Data are given as mean ± standard error of the mean (s.e.m.). Comparisons between two or more groups were conducted by using two-tailed analysis of variance (ANOVA) or paired t-test, as indicated in the figure legends. The Mann–Whitney U-test was used for data which were not normally distributed and/or had dissimilar variance. P-values are indicated by letter P, where P < 0.05 was selected as the criterion for statistical significance. No statistical methods were used to predetermine sample size. The experiments were not randomized and the investigators were not blinded to allocation during experiments and outcome assessment. No pre-established criteria were used to include/exclude samples or animals.

All cell lines were routinely checked for mycoplasma contamination using the Venor® GeM Classic kit (Minerva Biolobs, Cat. No. 11-1025).

The FreeStyle™ 293-F cell line (ThermoFisher, Cat. No. R79007) was maintained in Freestyle 293 Expression Medium (ThermoFisher, Cat. No. 12338026). The XL1-Blue cell line (Agilent, Cat. No. 200228) was maintained in SB medium composed of 32 g/L tryptone (Oxoid, Cat. No. LP0042), 20 g/L yeast extract (Oxoid, Cat. No. LP0021), 5 g/L sodium chloride (Shanghai Sangon Biotech, Cat. No. ST1218—1 kg), and 20 μg/ml tetracycline (ShanghaiSangon Biotech, Cat. No. BS731—10 ml), pH 7.5. The culture was shaken at 37°C until an optical density (OD) of 0.5–0.7 was obtained at 600 nm.

HeLa DH cells (Sigma-Aldrich, Cat. No. 96112022) or HaCaT cells (Cell Lines Service GmbH, Cat. No. 300493) were seeded onto round glass coverslips (Fisher Scientific, Cat. No. FIS#12-542A) at different degrees of confluence, depending on the type of assay. Cells were maintained in Dulbecco's modified Eagle's medium (ThermoFisher, Cat.No. 41965039) containing 10% (v/v) FBS (Gibco-Invitrogen, Cat.No. 10270-106) and 1% penicillin/streptomycin (Gibco-Invitrogen, Cat.No. 15070-063). Twenty four hours after plating, a lipofectamine transfection reagent (Lipofectamine 2000, Cat. No. 11668-019) was used to transiently transfect HeLa DH or HaCaT cells with a construct comprising the coding region of Cx26, or its G45E or D50N mutants, fused or not to Venus (a circularly permuted yellow fluorescence mutant of the green fluorescent protein) at the C-terminal end (Beltramello et al., 2005). For some experiments, HeLa DH or HaCaT cells were transduced with a lentivirus expressing Cx26, or its G45E or D50N mutants, fused to Venus, either directly or after application of doxycycline.

Glass coverslips with adherent connexin-expressing HeLa DH or HaCaT cells were continuously superfused at 2 ml/min at 20–23°C with an extracellular solution (EXP) containing (in mM): 140 NaCl, 5 KCl, 10 HEPES, 2 sodium pyruvate, 4 tetraethylammonium chloride (TEA-Cl), 4 CsCl, 5 glucose, and a reduced (0.2 mM) Ca²⁺ concentration (pH 7.4, 323 mOsm).

Cells expressing Cx26–Venus were selected for patch clamp recordings by visual inspection using fluorescence microscopy, whereas cells expressing untagged Cx26 were selected based on their ability to uptake Lucifer Yellow (CH dilithium salt, Sigma-Aldrich, Cat. No. L0259) (DeVries and Schwartz, 1992; Saez and Leybaert, 2014) after 5 min of incubation in Ca²⁺-free (0 mM) EXP containing 1 mM of the dye.

Patch pipettes were filled with an intracellular solution containing (in mM): 115 KAsp, 10 NaCl, 10 KCl, 1 MgCl₂, 10 HEPES, 1 CaCl₂, and 5 BAPTA tetrapotassium salt (pH 7.2, 311 mOsm) and filtered through 0.22-mm pores (Millipore). Filled pipettes had resistances of 4–6 MΩ when immersed in EXP. Hemichannel currents were assayed in EXP under whole cell patch clamp recording conditions.

