Distinct moieties underlie biphasic H+ gating of connexin43 channels, producing a pH optimum for intercellular communication

Most mammalian cells can intercommunicate via connexin‐assembled, gap‐junctional channels. To regulate signal transmission, connexin (Cx) channel permeability must respond dynamically to physiological and pathophysiological stimuli. One key stimulus is intracellular pH (pHi), which is modulated by a tissue's metabolic and perfusion status. Our understanding of the molecular mechanism of H+ gating of Cx43 channels—the major isoform in the heart and brain—is incomplete. To interrogate the effects of acidic and alkaline pHi on Cx43 channels, we combined voltage‐clamp electrophysiology with pHi imaging and photolytic H+ uncaging, performed over a range of pHi values. We demonstrate that Cx43 channels expressed in HeLa or N2a cell pairs are gated biphasically by pHi via a process that consists of activation by H+ ions at alkaline pHi and inhibition at more acidic pHi. For Cx43 channel‐mediated solute/ion transmission, the ensemble of these effects produces a pHi optimum, near resting pHi. By using Cx43 mutants, we demonstrate that alkaline gating involves cysteine residues of the C terminus and is independent of motifs previously implicated in acidic gating. Thus, we present a molecular mechanism by which cytoplasmic acid–base chemistry fine tunes intercellular communication and establishes conditions for the optimal transmission of solutes and signals in tissues, such as the heart and brain.— Garciarena, C. D., Malik, A., Swietach, P., Moreno, A. P., Vaughan‐Jones, R. D. Distinct moieties underlie biphasic H+ gating of connexin43 channels, producing a pH optimum for intercellular communication. FASEB J. 32, 1969–1981 (2018). www.fasebj.org

Cells of most human tissues-with the notable exception of blood cells and skeletal muscle cells-are electrically and metabolically coupled by means of gap junctional channels, assembled from connexin (Cx) proteins. The hexameric channels permit cell-to-cell solute and ion flow. This function plays a critical signaling role (1) that is particularly important for the spread of electric current in excitable tissues. The biological importance of gap junctional communication necessitates a means of regulating junctional permeability and conductance. Acute Cx channel regulation is typically exercised via post-translational modifications and may involve cellular metabolites and/ or electrophysiologic maneuvers. Moreover, aberrant forms of Cx channel regulation have been implicated in pathologic states (2,3), such as cardiac arrhythmias.
Among solutes that permeate Cx-assembled channels are H + ions, the end products of metabolism. H + ions are produced at a rate that reflects the tissue's metabolic activity. They can feedback potently on cellular function via an array of protonation reactions with proteins. Essentially, all cell types are equipped with a molecular apparatus for maintaining favorable intracellular pH (pH i ). Excess acid is commonly transferred from cells to the nearest functional blood capillary (4) via membrane transport proteins, such as H + -monocarboxylate transporters and Na + /H + exchangers (NHEs). In addition, permeation of H + ions through gap junctions allows pH i to equilibrate spatially among cells, such as those of the working myocardium. Channel-facilitated H + dissipation reduces the spatial heterogeneity of pH i , thereby helping to unify tissue-level function, such as myocardial contractility. In contrast, some clinical conditions-for example, myocardial ischemia-can trigger abnormally large decreases of tissue pH i ; permitting a large and localized intracellular acid load to spread into surrounding tissue would risk inflicting undue damage on cells that are coopted to share the pH i disturbance. Instead, gap junctional channels tend to close by sensing low pH i . A 1980s report first described an inhibitory effect of intracellular acidification on cell-to-cell coupling (5,6). Subsequent expression studies on Cx43 channels have linked this to an inhibition by H + ions, which relies on an interaction between the cytoplasmic C terminus of the Cx43 protein (residues 261-300 and 374-382) (7,8) with its intracellular loop (a protonatable histidine residue) (9). Moreover, these domains are influenced by phosphorylation (10) and interactions with the cytoskeleton (11), which allows for additional fine tuning of Cx43 channel pH i sensitivity. More recently, an additional pH i control of gap junctional conductance and permeability has been described. Inhibition of electrical and solute coupling between mammalian ventricular myocytes-where Cx43 is the dominantly expressed gap junctional isoform-has been demonstrated at both low and high pH i . Ventricular coupling is thus modulated by pH i in a biphasic manner, with peak conductance attained at pH i ;6.9, which is mildly acidic relative to normal resting pH i (12). The molecular structures that underpin gap junctional block at high pH i are currently unknown.
Here, by using heterologously expressed Cx43 channels, we confirm that alkaline-that is, high-pH i reversibly and robustly reduces gap junctional communication, probed electrophysiologically and from measurements of cell-tocell H + ion permeation down a photolytically evoked gradient of [H + ] i . Furthermore, by using mutants of Cx43, we show that the C terminus of Cx43 is involved in alkaline gating and that this process is independent of the molecular apparatus responsible for channel closure at acidic-that is, low-pH i . We present an updated model of the mechanism of biphasic gating of Cx43 channels by H + ions. Our model explains the phenomenon of optimal Cx43 channel permeability in terms of the ensemble of inhibitory and activatory effects of H + ions operating over distinct pH i ranges.

