Clustering of connexin 43–enhanced green fluorescent protein gap junction channels and functional coupling in living cells

Abstract

Communication-incompetent cell lines were transfected with connexin (Cx) 43 fused with enhanced green fluorescent protein (EGFP) to examine the relation between Cx distribution determined by fluorescence microscopy and electrical coupling measured at single-channel resolution in living cell pairs. Cx43–EGFP channel properties were like those of wild-type Cx43 except for reduced sensitivity to transjunctional voltage. Cx43–EGFP clustered into plaques at locations of cell–cell contact. Coupling was always absent in the absence of plaques and even in the presence of small plaques. Plaques exceeding several hundred channels always conferred coupling, but only a small fraction of channels were functional. These data indicate that clustering may be a requirement for opening of gap junction channels.

Direct electrical and chemical cell–cell communication mediated by gap junctions (GJs) is widespread in most cell types in organisms from coelenterates to mammals. It is well established that, in vertebrates, the multigene family of connexins (Cxs) encodes the protein subunits that comprise GJ channels (1, 2). Mutations in Cxs have now been shown to be responsible for several hereditary human diseases including the X-linked form of Charcot–Marie–Tooth demyelinating diseases (3, 4), nonsyndromic sensorineural deafness (5), erythrokeratodermia (6), and congenital cataractogenesis (7, 8).

Ultrastructurally, GJs have been identified as areas of close membrane apposition that, in freeze-cleaved replicas, can be seen to consist of tightly clustered particles, with each particle representing a GJ channel (9, 10). Although clustering of GJ channels into plaques is the primary basis for their ultrastructural identification, variations in GJ channel packing have been described and interpreted as representing pleiomorphisms (11, 12) or differences in functional state (13–15). An early study of reaggregating Novikoff hepatoma cells showed that development of electrical coupling and the appearance of clusters of particles in the membrane were correlated (16). Because electrical coupling did not develop synchronously among cell pairs and ultrastructure was correlated with the mean assessed electrical coupling, the functional capacity of any identified cluster of particles was not determined directly, nor was the minimal structure required for coupling established. Whether dispersed channels can provide electrical coupling between cells has not been resolved.

Recently, a fusion protein consisting of Cx43 with enhanced green fluorescent protein attached to its carboxyl terminus (Cx43–EGFP) was transfected into mammalian cells and was shown to be transported to the cell surface and assembled into functional GJs at locations of cell–cell contact (17). In this study, we make use of the Cx43–EGFP fusion protein to correlate the distribution of Cx43–EGFP fluorescence and electrical coupling simultaneously in living cell pairs. We used communication-incompetent Neuro-2a (N2A), HeLa, and RIN cell pairs stably transfected with Cx43–EGFP. Cell pairs allow direct measurement of junctional conductance (g_j), and the dual whole-cell patch-clamp technique provides resolution of electrical coupling at the level of a single GJ channel.

Materials and Methods

Cell Lines, Culture Conditions, and Transfection with cDNA Encoding Cx43–EGFP.

The construction of Cx43–EGFP cDNA is described by Jordan et al. (17). Transfection of N2A, HeLa, and RIN cells with Cx43–EGFP cDNA was performed by using Opti-MEM1 medium containing 10 μl of Lipofectamine (2 mg/ml) and 1 μg of plasmid DNA on cells grown to 50–75% confluence. For selection of stable transfectants, cells were trypsinized and plated at dilutions of 1:25 and 1:40 in the presence of 0.3–1.0 mg/ml G418. Individual colonies were screened for Cx43–EGFP expression by fluorescence microscopy and expanded into clonal cell lines. N2A, HeLa, and RIN cells stably transfected with Cx43–EGFP were grown in Eagle's medium, Dulbecco's medium, and RPMI medium 1640, respectively, supplemented with 10% (vol/vol) FBS. All transfection and culture reagents were obtained from Life Technologies (Grand Island, NY).

Electrophysiological and Fluorescence Measurements.

