Human ClC-3 is not the swelling-activated chloride channel involved in cell volume regulation.

Volume regulation is essential for normal cell function. A key component of the cells' response to volume changes is the activation of a channel, which elicits characteristic chloride currents (I(Cl, Swell)). The molecular identity of this channel has been controversial. Most recently, ClC-3, a protein highly homologous to the ClC-4 and ClC-5 channel proteins, has been proposed as being responsible for I(Cl, Swell). Subsequently, however, other reports have suggested that ClC-3 may generate chloride currents with characteristics clearly distinct from I(Cl, Swell). Significantly different tissue distributions for ClC-3 have also been reported, and it has been suggested that two isoforms of ClC-3 may be expressed with differing functions. In this study we generated a series of cell lines expressing variants of ClC-3 to rigorously address the question of whether or not ClC-3 is responsible for I(Cl, Swell). The data demonstrate that ClC-3 is not responsible for I(Cl, Swell) and has no role in regulatory volume decrease, furthermore, ClC-3 is not activated by intracellular calcium and fails to elicit chloride currents under any conditions tested. Expression of ClC-3 was shown to be relatively tissue-specific, with high levels in the central nervous system and kidney, and in contrast to previous reports, is essentially absent from heart. This distribution is also inconsistent with the previous proposed role in cell volume regulation.

The maintenance of a constant cell volume in the face of fluctuating intra-and extracellular osmolarity is essential for normal cell function. Following cell swelling upon exposure to hypotonic solution, animal cells restore their volume toward its original value by activation of channels and transporters in the plasma membrane: the loss of K ϩ and Cl Ϫ ions and organic osmolytes, followed by obligatory loss of water, leads to regulatory volume decrease (RVD) 1 (2).
Although a cell swelling-activated chloride current (I Cl, Swell ) required for RVD has been carefully characterized in several cell types, the molecular identity of the channel has not yet been established (3)(4)(5). This is primarily due to three experimental limitations. First, the current is ubiquitous, such that cell lines exhibiting little or no current necessary for expression cloning are not available. Second, no specific, high affinity blockers of the current are known. Third, the magnitude and the rate of activation of the currents are readily perturbed by both endogenous and exogenous factors making quantitative analysis difficult. The latter issue is illustrated by the fact that P-glycoprotein, proposed as a candidate for the cell swellingactivated chloride channel, was subsequently shown to be a regulator of endogenous channel activity (6,7), whereas another candidate, pICl n , was also shown not to be the channel and its precise role is still uncertain (8,9). Another swellingactivated chloride channel expressed in many cell types, ClC-2, has also been suggested as contributing to RVD (10). However ClC-2 generates a current with biophysical and pharmacological characteristics that differ significantly from I Cl, Swell (11)(12)(13), and in a human intestinal epithelial cell line it has been shown that ClC-2 does not contribute to RVD (13).
Recently, a series of studies has led to the suggestion that ClC-3 is the swelling-activated chloride channel responsible for I Cl, Swell . The gene coding for ClC-3 was first cloned from rat by a homology-based cloning strategy; its predicted amino acid sequence is similar to other ClC channels (14,15). The human cDNA was subsequently cloned and sequenced from fetal brain (16), and the guinea pig version was cloned and sequenced from cardiac myocytes (1). Expression of gpClC-3 was reported to increase significantly I Cl, Swell (1,17). This view was supported by antisense experiments where reduction of ClC-3 expression was reported to decrease I Cl, Swell (18).
Subsequently, the role of ClC-3 in I Cl, Swell has been challenged. Attempts to replicate experiments with gpClC-3 and hClC-3, either in Xenopus oocytes or in mammalian cell lines (HEK293 and NIH3T3), were unsuccessful (19). In another study, expression of rClC-3 in CHO cells was reported to generate a Ca 2ϩ -sensitive chloride channel (15). Further complexity was introduced by the suggestion that short and long versions of ClC-3 could potentially be generated in vivo using two different translation initiation sites, identical except for an additional 58 amino acids at the N-terminal of the long version (20). It was reported that expression of the short and long versions generated, in Chinese hamster ovary-K1 cells, distinct currents that are ClC-5-like and I Cl, Swell -like, respectively (20), although the ion selectivity of the ClC-5-like currents is reported both as I Ϫ Ͼ Cl Ϫ and Cl Ϫ Ͼ I Ϫ (21 The tissue distribution of ClC-3 is also ambiguous. Kawasaki et al. (14) reported mRNA coding for ClC-3 mainly in brain and kidney, but not in heart. In contrast, Shimada et al. (20) reported high levels of expression of ClC-3 protein in the liver, and Britton et al. (22) high levels in heart. Furthermore, ClC-3 was reported in bovine non-pigmented ciliary cells to be localized mainly to the nucleus (18), yet to the canalicular membrane in hepatocytes (20).