Cx43 expression and cell culture
HeLa (CRM-CCL2; American Type Culture Collection, Manassas, VA, USA) and N2a (CCL-131; American Type Culture Collection) cells were transfected with cDNA for rat Cx43 in pcDNA3.1 vectors (13). Truncated Cx43 (Cx43m257HT) was generated by introducing a stop codon at residue 258 into the cloning site of the bicystronic vector, pIRES2 (Clontech, Mountain View, CA, USA) that provides resistance to Geneticin (G418; Mutagenesis Kit; Thermo Fisher Scientific, Waltham, MA, USA) and includes a histidine tag (His) 6 . Mutants with $1 cysteine-toalanine mutations at the C tail were generated by PCR using specific primers (Supplemental Fig. 1), and the gene was introduced into a pcDNA vector with G418 resistance. Mutant constructs that have all serine residues of the PKC epitope (356-389) substituted to alanine were prepared by PCR using specific primers, similar to those shown in Supplemental Fig. 1. This phospho-null mutant gene was introduced into a pcDNA vector with hygromycin resistance. Transfected cells were selected with 150 mg/ml hygromycin B or 800 mg/ml G418. Cells were grown in high-glucose DMEM that was supplemented with 10% fetal bovine serum, penicillin, streptomycin, glutamine, and normocin (Thermo Fisher Scientific) at 37°C and 5% CO 2 . Cell pairs were obtained by allowing for 1 division cycle (8-to 24-h culture for HeLa; 72-h culture for N2a).

Electrophysiology
Dual whole-cell voltage clamp was performed on cell pairs by using 3-to 5-MV pipettes that were filled with intracellular solution (in mM: 143 CsCl, 10 NaCl, 5.5 glucose, 1 MgCl 2 , 3 HEPES, pH 7.1). Measurements were performed at room temperature. A continuous protocol of 10-mV/10-ms hyperpolarizing pulses was applied sequentially to each cell of the pair at 0.5 Hz by using an analog stimulator (Winston Electronics, St. Louis, MO, USA). For longer-lasting experiments-namely, those that involve studies of Cys-to-Ala mutants-cell triplets were used for double voltage-clamp experiments as these produced more stable recording over many minutes compared with cell pairs. Two voltage-clamp amplifiers (Warner 201A) were used to determine trans-junctional current (I j ). Junctional conductance (G j ) was calculated by dividing I j by the applied voltage. Signals were acquired at 5 kHz and filtered at 1 kHz. Experiments were conducted at room temperature and results were normalized to the value at the start of the experiment. Single channel-unitary currents were obtained after adding 2 mM halothane to the bathing solution to reduce channel open probability, which makes it possible to record single-channel events from cell pairs. This is critical because junctional plaques will inevitably contain many Cx channels, which would be a confounding factor in measuring single-channel properties. A driving force of 30-60 mV was applied to calculate channel-unitary conductance. Current signals were filtered at 100 Hz. pH i was monitored by using an epifluorescence system coupled to a Nikon inverted microscope (Nikon, Tokyo, Japan) using 5-(and-6)-carboxy seminaphtharhodafluor-1 (SNARF-1) (14) loaded into cells as their acetoxymethyl ester (AM; 10 mM) for 10 min (Molecular Probes, Eugene, OR, USA). Excitation at 515 nm was provided by a mercury-arc lamp, and fluorescence at 640 6 20 and 580 6 20 nm was collected by 2 photomultiplier tubes that were equipped with band-pass filters. The fluorescence emission ratio (640/580) was digitized at 5-10 kHz (Digidata 1322A; Molecular Devices, Sunnyvale, CA, USA). Emission ratio was calibrated as previously described (14). pH i imaging, buffering capacity, and H + flux measurement HeLa or N2a cells-AM loaded with 5-(and-6)-carboxy SNARF-1 (10 mM for 10 min)-were imaged confocally (514-nm excitation; emission at 630-650 and 580-600 nm) and calibrated as previously described (14). Hepes-buffered superfusates contained (mM): 20 HEPES, 135 NaCl, 4.  (14).

H + uncaging and apparent junctional H + -permeability measurements
Superfusates contained 1 mM 2-nitrobenzaldehyde, a photolabile, membrane-permeable H + donor substance (15). Uncaging was performed in a 5 3 5-mm region of interest in one cell of a pair-referred to as cell 1 -by 351 nm UV-light-3 flashes-every 3.6 s. The rate of uncaging was estimated from the product of the change in pH i , evoked by a triple-flash event, and buffering capacity (Supplemental Fig. 4). Per uncaging event, 0.3 mM H + ions are released (equivalent to 5 mM H + /min). Acid extrusion by the membrane transporters, NHE1 and Na + -HCO 3 2 , cotransport was inhibited by 30 mM dimethyl amiloride and 150 mM 4,49-diisothiocyano-2,29-stilbenedisulfonic acid included in superfusates. Apparent junctional H + -permeability (P H app ) was calculated from the initial time course of H + permeationconsisting of 40 s of baseline and the first 54 s of H + uncaging-by fitting a diffusion permeation algorithm (16). pH i manipulation pH i was manipulated by superfusion with solutions in which weak acid or base osmotically replaced NaCl and extracellular pH was kept constant at 7.4, unless indicated otherwise. Intracellular acid and alkaline loads were achieved with 40-160 mM Na + acetate and 10-40 mM trimethylamine (TMA), respectively.

Ca 2+ imaging
[Ca 2+ ] i was measured in cell pairs that were AM-loaded with Fluo3 (10 mM; Molecular Probes), excited at 488 nm, and with emitted fluorescence collected at .505 nm. After subtracting background fluorescence, Fluo3 time course signal was normalized to baseline (F/F 0 ).