For simultaneous electrophysiological and fluorescence recording, cells were grown on thin (No. 0) coverslips (Clay Adams) and transferred to an experimental chamber mounted on the stage of an inverted microscope. The bathing medium consisted of a modified Krebs–Ringer solution containing (in mM): 140 NaCl, 4 KCl, 2 CaCl₂, 1 MgCl₂, 5 Hepes, 5 glucose, and 2 pyruvate (pH 7.4). Patch pipettes were filled with a solution containing (in mM): 10 NaCl, 140 KCl, 0.2 CaCl₂, 1 MgCl₂, 3 MgATP, 2 EGTA, and 5 Hepes (pH 7.2). Junctional conductance was measured by using dual whole-cell patch clamp (18, 19). Voltages and currents were recorded on videotape with a data recorder (VR-100, Instrutech, Mineola, NY) and were digitized subsequently with a MIO-16X A/D converter (National Instruments, Austin, TX) and our own acquisition software.

Fluorescence of Cx43–EGFP was monitored with a MERLIN imaging system (LSR, Cambridge, U.K.) equipped with an UltraPix FE250 cooled digital camera (12 bit) and a SpectraMASTER high-speed monochromator. The wavelength used for excitation of EGFP was 480 nm. For emission, we used a 520 ± 20-nm filter and a beam splitter of 495 nm (Chroma Technology, Brattleboro, VT). Fluorescence intensity of plaques was measured with a fixed illumination intensity, a ×100 oil objective, and a 3-sec exposure time.

Dye Transfer Measurements.

Tracer flux was assessed qualitatively by using Lucifer Yellow and 4′,6-diamidino-2-phenylindole (DAPI; Molecular Probes). In each experiment, dye was delivered to one cell by establishing a whole-cell recording with a pipette filled with 0.1% dye in pipette-filling solution. A gigohm seal was initially established with the recipient cell, but to prevent dye loss caused by dialysis with the patch pipette, the patch was not ruptured. Dye transfer was monitored by acquiring images every 5 sec (0.5-sec exposures) over a period of 5–15 min, after which whole-cell recording was established in the recipient cell to measure g_j.

Results

Cx43–EGFP Distribution, Electrical Cell–Cell Coupling, and Dye Transfer.

We examined the distribution of Cx43–EGFP fluorescence and measured g_j simultaneously in N2A (n = 45), HeLa (n = 35), and RIN (n = 45) cell pairs. In all three cell types, Cx43–EGFP fluorescence was distributed diffusely in the cell membrane except in areas of cell–cell contact where there was brighter punctate fluorescence (Fig. 1), resembling junctional plaques viewed by immunofluorescence with Cx-specific antibodies (17). The junctional plaques varied considerably in size and number among cell pairs. Cell pairs with diffuse membrane staining (n = 32) were devoid of electrical coupling. An example of a fluorescent image of an N2A cell pair without plaques is shown in Fig. 1A together with a dual whole-cell patch recording illustrating the complete absence of electrical coupling. Most cell pairs containing plaques at areas of cell–cell contact (n = 77 from N2A, HeLa, and RIN combined) were electrically coupled (Fig. 1B). For plaques larger than ≈0.2 μm in diameter, the value of g_j increased with the number and size of plaques. Cells with small plaques were not always coupled, suggesting that a minimal plaque size was required for GJ channels to open (see below). A similar relation between the size of plaques and electrical coupling was observed in all three cell types. Several instances of endogenous coupling were observed in HeLa and N2A cells and were easily distinguished from Cx43–EGFP coupling by voltage dependence and by single-channel conductance (20, 21).

Figure 1.

