Gating, permselectivity and pH-dependent modulation of channels formed by connexin57, a major connexin of horizontal cells in the mouse retina

Abstract

Mouse connexin57 (Cx57) is expressed most abundantly in horizontal cells of the retina, and forms gap junction (GJ) channels, which constitute a structural basis for electrical and metabolic intercellular communication, and unapposed hemichannels (UHCs) that are involved in an exchange of ions and metabolites between the cytoplasm and extracellular milieu. By combining fluorescence imaging and dual whole-cell voltage clamp methods, we showed that HeLa cells expressing Cx57 and C-terminally fused with enhanced green fluorescent protein (Cx57-EGFP) form junctional plaques (JPs) and that only cell pairs exhibiting at least one JP demonstrate cell-to-cell electrical coupling and transfer of negatively and positively charged dyes with molecular mass up to ∼400 Da. The permeability of the single Cx57 GJ channel to Alexa fluor-350 is ∼90-fold smaller than the permeability of Cx43, while its single channel conductance (57 pS) is only 2-fold smaller than Cx43 (110 pS). Gating of Cx57-EGFP/Cx45 heterotypic GJ channels reveal that Cx57 exhibit a negative gating polarity, i.e. channels tend to close at negativity on the cytoplasmic side of Cx57. Alkalization of pH_i from 7.2 to 7.8 increased gap junctional conductance (g_j) of ∼100-fold with pK_a= 7.41. We show that this g_j increase was caused by an increase of both the open channel probability and the number of functional channels. Function of Cx57 UHCs was evaluated based on the uptake of fluorescent dyes. We found that under control conditions, Cx57 UHCs are closed and open at [Ca²⁺]_o=∼0.3 mm or below, demonstrating that a moderate reduction of [Ca²⁺]_o can facilitate the opening of Cx57 UHCs. This was potentiated with intracellular alkalization. In summary, our data show that the open channel probability of Cx57 GJs can be modulated by pH_i with very high efficiency in the physiologically relevant range and may explain pH-dependent regulation of cell–cell coupling in horizontal cell in the retina.

Connexins (Cxs), a large family of membrane proteins, form gap junction (GJ) channels that provide a direct pathway for electrical and metabolic signalling between cells (Harris, 2001; Sohl et al. 2005). The docking of two apposed hemichannels (AHCs) at the cell–cell contact forms a GJ channel. Connexin-based hemichannels found at the non-junctional plasma membrane, called unapposed hemichannels (UHCs), connect the cell interior with the extracellular milieu. They have been implicated in physiological functions including paracrine intercellular signalling and induction of cell death under pathological conditions (Saez et al. 2003). Functional properties of Cx-based channels vary notably depending on the Cx isoform and can be modulated by voltage, phosphorylation, [Ca²⁺]_i, pH, arachidonic acid and many other lipophilic uncouplers (Baldridge et al. 1987; Bennett et al. 1991; Bukauskas & Verselis, 2004; Jouhou et al. 2007b).

The retina is a highly organized laminar structure with three nuclear layers where cell bodies are located and two plexiform layers where the dendritic interactions take place. Seven major cell types compose the retinal layers and most of them exhibit gap junctional communication (Vaney & Weiler, 2000). Neurobiotin cell–cell transfer studies have revealed extensive coupling between horizontal, bipolar, amacrine or photoreceptor cells (McMahon et al. 1989; Becker et al. 1998; Hombach et al. 2004). Several Cx isoforms, such as Cx30.2, Cx36, Cx37, Cx43, Cx45 and Cx57, have been reported to be expressed in the mouse retina with a cell type-dependent expression pattern (Sohl et al. 2005; Kreuzberg et al. 2008). In addition, it was concluded that Cx26 is expressed in the horizontal cells of carp (Janssen-Bienhold et al. 2001) but not mice (Deans & Paul, 2001).

The first excitatory synaptic relay within the retinal circuit is located in the outer plexiform layer where photoreceptors project to horizontal and bipolar cells. The lateral contacts between photoreceptors and horizontal cells are thought to be the origin of a feedback pathway that modulates light adaptation and luminance contrast, and forms the basis of the centre–surround antagonism of bipolar and ganglion cells (Kuffler, 1953); for review see (Kamermans & Spekreijse, 1999). Several hypotheses for this feedback have been proposed including a GABA-dependent pathway (Kaneko & Tachibana, 1986), changes in the concentration of hydrogen ions in the synaptic cleft of the photoreceptor terminal (Hirasawa & Kaneko, 2003; Davenport et al. 2008; Vessey et al. 2005; Cadetti & Thoreson, 2006; Jouhou et al. 2007a) and hemichannel-mediated ephaptic inhibition of photoreceptors (Kamermans et al. 2001; Kamermans & Fahrenfort, 2004). It was reported that horizontal cells form a large syncytial network in the mouse retina via GJs formed for the most part of Cx57 (Hombach et al. 2004; Ciolofan et al. 2007; Janssen-Bienhold et al. 2009). This gap junctional communication between horizontal cells has been considered a structural substrate for a large receptive field (Hampson et al. 1992). Gap junctional communication between horizontal cells can be modulated by multiple neuromodulators (McMahon, 1994; Mills & Massey, 1995; Weiler et al. 2000), arachidonic acid (McMahon, 1994), [Ca²⁺]_o (Teranishi et al. 1983; 1984; Gallemore et al. 1994; Rabl & Thoreson, 2002) and pH (DeVries & Schwartz, 1989; Hampson et al. 1992; Jouhou et al. 2007b). However, studies with Cx57 knockout mice have suggested that Cx57 expression may not be critical for spatial tuning and light adaptation of ganglion cells (Dedek et al. 2008) although receptive field size and resting potential of horizontal cells are affected (Shelley et al. 2006). Despite extensive studies performed on this matter, the nature of the feedback and role of Cx57 in the mouse retina remains a subject of discussion.

We analysed HeLa transfectants expressing wild-type Cx57 or Cx57 fused to its C-terminus with an enhanced green fluorescent protein (Cx57-EGFP). Originally, it was reported that Cx57 is coded by a single exon and HeLa cells expressing Cx57 were generated based on the predicted sequence. Their electrical and permselectivity properties were addressed in Manthey et al. (1999). Later, DNA studies in mouse horizontal cells revealed that Cx57 is coded by a second exon spliced to the coding region of the first exon causing a shorter carboxy-terminal region (Hombach et al. 2004). Using combined fluorescence imaging and dual whole-cell patch clamp methods, we examined the voltage gating, permselectivity and pH-dependent modulation of GJ channels in HeLa cells expressing the corrected sequence of Cx57. Functional properties of Cx57 UHCs were evaluated based on dye uptake during time-lapse imaging over ∼5 h. We found similar voltage gating properties as previously reported (Manthey et al. 1999) but with higher single channel conductance. Under control conditions at pH_i 7.2, junctional conductance (g_j) was very low, and g_j increased ∼100-fold during relatively weak intracellular alkalization. We found that this g_j increase was caused by a reduction of V_j-gating sensitivity of AHC and an increase in the number of functional GJ channels. In addition, dye uptake through Cx57 UHCs increased at moderately low levels of [Ca²⁺]_o (∼0.3 mm or below) and is potentiated by intracellular alkalization.

Methods

Cloning of retinal Cx57

A pMJ-Cx57eGFP vector was generated for transfection of HeLa cells. Briefly, a NcoI/BamHI mouse Cx57 PCR fragment (Position 1 to 1476) (Hombach et al. 2004) was cloned into the NcoI/BamHI sites of pMJGreen (Maxeiner et al. 2005), which contained the CMV promoter, a polyadenylation signal and a gene that conferred resistance to puromycin. To generate the pMJ-Cx57-IRES-eGFP vector, the NcoI/BamHI fragment of Cx57 and an EcoRI/NotI fragment of IRES-eGFP from pIRES-eGFP-zeo+ (Dobrowolski et al. 2007) were cloned into the NcoI/NotI sites of pMJ-Green.

Cells and culture conditions

Experiments were performed using HeLa cells (human cervix carcinoma cells, ATCC CCL2). Cells were grown in Dulbeco's modified Eagle's medium supplemented with 8% fetal calf serum, 100 μg ml⁻¹ streptomycin and 100 units ml⁻¹ penicillin. Cells were maintained in the CO₂ (5%) incubator at 37°C. For the transfection procedure we used 2 μg of pMJ-Cx57eGFP or pMJ-Cx57-IRES-eGFP DNA and Lipofectamine2000 (Invitrogen, Carlsbad, CA, USA) following the manufacturer's protocol. Clones were selected after cells were grown in the presence of puromycin, 1 μg ml⁻¹ (Sigma, Dreisenhofen, Germany).

