Preferential association with ClC-3 permits sorting of ClC-4 into endosomal compartments

ClC-4 is an intracellular Cl−/H+ exchanger that is highly expressed in the brain and whose dysfunction has been linked to intellectual disability and epilepsy. Here we studied the subcellular localization of human ClC-4 in heterologous expression systems. ClC-4 is retained in the endoplasmic reticulum (ER) upon overexpression in HEK293T cells. Co-expression with distinct ClC-3 splice variants targets ClC-4 to late endosome/lysosomes (ClC-3a and ClC-3b) or recycling endosome (ClC-3c). When expressed in cultured astrocytes, ClC-4 sorted to endocytic compartments in WT cells but was retained in the ER in Clcn3−/− cells. To understand the virtual absence of ER-localized ClC-4 in WT astrocytes, we performed association studies by high-resolution clear native gel electrophoresis. Although other CLC channels and transporters form stable dimers, ClC-4 was mostly observed as monomer, with ClC-3–ClC-4 heterodimers being more stable than ClC-4 homodimers. We conclude that unique oligomerization properties of ClC-4 permit regulated targeting of ClC-4 to various endosomal compartment systems via expression of different ClC-3 splice variants.

mostly because of unclear subcellular localization of these two transport proteins.
The subcellular localization of ClC-4 was studied in multiple preparations, with conflicting results. Although Okkenhaug et al. (10) described endoplasmic reticulum (ER) 3 localization of ClC-4 upon heterologous expression in mammalian cells, endogenous ClC-4 inserts into the early endosomes and into the apical membrane of Caco-2 cells, a model system for human enterocytes (11). Moreover, ClC-4 contributes to acidification and to transferrin receptor trafficking in endosomes of CaCo-4, COS-7, and LLCPK-1 cells (12). These separate results might be due to the formation of heterodimers with other CLC transporters that target ClC-4 to intracellular compartments inaccessible to homodimers. Earlier reports have already described heterodimerization of ClC-4 with ClC-3 or ClC-5 (11,13). ClC-5 targets to recycling endosomes (14), and alternative splicing of ClC-3 results in various localizations of different variants, including recycling endosomes (15), the Golgi (16), synaptic-like microvesicles (17), and synaptic vesicles (5,18).
Here we investigated the subcellular localization of ClC-4 after expression in cultured mammalian cells. We observed targeting of ClC-4 into diverse endosomal pathways by forming heterodimeric assemblies with ClC-3 that are more stable than ClC-4 homodimers.

Identification of sequence signals responsible for the subcellular targeting of ClC-4
Alternative splicing of Clcn4 was reported to result in translation of two distinct ClC-4 isoforms that differ in the aminoterminal end of the protein (12), with a shorter splice variant (687 amino acids) lacking ϳ50% of helix A of ClC-4. Although expression of the long isoform as a fluorescent fusion protein in HEK293T cells generates robust whole-cell fluorescence levels and currents (amplitude at ϩ175 mV, 1.20 Ϯ 0.4 nA, n ϭ 5) with defined time and voltage dependences in whole-cell patch clamp experiments (1,2,19), we observed significantly reduced expression levels and absent whole-cell currents in cells expressing the short isoform (at ϩ175 mV, 0.03 Ϯ 0.04 n〈, n ϭ 11). Because all functional CLC channels and transporters exhibit an intact helix A and because we were unable to express the short ClC-4 isoform at comparable protein levels as other constructs, we restricted ourselves to the long ClC-4 in the following analyses.