The antibody was dissolved in EXP and delivered through a glass micropipette pulled to a 8 μm diameter tip from B150F glass (World Precision Instruments, Sarasota, USA). During antibody delivery, the superfusion was stopped. To block connexin hemichannels, CaCl₂ was added at the indicated final concentrations to EXP, whereas ZnCl₂ was added at 100 μM final concentration.

HaCaT cells were transduced with a doxyxycline-inducible lentivirus expressing Cx26 and plated at a density of 1.5 × 10³ cells/well in 96-well plates 24 h before the experiment. Cx26 expression was induced by doxycycline addition to the culture medium 8 h before the measurement of ATP release. Complete DMEM was removed and cells were washed once with serum-free DMEM. Plates were then incubated for 30 min at 37°C and 5% CO₂. In order to test the ability of the antibody to inhibit ATP release through Cx26 hemichannels, abEC1.1 was subsequently added to the wells to a final concentration of 400 nM, and cells were incubated in the above-mentioned conditions for an additional 30 min. Immediately before starting the ATP release-stimulation protocol, cells were washed once with a solution (NCS) containing a normal (1.8 mM) Ca²⁺ concentration and (in mM): 137 NaCl, 5.36 KCl, 0.81 MgSO₄, 0.44 KH₂PO₄, 0.18 Na₂HPO₄, 25 HEPES, and 5.55 Glucose (pH7.3). Cells were then washed a second time either with the same NCS or with a solution (ZCS) containing 0 mM Ca²⁺ concentration and (in mM): 137 NaCl, 5.36 KCl, 0.44 KH₂PO₄, 0.18 Na₂HPO₄, 0.1 EGTA, 25 HEPES, and 5.55 Glucose (pH7.3). To stimulate the release of ATP through Cx26 hemichannels, cells were incubated in NCS, ZCS, or ZCS supplemented with abEC1.1 (400 nM) for 10 min at 37°C. The amount of ATP released in these conditions was measured with a bioluminescent ATP assay kit (Molecular Probes-A22066) and bioluminescence was measured by a luminometer (Perkin Elmer Victor Light 1420). For each experiment, ATP standard curves were generated using serially-diluted concentrations of ATP and were used to convert measurements of luminescence signals into ATP concentrations.

LDH release was estimated by measuring optical absorbance (A) at 490 nm, using a Cytotoxicity LDH Assay Kit-WST (Dojindo Molecular Technologies, Cat. No. CK12) according to the manufacturer's instruction. Briefly, HaCaT cells were seeded in 96-well plates at 8,000 cells per well density and cultured in 100 μl DMEM medium (HyClone, Cat. No. SH30022.01) without sodium pyruvate and penicillin/streptomycin antibiotics, and supplemented with 5% FBS (Gibco, Cat. No. 10099141). For each sample, A was measured by a PerkinElmer Enspire plate reader. Untransfected cells incubated in normal medium were used as minimum absorbance (A_min) controls, whereas untransfected cells exposed to 1% Triton X-100 for 30 min before testing were used as maximum absorbance (A_max) controls. Percent LDH release (R) was estimated by substituting A-values into the following formula: R=A−AminAmax−Amin×100 and averaging the results over homogenous sets of eight wells.

A peptide corresponding to residues 41–56 of Cx26 (KEVWGDEQADFVCNTL = pepEC1.1) was synthesized and labeled with biotin at the N-terminus (Chinese Peptide Company, Hangzhou Economic and Technological Development Zone, China).

