The syndromic deafness mutation G12R impairs fast and slow gating in Cx26 hemichannels

A family of proteins called connexins forms vertebrate gap junction channels and hemichannels. They are encoded by a multigene family that contains 21 members in humans (Söhl and Willecke, 2003). Connexins have four transmembrane segments (TM1–4) connected by two extracellular loops (EL1–2) and one intracellular loop, with cytosolic N- and C-terminal segments. Hemichannels are formed by oligomerization of six connexin monomers and communicate between the cytoplasm and the extracellular milieu. At plasma membrane appositional zones, hemichannels can dock to compatible hemichannels in the adjacent cell to form gap junction channels, providing a direct pathway for intercellular communication (Goodenough et al., 1996; Gaietta et al., 2002; Segretain and Falk, 2004). Connexin hemichannels are regulated by pH (Bevans and Harris, 1999), phosphorylation (Lampe and Lau, 2000), and divalent cations (Verselis et al., 1994; Verselis and Srinivas, 2008) and exhibit two voltage-dependent gating mechanisms: loop (slow) gating, which closes hemichannels at hyperpolarizing voltages, and Vj (fast) gating, which drives hemichannels into a subconductance state at depolarizing voltages (Trexler et al., 1996; Bargiello et al., 2012; Oh and Bargiello, 2015).

Mutations in the Cx26 gene are the main cause of genetic deafness (Bergoffen et al., 1993; Kelsell et al., 1997, 2000; Arita et al., 2006; Arora et al., 2008; Dobrowolski and Willecke, 2009), with a prevalence of ∼50% of cases worldwide (Kelsell et al., 1997). Some mutations in Cx26 produce syndromic deafness in which severe deafness is accompanied by extensive damage in other tissues, such as in keratitis ichthyosis deafness (KID) syndrome, in which patients also present skin abnormalities (such as palmoplantar keratoderma and erythrokeratoderma) and strong inflammation of the cornea (Yotsumoto et al., 2003; van Steensel et al., 2004; Arita et al., 2006; Mazereeuw-Hautier et al., 2007). To date, there is agreement that nonsyndromic deafness-associated Cx26 mutations reduce gap junction channel function, the degree of which correlates well with deafness severity. In contrast, Cx26 mutations associated with syndromic deafness produce hyperactive or leaky hemichannels in their homomeric (Gerido et al., 2007; Lee et al., 2009; Sánchez et al., 2010, 2013, 2016) and heteromeric (García et al., 2015) configurations. KID-associated mutations cluster at the N terminus and the TM1/EL1 segment of Cx26 in amino acid residues that face the channel pore (García et al., 2016). These regions are critical for regulation of gating. Previous studies have shown that KID-causing mutations in the TM1/EL1 segment affect hemichannel voltage dependence and permeability and impair regulation by extracellular Ca²⁺ ([Ca²⁺]_o; Gerido et al., 2007; Lee et al., 2009; Sánchez et al., 2010, 2013, 2016). It has recently been reported that the N-terminal mutation N14K selectively disrupts channel slow gating without apparent changes in its fast gating (Sánchez et al., 2016). The finding that N14K mutation does not affect the fast gating mechanism is intriguing because amino acids in the N terminus have been previously implicated in this gate (Trexler et al., 1996; Maeda et al., 2009; Bargiello et al., 2012).

Among the syndromic deafness-associated mutations located in the Cx26 N terminus, G12R is a good candidate to investigate the mechanisms underlying the observed gain of hemichannel function. Indeed, glycine at position 12 is conserved among β connexins, and mutations at this position have been found in patients suffering from erythrokeratodermia variabilis (Cx30.3 GJB4; Cx31 GJB3; Richard et al., 1998) and Charcot-Marie-Tooth disease (Cx32 GJB1; Bergoffen et al., 1993).

Here, we studied the biophysical properties of the G12R mutant expressed in Xenopus laevis oocytes and the structural changes induced by the mutation by molecular dynamics simulations, thus providing insight into the molecular mechanisms underlying the gain of hemichannel function of G12R. Remarkably, we found that the G12R mutation strongly impairs the fast gate and modifies the slow gate kinetics of hemichannels. At the single-hemichannel level, this mutation greatly decreases the probability of finding the hemichannel in the subconductance state that is characteristic of the fast gate activation. We propose that at the molecular level, the mutation allows an interaction between R12 and R99, fixing the N terminus in a new position toward the cytoplasm, thereby preventing its action as a blocking gating particle. Finally, these results favor the idea that the fast and slow gates are functionally coupled. Knowledge of the structural and functional bases for the gain of hemichannel function of Cx26 syndromic mutants can lead to development of novel therapeutic strategies.

