Sorting Motifs of the Endosomal/Lysosomal CLC Chloride Transporters*

The CLC protein family contains plasma membrane chloride channels and the intracellular chloride-proton exchangers ClC-3–7. The latter proteins mainly reside on the various compartments of the endosomal-lysosomal system where they are involved in the luminal acidification or chloride accumulation. Although their partially overlapping subcellular distribution has been studied extensively, little is known about their targeting mechanism. In a comprehensive study we now performed pulldown experiments to systematically map the differential binding of adaptor proteins of the endosomal sorting machinery (adaptor proteins and GGAs (Golgi-localized, γ-ear containing, Arf binding)) as well as clathrin to the cytosolic regions of the intracellular CLCs. The resulting interaction pattern fitted well to the known subcellular localizations of the CLCs. By mutating potential sorting motifs, we could locate almost all binding sites, including one already known for ClC-3 and several new motifs for ClC-5, -6, and -7. The impact of the identified binding sites on the subcellular localization of CLC transporters was determined by heterologous expression of mutants. Surprisingly, some vesicular CLCs retained their localization after disruption of interaction sites. However, ClC-7 could be partially shifted from lysosomes to the plasma membrane by combined mutation of N-terminal sorting motifs. The localization of its β-subunit, Ostm1, was determined by that of ClC-7. Ostm1 was not capable of redirecting ClC-7 to lysosomes.

The CLC 2 family of chloride transport proteins comprises nine members in mammals (1). Although four of these are plasma membrane-residing chloride channels, the other five, ClC-3-7, localize to distinct, yet partially overlapping compartments of the endo-lysosomal pathway and to other specialized vesicles of the late biosynthetic pathway like synaptic vesicles (1)(2)(3)(4)(5). On these organelles, they are involved in enabling luminal acidification and/or chloride accumulation (6 -10). Despite the pivotal role of CLC proteins in endo-lysosomal function and their involvement in diverse pathologies in mouse models and human genetic disease (1), little is known about the sorting steps by which they reach their subcellular destinations.
Sorting of endo-lysosomal transmembrane proteins is usually mediated by cytosolic motifs that are recognized by adaptor proteins. These recruit further components of the protein transport machinery, such as clathrin (11,12). The dileucine motif (DE)XXXL(LI) and the tyrosine-based motif YXX⌽ (with ⌽ being a bulky, hydrophobic amino acid) are recognized by the adaptor protein (AP) complexes AP-1-4. The cell uses different APs to recruit cargo proteins for specific transport routes such as AP-2 for endocytosis from the plasma membrane, AP-1 for transport between the trans-Golgi network and early endosomes, and AP-3 for sorting from early endosomes and the trans-Golgi network to late endosomes (13). The role of AP-4 is less understood. Another sorting signal, which is recognized by adaptors of the Golgi-localized, ␥-ear containing, Arf-binding (GGA) family, is the dileucine motif DXXLL (14). The three mammalian GGAs mediate sorting at the trans-Golgi network by binding to the DXXLL motif of cargo and recruiting clathrin for transport to early or late endosomes. It is unclear whether different GGAs (GGA1, -2, -3) are involved in different sorting events (15).