Streptavidin-coated magnetic beads (Pierce, Cat. No. 88817) were mixed in Tween phosphate buffer (PBST) with biotinilated pepEC1.1 and screened against a scFv combinatorial library expressed in phage (Zhang et al., 2012). For the phage panning process (Barbas et al., 2004; Lee et al., 2007), specific clones were enriched by binding to immobilized pepEC1.1, followed by elution with Glycine-HCl (pH 2.2) and repropagation of phage in XL1-Blue cells. After four rounds of panning, 150 colonies were picked and analyzed by phage ELISA (see below). Positive colonies were sequenced and analyzed using the international ImMunoGeneTics information system® (Lefranc et al., 2015). Complementarity determining regions 3 (CDR3s) were aligned with Clustal X (Larkin et al., 2007) and a phylogenetic tree was constructed using CLC Genomics Workbench 8.0 (CLCbio). The genes encoding the scFv candidates, identified by analysis of the phylogenetic tree, were cloned into a modified pFUSE expression vector (Invivogen, Cat. No. pfuse-hg1fc2) to obtain scFv-Fc fusion proteins. 293-F cells were transfected with the scFv-Fc vectors and the expressed fusion proteins were purified using HiTrap Protein A HP columns (GE Healthcare, Cat. No.17-0403-03) with ÄKTApurifier 100 (GE Healthcare). After purification, the buffer was exchanged to PBS (pH 7.4). The purified scFv-Fc proteins were kept either at 4°C, for short-term storage, or at −80°C for longer-term conservation.

The Cx26 coding sequence fused with 10^*His and flag tag was cloned into a pFastBac1 baculovirus transfer vector and expressed using the Bac-to-Bac® Baculovirus Expression System (Life Technologies, Cat. No. 10359-016, 10360-014, 10584-027, 10712-024). The baculovirus infected Sf9 cells were cultivated in an ESF 921 insect cell culture medium (Expression Systems, Cat. No. 96-001-01) at 27°C for 48 h. The recombinant Cx26 protein was then purified using a modified protocol described previously (Dupont et al., 1991; Stauffer et al., 1991).

Alternatively, Hela DH cells transfectants expressing Cx26-Venus were used to isolate total cellular protein. Around 0.1 μg of protein supernatant was separated in a denaturating 10% polyacrylamide gel and transferred to a nitrocellulose membrane by electroblotting.

After blocking with 5% milk powder in washing buffer (TBST, composed of 10 mM Tris-HCl, 150 mM NaCl, 0.1% Tween-20, pH 7.5) for 1 h at room temperature, the nitrocellulose membrane was incubated overnight at 4°C with scFv-Fc fusion proteins (1 mg/ml, dilution factor 1:100). Membranes were washed three times with TBST and after 1 h they were incubated with goat anti-human horseradish peroxidase (HRP)-conjugated antibody (dilution factor 1:2,000; Sigma, Cat. No.A0170), which recognizes the Fc domain of the scFv-Fc fusion proteins. Finally, membranes were incubated with SuperSignal West Picochemiluminescence substrate (Pierce, Cat. No. 34087), which reacts with HRP, and examined using Bio-Rad ChemiDoc MP (Biorad).

Ninety-six-well ELISA plates (Corning Costar, Cat. No. 3690) were filled with 25 μl per well of a solution containing avidin (1 μg/well; Pierce, Cat. No. 21121) dissolved in CBS buffer (pH 9.0). Plates were incubated at 4°C overnight or 37°C for 1 h. After washing with PBST buffer (0.05% TWEEN20 in PBS), 25 μl of a pepEC1.1 solution (0.05 μg/well, diluted in PBS) was added to each well and incubated for 1 h at 37°C. Wells were washed twice and blocked with 150 μl per well of M-PBST (5% milk in PBST) and incubated for 1 h at 37°C. After discarding the blocking solution, 25 μl of abEC1.1 solution (1 mg/ml, diluted in M-PBST buffer, dilution factor 1:100) were added and incubated for 1 h at 37°C, then washed eight times. Each well then received 50 μl of a solution containing goat anti-Human horseradish peroxidase-conjugated antibody (Sigma, Cat. No. A0170-1 ml) diluted in M-PBST buffer (dilution factor 1:5,000). Plates were incubated for 1 h at 37°C and then washed eight times. Finally, 50 μl of substrate ABTS solution (Roche, Cat. No. 11684302001) were added to each well and incubated for 20 min at room temperature. OD was read out at 405 nm on a PerkinElmer Enspire plate reader.