The cDNA for hCx26 (OriGene) was subcloned in the pGEM-HA vector (Promega) for expression in X. laevis oocytes. Mutations of hCx26 were produced with Quick-Change II Site-Directed Mutagenesis kits (Agilent Technologies). Site-directed mutagenesis was confirmed by sequencing at Macrogen. Nhe1-linearized hCx26 WT and mutant DNAs were transcribed in vitro to cRNAs using the T7 Ultra mMessage mMachine kit (Ambion). RNA integrity was confirmed by running the samples on an agarose gel. The RNA concentration was determined using UV spectrophotometer (UV-1700; Pharma Spec). RNA stock was diluted in RNA-free water to a final concentration of 1 µg/µl and stored at −80°C. The follicular membrane of the oocytes was removed by digestive treatment with collagenase type II (Worthington Biochemical Corp.). To efficiently reduce endogenous Cx38 expression, an antisense oligonucleotide against Cx38 (1 µg/µl; 5′-GCTTTAGTAATTCCCATCCTGCCATGTTTC-3′; Integrated DNA Technologies) was injected 4 h after harvesting the oocytes (Ebihara, 1996). Defolliculated oocytes were injected with 41 nl of the desired cRNA using a Drummond Scientific Nanoinjector and maintained at 18°C in ND96 medium (in mM): 96 NaCl, 2 KCl, 1.8 CaCl₂, 1 MgCl₂, and 5 HEPES, pH 7.5, adjusted with NaOH. Macroscopic currents were evaluated 24 or 48 h after cRNA injection in oocytes bathed in an extracellular solution containing 1.8 mM Ca²⁺. Oocytes were injected with 30 nl BAPTA (1,2-bis(o-aminophenoxy)ethane-N,N,N′,N-tetraacetic acid) at 5 µM to overcome the endogenous activation of Ca²⁺-activated chloride channels. Oocytes were maintained in ND-96 for 30 min for recovery. All experiments were performed at room temperature (20–22°C).

Macroscopic Cx26 hemichannel currents expressed in X. laevis oocytes were recorded using the two-electrode voltage-clamp technique with an oocyte clamp (OC-725C; Warner Instruments). Micropipettes housing the two Ag/AgCl electrodes to measure intracellular potential (voltage electrode) and inject currents (current electrode) into the oocyte were pulled to a resistance of 0.5–1.5 MΩ and filled with 3 M KCl. The bath solutions contained (in mM) 118 NaCl, 2 KCl, and 5 HEPES, pH 7.4, with variable Ca²⁺ concentration from 0.01 to 10 mM. The bath was connected to two 3-M KCl pools using agar salt bridges. The Ca²⁺ concentrations in the bath were changed using a custom-made perfusion system. Currents were recorded 1–3 d after cRNA injection. Currents were low-pass filtered at 200 Hz and sampled and digitized at 1 kHz. Membrane currents were recorded from oocytes under two-electrode voltage clamp in response to depolarizing voltage steps from a holding potential of −80 mV, stepped in 10- or 20-mV increments from −60 mV to 50 or 60 mV, and returned to −80 mV.

Single-hemichannel recording was done using cell-attached and inside-out patch clamp. Removal of the oocyte vitelline membrane of selected oocytes was performed manually using forceps (Word Precision Instruments). Oocytes were placed in a recording chamber containing the internal patch solution, which consisted of (in mM) 140 KCl, 1.8 CaCl₂, 1 MgCl₂, 5 HEPES, and 3 EGTA, pH adjusted to 8 with KOH. The bath solution was connected to the 3-M KCl ground solution pool via an agar bridge. Patch micropipettes were made of glass borosilicate capillaries (World Precision Instruments) pulled on a horizontal puller (Sutter Instrument Co.) and fire polished using a custom microforge to a resistance ranging between 500 kΩ and 1 MΩ when filled with internal patch solution. Patch clamp was performed using an Axopatch 200B amplifier (Axon Instruments; Molecular Devices). The voltage command and current output were filtered with eight-pole Bessel low-pass filters (Frequency Devices). The −3-decibel frequency was 200 Hz for the current input, which was digitized at 1 kHz using a 16-bit A/D converter data acquisition system (Digidata1322; Axon Instruments; Molecular Devices). Recordings were performed holding the voltage at the desire potentials using Clampex 8 acquisition software (Axon Instruments; Molecular Devices).

