Gap junction permeability: selectivity for anionic and cationic probes

Abstract

Gap junction channels formed by different connexins exhibit specific permeability to a variety of larger solutes including second messengers, polypeptides, and small interfering RNAs. Here, we report the permeability of homotypic connexin26 (Cx26), Cx40, Cx43, and Cx45 gap junction channels stably expressed in HeLa cells to solutes with different size and net charge. Channel permeability was determined using simultaneous measurements of junctional conductance and the cell-cell flux of a fluorescent probe. All four connexins allowed passage of both cationic and anionic probes, but the transfer rates were connexin dependent. The negatively charged probes [Lucifer yellow (LY; median axial diameter 9.9 Å, charge −2), carboxyfluorescein (CF; 8.2 Å; −2), and Alexa Fluor350 (AF350, 5.4 Å; −1)] exhibited the following permeability order: Cx43 > Cx45 > Cx26 > Cx40. In contrast, for the positively charged species permeability, the orders were as follows: Cx26 ≈ Cx43 ≈ Cx40 ≈ Cx45 for N,N,N-trimethyl-2-[methyl-(7-nitro-2,1,3-benzoxadiol-4-yl) amino] ethanaminium (NBD-m-TMA; 5.5 Å, +1) and Cx26 ≥ Cx43 ≈ Cx40 > Cx45 for ethidium bromide (10.3 Å, +1). Comparison of probe permeability relative to K⁺ revealed that Cx43 and Cx45 exhibited similar permeability for NBD-m-TMA and AF350, indicating weak charge selectivity. However, lesser transfer of CF and LY through Cx45 relative to Cx43 channels suggests stronger size-dependent discrimination of solute. The permeability of NBD-m-TMA for Cx40 and Cx26 channels was approximately three times higher than to anionic AF350 despite the fact that both have similar minor diameters, suggesting charge selectivity. In conclusion, these results confirm that channels formed from individual connexins can discriminate for solutes based on size and charge, suggesting that channel selectivity may be a key factor in cell signaling.

connexins form intercellular channels and have been shown to be essential for electrical and chemical cell-cell communication. In humans at least 21 connexin family members have been identified (32). Each gap junctional channel is formed by two opposing hemichannels within the plasma membranes of two adjacent cells. Numerous studies (8, 9, 14, 16–18, 22, 35) have demonstrated that gap junction channels composed of different connexin isoforms have distinct properties, varying in structure, pore size, voltage-dependent gating, open probability, and permeability.

The ability of connexins to pass selectively different molecules is of a particular interest, as these channels properties are essential and probably rate limiting in the diffusion of biologically active molecules, various metabolites, or small interfering RNA within a syncytium (7, 22, 35, 37). Previous studies (5, 20, 27, 29) have documented that gap junctions are permeable to second messengers involved in cell signaling, such as phosphatidylinositol 1,4,5-trisphosphate (IP₃) or cAMP (5, 20, 27, 29). In addition, differences in permeability of gap junction channels formed by different connexins were also observed. Recently, we have demonstrated that connnexin43 (Cx43) exhibits ∼10 and ∼7 times higher permeability to cAMP than Cx40 and Cx26, respectively (22).

Several studies have determined the single channel permeability of different connexins for large fluorescent probes. For example, the single channel permeability of Lucifer yellow (LY) to Cx43, Cx45, Cx40, and Cx26 has been reported (16, 22, 25, 31, 35). Eckert (11) demonstrated the differences in single channel permeability for LY and calcein in cells expressing Cx43 and Cx46. Weber et al. (45) reported connexin-dependent selectivity for Alexa probes of different sizes for six connexin isoforms expressed in oocytes. Rackauskas et al. (31) compared the diffusion of LY and AF350 through homotypic and heterotypic gap junctions composed of Cx40, Cx45, Cx43, and Cx30.2. These data suggest that, besides the size-dependent selectivity, the net charge of molecule is also very important determinant of solute permeability (2, 10, 26, 33, 41, 42), although more recently Heyman and Burt (19) reported that Cx43 has no charge-based selectivity.

In this study, the double patch-clamp method combined with perforated patch-clamp technique was used to measure gap junction conductance, while simultaneously monitoring intercellular transfer of fluorescent probes, and a single channel permeability of anionic and cationic dyes to Cx40, Cx43, Cx45, and Cx26 was determined. We demonstrated that Cx43 and Cx45 gap junction channels exhibit weak or no selectivity towards charge of solutes of similar size, while Cx26 and Cx40 clearly show preference toward probes of positive net charge. The important signaling molecules such as cAMP, IP₃, and various metabolites like ATP, ADP, and AMP have comparable molecular weight and net charge to the probes used here and the data obtained in this study may reveal connexin selectivity patterns for such metabolites and second messengers as well.

METHODS

Cells and culture conditions.

Experiments were performed on HeLa cell pairs stably expressing wild-type mCx43, mCx45, rCx40, and hCx26. HeLa cells were grown in DMEM (GIBCO-BRL), supplemented with 10% FCS (Hyclone), 100 μg/ml streptomycin (GIBCO-BRL), and 100 U/ml penicillin (GIBCO-BRL). The media also contained 100 μg/ml hygromycin (Sigma) or 1 μg/ml puromycin (Sigma). The cells were passaged weekly, diluted 1:10, and kept at 37°C in a CO₂ incubator (5% CO₂-95% ambient air). Culture conditions for these cells in more detail have been described previously (36, 38). Electrophysiological measurements and dye flux studies were carried out on cells cultured on the glass coverslips for 24–60 h.