Cx43–EGFP fluorescence forms junctional plaques correlated with electrical coupling and dye transfer. (A) N2A cell pair showing only diffuse Cx43–EGFP fluorescence staining in appositional and other membrane areas (Left) was not electrically coupled (Right). Voltage steps applied to one cell (transjunctional voltage, V_j, Lower trace) produced no changes in junctional current (I_j) recorded in the unstepped cell. (B) A cell pair showing two junctional plaques and diffuse membrane staining elsewhere (Left, fluorescent and phase-contrast images) was electrically coupled as indicated by the large changes in I_j in response to an applied V_j. Coupling conductance was ≈12 nS. (C and D) Electrically coupled HeLa cells containing Cx43–EGFP plaques show Lucifer Yellow and DAPI cell–cell transfer. (Left) Phase. (Center) Fluorescence before dye injection. (Right) After dye injection. (C) A cluster of three cells. Lucifer Yellow was injected into cell 2 and transferred to cell 1 but not cell 3. A large (arrow) and a small plaque could be seen between cells 1 and 2; g_j = 15 nS. No plaques were evident between cells 2 and 3. The fluorescent image (taken 5 min after dye loading of cell 2) showed spread to cell 1 but not cell 2. (D) A cluster of three cells. DAPI was injected into cell 2 and transferred to cell 1 but not cell 3. Four junctional plaques could be seen between cell 1 and cell 3; g_j = 40 nS; however, no plaques were seen with cell 2. The fluorescent image taken 14 min after dye loading of cell 3 showed labeling of the nuclei of cells 1 and 3 but not cell 2.

To examine permeability qualitatively, we tested transfer of Lucifer Yellow and DAPI, negatively and positively charged dyes, respectively. Fig. 1 C and D shows images of different clusters of three HeLa cells in which plaques were present between two of the three cells (arrows). In both cases, Lucifer Yellow and DAPI transferred to the cell that participated in plaque formation with the injected cell, whereas neither dye transferred to the cell in which plaques were absent; g_j between the dye-coupled cells exceeded 35 nS in both of these cases. DAPI had strong nuclear localization, and to see dye transfer, the site of DAPI loading had to be close to the junctional membrane. Similar results were obtained in all three cell types with g_js ranging from 11 to 60 nS. All dye-coupled cell pairs showed full electrical uncoupling with heptanol (not shown), indicating that GJs and not cytoplasmic bridges mediated coupling. Dye coupling was the same as that observed with wild-type (wt)Cx43 (data not shown), indicating that attachment of EGFP to the carboxyl terminus of Cx43 does not alter charge selectivity as qualitatively assessed by dye transfer.

Biophysical Properties of Cx43–EGFP Channels.

GJs can be closed when a transjunctional voltage, V_j, is applied between cells (22, 23). For wtCx43, steady-state g_j is maximal when V_j is near zero and decreases for either polarity of V_j to an apparent plateau or residual g_j. The residual g_j is explained by incomplete closure of GJ channels by the V_j gating mechanism as described (24–26). Larger V_js close wtCx43 channels completely and with a slower time course (data not shown), but the different voltage sensitivity and slow kinetics of this gating suggest that it is a different mechanism. For Cx43–EGFP, V_j gating to the residual g_j was absent, leaving only gating characterized by full channel closure and slow kinetics. The g_j–V_j relation of Cx43–EGFP GJs is shown in Fig. 2A. The data are shown as filled circles taken from 19 different Cx43–EGFP expressing cell pairs from all three cell types. Cx43–EGFP junctions in all cell types had similar V_j dependence (data are superimposed). V_js up to ≈±50 mV showed no appreciable change in g_j; beyond that, g_j decreased gradually to near zero at ±120 mV. The solid line is a fit of the data to the Boltzmann equation (see Fig. 2 legend). For comparison, the dashed line shows a fit of data for wtCx43 (data points not shown) to the Boltzmann equation of the same form. Fig. 2B shows an example of the slow time course of decay and recovery of junctional current in a Cx43–EGFP expressing cell pair subjected to a long-duration V_j; full recovery (not shown) took ≈5 min.

Figure 2.