Calcium dose–response measurements were performed essentially as described previously (Lopez et al., 2013). Ca²⁺ dose–response experiments were performed in batches of oocytes showing low or no expression of endogenous currents when bathed in an extracellular solution containing Ca²⁺ concentrations ranging from 0.01 to 10 mM. Because of the slow kinetics of WTCx26 activation and deactivation, Ca²⁺ dose–response experiments were obtained by assessing the peak tail currents after reaching current saturation during a depolarizing pulse from −80 to 0 mV, at different [Ca²⁺]_o concentrations. To minimize the damage caused to the oocyte by the opening of hemichannels under these experimental conditions, no more than four different Ca²⁺ concentrations, including 0.01 mM Ca²⁺, were tested to prevent damage to the oocytes and for the reproducibility of the currents measured in response to Ca²⁺ changes. Data were adjusted to a Hill equation of the following form:

where the fractional current is I/I_max, I_max is the maximal tail current measured at 0.01 mM Ca²⁺, I is the instantaneous tail current recorded at a particular Ca²⁺ concentration measured at −80 mV after a 40-s depolarization to 0 mV, K_d is an apparent Ca²⁺ dissociation constant (in mM), and n is the Hill coefficient.

Deactivation time constants from tail currents were determined by fitting instantaneous tail currents measured up to 10 s after reaching steady state to single exponential functions using Clampfit 10.4 software (Molecular Devices).

The WTCx26 hemichannel model was constructed using the coordinates of the crystallographic structure of the Cx26 gap junction channel (Maeda et al., 2009; Protein Data Bank accession no. 2ZW3). Coordinates of residues Met1, intracellular loop, C-terminal domain, and side chains for residues K15, S17, and S19 were lacking from the x-ray structure. All missing residues in the crystal structure of Cx26 were modeled using Modeler (Martí-Renom et al., 2000), keeping the original x-ray coordinates fixed and considering the hexameric symmetry of the hemichannel. The models were evaluated using a statistical potential algorithm derived exclusively from membrane proteins (Postic et al., 2015). Hydrogen atoms were assigned using the psfgen module of the Visual Molecular Dynamics software, and protonation states were assigned according to the pKa values predicted by PROPKA 3.0 (Olsson et al., 2011). Disulfide bonds identified in the crystal structure were created between C53–C180, C60–C174, and C64–C169 in each connexin subunit. The WTCx26 hemichannel was then inserted into a POPC (1-palmitoyl-2-oleoylsn-glycero-3-phosphocholine) lipid membrane. The protein domains were arranged with regard to the hydrocarbon core of the lipid bilayer using the Orientations of Proteins in Membranes database (Lomize et al., 2006). As previously described (Kwon et al., 2011), a 150 × 150 × 110–Å box including the protein, lipids, TIP3 water molecules, and 150 mM KCl was generated using the membrane builder module of CHARMM-GUI. The CHARMM-22 and CHARMM-36 force fields were used for protein and lipids, respectively. An initial system equilibration of 375 ps at 303 K was performed using the protocol developed by Woolf and Roux (1996). To obtain a relaxed system, initial force constraints were gradually reduced using constant number of particles, volumen, and temperature (NVT) dynamics. Then, the system was equilibrated for 100 ns at 310 K using constant number of particles, pressure, membrane surface tension, and temperature (NPnAT) dynamics with no constraints. Both methods were applied using the Langevin temperature control. From the final frame of equilibration, we started four independent molecular dynamics simulations of 20 ns. We started these simulations using different seeds and restarting the velocities from a Maxwell distribution. NPnAT dynamics were used for data collection and statistical analysis, with set points of 1 atm and 310 K using the Nose–Hoover method and Langevin dynamics for temperature and piston fluctuation control, with a damping coefficient of 1 ps. This simulation protocol was repeated three independent times starting from the crystal structure, giving n = 12 simulations for data analysis. All simulations were performed using NAMD 2.9. The particle mesh Ewald (PME) method was used for full long-range electrostatics within a relative tolerance of 10⁻⁶. A cutoff distance of 12 Å was applied to real-space Ewald interactions, with a smooth switching function applied between 10 and 12 Å to account for Van der Waals interactions. The SHAKE algorithm was applied to constrain bond lengths to all hydrogen atoms.