For the predominantly plasma membrane-localized ClC-2, two sorting motifs have been identified, a dileucine motif in the C terminus that recruits AP-1B for sorting to the basolateral membrane of epithelial cells (16) and a tyrosine-based motif in a cytoplasmic loop between helices D and E that mediates recycling between endosomes and the plasma membrane (17). Barttin, the ␤-subunit of ClC-K channels at the plasma membrane, carries an amino acid sequence (PQPPYVRL) likely to be involved in endocytosis (18). However, it is not clear whether the critical tyrosine belongs to a functional YXX⌽ sorting motif or the sequence functions as a so-called PY motif for ubiquitylation-regulated internalization as in the case of the sodium channel ENaC (19). So far no "conventional" AP or GGA binding motifs have been described for intracellular CLCs. ClC-3, which localizes to endocytic compartments (2, 4, 10, 20 -23) as well as to synaptic vesicles and synaptic-like microvesicles (2,3,24), has been shown to interact via its N terminus with AP-1, AP-2, and clathrin (23). The interaction with clathrin was shown to be dependent on an acidic amino acid stretch with two dileucines, whereas that with AP-2 did not require this stretch. Targeting of ClC-3 to synaptic vesicles and synapticlike microvesicles has been reported to require AP-3 (3). Various motifs different from AP or GGA interaction sequences have been shown to be important for the subcellular localization of endo-lysosomal CLCs. A splice variant of ClC-3, ClC-3B, exhibits a PDZ binding motif at its extreme C terminus (25) that recruits ClC-3B to the Golgi complex via an interaction with the PDZ domain-carrying Golgi protein GOPC (26). For ClC-4, which has mostly been reported to localize to endosomes (4,27) but also to the endoplasmic reticulum (ER) (28), an N-terminal amino acid stretch is reportedly involved in ER retention when expressed heterologously (28). The predominantly endosomal ClC-5 (4, 27, 29 -32), a small portion of which is endogenously found also in the plasma membrane (30,33), bears a C-terminal-located PY motif (34). Although not required for in vivo ClC-5 function (33), studies in Xenopus oocytes and cultured opossum kidney cells revealed that ubiquitin ligases bind this motif and ubiquitylate ClC-5, stimulating its internalization from the cell surface (34,35). Although endogenous ClC-6 localizes to late endosomes (36), heterologously expressed ClC-6 has been found on early and recycling endosomes (37). A basic amino acid stretch in ClC-6 seems to be involved in the recruitment of ClC-6 into detergent-resistant membranes that may play a role in its subcellular localization (37). For ClC-7 or its ␤-subunit Ostm1 (38), no motifs responsible for their lysosomal localization (38 -40) have been reported so far. ClC-7 is targeted to lysosomes in the absence of Ostm1, whereas Ostm1 requires ClC-7 to be exported from the ER (38). The aim of this study was to systematically identify the cytosolic sorting motifs of all endosomal/lysosomal CLCs and to investigate their role in subcellular sorting.
To express hClC-5 in Xenopus oocytes, the cDNA was cloned into pTLN (44). An HA epitope was inserted between amino acids Glu 107 and Val 108 (extra-cytosolic loop between helices B and C) or at the C terminus by PCR mutagenesis. Point mutations were introduced by PCR with primers carrying the respective mutation. All constructs were confirmed by sequencing the complete ORF.
GST Pulldown Assays-GST fusion proteins were expressed in Escherichia coli (BL21, DE3) for 5-6 h at 25°C after induction with 0.12 mg/ml isopropyl-␤-D-thiogalactopyranoside before pelleting the cells by centrifugation at 5000 ϫ g. Cells were lysed by sonification in PBS supplemented with 0.5 mg/ml AEBSF (Roche Applied Science), protease inhibitor mixture (Complete, Roche Applied Science), and lysozyme (Sigma) and subsequent incubation with 1% (w/v) Triton X-100 on ice. GST fusion proteins were affinity-purified from a 20,000 ϫ g supernatant by a 2-h incubation with glutathione-Sepharose (GE Healthcare) under constant agitation at 4°C and subsequent washing with PBS. Purity and concentration was estimated by Coomassie staining after SDS-PAGE with BSA as standard.