Relative affinity and specificity of scFv-phages and soluble scFvs was assessed against the pepEC1.1 antigen. To this end, 10 μg/ml of pepEC1.1 were used to coat a microtiter plate at 4°C overnight. Any remaining binding sites were blocked with Blotto (5% w/v of Bovine Serum Albumin in PBST; bovine serum albumin was purchased from Thermo Fisher, Cat. No. 23210). Approximately 25 μl per well of scFv-phage or soluble scFv supernatant from overnight cell cultures was added and incubated for 1 h at 37°C. For scFv-phage ELISA (Zhang et al., 2012), after washing, 25 μl of anti-M13 mAb horseradish peroxidase (HRP) conjugate (GE Healthcare, Cat. No. 27-9421-01) diluted 1:1000 in Blotto was added for 30 min at 37°C. For ELISA using soluble scFv, anti-Human horseradish peroxidase conjugate (Sigma, Cat. No. A0170-1 ml) in Blotto was added and incubated for 30 min at 37°C. Detection was accomplished by adding 50 μl of ABTS solution (Roche, Cat. No. 11684302001) and absorbance was measured at 405 nm.

Crystallization was carried out using the hanging drop vapor diffusion method. Crystals of the scFv domain of abEC1.1 grew in the drop consisting of 0.2 μL of the protein solution (8 mg/mL) and 0.2 μL of the reservoir solution (0.1 M HEPES, pH7.5, 3 M NaCl). Crystals were cryoprotected using the reservoir solution supplemented with 20% v/v glycerol and then flash cooled into the liquid nitrogen. Diffraction data were collected at the BL18U beam line of Shanghai Synchrotron Radiation Facility and processed with the HKL-3000 suite (Minor et al., 2006). The structure of the scFv domain of abEC1.1 was solved with the MR method using the programs Phaser (McCoy, 2007). The Fv fragment of a human antibody structure (PDB ID 2GHW) was used as the search model. Structure refinement was carried out using Refmac5 (Murshudov et al., 2011) and model building was facilitated with the program Coot (Emsley and Cowtan, 2004).

The atomistic model of V_H and V_L (without linker) derived from the crystal structure was docked to the extracellular domain of a published model of Cx26 hemichannel embedded in the plasma membrane (Zonta et al., 2012) using the ClusPro 2.0 server (Comeau et al., 2004) in the antibody docking mode (Brenke et al., 2012). Among the 50 docking configuration generated by the software, we selected the only one in which the three CDRs of V_H faced the EC1 loop of Cx26, and we tested its stability by performing a 70 ns molecular dynamics simulation using the Gromacs 4 package (Pronk et al., 2013) and the CHARMM 36 force field (Best et al., 2012). After molecular dynamics thermalization, we obtained a stable configuration in which V_H and V_L interacted with three adjacent protomers of the Cx26 hemichannel. Therefore, we added a second symmetrically docked pair of V_H and V_L and obtained a new configuration in which the pore lumen was completely covered.

The atomistic model system, containing the Cx26 hemichannel inserted in a phospholipid bilayer and the pair of docked antibodies, underwent a short energy minimization in vacuum, and was subsequently solvated with full atom TIP3P water, containing Cl⁻ and K⁺ ions at a concentration of ~0.15 M in order to mimic a physiological ionic strength. After solvation, the total number of atoms was around 253,000. We then performed an equilibrium molecular dynamics simulation under periodic boundary conditions at constant pressure for an additional 200 ns.

Temperature T was kept fixed at 300 K in all simulations, and, where stated, the pressure P was fixed at 1 atm using Berendsen thermostat and barostat (Berendsen et al., 1984). Fast smooth Particle-Mesh Ewald summation (Darden et al., 1993) was used for long-range electrostatic interactions, with a cut off of 1.0 nm for the direct interactions.

Based on the crystal structure of Cx26 gap junction channels (Maeda et al., 2009) and a related molecular dynamics model of the Cx26 hemichannel (Zonta et al., 2012), we synthesized a bait peptide, which we named pepEC1.1, corresponding to residues 41–56 in the EC1 domain of Cx26. We investigated the conformation of pepEC1.1 in solution by circular dichroism (CD) and nuclear magnetic resonance (NMR) spectroscopy, and determined that the peptide did not fold in solution, therefore we screened the scFv phage library (Zhang et al., 2012) against pepEC1.1.