To resolve the controversy surrounding ClC-3 we generated cell lines expressing the long and short versions of hClC-3-GFP, with hClC-5-GFP as control, in HEK293 cells. Neither the long nor the short version of hClC-3-GFP affected the swelling-activated currents significantly or influenced cell volume regulation. Indeed, ClC-3 generated no detectable chloride currents when overexpressed in response to change in cellular Ca 2ϩ concentrations. The findings presented here exclude ClC-3 as the channel responsible for I Cl, Swell and cell volume regulation.

Generation of Cell Lines Expressing hClC-3-GFP and hClC-5-GFP-
The original cDNA sequence coding for ClC-3 predicted a protein of 760 amino acids (the short version) (14). Subsequently, an upstream ATG codon was identified, and a protein with an additional 58 amino acids at the N terminus was predicted (the long version) (20). It has been suggested that both versions may be expressed (20). To study the localization of hClC-3, we generated cell lines expressing both the short and long versions fused to GFP at the C terminus. It has previously been reported that GFP tags at the C terminus of ClC channels do not impair function (23). As a control for function, a cell line expressing hClC-3 with no GFP tag was also generated. Plasmids containing overlapping cDNA fragments encoding hClC-3 were obtained from Dr. Borsani (16). Appropriate restriction fragments were excised from these plasmids, or generated from these plasmids by PCR, and ligated together to generate a full-length CLCN3 cDNA. The PCR primer at the 5Ј-end (5Ј-AGTTGGAACGCTAGCCACCATGACAAATGGAGGCAGC-3Ј) was designed to introduce a consensus Kozak sequence and a unique NheI restriction site upstream of the start codon. The full-length cDNA was inserted as a NheI-XhoI fragment into the expression vector pCIneo (Promega) to generate the expression plasmid pCI-ClC-3. The cDNA was verified by sequencing. To generate the long version, the missing coding sequence was cloned from human primary lung fibroblasts by a reverse transcriptase reaction (Omniscript, Qiagen) using random hexamer primers (Roche Molecular Biochemicals), followed by PCR using primers specific for ClC-3: the upstream promoter was 5Ј-CGAGATA-ATGCTAGCCCACCATGGAGTCTGAGCAGCTGTTCCAT-3Ј and at the downstream promoter 5Ј-AGAACTGTTAATGGATCCTCCATTTGT-CAT-3Ј. The PCR fragment was ligated as an NheI-BamHI fragment into pCI-ClC-3, which had been modified by a PCR-based insertion of a silent BamHI restriction site downstream of the short version start codon. The resultant plasmid was designated pCI-ClC-3long. To generate the mutant K579N, the Altered Sites mutagenesis system (Promega) with an oligonucleotide containing the desired mutation together with a silent AflII restriction site (5Ј-GAAGCACACATCCGACTTA-AGGGATACCCTTTCTTG-3Ј) was used.
Plasmids encoding hClC-3-GFP fusion proteins were generated by a two-step cloning procedure. First, the stop codon was substituted with an XhoI site by PCR using primer 5Ј-CCTCATCTGTGACTCGAGGTT-GAACATTATTG-3Ј. The EGFP-fragment from the pIRES2-EGFP plasmid (CLONTECH) was then inserted as a NcoI-NotI fragment into plasmid pCI-ClC-3 after digestion with XhoI and NotI, using two short linker oligonucleotides to mediate ligation between the XhoI and the NcoI ends (TCGAGTCTAGAGCCAC and CATGGTGGCTCTAGAC). The plasmid encoding hClC-5-GFP was a generous gift of Prof. T. Jentsch, Hamburg, and was based on the same pCIneo vector and EGFP gene as the hClC-3 fusions generated here.
Antibody Production-A polyclonal antibody against ClC-3, called D1, was produced by injecting a synthetic peptide MTNGGSINSSTH-LLD (corresponding to amino acids 59 -73 of the long version of hClC-3; amino acids 1-15 of the short version) coupled to bovine serum albumin, together with Freund's adjuvant, into rabbits at four time points over a 5-month period (Regal Group Ltd., UK). This peptide was selected because it is specific for ClC-3, and we showed that the antibody D1 does not cross-react with overexpressed ClC-4 or ClC-5 in Western blots (data not shown). Sera from bled animals were affinity-purified using N-hydroxysuccinimide-activated Sepharose (Amersham Pharmacia Biotech) coated with the immunogenic peptide. The eluates were analyzed by SDS-PAGE and Coomassie Blue staining. The IgG-containing fractions were then dialyzed and concentrated using spin columns (Millipore) with a cut-off size of 70 kDa.