For a single pulldown experiment, 1.0 -1.5 ml of lysate of a 10-cm dish of confluent HeLa cells in PBS supplemented with 1% (w/v) Triton X-100, 0.5 mg/ml AEBSF (Pefabloc SC), protease inhibitor mixture (Complete), and 1 mM Na 3 VO 4 was centrifuged at 10,000 ϫ g for 10 min, and the supernatant was incubated with roughly 50 g of GST fusion protein coupled to Sepharose for 2 h under constant agitation at 4°C. After 4 washes with PBS, supplemented with 0.1% (w/v) Triton X-100, 0.5 mg/ml AEBSF, and protease inhibitor mixture (Complete), bound protein was eluted by incubation in SDS sample buffer at 55°C for 15 min. After sedimenting the Sepharose beads, the eluate was separated by SDS-PAGE and probed by immunoblot. Pulldown assays from mouse brain or kidney lysate were performed equivalently, with PBS replaced by HEPES-buffered saline. For tissue lysate preparation, two brains or four kidneys of WT C57Bl/6 mice were homogenized in 10 ml of HEPESbuffered saline with 0.5 mg/ml AEBSF, protease inhibitor mixture (Complete), and 1 mM Na 3 VO 4 . The supernatant of a 10-min centrifugation at 10,000 ϫ g was supplemented with 1% final Triton X-100. 1.5 ml of the supernatant of 2 further centrifugation steps (15 min at 10,000 ϫ g and 20 min at 50,000 ϫ g) was used per pulldown as described above.
Expression in Cell Culture and Fluorescence Microscopy-Plasmid DNA encoding the respective construct was transfected using FuGENE6 (Roche Applied Science) according to the manufacturer's instruction, and cells were grown in a humidified 5% CO 2 incubator at 37°C for further 24 -48 h before fixation with 4% paraformaldehyde in PBS for 15 min. For immunostaining, cells were incubated with 30 mM glycine in PBS for 5 min and permeabilized with 0.1% saponin in PBS for 10 min. Both primary and AlexaFluor-coupled secondary (Molecular Probes) antibodies were applied in PBS, 0.05% saponin supplemented with 3% BSA. Images were acquired with an LSM510 laser scanning confocal microscope equipped with a 63 ϫ 1.4 NA oil immersion lens (Zeiss).
Voltage Clamp Analysis and Surface Expression Assay in Xenopus laevis Oocytes-Capped cRNA was transcribed from constructs in pTLN linearized with MluI (extra-cytosolic HA) or HpaI (C-terminal HA) using the mMessage mMachine kit (Ambion) and SP6 polymerase. 20 ng of cRNA were injected into defolliculated oocytes. Oocytes were kept at 17°C in ND96 (96 mM NaCl, 2 mM KCl, 1.8 mM CaCl 2 , 1 mM MgCl 2 , 5 mM HEPES, pH 7.5) for 2 days before analysis. Two-electrode voltage clamp measurements were performed at room temperature using a TurboTec10C amplifier (npi electronic) and pClamp10 software (Molecular Devices) as described previously (34). Surface expression of HA-tagged ClC-5 protein was determined as described previously (34,45).

Binding of Endosomal Sorting Machinery to N-and C-terminal Cytosolic Domains of Endosomal/ Lysosomal CLCs-To screen for interactions of intracellular
CLCs with the endosomal/lysosomal transport machinery, we generated recombinant GST fusion proteins of their cytosolic N-and C-terminal domains (NT and CT, respectively) and of the cytosolic CT of the ClC-7 ␤-subunit Ostm1, a protein with a single transmembrane span. In addition, we included the cytosolic domains of the plasma membrane-residing channel ClC-1 and GST alone. Pulldown experiments from HeLa cell lysate were performed with equal amounts of the purified, immobilized fusion protein (supplemental Fig. S1). Immunoblotting of bound protein against subunits of all four AP complexes, GGA proteins, and clathrin revealed numerous interactions between the CLC domains and sorting machinery components (Fig. 1). Pulldown experiments from homogenates of mouse kidney and brain yielded similar binding patterns for clathrin and AP complexes (not shown). Interactions with GGAs, however, could not be probed in those tissues as our antibodies did not recognize the mouse proteins.