ClC-3 targets ClC-4 into endosomal compartments
nor ClC-4 ClC-3 exhibit capacitive currents with amplitudes corresponding to ClC-3 (2). Differences in the subcellular distribution are reflected by distinct macroscopic current amplitudes (at ϩ175 mV, ClC-4 ϭ 1.20 Ϯ 0.04 nA, n ϭ 5, ClC-3 ClC-4 ϭ 1.09 Ϯ 0.4 nA, n ϭ 5, ClC-3 ClC-4 ClC-3 ϭ 1.00 Ϯ 0.35 nA, n ϭ 5, ClC-4 ClC-3 ϭ 3.22 Ϯ 0.9 nA, n ϭ 5; Fig. 1, F and G). Whole-cell currents exclusively originate from transporters in the surface membrane and thus report on changes in the numbers of transporter with this localization. Because altered expression levels also have a direct impact on transporter numbers in all subcellular compartments, whole-cell currents do not only reflect the ability of the expressed transporter to insert into the plasma membrane and in the individual transport rates but also depend on protein expression levels, which are often affected by alterations in the primary sequence. Changes in whole-cell currents thus only provide a qualitative measure for surface membrane insertion of the tested transporter chimeras. However, the subcellular localization of WT ClC-3a prevents measurable surface insertion under all expression levels obtained so far. Thus, current amplitudes exceeding background currents are a clear indication for a change in subcellular localization in a chimera approach using ClC-3a and ClC-4. Transplanting the carboxyl terminus of ClC-3 (ClC-4 ClC-3 ) increases surface membrane insertion (Fig. 1C) and macroscopic current amplitudes (Fig. 1F) to values above WT ClC-4 ( Fig. 1, F and G). In agreement with their subcellular localiza-    and the corresponding protein topology pictures to show part of the protein sequence that was substituted (green). For confocal images, cells were either cotransfected with the ER marker calnexin or the lysosomal marker LAMP1. Scale bars ϭ 10 m. E, box plot of Manders coefficients of the co-localization analysis for ClC-3 chimeric proteins and corresponding intracellular marker (ClC-3a/LAMP1, n ϭ 8; ClC-3/Calnexin, n ϭ 7; ClC-4 ClC-3/LAMP1, n ϭ 6; ClC-3 ClC-4 /Calnexin, n ϭ 11; ClC-4 ClC-3 ClC-4 /Calnexin, n ϭ 5), with n being the number of fields of view containing at least three to five co-transfected cells. All ClC-3a chimeric constructs show MC values that were significantly different from those measured in ClC-3a. F, representative whole-cell recordings from HEK293T cells expressing ClC-3 or its mutant variants. G, voltage dependence of mean current amplitudes for cells expressing WT or various ClC constructs obtained from recordings as shown in F (ClC-3a, n ϭ 4; ClC-4 ClC-3, n ϭ 8; ClC-3 ClC-4 , n ϭ 4; ClC-4 ClC-3 ClC-4 , n ϭ 6). Values are given as means Ϯ S.D., with n giving the number of cells. ***, p Ͻ 0.001, t test.

, C and E-G).
Exchanging the ClC-3 amino terminus with its ClC-4 counterpart removes a clathrin-binding dileucine motif (20,21) so that ClC-4 ClC-3 can insert into the surface membrane ( Fig. 2, B, F, and G). Taken together, these results indicate that ClC-4 lacks the endosomal targeting sequences of ClC-3 and that a carboxyl-terminal ER retention signal is responsible for the subcellular localization of ClC-4.
Removal of this signal resulted in ER exit and surface membrane insertion of ClC-4, increased the macroscopic current amplitudes (at ϩ175 mV, ClC-4 ClC-3(Linker) ϭ 2.73 Ϯ 0.70 nA, n ϭ 5), and its insertion into ClC-3 promote ER retention (data not shown). We conclude that the CBS linker region contains the ER retention signal of ClC-4.

ClC-4 exhibits unique oligomerization behavior
Co-expression studies in mammalian cell cultures (Fig. 4) as well as the experiments in cultured astrocytes (Fig. 5) reveal an

ClC-3 targets ClC-4 into endosomal compartments
almost complete change in subcellular localization of ClC-4 in the presence of ClC-3. Although HEK293T cells expressing ClC-4 alone show prominent ER staining, there is virtually no ER fluorescence staining in cells co-expressing ClC-4 together with various ClC-3 isoforms. This is at first glance surprising. CLC channels and transporters are dimers, and co-expression of ClC-3 and ClC-4 is expected to result in the appearance of ClC-3 and ClC-4 homodimeric as well as ClC-3-ClC-4 heterodimeric transporters. In the case of equal association probabilities, homo-and heterodimeric transporters will be binomially distributed, and comparable expression levels of ClC-3 and ClC-4 in co-transfected HEK293T cells thus predict 25% ClC-4

