min at RT. Then, samples were washed twice with PBS and fixed with 4% PFA for 30 min at RT. After fixation, cells were washed and mounted with anti-fade mounting medium Fluoromount-G (Sigma-Aldrich). All the experiments were kept in dark till assessment. The cells were examined and acquired in an upright confocal laser-scanning Nikon Eclipse C1-Plus microscope (Nikon Instruments) using the EZ-C1 software and were analyzed with the FIJI software (NIH). Cx30 in this experiment was tagged with mCherry (Ex/Em λ= 587/610), which shares similar spectrum with PI (Ex/Em λ= 535/617). The difference between Cx30-mCherry and PI was assessed by morphology, since PI stains the nuclei and cytoplasm with fade red in healthy cells and strong fluorescent red in necreotic cells, which are also round.

Intracellular Ca ²⁺ signal measurement. Basal intracellular Ca²⁺ signal was evaluated in HeLa Cells using Fura-2 (Invitrogen). Cells seeded on 25 mm coverslips were loaded with 5 µM Fura-2 diluted in serum-free medium for 30 min at 37°C and washed two times with Hank´s solution. Coverslips were placed in the recording chamber filled with 1 ml of Hank´s solution. The imaging protocol involved data acquisition of light emission at 510 nm due to excitation at 340 nm and 380 nm. The ratio was obtained by dividing the emission fluorescence image value at 340-nm by the 380 nm excitation on a pixel-by-pixel base (Ca²⁺ signal = F340 nm/F380 nm). Extracellular [Ca²⁺]_e was changed every 3 min from nominal Ca²⁺-free solution to 10 μM, 100 μM, 500 μM, 1 mM, 1.5 mM Ca²⁺ influx at different [Ca²⁺]_e was estimated from changes in [Ca²⁺]_i rate of increments in every [Ca²⁺]_e. Experiments were performed in an upright Nikon Eclipse TE 2000U microscope (Nikon Instruments) using the NIS Elements software and were analyzed with the FIJI software (NIH).

Molecular dynamics. The Cx26WT/Cx30 and Cx26S17F/Cx30 putative heteromeric HC models were constructed using the coordinates of crystallographic structure from the cryo-EM structure of Cx50 (PDB code: 7JJP; (Flores et al., 2020); as a template. This structure has a resolution of 1.9 Å, allowing us to accurately model the sidechain of residues. Missing residues and sections from original coordinates were modeled using MODELER (Martí-Renom et al., 2000). The stoichiometry used was Cx26:Cx26:Cx30:Cx30:Cx26:Cx30 because it is the most abundant composition found in vitro (Naulin et al., 2020). The best models were chosen using the MAIDEN program (Postic et al., 2015). Each HCs were inserted into a 1-palmitoyl-2-oleoylsn-glycero-3-phosphocholine (POPC) lipid membrane considering the spatial arrangements of the protein with respect to the hydrocarbon core of the lipid bilayer, as obtained from the OPM database (Lomize et al., 2006). A 150 × 150 × 110Å box, including the protein, lipids, TIP3 water molecules, and 150 mM KCl, was generated using the INFLATEGRO technique (Kandt et al., 2007). The CHARMM-22 and the CHARMM-36 force fields (MacKerell et al., 1998) were used for protein and lipids, respectively. Initial force constraints were used in NVT dynamics, but these were gradually reduced. The system was then equilibrated for 50 ns at 310 K using NPT dynamics without any constraints. Three independent MD simulations of 20 ns were performed by starting the simulation from different seeds, using NPT dynamics for data collecting and statistical analysis with set points of 1 atm and 310 K using the Nose–Hoover thermostat and Parrinello-Rahman barostat. All simulations were performed using Gromacs 2019 (Abraham et al., 2015). The PME method was used for full long-range electrostatics within a relative tolerance of 1 x 10-6. A cutoff distance of 12Å was applied to real-space Ewald interactions, with a smooth switching function applied between 10 and 12Å to account for Van der Waals (VdW) interactions. The SHAKE algorithm was applied to constrain bond lengths to all hydrogen atoms. All the simulations were performed on the infrastructure of the Computational Biology Laboratory at Fundación Ciencia & Vida (Santiago, Chile).