The averages and distributions shown for the N terminus (Figs. 6, S1, S2, S3, and S4) were taken from the production simulations of 20 ns. The data for each monomer were calculated and compiled for histogram analysis, using the same amount of bins and the same limits. Curves were smoothed using the Lowess method.

Fig. S1 shows the geometrical distance of the N terminus of WTCx26 or G12R with reference to the center of mass of the protein. Fig. S2 shows the analysis of the local bending angles of TM2 near the proline kink in search for structural changes induced by G12R. Fig. S3 shows the root mean square deviation analysis derived from molecular dynamics simulations. Fig. S4 shows the correlation between the N-terminal displacement and side chain of R12 and also demonstrates no trapping in a local minimum in our simulations. Fig. S5 represents ensemble averages of single hemichannels for WTCx26, G12R, and G12/R99A and shows that single-hemichannel activity recapitulates the macroscopic current in all conditions.

To investigate whether the G12R mutation modifies the biophysical properties of Cx26 hemichannels, we expressed WTCx26 and the G12R mutant variant in X. laevis oocytes and performed electrophysiological recordings. Depolarization of oocytes expressing WTCx26 to 0 up to 60 mV from a holding potential of −80 mV induced a slowly activating current that reaches a maximum value of ∼30 mV and then decreases at higher voltages (Fig. 1, a and c). The decrease in the amplitude of the macroscopic current is the result of the activation of the fast gating mechanism (Trexler et al., 1996). When the same voltage protocol was applied to G12R-expressing oocytes, the induced macroscopic currents were larger than those produced by WTCx26 hemichannels, and the current amplitude increased with successive depolarizing voltage steps (Fig. 1, b and c).

To discard the possibility that large and prolonged depolarization was activating endogenous currents in oocytes that could mask the activation of the fast gate in G12R hemichannels, we performed a fast gating voltage protocol (Pinto et al., 2016). When depolarizing pulses were applied from a holding voltage of 30 mV, macroscopic WTCx26 hemichannel currents showed an initial rise followed by decay to a steady state, a clear demonstration that the fast gate is operative (Fig. 2, a and c). In contrast, G12R hemichannel currents increased with depolarizing voltages that in the WTCx26 hemichannels promote a clear macroscopic current decrease, suggesting that their fast gate is defective (Fig. 2, b and c). To confirm these observations, we performed single-hemichannel recordings from inside-out patches containing WTCx26 or G12R hemichannels (Fig. 3). Acquisition was performed using the fast gate voltage protocol. Ensemble averages from single hemichannels recordings in WTCx26 or G12R hemichannels mimic the kinetics of the currents observed at the macroscopic level. Accordingly, the analysis of the time constant deactivation kinetics in WTCx26 at macroscopic and single-hemichannel levels gave similar values (Fig. 3, a and b). Altogether, these results strongly indicate that the fast gate is defective in G12R hemichannels.

We observed differences in the kinetics of the deactivation between WTCx26 and G12R hemichannel currents (Fig. 1). Because channel closure at hyperpolarizing membrane potentials is regulated by the slow gating mechanism (Verselis et al., 2009), we determined the current deactivation time constants of WTCx26 and G12R hemichannels at different concentrations of [Ca²⁺]_o. The deactivation time constant of WTCx26 hemichannels was ∼44 s at 0.01 mM [Ca²⁺]_o and decreased to 2 s at 10 mM [Ca²⁺]_o (Fig. 4, a and c), in agreement with the inverse linear relationship between the time constant of current decay and the logarithm of [Ca²⁺]_o reported previously for WTCx26 hemichannels (Lopez et al., 2013). In contrast, G12R hemichannels exhibited a deactivation time constant that was essentially independent of the [Ca²⁺]_o; its value was 7 ± 0.5 s in 0.01 mM [Ca²⁺]_o and 5 ± 0.7 s in 1.8 mM [Ca²⁺]_o (Fig. 4, b and c). Ensemble averages from single hemichannels recordings in WTCx26 or G12R exposed to 1 mM Ca²⁺ mimic both the kinetics and deactivation time constants obtained from macroscopic hemichannel currents (Fig. 4, e and f), confirming that the slow gate is also impaired by the N terminus mutation G12R.