Electrophysiological recordings and analysis

Experiments were performed in a modified Krebs–Ringer (MKR) solution containing (in mm): NaCl, 140; KCl, 4; CaCl₂, 2; MgCl₂, 1; glucose, 5; pyruvate, 2; Hepes, 5 (pH 7.4). Electrodes were filled with a pipette solution containing (in mm): KCl, 140; NaAsp, 10; MgATP, 3; MgCl₂, 1; CaCl₂, 0.2; EGTA, 2; Hepes, 5 (pH 7.2). For simultaneous electrophysiological and fluorescence recordings, cells were grown onto glass coverslips and transferred to an experimental chamber mounted on the stage of an inverted microscope (Olympus IX70) equipped with a fluorescence imaging system. Cells were perfused with MKR solution at room temperature. Junctional conductance (g_j) was measured in selected cell pairs using the dual whole-cell patch clamp system. Briefly, each cell of a pair was voltage clamped independently with a separate patch clamp amplifier. Transjunctional voltage (V_j) was induced by stepping the voltage in cell-1 (ΔV₁) and keeping the other constant, V_j=ΔV₁. Junctional current (I_j) was measured as the change in current in the unstepped cell-2, I_j=ΔI₂. Thus, g_j was obtained from the ratio −I_j/V_j, where the negative sign indicates that the junctional current measured in cell-2 is oppositely oriented to the one measured in cell-1. To avoid effects of series resistance on g_j measurements (Wilders & Jongsma, 1992), we selected cell pairs exhibiting typically small junctional plaques (JP), and maintained the pipette resistances below 5 MΩ. Typically, cell pairs without JPs showed no coupling. Signals were acquired and analysed using custom-made software (Trexler et al. 1999) and A/D converter from Axon Instruments.

Fluorescence imaging and dye transfer studies

Fluorescence signals were acquired using an ORCA digital camera (Hamamatsu Corp., Bridgewater, NJ, USA) with UltraVIEW software for image acquisition and analysis (Perkin Elmer Life Sciences, Boston, MA, USA).

For dye transfer studies, a given dye was introduced into cell-1 of a pair through a patch pipette in whole-cell voltage clamp mode. Dyes used include (molecular mass of the fluorescent ion, valence): Lucifer yellow (LY) (443, −2), Alexa Fluor-350 (AF³⁵⁰) (326, −1), ethidium bromide (EtBr) (314, +1), DAPI (279, +2) and propidium iodide (PrI) (415, +2) (Invitrogen, Eugene, OR, USA). Typically, this resulted in a rapid loading of cell-1 followed by dye transfer to the neighbouring cell-2. In our earlier studies of single channel permeability of GJs, a whole cell recording in the dye recipient cell (cell-2) was established ∼6–10 min after opening the patch in cell-2. This allowed us to measure g_j and avoid dye loss due to its diffusion from cell-2 to the patch pipette.

To account for changes of fluorescence intensity in both cells as well as g_j we used the Goldman–Hodgkin–Katz (GHK) equation (Hille, 2001) adapted for junctional permeability (P_j) measurements (Verselis et al. 1986). The transjunctional flux (J_j) of dye through gap junction channels is determined by change in dye concentration in cell-2 (ΔC₂) over time interval (Δt) so that:

(1)

where vol₂ is the volume of cell-2. Then, according to a modified GHK equation (Verselis et al. 1986), the total junctional permeability (P_j) can be described as follows:

(2)

where z is the net charge of dye molecule, F is Faraday's constant, R is the gas constant, T is absolute temperature, and C₁ and C₂ are dye concentrations in cell-1 (dye-donor) and cell-2 (dye-recipient), respectively. In the present study, we examined P_j in the absence of a transjunctional electrical field (V_j= 0 mV). Therefore, eqn (2) can be transformed as follows:

(3)

As we indicate below, g_j of Cx57-EGFP junctions at control (normal) conditions was relatively low and to improve dye transfer, we used ammonium chloride. Under these conditions we needed to measure g_j continuously. However, when both patch pipettes are in the whole-cell recording mode, a fraction of dye molecules that are transferred to the cell-2 through GJ channels will diffuse to the patch pipette on cell-2. This can be accounted by writing an equation in which a change of the concentration in the cell-2 (ΔC₂) over time Δt is described as follows:

(4)

where P_p is the permeability of dye from cell-2 to pipette-2 (see eqn (7)). Then from eqn (4) it follows:

(5)

Cell volume was approximated as a hemisphere. The diameter of a hemisphere was determined by averaging the longest and the shortest diameters of the cell; on average, the volume of examined HeLa cells was ∼1800 μm³. Assuming that dye concentration (C) is directly proportional to fluorescence intensity (FI) (C=k× FI), which for a number of anionic dyes at concentrations below 1 mm is supported by our studies as well as others (Valiunas et al. 2002; Ek-Vitorin & Burt, 2005), then eqn (5) can be modified as follows:

(6)

where ΔFI₂= FI_2,n+1− FI_2,n is the change in FI in cell-2 over the time, Δt= (t_n+1−t_n); n is nth time point in the recording. P_p depends mainly on the size of the tip of the patch pipette. To estimate P_p we blocked gap junctional conductance by heptanol or another gap junction channel blocker and measured the kinetics of FI₂ decay over time. Under blocking conditions, P_j= 0 and from eqn (6) it follows that:

(7)

We used this equation to estimate P_p under blocking conditions of gap junctional communication and our data show that it varies in the range of ∼1–3 × 10⁻¹¹ cm³ s⁻¹. For more details see the section describing P_j measurements and Fig. 7B. Single channel permeability (P_γ) can be found by dividing P_j by the number of functional channels (N_f=g_j/γ, where γ is the single channel conductance) at any given time:

(8)

Figure 7. Permeability of Cx57-EGFP gap junction channels to dyes.

A, representative images illustrating LY (a), AF³⁵⁰ (b), EtBr (c) and DAPI (d) but not PrI (e) permeate GJ channels. Images were recorded ∼10 min after opening the patch in cell-1 with pipette filled with dye (indicated by two asterisks). Dye acceptor cell is shown by one asterisk. B, Cx57-EGFP single channel permeability to AF³⁵⁰. Two consecutive applications of NH₄Cl (10 mm) were used to increase g_j. The grey bar shows CO₂ application used to measure P_p that characterizes AF³⁵⁰ permeability from cell-2 (dye recipient cell) to patch pipette-2. Repeated V_j ramps of ±12 mV were used to measure I_j (see the inset). The patch to cell-1 was open at the beginning of the recording. The vertical arrow shows the moment when the patch was open in cell-2 to measure I_j. FI₁ and FI₂ traces show dynamics of fluorescence intensities in cell-1 and cell-2, respectively. P_γ trace shows the single channel permeability.

To increase dye detection sensitivity, which is particularly important in cases where coupling is weak and/or channel permeability is low, time-lapse imaging was performed as follows: the entire visible field was exposed to excited light followed by focused excitation light directed only at the dye-recipient cell. The latter allowed us to avoid emission light scattering from the dye-donor cell as well as from the dye-filled pipette, which can obscure dye transfer to the recipient cell in cases where permeability is low or can give the appearance of dye transfer when it is, in fact, absent. Our estimates show that when using this approach, the sensitivity of dye transfer measurements increases over 100-fold compared with traditional methods when both cells are exposed to excitation light.

To minimize dye bleaching, we performed time-lapse imaging exposing cells to a low-intensity light for ∼0.5 s every 6 s or more. We also used low dye concentrations in a pipette solution, typically 0.1 mm and below, which minimized photo toxicity, but still provided satisfactory fluorescence intensities. We found no difference in fluorescence intensity over time when images were acquired once every 20 s or every 6 s, indicative that dye bleaching was minimal. We have used a similar methodology for evaluation of single channel permeability of connexins 30.2, 40, 43 and 45 (Rackauskas et al. 2007).