Study design and statistical analysis. Sample sizes were not statistically pre-determined, but they were estimated based on previous studies. Considering that in vitro experiments were used in transfected cells, they can be considered technical replicates. Despite that, experiments were performed in four or more different independent cultures for every method, and more than four coverslips per day of the experimental condition were considered. Every double-transfected cell in the microscopy field of view was analyzed. For cochlear explants, only two animals were used per condition, and 120 ROIs positioned around the nuclei area were analyzed in different cell types. The researcher was not blinded during the experiments, but the experiments were analyzed randomly.

Results are expressed as means ± SEM, for most results data were normally distributed according to the Kolmogorov-Smirnov test. In some experiments, outliers were removed after identification using the ROUT Method (GraphPad Software Inc, La Jolla, CA, United States) using Q = 1%. Statistical comparisons were performed using a two-tailed t-test comparison between groups, and results were expressed in the graphs as *p < 0.05; **p < 0.005, ***p < 0.001; ****p < 0.0001 accordingly. When data were not distributed normally, a non-parametric t-test was used (Mann-Whitney test). Statistical analysis for multiple comparisons was corrected by the Bonferroni-Dunn method. All data analyses of every experiment were done in Excel (Microsoft Inc.) and GraphPad Prism 6 software (GraphPad Software Inc.).

In the present work, we found that the expression of Cx26S17F in cochlear explants caused the formation of hyperactive HCs and less-functional GJCs in supporting cells of the organ of Corti, was associated with damage in sensory hair cells stereocilia, and would explain the deafness phenotype. Further study will be required to demonstrate this possibility. The hyperactive HCs are insensitive to blockage by classical HCs blockers, like extracellular Ca²⁺ and CBX. We observed the formation of hyperactive HCs and non-functional GJCs in cells co-expressing Cx26S17F and Cx30, which, together with molecular (PLA) and co-localization analysis, suggest the formation of heteromeric channels between Cx26S17F and Cx30 in HeLa cells. These putative disease-associated heteromeric HCs may present reduced sensitivity to extracellular Ca²⁺, which prevents closure of the HCs and allows a significant influx of Ca²⁺ into cells resulting in some cellular damage, but not necessarily cell death. Molecular dynamic experiments in hypothetic heteromeric HCs formed by Cx26S17F and Cx30 suggest that altering extracellular Ca²⁺ binding sites and gating mechanisms can produce HCs hyperactivity.

To study the consequences of KID mutations in-vivo, two transgenic mouse lines have been developed (Schütz et al., 2011) (Mese et al., 2011), and these studies have focused mainly on the skin phenotype (García et al., 2016a). Here, we used the same Cx26S17F mouse line as Schütz et al. (2011), but it was crossbred with CMV-Cre mice to induce a KID phenotype (KI Cx26S17F) in cochlea and skin. Moreover, the expression of the Cx26S17F variant was accomplished by in vitro treatment with recombinant TAT-Cre of cochlea explants from the Cx26S17F mouse line (cKI Cx26S17F). These models revealed that expression of a deafness syndromic mutant, Cx26S17F, causes the formation of hyperactive HCs and less functional GJCs in supporting cells of the organ of Corti. In fact, in situ expression of Cx26S17F in supporting cells in explants of the organ of Corti (cKI Cx26S17F) was enough to induce degeneration of hair cell stereocilia, which indicates that supporting cells keep a dynamic control of hair cell function.

It has been proposed that Cx HCs play important roles in cochlear development (Wingard and Zhao, 2015), and in physiological or pathological cochlear function (Verselis, 2017). Thus, the expression of hyperactive HCs could severely impact the cell viability and function of cochlear supporting and skin cells since HCs, in part, mediate ATP release from these cells and modulate paracrine signaling between support and hair cells in the cochleae (Zhao et al., 2005) and activate an inward K⁺ current into supporting cells, which has been suggested to be important for the excitability of hair cells (Zhu and Zhao, 2010). A similar mechanism has been proposed for skin cells (Skinner et al., 1981; Sanchez and Verselis, 2014).

Because the expression of Cx26S17F in exogenous expression systems did not develop functional GJCs or HCs (García et al., 2015), it is possible that interaction of Cx26S17F with other WT Cxs co-expressed with mutant Cx26 in the cochlea, like Cx30, cause the formation of functional aberrant channels. Indeed, we observed a strong co-localization of Cx26S17F and Cx30 in several cochlear-supporting cells. Moreover, we can not discard that the differences in the magnitude of the effects induced by the expression of Cx26S17F on the DAPI uptake between different supporting cell types can be the consequence of differences in the expression ratio between Cx26 and Cx30, in fact Deiters cells express more Cx30 than Cx26 (Figure 1), and this is consistent with previous publications (Sun et al., 2005; Ortolano et al., 2008).