ClC-3 targets ClC-4 into endosomal compartments
homodimers. The absence of ER staining in co-transfected HEK293T cells thus suggests the preferential association of ClC-4 with ClC-3.
To test for differences in homo-and heterodimeric assembly rates, we studied oligomerization by high-resolution clear native gel electrophoresis (hrCNE) (28,29). The similarity of molecular masses makes the distinction of ClC-3b (91 kDa) and ClC-4 (84 kDa) by gel electrophoresis difficult. We therefore increased the size of fluorescently tagged ClC-3 by covalently attaching a maltose-binding protein (MBP) moiety (42.5 kDa) to the amino terminus. Subcellular distribution and transport properties of MBP-ClC-3b were assessed by confocal microscopy and by patch clamp recordings and found not to be affected by the maneuver (data not shown). ClC-3 and ClC-4 are both glycosylated (30), and the co-existence of multiple glycosylation states causes rather unfocused protein bands, making assignment of distinct protein conformations in hrCNE difficult. To reduce protein glycosylation, we mutated the asparagines at positions Asn-880 and Asn-883 in MBP-ClC-3b and at Asn-428 and Asn-431 in ClC-4 to glutamine.
Transporters were resolved by SDS-PAGE and hrCNE from whole-cell lysates and visualized by scanning the gels after expressing MBP-ClC-3b-eGFP or ClC-4-eGFP either individually or together in mammalian cells (Fig. 6). SDS-PAGE was used to compare relative expression levels, indicating similar expression levels of MBP-ClC-3b and ClC-4 under all tested conditions (Fig. 6A). When expressed alone, ClC-3 shows two distinct bands in hrCNE (Fig. 6B), a monomeric and a homodimeric band. In contrast, we observed a predominantly monomeric band with a faint homodimeric band in experiments expressing ClC-4 alone. In co-expression experiments, the formation of heterodimers produces a yellow band because of overlapping of ClC-4-mCherry and MBP-ClC-3-eGFP fluorescences (Fig. 6B).
A limitation of native gel electrophoresis in determining the native oligomeric state of proteins and in quantifying the stability of such oligomers is that solubilization steps might dissociate protein-protein interactions that exist in native membranes. We tested subunit association of the muscle chloride channel ClC-1 (31) as a control and resolved YFP-ClC-1 as a single dimer band in hrCNE (Fig. 6B) (32,33). This result demonstrates that solubilization and gel electrophoresis can conserve CLC dimers and that the different observed monomer/ dimer ratios are due to isoform-specific properties. However, it appears possible that the percentage of ClC-4 monomers is lower in native systems than after solubilization.

Discussion
ClC-3 and ClC-4 are two closely related intracellular chloride/proton exchangers that co-exist in neurons, glia, muscle, heart, and epithelial cells (3,5,7,36,37). Both mediate outwardly rectifying transport currents with significant currents only at very positive potentials (1,2,15,19,38,39). ClC-3 and ClC-4 both perform incomplete transport cycles that result in capacitive currents (2); however, there are significant differences in the relative number of complete and incomplete cycles. Although ClC-3 mainly performs such non-transporting cycles, ClC-4 is an effective coupled transporter with more than 10-fold larger transport efficiency. Despite its apparently low transport efficiency, genetic ablation of ClC-3 results in pronounced neurodegeneration in Clcn3 Ϫ/Ϫ animals (5-7). The consequences of ClC-4 ablation are rather benign (40), possibly because of the existence of functionally equivalent CLC transporters that are able to compensate for lacking ClC-4 transporters.
Whereas homodimers of various Clcn3 splice variants mediate chloride/proton exchange in the Golgi, recycling early and late endosomes in Clcn4 Ϫ/Ϫ cells, ClC-4 will remain in the ER of Clcn3 Ϫ/Ϫ cells. Our results suggest that expression of separate ClC-3 splice variants or other CLC transporters will result in cell type-specific trafficking and localization of ClC-4 (1, 10 -12).
Although predominantly localized to the endoplasmic reticulum, a fraction of homodimeric ClC-4 also inserts into the surface membrane in heterologous systems (1,2,15,19,38,39). In native cells, homodimerization and/or heterodimerization with ClC-3c may permit insertion into the plasma membrane (11). In contrast, other ClC-3 splice variants, such as ClC-3a and ClC-3b, are exclusively present in late endosomal vesicles without any measurable number of transporters in the surface membrane (15). We took advantage of distinct intracellular trafficking of ClC-3a and ClC-4 in a chimera approach to elucidate molecular determinants of intracellular targeting (Figs. 1  and 2). Despite measurable surface membrane insertion of ClC-3c and ClC-4 transporters, the main physiological function of ClC-3 and ClC-4 is assumed to be transport across intracellular membranes (41).
CLC channels and transporters are generally assumed to form dimers in which each subunit functions independently of its counterpart (42,43). In the Escherichia coli ClC, mutations have been identified that weaken the dimerization equilibrium and result in the formation of stable and functional monomers that are in dynamic equilibrium with dimeric transporters (44,45). Our experiments demonstrate that ClC-4 does not form stable dimers, likely resembling the behavior of such engineered monomeric CLCs. We compared relative amounts of homo-and heterodimeric ClC-3 and ClC-4 by hrCNE PAGE and found that ClC-4 is a rather unstable dimer. Comparison of the intracellular chloride/proton exchangers ClC-3 and ClC-4 with the muscle chloride channel ClC-1 illustrates a much