The commercial antibody against ClC-3, Anti-ClC-3, was obtained from Alomone and was raised against residues 592 to 661 of short ClC-3. The antibody against GFP, JL-8, was a monoclonal antibody from CLONTECH.
Confocal Immunofluorescence Imaging-Cells, grown at low density on glass coverslips coated with poly-L-lysine, were fixed with 4% formaldehyde in phosphate-buffered saline (PBS) with calcium and magnesium: 8.0 g/liter NaCl, 1.15 g/liter Na 2 HPO 4 , 0.2 g/liter KCl, 0.2 g/liter KH 2 PO 4 , 0.13 g/liter CaCl 2 -2H 2 O, 0.1 g/liter MgCl 2 ⅐6H 2 O) for 30 min at room temperature. Cell nuclei were stained with propidium iodide after treatment with RNase A (Sigma) (100 g/ml) for 3 min at 37°C. The coverslips were then mounted onto glass slides with mounting oil containing anti-fading agent (Vectashield, Vector). Images were taken by a MicroRadiance Bio-Rad confocal microscopy through a 63ϫ planapochromat 1.4 numerical aperture objective. The GFP signal was detected by excitation with the 488-nm line of an argon laser through a HQ515/30 emission filter. The signal from propidium iodide in the same cell section was detected by excitation using the 543-nm line of a helium-neon laser through an E600LP emission filter.
Biotinylation of Membrane Proteins-Cells grown on poly-L-lysinecoated 75-cm 2 flasks or 35-mm dishes to 80% confluency were incubated in 5 ml of reducing buffer (50 mM dithiothreitol, 1 mM EDTA in PBS) for 5 min at room temperature. The cells were then washed twice with 5 ml of 1 mM EDTA in PBS. Cells were labeled by incubation for 30 min at room temperature with 1 ml of biotin maleimide (Molecular Probes) dissolved in DMSO and diluted to 1 mM in EDTA buffer. The reaction was quenched by washing twice with 2 ml of 5 mM dithiothreitol in PBS and twice with PBS plus 1 mM EDTA. The labeled cells were lysed by addition of 1 ml of lysis buffer containing 50 mM Tris-HCl, pH 7.4, 150 mM NaCl, 1% Nonidet P-40 plus complete mini protease inhibitors (Roche Molecular Biochemicals). 1 mg of protein was incubated with 100 l of prewashed immobilized Ultralink neutravidin (Pierce) and gently rotated overnight at 4°C. The slurry was recovered by centrifugation, washed twice (by resuspension and centrifugation) with 0.5 ml of lysis buffer, twice with 0.5 ml of high salt buffer (0.5 M NaCl, 25 mM Tris-HCl, pH 7.5, 1% Nonidet P-40 in dH 2 O), and once with 0.5 ml of 50 mM Tris-HCl (pH 7.5). The bound, biotinylated proteins were eluted by incubating the washed slurry with 100 l of Laemmli sample buffer (50 mM Tris HCl, pH 6.8, 10% glycerol, 5% ␤-mercaptoethanol, 2% SDS, 0.05% bromphenol blue) for 10 min at 90°C and analyzed by Western blotting.
Protein Preparation and Western Blotting-Cell extracts for Western blotting were prepared by detaching cells with PBS containing 5 mM EDTA followed by lysis buffer (50 mM Tris-HCl, pH 7.4, 150 mM NaCl, 1% Nonidet P-40 and complete mini protease inhibitors; Roche Molecular Biochemicals). The lysate was sonicated to shear the DNA.
Proteins from tissues were obtained from whole organs removed from laboratory mice (C57/Bl6) immediately after killing and snap-frozen in liquid nitrogen. Frozen organs were stored at Ϫ80°C until further processing. To homogenize, the frozen organs were first pulverized with pestle and mortar in liquid nitrogen, and then resuspended in ice-cold lysis buffer with protease inhibitors (described above) and treated for 30 s using an Ultra-Turrax homogenizer. Where necessary, proteins were deglycosylated by resuspending the cell lysate in a 50 mM sodium phosphate buffer (pH 7.5) and with 5 units/l PNGase F (New England BioLabs) at 37°C for 1 h with occasional gentle mixing. To digest DNA, Benzonase (Sigma Chemical Co.) was added to a final concentration of 5 units/l and incubated at 37°C for 5 min prior to addition of Laemmli sample buffer.
For electrophoresis, samples were mixed with equal amounts of Laemmli sample buffer and incubated at room temperature for 10 min.
