The chloride channel ClC-4 contributes to endosomal acidification and trafficking.

Mutations in the gene coding for the chloride channel ClC-5 cause Dent's disease, a disease associated with proteinuria and renal stones. Studies in ClC-5 knockout mice suggest that this phenotype is related to defective endocytosis of low molecular weight proteins and membrane proteins by the renal proximal tubule. In this study, confocal micrographs of proximal tubules and cultured epithelial cells revealed that the related protein ClC-4 is expressed in endosomal membranes suggesting that this channel may also contribute to the function of this organelle. In support of this hypothesis, specific disruption of endogenous ClC-4 expression by transfection of ClC-4 antisense cDNA acidified endosomal pH and altered transferrin trafficking in cultured epithelial cells to the same extent as the specific disruption of ClC-5. Both channels can be co-immunoprecipitated, arguing that they may partially contribute to endosomal function as a channel complex. These studies prompt future investigation of the role of ClC-4 in renal function in health and in Dent's disease. Future studies will assess whether the severity of Dent's disease relates not only to the impact of particular mutations on ClC-5 but also on the consequences of those mutations on the functional expression of ClC-4.

Mutations in the gene coding for the chloride channel ClC-5 cause Dent's disease, a disease associated with proteinuria and renal stones. Studies in ClC-5 knockout mice suggest that this phenotype is related to defective endocytosis of low molecular weight proteins and membrane proteins by the renal proximal tubule. In this study, confocal micrographs of proximal tubules and cultured epithelial cells revealed that the related protein ClC-4 is expressed in endosomal membranes suggesting that this channel may also contribute to the function of this organelle. In support of this hypothesis, specific disruption of endogenous ClC-4 expression by transfection of ClC-4 antisense cDNA acidified endosomal pH and altered transferrin trafficking in cultured epithelial cells to the same extent as the specific disruption of ClC-5. Both channels can be co-immunoprecipitated, arguing that they may partially contribute to endosomal function as a channel complex. These studies prompt future investigation of the role of ClC-4 in renal function in health and in Dent's disease. Future studies will assess whether the severity of Dent's disease relates not only to the impact of particular mutations on ClC-5 but also on the consequences of those mutations on the functional expression of ClC-4.
There are nine members of the ClC family of chloride channels in mammals, and several members have been implicated in congenital diseases. For example, mutations in ClCN1 cause congenital myotonia (1)(2)(3)(4); mutations in ClCN2 cause idiopathic generalized epilepsy (5); ClCN5 is mutated in Dent's disease, a renal disease characterized by low molecular weight proteinuria, hypercalciuria, and in some severe cases renal failure (6 -11); and mutations in ClCN7 are associated with osteopetrosis (12). Each of these particular chloride channels belongs to distinct subgroups of the ClC family, defined on the basis of their degree of sequence similarity. ClC-1 is grouped with ClC-2, ClC-Ka, and ClC-Kb (7). ClC-7 is most closely related to ClC-6 (7). ClC-5, ClC-4, and ClC-3 form a distinct subgroup, sharing close to 80% sequence identity.
The subgroup of ClC channels, including ClC-3, ClC-4, and ClC-5, is thought to function in intracellular compartments.
For example, ClC-3 is localized in synaptic vesicles in neurons where it has been shown to contribute to vesicular acidification probably by providing an electrical shunt permissive to V-type ATPase activity (13). In non-neuronal tissue, the shorter isoform (ClC-3A) has been localized to late endosomes and lysosomes where it is presumed to function in regulating the pH of these compartments (14,15). The function of the longer Golgilocalized isoform (ClC-3B) has yet to be determined (14).
ClC-5 is localized primarily in early endosomes in native tissues of rodent renal proximal tubules and in various heterologous expression systems (16,17). Disruption of Clcn5 in mice leads to defective fluid phase and receptor-mediated endocytosis by the renal proximal tubule, arguing that ClC-5 contributes to endocytosis in vivo (16, 18 -20). On the basis of these studies, it has been hypothesized that ClCN5 mutations may lead to low molecular weight proteinuria and hypercalciuria in patients with Dent's disease because of defective internalization of protein and membrane receptors from the apical surface of the renal proximal tubule (19). Furthermore, it was proposed that ClC-5 normally contributes to endocytosis by facilitating endosomal acidification, a phenomenon essential for appropriate vesicular trafficking (21)(22)(23). As for ClC-3A, it has been suggested that ClC-5 may provide an electrical shunt for charge dissipation, thereby permitting endosomal acidification through the action of V-type ATPases (15,24). In fact, endosomal vesicles purified from Clcn5 knockout mice exhibited slower rates of acidification than vesicles purified from their wild type siblings (18). However, a direct contribution of ClC-5 to this function in vivo has not yet been demonstrated.
ClC-4 shares 78% sequence identity with ClC-5, and these two channels exhibit almost identical channel properties when studied in heterologous expression systems (25). However, unlike ClC-5, very little is known about the native expression and function of ClC-4 in epithelial cells. Recently, we generated a specific antibody against ClC-4, and we showed that in rodent and human intestinal epithelia, ClC-4 channels co-localize with the cystic fibrosis transmembrane conductance regulator in the apical membrane and in subapical vesicles (26). Significantly, a proportion of the intracellular vesicles bearing ClC-4 appeared to co-localize with the endosomal marker EEA1 (26), raising the possibility that ClC-4 may participate with ClC-5 in regulating the function of this organelle.