Voltage dependence of Cx43–EGFP GJs. (A) Junctional conductance depended on V_j. The solid line is a fit of the data to a Boltzmann equation of the form G_j = (1 − G_min)/{1 + exp[A*(V_j − V₀)]} + G_min, where G_j is g_j normalized to the maximum value, g_max, at V_j = 0; G_min is the normalized residual g_j; A is a measure of voltage sensitivity in mV⁻¹; and V₀ is the V_j at which G_j is halfway between the maximum and minimum values. The Boltzmann parameters were V₀ = 81 ± 4 mV and A = 0.08 ± 0.02 mV⁻¹, for positive V_j, and V₀ = −82 ± 4 mV and A = 0.09 ± 0.02 mV⁻¹, for negative V_j. G_min was close to zero for both polarities of V_j. (B) An example of the slow time course with which I_j nears zero in response to a long-duration (2.5-min) V_j of −90 mV (HeLa cell pair). Repeated −20-mV V_j pulses, applied at 1-sec intervals before and after the long-duration V_j step, show a slow g_j recovery. (C) V_j-induced gating of Cx43–EGFP GJs at the single-channel level in a HeLa cell pair that was coupled by nine functional channels. A V_j of −100 mV was applied for a duration of 60 sec. Seven transitions ≈110 pS in size were evident (see Inset histogram of I_j amplitude). There was no residual g_j at the end of the V_j step. (D) Linearity of the I_j–V_j relationship at the single-channel level. V_j ramps of ±100 mV and 2-sec duration were applied at regular intervals preceded by brief (200-msec) ±40-mV V_j steps. The arrow indicates spontaneous opening of a single Cx43–EGFP channel between a HeLa cell pair.

Gating of individual Cx43–EGFP channels could be seen in poorly coupled cell pairs (Fig. 2C). Consistent with the macroscopic properties of Cx43–EGFP, application of V_j caused a decline in junctional current to zero, and only one size of current transition was evident (Fig. 2D). An all-points amplitude histogram taken from the indicated segment of the junctional current record shows multiple, evenly spaced peaks that give a transition amplitude of ≈110 pS. This conductance corresponds to the fully open channel conductance of wtCx43 (27). To examine the I–V relation of a single Cx43–EGFP channel, voltage ramps from −100 mV to +100 mV were applied to a cell pair expressing one functional channel (Fig. 2D). Open channel current was nearly linear, with voltage yielding a slope conductance of ≈110 pS, the same as for wtCx43 (data not shown). These data indicate that attachment of EGFP to Cx43 does not interfere with ion flux through the fully open channel over a wide range of V_j.

GJ Plaques and Functional Channels.

Estimation of fluorescence of a single Cx43–EGFP channel.

To relate g_j to the number of channels contained in a GJ plaque, we first determined fluorescence per unit area in larger plaques viewed en face where fluorescence was uniform in the center. Then, from channel packing density previously determined for GJs (28, 29), we could estimate fluorescence per channel. For these measurements, cell pairs containing relatively large junctional plaques were selected (Fig. 3). In the side-by-side view of an N2A cell pair (Fig. 3A), a large plaque was visible at the junction between the cells. This cell pair was then detached from the coverslip, by using a gentle stream of bath solution directed from a flow pipette, and manipulated into a vertical position to allow visualization of the plaque en face (Fig. 3B). Two-dimensional (Fig. 3C Inset) and three-dimensional (Fig. 3C) plots indicate that fluorescence intensity was essentially uniform in the center of the plaque. The largest of three plaques between a HeLa cell pair viewed en face also showed uniform intensity over the central region (Fig. 3 D and E). Fluorescence intensity in the center of large plaques was the same in all three cell types. The fall-off at the edges of large plaques and lower peak values in small plaques are likely to reflect fluorescence diffraction rather than looser particle packing.

Figure 3.

Fluorescence intensity reaches a plateau in the center of en face images of larger Cx43–EGFP plaques. Schematic drawings (Left) show orientations of the cell pairs and light paths for excitation (upward arrows) and emission (downward arrows). (A and B) Side-by-side and en face views of a Cx43–EGFP plaque in an N2A cell pair. (C) Two- and three-dimensional pseudocolor plots of fluorescence intensity in the square shown in B show a plateau in the central region of the plaque. (D) En face views of large and small plaques in a HeLa cell pair. (E) Plot of fluorescence intensity along the line shown in D crossing two of the three plaques. Fluorescence intensity is normalized to the average plateau value in the center of the largest plaque. Peak fluorescence was smaller in small plaques, but fall-off at edges was about the same. Pseudocolor in C and D was adjusted to maximize the range of fluorescence intensity.