ClC-1 interacted through its N terminus with AP-2 ( Fig. 1A). ClC-3-NT bound clathrin, as reported previously (23). In contrast to this previous study, we did not detect binding of AP-1 or -2 to ClC-3 (Fig. 1A). Neither did ClC-3 fusion proteins bind AP-3, as would have been expected from the reported role of AP-3 in targeting ClC-3 (3). ClC-4 bound clathrin weakly with its N terminus (Fig. 1A), but no binding to AP or GGA proteins was found with either its N or C terminus (Fig. 1, A and B). ClC-5-NT pulled down AP-1 and AP-2 as well as clathrin (Fig. 1A), whereas none of the investigated proteins was bound by the ClC-5 C terminus (Fig.  1B). The N terminus of ClC-6 interacted weakly with AP-2 and more strongly with AP-3 (Fig. 1A). AP-3 also bound to the ClC-6 C terminus, albeit less strongly (Fig. 1B). Both the N and C termini of ClC-7 bound AP-2 and AP-3 ( Fig. 1, A-C), and its N terminus also bound AP-1 (Fig. 1, A and C). AP-1 and AP-4 binding to the C terminus of ClC-7 appeared weak because it was not detected in all pulldown assays (compare Figs. 1B and 2G). The N terminus of ClC-7 also pulled down GGA2 more efficiently than GGA3 (Fig. 1A) and GGA1 (not shown). The C terminus of the ClC-7 ␤-subunit Ostm1 bound AP-2 and AP-3 but not GGA2 or GGA3 (Fig. 1C). These interactions were also found when pulldown experiments were performed with lysates of murine adult fibroblasts lacking ClC-7 (39) or Ostm1 (38) (not shown), excluding a possible indirect interaction of Ostm1-CT mediated by bound ClC-7 and vice versa.
Identification of the Sorting Motifs Mediating Clathrin and Adaptor Binding-To investigate the interactions in more detail, we set out to identify the respective binding motifs. The N and C termini of CLCs contain numerous potential tyrosinebased or dileucine sorting motifs that might mediate the binding of APs or GGAs ( Fig. 2A). To test for the actual role of the potential motifs in intracellular CLCs, we replaced the tyrosines of tyrosine-based motifs and the leucines (or leucine-isoleucines) of dileucine motifs, respectively, by alanines in our GST fusion proteins and tested their impact on adaptor binding in pulldown experiments.
The N-terminal domain of ClC-3, which bound clathrin but not APs or GGAs, contains a potential tyrosine-based motif at Tyr 27 ( 27 YDDF, Fig. 2A). However, previous work (23) showed that clathrin did not bind to this motif but to a more N-terminal stretch containing acidic residues and two dileucines ( 13 LLDLLDE, Fig. 2A). Our pulldown experiments confirmed that the strong binding of ClC-3-NT to clathrin (Fig. 1A) was virtually abolished when all four leucines of this stretch were replaced by alanines (Fig. 2B).
The N terminus of ClC-6, which bound AP-2 and AP-3 ( Fig.  1A), contains three potential tyrosine-based motifs ( 48 YESL, 61 YLEV, and 76 YEAV) ( Fig. 2A). Because the crystal structure of the prokaryotic EcClC-1 protein suggests that the third motif might extend into the first membrane-spanning helix B (43), we focused on the first two motifs. Mutating Tyr 48 to Ala strongly reduced AP-2 and AP-3 binding, with a weaker effect seen with the Y61A mutant (Fig. 2D). Combining both mutations virtually abolished binding (Fig. 2D). Recently, the basic amino acid stretch 71 KKGRR ( Fig. 2A) was implicated in the localization of heterologously expressed ClC-6 to early and/or recycling endosomes, an effect that may involve lipid rafts (37). When this stretch had been mutated to 71 AAGAA, ClC-6 was trafficked to late endosomes/lysosomes (37). Introducing this mutation in our GST-ClC-6-NT seemed to affect binding of AP-2 and AP-3 only slightly (Fig. 2D).