Proteins were separated by SDS-PAGE on 7.5% polyacrylamide gels at 15-40 V/cm for ϳ1.5 h in a Bio-Rad chamber. The separated proteins were then transferred to polyvinylidene difluoride membranes (0.45-m pore size, Immobilon-P, Millipore) using a semi-dry blotting apparatus (Anachem) as described by the manufacturer. After blocking with PBS containing 4% dried skimmed milk and 0.02% Tween 20 for at least 30 min, the membranes were incubated overnight at 4°C with diluted antiserum (1/1000, JL-8 CLONTECH; 1/200 Anti-CLC-3, Alomone and 1/100 D1) in the blocking buffer. Membranes were washed three times for 15 min with PBS containing 0.02% Tween 20, incubated with the horseradish peroxidase-conjugated secondary antibody (antimouse or anti-rabbit-IgG, as appropriate; Dako) for 1 h at room temperature, washed three times for 15 min each, and finally the horseradish peroxidase signal detected using the ECL chemiluminescence system (Amersham Pharmacia Biotech).
Electrophysiology-Chloride currents were measured in whole cell recording mode of the patch-clamp technique as described previously (7). To prevent voltage offset when bath Cl Ϫ concentration was changed, the bath reference electrode was connected through an agar bridge that maintains a constant Cl Ϫ concentration in the immediate vicinity of the Ag/AgCl electrode. The extracellular (bathing) isotonic solution contained 100 mM NMDGCl, 0.5 mM MgCl 2 , 1.3 mM CaCl 2 , 10 mM HEPES titrated to pH 7.4 with Tris. The osmolarity was corrected to 300 mOsm with mannitol. The extracellular hypotonic solution, used to elicit swelling-activated currents, had the same composition as the isotonic solution except that the osmolarity was corrected with mannitol to 220 mOsm. Extracellular solutions with chloride substitution were obtained by replacing 100 mM NMDGCl with 100 mM of the respective anion salts (NaI, NaF, or NaBr), and osmolarity was adjusted with mannitol to 220 or 300 mOsm for hypotonic or isotonic solutions, respectively. The intracellular (pipette) solution was 100 mM NMDGCl, 1.2 mM MgCl 2 , 1 mM EGTA, 10 mM HEPES, 2 mM ATP titrated to pH 7.4 with Tris, and osmolarity was adjusted to 280 mOsm with mannitol. Data are expressed as mean Ϯ S.E. (n, number of cells). Statistical analyses were performed by non-paired t test; statistical significance was accepted for p Ͻ 0.05(*) or p Ͻ 0.01(**).
Cell Volume Measurements-Cells were grown on poly-L-lysinecoated coverslips and loaded, before the experiment, for 5 min with 2.5 M Calcein-AM (Molecular Probes) in isotonic solution (70 mM NaCl, 5 mM KCl, 0.5 mM MgCl 2 , 2 mM CaCl 2 , 5.5 mM glucose, 10 mM Hepes buffer, pH 7.4, and osmolarity was adjusted to 320 mOsm with mannitol). After extensive washing with isotonic solution not containing calcein-AM, cells were transferred to a perfusion chamber set on the stage of the microscope. Experiments were performed at room temperature. Cells were imaged by a MicroRadiance Bio-Rad confocal microscopy through a Plan-Neofluor 40 ϫ 1.3 numerical aperture oil immersion objective (Zeiss), and the dye was excited with a 488-nm line of an argon ion laser. The emitted fluorescence was collected through a QH500LP filter. An optical section passing through the cells was acquired every 30 s, and the average fluorescence signal from a representative cellular area was plotted along the time. To calibrate the fluorescence signal, brief exposures to 15% hypotonic and 15% hypertonic solutions, obtained by adjusting the amount of mannitol, were used (25). The cells were then perfused, for at least 5 min, with an isotonic solution containing 10 g/ml Gramicidin (Sigma) and where NaCl was substituted with 70 mM NMDGCl and then challenged with a 40% hypotonic solution, obtained by reduction of mannitol concentration. The signal was then analyzed and converted in volume measurement and the percentage of RVD was calculated (6,25). The fluorescence signal due to the GFP tag, present in the cells expressing ClC-3-GFP, was negligible in comparison to the signal due to calcein and was shown to be nonosmotically sensitive, so it could readily be accounted for in the calibration of the signal.

