Corresponding author M. Duszyk: Department of Physiology, University of Alberta, 7–46 Medical Sciences Building, Edmonton, Alberta, Canada T6G 2H7. Email: email@example.com
1The predominant Cl− channel in bovine tracheal epithelial cells has a conductance of ∼71 pS and accounts for more than 80 % of the total chloride conductance. We examined the effects of protein-modifying reagents on channel function and found that amino groups are critically involved in gating.
2Patch clamp studies showed that lysine-specific reagents, such as dimethyl adipimidate (DMA), significantly increased the channel open probability, but not its conductance. This suggests that modified residues are involved in the gating mechanism, but are distant from the channel permeation pathway.
3Kinetic analysis of channel activity showed that histograms of open and closed durations could be well fitted by double exponential distributions, suggesting that the channel has at least two open and two closed states. DMA did not change the number of open or closed states, but increased channel mean open time.
4Since membrane impermeant reagents were effective only from the extracellular side, we conclude that lysine residues in the extracellular domain of the channel are critically involved in gating. These residues may present an important target for site-directed mutagenesis and pharmacological activation of Cl− channels in epithelial cells.
Functional modification of ion channels by reagents that interact with specific amino acids may provide clues regarding the channel structure-function relationship (Spires & Begenisich, 1992; Varnum et al. 1995). As more information about the primary structure of ion channels becomes available, chemical modification methods may help map the location of essential residues in ion channel function. Modification of channel proteins may either reduce ionic current (Drews & Rack, 1988), or cause persistent channel activation (Varnum et al. 1995).
In the present study, we have identified the amino acid residues crucial for normal Cl− channel function in bovine tracheal epithelial cells. Bovine cells have frequently been used to study transepithelial ion movement, since they express DIDS-sensitive Cl− channels and amiloride-sensitive Na+ channels (Durand et al. 1986; Valdivia et al. 1988). The predominant Cl− channel has a linear current-voltage relationship with a conductance of ∼71 pS, it is activated by cAMP-dependent phosphorylation but not by Ca2+ and accounts for more than 80 % of the total Cl− conductance (Valdivia et al. 1988). Its halide permeability sequence (Cl− > Br−≥ I−) resembles a mid-range Eisenman sequence for anions (Wright & Diamond, 1977), indicating that permeation through the pore is a combination of dehydration and weak interaction with an internal binding site.
Here we report that the activity of the epithelial Cl− channel is significantly increased by modification of lysine residues located in the extracellular domain of the channel.
Bovine tracheal epithelial cells
Cattle tracheae were obtained from a local slaughterhouse. The tracheae were removed 10–15 min after death and immediately placed on ice for transport to the laboratory. Epithelial cells were obtained by enzyme digestion and cultured as described previously (Duszyk et al. 1995). Briefly, strips of mucosa were treated with 0.1 % protease and 0.1 % DNase in calcium-free minimum essential medium (MEM) at 4°C for 16–24 h. The enzymes were neutralized with 10 % fetal bovine serum and the cells were plated on collagen-coated Falcon plates. Patch clamp experiments were performed on isolated cells, usually between 1 and 5 days after plating.
Apical membrane vesicles were prepared by scraping bovine tracheae in a buffer consisting of (mm): 60 mannitol, 0.1 MgCl2 and 5 Tris-Hepes (pH 7.4), supplemented with protease inhibitors (μg ml−1): 25 aprotinin, 10 leupeptin and 10 pepstatin A (Sigma). Scrapings were homogenized and sequentially centrifuged at 3000 g for 10 min to remove cellular debris and nuclei, at 10 000 g for 10 min to remove mitochondria, and at 37 000 g for 40 min to collect plasma membranes. MgCl2 (10 mm) was added to the suspension of plasma membranes and the solution was stirred gently for 1 h at 4°C to separate apical from basolateral membrane proteins. Aggregated (basolateral) membranes were then spun down (at 6500 g for 12 min) and the membranes remaining in the supernatant were pelleted (at 100 000 g for 60 min) to produce an apical membrane fraction. Apical membrane protein purity was assessed from the enrichment of alkaline phosphatase (24-fold over the initial homogenate). Similarly, there was a 3-fold enrichment in the activity of Na+,K+-ATPase in this fraction. Apical membrane proteins were subsequently solubilized for 60 min in a buffer consisting of 150 mm NaCl, 175 μg ml−1 phenylmethylsulphonyl fluoride, 1 mm EDTA, 10 mm 3-((3-cholamidopropyl)-dimethylammonio)-1-propanesulphonate (CHAPS) (pH 7.4), and then spun for 60 min at 43 000 g. The supernatant containing the detergent-solubilized proteins was stored at −80°C at a protein concentration of about 3 mg ml−1.