To study whether Cx57 forms functional UHCs, we examined an uptake of fluorescent dyes using time-lapse imaging over ∼5 h. Cells were plated in a 35 mm thin glass-bottomed (∼0.1 mm) Petri dish and placed on the top of ×40 or ×60 oil objective inside the CO₂ microincubator adapted to fit on the stage of an inverted microscope. Temperature was maintained at 37°C and the microincubator was filled with humidified air when cells were exposed to MKR solution and with air containing 5% CO₂ when cells were exposed to Dulbecco's modified Eagle's medium (DMEM) with 8% fetal bovine serum (FBS). Time-lapse imaging over several days showed that cells grew normally with a proliferation rate similar to that in a standard CO₂ incubator. This set-up has been used previously to examine the function of Cx30.2, Cx31.9, Cx40, Cx45 and Cx43 UHCs (Contreras et al. 2003; Bukauskas et al. 2006). Repeated series of images were taken with an exposure time of 500 ms by using different excitation and emission filter sets; interlapse interval was ∼5–10 min. Selected filters were housed in Sutter filter wheels (LAMBDA 10-2) and their positioning was controlled by the imaging software.

Data analysis and statistics

The analysis was performed using SigmaPlot software and averaged data are reported as the mean ±s.e.m.

Results

Electrical cell–cell coupling and voltage gating

Experiments were performed using spontaneously preformed HeLaCx57-EGFP cell pairs exhibiting one or several small JPs by measuring g_j with a double-voltage clamp method. We found that cell pairs exhibiting JPs were electrically coupled and demonstrated full uncoupling under exposure to heptanol (2 mm), octanol (0.5 mm) or CO₂. Thus, Cx57-EGFP GJ channels show similar properties to other members of the connexin family.

In examined cell pairs, JPs varied in size from small puncta with a diameter less than 1 μm to very large JPs up to 5 μm or more in length occupying almost the entire region of cell–cell contacts. Figure 1A and B shows two cells pairs with small (A) and large (B) JP in the contact region between two cells. At control conditions, g_j varied between ∼0.01 and 2 nS with a mean value of ∼0.39 ± 0.18 nS (n= 39). There was no coupling in cell pairs without JPs (n= 4; not shown) and, in general, g_j was higher in cell pairs with larger JPs. On average, g_j values in HeLaCx57-EGFP cell pairs were ∼500-fold lower than in HeLaCx43-EGFP cell pairs (Bukauskas et al. 2000) exhibiting JPs of similar size, which allowed us to assume that only a small fraction of Cx57 GJ channels are functional.

Figure 1. Representative images showing junctional plaques in cell pairs expressing Cx57-EGFP.

A–B, fluorescence images of cell pairs with small (A) and large (B) JPs (see arrows). C, fluorescence image of the region where cells overlap and form large JP oriented parallel to the focal plane. Areas encircled with dashed lines are regions of interest (ROI) where fluorescence intensity was measured.

Under control conditions, the averaged g_j was very low and it was difficult to obtain stable recordings over the time needed to study V_j-gating. As we show below, g_j of Cx57-EGFP cell pairs depends on pH_i. To increase quality of the records illustrating V_j-gating, we increased pH of the pipette solution from 7.2 to 7.3. This increased g_j up to several nanosiemens allowing us to study V_j-gating relatively stably. Figure 2B shows an average g_j–V_j plot from four consecutive measurements in one Cx57-EGFP cell pair by applying V_j ramps that change from 0 to −110 mV and from 0 to +110 mV (Fig. 2A). It shows a symmetric bell shaped dependence. Similar g_j–V_j plots were obtained in seven other cell pairs. To determine gating parameters, we fitted the g_j–V_j dependence with a stochastic four state model of contingent gating (Paulauskas et al. 2009); below it is called as the model of contingent gating. In the model, each AHC contributes one fast voltage-sensitive gate. Transitions between open and closed states of AHC (gates) are described by the equilibrium constant (K_i), which depend on A_h, the coefficient characterizing sensitivity to voltage across AHC (V_h), and V_h,o, which is the value of V_h at which K_i= 1. When both AHCs are open (O–O), each with conductance γ_h,o, then the GJ channel is in the fully open state with conductance γ_o determined by two γ_h,o values in series. When both AHCs are closed (C–C), each with conductances γ_h,res, then the GJ channel is in the residual state with conductance γ_res determined by two γ_h,res values in series. Intermediate states (C–O and O–C) contributed conductances determined by γ_h,o and γ_h,res in series. Therefore, two voltage gates in series control the gating of GJ channels in a contingent manner, when gating of one AHC depends on the state of the other AHC. We assumed that for Cx57 hemichannels γ_h,o= 114 pS (twice higher than conductance of GJ channel, γ_o= 57 pS; see below). Conductance of the residual state, γ_h,res, we left variable and evaluated from the fitting process. This was done due to the fact that the model accounts for only the fast gating mechanism while GJ channels also contain the slow gating mechanism, which closes channels fully (Bukauskas & Verselis, 2004). The ratio between activation of the fast and the slow gating mechanisms is unknown, and the fitting process should return some averaged value between γ_h,res and zero conductance. The grey line in Fig. 2B shows the fitting curve of the g_j–V_j plot using the following set of parameters of the model: A_R= 0.054 mV⁻¹; A_L= 0.057 mV⁻¹; V_oR=−9 mV; V_oL=−10 mV; N_f= 270, where the R and L subscripts describe the gating parameter for the right- and left-side AHC, respectively, and N_f is the number of functional GJ channels. Least-squares criteria were used to optimize the fitting. Figure 2C and D shows dependencies of different parameters on V_j: (1) g_h,R–V_j and g_h,L–V_j dependencies of left- and right-side AHCs, where g_h,R and g_h,L integrate conductances of both open and closed AHCs (grey and black lines, respectively; in Fig. 2C); (2) the probability of GJ channels to dwell in the O–O state (P_o-o, black line; in Fig. 2D); (3) the probability of GJ channels to dwell in the C–O, O–C or C–C state (P_o-c, P_c-o and P_c-c, grey lines; in Fig. 2D). If all 270 GJ channels are fully open (P_o-o= 1), then g_j should be equal 15.4 nS. However, due to low P_o-o (∼0.14), measured g_j was much lower (∼2.1 nS). The C–O, O–C and C–C states also contribute to the measured g_j but to a much lower extent due to their lower unitary conductances.

Figure 2. Voltage gating of Cx57-EGFP gap junction channels.

A–B, g_j–V_j dependence (B) measured in Cx57-EGFP cell pair by applying 60 s long V_j ramps (A). The fitting curve shown in grey (B) was obtained by using the model of contingent gating. C, g_h–V_j dependencies of left- and right-side AHCs (grey and black lines, respectively) integrating conductances of open and gated AHCs. D, probability of channels to dwell in different states (O–O, C–O, O–C and C–C) depending on V_j. Black line shows P_o-o, while grey lines show P_c-o, P_o-c, and P_c-c. E, an example of V_j-gating of Cx57-EGFP cell pair. The bottom trace shows dynamics of g_j calculated from V_j and I_j traces. g_j decayed in response to V_j step of +75 mV. Measurements of g_j during repeated V_j ramps (±15 mV) show that g_j recovery after V_j step has two distinct phases: fast and slow. Dashed lines in the inset show the fitting curves to a single exponential function of the data in periods corresponding to the fast and slow phases of g_j recovery. Time constants of the fast and the slow phases were ∼0.3 and 2.6 s, respectively.

Figure 2E shows an example of I_j recorded in response to repeated V_j ramps (±15 mV) and step (+75 mV) in a Cx57-EGFP cell pair (pH_i 7.3). In response to a V_j step, g_j declined from ∼8.4 to 2 nS over ∼2–3 s (bottom trace). Repeated V_j ramps allow the tracking of g_j recovery after each V_j step. The enlarged g_j trace in the inset distinguishes the fast and slow phases in the g_j recovery with a time constant of ∼0.3 s and 2.6 s, respectively. In concert with our earlier reports (Bukauskas & Verselis, 2004), these data allow us to assume that the fast phase reflects the fraction of GJ channels closed by the fast gate and the slow phase reflects channels closed by the slow gate. Therefore, those data support the view that both fast and slow gating mechanisms are involved in the voltage gating of Cx57-EGFP GJs.