Electrophysiological measurements.

The glass coverslips with attached cells were placed to the perfusion chamber mounted on the stage of an inverted microscope (Olympus IMT-2) with fluorescence imaging system. The external solution was composed of the following (in mM): 137.7 NaCl, 5.4 KCl, 1 MgCl₂, 2 CaCl₂, 5 HEPES, 10 glucose 2 mM of CsCl, CdCl₂, and BaCl₂ were added with pH adjusted to 7.4 with NaOH. Patch pipettes (1–2.5 MΩ) were pulled from glass capillaries (code 60300; A-M Systems) with a horizontal puller (Sutter Instruments) and filled with internal solution, which contained the following (in mM): 120 K⁺ aspartate⁻, 10 NaCl, 3 MgATP, and 5 HEPES (10 EGTA at ∼8 pCa) with pH adjusted to 7.2 with KOH and then filtered through 0.22-μm pore filter.

Gap junction conductance was measured in HeLa cell pairs using a double voltage patch-clamp method. Each cell of a pair was voltage clamped at the same potential by two separate patch-clamp amplifiers (Axopatch 200B; Molecular Devices). To record junctional conductance avoiding channel gating, the brief voltage steps (±10 mV, 10 ms) were applied to one cell of a pair, whereas the other cell was held at constant voltage and the junctional currents were recorded from the unstepped cell. High junctional conductance values were corrected for series resistance as defined by van Rijen et al. (39). All experiments were performed at room temperature (∼22°C). To prevent dye washout, the acceptor cell was held in perforated patch-clamp mode (35). For perforated patch experiments, 30–50 μM of β-escin were added to the pipette-filling solution (15).

Dye flux analysis.

LY (Molecular Probes), 5-(and-6)-carboxyfluorescein (CF; Molecular Probes), ethidium bromide (EthBr; Molecular Probes), Alexa Fluor350 (AF350; Invitrogen, Molecular Probes), and N,N,N-trimethyl-2-[methyl-(7-nitro-2,1,3-benzoxadiol-4-yl) amino] ethanaminium (4) (NBD-m-TMA; obtained from Dr. S.Wright, Dept. Physiology, University of Arizona) were dissolved in the pipette solution at concentration of 1 mg/ml. Transfer of dyes through gap junction channels was investigated introducing fluorescent probes via patch pipette to the one cell of a pair and monitoring change of fluorescence in both cells over the time, simultaneously measuring junctional conductance. Modeling using Accelrys DS Visualizer (Accelrys Software) showed that probes have nonspherical shape. Thus the median axial diameter was considered to be the most diffusion limiting and was used for comparison of the probes. The net charge, molecular weight (without counter-ion), and size of the fluorescent probes are shown in Table 1.

Table 1.

Characteristics of the fluorescent probes

	Charge	Molecular Weight, g/mol	Dimensions, Å*
LY	−2	443	12.2 × 9.9 × 2.6
CF	−2	376	10.2 × 8.2 × 7.3
AF350	−1	326	14.0 × 5.4 × 2.6
NBD-m-TMA	+1	280	12.2 × 5.5 × 4.3
EthBr	+1	314	10.8 × 10.3 × 4.3

Net charge, molecular weight (without counter-ion), and dimensions of the fluorescent probes.

LY, Lucifer yellow; CF, carboxyfluorescein; AF350, Alexa Fluor350; NBD-m-TMA, N,N,N-trimethyl-2-[methyl-(7-nitro-2,1,3-benzoxadiol-4-yl) amino] ethanaminium; EthBr, ethidium bromide.

^*Median axial diameter of the probe was assumed to be the most diffusion limiting dimension.

Fluorescent dye cell-to-cell spread was monitored using the digital CCD camera PixelFly (12-bit; The Cooke) and HRm Axiocam (14-bit; Carl Zeiss). The averaged fluorescence intensities in recipient and donor cells were measured over the time and corrected by subtracting background fluorescence intensity. Background fluorescence was measured in an area on the slide without cells. The relative fluorescence intensity was calculated as a ratio of fluorescence intensity in recipient cell over fluorescence intensity in injected cell after 12 min of application of the dye.

Single channel permeability (P_γ) was calculated as described previously (21). Briefly:

$${\text{P}}_{\gamma }={\text{P}}_{junction}/n,$$ 1

where P_junction is total junctional permeability and n is the number of the active channels, calculated as:

$$n={\text{g}}_{\text{j}}/{\text{γ}}_{\text{j}},$$ 2

where g_j is total junctional conductance and γ_j is single channel conductance.

The unitary conductances of the channels using our pipette filling solution containing 120 mM K⁺ aspartate^- were measured as previously reported: 55 pS for homotypic Cx43, 25pS for Cx45, 125pS for Cx40, and 110pS for Cx26 (22, 25, 35). These values are also consistent with other studies if corrected for the salt concentration used here (33, 41). Aspartate is used as the major intracellular anion to achieve intracellular chloride concentrations of 10 mM.