Because the Ca²⁺ sensitivity of the deactivation kinetics is altered in G12R hemichannels, we compared the apparent affinity constant (K_d) of WT and Cx26 mutant hemichannels for Ca²⁺. The estimated K_d (mM) for Ca²⁺ was 0.31 ± 0.06, with a Hill coefficient (n) of 0.89 ± 0.12 for WTCx26 hemichannels (Fig. 4 d, open circles). Remarkably, the K_d for Ca²⁺ and n for G12R hemichannels were 0.34 ± 0.07 and 1.01 ± 0.15, similar to WTCx26 values (Fig. 4 d, red circles). These results are intriguing because the time constant of the deactivation is strongly Ca²⁺ dependent for WTCx26 hemichannels but not for G12R hemichannels, which suggests that either the kinetics of the voltage-dependent deactivation of G12R hemichannels is strongly accelerated in the absence of Ca²⁺ or that the current has not reached a steady state by the end of the depolarization pulse to 0 mV.

We have previously demonstrated that the relative rate of ethidium uptake is several-fold greater for G12R than for WTCx26 hemichannels (García et al., 2015). Thus, the increased macroscopic hemichannel activity induced by G12R could be explained by augmented single-hemichannel conductance and/or an increased hemichannel open probability. To distinguish between these, we used cell-attached or inside-out macropatches from X. laevis oocytes expressing either WTCx26 or G12R. At 20 mV, WTCx26 hemichannels (Fig. 5 a) exhibited a high open probability with few and brief transitions to a subconductance state (Or) occurring in WTCx26. In WTCx26 hemichannels, these transitions became more evident and longer when the patch membrane was depolarized to 40 mV, associated with reduction of the open probability of the fully open state (Fig. 5 a, trace at 40 mV). After the onset of the depolarizing step at 60 mV, single WTCx26 hemichannels enter a subconductance state, briefly visiting the open state afterward, then return to the subconductance state, and finally reach the closed state, dramatically reducing the open probability of the fully open state (Fig. 5 a, trace at 60 mV). In contrast, G12R hemichannel single-channel currents very seldom visit this subconductance state; instead, direct transitions between the fully open and closed states can be observed at any of the voltages tested, revealing defective fast gating activation in G12R hemichannels (Fig. 5 b). In contrast to WTCx26 hemichannels, both the open probability and mean open time of the fully open state increase with increasing voltages (Fig. 5 b). The single-channel behaviors of WTCx26 and G12R hemichannels are in agreement with the kinetics of the macroscopic currents (Figs. 2, a and b).

Linear regression analysis of the single hemichannel current versus voltage relationship revealed that the single-channel conductance of the fully open G12R hemichannels did not differ from that of WTCx26 (335 ± 9 vs. 334 ± 6 pS, respectively). These results demonstrate that the gain of function of the mutant hemichannel is caused by a marked increase in the single-hemichannel open probability.

To determine the molecular mechanisms underlying G12R hemichannel functioning, we performed molecular dynamics simulations. These simulations evince that the all-frame averages of the inner topology and radius of the pore along the z axis were similar between WTCx26 and G12R hemichannels (Fig. 6, a and b). The minimum pore diameter was ∼5.5 Å and 5.6 Å for WTCx26 and G12R, respectively, and occurred at the level of the N terminus. However, this domain appears slightly displaced toward the cytoplasm in G12R hemichannels (Fig. 6 b, arrow between blue dashed lines). Correlation analysis between the N-terminal displacement and the side-chain angle of G12R in relation to the z axis shows no trapping on a local minimum during our simulations (Fig. S4).

To determine the basis for this displacement, we analyzed the position of the N terminus in WT and mutant Cx26 by measuring the geometrical distance between the centers of mass of the N terminus with respect to the center of mass of the whole structure. The distribution of distances in WTCx26 showed the main peak at 5.7 Å and a small peak at 7.5 Å (Fig. S1, black line). In contrast, G12R exhibits the main peak at 7.5 Å, a lower peak at 5.2 Å, and several low-frequency peak tails that can be observed up to 12 Å (Fig. S1, red line). This analysis confirms that the N terminus of G12R hemichannels is displaced toward the cytoplasm, in agreement with the pore radius profile analysis (Fig. 6 b). Moreover, the asymmetrical shape of the peak distribution in G12R indicates a more asymmetrical dynamic behavior of the monomers than in WTCx26. These results indicate that the N terminus of G12R hemichannels appears to be closer to the cytoplasm and is capable of acquiring a different molecular conformation than in the WT hemichannel.