Single gap junction channel

In a cell pair exhibiting low coupling at pH_i 7.2, I_j recordings frequently revealed unitary gating events (n= 5). Figure 3A demonstrates I_j transitions between the open state with conductance (γ_o) and the substate, also called residual state, with conductance (γ_res) during a V_j step of −95 mV. The inset in Fig. 3A shows the calculated g_j for the time window between 64 and 78 s, where several unitary gating events took place. The frequency histogram (top-right) revealed two major peaks of 11.0 ± 0.9 and 57.0 ± 1.4 pS, which presumably correspond to γ_res and γ_o, respectively. In Fig. 3A can be seen simultaneous opening of two GJ channels (see arrow). Theoretically, the open channel probability (P_o) of Cx57 can be calculated dividing the nominal open probability (NP_o) by the number of functional GJ channels (N_f) estimated from simultaneous opening events from the I_j record. Following this concept, we found that at V_j values in between of 80 and 100 mV P_o= 0.06 ± 0.02 (n= 4). However, P_o depends on applied V_js and cannot be measured at V_j= 0 mV. To overcome these problems at least in part, we used the model of contingent gating (Paulauskas et al. 2009). We simulated unitary gating events similar to those shown in Fig. 3A. Figure 3B shows the simulated I_j record using parameters close to those obtained during fitting of the g_j–V_j plot shown in Fig. 2B, which were obtained at a similar pH_i (7.2 vs. 7.3). We found that unitary events shown in B can be observed by measuring I_j in a cell pair containing 200 functional GJ channels exhibiting V_h,o=−20 mV and A_h= 0.085 mV⁻¹. Figure 3C shows the dependencies of probabilities of channels to dwell in different states on V_j. Therefore, even though N_f= 200, only two simultaneous opening events were detected (see arrows) during a 100 s period of simulation because P_o-o at V_j= 95 mV was very low (∼0.002).

Figure 3. Unitary gating events in HeLaCx57-EGFP cell pairs.

A, an example of I_j record during V_j step of −95 mV in a cell pair with low coupling at control conditions (∼0.1 nS). Most gating transitions are between open (γ_o) and residual states (γ_res). The inset shows calculated g_j for the time window between 64 and 78 s. Frequency histogram (top-right) reveals two major peaks of 11 ± 0.85 and 57 ± 1.35 pS, which correspond to γ_res and γ_o, respectively. B–C, simulation of V_j-gating in the junction containing 200 homotypic GJ channels using the model of contingent gating. I_j trace in B was obtained in response to V_j step of −95 mV (upper trace). C, probability of channels to dwell in different states (O–O, C–O, O–C and C–C) depending on V_j. The black line shows P_o-o (see the inset), while grey lines show P_c-o, P_o-c, and P_c-c. D, I_j recording from cell pair exhibiting unitary gating events in response to three consecutive V_j ramps of ±80 mV. E, I_j–V_j plot of data showed in D. Continuous lines are drawn over I_j traces corresponding to one and two open channels, and dashed lines correspond to the residual state.

The I_j record from another cell pair exhibiting single channel events in response to three consecutive V_j ramps of ±80 mV is shown in Fig. 3D. The continuous lines are drawn over the I_j values corresponding to the open states of one and two channels. The linear character of the I_j over V_j indicates that the open state shows no rectification (Fig. 3E). In this cell pair, we observed gating transitions of ∼43–49 pS to the residual state (see dashed line), as well as transitions of ∼57 pS to the close state (see arrow). We assume that the transitions of 43–49 pS in amplitude were operated by the fast gating mechanism and the transitions of ∼57 pS in amplitude were operated by the slow or the ‘loop’ gating mechanism (Bukauskas & Verselis, 2004). Therefore, gating to γ_res at the single channel level, as well as data shown in Fig. 2B allow us to conclude that tagged EGFP does not suppress function of the fast gating mechanism of Cx57, as it was reported for Cx43 (Bukauskas et al. 2001).

In summary, we found that an estimation of unitary conductances was much more challenging for Cx57 compared to many other Cx isoforms that have been studied (Bukauskas & Verselis, 2004). We assume that due to a very low P_o-o under control conditions, a substantial number of GJ channels should be functional to observe unitary gating events.

pH_i-dependent modulation of the cell–cell coupling and voltage gating

Figure 4A shows the dynamics of I_j and g_j at V_j= 0 mV measured by applying repeated V_j ramps of ±20 mV (see the inset). To change pH_i without changing pH_o, we used ammonium chloride (NH₄Cl) added to the modified Krebs–Ringer (MKR) solution (Swietach & Vaughan-Jones, 2005). Application of NH₄Cl (10 mm), which typically raised pH_i for ∼0.6 units, increased g_j from 0.8 to ∼30 nS in ∼100 s. In 31 examined cell pairs application of NH₄Cl (10 mm) increased g_j on average from 0.41 ± 0.15 to 43.6 ± 6.7 nS.

Figure 4. pH_i-dependent modulation of junctional conductance.

A, dynamics of I_j and g_j measured by applying repeated V_j ramps of ±20 mV (enlarged I_j responses to V_j ramps are shown in the inset) in response to applications of NH₄Cl (10 mm). B, g_j and pH_i measurements in one cell pair after application of 1, 3, 7 and 15 mm of NH₄Cl. C, summarized g_j–pH_i relationship in which the continuous line is a fitting curve to the Hill equation; pK_a= 7.41 ± 0.02 and the Hill coefficient of 5.9 ± 0.5 (n= 6). Filled circles are from the experiment shown in B.

To find out how g_j depends on pH_i, we loaded cells with BCECF (for more details see http://probes.invitrogen.com), a fluorescent indicator for pH_i, and applied different concentrations of NH₄Cl. We measured g_j values at different concentrations of NH₄Cl (Fig. 4B). Figure 4C summarizes the g_j–pH_i relationship. Data fitting with Hill's equation revealed that pK_a= 7.41 ± 0.02 and the Hill coefficient, n= 5.9 ± 0.5 (n= 6), where pK_a is the pH at half maximum of g_j, and n characterizes cooperativity of the H⁺ ions and their binding sites on the GJ channel.

To determine whether intracellular alkalization affects V_j-gating, we examined changes of g_j–V_j dependence in nine experiments. To reduce the time needed to measure g_j–V_j dependences after application of increasing concentrations of NH₄Cl, we examined V_j-gating by applying slow voltage ramps of one polarity, as shown in Fig. 5A. We found that in parallel with g_j increase, the V_j-gating sensitivity was substantially reduced. Figure 5B shows that application of 1 and subsequently 6 mm of NH₄Cl increased g_j from ∼0.9 to 4 and 27 nS, respectively. Grey numbers (1–7) attached to g_j–V_j plots shown in Fig. 5B–D correspond to the numbers attached to V_j ramps shown in Fig. 5A. Continuous grey lines in Fig. 5B show fitting curves of g_j–V_j plots by using the model of contingent gating (Paulauskas et al. 2009). During the fitting process, we assumed that each experimental g_j–V_j plot has a symmetric counterpart at positive V_j values and that for both hemichannels γ_h,o= 114 pS. Figure 5C shows dependencies of P_o-o on V_j for g_j–V_j plots shown in B. Under control conditions P_o-o at V_j= 0 mV was very low (∼0.02) and increased with application of NH₄Cl, approaching 1 for g_j–V_j plot no. 3 and remaining at this level for all other g_j–V_j plots. In addition, P_o-o–V_j plots became less V_j sensitive. The fitting of the data with the model allowed us to define V_h,o and the number of functional GJ channels for each g_j–V_j plot. Figure 5D shows that over time V_h,o increased from −16 to 98 mV (filled circles) and N_f increased from 220 to 490 GJ channels (open circles). Therefore, an increase of g_j during alkalization is caused by changes of both P_o-o and N_f. The same tendency under intracellular alkalization was found in all nine other experiments including the one shown in the online supplementary figure, Fig. S1.

Figure 5. Dynamics of V_j-gating during intracellular alkalization in HeLaCx57-EGFP cell pairs.

A, I_j recording in response to repeated V_j ramps from 0 to −110 mV before and during application of NH₄Cl (1 and 6 mm). Voltage steps of −20 mV were used to measure g_j in between voltage ramps. The insets show I_j responses on an enlarged scale to voltage ramps before and after application of 1 mm NH₄Cl. B, g_j–V_j dependencies calculated from V_j and I_j records shown in A. Grey lines show fitting curves obtained by using the model of contingent gating for each experimental g_j–V_j plot. Grey numbers in B–D attached to different plots correspond to numbers of V_j ramps shown in A. C, P_o-o dependence on V_j for different g_j–V_j plots. D, changes of V_h,o (open circles) and N_f (filled circles) over time during recordings shown in (A). Continuous lines show regression curves of the second order (continuous lines).