After two rounds of panning (see section Materials and Methods), an increased enzyme-linked immunosorbent assay (ELISA) signal was observed for the pepEC1.1 group compared to the control group (bovine serum albumin, BSA); the scFvs which specifically bound pepEC1.1 were enriched after four rounds of panning (Figure 1A). Seven hundred phage-infected clones were picked randomly from the third and fourth round of panning and tested for their ability to bind pepEC1.1 using phage ELISA. One hundred and fifty phage clones were selected from the original 700 clones as positive binders (pepEC1.1/control >2) (Figure 1B). Positive colonies were sequenced, complementarity-determining regions 3 (CDR3) were aligned and a phylogenetic tree was constructed (Figure 1B, inset). Based on the analysis of the phylogenetic tree, different scFv sequences were converted to the scFv-Fc format, in which the scFv domain is fused in one polypeptide chain to the hinge and fragment constant (Fc) domain of human immunoglobulin G1 (IgG1) (Bujak et al., 2014). The binding affinity of the scFv-Fc recombinant antibodies for pepEC1.1 was quantified by ELISA, and the most promising scFv-Fc candidate (Figure 1C) was named abEC1.1.

In western blot analyses of Cx26 protein fused with a 10^*His and flag tag (29.4 kDa) and purified from a Sf9 expression system (see section Materials and Methods), abEC1.1 showed a band between 25 and 32 kDa (Figure 1D, left). Likewise, using total cellular proteins obtained from HeLa DH cells (a clone virtually devoid of native connexins) transfected with Cx26-Venus (Beltramello et al., 2005), abEC1.1 showed a band between 46 and 58 kDa (Figure 1D, right), in agreement with the expected molecular weight of Cx26-Venus (52 kDa).

To determine whether abEC1.1 binding to Cx26 interferes with hemichannel function, we performed patch clamp experiments on HeLa DH cells expressing Cx26-Venus. In low extracellular Ca²⁺ conditions (0.2 mM), which favor the opening of connexin hemichannels (Saez et al., 2005; Fasciani et al., 2013; Saez and Leybaert, 2014; Bennett et al., 2016), the dose-inhibition response curve of membrane conductance, normalized to pre-antibody application levels, vs. abEC1.1 concentration yielded a half maximal inhibitory concentration (IC₅₀) of ~40 nM (Figures 2A,B). The inhibitory effect of abEC1.1 on Cx26 hemichannels was incomplete, leaving a residual conductance of 17.5% ± 4.1% (n = 4) at saturating concentrations (952 nM). Similar results were obtained in HeLa DH cells expressing Cx26 not fused to Venus (residual conductance at 952 nM: 15.9 ± 3.4%, n = 4, P = 0.8, ANOVA).

The currents affected by the antibody were inhibited by extracellular Ca²⁺, in a dose-dependent manner, and abrogated by Zn²⁺ (100 μM; Figure 2C), a non-specific Cx26 hemichannel blocker (Sanchez et al., 2014; Levit et al., 2015).

In the mammalian cochlea, ATP is released under various conditions (Housley et al., 2009) from the organ of Corti (Wangemann, 1996; Zhao et al., 2005) and the lateral wall (Munoz et al., 2001; Chen et al., 2015). Here, we used organotypic cochlear cultures from P5 mice (Beltramello et al., 2005). In this model system, which is known to express endogenous Cx26 (Ortolano et al., 2008), the release of ATP from connexin hemichannels plays a crucial role in intercellular Ca²⁺ signaling both in the lesser epithelial ridge (Anselmi et al., 2008; Majumder et al., 2010; Ceriani et al., 2016) and in the greater epithelial ridge (Tritsch et al., 2007; Schutz et al., 2010; Rodriguez et al., 2012; Wang and Bergles, 2015; Johnson et al., 2017). Incubating cochlear organotypic cultures with abEC1.1 for 15–20 min severely but reversibly reduced the frequency of spontaneous Ca²⁺ transients in the greater epithelial ridge, as well as the area of tissue invaded by each event (Figures 3A–D and Supplementary Video 1).