The primary goal of the present project was to assess the role of ClC-4, relative to ClC-5, in endosomal trafficking in epithelial cells and to determine the mechanism underlying this putative function. We show for the first time that the specific depletion of ClC-4 reduces the rate of transferrin receptor recycling and that this reduction is associated with a defect in endosomal acidification. Hence, normal endosomal trafficking in epithelial cells may require the functional expression of ClC-4 as well as ClC-5.

Constructs-
The antisense murine ClC-4 was generated as described previously (26). Similarly, the antisense human cDNA ClC-5 (ClC-5 cDNA was a gift from Dr. T. J. Jentsch) was generated by cloning the ClC-5 open reading frame with BamHI (5Ј) and EcoRI (3Ј) linkers on the forward and reverse primers, respectively. As described previously, the resulting PCR fragment was subcloned into pcDNA3.1 (Ϫ) to create the antisense plasmid. The HA 1 tag was inserted onto the amino terminus of hClC-5 with the help of BamHI-ATG-HA-hClC-5 oligonucleotide (ϩ) and hClC-5 EcoRIstop (Ϫ)oligonucleotides. The subsequent PCR product was then subcloned into pCDNA 3.1 (ϩ) with the BamHI and EcoRI linkers for eukaryotic expression. Pfu enzyme (Stratagene) was used for PCR, and sequencing was done with three different oligonucleotides that spanned the hClC-5 open reading frame.
Tissues-Kidney tissues were obtained from adult male rats (Wistar) fed a standard diet. For immunofluorescence staining, we infused fluorescein isothiocyanate (FITC)-labeled dextran, 10 kDa (1.75 mg/100 g body weight; Molecular Probes, Leiden, The Netherlands), dissolved in 0.5 ml of 0.9% NaCl, into the penile vein of a rat over a period of 30 s as described previously (27). Eight minutes after the injection of FITCdextran, the infrarenal abdominal aorta was cannulated, and the rats were subjected to perfusion-fixation for subsequent visualization of FITC-dextran. Rats were perfused with 2% paraformaldehyde in phosphate-buffered saline (PBS) for 3 min and subsequently with a 18% sucrose solution in PBS for another 3 min. Kidneys were removed and cut along the longitudinal axis and incubated in a 18% sucrose/PBS solution, containing 0.02% sodium azide for 24 h at 4°C. Tissues were then frozen in liquid nitrogen and stored at Ϫ80°C until use.
Immunoblotting-Expression of ClC-4 and ClC-5 protein in Caco-2, CHO cells, and rat tissue was determined by immunoblotting as described previously (28,29). The anti-ClC-4 polyclonal antibody was generated against a GST fusion peptide containing amino acid residues 1-52 of mouse ClC-4. The antiserum was pre-absorbed to a GSTcoupled matrix to remove anti-GST antibodies. The signal recognized by anti-ClC-4 antibody is specific for ClC-4 as it is competed by the fusion peptide of the amino terminus of ClC-4 but not a fusion peptide containing the amino terminus of the closely related channel protein ClC-5 (26). Furthermore, immunoreactive protein, migrating as a 90-kDa polypeptide in Western blots, is absent in immunoblots of tissue obtained from Clcn4 null mice. The anti-ClC-5 polyclonal antibody (a gift of Dr. T. Jentsch, Hamburg, Germany) was generated against a synthetic peptide (CKSRDRDRHREITNKS) representing a part of the amino terminus of ClC-5. This antibody was originally characterized and shown to be specific by Gunther et al. (16). Immunoreactive protein was detected using the ECL system (Amersham Biosciences).
Culture and Transfection of Cells-Caco-2, CHO, and LLCPK1 cells were obtained from the American Type Culture Collection (Manassas, VA). Cells were transfected using Lipofectin (Invitrogen), as described previously (26). For immunofluorescence studies, cells were used 1 day following transfection.
Transferrin-Biotin Endocytosis Assay-Receptor-mediated endocytosis was measured by assaying the uptake of biotinylated human transferrin (Molecular Probes, Leislen, The Netherlands) by Caco2 cells. Cells were seeded onto 55-mm Petri dishes and grown to 50% confluency prior to transfection. 42 h post-transfection, cells were first serumstarved for 1 h and then incubated in serum-free media supplemented with 50 g/ml biotinylated-transferrin for 20 min at 37°C. Cells were then washed twice with cold PBS, pH 4.5, to remove surface-associated transferrin-biotin. Cells were lysed using SDS sample buffer and subjected to SDS-PAGE analysis. Internalized transferrin was detected using Extravidin (Sigma) and quantitated by band densitometry analysis using ImageJ software.