Statistics
Summarized electrophysiologic results of changes in G j are expressed as means 6 SD, whereas summarized permeability results are expressed as means 6 SEM. A paired Student's t test was used to test the significance between results obtained with each cell pair serving as its own control. An unpaired t test was used to test significance between results that were obtained on different cells pairs. Values of P , 0.05 were considered significant.

RESULTS
Cells transfected with the Cx43 gene become electrically uncoupled at low and high pH i Gja1, the gene that codes for rat Cx43 protein, was stably transfected into HeLa and N2a cells. Cx43 protein was confirmed by Western blot analysis and immunofluorescence (Fig. 1A). Cx43-positive plaques formed at the interface between cells. Junctional conductance (G j )-calculated from double whole-cell voltageclamp pulses (Fig. 1B)-was abrogated by the addition of the gap junctional blocker, b-glycyrrhetinic acid (bGA; 60 mM). These observations are attributable to the transfected gene because wild-type HeLa and N2a cells lack endogenous Cx43 (Fig. 1A) and have no detectable junctional conductance.
Effects of intracellular acidosis were studied in HeLa and N2a cells that were transfected with Cx43. Superfusion with 80 mM Na + acetate produced a prompt acidification (fall of pH i ), followed by a small degree of recovery (Fig. 1C, middle) that was attributable to acid extrusion by transporters, such as NHE (flux analysis in Supplemental Fig. 2). Intracellular acidification first produced a small transient rise in G j , followed by a delayed and slower decrease that reached a nadir of ;70% of the control value (Fig. 1C, bottom). Withdrawal of acetate alkalinized pH i and evoked a rise in G j , which was indicative of a reversible block (Fig. 1C, bottom). These G j changes are additionally quantified in Fig. 1D. Of note, the G j response lagged behind the pH i change by tens of seconds, which indicates a time dependence of acid-evoked block and its subsequent recovery (illustrated as a hysteresis loop in Fig. 1E).
The effect of raising pH i with 20 mM TMA was determined in separate experiments on Cx43-transfected N2a and HeLa cells. Raising pH i produced a monophasic, albeit delayed, fall in G j to ;40% of control conductance (Fig. 1F, bottom). As with the response to acidosis, the effect of raising pH i was reversible (Fig. 1F) and is additionally quantified in Fig. 1G. The G j response, again, was delayed relative to the rapidly imposed pH i maneuvers, as illustrated in the hysteresis plot shown in Fig. 1H. Overall, changes of pH i in either direction from the resting value of ;7.1 reduced junctional conductance, but with a significant time delay. One interpretation of this result would be an activatory effect on G j of Cx43 channel protonation over the alkaline pH i range, but an inhibitory effect of protonation over a more acidic range, with a net transition that occurred near pH i ;6.9-the latter would explain the rise, then fall of G j upon acidification from pH i ;7.1 (see Fig. 1C).
To explore the mechanism of pH i sensitivity of G j , Cx43 single-channel conductance was recorded in HeLa cells-in the presence of 2 mM halothane, which rapidly blocks the majority of channels ( Fig. 2A). Cx43 channels typically exhibit one maximal conductance state that corresponds to the dephosphorylated Ser368 open state (o, ;120 pS), and a residual conductance state (r, ;30 pS) observed upon application of a trans-junctional voltage (13). In addition, a number of events are observed near 60 pS, considered to be the unitary conductance of phosphorylated connexin. Acidosis-attained with 80 mM Na + acetate (

Probing Cx43 channel permeability from measurements of H + ion permeation between cells
Cx43 channel gating was also interrogated from measurements of cell-to-cell permeability to H + ions. To drive a net flux of H + ions through gap junctions, one cell of a pair was acidified by a series of photolytic uncaging reactions that involved the membrane-permeable H + donor, 2nitrobenzaldehyde, which was dissolved in superfusates at 1 mM (15) (Fig. 3A, left). The diffusive spread of uncaged H + ions was then monitored by imaging pH i in the cell pair at intervals between local uncaging events. To obtain an estimate of P H app , [H + ] i time courses in the source and recipient cells were best fitted to a permeation algorithm. It is well established (12,17) that H + ions permeate through connexin channels aboard small, mobile buffer molecules; therefore, calculated P H app is a dual function of the channels' permeability state and the availability of permeant buffers for the shuttling of H + ions through gap junctions.