The single channel permeability to the probe then can be calculated as:

$${\text{P}}_{\gamma }=\left[{\text{vol}}_2\times \Delta {\text{C}}_2/\left(\Delta \text{t}\left\{\left[\left({\text{C}}_{1,n}-{\text{C}}_{2,n}\right)+\left({\text{C}}_{1,n+1}-{\text{C}}_{2,n+1}\right)\right]/2\right\}\right)\right]/n,$$ 3

where C_1,n and C_2,n are the dye concentrations at the beginning of the recording in donor cell (cell 1) and recipient cell (cell 2), respectively; C_1,n+1 and C_2,n+1 are the dye concentration in the cell 1 and cell 2 at t minutes later; Δt is the time between two successive measurements; vol₂ is the volume of recipient cell; ΔC₂ is the concentration change in the recipient cell during the period of time Δt. [(C_1,n − C_2,n) + (C_1,n=1 − C_2,n+1)]/2 represents the averaged gradient during the period of time Δt. The volume of examined HeLa cells was assumed to be ∼1.8 pL, which is consistent with estimations of others (20, 31).

Over the range of probe concentration fluorescence is linearly proportional to concentration consistent with previous published data (FI) (19, 31, 35, 37), and the single channel permeability can be calculated following:

$${\text{P}}_{\gamma }=\left[{\text{vol}}_2\times \Delta {\text{FI}}_2/\left(\left\{\Delta \text{t}\left[\left({\text{FI}}_{1,n}-{\text{FI}}_{2,n}\right)+\left({\text{FI}}_{1,n+1}-{\text{FI}}_{2,n+1}\right)\right]/2\right\}\right)\right]/n,$$ 4

where ΔFI₂ = (FI_2,n+1 − FI_2,n) is the change of fluorescence intensity in the recipient cell over the time Δt = (t_n+1 − t_n); and FI_1,n, FI_1,n+1, FI_2,n, and FI_2,n+1 are fluorescence intensity measurements at times t_n and t_n+1 in donor and recipient cells, respectively.

To reduce the effect of photo-bleaching, fluorescence in the cells was measured at 1- to 2-min intervals with exposure times of 0.2–1.5 s. As shown previously (34), the fluorescence intensity is linear over the relevant concentration range and single exposures only represented a small loss in signal intensity (<1%). The last recording was taken 12 min after breaking whole cell patch in donor cell and injecting fluorescent probe.

P_γ reflects the probe flux through single channel. However, different connexins exhibit a wide range of single channel conductances and to compare them the ratio between dye flux and the permeability of the channel to ubiquitous monovalent ion K⁺ was calculated as described previously (22, 35). Briefly, the number of molecules (N_γ) that diffuse through the single gap junction channel per second can be calculated as:

$${\text{N}}_{\gamma }=\left({\text{vol}}_2\times \Delta {\text{C}}_2\times \text{Na}\right)/\left(\Delta t\times n\right),$$ 5

where vol₂ is the volume of the recipient cell; ΔC₂ is the estimated change in probe concentration in recipient cell over the time Δt; Na is Avogadro's number; and n is number of channels (see Eq. 2).

Based on studies for the anion-to-cation conductance ratio (2, 32, 42), we assumed that the major current carrier is K⁺, while aspartate is assumed to be an insignificant current carrier. K⁺ flux per channel per second was calculated by dividing the current passing connexin channel by the elementary charge (e = 1.6 × 10⁻¹⁹ C). It was assumed that the 23.4-mV step is roughly equivalent to concentration gradient of × 10. For calculation of probe-to-K⁺ permeability ratios, the concentration differences between K⁺ and fluorescent probes were evaluated and were corrected to the same concentrations.

Signal recording and analysis.

Current and voltage signals were digitized with a 16-bit A/D-converter (Digidata 1322A; Molecular Devices) and stored with a personal computer. Data acquisition and analysis were performed with pClamp9 software (Molecular Devices).

Curve fitting and statistical analyses were performed using SigmaPlot (Systat Software) and GraphPad Prism (GraphPad Software). The difference was considered statistically significant when P was < 0.05. Results are presented as means ± SE.

RESULTS

To compare the permeability of homotypic gap junction channels formed by Cx43, Cx45, Cx40, and Cx26 the following fluorescent probes of different net charges and sizes (Table 1) were used: carboxyfluorescein, LY, AF350, EthBr, and NBD-m-TMA. Double patch-clamp experiments were performed on HeLa cell pairs stably expressing mCx43, mCx45, rCx40, or hCx26, where one pipette was held in whole-cell mode and was used to deliver a fluorescent probe of known concentration into the donor cell, while another cell was kept in perforated patch mode (35). This approach allowed measurement of cell-to-cell transfer of fluorescent probes while simultaneously monitoring transjunctional conductance (g_j) between the cells without significant washout of the probe from the recipient cell (35). The changes in fluorescence intensity in both cells were monitored over time (Fig. 1A). Figure 1 shows an example of AF350 diffusion in HelaCx45 cell pair. Figure 1B shows the junctional current measured in response to 10-mV voltage steps for the same cell pair. Fluorescence intensity (F) was measured at specific time intervals in both donor cell (cell 1) and recipient cell (cell 2; see Fig. 1C). These parameters were used to estimate single channel permeability (P_γ) to the probe (see methods). The steady-state fluorescence intensity detected in the donor cell (cell 1) was assumed to correspond to the probe concentration in the pipette. P_γ was calculated for every two consecutive measurements and remained essentially constant during the course of the experiment (Fig. 1D). Routinely, at the end of each experiment perfusion with CO₂ (100%), the saturated solution was applied to uncouple cell pairs and to prove that dye diffusion occurred through gap junction channels (35).

Fig. 1.