Endosomal pH measurements-Endosomal pH was determined using fluorescence ratio imaging as described (30, 31). Caco cells plated on coverslips were co-transfected with enhanced blue fluorescence protein and either vector control or ClC-4 or ClC-5 antisense and serum-starved overnight. Endosomes were loaded with 150 g/ml FITC-transferrin in sodium-rich saline (140 mM NaCl, 5 mM KCl, 5 mM glucose, 1 mM CaCl 2 , 1 mM MgCl 2 , 10 mM HEPES, pH 7.4) for 30 min. Coverslips were washed and pulsed briefly with PBS (pH 5.0) to remove surface-bound FITC-transferrin before transferring to Leiden chambers for analysis. Transfected cells were identified using UV filters using a ϫ40 oilimmersion objective on a thermostatted Leiden holder of a Zeiss IM-35 microscope. Internalized FITC-transferrin was then excited at alternating wavelengths of 490 (700 ms) and 440 nm (100 ms) using a Sutter filter wheel. The fluorescent light was directed to a 535-nm emission filter placed before a cooled CCD camera used for fluorescent detection using 8 ϫ 8 binning. Image acquisition was controlled using the Metafluor software (Universal Imaging Corp.). Images were captured at 2-min intervals. The fluorescence ratio versus pH was calibrated by equilibrating the cells in K ϩ -rich medium (140 mM KCl, 10 mM NaCl, 5 mM glucose, 1 mM CaCl 2 , 1 mM MgCl 2 , 10 mM HEPES, pH 7.4) adjusted to varying pH values (ranging from 7.3 to 5.7) containing 10 M of nigericin and monensin. Endosomes were defined as regions of interest, and pixel intensity values were determined for resting endosomes and defined calibration pH acquisitions following background subtraction for both 490 (pH-dependent) and 440 nm (pH-independent). Calibration curves of the fluorescence ratio (490:440) were then plotted against pH, and endosomal pH values were extrapolated from the curve.
To study the localization of ClC-4 and ClC-5 proteins relative to transferrin-positive compartments, cells were serum-starved for 1 h at 37°C in free serum medium (Invitrogen) and then loaded with 50 g/ml transferrin conjugated to iron and tetramethylrhodamine (transferrin-Fe 2ϩ -Rhd, Molecular Probes, Leiden, The Netherlands) for 60 min. Following fixation, cells were labeled with ClC-4 or ClC-5 antibodies for subsequent immunolocalization. For kinetic analyses of transferrin internalization in control and ClC-4-and ClC-5-depleted (antisensetreated) cells, cells were serum-starved as above and then pulse-labeled for 2.5, 5, 10, 20, 40, and 60 min with 50 g/ml transferrin-Fe 2ϩ -Rhd. To examine recycling, cells were pulse-labeled with 50 g/ml transferrin-Fe 2ϩ -Rhd for 1 h and then washed and chase for 5, 20, 40, 60, and 80 min in medium containing 10% fetal calf serum. At the end of each labeling and chase interval, cells were washed and then fixed in cold 4% paraformaldehyde. Slides were viewed with a ϫ100 objective on a Carl Zeiss LSM 510 equipped with an Axiovert 100 confocal microscope.
Quantification of Immunofluorescence-GFP-transfected cells were analyzed with respect to ClC channel-specific immunofluorescence (labeled with Cy3-conjugated secondary antibodies) or rhodamine-transferrin. Although cross-talk of the fluorophores into the wrong detectors was negligible, GFP fluorescence was specifically subtracted prior to Cy3 or rhodamine immunofluorescence measurements. All images were acquired using the palette function of LSM510 confocal acquisition software, which allows verification that the fluorescence intensity is in the linear range. The linear range is defined as 0 units ϭ white and 255 units ϭ black. The Cy3 or rhodamine-specific immunofluorescence was converted to a gray scale image, and the background (determined from the mean intensity detected in cell nuclei) was subtracted prior to quantitation. Mean pixel intensity of the gray scale image was measured (Scion Image software) for the outlined region of the whole cell or for the perinuclear region (by superpositioning a box at 8 locations around the nucleus). Data were plotted, and the half-times for transferrin internalization or recycling were determined by fitting logarithmic or monoexponential decay functions, respectively, to data acquired at different time points using Origin software (Microcal Origin).
Electron Microscopy-CHO cells co-transfected with both Rab5a-  The asterisk indicates that both antisense constructs cause a significant reduction in transferrin-Fe-RhD accumulation (p Ͻ 0.0001). These data were obtained from three separate transfections for each construct.