We next explored which interactions in G12R might contribute to stabilizing the altered conformation of its N terminus. Because of physical proximity, the more likely interactions of the N terminus might take place with the intracellular loop. To explore this possibility, we sought to determine potential contacts between the mutated position 12 and an extended intracellular loop (from residue 90 to 140). We measured the distance between the guanidyl moiety and every residue in the extended intracelluar loop. The arginine 12 from five of the six protomers in a G12R hemichannel exhibited proximity with several residues in the intracellular loop (Table 1). These residues were A92, M93, V95, Y97, R99, and H100, with R99 being the most frequent (Fig. 6 c). These possible contacts are probably mediated by van der Waals interactions because no hydrogen bonds or salt bridges were observed. These contacts were absent in WTCx26, could represent a source for the stabilization of the displaced N terminus, and can be related to the change in the open probability of G12R hemichannels observed in single-hemichannel recordings.

Based on the in silico data, we hypothesized the R12–R99 interaction in G12R hemichannels “trapped” the N terminus in a position that impeded the conformational changes required for effective fast gating, leading to an increase in the open probability. To test this possibility, we replaced R99 on G12R by D (to strengthen the interaction), K (a conservative mutation), or A (to weaken the interaction) and determined the hemichannel properties of these mutants. Examples of macroscopic currents (Fig. 7, left) and single-hemichannel recordings obtained from excised inside-out or gap-free patches held at 40 mV (Fig. 7, right) show effective activation of the fast gate in WTCx26 revealed as a transition to the subconductance state (Fig. 7 a, Or) that is impaired in G12R hemichannels (Fig. 7 b). No hemichannel currents were detected in oocytes expressing G12R/R99D, but small conductance from endogenous channels can be appreciated at the single-channel level (Fig. 7 c). In contrast, a conservative mutation in R99, R99K (G12R/R99K), produces macroscopic and single-hemichannel currents that resemble those of G12R (Fig. 7, d and g). G12R with R99 replaced by A (G12R/R99A) generated macroscopic hemichannel currents that decrease at the end of the test pulse at voltages ≥40 mV, similar to WTCx26 hemichannels (Fig. 7, e and g). At the single-channel level, G12R/R99A hemichannels exhibit short-lived transitions to the fully open state from the open substate and return to the subconductance state, where these hemichannels stay for several seconds as in WTCx26 (Figs. 7 e and S5). As a control, we also substituted R99 with A in WTCx26 (R99A) to evaluate whether this substitution alters the fast gate activation. This mutant generated functional macroscopic hemichannel currents that exhibit effective, fast gating. Interestingly, the traces obtained at 60 mV showed a marked decrease in macroscopic current amplitude compared with those observed in WTCx26 (Fig. 7, f and g). Thus, the biophysical behavior of G12R/R99A hemichannels resembles that of WTCx26 hemichannels and strongly indicates that R12–R99 interactions are responsible for the lack of fast gating activation in G12R.

In connexin proteins, the N terminus is critical for structural and functional properties because it participates in oligomerization, channel closure, and voltage dependence (Verselis et al., 1994; Lagree et al., 2003; Kyle et al., 2008; Bargiello et al., 2012). Accordingly, we observed that the G12R mutation evoked hemichannel currents that did not reach a maximum and then a decrease at depolarizing voltages ≥40 mV, indicating that the fast gating mechanism is inoperative. An altered fast gating mechanism promoted by the G12R mutation was further confirmed by single-hemichannel recording experiments showing that the G12R hemichannels very seldom enter the subconductance state (Fig. 5). In addition, the G12R hemichannel closure is faster in response to hyperpolarizing voltage pulses, and the deactivation time constant becomes insensitive to [Ca²⁺]_o, suggesting that the slow gate is also critically affected by this mutation.

The augmented current amplitude, together with a deactivation time constant that was almost independent of [Ca²⁺]_o concentrations, indicates that the G12R mutation modifies the slow gating kinetics (Pfahnl and Dahl, 1999). Ca²⁺ binds to residues located at the parahelix region (Zonta et al., 2014; Bennett et al., 2016; Lopez et al., 2016; Nielsen et al., 2017) to modulate the slow gate and stabilize the closed state of hemichannels (Trexler et al., 1996; Verselis and Srinivas, 2008; Lopez et al., 2013, 2016). Nevertheless, the mechanism of [Ca²⁺]_o action to stabilize the closing of hemichannels is controversial because it was recently proposed that the binding of Ca²⁺ does not cause conformational changes but instead creates an electrostatic “seal” that hinders the flux of ions through the pore (Bennett et al., 2016). This study contrasts with a previous one based on atomic force microscopy showing that, upon Ca²⁺ binding, hemichannels undergo large structural rearrangements that substantially reduce the pore size (Müller et al., 2002). Recently, a mechanism of gating by Ca²⁺ in connexin hemichannels has been reported. In this work, Lopez et al. (2016) demonstrate that the negatively charged residues D50 and E47 are essential for the stabilization of the closed state induced by Ca²⁺ binding to the Cx26 hemichannel. Our findings, that the K_d for Ca²⁺ in G12R hemichannels remained similar to that of WTCx26 (Fig. 3 c) and that both the pore diameter and the conformation of the parahelix were not affected in the mutant, suggest that the conformation of the putative Ca²⁺ binding site was not altered by the introduction of this charged residue. Likewise, the reduction of the deactivation time constant induced by this mutation strongly suggests that G12R did not uncouple Ca²⁺ binding from hemichannel closing.