To obtain plaque fluorescence per unit area, light emission was integrated over a given area of uniform fluorescence in the center of a plaque. Background fluorescence was measured from the same area outside the cell. We used plaques >2 μm in diameter to assure that fluorescence in the center of the plaque reached a plateau corresponding to full intensity. We evaluated integrated plaque fluorescence (background subtracted), L_p, in arbitrary fluorescent units. In all experiments, we used the same objective, the same intensity of illumination, and the same exposure time for fluorescence imaging. We estimated the fluorescence produced by single GJ channel, L_ch, from L_p and density of channel packing. We assumed that each Cx43–EGFP channel occupied 100 nm² (corresponding to 10 nm center-to-center in a square array) as an approximation of values seen in electron microscope studies of junctions in fixed tissues (28, 29) or of isolated Cx43 junctions imaged in aqueous media by atomic force microscopy (30). This channel density gave L_ch = 9.6 × 10⁻³ ± 0.7 × 10⁻³ arbitrary fluorescent units (n = 7).

Correlation between number of channels in plaques and junctional conductance.

To correlate the number of channels contained in a plaque with g_j, we selected N2A, HeLa, and RIN cell pairs containing only a single small plaque visible within the area of apposition (Fig. 4A). To evaluate the number of channels in a plaque, we measured the total fluorescence of a plaque, L_tot, defined as fluorescence measured from an ROI containing the entire plaque minus background fluorescence and divided by L_ch. As L_tot is critical in estimating the number of channels in a plaque, we examined the dependence of L_tot on focus and on size of the ROI. We measured L_tot of plaques ≈1–2 μm in diameter at different z-axis values above sharpest focus (z = 0) and with ROIs of various diameters (Fig. 4B). When the diameter of the ROI was approximately five or six times larger than the plaque diameter (plots with filled and open circles), L_tot was nearly constant with z values up to 2 μm. Decreasing the diameter of the ROI reduced the measured L_tot progressively at larger values of z. With an ROI comparable to that of the plaque (open diamonds), L_tot was reduced even at z = 0, and small changes in focus significantly affected L_tot. Very large ROIs (not shown) decreased signal to noise. The ROI for each plaque was obtained by adjusting d until L_tot remained unchanged with refocusing over a range of ±2 μm. L_tot did not differ whether measured in lateral or en face views of plaques up to 2 μm in diameter.

Figure 4.

Relation between the size of Cx43–EGFP plaques and g_j. Data are summarized from N2A, HeLa, and RIN cell pairs with single plaques. (A) Side-by-side view of an N2A cell pair containing a single junctional plaque ≈0.3 μm in diameter. g_j in this cell pair was 230 pS, which corresponds to two functional channels. (B) Examination of the effect of focus (z-axis position) and size of the region of interest (ROI) used to measure fluorescence on the total measured fluorescence (L_tot). Diagrams of a cell pair containing a single plaque (thick vertical bar) illustrate the experimental procedures used. The Upper drawing is a side view, in which the solid horizontal line represents the coverslip and the dashed horizontal lines represent different z-axis positions (z = 0 is the position giving the sharpest image of the plaque). The Lower drawing is the view as seen through the microscope. The concentric circles represent ROIs of different diameters, d. L_tot as a function of z is plotted for ROIs of varying diameters (indicated to the right of each plot). The cell pair from which the data were obtained had a single plaque ≈1 μm in diameter. The objective was moved upward in 0.5-μm steps starting at z = 0. (C) Relation between the number of channels in a plaque and g_j. Data were obtained from the three cell types: HeLa (circles), N2A (triangles), and RIN (squares). (Inset) An expanded view of data with g_j between 0 and 0.8 nS is shown. g_j was measured at small V_js (<50 mV). (D) Plot of the calculated fraction of functional channels in a plaque vs. g_j from the data presented in C. The solid line is a linear fit of the data for g_js exceeding 4 nS. The dashed lines represent 99% confidence intervals.