Numerous studies have reported that connexins have half-lives of 1.5–3.5 h (Thomas et al., 2005; Esseltine and Laird, 2016), and that antibodies targeted to the extracellular domain of connexins may interfere with the assembly of gap junction channels (Riquelme et al., 2013). Therefore, we assayed dye transfer among non-sensory cells of the lesser epithelial ridge after incubating cochlear organotypic cultures for 4 h with abEC1.1 at a saturating concentration (952 nM). The cell-to-cell spreading of Lucifer Yellow over time was not hampered by the antibody (Figures 3E,F and Supplementary Video 2), indicating that abEC1.1 does not affect dye transfer through gap junction channels in this tissue.

The presence of gap junctions in these dye coupled cells was confirmed by immunolabeling detergent-permeabilized cochlear organotypic cultures with a commercial monoclonal anti-Cx26 antibody (Figure 4A). In non-permeabilized cultures, we detected abEC1.1 immunoreactivity in the apical plasma membrane (Figure 4B), consistent with the capacity of abEC1.1 to recognize surface-exposed Cx26 but not gap junction channels. These results are similar to those previously obtained with CELAb polyclonal antibodies (Clair et al., 2008) in organotypic cochlear cultures (Anselmi et al., 2008; Majumder et al., 2010). abEC1.1 immunoreactivity was eliminated by pre-incubating the antibody with its cognate peptide, pepEC1.1 (Figure 4C).

To examine potential cytotoxic effects of the antibody, cochlear organotypic cultures were incubated overnight in serum-free culture medium supplemented with abEC1.1 (952 nM). A subsequent live/dead fluorescence assay (Figure 5) showed an overwhelming proportion of viable cells, indistinguishable from controls not exposed to the antibody. In contrast, cell viability in these organotypic cultures was compromised after just 1 h of incubation with mefluoquine, a small-molecule analog of the anti-malarian drug quinine, which requires a concentration of 100 μM (used in these assays) to inhibit Cx26 hemichannels by ~70% (Levit et al., 2015), i.e., less than the maximal inhibition imparted by abEC1.1 at (non-toxic) concentrations <1 μM.

To explore further the potential applications of this antibody, we performed patch clamp experiments in a keratinocyte cell line from adult human skin (HaCaT cells) which maintain full epidermal differentiation capacity (Boukamp et al., 1988). Wild type HaCaT cells exhibited negligible whole cell currents (<200 pA) in response to voltage commands in 0.2 mM extracellular Ca²⁺, whereas currents >2 nA were measured in HaCaT-Cx26, i.e., HaCaT cells in which stable expression of Cx26 was induced by lentiviral infection (Figure 6A). abEC1.1 (952 nM) inhibited these currents by > 80% (Figure 6B), consistent with the results obtained in HeLa DH transfectants.

To test antibody cytotoxicity in keratinocytes, we used a LDH release assay. LDH is a stable enzyme, present in all types of cells, which is normally retained in the cytoplasm, but is released into the medium in the presence of cytotoxic agents (see section Materials and Methods). abEC1.1, applied overnight at saturating concentrations (115, 400, and 952 nM) reduced slightly LDH release in both wild type and HaCaT-Cx26 cells, whereas incubation with mefluoquine (100 μM) for 1 h increased LDH release by >9-fold compared to control conditions (Figure 6C). These results indicate that abEC1.1 can be safely applied to cultured keratinocytes, whereas mefluoquine is toxic. In accord with the inhibition of ATP-dependent spontaneous Ca²⁺ signaling activity in cochlear organotypic cultures, abEC1.1 (400 nm) reduced also the release of ATP from cultured keratinocytes exposed to a Ca²⁺-free medium for 10 min (Figure 6D; see section Materials and Methods for further details).