P H app measurements were first performed on wild-type or Cx43-transfected N2a cells. Transfected N2a cells had bGA-inhibitable junctional conduction (Fig. 3B), which confirmed the functional expression of Cx43. Uncaging experiments were performed on cells that were superfused with CO 2 /HCO 3 2 -free (Hepes) superfusates to measure H + ion permeation aboard intrinsic mobile buffers (Fig.  3C). Acid extrusion by membrane transporters was blocked with 30 mM dimethylamiloride, an NHE1 inhibitor. In agreement with electrophysiologic recordings, wild-type cells displayed no evidence for H + ion permeation between cells (Fig. 3D, left), but significant transmission was observed between Cx43-transfected cells (Fig.  3D, middle), which was inhibitable with bGA ( Fig. 3D, right). P H app data are summarized in Fig. 3Fi. In a second series of experiments, P H app was probed in the presence of 5% CO 2 /22 mM HCO 3 2 buffer, which introduced additional mobile buffering into the Wild-type HeLa and N2a lack Cx43 endogenous. Multiple bands correspond to Cx43 phosphorylation states. Cx43 localizes at the interface between cells (white arrows). Cells were stained by using anti-Cx43 Ab (red). Nuclei were labeled with DAPI (purple). Scale bar, 25 mm. B) Junctional conductance (G j ) assessed by double whole-cell voltage clamp on cell pairs. During alternating voltage pulses, junctional currents (I j ) are a readout of cell-to-cell coupling and I j + I nj represent the junctional plus transmembrane currents in the stimulated cell. C ) Example experiment showing effect of superfusion with 80 mM Na + acetate. Junctional currents (increase indicates stronger coupling; top). Time course of pH i measured by using SNARF-1, showing rapid acidification, secondary partial recovery, and a rebound alkalinization upon washout (middle). Low pH i reduced G j rapidly (bottom). Note the small initial transient increase in G j , followed by a fast decrease that reached ;70% of initial value. D) Average data (mean 6 SD) for G j at baseline conditions, reduction after 3 min of acetate, and recovery after 5 min washout (n = 5, HeLa and N2a cell pairs). E ) Representative plot of pH i vs. G j . After the start position (all white circles), there is a robust reduction in pH i , whereas G j remained constant. This is followed by an increase in G j , then a decrease. Arrow indicates the net change in G j and pH i during the protocol. During washout (all white circles), pH i and G j recover toward initial levels. F ) Example experiment showing effect of superfusion with 20 mM TMA. I j (top). Time course of pH i showing rapid alkalinization, secondary partial recovery, and a rebound acidification upon washout (middle). Alkaline pH i reduces G j toward ;50% of the initial G j value (bottom). G) Average (mean 6 SD) G j at baseline conditions, after 3 min perfusion of TMA, and recovery after 5 min of washout (n = 5 N2a cell pairs). H ) Representative plot of pH i vs. G j plot. Arrow indicates net change in G j and pH i near the steady state. *P , 0.05.
cytoplasm. Acid extrusion was blocked pharmacologically with 30 mM dimethylamiloride and the HCO 3 2 transport inhibitor, 4,49-diisothiocyano-2,29-stilbenedisulfonic acid (150 mM; Supplemental Fig. 2). P H app was 2-fold higher under this buffering regime (Fig. 3E, Fii), which was consistent with the junctional permeation of additional mobile buffers provided by cytoplasmic CO 2 /HCO 3 2 (15). Paracellular diffusion of CO 2 was ruled out on the basis of the absence of cell-to-cell H + ion transmission in wild-type cells and the complete block of transmission by bGA in Cx43expressing cells (Fig. 3Fii). Similar results were obtained with HeLa cells (Supplemental Fig. 3). H + ion permeation time courses were analyzed in terms of junctional H + flux (Supplemental Fig. 4). For an uncaging rate of ;5 mM H + /min, at pH i ;6.95, the H + ion flux through Cx43 channels was 1.0 mM H + /min in Hepes buffer and 1.7 mM H + /min in CO 2 /HCO 3 2 buffer (Supplemental Fig. 5).

Biphasic regulation of Cx43 channels' H + ion permeability by cytoplasmic H + ions
In the next series of experiments, Cx43 channel activity was characterized over a range of pH i values by using the protocol for measuring P H app in Cx43-transfected HeLa cells. Experiments were performed in Hepesbuffered superfusates to ensure that only the intrinsic mobile buffers are responsible for shuttling H + ions between cells. Use of weak acids or bases to manipulate pH i -as was done in electrophysiologicl recordings (Fig. 1)-was now avoided to eliminate the possibility of exogenous weak acid-base-facilitated H + ion permeation, which may, in principle, supplement the effect of intrinsic buffers. Instead, pH i was manipulated by means of a prior solution maneuver that shifted pH i to a new level (prepulse). Cells were superfused with a weak acid (or base) for 5 min, then returned to normal solution. pH i then rebounded to an alkaline (or acidic) level. At the displaced level of pH i , measurements of P H app could be made with no interference from the exogenous weak acid-base because, at that point, the exogenous substance had been washed away. Superfusion with 30 mM NH 4 Cl, followed by washout in normal Tyrode-containing 30 mM dimethylamiloride (NHE1 inhibitor), induced a sustained intracellular acid load (Fig. 4A), whereas superfusion with 80 mM acetate, followed by washout with normal Tyrode at pH 8.4, induced an intracellular alkali load (Fig. 4B). (Note: alkaline superfusates block any membrane transporters that might otherwise acidify cytoplasm.) In calculating P H app , the slow recovery of pH i from an acid or alkali load was corrected by extrapolating the trend line that was measured before photolysis. Cellto-cell H + ion transmission was reduced substantially at acidic and alkaline pH i , which confirmed biphasic pH i gating (Fig. 4A, B).
To assess whether the inhibitory effect of acute acidosis and alkalosis on P H app was reversible, repeat measurements were made once pH i had stabilized nearer to resting levels (N2A cells). As expected from a reversible H + -dependent block, P H app was restored to control levels upon pH i recovery from an acidic (Fig.  4C) or alkaline load (Fig. 4D). In summary, measurements of the pH i dependence of P H app confirm the observations-made using an electrophysiologic approach (Fig. 1)-that Cx43 channels are gated by low and high pH i within minutes and that this process is fully reversible. A plot of P H app vs. pH i -x-coordinate taken as the pH i in the source cell averaged over a 1-min period of uncaging-demonstrates that acidification from 6.9 to 6.6 decreased Cx43 P H app by 80%, whereas alkalinization from 6.9 to 7.3 caused P H app to decrease by 60% (Fig. 4E).