GFP and HA-ClC-5 cDNA were fixed in 4% paraformaldehyde and 0.1% glutaraldehyde in 0.1 M phosphate buffer, pH 7.4. They were then harvested with a rubber policeman, drawn off the culture dish, and lightly centrifuged into a mm 3 pellet in a microcentrifuge tube. They were fixed for an additional 4 h and washed in PBS containing 20 mM azide and stored at 4°C until further processing. Prior to cryo-ultramicrotomy the cells were infiltrated with a solution of 10% gelatin in PBS at 37°C, allowed to gel at 4°C, and infused with 2.3 M sucrose overnight. Blocks of gelatin containing cells were attached to aluminum pins, frozen in liquid nitrogen, and cut at Ϫ95°C on a diamond knife ϳ100 nm thick using a Leica Ultracut R with a cryo-chamber (Leica Canada, Willowdale, Ontario, Canada). Sections were transferred to Formvar-coated grids in a drop of molten sucrose, and the aldehyde residues were blocked with a solution of PBS containing 0.15% glycine and 0.5% BSA. Samples were rinsed several times with PBS with just BSA prior to incubation in a polyclonal rabbit antibody against ClC-4 for 1 h. Grids were then washed in PBS/BSA thoroughly and incubated for another hour in goat anti-rabbit IgG 10-nm gold complexes (Amersham Biosciences). Again after rinses in PBS/BSA samples were incubated for 1 h in a monoclonal antibody against GFP. This was followed by more washes and another 1 h of incubation in goat anti-mouse IgG 5-nm gold complexes (Amersham Biosciences). Grids were thoroughly washed with PBS followed by distilled water. They were then stabilized in a thin film of methylcellulose containing 0.2% uranyl acetate. Controls included the omission of the primary antisera or the IgG gold complexes. Specimens were examined and images acquired with a JEOL JEM transmission electron microscope (JEOL, Peabody, MA) equipped with a digital camera (AMT Corp. Danvers, MA).
Statistics-Statistical analyses, analysis of variance followed by Bonferoni's non-paired "t" test, were conducted using Prism software, and p values of 0.05 or less were considered significant.

ClC-4 Is Expressed in the Renal Proximal Tubule-Previ-
ously, we showed that a polyclonal, affinity-purified antibody directed against the amino terminus of ClC-4 labels a 90 -97-kDa protein in immunoblots of rat brain tissue. This band was competed using the amino terminus of ClC-4, and not the amino terminus of the related protein ClC-5, attesting to its specificity (26). In Fig. 1A, we show that only this band is detected in whole mouse brain and is completely absent in immunoblots performed using brain tissue obtained from two Clcn4 null mice, providing compelling evidence in support of the specificity of this antibody. Clcn4 null mice were generated by breeding two strains of mice, Mus spretus and C57BL/6J, as described previously by Rugarli et al. (32). Clcn4 is located on the X chromosome in the M. spretus strain and on chromosome 7 in the C57BL/6J strain permitting the generation of a population of male mice devoid of Clcn4 through backcrossing C57CL/6J ϫ M. spetus F1 females with M. spretus males. To date, however, only two Clcn4 null mice have been produced during 6 months due the difficulty involved in breeding these two strains. Therefore, the following in vivo studies of ClC-4 function were performed in rats or using an antisense strategy in established intestinal and renal cell lines.
In Fig. 1B, we show that ClC-4 protein is expressed as a 90 -97-kDa protein in immunoblots of total rat kidney lysates. Confocal micrographs (Fig. 1C) reveal that ClC-4-specific immunofluorescence is expressed in the epithelium of the proximal tubule. The apical aspect of the epithelium of the proximal tubule was defined visually by in vivo injection of fluoresceinlabeled dextran (FITC-dextran), known to be internalized by this tissue via fluid phase endocytosis. As described in previously published studies (27), 8 min after injection, FITC-dextran will have been filtered through the glomerulus and internalized across the apical membrane of the epithelium of the proximal tubule by endocytosis. After this perfusion time, the kidneys were fixed in situ and flash-frozen. Confocal micrographs of frozen sections confirmed that FITC-dextran (27) can be detected close to the apical membrane and in sub-apical vesicles of the proximal renal tubule epithelium. These same sections were labeled using the ClC-4 antibody described above, and we observed that the pattern of ClC-4-specific immunofluorescence overlapped with that of FITC-dextran (Fig.  1C). These findings are similar to those reported for the related protein ClC-5 (16) and suggest that both channels are expressed in the proximal tubule, probably in the apical membrane and sub-apical membrane vesicles.
Expression of ClC-4 or ClC-5 Channels Can Be Specifically Depleted in Caco-2 Cells Using an Antisense Strategy-We used an antisense strategy to determine the relative functional expression of ClC-4 and ClC-5 endogenously expressed in epithelial cells. As both channel proteins are endogenously expressed in the Caco-2 cell line (26), we initiated our comparative studies using this intestinal, epithelial cell line. Control experiments, using cDNA coding for green fluorescent protein, revealed that the efficacy of Caco-2 cell transfection is low with ϳ10 -15% of cells being transfected. Therefore, studies of the consequences of ClC-4 and ClC-5 antisense expression require single cell assays, using green fluorescent protein expression to indicate the transfected cells. Our previously published data (26) supported the efficacy of ClC-4 antisense cDNA expression in specifically depleting ClC-4 protein. We found that antisense ClC-4 cDNA transfection caused a marked inhibition of ClC-4specific immunofluorescence (relative to control), with only a minor effect on ClC-5-specific immunofluorescence.