Dephosphorylation and phosphorylation of membrane proteins
Apical membrane proteins (2 mg) were treated with alkaline phosphatase (Sigma, calf intestine, 200 units (mg protein)−1) for 3–-40 min. After incubation, proteins were sedimented at 100 000 g for 45 min at 4°C and the pellet was resuspended in solubilization buffer. For re-phosphorylation experiments a catalytic subunit of protein kinase A (Sigma, 200 units (mg protein)−1) was added in combination with 1 mm ATP and 2 mm MgSO4 to 2 mg of dephosphorylated apical membrane proteins. The mixture was incubated for 15 min at room temperature (21°C), and the proteins were sedimented and resuspended as described above.
The chloride-sensitive dye, 6-methoxy-N-(3-sulphopropyl)quinolinium (SPQ, Molecular Probes), was loaded into vesicles at the time of protein reconstitution (Dunn et al. 1989). Asolectin lipids (Sigma) were resuspended at 50 mg ml−1 in a buffer containing (mm): 100 sodium gluconate, 45 N-methyl-D-glucamine gluconate (NMG-gluconate), 5 potassium gluconate and 10 Hepes (pH 7.4), with 0.02 % NaN3 and 1.5 % CHAPS. Lipid aliquots were mixed with proteins and 100 μl of 100 mm SPQ (final volume 1 ml, containing 2 mg protein, 10 mg lipid and 10 mm SPQ). The mixture was incubated for 20 min at 4°C, followed by the removal of CHAPS and extravesicular SPQ by gel filtration on a Sephadex G-50–80 column (1.5 × 90 cm2). During reconstitution, SPQ was trapped within the vesicles, and the dye-loaded vesicles eluted in the void volume. The diameter of the vesicles was 184 ± 47 nm (mean ±s.d., n= 8 preparations), as estimated by laser light scattering using a Brookhaven BI-90 particle size analyser.
SPQ fluorescence was excited at 350 nm and the emission was measured at 440 nm with an SLM 8000C spectrofluorimeter (SLM, Urbana, IL, USA). The intravesicular solution contained (mm): 100 sodium gluconate, 5 potassium gluconate, 45 NMG-gluconate, 10 Tris-Hepes (pH 7.4), while the external solution contained 100 sodium gluconate, 50 KCl and 10 Tris-Hepes (pH 7.4). Under these ionic conditions the membrane potential was clamped at +60 mV, creating a favourable electrochemical gradient for Cl− flux. Chloride flux experiments were performed with vesicles containing ∼50 μg protein. Flux was calculated as the initial rate of change of intravesicular Cl− concentration after addition of 4 μm valinomycin (Sigma) which clamped the vesicles at the K+ equilibrium potential (Illsley & Verkman, 1987). A single exponential was fitted to the data and since the initial rate of change of intravesicular Cl− concentration is directly proportional to Cl− flux, the two terms are used interchangeably. To compensate for variations in absolute Cl− flux between different vesicle preparations, each value in an individual data set was expressed relative to the mean control value for that set, and then the relative change in Cl− flux was calculated from all sets. All experiments were performed at 21°C.
Patch clamp recording and data analysis
Pipettes were made from borosilicate microfilament glass. Tips were coated with Sylgard (Dow Corning) and fire polished. Pipette resistance was ∼12 MΩ and offset potentials were measured and corrected before forming a seal. The pipette solution contained (mm): 140 choline chloride, 1 MgCl2, 1 CaCl2, 10 Hepes (pH 7.4). The bath solution contained (mm): 140 NaCl, 1 MgCl2, 1 CaCl2, 10 Hepes (pH 7.4). All potentials are reported relative to zero in the extracellular solution, and positive currents are outward throughout.
Single channel currents were recorded from excised patches in the inside-out configuration using a model EPC-7 amplifier (Adams & List Associates, Ltd, Great Neck, NY, USA). Current signals were filtered using a 10 kHz bandwidth Bessel filter, fed into a digital VCR recorder adaptor (PCM-2, Adams & List Associates, Ltd) and stored on videotape. A digital computer sampled the data at 10 kHz with a 12 bit analog-to-digital convertor. During data processing, the recordings were usually further filtered using a Gaussian digital filter (either 5 or 1 kHz).