To investigate whether changes in the phosphorylation of serine, particularly residue Ser368, underpin the decrease in G j at acidic or alkaline pH i , additional experiments were performed on Cx43 mutants [Cx43(S-A) 6 ] that had all serine residues of the PKC epitope, including Ser368, substituted by alanine, i.e., a phospho-null variant. This mutant, expressed in HeLa cells, retained the response to acetate and TMA, which argues against a role for phosphorylation changes in pH i dependence (Fig. 4E).
At least part of the pH i dependence of P H app may be attributed to changes in the availability of mobile buffers-that is, the fraction of buffering capacity held on small molecules (F = b mobile /b total ), such as CO 2 /HCO 3 2 , phosphates, and dipeptides that shuttle H + ions through Cx43 channels (17). Mathematically, P H app can be deconvoluted as the product of F and P mobile (pH-gated channel permeability to mobile buffers). In principle, both P mobile and F can display pH dependence, but only the former relates to Cx43 channel activity. To confirm that the pH i sensitivity of P H app (Fig. 4E) reflects, at least in part, the H + gating of Cx43 channels, it is necessary to rule out a biphasic pH i sensitivity of F. This was evaluated by considering 2 models. In the first, b mobile was assumed to be a constant fraction of b total -that is, F would take a pH-independent constant value (0 , F ,1) and, consequently, the observed pH i dependence of P H app would purely be a phenomenon of the pH i sensitivity of P mobile . In an alternative model, b mobile and b total are assumed to have opposite pH dependence-that is, F is not constant. In most cells, including N2a cells, the pH i dependence of intrinsic b has a negative slope over the physiologic pH i range [Supplemental Fig. 4; in N2a: b total = 95211 3 pHi (millimolar)]. This negative relationship arises because fixed buffers, mainly proteins, which dominate the intrinsic buffer pool demonstrate peak buffering in the acidic range. In contrast, less abundant mobile buffers typically have a more alkaline pH i optimum (18). Effectively, F would be a positive function of pH i , thereby eliminating the possibility that alkaline inhibition of P H app is caused by a fall in F; however, the reduction in P H app that was observed at low pH i may still relate to a decrease in F. This, however, is not quantitatively consistent with experimental observations. A decrease in pH i from 7.3 to 6.6 was associated with an 80% decrease in P H app , which cannot be explained by a decrease in F alone because buffering capacity is noncooperative and, therefore, not associated with a steep pH i dependence. In summary, the biphasic shape of the pH i -P H app relationship, at least in part, must be a result of H + gating of Cx43 channels. Furthermore, the alkaline part is wholly attributable to an H + gating of Cx43 channels.
Biphasic H + gating of Cx43 channels is not explained by pH-evoked Ca 2+ signals Changes in pH i can displace Ca 2+ ions from intracellular buffers or subcellular compartments (19). As robust C) Schematic representation of the different coupling states from which permeability data are obtained (wild-type, Cx43 expressor, and Cx43 expressor inhibited with bGA). D) Cell-to-cell H + ion permeation measured in Hepes-buffered superfusates in N2a cells (Cx43-negative). To highlight pH i changes in the recipient cell (cell-2), the dashed line extrapolates baseline. Substantial acidification of cell-2 is observed in Cx43 transfectants and is bGA sensitive. E) Cell-to-cell H + ion permeation measured in CO 2 /HCO 3 2 buffered superfusates. For panels D, E, the shaded regions correspond to the period we used to fit data using the mathematical models. Purple bands correspond to the uncaging period (flash photolysis). F ) Summary of P H app data. All H + uncaging experiments were performed in the presence of 30 mM dimethylamiloride to inhibit acid extrusion by the NHE. Experiments in CO 2 /HCO 3 2 buffer were also performed in the presence of 150 mm 4,49-diisothiocyano-2,29-stilbenedisulfonic acid. # Significantly different from 0 (P , 0.05). *P , 0.001. cytoplasmic Ca 2+ signals are known to produce Cx43 channel closure (5), the acidic responses of G j and P H app may, in principle, involve an increase in [Ca 2+ ] (20). To investigate whether the pH i sensitivity of Cx43 function had an underlying Ca 2+ -dependent component, pH i gating responses were measured in cells that were preloaded with the Ca 2+ buffer, BAPTA (AM ester; 100 mM), which minimized any acidosis-evoked changes in [Ca 2+ ] (Fluo3 fluorescence). Prepulsing with 30 mM NH 4 Cl or 80 mM Na + acetate transiently increased Fluo3 fluorescence, but this response was ablated in cells that were preloaded with BAPTA (Fig. 5A). The effect of BAPTA on the pH i dependence of P H app was investigated by using the photolytic protocol performed in CO 2 /HCO 3 2 -free superfusates. H + uncaging induced an increase in cytoplasmic [Ca 2+ ] (Fig. 5B), but cells that were pretreated with BAPTA showed no acid-evoked [Ca 2+ ] response, yet retained pH sensitivity of P H app (Fig. 5C, E). As additional confirmation for the absence of a meaningful Ca 2+ -dependent component, the inclusion of 5 mM BAPTA in the pipette solution had no effect on G j response to 80 mM Na + acetate or 40 mM TMA (Fig. 5D).