It is always a concern that the increase in g_j can lead to some reduction of V_j-gating sensitivity due to a series resistance effect (Wilders & Jongsma, 1992). In regards to this, we would like to emphasize that even at equal g_j,max values in no. 6 and no. 7 (Fig. 5B) g_j–V_j plots (∼27 nS) they show a substantial difference in sensitivity to V_j-gating. Similar differences in V_j-gating sensitivity can be seen in the supplemental Fig. S1 between g_j–V_j plot no. 8 and plots nos 9–11 exhibiting the same g_j,max. Although it is not possible to exclude fully an effect of series resistance on V_j-gating, our data allow us to conclude that intracellular alkalization indeed reduces sensitivity of V_j-gating.

Gating of Cx57-EGFP/Cx45 heterotypic gap junction channels

Gating polarity, which defines whether AHCs close at negative or positive potential on their cytoplasmic side, is one of the fundamental properties of GJs that cannot be determined in homotypic GJs. This can be done in cell pairs where Cx57 forms heterotypic channels with another Cx isoform with the already established gating polarity; e.g. Cx45 exhibits a negative gating polarity (Bukauskas et al. 2002b). We examined gating properties of heterotypic junctions in a coculture of HeLaCx57-EGFP and HeLaCx45 cells pre-labelled with DAPI. This allowed us to select heterologous cell pairs for the experiments. Figure 6A shows that under control conditions (pH_i 7.2) as well as during application of NH₄Cl, I_j response evoked by voltage ramps applied to the cell expressing Cx45 was asymmetric. We observed this I_j/V_j asymmetry in all five experiments and found that application of NH₄Cl (2 mm) increased g_j 2.0 ± 0.3-fold (n= 5). Figure 6B (black dots) shows the averaged g_j–V_j dependence that was obtained by overlapping normalized g_j–V_j plots (n= 8) measured under control conditions. The white line is a fitting curve obtained from the model of contingent gating assuming that γ_h,Cx57= 114 pS (see above), γ_h,Cx45= 57 pS (Valiunas, 2002) and gating polarity of Cx57 is negative, i.e. the same as for Cx45. Assuming that the gating polarity of Cx57 is positive, then the simulated g_j–V_j plot (see black line obtained using the same parameters as for the white line) deviates substantially from the experimental data. Thus, we demonstrate that Cx57 channels exhibit a negative gating polarity.

Figure 6. Voltage gating of Cx57-EGFP/Cx45 heterotypic gap junction channels.

A, I_j response in cells expressing Cx57-EGFP evoked by V_j ramps applied to the cell expressing Cx45 shows I_j/V_j asymmetry. Application of NH₄Cl (2 mm) increased g_j∼2-fold. B, an averaged g_j–V_j dependence (black dots) that was obtained by overlapping normalized g_j–V_j plots measured under control conditions. The white line inside experimental points is a fitting curve obtained from the model of contingent gating assuming that the gating polarity of Cx57 is negative, i.e. the same as for Cx45. The black line shows the simulated g_j–V_j plot with the same parameters of the model for white line but assuming that the gating polarity of Cx57 is positive. C, I_j recordings in cell expressing Cx57-EGFP that exhibit unitary gating events in response to positive and negative V_j steps applied to the cell expressing Cx45, and the histogram of g_j data obtained from the I_j/V_j ratio, suggesting that unitary conductance of Cx45–Cx57 heterotypic junction is ∼49 pS.

Figure 6C shows I_j recordings in a heterologous cell pair expressing Cx57-EGFP that exhibit unitary gating events in response to positive and negative V_j steps applied to the cell expressing Cx45. The histogram of g_j data shows that the distance between the peaks is ∼49 pS. Therefore, unitary conductance of the Cx45–Cx57 heterotypic junction is, as expected, in between the single channel conductances of Cx45 and Cx57 homotypic junctions, i.e. 32 (Bukauskas et al. 2002b) and 57 pS, respectively.

The major goal of these studies was to establish the gating polarity of Cx57. In doing so, we found that Cx45–Cx57 heterotypic GJs under control conditions exhibit junctional coupling of ∼2 nS i.e., substantially higher than in Cx57 homotypic junctions. In addition, alkalization increased g_j of Cx45–Cx57 heterotypic junctions assuming that this was caused mainly by Cx57 hemichannel. However, we did not pursue this direction more in depth because we do not know either the exact g_j–pH dependence of Cx45 or whether heterotypic GJs configuration modulates sensitivity to pH or V_j-gating.

Permeability studies

To assess the permeability of GJ channels, we measured dye transfer and g_j in HeLaCx57-EGFP cell pairs exhibiting at least one JP. In these studies, the pipette filled with dye is typically attached to cell-1 and the pipette without dye attached to cell-2. The patch in cell-1 was opened to fill it with dye and observe dye transfer to cell-2. Approximately 10 min later, the patch in cell-2 was open to measure g_j in dual whole-cell voltage clamp mode. This allowed us to avoid problems related to dye flux into the pipette of cell-2 during dye transfer measurements.

We demonstrate that Cx57-EGFP GJ channels are permeable to LY, AF³⁵⁰, EtBr and DAPI, but not PrI (Fig. 7A). At least five cell pairs were examined for each dye. In the experiments where no dye transfer occurred during the first 5 min of recording, cells were exposed to 5 mm of NH₄Cl, which typically increases g_j. We found difficulties detecting LY and PrI transfer under normal perfusion conditions, and only after application of NH₄Cl for ∼30 min could we conclude that LY but not PrI permeates Cx57-EGFP GJ channels.

To quantify dye permeability, we only used dyes that remain largely unbound in the cytoplasm and do not change their quantum efficiency after binding. Most of the positively charged dyes bind to nucleotides and change their quantum efficiency substantially. To assess the junctional permeability (P_j) of negatively charged AF³⁵⁰, we measured changes in fluorescence intensities in both cells, FI₁ and FI₂, over time. As indicated above and in Fig. 7A, AF³⁵⁰ permeates Cx57-EGFP GJ channels, but it was problematic to quantify this permeability due to the low g_j. To overcome this problem, we increased g_j by applying NH₄Cl. Figure 7B shows one of six experiments in which we measured P_j and single channel permeability (P_γ) by combining electrophysiological and imaging techniques. Soon after opening the patch in cell-1, fluorescence intensity started to increase in cell-1 but there were no detectable changes in cell-2. At ∼120 s, the patch was opened in cell-2 (see vertical arrow) allowing the measurement of junctional conductance by applying repeated ramps of ±12 mV; an example of I_j response to V_j ramps is shown in the top inset. Initially g_j was around 0.1 nS suggesting a very low P_o at V_j= 0 mV. To increase g_j, cells were exposed two times to NH₄Cl (10 mm; see horizontal bars). Each application resulted in an increase of I_j followed by FI₂ increase. The second application of NH₄Cl was followed by exposure to 100% CO₂ (see grey horizontal bar) that resulted in full uncoupling.

We used the time window of 2050–2260 s (see the inset on FI₂ trace), when junctional conductance was blocked by CO₂, to estimate the parameter P_p, characterizing dye permeability from cell-2 to pipette-2. For more details see Methods and eqn (7). We found that in this experiment P_p= 2.2 × 10⁻¹¹ cm³ s⁻¹. P_p depends presumably on the size of the open patch at the very tip of the patch pipette and among the six experiments it varied between 1.2 and 2.9 × 10⁻¹¹ cm³ s⁻¹. The bottom trace in Fig. 7B shows calculated P_γ by using eqn (8). To find P_γ, we used the data inside grey rectangles, i.e. during periods with noticeable g_j values. We found that on average P_γ= 0.95 ± 0.09 × 10⁻¹⁵ cm³ s⁻¹ (n= 6). All P_γ measurements were performed at repeated ramps of small amplitude and with the absence of a permanent V_j, and therefore transjunctional electrical field did not affect P_γ. As it was reported earlier, AF³⁵⁰, LY and cAMP do not permeate the residual state of Cx43 and Cx46 (Bukauskas et al. 2002a; Qu & Dahl, 2002). Thus, we assume that AF³⁵⁰ does not permeate the residual state of Cx57 GJ channels that exhibit even lower conductance than Cx43 or Cx46 and measured P_γ should indeed reflect permeability of the open state.

In summary, P_γ of Cx57-EGFP GJ channels is 24-fold higher than that of Cx30.2 and 6-, 35- and 90-fold lower than those of Cx45, Cx40, and Cx43 GJ channels, respectively (see Rackauskas et al. (2007) for P_γ values of Cx30.2, Cx45, Cx40 and Cx43 GJs). The ratios between γ_o values of Cx57 and Cx30.2, Cx45, Cx40 and Cx43 GJ channels are ∼6, 1.7, 0.4 and 0.5, respectively. Thus, permeability of GJs to metabolites is not directly proportional to their unitary conductances.