The confocal micrographs in Fig. 2 show that antisense ClC-5 cDNA dramatically reduces ClC-5-specific immunofluorescence ( Fig. 2A), with minor effects on ClC-4-specific immunofluorescence (Fig. 2B). These images provide qualitative evidence that antisense ClC-5 cDNA expression leads to the specific depletion of ClC-5 protein. ClC-5-specific immunofluorescence was quantitated in transfected (GFP-expressing) cells using Scion Image software (Fig. 2, bar graphs). We found that ClC-5-specific immunofluorescence was significantly reduced (41.2 Ϯ 2.8, n ϭ  (Table I). These results shown in the bar graphs and in Table I were obtained from three independent trials (different transfections) for each construct with 15-20 cells analyzed for each trial.
Having shown that antisense ClC-4 and antisense ClC-5 were capable of specifically depleting ClC-4 and ClC-5 proteins, respectively, we then assessed the efficacy of antisense transfection in depleting the functional expression of these channels. As expected for channels that traffic to and from the cell surface, the channel function of ClC-4 and ClC-5 can be measured at the plasma membrane by patch clamp electrophysiology (25,26). Expression of either ClC-4 or ClC-5 is associated with the appearance of depolarization-activated chloride-selective conductance paths (25,26). The biophysical properties of these channels are virtually identical with respect to their voltage dependence, kinetics of activation, and ionic selectivity. Furthermore, these channels cannot be distinguished pharmacologically, as both are insensitive to classic chloride channel inhibitors (25). As shown previously, Caco-2 cells endogenously express depolarization-activated chloride conductances with properties consistent with the functional expression of ClC-4 and/or ClC-5. Specific ''knock-down'' of ClC-4 using antisense cDNA was shown to reduce significantly depolarization-activated chloride-selective currents (26). In the present work we show that specific knock-down of ClC-5 also reduced these endogenous depolarization-activated currents (Fig. 3), providing quantitative evidence that both channels contribute to this function. Current density (at ϩ100 mV) in antisense ClC-5 cDNA-transfected cells was 4.4 Ϯ 0.8 (n ϭ 9) versus 9.4 Ϯ 1.5 pA/pF (n ϭ 6) in vector-transfected cells (p Ͻ 0.0125). Together with our previous studies, these findings suggest that our an- The reduction in Tfn-Rhd uptake caused by aClC-4 and aClC-5 transfection is statistically significant (*, both with p values Ͻ0.0001). These data were obtained from three separate transfections for each construct. C, analysis of the kinetics of Tfn-Rhd uptake in vector-(square), aClC-5-(triangle), and aClC-4-transfected (circle) LLCPK1 cells. Cells were fixed at various intervals after Tfn-Rhd addition (0 -60 min) for immunofluorescence quantitation using Scion-Imaging software. These data were fit using a logarithmic function using Prism software (r 2 Ͼ 0.9). D, analysis of the kinetics of Tfn-Rhd recycling in vector-(square), aClC-5-(triangle) and aClC-4-transfected (circle) LLCPK-1 cells. LLCPK-1 cells allowed to take up Tfn-Rhd for 1 h prior to initiation of chase periods of various durations: 0, 5, 20, and 40 min. Intracellular Tfn-Rhd fluorescence remaining after chase was quantitated using Scion-Imaging software. Recycling data from vector-transfected cells were fit with a monoexponential decay using Prism software (r 2 ϭ 0.92). Mean values Ϯ S.E. are shown. Tfn-Rhd recycling was delayed by 5 min in antisense-transfected cells. Eventually, after 80 min, ϳ80% of the transferrin signal has been effluxed from vector-and antisense-transfected cells (data not shown). On average, 40 -50 cells from two separate independent trials were analyzed for each time point in the pulse (C) and in the chase (D) experiments.
tisense ClC-4 and ClC-5 constructs are effective in depleting functional expression of these channels.
ClC-4 and ClC-5 Participate in Trafficking of the Transferrin Receptor-ClC-5 has been implicated in endosomal trafficking of membrane proteins (8, 10, 18 -20), and our goal is to assess the relative role of ClC-4 in this function. However, the specific trafficking events that require ClC-5 function have not been determined. Therefore, we designed experiments to assess the role of ClC-4 and ClC-5 in transferrin trafficking via the transferrin receptor, as this pathway has been studied extensively (34 -36). In Fig. 4A, we show confocal images of ClC-4 and ClC-5 localization relative to endosomes loaded with rhodamine-labeled transferrin for 60 min. At this time, transferrin is expected to label both early and recycling endosomes (34,36,37). ClC-5-specific immunofluorescence in Caco-2 cells appears as small punctate structures. Although the pattern associated with rhodamine-labeled transferrin is more extensive, most of ClC-5-positive structures overlap with those structures containing rhodamine-labeled transferrin internalized via the receptor-mediated endocytosis pathway (merged image). In Fig.  4B, we show that ClC-4-specific immunofluorescence also appears in punctate structures and most of these structures overlap with vesicular structures bearing rhodamine-transferrin (merged image). These images suggest that both ClC-5 and ClC-4 partially co-localize with transferrin-positive endosomes, including both early and recycling endosomes.