Detection of gap junction channel permeability. A: cell-to-cell transfer of Alexa Fluor350 (AF350) in HeLaCx45 pair. Dual whole cell and perforated patch clamp has been used to assess gap junctional conductance and to deliver the fluorescent probe to the donor cell (cell 1) of a pair. Gradual increase of fluorescence intensity in recipient cell reflects dye diffusion from donor cell to recipient cell (cell 2; epifluorescence micrographs taken at 1, 5, and 12 min after dye application to the donor cell). Scale bar = 20 μm. B: voltage steps were delivered to the cell 1 (V₁ = ±10 mV ); cell 2 was held at 0 mV (V₂) and junctional current (I₂) was recorded. Junctional conductance was ∼8 nS. C: plots of normalized AF350 fluorescence intensity vs. time for injected (●) and recipient cell (□). Indicated parameters were used to calculated single channel permeability P_γ (see methods): F_1,n and F_2,n are fluorescence intensity in donor and recipient cell, respectively, at time t_n; F_1,n+1 and F_2,n+1 are fluorescence intensity at t_n+1; Δt is time interval between 2 successive measurements; and ΔF₂ is change of fluorescence intensity in recipient cell. D: single channel permeability (P_γ) of Cx45 channels to AF350 plotted vs. time. P_γ was calculated for each 2 consecutive measurements from same HeLaCx45 pair as in A and B.

All connexins tested allowed the passage of both anionic and cationic fluorescent dyes, but the fluxes of the probe were connexin-dependent. The flux data for LY, CF, EthBr, AF350, and NBD-m-TMA are summarized in Fig. 2 and Fig. 3, A and B, for Cx43, Cx40, Cx45, and Cx26. The relative fluorescence intensity (a fluorescence intensity ratio of the recipient cell divided by injected cell fluorescence) for each probe was plotted against the measured junctional conductance (g_j). Relative intensity was estimated at 12 min after the dye injection to the donor cell via patch-clamp pipette. Figure 2, A–C, shows plots for Cx43, Cx40, Cx45, and Cx26 for LY, CF, and EthBr, respectively. In all cases the transfer of the probe was essentially proportional to g_j.

Fig. 2.

Cell-to-cell diffusion of fluorescent dyes through homotypic gap junction channels. A–C: summary plots of relative fluorescence intensity for Lucifer yellow (LY; A), carboxyfluorescein (CF; B), and ethidium bromide (EthBr; C) vs. junctional conductance in cell pairs expressing Cx43 (●) and Cx40 (○) (left ) and Cx45 (▴) and Cx26 (▵) (right). Each data point represents relative fluorescence intensity (recipient cell fluorescence intensity over injected cell fluorescence intensity) obtained at the 12-min time point after the injection of dye. Solid lines correspond to the first-order regressions and dashed lines are 95% confidence intervals.

Fig. 3.

Intercellular transfer of N,N,N-trimethyl-2-[methyl-(7-nitro-2,1,3-benzoxadiol-4-yl) amino] ethanaminium (NBD-m-TMA) and AF350 through homotypic gap junction channels. A: relative fluorescence intensity of the cationic dye NBD-m-TMA plotted as a function of g_j in cell pairs expressing Cx43 (●) and Cx40 (○) (left) and Cx45 (▴) and Cx26 (▵) (right). B: relative fluorescence intensity of anionic dye AF350 vs. junctional conductance in cell pairs expressing Cx43 (●) and Cx40 (○) (left) and Cx45 (▴) and Cx26 (▵) (right). Solid lines correspond to the best fit to the first order regressions and dashed lines are 95% confidence intervals. C, left: normalized NBD-m-TMA fluorescence intensity vs. time in injected (▴, ■) and recipient (▵, □) cells in HeLaCx43 (■, □; g_j = 11 nS) and Cx40 (▴, ▵; g_j = 9 nS) pairs. Right: plot of normalized AF350 fluorescence intensity vs. time in injected (■) and recipient (□) cells in HeLa Cx43 (■, □; g_j = 12 nS) and Cx40 (▴, ▵; g_j = 19 nS) pairs.

Linear fits of the data to first-order regression yielded the following slopes: 0.0231 ± 0.0008/nS (Cx43), 0.0017 ± 0.0003/nS (Cx40), 0.0116 ± 0.0009 /nS (Cx45), and 0.0030 ± 0.0005/nS (Cx26) for LY. Comparison of the regression lines by analysis of covariance (GraphPad Prism) revealed that the differences between all the slopes for LY were significant. (P < 0.0001). The slopes for CF were 0.0239 ± 0.0017/nS (Cx43), 0.0026 ± 0.0006/nS (Cx40), 0.0161 ± 0.0026/nS (Cx45), and 0.0090 ± 0.0015/nS (Cx26) and were also statistically different (P < 0.0001). The permeability of gap junctions to LY (Fig. 2A) and CF (Fig. 2B) followed the same order: Cx43 > Cx45 > Cx26 > Cx40; however, LY, which has a larger molecular mass, was less permeable than CF in all cases.