RESULTS
Generation of Cell Lines Expressing hClC-3-To facilitate study of ClC-3 and its subcellular location, stable cell lines expressing short hClC-3-GFP and long hClC-3-GFP, and short hClC-3(N579K)-GFP and long ClC-3(N579K)-GFP, were generated in HEK293 cells (see "Experimental Procedures"). It has previously been shown for other ClC channels that a C-terminal GFP tag does not affect function (23). As controls, cell lines expressing hClC-5-GFP (which is known to localize to the plasma membrane and to endosomal compartments (26, 27)), and short hClC-3 without a GFP tag, were also generated. Fig.  1 shows Western blots, probed with anti-GFP antibody (panel a) or antibodies against ClC-3 (panels b and c), demonstrating expression of the appropriate proteins in these cell lines. Note that the fusion proteins remain intact with the GFP tag at the C terminus.
Localization of ClC-3-GFP-To localize the ClC-3-GFP fusion proteins, the cell lines were studied by confocal microscopy. Fig. 2 (top row of panels) shows three optical sections taken along the z axis of a typical cell expressing short hClC-3-GFP. The signal in green corresponds to the ClC-3-GFP fusion protein, and the red signal corresponds to the cell nucleus (DNA stained with propidium iodide). The majority of the hClC-3 protein was localized to an organelle in proximity to the nucleus, presumably the Golgi, and to a variety of differently sized intracellular vesicles scattered throughout the cytoplasm. However, a proportion of the protein was located in a rim around the periphery of the cell, presumably the plasma membrane. A very similar pattern of localization was also observed for short hClC-3 bearing the mutation N579K (middle row of panels) and for hClC-5-GFP (lower row of panels). The cells shown were representative of the large majority of the cells within a single clone and of several independent clones expressing each GFP fusion protein (data not shown).
ClC-3 Is Present on the Plasma Membrane-The data above suggest that a proportion of the ClC-3-GFP fusion protein is localized to the plasma membrane. However, because of the limited horizontal resolution of confocal microscopy, the data do not exclude the possibility that the protein is located in vesicles just beneath the plasma membrane. To demonstrate that the ClC-3-GFP fusion proteins are inserted in the plasma membrane, cells were labeled with the membrane-impermeant, thiol-reactive reagent biotin-maleimide (28). The biotinylated proteins were then isolated using immobilized neutravidin, and any GFP fusion proteins in the biotinylated protein fraction were detected by SDS-PAGE and Western blotting using anti-GFP antibodies. In the absence of biotinylation, no proteins were bound by the resin, showing that the labeling is specific (data not shown). A significant proportion of each of the GFP fusion proteins tested was accessible to biotin maleimide in intact cells, consistent with a plasma membrane location (Fig.  3, top panel a). Under these conditions, an intracellular control antigen known to be located just beneath the plasma membrane, the Ras pathway protein SHC (29) was not labeled, demonstrating that the biotin maleimide does not permeate the  (Fig. 3, bottom panel a). When cell membranes were permeabilized with saponin (Fig. 3b), there was, as expected, a significant increase in labeling of ClC-3-GFP proteins (top panel b), and SHC was now also labeled (bottom panel b). Similar results were obtained also with the clone expressing the native ClC-3 protein (data not shown). Thus, in the HEK cell lines a significant proportion of both the hClC-3 and hClC-5-GFP fusion proteins is located in the plasma membrane.
hClC-3 Is Not a Swelling-activated Channel-Having established that the hClC-3GFP fusion proteins were present in the plasma membrane, cell swelling-activated chloride currents were studied under the whole cell configuration of the patch clamp technique. Cells were maintained at the holding potential of 0 mV and stimulated by a standard protocol from Ϫ80 mV to ϩ120 mV in 40-mV steps. Fig. 4 shows representative current traces in isotonic solution (upper panels), and 7 min after exposure to a 30% hypotonic solution (lower panels) at which time the currents had reached steady state (Fig. 5a). Very low currents were recorded in isotonic conditions (Fig. 5b) except for cells expressing hClC-5. Following hypo-osmotic shock, all cell lines, including the parental HEK293 cells, exhibited a similar current, showing moderate outward rectification, fast activation, and time and voltage inactivation at ϩ120 mV (Fig. 4, bottom panel). No statistically significant difference in either the magnitude (Fig. 5c) or rate of activation of I Cl, Swell was observed between the different lines. Osmotically induced currents were blocked by addition of 10 M tamoxifen, an inhibitor of swelling-activated chloride channel (30) (data not shown).
Although data with other ClC channels show that a GFP tag at the C terminus has no effect on channel properties (23), we also confirmed that no difference in current magnitude or current characteristics were observed between HEK293 cells expressing ClC-3 and ClC-3-GFP fusions (Figs. 4 and 5).