The single-hemichannel conductance was not affected by the G12R mutation, in agreement with our molecular dynamic simulation results showing no radii differences along the pore between WT and mutant Cx26 hemichannels (Fig. 6 b). Reduction in single-channel conductance, impaired regulation by external Ca²⁺, and augmented permeability to Ca²⁺ have been reported for KID mutations D50N, A40V, and G45E, respectively (Sánchez et al., 2013), suggesting that the underlying mechanism induced by KID mutations differs between mutations. In addition, our results strongly indicate that the N terminus is an active component of both the slow and fast gating mechanisms, and that both gates are coupled, which contrasts with previous studies on KID-associated mutations that affected only the slow gating mechanism, including mutation N14K that is also located in the N terminus of Cx26 (Sánchez et al., 2010, 2013, 2016). The involvement of the N terminus in the slow gating mechanism was also suggested by Rubinos et al. (2012). In this work, the authors showed that the antimalarial drug N-benzylquininium inhibits Cx50 hemichannels when added either extracellularly or intracellularly. N-benzylquininium promotes a drastic reduction of the open probability of hemichannels and the stabilization of the closed state, likely by binding specifically to the N terminus of Cx50 (Rubinos et al., 2012). Our results further show that intrinsic modifications of the N terminus (G12R mutation) modified the fast and slow gating of the hemichannel.

Molecular dynamics simulations indicate that the mechanism leading to altered biophysical properties of G12R hemichannels involves conformational changes in the N terminus. These conformational changes displaced the N terminus through the cytoplasm and allowed new contacts with residues located in the TM2/intracellular loop, reducing the N terminus flexibility (Fig. S3). Previous studies have shown the importance of G12 for the bending of the N terminus into the pore (Oshima et al., 2007; Maeda et al., 2009). Recent findings using ¹H-NMR on peptides mimicking the N terminus amino acidic composition of WTCx26 and G12R suggest that G12R does not affect the bending but makes the N terminus more flexible and less structured than that of WTCx26 (Batir et al., 2016). Although this study is in apparent disagreement with our finding that the N terminus exhibits reduced mobility (Fig. S3), the structure of the N terminus in solution can differ considerably from the structure of the same peptide surrounded by other protein segments. Moreover, our experimental results support our hypothesis that the absence of the fast gating activation in G12R hemichannels relies on restricted movement of the N terminus favored by the N terminus/intracellular loop attachment through R12–R99 interaction. Arginine–arginine interactions are common at the interface of oligomeric proteins, supporting protein stabilization (Magalhaes et al., 1994; Neves et al., 2012). The repulsion between the positive charges could be counteracted through induced dipole interactions, quaternary interactions among a cluster of arginines, neighboring negatively charged ions, hydrogen bonding residues, or mediating water molecules (Neves et al., 2012). In our simulations, we did not observe hydrogen bonding or water molecules mediating this possible interaction, but it has been proposed that an induced dipole could explain the ubiquitous arginine–arginine contact (Neves et al., 2012).

Impaired fast gating caused by conformational changes has been reported previously by Ri et al. (1999), in which altering the bend angle of the TM2 at the proline-kink site promotes a subtle change in the conductance value for the subconductance state in Cx32 gap junction channels. These authors hypothesized that the change in the conductance is a consequence of a narrowing of the cytoplasmic entrance at the N terminus via the straightening of a TM2 bend angle, thereby impairing the fast gating (Ri et al., 1999). We tested whether G12R leads to alterations in the bending of the proline-kink by analyzing the bend angles of the TM2 during our simulations. No differences were found in the proline-kink between WTCx26 and G12R hemichannels (Fig. S2 and Table 2), nor was a pore narrowing at the cytoplasmic entry observed (Fig. 6, a and b).