It has been shown that positively charged fluorescent probes are permeable to gap junction channels (10, 14, 31, 35). The cationic dye NBD-m-TMA has been used previously to estimate gap junction permeability (12, 13) and charge-based selectivity of Cx43 (19). Another cationic dye EthBr exhibits pronounced binding to nucleic acids, and it is extensively used for DNA staining. In addition, EthBr when bound to DNA exhibits robust increase in fluorescence intensity (1, 23, 40). Because of these limitations, the exact assessment of the EthBr diffusion rate through the gap junctions is problematic. Therefore, the data obtained for cell-to-cell transfer of EthBr (Fig. 2C) were used only for qualitative comparison of different connexins. The best fit of the data to the first-order regression yielded the following slopes: 0.0089 ± 0.0014/nS (Cx43), 0.0106 ± 0.0013/nS (Cx40), 0.0052 ± 0.0018/nS (Cx45), and 0.0099 ± 0.0024/nS (Cx26) for EthBr transfer. As a result, Cx43, Cx40, and Cx26 exhibited statistically similar permeability to EthBr (P = 0.4025 − 0.7884), while Cx45 showed significantly restricted transfer of solute compared with Cx40 (P = 0.027)

AF350 and NBD-m-TMA are comparable in their molecular weight and their limiting diameter, but they possess opposite net charges. Therefore, the comparison of permeability to these two probes is particularly informative, as it allows one to distinguish between size and/or charge or both depending on the selectivity of the channel. Interestingly, all connexins tested in this study exhibited similar transfer of cationic NBD-m-TMA (Fig. 3A). Slopes of the first-order regression fit for NBD-m-TMA were as follows: 0.0431 ± 0.0026/nS (Cx43), 0.04231 ± 0.0028/nS (Cx40), 0.0387 ± 0.0039/nS (Cx45), and 0.0453 ± 0.0015/nS (Cx26). The statistical analysis of covariance revealed that there was no statistical difference between these slopes, i.e., NBD-m-TMA transfer was similar for all connexins tested (P = 0.1218 − 0.8344). In contrast, different connexins exhibited different permeability to the anionic probe AF350. In Fig. 3B, the plots of relative fluorescence intensity of AF350 in the recipient cell vs. g_j are shown. The plots of relative fluorescence intensity for AF350 shown in Fig. 3B demonstrate restricted diffusion of AF350 through Cx40 and Cx26 compared with Cx43 and Cx45. The differences between slopes of the first-order regression fits for Cx43 and Cx45 data or between slopes generated for Cx40 and Cx26 data points were not statistically significant (P = 0.1636 and P = 0.3467): 0.0420 ± 0.0032/nS (Cx43), 0.0333 ± 0.0038/nS (Cx45), 0.0180 ± 0.001/nS (Cx40), and 0.0209 ± 0.0028/nS (Cx26). However, the differences between the Cx43 slope and Cx40 or Cx26 slopes were very significant (P < 0.001 and P = 0.0033). Furthermore, the Cx45 slope was also significantly different than Cx40 and Cx26 slopes (P = 0.0011 and P = 0.0197).

Figure 3C shows single experiment examples with the relative fluorescence intensities for source cell and recipient cell plotted vs. time for NBD-m-TMA and AF350 for Cx43 and Cx40.

The simultaneous recording of fluorescence intensity of a probe and gap junction conductance allowed us to quantitatively estimate permeability of single channels. Figure 4 shows summary data for single channel permeability P_γ for all four connexins tested. The calculated mean P_γ (cm³/s × 10⁻¹⁵ or attoliter/s) values for Cx43 were as follows: 6.55 ± 0.81 (n = 8), 8.09 ± 0.74 (n = 7), 5.30 ± 0.23 (n = 7), and 4.12 ± 0.68 (n = 7) for NBD-m-TMA, AF350, CF, and LY, respectively. Cx45 yielded the following P_γ (cm³/s × 10⁻¹⁵) values: 3.57 ± 0.45 (NBD-m-TMA; n = 9), 3.81 ± 0.15 (AF350; n = 12), 1.26 ± 0.09 (CF; n = 9), and 0.89 ± 0.06 (LY; n = 12). The data show that both Cx43 and Cx45 had similar permeability orders to fluorescent dyes: AF350 ≈ NBD-m-TMA > CF > LY. Thus they exhibited similar permeability for both anionic (AF350) and cationic (NBD-m-TMA) probes that had comparable size, indicating weak or no charge-dependent selectivity of these channels. For Cx43, P_γ(AF350) was approximately two times higher than P_γ(LY), whereas such a comparison for Cx45 exhibited an approximately fourfold higher P_γ(AF350) value, suggesting that permeability through Cx45 channels were more affected by the increase in solute size compared with Cx43.

Fig. 4.

Single channel permeability. Summary of single channel permeability P_γ (cm³/s × 10⁻¹⁵) for Cx43, Cx45, Cx40, and Cx26 gap junction channels to LY (black bars), CF (light gray bars), AF350 (white bars), and NBD-m-TMA (dark gray bars).

In HelaCx40 cells, the P_γ (cm³/s × 10⁻¹⁵) for NBD-m-TMA was 21.25 ± 2.91 (n = 7), significantly higher (P < 0.001) than for probes of negative net charge: 7.19 ± 1.14 (AF350; n = 9), 1.13 ± 0.19 (CF; n = 7), and 0.75 ± 0.07 (LY; n = 6). Similarly, HeLaCx26 channels yielded the following P_γ (cm³/s × 10⁻¹⁵) values: 24.06 ± 2.39 (NBD-m-TMA; n = 8), 7.71 ± 1.04 (AF350; n = 7), 2.76 ± 0.47 (CF; n = 7), and 1.70 ± 0.28 (LY; n = 5). According the data above, Cx40 and Cx26 exhibited the following cell-to-cell transfer: NBD-m-TMA >> AF350 > CF > LY. In contrast to the results obtained for Cx43 and Cx45, on average P_γ(AF350) for Cx40 and Cx26 was approximately threefold lower than P_{γ(NBD-m-TMA)}, indicating a strong preference of these channels toward cationic probes. The cell-to-cell transfer of solutes possessing −2 net charge (CF and LY) was even more restricted and P_{γ(NBD-m-TMA)}-to-P_γ(LY) ratios were ∼15 and ∼27 for Cx26 and Cx40, respectively. Furthermore, the P_γ of Cx43 channels to LY was ∼2.4 and ∼5.5 times higher that those of Cx26 and Cx40, respectively, despite the fact that unitary conductances of Cx40 and Cx26 channels are approximately two times higher than unitary conductance of Cx43.