C tail of Cx43 is linked to the mechanism of alkaline gating of channel activity Earlier studies have implicated the C tail of the Cx43 protein in the response to low pH i (21,22). To investigate whether this Cx43 domain is also responsible for alkaline gating, a His-tagged Cx43m257 truncated mutant, referred to as tailless, was transfected into N2a cells (7). Immunostaining using His-tag Abs confirmed the correct targeting to the cell-cell interface (Fig. 6A). The mutant retained its sensitivity to bGA (Fig. 6B), but, unlike the full-length construct, cell-to-cell pair conductance demonstrated no response to alkaline pH i (Fig. 6C, D). Moreover, acidification or alkalinization of N2a cells that were transfected with the mutant did not inhibit H + ion transmission between coupled cells (Fig. 6E, F). In double whole-cell voltage-clamp experiments under superfusion with 20 mM TMA, tailless mutant channel gating did not demonstrate a residual state, but its main conductance state was preserved (Fig. 6G), as previously reported at resting pH i (13). In summary, the C tail of Cx43 is involved in mediating both acidic and alkaline gating of assembled Cx43 channels (Fig. 6H). showing reduced transmission. B) Resting pH i preadjusted to alkaline before H + uncaging by 80 mM Na + acetate prepulse (returned to normal Tyrode at pH 8.4 to slow pH i recovery). pH i time course (bottom) for estimating P H app shows significant reduction in transmission. C) Acid block was reversed after allowing 10 min of recovery from acidic pH i . D) Alkali block was reversed after allowing 10 min of recovery from alkaline pH i . Shaded regions correspond to the period we used to fit data using mathematical models. Purple bands correspond to the uncaging period (flash photolysis). E) Summary of reversibility results were fitted to a bell curve. Added to this data are the results obtained from acidosis and alkalosis experiments performed by using Cx43 mutants where all serine residues from the PKC epitope at the end of the C tail have been substituted by alanine residues (SA6). Activation and inhibition by H + ions are highlighted by gray arrows. # Significantly different from 0.
C tail cysteine residues are implicated in the mechanism of Cx43 channel gating by alkaline pH i Histidine residues of the C tail of Cx43 have been implicated in the gating of channel activity by low pH i (9), which is consistent with the residue's pK a of ,7.0. Other residues of the C tail with higher pK a are plausible candidates for the alkaline gating. Of these, cysteine, arginine and tyrosine have a pK a of .8 and are found in the Cx43 protein (Supplemental Fig. 1). A related connexin isoform, Cx45, which demonstrates greatly attenuated alkaline gating (Supplemental Fig. 7), contains no cysteine in its C tail; thus, these residues are plausible candidates for alkaline gating. To assess whether cysteines that are naturally found in the C tail of Cx43 are linked to alkaline inhibition, residues at positions 298, 271, or 260 were mutated to alanine in various permutations and their effect on G j and P H app was probed in N2a cells. Whereas cysteine, as a free amino acid, has a nominally alkaline pK a , which renders it suitable as an alkaline sensor, it is important to consider how adjacent residues in a motif affect its ionization state. The 3 residues that flank either side of Cys298 form a NSSCRNY motif, which has an alkaline calculated isoelectric point of pK a = 8.5. In contrast, the sequence that flanks Cys260 and Cys271-AKDCGSQ and FNGCSSP, Figure 5. Ca 2+ signals are not involved in the alkaline-gating mechanism. A) Prepulsing with 30 mM NH 4 Cl (left) or 80 mM acetate (right) normally increases [Ca 2+ ] (Fluo3), but this is attenuated in cell pairs that are AM-loaded with 100 mM BAPTA. B) BAPTA also ablates Ca 2+ response to H + uncaging. C ) BAPTA did not affect H + ion permeability, nor its pH i dependence. Shaded regions correspond to the period we used to fit data using mathematical models. Purple bands correspond to the uncaging period (flash photolysis). D) Inclusion of 5 mM BAPTA in the pipette solution during whole-cell voltage-clamp experiments had no effect on G j response to 80 mM acetate or 40 mM TMA. Controls to this experiment (BAPTA-free pipettes) are shown in Fig. 1. E) Comparison of P H app calculations between control (white circles) and BAPTA (green circles) experiments. Also included are results from nontransfected N2a cells (red circle) and Cx43transfected cells that were treated with 60 mM bGA (yellow circle). # Significantly different from 0.
respectively-have ensemble isoelectric points of 6.2 and 5.9, respectively. Thus, Cys260 and Cys271 are less likely to be alkaline sensors because the adjacent, negatively charged glutamate greatly shifts the apparent pK a of the motif toward acidic levels. For this reason, a Cys298Ala substitution was first Cx43 mutant constructed. As shown in Fig. 7Ai, the Cys298Ala mutant was insensitive to alkaline pH i (TMA superfusion), but retained sensitivity to acidic pH i (acetate superfusion). To investigate whether if either of the 2 other Cys residues have a supplementary effect, mutants with additive Cys-to-Ala substitutions were expressed. As shown in Fig. 7Aii, iii, these additional Cys substitutions had no additional effect on alkaline or acidic responses, which argues that Cys298 is likely to be the principal alkaline-sensing residue in Cx43 protein.