Despite the low efficiency with which the C terminus of ClC-6 pulled down AP-3 (Fig. 1B), it displays many potential sorting motifs ( Fig. 2A). The EKEDLL 706 sequence conforms to the (DE)XXXL(LI) consensus motif for AP binding, whereas DLTLL 795 conforms to the DXXLL consensus sequence for GGA binding. Both sequences are located in the stretch between the two cystathione ␤-synthetase domains (46,47), which is particularly long in ClC-6. In this interdomain stretch there are also three tyrosine-based motifs ( 717 YPNL, 774 YAEM, and 784 YPDI). Such a motif is also found at the very end of the protein ( 866 YQTI). Another potential Tyr-based motif ( 855 YEFL) is predicted to be positioned within an ␣-helix of the cystathione ␤-synthetase 2 domain and is not conserved in mouse (where the Tyr is substituted by Asn). We, therefore, did not investigate it further. We next replaced the key tyrosines and leucines, respectively, of the other motifs by alanines. Mutants in which the extreme C-terminal 866 YQTI motif was disrupted (Y866A), either by itself or together with mutations of other potential sorting motifs, did not bind AP-3 anymore (Fig. 2E, supplemental Fig. 2A). In contrast, various mutants in which the 866 YQTI motif was preserved (Fig. 2E), even those with all of Leu 705 -Leu 706 , Tyr 717 , Tyr 774 , Tyr 784 , and Leu 794 -Leu 795 substituted with alanines (supplemental Fig. 2A), still pulled down AP-3 similar to the WT C terminus. We conclude that specific binding of AP-3 to the ClC-6 C terminus depends on the 866 YQTI motif at the extreme C terminus. In comparison to the N-terminal fusion protein, however, AP-3 binding to ClC-6-CT is weak.
The N terminus of human ClC-7 contains a DXXLL consensus sequence for GGA binding (DDELL 69 ) ( Fig. 2A) that might underlie the binding of GGA2 and GGA3 (Fig. 1A). The binding of AP-1, AP-2, and AP-3 might be mediated by the potential (DE)XXXL(LI)-type motif EAAPLL 24 and/or the tyrosinebased motif with 94 YESL (Fig. 2A). Indeed, mutating leucines Leu 68 and Leu 69 to alanines specifically abolished the interaction with the GGA proteins, leaving AP binding unaffected (Fig.  2F). Conversely, replacing Leu 23 and Leu 24 by alanines abolished the pulldown of APs but not of GGAs. We, therefore, did not examine a putative role of the 94 YESL motif in AP binding.
As expected, combined disruption of both dileucine motifs inhibited binding of both APs and GGAs (Fig. 2F). The only conventional AP binding motif in the C terminus of ClC-7 is a canonical YXX⌽ motif at Tyr 715 ( 715 YPRF 718 , Fig. 2A). However, mutating this tyrosine to alanine did not reduce the binding of AP complexes to ClC-7-CT (Fig. 2G).  Fig. 1 to the respective fusion proteins, either WT or mutants in candidate binding motifs. Eluates from GST columns and 1% (0.5% in E) input are shown as controls. For Ponceau staining showing similar amounts of bait protein see supplemental Fig. S2. B, clathrin binding to the ClC-3 N terminus is almost abolished when all four leucines of the 13 LLDLL sequence are changed to alanines (4ϫLA). C, binding of clathrin and ␤2-adaptin (AP-2) in a pulldown assay from mouse kidney lysate to the N terminus of ClC-5, a mutant in which Leu 47 -Ile 48 of the ESTWALI 48 sequence was replaced by alanines (LI47/48AA) and to mutants in which Tyr-14 of the 14 YDDF motif was changed to alanine either alone (Y14A) or in addition to the LI47/48AA mutation (Y14A/LIAA). D, shown is binding of 2 (AP-2) and 3A (AP-3) from mouse brain lysate to the N terminus of ClC-6, either WT or mutants in the 48 YESL and 61 YLEV motifs (Y48A and Y61A), the double mutant (Y48A,Y61A), and the AAGAA mutant in the previously described 71 KKGRR motif (37). E, 3A (AP-3) binding in pulldown from HeLa cell extract with the C termini of ClC-3, -4, and -5 and the ClC-6 C terminus, either WT, the Y866A mutant in the 866 YQTI motif at the extreme C terminus, a mutant (LL705/706AA) replacing both leucines of the EKEDLL 706 motif for