Gap junction channels formed by different connexins exhibit different single channel conductances (Table 2), and the cell pairs that have identical gap junctional coupling vary greatly in numbers of functioning channels. To compare the connexin-dependent permeability of gap junctions to fluorescent probes, the permeability of dyes relative to the ubiquitous K⁺ ions was estimated (see methods). Summarized data of fluorescent probe-to-K⁺ ratios for Cx43, Cx45, Cx40, and Cx26 are summarized in Table 2.

Table 2.

Unitary conductance and relative permeability

	γ, pS	LY/K+	CF/K+	NBD/K+	AF350/K+	EthBr/K+*
Cx43	55	0.0285	0.0296	0.0532	0.0519	0.0110
Cx45	25	0.0143	0.0200	0.0478	0.0411	0.006
Cx40	125	0.0020	0.0033	0.0522	0.0223	0.0131
Cx26	110	0.0048	0.0111	0.0560	0.0258	0.0122

Unitary conductance (γ; pipette filled with 120 mM K⁺ aspartate⁻) and permeability of Cx26, Cx40, Cx43, and Cx45 to fluorescent probes relative to K⁺.

^*Because of extensive intracellular binding, EthBr data should be used only for qualitative comparison.

Figure 5 shows the plots of probe-to-K⁺ ratios vs. limiting diameter (median axial diameter) of the probe, demonstrating that all connexins have similar permeability to cationic NBD-m-TMA relative to K⁺ ions. AF350-to-K⁺ and NBD-m-TMA-to-K⁺ transfer ratios for Cx43 and Cx45 channels were similar, whereas Cx40 and Cx26 exhibited restricted passage of anionic probes, indicating that in cell pairs expressing Cx43 or Cx45 cell-to-cell diffusion of AF350 will be approximately two times higher than in cells with the same junctional conductance but expressing Cx40 or Cx26. Transfer of LY and CF through Cx40 and Cx26 channels was even more restricted if compared with Cx43 or Cx45, where Cx43 channel permeability to LY/K⁺ was ∼14 and ∼6 times higher than for Cx40 and Cx26, respectively. The diffusion of solutes of larger size was more restricted through Cx45 compared with Cx43 channels, but the similar transfer of AF350 to K⁺ and NBD-m-TMA to K⁺ indicates that Cx45 has same permeability for anionic and cationic probes of similar size.

Fig. 5.

Permeability of fluorescent probes relative to K⁺ ions. Permeability of cationic (□, ◊, ▵) and anionic (■, ♦, ▴) fluorescent probes relative to K⁺ for Cx43 (●), Cx45 (▴), Cx40 (♦), and Cx26 (■) channels vs. median axial diameter of the probe plotted on log scale.

DISCUSSION

Characterization of gap junction channel selectivity provides insights on the role of gap junction communication between different cells. Connexins have distinctive expression patterns throughout an organism, and the variety of connexin isoforms clearly suggests that besides intercellular electric coupling connexins have more complex functions. Our data demonstrate that connexins have different selectivity toward solute charge and that while all connexins tested here (Cx43, Cx45, Cx40, and Cx26) exhibit similar permeability for cationic probes, the diffusion of anionic solutes through homotypic Cx40 and Cx26 is significantly restricted. Cx43 and Cx45 tend to discriminate solutes mostly by their size without significant charge-dependent selectivity.

Connexins allow the electrical coupling between cells ensuring simultaneous and well-coordinated electrical signal propagation in a multicellular syncytium such as the ventricular myocardium. The significance of the gap junction isoforms is not only evident from the abundance of various diseases related to connexins malfunction/mutation, but it is also clearly demonstrated in knockout and knockin studies. For example, the replacement of one connexin isoform by the another leads to development abnormalities, whereas the replacement of Cx43 in mice by Cx32 can result in multiple defects in the testes, eyes, and mammary glands, indicating the failure of Cx32 to functionally replace Cx43 (28). It has been shown that Cx46 fails to substitute the loss of the Cx50 in eye development (46). It is apparent that different connexin isoforms are able to replace each other with regard to electrical coupling when one considers that the Cx43KOCx32KI mouse has a near normal EKG (28).However, the defects in other organs point to other properties not shared by all connexins. Similar conclusions may be easily drawn from the studies demonstrating that connexin mutations associated to different diseases do not necessarily result in nonfunctional channels (25, 43). The ability of connexins to pass second messengers such as cAMP, cGMP, or IP₃ has been well documented (5–7, 20, 22, 29); furthermore, it has also been demonstrated that not all connexins have the same potential to pass second messengers (5, 22). In addition, some mutations of connexins affect the passage of larger solutes without considerable changes in their single channel conductance (6). All these findings point to the importance of permeability of connexins to larger solutes.