Functional efficiency of Cx57 gap junction channels

Our data show that measured g_j was much lower than expected assuming that all GJ channels assembled into junctional plaques would be functional. To address this question quantitatively, we performed a series of experiments similar to those previously used to determine the functional efficiency of Cx43 GJs (Bukauskas et al. 2000). Initially, we searched for cell pairs showing JPs in contact regions where cells overlap and the JP can be seen en face, as illustrated in Fig. 1C. Imaging at different focal planes showed that some of the selected JPs were large, oriented parallel to the focal plane. To obtain fluorescence per unit area, light emission was integrated over a given area of uniform fluorescence and background fluorescence was measured outside the JP. We used JPs of several micrometres in diameter to ensure that fluorescence in the centre of the plaque, where the region of interest (ROI) is located, was homogeneous. We evaluated fluorescence per unit area in en face JPs (F_JP) in arbitrary fluorescent units, a.u. (background subtracted). In all experiments, we used the same objective, light intensity and exposure time for fluorescence imaging. We estimated the fluorescence produced by a single GJ channel (F_γ) using F_JP and the packing density of GJ channels previously described by other groups. We assumed that each Cx57-EGFP GJ channel occupied 100 nm² (corresponding to 10 nm centre-to-centre in a square array) as an approximation of values seen in electron microscope studies of junctions in fixed tissues (Caspar et al. 1977; Peracchia, 1977) or of isolated JP imaged in aqueous media by atomic force microscopy (Lal et al. 1995). On average F_JP was equal to 1980 ± 52 a.u. per 1 μm² (n= 12). From the ratio 1980 a.u./10000 GJ channels per 1 μm², we found that F_γ= 0.2 a.u. In addition, we found that HeLaCx43-EGFP also exhibited very similar F_γ when measured under the same conditions (Bukauskas et al. 2000).

In combined imaging and electrophysiological studies, we measured g_j and the total fluorescence intensity of JPs (F_T) independent of their spatial orientation. F_T was estimated by measuring the total fluorescence in the ROI enclosing JP as shown in Fig. 1B (background subtracted). The total number of physical GJ channels (N_T) present in each JP was determined from the ratio N_T=F_T/F_γ. When P_o= 1, the number of fully open GJ channels at any given time (functional channels =N_f) for each cell pair can be determined as the ratio of total junctional conductance at V_j= 0 mV to a single GJ channel conductance, N_f=g_j/γ. The fraction of operating GJ channels (K) can be determined as the ratio between functional channels and total number of channels at any given time, K=N_f/N_T=g_jF_γ/F_Tγ.

We found that application of NH₄Cl (10 mm) that typically raised pH from 7.2 to ∼7.8 increased g_j on average from 0.41 ± 0.15 to 43.6 ± 6.7 nS (n= 13). Our earlier studies showed that EGFP fluorescence increased when pH increased from 5.5 to ∼7 and was close to constant at a pH in the range of 7 to 9 (see supplements of Bukauskas et al. (2006)). Therefore, at pH_i in the range of 7.2 to 7.8, dependence of EGFP fluorescence on pH_i should not affect evaluations of F_T and F_γ.

Our data show that P_o=∼1 under exposure of cells to 10 mm of NH₄Cl (see Fig. 5C and Fig. S1D in the Supplemental Material). Thus, we can assume that all functional GJ channels are fully open with γ= 57 pS. Consequently, we found that under these conditions N_f= 43.6/0.057 = 765 channels and K= 0.023 ± 0.006 (n= 13), which is ∼6.5-fold lower than K reported for Cx43-EGFP (Bukauskas et al. 2000). An assessment of K at pH = 7.2 is more complicated because as we showed earlier P_o-o is very low and measured g_j is a sum of conductances of channels residing in the fully open state and the substate or residual state. Therefore, it would be too speculative to define K at pH 7.2. If some g_j increase during application of 10 mm of NH₄Cl is due to ∼2-fold increase in the number of functional GJ channels (see Fig. 5D and Fig. S1F), then K at control conditions should be equal to 0.023/2 =∼0.012. Thus, we speculate that under control conditions ∼1% of Cx57-EGFP GJ channels are functional.

Cx57 unapposed hemichannels

To study whether Cx57-EGFP forms functional UHCs, we examined the uptake of fluorescent dyes using time-lapse imaging (see Methods). A co-culture of HeLa parental and HeLaCx57-EGFP cells were plated in a 35 mm thin glass-bottomed Petri dish. The dish was placed inside the CO₂ microincubator on the top of 40× or 60× oil objectives. During imaging, cells were exposed to 5 μm of DAPI and 0.5 μm of EtBr or 2 μm of PrI. Emission wavelengths of DAPI and EtBr or PrI do not overlap. We did not find any statistically significant differences in EtBr and DAPI uptakes between HeLa parental and HeLaCx57-EGFP cells under exposure to either DMEM (see Fig. 8A–C) or to MKR solution containing 2 mm Ca²⁺. An exposure of cells to a normal MKR solution containing 5 mm of NH₄Cl did not affect the rate of dye uptake. Such concentration of NH₄Cl typically increases pH_i by ∼0.4 units.

Figure 8. Dye uptake in co-cultured HeLaCx57-EGFP and HeLa parental cells.

A–C, fluorescence images of cells recorded 60 min after exposure to DAPI (5 μm) and EtBr (0.5 μm) added to DMEM. There was no noticeable difference in uptake of DAPI (B) and EtBr (C) between two cells expressing (shown in green) and not expressing Cx57-EGFP. D–E, cells expressing Cx57-EGFP (green) but not HeLa parental cells demonstrate enhanced uptake of DAPI (blue) and become round in shape ∼25 min after exposure to a Ca²⁺-free (no added Ca²⁺) MKR solution containing 5 mm of NH₄Cl. F–G, dynamics of DAPI (F) and EtBr (G) uptake at [Ca²⁺]_o= 100 μm in MKR solution; FI was measured in arbitrary units (a.u.) by subtracting background intensity.

HeLaCx57-EGFP cells exposed to a Ca-free MKR solution showed ∼5-fold increase in the rate of DAPI uptake, while the rate of EtBr and PrI uptake remained unaffected. Ca²⁺-free conditions not only increased DAPI uptake but also induced cell swelling, which became evident after ∼1 h. NH₄Cl (5 mm) added to the Ca-free solution accelerated DAPI uptake and shortened the lag time to ∼30 min for cell swelling. HeLa parental cells did not show cell swelling or an increase in the rate of DAPI uptake. Figure 8D–E shows that cells expressing Cx57-EGFP (green) but not HeLa parental cells become round in shape, swell and demonstrate enhanced uptake of DAPI (blue) when exposed to Ca²⁺-free MKR solution containing 5 mm NH₄Cl for 25 min. Swelling of cells that lasted for ∼10–20 min typically led to a rapid increase in the uptake of EtBr and PrI, presumably due to damage of the plasma membrane integrity. Figure 8F–G shows the dynamics of DAPI (F) and EtBr (G) uptake at [Ca²⁺]_o= 0.1 mm in MKR solution. Figure 8F shows that the rate of DAPI uptake is ∼3-fold higher in cells expressing Cx57-EGFP than in cells not exhibiting EGFP signal. There was no statistically significant difference in EtBr uptake among expressing and non-expressing EGFP cells (Fig. 8G). At [Ca²⁺]_o= 0.3 mm, the rate of DAPI uptake increased approximately 1.5-fold and at [Ca²⁺]_o= 0.5 mm there was almost no effect on the rate of DAPI uptake (not shown). Therefore, our data show that Cx57-EGFP UHCs are mainly closed under normal conditions and open under lessening [Ca²⁺]_o to 0.3 mm or below. Interestingly, UHCs are permeable to DAPI (279 Da; z=+2) but not to dyes with higher molecular mass, EtBr (314 Da; z=+1) and PrI (415 Da; z=+2), suggesting that the size but not the net charge is a limiting factor for permeation of UHC.

Does tagged EGFP makes a difference?