The procedures used for data analysis were based on those described by Colquhoun & Sigworth (1985). The half-amplitude criterion was used as a threshold to distinguish between open and closed states. Event durations were corrected for filter rise time by a polynomial approximation (see Eq. 17 of Colquhoun & Sigworth, 1985). Distributions of open and closed times were created and the probability of the channel being in the open state was calculated. Only openings longer than the filter dead time were used to compute the mean channel current amplitude. Histograms of open and closed intervals were created using the square root of the number of events vs. the log bin width of event durations (Sigworth & Sine, 1987). Distributions were corrected for sampling promotion error using the probabilistic redistribution method of Korn & Horn (1988). The distributions were fitted with probability density functions of the form:
where k is the number of exponential components, and aj and τj are the fitted area and time constant, respectively, of each component j. The data were fitted using the maximum likelihood method (Colquhoun & Sigworth, 1985), assuming that each event had a duration equal to the mid-point of its bin. The number of significant components k was determined using the Akaike likelihood ratio test (Korn & Horn, 1988). An Akaike predictor value, PA/B, was determined from:
where LA and LB are the maximum likelihoods that the observed experimental data were drawn from the distributions predicted by models A and B, respectively, and nA and nB are the numbers of free parameters in both models. If PA/B > 0, then model A was ranked above model B; if PA/B < 0, then model B was ranked above model A. Likelihood estimates were obtained for three Markov chain models consisting of one, two and three exponentials.
Amino acid-specific reagents
The following chemicals were applied to SPQ-containing vesicles or to excised inside-out patches in 0.05–10 mm doses: diethylpyrocarbonate (DEP, primary target - histidine), dimethyl adipimidate (DMA, amino groups), methyl acetimidate (MA, amino groups), 5,5′-dithio-bis(2-nitrobenzoic acid) (DNTB, sulfhydryl groups), 4,4′-diisothiocyanatodihydrostilbene-2,2′-disulphonic acid (H2-DIDS, amino groups), 3,3′-dithio-bis(sulphosuccinimidyl propionate) (DTSSP, amino groups), dithiothreitol (DTT, disulphide reducing agent), N-ethoxycarbonyl-2-ethoxy-1,2-dihydroquinoline (EEDQ, carboxyl groups), N-ethylmaleimide (NEM, sulfhydryl groups) and phenylglyoxal (PGO, arginine). All chemicals were purchased from Sigma except DTSSP (Pierce). The reagents were added from stock solutions prepared fresh for each experiment and stored and applied in dim light to reduce photo-oxidation. After treatment vesicles were washed with SPQ-free buffer (at 60 000 g, 4°C for 10 min) and used the same day.
The data are expressed as means ±s.d. Student's t test was used for statistical analysis between two group means. Statistical differences among multiple group means were determined using one-way analysis of variance (ANOVA), and P < 0.05 was considered significant.
The effect of phosphorylation on Cl− flux in reconstituted vesicles
We used the Cl−-sensitive fluorophore SPQ to monitor Cl− conductance in liposomes containing reconstituted apical membrane proteins. Figure 1 presents the effects of protein phosphorylation on SPQ fluorescence quenching. Apical membrane proteins were either phosphorylated with the catalytic subunit of protein kinase A and ATP, or dephosphorylated with alkaline phosphatase and then reconstituted into liposomes. Dephosphorylation significantly reduced Cl− flux into liposomes, whereas phosphorylation increased its magnitude compared with untreated proteins (Fig. 1). However, when the liposomes containing phosphorylated proteins were treated with alkaline phosphatase, no significant reduction in Cl− flux was observed. Statistical analysis of these experiments indicates that phosphorylation increases Cl− flux (P < 0.05, n= 6 separate preparations), whereas dephosphorylation decreases Cl− flux (P < 0.01, n= 6). Dephosphorylation of proteins reconstituted into liposomes had no effect on Cl− flux (P > 0.05, n= 6). These experiments suggest that Cl− channel phosphorylation sites are not accessible to alkaline phosphatase after reconstitution into liposomes.
The effect of chemical reagents on Cl− flux
Figure 2A shows a summary of the effects of group-specific reagents on Cl− flux into vesicles containing apical membrane proteins. Phenylglyoxal (PGO), an arginyl-specific reagent, had no effect on Cl− flux, suggesting a lack of involvement of this residue in Cl− flux regulation. Similar results were obtained with the carboxyl group modifier EEDQ. The roles of disulphide bonds and sulfhydryl groups were examined by application of DTT and NEM, respectively. As with PGO and EEDQ, neither compound affected Cl− flux. These results were subsequently confirmed by modifications with two other cysteine-specific reagents, DNTB and DEP. However, modification of amino groups with MA, DMA or DTSSP clearly increased Cl− flux. Figure 2B shows the time course of intravesicular SPQ fluorescence after treatment with DTSSP.