The solute size-dependent selectivity of gap junctions is well known, but not all connexin permeability data may be explained by the size of compounds; consequently, in addition to size-dependent selectivity, the selectivity toward charge has been proposed (2, 10, 33, 42). The probes NBD-m-TMA and AF350 are similar in their molecular weight and estimated size but possess opposite charge. To compare different connexins we calculated the ratios between P_γ(AF350) and P_{γ(NBD-m-TMA)}. If the ratio is close to 1, then that indicates weak selectivity to the charge, while the ratio ≪1 shows that cationic molecules are transferred better than anionic ones. Thus for Cx43 and Cx45, the P_γ(AF350)-to-P_{γ(NBD-m-TMA)} ratios were 1.2 and 1.07, while for Cx40 and Cx26 they were 0.34 and 0.32, respectively, indicating that Cx43 and Cx45 permeability was not selective to the charge. In contrast, Cx40 and Cx26 showed selectivity favoring a probe of positive net charged over anionic AF350. Therefore, we conclude that Cx45 and Cx43 do not exhibit significant selectivity toward the charge of solutes. Our data for Cx43 are consistent with the results recently reported by Heyman and Burt (19), where charge-dependent selectivity of Cx43 was investigated simultaneously injecting cells with two fluorescent probes and comparing their diffusion. The AF350-to-NBD-m-TMA and AF350-to-LY transfer ratios of ∼1.0 and ∼2.9 were reported (19), and they are similar to the ratios obtained in this study (∼1.2 and ∼2.0, respectively). Previously, the lack of charge-based selectivity of Cx43 for relatively small solutes was determined using solutions of different ion composition (41, 44) and from reversal potentials (34).

In contrast to our results reported in this study, Cao et al. (10) suggested that Cx45 channels exhibit preference for positive compounds when cell-to-cell transfer of cationic dye DAPI (MW 279) and LY (MW 443) was compared. However, these probes differ considerably in their size and our data for CF and LY transfer also indicate that the diffusion through Cx45 channels is highly dependent on the size of the probe. For example, P_{γ(NBD-m-TMA)} was approximately four times higher than P_γ(LY) for Cx45, while for Cx43 the P_{γ(NBD-m-TMA)}-to-P_γ(LY) ratio was ∼1.6. Similarly, Weber et al. (45), using series of the size-graded Alexa dyes (all of negative net charge), demonstrated that increase in probe size caused considerable reduction in cell-to-cell transfer through Cx45, while only slight drop in diffusion rate through Cx43 channels was recorded. The data generated in this study show that Cx45 channels exhibit only weak or no selectivity on the basis of the charge of the probe and that the observed differences in the intercellular diffusion are determined rather by the size of the probe.

Heyman and Burt (19) indicated the NBD-m-TMA-to-AF350 permeability ratio for homotypic Cx40 as ∼15, while we report that cell-to-cell transfer of NBD-m-TMA was approximatley threefold faster. Although in both studies Cx40 was shown to favor the cationic molecules, it is unclear why there is a such difference in the reported NBD-m-TMA-to-AF350 transfer ratios. However, our results are consistent with the relative cation-to-anion conductance ratio of 3.4 (3, 42), which was calculated from the changes in rat Cx40 single channel conductance when KCl in pipette-filling solution was replaced with potassium glutamate. We report similar results for homotypic Cx26, where the NBD-m-TMA-to-AF350 transfer ratio was ∼3.3. Previously, Suchyna et al. (33) using equimolar substitution of KCl by potassium glutamate estimated that Cx26 channels are ∼2.6-fold more permeable to cations than to anions of relatively small size. Consequently, our data indicate that Cx40 and Cx26 exhibit similar relative cation/anion permeability for both small and larger solutes.

The comparison of single channel permeability of Cx40 and Cx26 to NBD-m-TMA with LY and CF, with net charge of −2, revealed even more profound restrictions of intercellular transfer of anionic probes. For example, Cx40 and Cx26 were ∼28 and ∼14 times more permeable to NBD-m-TMA than to LY. In addition, the conclusion that passage of anionic solutes through Cx40 and Cx26 is restricted is supported by our recent study demonstrating that diffusion of anionic cAMP (−1) through gap junctions follows a permeability order of Cx43 > Cx26 ≥ Cx40 (22). In this study, the slope of the linear regression of LY transfer in HeLaCx26 pairs plotted as a function of g_j was 0.0039/nS and it is consistent with the slopes of 0.0035/nS (22) and 0.0029/nS (25) reported previously. Mese et al. (25) also provided evidence that Cx26 channels in N2A cells are permeable to EthBr with the slope 0.021/nS, which is similar to the slope for EthBr transfer of 0.010/nS obtained here in HeLa cells.

A nonlinear relationship between g_j and the diffusion rate of NBD-m-TMA in NRK cells, expressing endogenous Cx43, or Rin cells, transfected with Cx43, was reported by others (12, 13). Ek-Vitorin et al. (13) have associated the nonlinearity of the dye spread with different phosphorylation levels of the channel to account for the large range of transfer rates and reported the range of single channel transfer rate of NBD-m-TMA between 6,600 and 2,000,000 molecules/channel/second (at a concentration gradient of 1 mM). However, the latter flux seems to be overestimated as if normalized to the same concentration, the diffusion of fluorescent probe is greater than the expected K⁺ flux/channel/second (∼100,000 K⁺ ions/channel/second at a 1 mM concentration gradient). The NBD-m-TMA diffusion rate for Cx43 reported in this study was much lower, i.e., ∼3,570 molecules/channel/second and cell-to-cell transfer of NBD-m-TMA in HeLa cells, and was directly proportional to the junctional coupling. For our normal experimental conditions, HeLa Cx43 cells exhibit the majority of unitary conductances that could be fit by single normal distribution, suggesting only one type of channel. Moreover, Ek-Vitorin et al. (13) speculated that the fully open state of Cx43 is heavily favored in HeLa cells and thus the impact of phosphorylation on channel permeability and substate behavior is low. However, we are not able to distinguish to which phosphoisoform these channels belong.