In order to determine the relative function of these channels in endosomal trafficking, we assessed whether depletion of these proteins affects transferrin receptor trafficking in Caco-2 cells (Fig. 5). Antisense cDNA was co-transfected with cDNA encoding GFP, and GFP fluorescence was used to identify transfected cells. Rhodamine-labeled transferrin uptake by antisense ClC-5 (Fig. 5A, lower panel) or antisense ClC-4-transfected cells (Fig. 5A, middle panel) was compared with that measured in vector-transfected cells (Fig. 5A, upper panel) as described under "Materials and Methods." We observed that transfection with either antisense ClC-4 (36 units Ϯ 2, n ϭ 100; p Ͻ 0.0001) or antisense ClC-5 (37.3 units Ϯ 2.5, n ϭ 74; p Ͻ 0.0001) caused significant reductions in transferrin accumulation (measured 20 min after transferrin addition to cultures) relative to control, vector-transfected cells (106.8 Ϯ 3.3 units, n ϭ 67, Table II). These results suggest that both ClC-4 and ClC-5 participate in transferrin receptor trafficking in Caco-2 cells. Furthermore, the degree of inhibition by either antisense construct was greater than would be expected if ClC-4 and ClC-5 proteins were independent (i.e. 50% inhibition, Table II). Hence, these data support a model wherein a proportion of ClC-4 and ClC-5 molecules may form a functional heteromeric complex that participates in endosomal trafficking.
The fidelity of our methods for reporting the relative rhodamine-transferrin uptake by control and antisense-transfected cells could be substantiated using a biochemical assay (Fig. 6). In this assay, biotinylated transferrin (Tfn-biotin) was added to transfected Caco-2 cell cultures, and after a 20-min interval, endocytosis was stopped, and cells were washed with weak acid to remove surface-associated Tfn-biotin. Internalized Tfn-biotin was assessed following cell lysis and analysis by SDS-PAGE. As for our measurements of Rhd-Tfn immunofluorescence, we found that antisense ClC-4 and antisense ClC-5-transfected cells accumulated significantly less Tfn-biotin than the control, vector-transfected cells, i.e. 58.5 Ϯ 4 and 69 Ϯ 0.5% of control (Fig. 6). The percent reduction caused by antisense expression in this biochemical assay is less than in the single cell immunofluorescence assay as the biochemical assay reflects uptake by both transfected and non-transfected cells. However, these biochemical data clearly support the utility and validity of our single-cell immunofluorescence assay in assessing the effect of ClC channel antisense expression on transferrin trafficking.
In order to probe the relative function of ClC-4 and ClC-5 in transferrin-mediated endocytosis and recycling in renal proximal tubule cells, we manipulated endogenous ClC-4 and ClC-5 expression in the LLPCK-1 cell line, a cell line that has been described previously (39) as a useful model of proximal renal tubule cells and shown to express both ClC-4 and ClC-5. As in the studies using Caco-2 cells, we found that there was a significant reduction in rhodamine-labeled transferrin fluorescence intensity in antisense-transfected cells relative to the vector-transfected cells after a 20-min pulse (Fig. 7A). At 20 min, transferrin accumulation was markedly decreased in ClC-4 (37.8 units Ϯ 3, n ϭ 84, p Ͻ 0.0001) and ClC-5 antisensetreated cells (41.7 units Ϯ 3, n ϭ 77, p Ͻ 0.0001) relative to control (vector-transfected) cells (103.4 units Ϯ 7.2, n ϭ 83) ( Fig. 7B and Table II).
The rate of transferrin receptor recycling to the cell surface and transferrin release was determined following initiation of the chase (replacing transferrin with serum-supplemented medium) in vector-transfected, ClC-4 or ClC-5 antisense cDNA-transfected cells (Fig. 7D). In vector-transfected LLCPK cells, greater than 50% of the transferrin recycled to the cell surface with rapid kinetics (t1 ⁄2 ϭ 2.4 min). This rapid rate of membrane recycling is consistent with measurements (t1 ⁄2 Ͻ5 min) obtained in the renal proximal tubule and certain cultured cells (34,36,40). On the other hand, there was a marked delay in rhodamine-transferrin recycling with no apparent recycling by 5 min of "chase" in ClC-4-or ClC-5-depleted cells. After this 5-min lag period, recycling in antisense-transfected cells commenced although it was still less efficient than in control cells at 40 min, with 39% transferrin recycled for both antisense ClC-4 and ClC-5 versus 72% in vector-transfected cells (p Ͻ 0.007). Overall, these studies suggest that in renal epithelial cells, ClC-4 and ClC-5 may contribute to multiple steps in transferrin receptor trafficking. However, these channels appear to play a primary role in a fast component of receptor recycling.