Modification of amino groups by stilbene disulphonates is known to inhibit Cl− channel activity. In particular, H2-DIDS has been shown to inhibit bovine tracheal Cl− channels reconstituted into lipid bilayers (Ran et al. 1992). We used this compound to investigate whether amino groups associated with channel activation are likely to be separate from those related to channel inhibition. Figure 3 shows that 50 μm H2-DIDS caused a reduction in Cl− flux when applied to liposomes containing native proteins, but it had no significant effect on Cl− flux after the liposomes were incubated with 10 mm DMA for 5–10 min.
Single channel studies
Figure 4A shows a typical recording of a Cl− channel from bovine tracheal epithelial cells. The channel had a linear current-voltage relationship with a conductance of 71.2 ± 5.4 pS (n= 14), and a voltage-independent open probability of 0.38 ± 0.27 in the range of ±80 mV (n= 5), in excised inside-out patches. Figure 4B shows recordings from the same inside-out patch after adding 10 mm DMA to the solution bathing the intracellular side. The current-voltage relationship for the recording in Fig. 4A is shown in Fig. 4C. There was no significant change in channel conductance after DMA treatment (74.1 ± 5.6 pS, n= 6), but the channel open probability increased significantly (0.85 ± 0.14, P < 0.01, Fig. 4D). Kinetic analysis of channel activity was performed on data obtained from patches where there was clearly only one channel present and where the noise level allowed a 5 kHz filter to be used (11 control and 4 DMA recordings). The number of transitions varied between 1485 and 12 861 events. An example of kinetic analysis is shown in Fig. 5. Distributions of open and closed states were fitted with the sum of exponential functions and the likelihood ratio test was used to determine the minimal number of exponential components (eqn (2)). In eight control recordings, the best fit to the open durations was obtained with the sum of two exponentials (Table 1), whereas three recordings were best fitted with the sum of three exponentials. For the closed durations, ten recordings were best fitted by the sum of two exponentials, whereas one recording was best fitted by the sum of three exponentials.
Table 1. Kinetic parameters for the double exponential Markov model of channel gating
Time constant (ms)
Time constant (ms)
The data were obtained from channels held at −40 mV and the area and time constant were determined using eqn (1). The number of transitions varied between 1485 and 12 861 events.
0.64 ± 0.09
0.14 ± 0.08
0.58 ± 0.11
0.18 ± 0.07
0.35 ± 0.08
3.62 ± 1.41
0.41 ± 0.08
7.98 ± 2.54
0.11 ± 0.05
0.36 ± 0.11
0.96 ± 0.05
0.11 ± 0.05
0.88 ± 0.14
12.32 ± 7.16
0.05 ± 0.04
4.07 ± 1.89
The distributions of open and closed durations of all four recordings after DMA treatment were best fitted by the sum of two exponentials (Table 1). Both time constants for open durations significantly increased (P < 0.01) compared with the controls. For closed durations the first time constant was not changed significantly, whereas the second time constant decreased by ∼50 %. These results suggest that the 71 pS channel has at least two open and two closed states and that DMA treatment extends the time that the channel remains in the open state. No relationship could be found between kinetic parameters and voltage, reflecting the lack of dependence of open probability on membrane potential.
Figure 6 shows a typical recording of the 71 pS Cl− channel in an excised inside-out patch, in the absence and presence of 1 mm DTSSP in the solution bathing the intracellular side. The mean channel conductance was 73.6 ± 7.2 pS, and the open probability was 0.41 ± 0.32 (n= 5). The presence of DTSSP did not significantly change the channel conductance (70.9 ± 6.4 pS) or the channel open probability (0.36 ± 0.29, P > 0.05). Since DTSSP is membrane impermeant and increases Cl− flux into vesicles, this result suggests that the residues modified by DTSSP are located on the extracellular side and are not accessible to modification in the excised inside-out patch.
Our results show that amino group-specific reagents increase the activity of the 71 pS Cl− channel. Since imidoesters such as MA and DMA are regarded as truly lysine-specific reagents (Makoff & Malcolm, 1981), this suggests that lysine residues are involved in control of Cl− channel activity. The possibility that the observed effects are due to modification of the amino group of membrane phospholipids is less likely as similar results were obtained in patch clamp and Cl− flux studies. Such phospholipids would have to be considered an integral part of the channel that is preserved after reconstitution into lipid membranes.