The finding that Cx26S17F affects the trafficking of Cx30 is consistent with other KID mutations in Cx26, such as G12R and D50N, that also present trafficking defects (Common et al., 2003; Meşe et al., 2007; García et al., 2015). Similar results were obtained in Cx31 mutations that cause other skin disorders (Meşe et al., 2007). These findings support a common mechanism of syndromic deafness mutations with skin phenotype that involves the mislocation of hyperactive or aberrant HCs. Recently, Defourny and Thiry (Defourny and Thiry, 2021) proposed that cadherin-based tricellular adherens junctions, a specialized cell-cell junctions formed at sites where three epithelial cells make contact, which is enriched in lipid rafts, promote the microtubule-mediated trafficking of Cx26/Cx30 heteromeric channels to the cell surface and further assembly into GJ plaques in cochlear supporting cells. We cannot discard that mutant heteromeric Cx26S17F/Cx30 channels miss target tricellular junctions causing less GJ plaque formation. Further studies will be necessary to clarify this point.

Although we did not directly demonstrate the formation of heteromeric channels between Cx26S17F and Cx30, the most plausible interpretation of data obtained by co-localization, PLA, functional assay (for HCs and GJCs), all support the formation of heteromeric channels between Cx26S17F and Cx30. This is consistent with previous studies that demonstrate the formation of heteromeric channels between Cx30 and Cx26 using co-immunoprecipitation, co-localization and functional studies (Sun et al., 2005; Yum et al., 2007; Ortolano et al., 2008). Our results indicate that even though S17F mutation changes the ability of Cx26 to interact with Cx43 (García et al., 2015), it does not significantly affect the interaction of mutant Cx26 with Cx30, supporting the notion that there are other Cx compatibility signals beside the N-terminal domain (Jara et al., 2012; García et al., 2015). In addition, the expression of Cx26S17F produces a trans-dominant-negative effect over GJCs formed by other Cxs, like Cx43 (García et al., 2015) or Cx30 as shown here. Confirming that a common feature in non-syndromic and syndromic deafness is that mutant forms of Cx can affect the function of co-expressed WT Cxs and result in loss of intercellular coupling or the formation of GJCs with altered permeability to second messengers such as IP₃, while maintaining the normal transfer of ionic current through GJCs (Gossman and Zhao, 2008; Yum et al., 2010; García et al., 2016b). Although reduced GJC activity can be critic for the proper function of the cochlea, the exact role of GJCs in cochlear function and development is yet unknown. The cochlear sensory epithelium is a poorly vascularized tissue (Pickles, 2015). Looking at a sagittal section of the scala media, the irrigated part of the tissue would only correspond to the stria vascularis, where the perilymph is generated from blood serum (Wan et al., 2013). Therefore, it has been suggested that the GJC network connecting supporting cells and the network in the lateral wall is the principal way for cells to obtain nutrients and molecules relevant to the correct functioning of the auditory tissue (Zhao et al., 2005; Johnson et al., 2017; Chen et al., 2018; Ceriani et al., 2019). As speculation, if the GJC network is not working properly, recycling of K⁺, intercellular IP₃, ATP, and Ca²⁺ diffusion, among other signaling molecules or metabolites, would be altered, affecting the proper functioning of this tissue (Piazza et al., 2007; Majumder et al., 2010). Consistently, some animal models of non-syndromic deafness present a lack or reduction in GJC function (Johnson et al., 2017).

Although the recording conditions used in this study to determine the functional state of HCs and GJCs are identical to most studies in the field, they are non-physiological; hence we cannot discard that some physiological parameters, like Cx26 sensitivity to CO₂, could potentially affect the results. For example, some Cx26 KID mutants, like N14K, lost their regulation by CO₂ (de Wolf et al., 2016). Cx26 has been demonstrated to be sensitive to CO₂ and an increase in CO₂ open HCs and closed GJ channels (Meigh et al., 2013; Nijjar et al., 2020) similar to our finding in cells expressing Cx26S17F/Cx30. The action of CO₂ on Cx26 has been proposed to occur via carbamylation of the residue K125 and also requires R104 (Meigh et al., 2013). Although, K125 is located in the intracellular loop of Cx26, it seems that CO₂ induces conformational changes that are transmitted to the N terminus to unfold it and plug the channel (Brotherton et al., 2022). Interestingly, KID mutations located in the NT of Cx26, like G12R and N14K, also produce aberrant interaction with residues in the intracellular loop (Sanchez et al., 2016; Garcia et al., 2018a). A similar mechanism could be proposed for Cx26S17F interacting with Cx30, probably by the formation of aberrant heteromeric channels that can alter the mechanism of gating conformation at the intracellular sites of connexons.