Previous studies (1) have suggested that the N579K mutation of ClC-3 suppresses outward rectification of the current and changes its halide selectivity. Expression of hClC-3(N579K)-GFP induced no change in either the magnitude of swelling-activated chloride currents (Fig. 5c) or its outward rectification (Fig. 4). The anion selectivity of currents recorded in cells expressing the mutated protein (Fig. 6) did not change when compared with the currents recorded from cells expressing the wild type protein: for all cell lines tested the halide permeability was: I Ϫ Ͼ Br Ϫ Ͼ Cl Ϫ Ͼ F Ϫ (Fig. 6).
In conclusion, neither the magnitude, rate of activation, nor characteristics of swelling-activated currents in HEK293 cells were affected by overexpression of any of the versions of hClC-3 tested. This is despite the fact that we demonstrated that the proteins were localized to the plasma membrane and that similar expression of a closely related protein, hClC-5 (see below), generates characteristic currents. These data provide strong evidence that ClC-3 as expressed has no channel activity.
Expression of ClC-3 Does Not Affect RVD-To further test the involvement of ClC-3 in volume regulation, we measured RVD following exposure to hypo-osmotic solution. The measurement of volume change was performed by acquiring an optical section from single cells loaded with the fluorescent dye calcein. Changes in intensity of fluorescence are proportional to changes in the concentration of the dye, which are strictly dependent on changes in cell water volume, allowing the measurement of relative changes in cell volume (6,25). Unexpectedly, a significant proportion of both hClC-3-GFP and the parental HEK293 cells, did not show RVD. The non-responsive cells readily converted calcein-AM to calcein, an indication of active esterases, and did not show leak of the dye indicating an intact plasma membrane. We do not have an explanation for the heterogeneity of response, although it was not ClC-3-dependent. The cells that did exhibit cell volume regulation showed no significant difference in RVD due to the expression of ClC-3 (Fig. 7B). Thus, ClC-3 does not appear to be involved in cell volume regulation.
hClC-5 Is Not Activated by Cell Swelling-hClC-5 is closely related to ClC-3, and a current associated with ClC-5 expression has been recorded previously in Xenopus oocytes and in transiently transfected HEK293 cells (19). We therefore used hClC-5 as a control. Like the ClC-3-GFP cells, in the hClC-5-GFP cell line a similar proportion of the fusion protein was The bottom set of panels shows biotinylated proteins detected with an anti-SHC antibody: This intracellular protein is only biotinylated once the membrane is permeabilized demonstrating that the non-permeabilized cells were intact and that, therefore, the biotinylation of the ClC-GFP fusions reflects their insertion in the plasma membrane.
ClC-3 Is Not the Swelling-activated Chloride Channel located in the plasma membrane (see above). However, in contrast to ClC-3-expressing cells, a typical strong outwardly rectifying current was measured in ClC-5-expressing cells in isotonic conditions (Fig. 8A), with an anion conductance at positive potentials: Cl Ϫ ϭ Br Ϫ Ͼ I Ϫ Ͼ F Ϫ (Fig. 8C). This is similar to currents previously reported for ClC-5, and different in activation, rectification, and ion selectivity from the endogenous swelling-activated currents seen in parental HEK293 cells and in cells expressing hClC-3. Exposure of hClC-5-expressing cells to hypo-osmotic solution elicited a typical swelling-activated current indistinguishable in magnitude and anion selectivity from the endogenous currents of HEK293 control FIG. 4. Swelling-activated chloride currents in ClC-3-expressing HEK293 cells. Representative chloride currents in iso-osmotic (top set of panels) and 7 min after exposure to 30% hypo-osmotic (bottom set of panels) solutions are shown. Cells were stimulated with square voltage pulses from Ϫ80 mV to ϩ120 mV in 40-mV steps from a 0-mV holding potential. a, control HEK-293 cells; b, HEK293-short ClC-3-GFP; c, HEK293-long ClC-3-GFP; d, HEK293-short ClC-3(N597K)-GFP; e, HEK293-long ClC-3(N597K)-GFP; f, HEK293-short CLC-3 (no GFP fusion).

FIG. 5. Mean chloride currents in ClC-3 and ClC-5-expressing cells.