ClC-4 and ClC-5 Contribute to the Regulation of Endosomal pH-It has been hypothesized that ClC-5 participates in endosomal trafficking because it functions to permit acidification of this compartment (16,23). Furthermore, in comparative in vitro studies of endosomes purified from wild type and ClC-5 knockout mice, it was shown that disruption of ClC-5 expression was associated with endosomal alkalinization (18). Therefore, we evaluated the relative importance of ClC-4 expression in endosomal pH regulation. Endosomal pH was determined in situ by fluorescence ratio imaging of endosomes from single cells by preloading the endocytic compartment with the pHsensitive fluorophore FITC conjugated to transferrin. We compared the pH of endosomes from cells with depleted ClC-4 or ClC-5 expression, by antisense transfection, versus vectortransfected Caco-2 cells (Fig. 8). As depicted in Fig. 8, the endosomal pH values of both ClC-4 and ClC-5 antisense-transfected cells were significantly more alkaline than vector control endosomal pH (6.78 Ϯ 0.07 and 6.90 Ϯ 0.06 versus 6.1 Ϯ 0.11, respectively; p Ͻ 0.05 for both antisense treatments). Therefore, both ClC-4 and ClC-5 contribute to the regulation of endosomal pH in living cells.
ClC-4 and ClC-5 Can Be Co-immunoprecipitated and Colocalize in Endosomes Expressing Rab5-The dominant effect of disrupting either ClC-4 or ClC-5 by antisense transfection on rhodamine-transferrin trafficking suggests that these two channels may be capable of forming a complex. To test this hypothesis we assessed whether the two channel-forming proteins could be co-immunoprecipitated. We found that our ClC-4 polyclonal antibody (but not preimmune sera) can co-immunoprecipitate ClC-5 protein from whole rat kidney (Fig. 9A), providing the first biochemical evidence that these two channels can assemble in nature. In order to further evaluate this concept, we reconstituted ClC-5, engineered to possess an aminoterminal HA tag, into ClC-5-deficient CHO cells (16) (Fig. 9B,  i). ClC-4 protein is expressed endogenously in these cells and can be detected by immunoblotting as a 90 -97-kDa protein (Fig.  9B, ii). We show in Fig. 9C that the polyclonal antibody directed against ClC-4 could co-precipitate HA-ClC-5 (Fig. 9C, i) and the antibody directed against the HA epitope co-precipitated both the HA-ClC-5 and ClC-4 (Fig. 9C, ii). Hence, these studies support the notion that ClC-4 can form a complex with ClC-5. As predicted from these biochemical findings, the confocal micrographs in Fig. 9D revealed that both ClC-4 and HA-ClC-5 are expressed in similar vesicular structures in CHO cells.
It was shown previously (16) in COS-7 cells that ClC-5 protein co-localizes with Rab5a-GFP bearing early endosomes. Hence, in order to assess further whether ClC-4 and ClC-5 co-localize in early endosomal membranes, we compared HA-ClC-5 and ClC-4-specific immunofluorescence in COS-7 cells expressing Rab5a-GFP (Fig. 10A). Rab5a is a key regulator of endocytosis and mediates endosomal membrane fusion (41,42) such that the overexpression of Rab5a leads to enlarged endosomal vesicles (43). As predicted on the basis of previous studies (16), we found that a subpopulation of swollen puncta expressing Rab5a-GFP also labeled with HA-ClC-5 (blue)-specific immunofluorescence (Fig. 10A). Furthermore, of those puncta expressing both Rab5a-GFP and HA-ClC-5, a significant proportion were labeled by ClC-4-specific immunofluorescence. In addition, localization of ClC-4 in Rab5a-GFP-containing endosomal compartments was detected in electron micrographs of transfected CHO cells double-labeled with immunogold directed against ClC-4 (10 nm grains) and GFP (5 nm). Large vesicular structures, reminiscent of the swollen puncta expressing Rab5a-GFP in the previous confocal images, were also evident in electron micrographs. These swollen vesicular structures were decorated with 5 nm gold particles, supporting the notion that they are associated with the expression of Rab5a-GFP. As shown in Fig. 10B, 10 nm gold particles (indicative of ClC-4 expression) are detected on Rab5a-GFP-labeled swollen vesicles. ClC-4-specific label was associated with 32 of 41 (78%) Rab5a-positive endosomal structures. On the other hand, only 10 of 48 (20%) Rab5a-negative endosomal structures were labeled with ClC-4-specific immunogold (Table III). These studies suggest that both ClC-4 and ClC-5 are partially localized in early endosomes. DISCUSSION In this report, we show that ClC-4 participates with ClC-5 in endosomal trafficking of the transferrin receptor in renal and intestinal epithelial cells. Our kinetic analyses argue that these channels possibly contribute to multiple steps in transferrin receptor trafficking but predominantly to a rapid recy-  A, confocal images of COS-7 cells transfected with Rab5a-GFP and double-labeled with antibodies directed against ClC-4 (red) and HA-ClC-5 (blue). White arrows indicate particular punctate structures wherein Rab5a, HA-ClC-5, and ClC-4 co-localize. B, i, electron micrographs of CHO cells double-labeled with ClC-4 antibody (Ab) (10-nm immunogold grains, arrows) and anti-GFP antibody (5-nm immunogold grains, arrowheads). Both ClC-4 and Rab5a-specific immunogold label can be detected on membrane or within 20 nm of swollen endosomes. ii, labeling in A is specific as 10-nm grains are absent when ClC-4 primary antibody was omitted. Scale bars, 100 nm. cling component. A specific role for ClC-5 in endosomal recycling was not identified in previous studies of ClC-5-deficient cells or tissues and may account in part for the decreased availability of receptors and transporters observed (19). Furthermore, we show that the mechanism through which ClC-4 and ClC-5 affect trafficking likely may relate to endosomal pH regulation. Bafilomycin, a potent inhibitor of V-type ATPases, reduces both the rate of transferrin receptor internalization and recycling because it causes endosomal alkalinization (by 0.7-1.0 pH units), a modification known to be inhibitory to ligand sorting from receptors and intracellular retention of recycling receptors (22,21). The in situ measurements of endosomal pH shown in this report indicate that disruption of endogenous ClC-4 and ClC-5 expression is associated with alkalinization of this vesicular compartment by ϳ0.7 pH units. Therefore, ClC-4 and ClC-5 chloride channel depletion may be inhibitory to transferrin receptor trafficking because these channels have a primary role in endosomal pH regulation.