ClC-3 targets ClC-4 into endosomal compartments
higher stability of ClC-1 dimers. This may prevent the formation of ClC-1-ClC-4 heterodimers in skeletal muscle. ClC-4 is more likely to oligomerize with ClC-3 than with another ClC-4 subunit (Fig. 6). This particular property of CLC transporters provides a potential biological significance of the co-existence and the different transport efficiencies of ClC-3 and ClC-4. In the absence of ClC-3, ClC-4 either remains monomeric or forms homodimers that are retained in the endoplasmic reticulum. With rising numbers of ClC-3, the number of ClC-3-ClC-4 in distinct target compartments will increase, resulting in significantly enhanced chloride/proton exchange in this particular cell organelle. However, at a certain expression level, homodimeric ClC-3 will start to dominate. At this point, chloride/proton transport will decrease and reach significantly lower transport rates at very high numbers of ClC-3 subunits. Because alternative splicing results in the translation of different ClC-3 isoforms with separate intracellular targeting (16), this mechanism will also permit regulation of the ClC-4 distribution between recycling endosomes, lysosome, synaptic vesicle, and Golgi by varying ClC-3 isoform expression.
Ion channels and transporters are usually assumed to exist as stable oligomers that are formed immediately after translation. In the case of the superfamily of voltage-gated cation channels, the subunits jointly form a central ion conduction pathway and are unstable as a monomer. The double-barreled architecture permits folding and function of monomeric CLC channels and transporters, and this feature is also the basis for dynamic equilibria between monomers and dimers (45).
Here we report the physiological consequences of an extreme example of such dynamic oligomerization. ClC-4 cannot form stable homodimers and exists therefore mainly in heterodimeric assemblies with ClC-3 that dictate its subcellular localization.
For confocal images or cellular electrophysiology experiments, we transfected 1-2 g of plasmid DNAs (CLC transporters and their chimeric proteins) into HEK293T cells either alone or in combination with fluorescent LAMP1 (a gift from Walther Mothes, Addgene plasmid 1817) (47), Rab7 (a gift from Richard Pagano, Addgene plasmid 12605) (48), or Bcl-2-Cb5 (a gift from Clark Distelhorst, Addgene plasmid 1800) (49) (0.1-0.2 g of DNA encoding intracellular marker proteins) either using Lipofectamine 2000 (Invitrogen) or the calcium phos-phate method. 24 -36 h post-transfection, cells were used for experiments. For each construct, two independent recombinants from the same transformation were examined and shown to exhibit indistinguishable functional properties For high-resolution clear native electrophoresis, we increased the molecular weight of ClC-3b by adding the coding region of MBP in-frame to its 3Ј end using an overlapping PCR strategy. Point mutations N880Q and N883Q for MBP-ClC-3b and N428Q and N431Q for ClC-4 were inserted using an overlapping PCR strategy as well. We additionally attached a His 10 tag sequence at the 5Ј ends of the coding sequences of MBP-ClC-3b-eGFP as well as of ClC-4-mCherry for immobilized metal affinity chromatography purification when needed. HEK293T cells were transfected with 5 g of FsY1.1 G.W./ His 10 -MBP-ClC-3b-eGFP or 5 g of FsY1.1 G.W./His 10 -ClC-4-eGFP alone or co-transfected with the two DNAs using the calcium phosphate method. In another set of experiments, 5 g of FsY1.1 G.W./His 10 -MBP-ClC-3b-eGFP and 5 g of FsY1.1 G.W./His 10 -ClC-4-mCherry were co-transfected to show the heterodimer in yellow because the fluorescences of eGFP and mCherry overlap. As a control, 4 g of pSVL/mYFP hClC-1 was transfected. All experiments were performed 24 h after transfection. Normalized fluorescence intensities were calculated from co-expression experiments by dividing the fluorescence intensity of the ClC-3-eGFP homodimer, the ClC-4-eGFP homodimer, or the ClC-3-ClC-4 heterodimer band by the total fluorescence intensity measured in the gel lane.