The mean steady-state peak chloride currents for the indicated cell types at ϩ120 mV are shown. a, the rates of chloride current activation were measured from 1 min prior to exposure to hypo-osmotic solution up to 10 min following hypo-osmotic solution and are plotted as a fraction of the maximal steady-state currents at ϩ120 mV. The data were fitted with a sigmoidal function. No significant difference was observed between the rates of activation of any of the cell lines generated. The data shown here are representative of only two cell lines, expressing short ClC3-GFP and ClC-5-GFP, for clarity. b and c, the mean currents in iso-osmotic (panel b, black bars) and after 7 min of exposure to 30% hypo-osmotic solution (steady state) (panel c, open bars) are shown. In panel c the open circles show the individual current readings obtained for different cells, demonstrating the heterogeneity between individual cells. Note the differences in pA/pF scales between b and c. No significant difference in currents between cell types were observed except for ClC-5 expressing cells under iso-osmotic conditions (**). These currents showed characteristics of previously described ClC-5 currents (see also Fig. 8). The number of cells recorded for each cell line (n) were: HEK293, n ϭ 9; short hClC-3-GFP, n ϭ 9; long hClC-3-GFP, n ϭ 5; short N579K hClC-3-GFP, n ϭ 5; short hClC-3, n ϭ 3; hClC-5-GFP, n ϭ 9. cells or hClC-3-GFP-expressing cells (Fig. 8, B and D). These data demonstrate that, unlike hClC-3, hClC-5 is constitutively active under the conditions used and elicits characteristic currents very different from the cell swelling-activated currents seen in HEK293 cells. These data provide a positive control for the absence of currents seen when expressing ClC-3 and also demonstrate that addition of a GFP tag to ClC-5 has no significant effect on current magnitude or characteristics.
Changes in Intracellular Calcium Do Not Activate hClC-3-The data above provide strong evidence against the hypothesis that ClC-3 is responsible for cell swelling-activated chloride currents, and also show that (unlike ClC-5) if ClC-3 is a chloride channel at all, it is not constitutively active under the recording conditions used in this study. The other chloride channels that have been characterized in epithelial cells are activated by change in intracellular Ca 2ϩ . We therefore studied cells expressing hClC-3 in response to elevations of intracellular Ca 2ϩ using the calcium ionophore A23187 (Fig. 9B). This stimulus elicited, in both ClC-3-expressing HEK293 and untransfected HEK293 cells (Fig. 9A), currents with properties similar to well-characterized epithelial Ca 2ϩ -activated chloride channels (31)(32)(33). These currents are easily distinguishable in kinetic characteristics from the cell-swelling activated current. Thus, hClC-3 does not appear to encode a calcium-activated chloride channel.
Tissue Distribution of ClC-3-Regulatory volume decrease is a general property of almost all cell types (2,34). Similarly, a typical I Cl, Swell has been identified in most cell types studied (4,5). Therefore, any candidate protein for I Cl, Swell might be expected to have a widespread tissue distribution. We examined expression of ClC-3 by Western blotting in a wide array of murine tissues using an antibody against the N terminus of ClC-3 (D1). Although the D1 antibody, like the Alomone antibody, is not absolutely specific for ClC-3, comparison of the two antibodies allowed us to unambiguously identify protein bands that correspond to ClC-3. It is important to note that ClC-3 runs as multiple bands in SDS gels. This may reflect the predicted long and short versions, because D1 recognizes both versions (Fig. 1), or may reflect different post-translation modifications. Very high levels of ClC-3 were detected in kidney and the central nervous system. However, for all other tissues examined (thymus, lung, liver and spleen, heart, skeletal muscle, upper and lower intestine, and testis) little or no signal was detected (Fig. 10A). The absence from heart is in contrast to a previous report (22) (Fig. 10B, right panel). To resolve this apparent discrepancy we also probed the blot with Alomone anti-ClC-3 antibody. This antibody (Alomone) did detect a band in cardiac tissue (Fig. 10B, left panel), but because this was not detected by antibody D1 it must reflect a lack of specificity of the Alomone antibody rather than the presence of ClC-3 in the tissue. In conclusion, the distribution of ClC-3 is consistent with a specific role of ClC-3 in kidney and central nervous system, rather than with a more general role in RVD. DISCUSSION Cell volume regulation is an important property of all cells, and characteristic cell swelling-activated chloride currents Chloride currents in cells expressing ClC-5-GFP were measured. A, the family of currents in isotonic conditions elicited by square pulses, ranging from Ϫ80 mV to ϩ120 mV in 40-mV steps from a holding potential of Ϫ40 mV. B, swelling-activated currents recorded after 7-min exposure to 30% hypo-osmotic solution. C, anion selectivity (Cl Ϫ ϭ Br Ϫ Ͼ I Ϫ Ͼ F Ϫ ) recorded for currents in isotonic conditions. D, anion replacement experiments recorded under hypo-osmotic conditions using a ramp protocol identical to that used in Fig. 6.   FIG. 9. Calcium-activated chloride currents in ClC-3 expressing cells. Currents were activated by perfusion with the calcium ionophore A23187 (10 M) in control HEK 293 cells (A) and in cells expressing short hClC-3-GFP (B). The top panels show the chloride currents in iso-osmotic media, and the lower panels show the currents after exposure to the calcium ionophore. Cells were held at 0 mV and pulsed from Ϫ80 mV to ϩ120 mV in 40-mV steps.