The pH of early endosomes (ϳ6.2-6.5 pH units) is normally acidic relative to the cytosol (pH 7.0 -7.3) (37,44), and acidification is mediated through V-type ATPases that function to pump protons into the organelle lumen. Inward chloride conduction has been shown to enhance acidification in purified endosomes by reducing the positive luminal potential against which the pump operates (45)(46)(47). Typically, the extent of organelle acidification reflects the chloride permeability of the organelle membrane (44), and our in situ pH measurements suggest that ClC-4 and ClC-5 contribute to the chloride permeability of endomembranes as disruption of their expression impairs endosomal acidification.
The biophysical properties and regulation of ClC-4 and ClC-5 currents have not been extensively characterized, but to date there is little to distinguish them. Friedrich et al. (25) reported that strongly outwardly rectifying chloride currents, sharing a NO 3 Ϫ Ͼ Cl Ϫ Ͼ Br Ϫ Ͼ I Ϫ conductance sequence, were conferred by expression of either ClC-4 or ClC-5 into HEK293 cells. Both ClC-4 and ClC-5 currents are activated by membrane depolarization to approximately ϩ40 mV and inhibited by extracellular (or luminal) acidification to pH 6.0 and pH 6.3, respectively. Furthermore, in the present report, we show that depletion of endogenous ClC-4 or ClC-5 reduces such depolarization in epithelial cells (26). One would predict on the basis of this characterization that neither of these channels would be active in endomembranes, which are typically hyperpolarized and have luminal contents with pH values of ϳ6.2-6.5. Because our assays of endosome function indicate that both channels do contribute to trafficking and pH regulation, it is likely that our understanding of the native regulation of these channels is incomplete. Therefore, further study of ClC-4 and ClC-5 chloride channel activity in situ in endomembranes of living cells is required to fully characterize the activation properties of these channels in this intracellular compartment.
Recently, Riordan and co-workers (14) showed that the two isoforms of ClC-3, ClC-3A and ClC-3B, can be co-immunoprecipitated. Our studies provide the first biochemical evidence that two different members of the ClC family of chloride channels may interact in nature. Specifically, we found that ClC-4 and ClC-5 could be co-immunoprecipitated from rat kidney lysate. Previously, heteromerization between ClC channel family members was assessed by electrophysiological methods. For example, it was shown that co-expression of ClC-1 and ClC-2 channels, each with distinct voltage-dependent activation properties, could lead to the appearance of chloride currents in Xenopus oocytes with "hybrid" gating properties (2), suggesting that heteromers could form between ClC-1 and ClC-2 to increase the functional diversity of this branch of the ClC chlo-ride channel family. Based on the similarity of their conductance and activation properties, it seems improbable that heteromerization between ClC-4 and ClC-5 would confer a distinct activity (25). If there is a functional advantage to the interaction of these two channels, it may relate to protein biosynthesis and/or stability, and these properties will be studied in detail in our future work.
The molecular basis for interaction between ClC-4 and ClC-5 remains unclear. The biochemical studies described in this report do not allow us to distinguish between a direct or indirect basis for interaction. However, in our future work, we plan to test the prediction that ClC-4 and ClC-5 interact directly using information provided by the high resolution structure of procaryotic ClC channels recently solved by x-ray diffraction (48). This structure supported previous biophysical, biochemical, and electron diffraction studies (33) and showed that ClCtype channels are dimeric, with each polypeptide forming a single pore of a double-barreled channel (48). The x-ray structure shows that the dimerization interface in the membrane is extensive, and we would predict that mutation of residues residing in the putative interfacial helices in ClC-5 would be expected to disrupt homodimerization and possibly ClC-4/ ClC-5 heterodimerization if assembly occurs at this interface.
Finally, mutations in ClCN5 are thought to cause proteinuria in Dent's disease because ClC-5 channels normally contribute to endocytosis of low molecular weight proteins from the lumen of the proximal renal tubule (8,10,16,38). The current studies show that ClC-4 also contributes to endosomal trafficking by epithelial cells of the renal proximal tubule and other absorptive epithelia and hence provide the rationale for future in vivo studies of the role of ClC-4 in renal function in health and disease. We speculate that the severity of Dent's disease may relate not only to the impact of particular mutations on the structure and function of ClC-5 but also on the consequences of those mutations on the functional expression of ClC-4.