Overall, these results can be interpreted in terms of a 2-site model: an inhibitory domain that, upon protonation, leads to channel closure over the acidic pH i range, and an activatory domain that, upon protonation, leads to channel opening over the alkaline pH i range (Fig. 7B). Figure 6. Role of Cx43 C-tail domain in pH i gating. A) Confocal image showing the expression of truncated Cx43 (Cx43m257) tagged with a Histag domain at the end of the C terminus (red). DAPI staining identified nuclei (blue). B) N2a cells expressing truncated Cx43 become uncoupled after the application of 60 mM bGA (2 separate experiments in red and blue). The effect was reversible (black). C ) Junctional current (I j ) does not change during 40 mM TMA perfusion during the application of voltage pulses shown in Fig. 1B. D) G j in HeLa cells expressing Cx43m257 (n = 3) or wild-type Cx43 (n = 12) during superfusion with 40 mM TMA. E ) Cartoon representation of Cx43m257 protein, highlighting the lack of C tail (left). Resting pH i was preadjusted to acidic level before H + uncaging by 30 mM NH 4 Cl prepulse returned to normal Tyrode that contained 30 mM dimethylamiloride to inhibit pH i recovery (middle). Resting pH i was preadjusted to the alkaline level before H + uncaging by 80 mM acetate prepulse returned to normal Tyrode pH 8.4 to slow pH i recovery (right). F ) Transmission of acid indicates the persistence of Cx43m257 permeation during acidosis (middle) or alkalosis (right) at levels that were comparable to wildtype Cx43 near resting pH i (left). G) Single channel events recorded from N2a-Cx43m257 cells in 2 mM halothane show open conductance that is similar to wild-type channels. H ) P H app data compiled from all experiments indicate a lack of sensitivity to acidosis or alkalosis of the tailless mutant (from left to right: n = 11, 5, 13). # Significantly different from 0.

DISCUSSION
The results of this study demonstrate that Cx43 channels, transfected into HeLa or N2a cells, replicate the pH idependent gating behavior that has been previously described in cardiac myocytes (12), a type of cell with naturally high Cx43 expression. Thus, heterologously expressed Cx43 channels are a good model with which to study the gating mechanism of Cx43 channels by H + ions. In addition, the similarity in biophysical behavior argues that the biphasic pH i dependence of cell-to-cell coupling demonstrated in cardiac myocytes is not a function of cellular context or molecular components that are unique to the myocardium, but, rather, an intrinsic property of the Cx43 protein. The biophysical properties of heterologously expressed Cx43 channels were interrogated electrophysiologically by measuring junctional conductance and imaging H + ion transmission to quantify a permeability constant. The former, double patch-clamp technique is a validated and robust method for determining connexin channel gating in response to various biophysical or chemical agents (13). In contrast, H + ion transmission-assessed by P H app -is a more recently developed approach (15) that probes connexin activity by using an independent readout and has the advantage of measuring coupling in intact cells that are not impaled by electrodes-for example, without the potential problem of intracellular dialysis.
To map the pH i dependence of Cx43 function, this dual measurement approach was applied for a range of pH i levels that were manipulated by using weak acids or bases. Previously, changes in pH i have typically been induced by altering CO 2 partial pressure (22), adjusting extracellular pH, or inducing either quasi steady-state pH shifts (21) or shifts of .1 pH unit (7,9,23). The major disadvantages of the above-mentioned approaches are the simultaneous modification of pH on either side of the cell membrane and the longer times that are necessary for attaining a target pH i change. These factors make it difficult to disentangle the effects on intra-and extracellular pH. Although some early reports have claimed that coupling is insensitive to external acidification (24), increased extracellular pH is now recognized to play a role in regulating Cx function by favoring disulfide bridge formation and Cx docking (25). This effect becomes more relevant in longer-lasting experiments during which pH i changes are induced slowly and within the time frame of Cx43 protein turnover (10). Our approach to clamping pH i involved the addition-or addition followed by washout-of weak acids-bases. In their uncharged forms, these molecules permeate the cell where they dissociate or combine with H + ions, thereby changing pH i without altering extracellular pH. In our protocols, pH i changes were induced in a matter of seconds, and their effects on Cx43 channel gating were examined simultaneously or within a few minutes.
By demonstrating a biphasic pH i sensitivity of Cx43 function (Fig. 4E), the results of this study add to the arguments against the canonical sigmoidal pH i dependence (21). Although it is widely recognized that profound intracellular acidification closes Cx43 channels (7,9,21), less is known about the effect of increasing pH i . Apart from our earlier study that described a dual response of the ventricular gap junction to H + ions (12) as well as some evidence in the literature of an alkali block of Cx43 channels (21), only 2 contradictory observations have been made for Cx43 channel activity at increased pH i . One of these reported an 85% G j decrease at pH i 8.1 in Novikoff cells that endogenously expressed Cx43 (26), whereas the other describes a 10% G j increase at high pH i for Cx43 expressed in Xenopus oocytes (23). The reasons for the discrepancy between our work and the latter study are not clear, although the slow induction of alkalinization and high extracellular pH may be relevant. In addition, Xenopus oocytes express Cx38 channels (27) that may contribute to the alkaline-induced increase in coupling.
A noteworthy feature of the G j response to pH i is its significant time delay (;10 or more), as shown in Fig. 1C, F. This delay was also reported in earlier work (21,28,29) and may suggest a pH i -evoked cooperative conformational change that involves all 6 monomers of a Cx channel or the participation of accessory diffusible molecules that are also present in expression systems (30). The time scale of conductance responses cannot, per se, exclude a role for protein internalization in the underlying mechanism. Although fast internalization (seconds) has been described for some membrane proteins, such as synaptic vesicle proteins (31), the internalization of gap junction plaques involves a more complicated process (32). Previous studies have demonstrated that Cx43 internalization can be induced during a 30-min period of acute ischemia (33), but, paradoxically, this occurred with no apparent change in junctional conductance because of a compensatory delivery of new hemichannels to the junctional plaque (34,35).