The fact that the 71 pS Cl− channel is regulated by PKA-dependent phosphorylation allowed us to demonstrate the orientation of the channels after reconstitution into liposomes. Dephosphorylation of channel proteins before reconstitution reduced Cl− flux, confirming that PKA-dependent Cl− channels contribute the main fraction of Cl− flux. However, dephosphorylation of channel proteins after reconstitution into liposomes had no significant effect on Cl− flux. This suggests that phosphorylation sites on channel proteins are not accessible to alkaline phosphatase after reconstitution, and are probably located inside the liposome. This conclusion is in agreement with other experimental observations suggesting that proteins are incorporated into lipid bilayers in a unidirectional orientation determined solely by protein features (Brunen & Engelhardt, 1993). Additionally, studies of protein orientation in liposomes indicated that between 75 and 95 % of all proteins preserved their original orientations in the cell membrane (Gennis, 1989). Therefore we assumed that channel orientations in vesicles were similar to their original orientations in the native cells.
The different effects of DMA and DTSSP on channel function in excised patches is probably related to the fact that DMA is membrane permeant (Han et al. 1984), while DTSSP is membrane impermeant (Varnum et al. 1995). This suggests that the modified lysine residues are located on the extracellular surface of the cells. It is interesting to note that while Cl− flux was significantly reduced by 50 μm H2-DIDS in liposomes containing native proteins, this blocker was ineffective after modification of Cl− channels with DMA. This suggests either that the binding site for H2-DIDS after DMA modification is occupied by DMA, or that DMA modification causes conformational changes in channel proteins that prevent the binding of H2-DIDS.
DMA and DTSSP are bifunctional reagents that introduce cross-linking of lysine residues, and are often used to identify spatial arrangements in proteins (Han et al. 1984). Both reagents increased the Cl− flux, suggesting that the Cl− channel has two lysine residues amenable to cross-linking reactions. However, the mono-functional imidoester MA exerted a similar effect, indicating that only one lysine residue has to be modified to activate the Cl− channel.
Patch clamp studies have shown that the channel open probability, but not its conductance, increased after DMA treatment. This indicates that the modification site is an important part of the channel gating mechanism and is probably distant to the channel permeation pathway. Kinetic analysis of channel activity indicated that the distributions of open and closed durations were best approximated by a sum of two components, suggesting that the channel has a minimum of two open and two closed states (Korn & Horn, 1988). After DMA modification, both distributions were also best approximated by a sum of two components but the fitting parameters were significantly changed. As a result, the channel spent significantly more time in the open state than before DMA treatment. This effect is similar to the stimulation of CFTR Cl− channels by covalent modification of the regulatory domain with NEM (Cotten & Welsh, 1997). These authors concluded that the conformational change of this domain may be a key feature that regulates channel activity.
Full understanding of ion channel function requires molecular information on channel sequence and structure and their relationship to function. Recently, Cunningham et al. (1995) cloned a putative Ca2+-dependent Cl− channel from bovine trachea. The lack of amino acid sequence information in the case of the 71 pS Cl− channel prevents us from relating observed effects of amino group modification to precise locations on the channel protein. However, several studies have shown that amino acid residues involved in channel gating may be located in the extracellular domain. For example, Lopez-Barneo et al. (1993) have shown that mutations near the outer mouth of the Shaker K+ channel pore produced drastic changes in C-type inactivation kinetics. The authors proposed a ‘foot-in-the-door’ model of channel gating, which could reflect the interaction of ions with amino acids in the channel pore.
Modification of amino groups by imidoesters and DTSSP produces large kinetic changes in channel function, suggesting that these residues may be important candidates for site-directed mutagenesis studies. The fact that these residues are accessible from the extracellular side may play an important role in manipulation of Cl− channels in intact cells. Therefore, our work may have clinical implications in diseases characterized by reduced ion channel permeability, such as cystic fibrosis. Although the lack of protein specificity of amino group-specific reagents prohibits their therapeutic use, their stimulatory effects may suggest other pharmacological approaches by which channel activity can be augmented through modification of the extracellular domain.
We thank Dr S. Dunn for advice in preparation of liposomes, and Dr M. Radomski (Pharmacology) for helpful discussions. This work was supported by the Medical Research Council of Canada and the Canadian Cystic Fibrosis Foundation.