We estimated the number of channels within a plaque by fluorescence and measured g_j by dual whole-cell patch recording in 55 cell pairs containing single plaques. The number of channels contained in a plaque is plotted vs. g_j in Fig. 4C; each open symbol represents a different cell pair. A single point at the origin (filled circle) represents the 32 additional cell pairs in which plaques and electrical coupling were absent. In 11 cell pairs with small plaques with estimated numbers of channels ranging from 90 to 330, there was no electrical coupling. In five other cell pairs with plaques containing ≈200–400 channels, there was weak coupling (0.05–0.7 nS averaged over time; Fig. 1C Inset). For the four lowest conductances, the junctional current showed single-channel transitions resulting from gating, but the peak conductance indicated that coupling in these cell pairs was mediated by one or two channels. Cell pairs with plaques containing ≥500 channels were coupled by a conductance ≥4 nS, i.e., ≈≥35 open channels, and g_j increased as the plaque size increased. These data suggest that there is a critical plaque size of 200–400 channels at which Cx43–EGFP GJ channels become capable of opening.

The fraction of open channels in a plaque, F, was obtained from g_j/γn, where γ is single-channel conductance (110 pS) and n is the total number of channels. Fig. 4D shows a plot of the fraction of open channels vs. g_j calculated from the data in Fig. 4C. The fraction of open channels increased abruptly to ≈0.1 just above the critical size for function (200–400 channels) and then changed only modestly for further increases in size (m = 0.002/nS between 4 and 20 nS, where m is the slope). Larger junctions were not studied, because access resistance could cause error in the measurement of g_j (31). Preliminary observations of small plaques with a maximum of one or two channels open at one time suggest that open probability, P_o, is ≈0.5 for the channels that are capable of opening. Noise analysis of junctional currents may allow estimation of P_o in larger plaques and reveal whether P_o changes with plaque size.

Discussion

Although it has long been recognized that GJ channels exist in plaques that represent aggregates of channels (and presence in aggregates has been an important criterion for identification of GJ channels), the relationship between clustering and channel opening has not been established. Dispersed channels presumably do exist, but their relative number and their contribution to cell–cell coupling has not been evaluated. The possibility that dispersed channels are functional is suggested by structural studies that have shown particles in loose aggregates or rows (12, 32). Coupling by dispersed channels is also suggested by the presence of electrical coupling between cells lacking morphologically identifiable plaques (33). Recordings of single GJ channels with the double whole-cell patch-clamp technique without treatment with blocking agents might suggest that functional GJ channels exist in isolation, an inference that the present study shows is unreliable and certainly not valid for Cx43–EGFP junctions.

In this study, Cx43–EGFP was used to examine the relationship between clustering of GJ channels and electrical coupling. Selection of cell pairs with nonjunctional membrane fluorescence, but no identifiable GJ plaques or with only a single plaque within the region of cell–cell contact, allowed us to show that formation of clusters containing several hundred Cx43–EGFP channels is a prerequisite for channel opening. We observed that coupling was absent even when there were small plaques but was always present when there were larger plaques. Cells with plaques containing 200–400 channels were either not coupled or coupled by one or a few channels. The number of channels was estimated from area per channel of wt junctions and the fluorescence per unit area of large plaques where diffraction at the edges could be ignored. Although no freeze-fracture studies have been done on GJs composed of Cx43–EGFP, Jordan et al. (17) showed that Cx43–EGFP had wtCx43 characteristics in terms of trafficking to the cell surface and assembling into ultrastructurally normal-appearing GJ plaques. Thus, the assumption of packing density comparable to that of wt junctions is not unreasonable.