It has been proposed that some KID syndrome mutations may activate a pre-existing cryptic splice site in Cx26 mRNA in conjunction with a fluorescent protein tag (i.e. GFP) for chimeric construction (Cook et al., 2019) that may have toxic effects. However, previous study with this mutant using Cx26S17F cDNA constructs with or without GFP tag show the expression of full length Cx26S17F by sucrose gradients resolution and western blot analysis (Garcia et al., 2015). In addition, the data obtained in cochlear explants and HeLa cells are consistent with the increase in whole-cell HC currents observed in oocytes after co-injection of Cx26S17F and Cx30 mRNAs, suggesting that the gain-of-HC activity observed was not a consequence of the expression of splice variants of Cx26.

The co-expression of Cx26S17F with Cx30 produce HCs that did not respond properly either to extracellular Ca²⁺ or La³⁺, suggesting an abnormal gating by extracellular cations, which agrees with the functional properties of other KID mutations (García et al., 2016b). It has been proposed that Ca²⁺ ions are coordinated with residues D50 and K61 such that when Ca²⁺ is bound to these residues, it destabilizes the open state of HCs, favoring a closed conformation (Lopez et al., 2016). In addition, we and others have shown that the mutation G12R, N14K, and N14Y, which are also located at the NT, alters the fast and slow voltage gating, renders HCs with increased open probability, and poor regulation by [Ca²⁺]_e (Sanchez et al., 2016; García et al., 2018b). Although we cannot be certain that the same mechanism applies to the S17F mutation, both mutations have several functional hallmarks in common. The fact that the hyperactive HCs formed by Cx26S17F/Cx30 cannot respond to HC blockers such as La³⁺ or CBX suggests that the general closing mechanism of this channel is severely affected.

As an initial attempt to understand the loss of Ca²⁺ regulation by these putative heteromeric Cx26S17F/Cx30 HCs, we performed molecular dynamics simulation in WT and mutant heteromeric HCs. We found that Cx26S17F/Cx30 HCs present changes in inter-monomer salt bridges between extracellular-side residues of HCs, including D50, K61, E42, and R75, that are relevant for extracellular Ca²⁺ coordination that keeps HCs closed (Bennett et al., 2016; Lopez et al., 2016; García et al., 2018b). Thus, these molecular interaction network changes are consistent with the observed changes in extracellular Ca²⁺ sensitivity in Cx26S17F/Cx30 heteromeric HCs. However, molecular dynamic study was generated using as template the structure of Cx50/Cx46 cryo-EM from Flores et al. (2020) and not the last structure available of Cx6 (Brotherton et al., 2022), therefore, we cannot discard an alternative explanation to the hypothesis derived from our molecular dynamic analysis. Further studies using molecular dynamic simulation with new Cx26 structure available (Brotherton et al., 2022) and mutagenesis of the amino acid residues identified in Cx26S17F/Cx30 channels that may affect extracellular Ca²⁺ binding will be necessary to clarify this point.

Consistently, we found that cells expressing Cx26S17F/Cx30 HCs present higher [Ca²⁺]_i than cells expressing Cx26WT/Cx30 HCs, which is similar to previous findings in cells co-expressing Cx26S17F/Cx43 (García et al., 2015). This is relevant because persistent increases in [Ca²⁺]_i may lead to cell death, which can be part of the KID´s pathological features (Sánchez et al., 2010; Terrinoni et al., 2010; Decrock et al., 2011; Vinken et al., 2011). Moreover, increased [Ca²⁺]_i may be required for HC opening and GJC closure through interaction with Ca²⁺/calmodulin (CaM) that has been observed at around 500 nm [Ca²⁺]_i (de Vuyst et al., 2006; Peracchia, 2020). Although it is still unknown whether CaM binds Cx26 or Cx30, putative binding sites need to be explored.

This work strongly supports the hypothesis that the failure in “fine-tuning” of HCs caused by the Cx26S17F mutation undermined the cochlear and skin homeostasis and induced hearing loss and severe skin disease. Therefore, pharmacological targeting of hyperactive HCs can have therapeutic potential for skin disease and deafness conditions in KID syndrome.