Recently, Sánchez et al. (2016) reported that the syndromic mutation N14K, localized at the N terminus, stabilized the open configuration of hemichannels by impairing the loop gating. The authors propose that a cation–π interaction between N14K and His100 impairs the slow gating without affecting the fast gating mechanism. This finding contrasts with our results showing that G12R virtually eliminates the fast gating. It is surprising that despite the proximity of N14K and G12R at the N terminus, the basis for the gain-of-function phenotype caused by these two mutations appears to differ significantly. Although neither N14 nor G12 is part of the α-helix, G12 is next to the last amino acid of the α-helix (G11). Thus, it is expected that a change from N14 to K would not greatly modify the structure of the N terminus, whereas a change from G12 to R should eliminate the α-helical bending. Moreover, it is possible that the G12R change favors a stronger and stable interaction between the N terminus and the intracellular loop than the ones produced by the N14K mutant, explaining why an N14K substitution would still allow the movement of the N terminus during the fast gating process.

The possibility of a functional interaction between the N terminus and the intracellular loop is also suggested by the recently reported cryo-electron microscopy structure of Inx-6 that reveals a structural association between these segments (Oshima et al., 2016). Notwithstanding that connexins and innexins have poor sequence homology, it is quite remarkable that both Inx6 and Cx26 have similar quaternary structures (Oshima et al., 2016). Thus, the suggested functional interaction between the N terminus and the intracellular loop in connexin cannot be excluded. Likewise, we note that replacing R99 with A in WTCx26 produced functional hemichannels with a fully functional fast gate (Fig. 7, f and g). However, the large decrease in current we observed in the case of this mutant when applying voltages >40 mV is probably caused by a fast gate with speedier kinetics and a long-lived subconductance state compared with the WTCx26 hemichannels; i.e., regarding the fast gate, the R99A mutant is a gain-of-function mutant.

In summary, we elucidated the molecular determinants underlying the gain of hemichannel function in the syndromic mutant, G12R. The basis for its increased function is the larger open probability rather than differences in single-channel conductance. These changes alter the position and the bending angle of the N terminus α-helix because of the rigidity that the substitution of G by R confers to this segment (Fig. 8 b). Based on our molecular modeling and the experimental results, we propose that the interaction between R12 and R99 traps the N terminus in a rigid position, preventing the N terminus from performing its function as blocking gating particle (Fig. 8 b, inset).

We thank Dr. José Ignacio García-Palacios and Dr. Viviana Berthoud for helpful discussions on the manuscript. We thank Claudia Pareja-Barrueto for helpful discussions on molecular dynamic simulations. We thank the National Laboratory of High Performance Computing–Chile center for providing computing time for molecular dynamic simulations, Powered@NLHPC.

This work was supported by Programa de Atracción e Inserción de Capital Humano Avanzado a la Academia (PAI 79170081) to I.E. García, Fondo Nacional de Desarrollo Científico y Tecnológico (3150634 to I.E. García, 1171240 to A.D. Martínez, 3140590 to G.F. Contreras, 1180434 to C. González, 1160574 to T. Pérez-Acle, and 1150273 to R. Latorre), Beca Comisión Nacional de Investigación Científica y Tecnológica–Program Formación de Capital Humano Avanzado/Doctorado Nacional/2017 to B.I. Pinto, Programa de Financiamiento Basal PFB16 Fundación Ciencia para la Vida, and the National Institutes of Health (RO1-GM099490 to J.E. Contreras). This work was partially supported by the Air Force Office of Scientific Research under award number FA9550-16-1-0384 to R. Latorre. The Centro Interdisciplinario de Neurociencia de Valparaiso is a Millennium Institute supported by the Millennium Scientific Initiative of the Chilean Ministry of Economy, Development, and Tourism (P029-022-F).

The authors declare no competing financial interests.

Author contributions: I.E. García, R. Latorre, A.D. Martínez, and C. González conceived this work. I.E. García, F. Villanelo, O. Alvarez, R. Latorre, and A.D. Martínez cowrote and coedited the final version of the manuscript. I.E. García, O. Alvarez, and R. Latorre designed the experimental strategies. I.E. García performed all the electrophysiological experiments. I.E. García, F. Villanelo, A. Pupo, and T. Pérez-Acle performed molecular dynamic simulations. All authors contributed to data analysis.

Merritt C. Maduke served as editor.