Moreover, the process of reinsertion-that is, reversal of internalization-is expected to be slower because it has to allow sufficient time for the proper docking of juxtaposed hemichannels. Cx43 recycling has been reported to last 30 min (36), have a half-life of ;2 h in cultured cardiomyocytes (10,37), or require ;2 h to complete during mitosis (38). In contrast, our data (Fig.  1) show a similar response time for inhibition-TMA or acetate-and subsequent recovery (removal of TMA/ acetate), which is not consistent with the time scale of Cx43 internalization and reinsertion. Together with evidence for complete and fast reversibility, our findings point toward a gating mechanism.
Our results provide insight into the molecular mechanism that underpins the biphasic effect of H + ions on Cx43 channels. The absence of an apparent effect of pH i changes on single-channel conductance (Fig. 2)-also shown previously by others (39)-argues against a pH-dependent alteration in the assembly of Cxs around the central pore, one possible form of gating. The small shift toward 60 pS in the single-channel event distribution during alkalosis may suggest an effect of an altered phosphorylation state; however, experimental data presented in Fig. 7 on the alkaline response of phospho-null Cx43 mutants argue against this. In these experiments, HeLa cells expressed mutant Cx43 with all serine residues [Cx43(S-A)6] of the PKC epitope site substituted for alanine-from 364 to 373, including Ser368, which is responsible for the unitary conductance shift by phosphorylation. These mutants reached levels of G j uncoupling and H + flux that were similar to those of wildtype Cx43, which indicated that changes in phosphorylation state cannot explain the response to alkaline or acidic pH i . Moreover, Western blots presented in Supplemental Fig. 7 indicate that the ratio between phosphorylated and dephosphorylated Cx43 remains unchanged after a period of intracellular alkalosis or acidosis. The persistence of biphasic pH i gating after loading cells with the Ca 2+ buffer, BAPTA, argues against the involvement of acid-evoked [Ca 2+ ] signals as intermediates of gating (40). Of note, at least a partial involvement of Ca 2+ ions in the acid-gating mechanism has been proposed to take place in cardiac myocytes and Novikoff hepatoma cell pairs (12,41), yet this may be a function of cellular context. Previous studies have indicated that changes in Cx43 phosphorylation state at serine residues of the C terminus can influence Cx assembly, their gating, and half-life (42); however, our results demonstrate that a 5-min treatment of Cx43-transfected N2a cells with 80 mM Na + acetate or 30 mM NH 4 Cl did not affect the migration pattern of Cx43 immunoreactivity, which argues against any major shift in phosphorylation state (Supplemental Fig. 7), although certain single-residue changes in Cx43 phosphorylation, such as at Ser368, can occur without detectable migration shifts (42). However, a post-translational change at Ser368 would become apparent from a reduction in unitary conductance (43), which was not observed in our recordings at acid or alkali pH i (Fig. 2). In summary, Ca 2+ signals and phosphorylation state do not seem to play a role in rapid, pH i -induced gating behavior. Instead, the gating mechanism is likely to be an inherent property of the Cx43 protein, which is recapitulated when the channel is assembled in expression systems.
Our measurements on cells that were transfected with Cx43 mutants suggest that C-tail cysteine residues are involved with the mechanism of alkaline gating. First, a tailless Cx43 mutant loses its response to alkaline pH. This observation resembles the behavior of the ball-and-chain inactivation that has been observed in many types of ion channel. Second, substituting cysteine, a residue with an alkaline pK a , with nontitratable alanine attenuates alkaline gating. This is the first report, to our knowledge, to implicate these C-tail cysteines in Cx channel gating by pH i , although residues, such as Cys271, have been linked to redox responses (44) and S-nitrosylation by nitric oxide (45). Third, channels formed of Cx45, an isoform that is related to Cx43 but that lacks these critical cysteine residues, demonstrates an attenuated alkaline response compared with Cx43 channels (Supplemental Fig. 6).
The C-tail domain of Cx43 is also responsible for the inhibitory effect of H + ions, but here, the residue that underlies the inhibitory effect is a histidine that is located at the intracellular loop, with an accordingly lower pK a (9,21,22). The structural basis for the inhibition involves a stabilizing effect of histidine protonation on the a-helical order of the cytoplasmic loop that then favors intramolecular interactions (9,46).
The superimposition of an activatory effect of H + ions-titrated from an alkaline pH i -and an independent inhibitory effect-occurring at more acid pH i -produces an overall pH i range that permits physiologic cell-to-cell communication. The finding that peak Cx43 channel function occurs in the range of 6.9-7.0 pH i (Fig. 4E) indicates that modest acidification from a resting value of ;7.2 will strengthen cell-to-cell coupling. This response would favor syncytial dissipation of any pH i gradients-thereby unifying pH i -dependent processes-and facilitate tissue functions that rely on cell-to-cell communication, such as the transmission of electrical signals in the heart. The basis for this response is explained by H + activation, which underlies the alkaline range of the biphasic pH i sensitivity of the Cx43 channel. A more profound acidification leads to uncoupling, which can be interpreted as a protective strategy to mitigate for the potential consequences of toxic levels of acid spillover into neighboring cells. In the heart and brain, for instance, this mechanism would help protect cells around ischemic areas and reduce tissue damage propagation (2,3).
In summary, we demonstrate the phenomena of H + activation and H + inhibition of Cx43 channels as independent channel gating mechanisms. The ensemble of H + activation and inhibition produces a biphasic pH i dependence of Cx43 channel function. This bell-shaped pH i sensitivity ensures that gap junctions are responsive to changes in cytoplasmic acid-base chemistry, which produces a physiologic range of pH i that permits intercellular communication in tissues, such as the heart and brain.