FIG. 10. Expression of endogenous ClC-3 in mouse tissues. A,
Western blot of mouse tissues probed with anti-ClC-3 antiserum D1. 100 g of protein was loaded in each lane. The mobility of molecular mass markers is indicated. (heart, skeletal muscle, upper and lower intestine, and testis are not shown). The same protein bands were also detected using the Alomone antibody demonstrating they do correspond to ClC-3 (data not shown). B, ClC-3 is not expressed in murine heart. The blot was probed with the Alomone anti-ClC-3 antibody (left panel) and the anti-ClC-3 polyclonal antibody D1 (right panel). The band of around 90 kDa in heart detected with the Alomone antibody is not detected using the D1 antibody, demonstrating that it is a consequence of a non-specific interaction.
play an important role in RVD in most cell types. Despite intensive study, the molecular identity of the channel(s) underlying I Cl, Swell has been elusive. Recently, ClC-3 has been proposed as a candidate for I Cl, Swell with a role in cell volume regulation (1,17,18,22). However, this conclusion has been questioned (19) and further complicated by the identification of two potential isoforms, a short and long version of ClC-3 (20). In this study we have rigorously assessed whether ClC-3 is responsible for I Cl, Swell and conclude that ClC-3 is neither responsible for I Cl, Swell nor plays a role in RVD.
Cell lines overexpressing hClC-3 were generated. The use of GFP fusions and extracellular biotinylation were used to demonstrate that a significant proportion of ClC-3 was inserted in the plasma membrane. Neither the short nor long form of ClC-3, when overexpressed at the surface of HEK293 cells, had any significant influence on I Cl, Swell or on cell volume regulation. Furthermore, overexpression of a mutant ClC-3, N579K, that has previously been reported to change both the anion selectivity and the rectification of I Cl, Swell (1) had no effect on the characteristics of the swelling-activated currents. As a positive control, cell lines generated in parallel over-expressing ClC-5 exhibited characteristic ClC-5-like currents. Finally, we generated antibodies against ClC-3 and showed that the protein is expressed in a highly tissue-specific fashion, a result incompatible with a general role in RVD or in underlying I Cl, Swell .
We conclude that the previous confusion about the role of ClC-3 is likely due to two experimental difficulties. First, we have shown that the commonly used commercially antibody (Alomone) is not specific to ClC-3 and cross-reacts with other cellular proteins of similar apparent molecular weight. The generation of a new antibody against ClC-3 has also allowed us to clarify the tissue distribution. It appears that ClC-3 is relatively restricted in its distribution, being expressed primarily in kidney and the brain. Second, most or all mammalian cells possess swelling-activated chloride channels: the magnitude of these "background" currents can vary substantially from cell to cell, and, perhaps more importantly, their magnitude is sensitive to manipulations such as transfection and cell confluency (data not shown). Thus, unless a very large number of cells is studied in a controlled and rigorous fashion, and no selection of cells with high current is involved, misleading conclusions can be drawn. Indeed, in preliminary studies using transient transfections we demonstrated higher currents in cells expressing ClC-3 (35), yet, on more detailed analysis, it became apparent that this effect was a result of experimental manipulation of the cells rather than expression of ClC-3 itself.
If ClC-3 is not responsible for I Cl, Swell and has no role in RVD, what is its cellular role? Expression of GFP fusion proteins showed the majority of hClC-3 is located in Golgi or intracellular vesicles. In contrast to a published report (18), no ClC-3 was localized to the nucleus: This previous result was likely an unfortunate consequence of lack of antibody specificity. A similar pattern of subcellular expression was seen for ClC-5-GFP, a protein known to be located and functional in intracellular vesicles in the kidney (26,27). This suggests that ClC-3 may also have a role in intracellular membranes.
When expressed at the cell surface, neither the long or short version of ClC-3 generated chloride currents (whereas ClC-5 as a positive control did), and no current was elicited in response to changes in intracellular calcium or cell volume. A report that short ClC-3 generates a ClC-5-like current with a selectivity of Cl Ϫ Ͼ I Ϫ (21) is countered by a second report from the same team, which suggests that the short version generates currents with the opposite selectivity (I Ϫ Ͼ Cl Ϫ ) (20). We observe no such currents generated by either the long or short version. Thus, the conditions or accessory proteins required to activate ClC-3 are unknown. Either way, these results are incompatible with ClC-3 being responsible for known chloride currents at the plasma membrane and are consistent with an intracellular role.