To study whether tagged EGFP affects properties of Cx57 GJ channels and UHCs, experiments were performed in HeLa cells transfected with the pMJ-Cx57IRES-eGFP vector resulting in co-expression of wild-type Cx57 (Cx57WT) and untagged EGFP that distributes virtually homogeneously inside the cells. All six examined cell pairs showed a g_j increase after application of NH₄Cl and its recovery to initial values during the washout period. Figure 9A shows that in one of these cell pairs shortly after application of NH₄Cl (5 mm) g_j increased from ∼0.2 nS and reached a steady state at ∼8 nS. The g_j was measured by applying repeated ±20 mV ramps to one cell of the cell pair. The inset shows I_j responses to voltage ramps during an initial period of NH₄Cl application. I_j responses to positive and negative ∼70 mV voltage steps show moderate (∼25%) decay. Studies of g_j–V_j dependence by using long voltage ramps also revealed a decrease in V_j gating sensitivity under exposure to NH₄Cl. Figure 9B demonstrates that in response to 60 s long V_j ramps from 0 to −100 mV, the I_j trace shows well expressed V_j-gating under control conditions at pH_{pipette−sol.}= 7.3 compared to a nearly linear change under exposure to NH₄Cl. We observed no difference between a single GJ channel conductance of Cx57-EGFP and Cx57WT co-expressed with an untagged EGFP (not shown). In addition, imaging of cocultures of HeLa parental and HeLaCx57 cells revealed no difference in EtBr and DAPI uptake under normal conditions and a significant increase in dye uptake only by cells expressing Cx57WT under exposure to Ca²⁺-free medium (not shown). In summary, we conclude that Cx57WT and Cx57-EGFP exhibit similar V_j-gating, g_j–pH_i relationship and function of UHCs depending on [Ca²⁺]_o.

Figure 9. pH-dependent modulation of g_j in HeLa cells expressing wild-type Cx57 and untagged EGFP.

A, intracellular alkalization with NH₄Cl (5 mm) increased g_j from ∼0.2 nS to ∼8 nS. MKR solution containing 2 mm of CaCl₂ was used in this experiment. g_j was measured by applying repeated V_j ramps of ±20 mV. The inset shows I_j responses to V_j ramps during an initial period of NH₄Cl application. I_j responses to positive and negative ∼70 mV voltage steps show moderate (∼25%) decay. B, I_j trace measured in response to 60 s long V_j ramps from 0 to −100 mV shows well expressed V_j-gating under control conditions at pH_{pipette−sol}= 7.3, while under exposure to NH₄Cl, the I_j–V_j relationship is almost linear indicating almost no V_j-gating.

Discussion

Different hypotheses have been proposed to explain how GJ channels and UHCs in horizontal cells are involved in the processing of light-modulated signals in the retina. Despite extensive studies performed on this matter, the role of Cx57, the major Cx expressed in mouse horizontal cells, remains disputable.

Biophysical properties and pH-dependent modulation of Cx57 gap junction channels

Combined fluorescence imaging and dual whole-cell patch clamp methods allowed us to conclude that HeLa cells expressing Cx57 exhibited relatively weak electrical cell–cell coupling under control conditions and that only a small fraction of GJ channels assembled into JPs function at any given time (∼1%). Several factors can determine this small ratio: (1) some GJ channels assembled into JPs may not be functional, e.g. GJ channels that are in the process of internalization are closed permanently, (2) newly assembled GJ channels need some time to mature, (3) open channel probability is very low, (4) interaction with other proteins that may modulate function of GJs, etc.

It is commonly accepted that the cytoplasm is slightly acidified (∼0.2 pH units) compared to extracellular milieu. We found that at pH_i 7.2 Cx57 GJ channels exhibit very low P_o. By how much does pH_i in horizontal cells differ from 7.2? There are only a few reports showing that pH_i in isolated horizontal cells under normal conditions is between 7.3 and 7.4 (Haugh-Scheidt & Ripps, 1998; Takahashi et al. 1993; Dixon et al. 1996). In addition, pH_i can vary substantially among different tissues and cell types, and numerous factors can affect pH_i. For example, synchronous neuronal spiking at high frequency can cause changes in intracellular and extracellular pH within seconds (Chesler, 2003). Changes in pH_o (which may exert an effect on pH_i) have also been suggested to play a role in the modulation of coupling between horizontal cells in retinas of rabbits, fish and mice (DeVries & Schwartz, 1989; Hampson et al. 1992; Jouhou et al. 2007b). Depolarization of fish horizontal cells reduced pH_o in their immediate surrounding by activating the release of H⁺ via the voltage-sensitive H⁺-pump (V-ATPase) supporting the hypothesis that the feedback from horizontal cells to photoreceptors could be proton mediated (Jouhou et al. 2007b). Moreover, a significant reduction of dye transfer in horizontal cells of the rabbit retina has been described when pH_o was reduced to 7.2 (Hampson et al. 1994). Our data show that the pK_a of Cx57 GJs is 7.41 ± 0.02, which is close to the pH_i measured experimentally in horizontal cells, and the Hill coefficient is 5.9 ± 0.1, which is close to a number of Cxs in the hemichannels. This suggests that the g_j between horizontal cells can be efficiently regulated by small changes in pH_i around 7.4 where the steepness of the g_j–pH_i relationship is maximal (Fig. 4C). We found that ∼100-fold increase of g_j when pH_i increased from ∼7.2 to 7.8 was caused by an increase of P_o-o and N_f (see Fig. 5C–D and supplemental Fig. S1D–E).

We demonstrate that Cx45–Cx57 heterotypic junctions exhibit an increase of coupling in response to alkalization as well. We assume that for this heterotypic junction Cx57 AHC is a limiting factor determining this g_j increase because it demonstrates higher sensitivity to pH_i than Cx45 exhibiting pK_a=∼7 (Stergiopoulos et al. 1999). If P_o of Cx57 AHC is equal to 0.1 (see P_o-c or P_c-o traces in Fig. 3C) and we assume that P_o of Cx45 AHC is equal to 1 (for reasons of simplicity), then P_o of Cx45–Cx57 GJs should be ∼0.1 (0.1 × 1), while P_o of Cx57 homotypic junction should be ∼0.01 (0.1 × 0.1). This may explains why Cx45–Cx57 heterotypic junctions exhibit higher g_j than Cx57 homotypic junctions under control conditions.

At this point it would be difficult to envision the mechanism of pH action on g_j. For Cx43 it was predicted that its C-terminus is involved in the pH-dependent closing of GJ channels and the pK_a for both Cx40 and Cx43 is equal to 6.7 (Morley et al. 1997; Gu et al. 2000). Interestingly, compared to Cx43, Cx57 has 12 more histidine residues in the C-terminus, which may explain the enhanced sensitivity to pH_i. However, the Hill coefficient for Cx43 (n= 5.5) (Stergiopoulos et al. 1999) does not differ evidently from Cx57 (n= 5.9) and the molecular events that determine pK_a and the Hill coefficient for Cx57 are yet to be resolved. Protonation of histidine residues in the second half of the cytoplasmic loop and its interaction with the C-terminus may also cause pH-sensitive gating (Delmar et al. 2004). There is one histidine residue in the second half of the Cx57 cytoplasmic loop surrounded by several lysine residues (not typical for Cx43) that may also be involved in pH-sensitive gating. However, further studies must be carried out to relate pH sensitivity of Cx57 with its structure.

Earlier studies of Cx57 with the incorrect sequence (longer C-terminus) (Manthey et al. 1999) also showed surprisingly low cell–cell coupling with g_j in the range of 0.2–2.5 nS. At that time the g_j–pH_i dependence was not examined, but equally low g_j as well as similarity in V_j-gating allows us to assume that Cx57 with a longer C-terminus has a similar g_j–pH_i relationship to Cx57 with the corrected C-terminus.

Our single GJ channel conductance measurements of the open state (57 pS) are in concert with γ_o values of 50–60 pS found in horizontal cells of teleost (McMahon et al. 1989) and bass retinas (Lu & McMahon, 1996). Single GJ channel conductance of Cx45–Cx57 heterotypic junction is, as expected, in between γ_o values of Cx45 and Cx57 homotypic junctions, i.e. 32 (Bukauskas et al. 2002b) and 57 pS, respectively. We observed no difference between γ_o values of Cx57-EGFP and Cx57WT co-expressed with an untagged EGFP (not shown). However, measured γ_o is higher than that reported earlier (Manthey et al. 1999) for Cx57 with the longer C-terminus, ∼28 pS. At this point, we expect that the difference in the sequence of the C-terminus between old (incorrect) and new (corrected) versions of Cx57 explains the difference in the unitary conductance. In addition, tagged-EGFP did not affect gating properties of Cx57 GJ channels. This conclusion is based on the presence of fast and slow phases during g_j recovery from V_j-gating and the residual conductance measured macroscopically (see Fig. 2E) and at the single channel level (Fig. 3A–D). Data obtained by examining Cx57–Cx45 heterotypic junctions (Fig. 6) allowed us to conclude that Cx57 GJ channels have a negative gating polarity, i.e. the same as Cx30.2 (Kreuzberg et al. 2005), Cx36 (F. F. Bukauskas, unpublished observations) and Cx45 (Bukauskas et al. 2002b) that are also expressed in the retina. This suggests that Cx57 UHC also exhibit a negative gating polarity and their open probability tends to increase by increasing positivity on the cytoplasmic side.