In addition to trafficking and ultrastructural appearance in thin section, properties unaffected by attachment of EGFP to the carboxyl terminus of Cx43 include single-channel conductance, permselectivity qualitatively evaluated by dye transfer, the “slow” form of gating induced by large V_js, and sensitivity to block by heptanol. The latter two processes have been ascribed to the action of a gating mechanism that characteristically closes GJ channels and hemichannels to a fully closed state via slow gating transitions (25, 34, 35). Compared with wtCx43 channels, Cx43–EGFP channels showed loss of V_j gating to the residual subconductance state, which is a fast transition, consistent with there being two gating mechanisms. We have no explanation for the loss of this form of V_j gating. Although V_j gating is affected, preliminary studies of the rate of appearance of functional channels after placing cells into contact show no obvious differences in cells expressing wtCx43 or Cx43–EGFP (data not shown). This finding and the very similar ultrastructural appearance of GJ plaques composed of Cx43–EGFP suggest that it is unlikely that Cx43–EGFP channels require clustering and that wtCx43 channels do not. However, we cannot completely rule out the possibility that the EGFP tag reduces the packing density and/or overall efficiency of Cx43–EGFP assembly into GJs.

Even in plaques that were large enough to mediate coupling, only a small fraction of Cx43–EGFP channels were open at the same time. The fraction of channels that were capable of opening would be larger by a factor 1/P_o. Preliminary studies with cell pairs containing one or two functional channels showed P_o ranging from 0.5 to 1.0. Thus, if P_o were the same in larger plaques, the fraction of active channels would be about 0.1–0.2. Smaller values of P_o would imply a larger fraction of active channels. In small plaques close to the critical size for function, we estimated that fewer than 2% of the channels were active, and this fraction increased to 10–20% in plaques with diameters of ≈0.5 μm (≈2,000 channels). This fraction may increase further as plaque size increases. Counts of GJ particles at club endings on the Mauthner cell and measurements of electrical excitatory postsynaptic potential amplitudes indicate either that most GJ channels are closed at these junctions or that their single-channel conductance is very low (36).

The active channels may reside in central regions of plaques because of interactions within the cluster, leaving inactive channels at the periphery, which is likely the region of growth. As a plaque enlarges, the number of channels grows with the square of the circumference, which could give rise to a greater proportion of functional channels. Although the small fraction of active channels suggests that channels within a plaque differ, all channels may be the same and cycle between active and inactive states on a time scale slower than we were able to detect. Phosphorylation and dephosphorylation are processes that might mediate such cycling. We did not evaluate g_j in plaques larger than ≈0.5 μm because of possible errors caused by series resistance (31).

Although we can conclude that there are no dispersed functional Cx43–EGFP channels, we cannot distinguish whether they do not form, i.e., the hemichannels do not align and dock or whether they form but do not function. The demonstration that some hemichannels can be functional when unapposed (34, 37–39) and the presence of diffuse membrane fluorescence with Cx43–EGFP suggest that Cx hemichannels are assembled and present in the plasma membrane before docking and channel opening. Because we did not observe aggregation of Cx43–EGFP outside of areas of cell–cell contact, undocked Cx43–EGFP hemichannels do not seem to have an intrinsic tendency to aggregate. Thus, it may be that other proteins, such as adhesion molecules, localize to areas of cell–cell contact and promote Cx aggregation or clustering. GJ formation may be initiated by recruitment of hemichannels to areas of cell–cell contact, thereby increasing the likelihood of hemichannel docking across the gap between neighboring cells. A central role for cell adhesion in GJ formation has been proposed (40–42). Alternatively, aggregation may be a consequence of hemichannel docking. In this case, dispersed hemichannels would dock by chance and serve as nucleation sites for plaque formation. Hemichannel docking could promote aggregation by conformational changes in Cx subunits that expose adhesive domains or by physicochemical forces that make it energetically favorable for docked hemichannels to aggregate (43).

GJs are unique among ion channels in having principal structural components contributed from neighboring cells. The leading hypothesis proposed for formation involves the docking of assembled hemichannels. Herein, we propose that GJ channel aggregation or clustering may also be necessary for channel opening.

Abbreviations

Cx

connexin

EGFP

enhanced green fluorescent protein

GJ

gap junction

N2A

Neuro-2a

DAPI

4′,6-diamidino-2-phenylindole

wt

wild type

ROI

region of interest