Rackauskas et al. (31) reported that homotypic rodent Cx43, Cx45, and Cx40 channels allow passage of both negative (LY and AF350) and positive (EthBr, DAPI, and propidium iodide) dyes. In the case of positive dyes, the single channel permeability was not quantitatively assessed because of the binding of positive probes to nucleic acids. Calculated single channel permeability P_γ for negative dyes (LY and AF350) in Rackauskas et al. (31) study was on average about five to seven times higher for Cx43 [P_γ(LY) of 24.6 × 10⁻¹⁵ cm³/s and P_γ(AF350) of 86.0 × 10⁻¹⁵ cm³/s] and for Cx40 [P_γ(LY) of 6.9 × 10⁻¹⁵ cm³/s and P_γ(AF350) of 33.1 × 10⁻¹⁵ cm³/s] than the P_γ reported here. However, the P_γ values obtained for Cx45 are similar in both studies: 1.1 vs. 0.89 for LY and 5.5 vs. 3.8 for AF350. The discrepancies in the results are most likely due to differences in imaging techniques and methodological approaches, as we have monitored diffusion of the probe for 12 min, while Rackauskas et al. (31) studied permeability the first ∼3–4 min after dye was injected. Despite the differences in calculated P_γ values, both studies demonstrated that Cx43 channels are more permeable to both LY and AF350 than Cx40, despite the fact that the single channel conductance of Cx40 is much higher. Cx43 exhibited ∼2.6-fold higher permeability for AF350 and ∼3.6 higher permeability for LY (31), respectively, than Cx40, while our data indicate that Cx43 is ∼1.2-fold more permeable to AF350 and ∼5.5-fold more permeable for LY than Cx40. The reported P_γ(AF350) was 3.5 times higher for Cx43 and ∼4.8-fold higher for Cx40 than P_γ(LY) (31), while in this study we obtained P_γ(AF350)-to-P_γ(LY) permeability ratios of ∼2.0 and ∼8.9 for Cx43 and Cx40, respectively. In addition, Eckert (11) reported single channel permeability to LY through rat Cx43 channels expressed in HeLa cells and in rat mammary tumor BICR/M1R_k cells expressing endogenous Cx43. The P_γ(LY) values obtained from HeLaCx43 cells pairs were of the wide range and varied between 2.1 × 10⁻¹⁵ and 80 × 10⁻¹⁵ cm³/s, with an average of 17 × 10⁻¹⁵ cm³/s, so the P_γ(LY) of 4.1 × 10⁻¹⁵ cm³/s calculated in this study falls in the reported range. Here estimated LY flux through Cx43, Cx40, and Cx45 channels relative to potassium was 0.0285, 0.002, and 0.014, respectively, which is consistent with LY-to-K⁺ ratios reported previously: 0.028 for Cx43, 0.0025 for Cx40 (35), and 0.010 for Cx45 (16).

In previous reports, Valiunas et al. (35) and Kanaporis et al. (22) used the Levitt continuum equation (24) for multiple ion occupancy to determine pore size. This model is based on channel geometry, ion and channel wall interactions, and ion to ion interactions. Application of this model as illustrated in Valiunas et al. (35) reveals that for Cx43 and Cx45 the pore sizes derived from Levitt equation were similar for LY and CF, 1.3 and 1.1 nm, respectively. These values were higher than those calculated for NBD and AF350, 0.8 and 0.79 nm, respectively. The Levitt equation (24) for determining the effective diffusion coefficient within a channel assumes a spherical solute. None of the probes meets this requirement. In particular, the dimensions of AF350 and NBD molecules given in Table 1 represent the most extreme deviation from a sphere-like structure. The rod-like shapes and the largest apparent diameter of both AF350 (14 Å) and NBD (12.2 Å) are close to pore size and perhaps limiting free entrance into the pore. There is perhaps an orientation dependence for these solutes. Cx40 and Cx26 showed the same trends as Cx43 and Cx45, i.e., larger apparent pore size for LY and CF than AF350 or NBD. The calculated differences in pore size from the probes such as CF vs. AF350 are an indicator that size alone is not the dominant determinant of permeability within gap junction channels, selectivity arising from solute-solute interactions and solute channel-wall interactions are also determinants of permeability.

The emergence of cellular therapies utilizing immune-privileged delivery cells such as adult mesenchymal stem cells has as a cornerstone the ability of those cells to integrate into target tissues and organs. Paramount among the features of those cells is the ability to form gap junctions with target cells in vivo (30). Understanding of connexin perm-selectivity may be invaluable not only for defining physiological role of gap junctions but also for the designing cell-based therapies for delivery of specific drugs or small interfering RNAs. In vitro studies have shown that connexins are capable of transfer solutes of varying size and charge and illustrate that rod-shape solutes such as small interfering RNAs are permeable to gap junction (37, 47). It is now clear that gap junctions are becoming the necessary link for cellular delivery system with the ability to affect inherited and genetic diseases.

GRANTS

This work was supported by Amercian Heart Association Grant AHA-0335236N (to V. Valiunas), National Heart, Lung, and Blood Institute Grant HL-28958, NYSTEM institutional support, and National Institute of General Medical Science Grant RO1-GM-88180 (to P. R. Brink).

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the author(s).