Our permeability studies show that Cx57 GJ channels are permeable to LY, AF³⁵⁰, EtBr and DAPI, but not to PrI. Thus, Cx57 GJ channels are largely non-selective for the net charge of the permeable molecule; rather, the size of the dye determines its permeability. For example, both DAPI and PrI have two positive charges, but PrI is greater in molecular mass (415 vs. 279 Da for ionic forms) and does not permeate Cx57 GJs. Our data show, that the single Cx57-EGFP GJ channel exhibits 24-fold higher permeability for AF³⁵⁰ than mCx30.2 and 6-, 35- and 90-fold lower than Cx45, Cx40, and Cx43 GJ channels, respectively (Rackauskas et al. 2007). Interestingly, despite the fact that single channel conductance of Cx57 GJs is ∼2-fold higher than that of Cx45, P_γ for AF³⁵⁰ is ∼6-fold lower. The lack of correlation between single channel conductance and permeability may imply that in Cx57, GJ channel constriction site(s) are narrower resulting in lower permeability to molecules comparable in size with numerous metabolites.

Potential implications for the Ca²⁺-dependent modulation of Cx57 unapposed hemichannels

We demonstrate that Cx57 UHCs can open after reducing [Ca²⁺]_o at moderate levels, ∼0.3 mm. This reduction of [Ca²⁺]_o can be physiologically relevant in the narrow interstitial space of the ribbon synaptic complex where photoreceptors interact with bipolar and horizontal cells. Moreover, our data show that the opening of Cx57 UHCs by low [Ca²⁺]_o could be potentiated by intracellular alkalization. Recently, no Cx57 immunoreactivity was found in dendritic tip membranes of horizontal cells. It was concluded that Cx57 UHCs are absent in the triad synapse of photoreceptors, bipolar cells and horizontal cells (Janssen-Bienhold et al. 2009). Typically, UHCs residing in the plasma membrane tend to distribute diffusely; otherwise they should be anchored to scaffolding proteins or their diffusion should be limited by barriers similar to those present in endothelial cells. If neither of these factors was demonstrated, it would be difficult to exclude the possibility that their concentration is below the limit of resolution for immunoreactivity studies.

In the dark, photoreceptors are depolarized, release glutamate continuously, and keep the horizontal cells depolarized. It is well known that depolarization can increase the open probability of UHCs (Contreras et al. 2003). In addition, the reduction of [Ca²⁺]_o in the synaptic cleft, induced by Ca²⁺ entry in the photoreceptors during dark (Rabl & Thoreson, 2002), may further enhance the open probability of UHCs. Therefore, Cx57 UHCs may open during dark and increase an exchange of ions and metabolites between the cytoplasm and the extracellular milieu. Light stimulation hyperpolarizes photoreceptors, reduces glutamate release and hyperpolarizes horizontal cells. Presumably, hyperpolarization of horizontal cells should tend to close UHCs due to their negative gating polarity shown in this study (see above).

It has been proposed that horizontal cells generate a negative feedback signal to photoreceptors upon light stimulation (Kaneko & Tachibana, 1986; Kamermans et al. 2001; Hirasawa & Kaneko, 2003; Vessey et al. 2005; Cadetti & Thoreson, 2006; Jouhou et al. 2007a; Davenport et al. 2008). This negative feedback could be abolished experimentally by gap junction blockers (Kamermans et al. 2001; Verweij et al. 2003; Pottek et al. 2003). However, the loss of Cx57 expression in Cx57 knockout mice had no influence on this negative feedback (Shelley et al. 2006), and caused changes in neither spatial tuning nor light adaptation of retinal ganglion cells (Dedek et al. 2008). Data demonstrating possible compensatory changes in the Cx57 knockout mice are lacking. Hence, involvement of Cx57 UHCs in the feedback from horizontal cells to photoreceptors in the wild-type mouse retina remains disputable.

We show that Ca²⁺-free conditions not only increased the rate of DAPI uptake but also induced cell swelling, which was followed by necrotic transformations. The latency for cell swelling and necrosis decreased from 1 h to 30 min after adding NH₄Cl to the Ca²⁺-free MKR solution. We found that it was more difficult to culture and maintain stable cell lines expressing Cx57-EGFP or Cx57WT than cell lines expressing other Cx isomers, such as Cx26, Cx32, Cx36, Cx40, Cx43, etc. During the trypsinization procedure, when the cells are exposed to very low [Ca²⁺]_o, a high fraction of cells expressing Cx57-EGFP or Cx57 swell and become necrotic. Only a relatively small fraction of cells, typically those forming numerous JPs, survived the trypsinization procedure. We assume that in cells forming more JPs, a lesser fraction of UHCs are in the plasma membrane, and therefore fewer UHCs can be open during the trypsinization procedure. We found that adding 0.5 mm Ca²⁺ to the trypsin solution improved cell survival. Thus, our data show that Cx57 UHCs are mainly closed under normal conditions and open at moderately low [Ca²⁺]_o. A future reduction of [Ca²⁺]_o up to several micromolar leads to an extensive opening of Cx57 UHCs resulting in cell death. In addition, among HeLa cell lines stably expressing different Cxs isoforms fused with GFPs, only HeLaCx57-EGFP cells show necrotic transformations under Ca²⁺-free conditions or during trypsinization. Therefore, we suggest that the increase in dye uptake under reduction of [Ca²⁺]_o can be mainly attributed to the opening of Cx57 UHCs.

Our data support the notion that Cx57 forms functional UHCs in horizontal cells of mouse retina similarly to Cxs expressed in horizontal cells of catfish, skate and zebrafish retinas (DeVries & Schwartz, 1992; Malchow et al. 1993; Shields et al. 2007; Zoidl et al. 2004). Strong pH_i-gating of Cx57 GJs in a physiologically relevant range provides evidence to support experimental data showing pH regulation of centre–surround antagonism and the modulation of the receptive field size. Finally, our data show that there is no need to reach an extremely low concentration of [Ca²⁺]_o to open Cx57 UHCs.

Glossary

Abbreviations

A_h

coefficient characterizing the steepness of g_j decay of AHC over V_j

AHC

apposed hemichannel

AF³⁵⁰

Alexa Fluor-350

F_JP

fluorescence per unit area in en face JPs

F_γ

fluorescence produced by a single GJ channel

F_T

total fluorescence of JP

γ_h,o

unitary conductance of AHC at the fully open state

γ_h,res

unitary conductance of AHC at the residual state

γ_o

unitary conductance of GJ channel at the fully open state

γ_res

single channel conductance at the residual state

JP

junctional plaque

g_j

transjunctional conductance

GJ

gap junction

AHC_R and AHC_L

right- and left-side AHCs

K

fraction of operating channels at any given time

MKR

modified Krebs–Ringer solution

N_T

total number of physical channels in a JP

N_f

number of functional channels

O–O

both AHCs are in the open state

C–C

both AHCs are in the closed state

O–C and C–O

one AHC is in the open state, while the other AHC is in the closed state

P_j

junctional permeability

P_γ

single channel permeability

ROI

region of interest

P_o-o, P_o-c

P_c-o and P_c-c, probabilities of channels to dwell in O–O, O–C, C–O and C–C states, respectively

UHC

unapposed hemichannel

V_h,o

V_j at which equilibrium constant between states of AHCs are equal to 1

V_h

voltage across AHC

V_j

transjunctional voltage

Author contribution

N.P.-P., S.S., V.A.S., K.W. and F.F.B. designed the experiments; N.P.-P. and F.F.B. performed most of the experiments; S.S. made all constructs; N.P.-P. and F.F.B. planned and performed the analysis of the data; N.P.-P., S.S., V.A.S., K.W. and F.F.B. wrote, revised and approved the final version of the paper. All experiments were carried out at the Department of Neuroscience, Albert Einstein College of Medicine.