Cellular uptake of fluorescent tracers allows real time measurements of membrane permeability properties via hemichannels (Saez and Leybaert, 2014). Lucifer Yellow uptake was undetectable in wild type HaCaT cells incubated for 5 min in nominally Ca²⁺-free (0 mM) extracellular medium (Figure 7A), but was prominent in HaCaT–Cx26 cells (Figure 7B). Dye uptake was inhibited either by pre-incubating HaCaT–Cx26 cells with abEC1.1 (952 nM) for 20 min in 0 mM Ca²⁺ (Figure 7C), or by maintaining the cells in 2 mM Ca²⁺ (Figure 7D). These results indicate that abEC1.1 abrogates dye uptake through connexin hemichannels.

The D50N mutation of Cx26 causes the most common form of KID/HID, whereas the G45E mutation has been linked to lethal forms of the syndrome (Martin and van Steensel, 2015). We induced transient expression of G45E or D50N mutant hemichannels in HaCaT cells. As expected, most cells displayed visible sign of stress within few hours post transfection. Nonetheless, we succeeded to perform patch clamp experiments on some of them. In 0.2 mM extracellular Ca²⁺, Cx26 G45E hemichannel currents were similar to wild type Cx26 currents, whereas Cx26 D50N exhibited faster tail current kinetics, as previously reported for hemichannels expressed in Xenopous oocytes (Sanchez et al., 2013). Importantly, both G45E and D50N, continued to show inhibition by abEC1.1 (Figure 8).

Considering abEC1.1 a promising candidate, we solved the crystal structures of its scFv domain (Figure 9A; the statistics of data collection and structure refinement are summarized in Table 1). Five of the seven amino acids that link V_H to V_L were not solved, however the structure is compatible with that of a diabody (Holliger et al., 1993; Perisic et al., 1994). Therefore, abEC1.1 is expected to assemble as diabody-Fc, i.e a dimer of two scFv-Fc polypetides where V_H and V_L from one scFv, pair with the complementary domains of a second scFv (Wu et al., 2001).

To gain insight into the interaction of the antibody with its target, we combined the crystal structure of the scFv domain (Figure 9A) with an atomistic model of the Cx26 hemichannel (Zonta et al., 2012). In agreement with the results of Figure 2A, we simulated the docking two scFv diabodies to the extracellular domain of the hemichannel and used molecular dynamics to thermalize the interacting systems. In this computational model, three CDR residues of the abEC1.1 V_H (Y34 in CDR1, H55 in CDR2, and W104 in CDR3) interacted with N56, T57, and L58 residues of Cx26 at the apex of the EC1 loop (Figure 9B and Supplementary Video 3).

To test this prediction, we mutated one by one the three putatively critical amino acids of abEC1.1. Patch clamp assays confirmed a significantly reduced blocking power of abEC1.1 mutants compared with wild type abEC1.1 (Figure 10). We performed also computer-simulated electrophysiological experiments. To speed up these time-consuming simulations, we removed the V_H and V_L domains of each bound scFv diabody that did not interact with the hemichannel (Supplementary Video 4). The starting configuration was stable within the simulated time window (150 ns) also in the presence of an external electrical potential. Small monatomic ions escaped through narrow holes at the interface between connexins and bound antibodies, as the latter did not create a perfect seal at the extracellular mouth of the hemichannel. Nevertheless, the bound antibodies reduced the simulated ionic current passing through the pore to 42 ± 14% of its control value (Figure 11A), in qualitative agreement with the experimental results (Figures 2A,B). In this model, the inhibitory effect of the antibodies is associated with a reduction of the hemichannel pore diameter following antibody binding to the EC1 epitope described above (Figures 11B,C). In agreement with these experimental results, molecular dynamics simulations showed that ATP escapes from the extracellular pore vestibule of the free hemichannel in few nanoseconds, whereas bound antibodies impede ATP release to the extracellular milieu (Supplementary Video 5).

Altogether, these results indicate that abEC1.1 inhibits connexin hemichannels by binding an epitope at the apex of the EC1 loop.

RAL is a founder of Zebra Biologics Inc.; GY is a partner of Zebra Biologics Inc. The other authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.