Preparation of WT and Clcn3 ؊/؊ astrocytes
Astrocytes were prepared from WT or Clcn3 Ϫ/Ϫ (5) litter mates at postnatal day 1 after killing by decapitation. Hippocampi were dissected and mechanically dissociated to produce single cells by passing it through a 100-m pore mesh. The cell suspension was centrifuged at 1500 ϫ g for 10 min and resuspended in DMEM containing 10% FBS, 100 units/ml penicillin, and 100 g/ml streptomycin (DM10 medium). The cells were then plated in 5-ml flasks precoated with 0.5 mg/ml of collagen, rat tail type II, and then cultured at 37°C in a 95% air/5% CO 2 incubator. After growing to confluence (7-8 days), astrocytes were detached using trypsin-EDTA and collected by pelleting down using centrifugation (1500 ϫ g for 10 min). Cell pellets were resuspended by careful pipetting using DM10 medium, plated on glass coverslips, and cultured at 37°C in a 95% air/5% CO 2 incubator for 5 days more. One day after plating, cells were transduced using lentiviral particles and used for imaging 3-4 days later.

ClC-3 targets ClC-4 into endosomal compartments
essential amino acids, and 100 mM GlutaMAX). 48 h later the virus was harvested, filtered (0.2-m PVDF membrane, Millipore), and concentrated by ultracentrifugation. The viral particles were immediately frozen and stored at Ϫ80°C. Lentiviral particles were added directly to the medium containing primary astrocytes.

Confocal imaging and co-localization analysis
Images were acquired with a Leica TCS SP5 II inverted microscope (Leica, Wetzlar, Germany) using a ϫ63 oil immersion objective from living cells in PBS containing Ca 2ϩ and Mg 2ϩ (Gibco) at room temperature (22-24°C). The plasma membrane marker CellMask TM (Invitrogen) was added to the culture medium before imaging according to the instructions of the manufacturer. eGPF, mRFP, or mCherry fluorophores were excited with a 488-nm argon laser, mRFP or mCherry with a 594-nm helium-neon laser, and CellMask TM at 633 nm. Emission signals were detected after filtering with 500-to 550-nm, 600-to 640-nm, and 640-to 700-nm bandpass filters, respectively. Confocal images were assembled for publication in ImageJ 1.44p software (National Institutes of Health, Bethesda, MD) (54).
Co-localization was quantified by calculating the MC using the JACoP plugin (55) embedded in the ImageJ 1.44p software (National Institutes of Health) (54). At least five different fields of view (containing at least three co-transfected cells each) for each construct from at least two independent co-transfections were analyzed. For all tested constructs, transfected cells exhibit rather homogenous levels of whole-cell fluorescence intensities, and we excluded cells from analysis that displayed either very low or very high expression levels upon visual inspection.
In experiments testing for ER localization of chimeric transporters, we co-expressed either calnexin-mCherry (56) or Bcl-2-eGFP (49), depending on the fluorescent protein to which the tested transporter was linked. We evaluated the equivalence of the marker proteins in co-expression experiments of calnexin-mCherry and eGFP-Bcl-2. Such experiments demonstrated perfect co-localization (MC ϭ 0.99 Ϯ 0.002, n ϭ 8).

High-resolution clear native electrophoresis
Transfected HEK293T cells were washed with ice-cold phosphate-buffered saline and lysed in a buffer containing 0.1 M sodium phosphate (pH 8.0), 0.5% digitonin, protease inhibitors, and 20 mM iodoacetamide for 15 min and transferred into a reaction tube. After a clear spin at 4°C, an aliquot of the resulting whole-cell lysate (approximately 10 g) was loaded on a native gel. 4 -14% acrylamide gradient gels were prepared as described previously (29,30). The anode buffer contained 25 mM imidazole/HCl (pH 7.0) and cathode buffer (50 mM Tricine and 7.5 mM imidazole (pH 7.0)). Cathode buffer was supplemented with the anionic detergent sodium deoxycholate (0.04%) and the non-ionic detergent dodecyl maltoside (0.008%) (30). Gels were run in the cold (8°C), and the original voltage was set to 100 V. After 1 h, the voltage was raised to 150 V. The electrophoresis was stopped after an additional 2 h. Gels were scanned on a fluorescence gel scanner (Typhoon FLA 9500, GE Healthcare) at 100-m resolution. eGFP and mYFP were excited at 473 nm, and their emission was recorded using a 530/20 bandpass filter. mCherry was excited at 532 nm, and the emission was recorded using a 575-nm long pass filter. Gel images were quantified using Fiji software as described previously (57). The two channel recordings were analyzed and quantified separately using the built-in tools for gel analysis.

Data analysis
Data analysis was performed using a combination of FitMaster (HEKA), Origin (OriginLab), SigmaPlot (Systat Software), and Excel (Microsoft) software. All data are presented as mean Ϯ S.D., and statistical analysis was done by Student's and Mann-Whitney rank-sum test, with significance levels as follows: *, p Ͻ 0.05; **, p Ͻ 0.01; ***, p Ͻ 0.001.