AP binding by alanines, a similar mutant (LL794/795AA) in the DLTLL 795 consensus sequence for GGA binding, and fusion proteins containing these mutations in various combinations. F, shown is a pulldown assay from HeLa cell extract with the wild type N terminus of human ClC-7 (WT), a mutant with both leucines of EAAPLL 24 changed to alanine (LL23/24AA), a similar mutant of the DDELL 69 sequence (LL68/69AA), and a mutant combining these mutations (2ϫLLAA). G, shown is a pulldown assay from mouse brain lysate with the C terminus of hClC-7, either WT, or a mutant (Y715A) in the 715 YPRF 718 motif. H, shown is a pulldown assay from HeLa cell extract with the C terminus of Ostm1 (aa 307-338) and various fragments thereof: a membrane-proximal fragment (aa 307-328), an extreme C-terminal fragment (aa 319 -338), and mutants changing Leu 318 -Ile 319 to dialanine either in the complete C terminus (aa 307-338-LIAA) or in the membrane-proximal fragment (aa 307-328-LIAA). NOVEMBER 5, 2010 • VOLUME 285 • NUMBER 45

JOURNAL OF BIOLOGICAL CHEMISTRY 34541
Although it weakly bound AP-2 and AP-3, the cytosolic C terminus of Ostm1 does not display any consensus tyrosine-based or dileucine motifs. To narrow down the position of the binding site, we generated two overlapping constructs with the first part of the C terminus (Ser 307 -Thr 328 ) and the extreme C-terminal part (Ile 319 -Thr 338 ), respectively, both fused to GST. The sequence between Ser 307 and Thr 328 was sufficient to pull down AP-2 and AP-3 as efficiently as the full C terminus, whereas no binding to the C-terminal fragment Ile 319 -Thr 338 was detected (Fig. 2H). The interaction was not dependent on the dipeptide LI 319 (Fig. 2H).

Internalization of Heterologously Expressed ClC-5 Is Largely Independent of Its N-terminal AP-2
Binding Motif-The 14 YDDF motif in the N terminus of ClC-5 binds AP-2 (and other adaptors, Figs. 1A and 2C) and might, therefore, be involved in its endocytosis from the plasma membrane. To test this hypothesis, we introduced the Y14A mutation into a full-length ClC-5 construct and expressed it in CHO cells, which have negligible endogenous levels of ClC-5 (48). Immunofluorescence microscopy did not reveal obvious differences in the subcellular localization between WT and Y14A ClC-5 (Fig. 3A). Both proteins displayed some plasma membrane localization in addition to intracellular vesicular staining. We tested whether we could detect stronger plasma membrane staining with a mutant in the C-terminal "PY-like" internalization motif (34) (Fig. 2A). This motif mediates E3 ubiquitin ligase-mediated internalization of ClC-5 in heterologous expression systems (34,35). However, we could not observe an increase in plasma membrane expression of this Y672E mutant by immunofluorescence (Fig. 3A).
Even 2-fold increases in plasma membrane expression are difficult to detect by immunofluorescence. For a more reliable, quantitative comparison of cell surface expression between WT and mutants, we introduced an HA epitope into the extracytosolic loop between helices B and C of ClC-5 (34). As a negative control we appended an HA tag to the cytosolic C terminus. All constructs were expressed to similar extents in Xenopus oocytes (Fig. 3B). The extracellular presence of the epitope was quantified using an HA antibody in a chemiluminescence assay (34,45). These experiments revealed that the surface expression of ClC-5 Y14A and WT ClC-5 were indistinguishable, whereas the PY mutant ClC-5 Y672E showed a roughly 2-fold higher surface expression (Fig. 3C) as in our previous work (34). Likewise two-electrode voltage clamp measurements of oocytes expressing the Y672E mutant yielded about 2-fold higher currents than those expressing either WT or Y14A ClC-5 (Fig. 3D). We conclude that unlike the C-terminal PQPPYVRL 675 motif, the N-termi-nal 14 YDDF motif does not play a significant role in plasma membrane localization of ClC-5.
ClC-6 Sorting through Its Cytosolic Domains-To investigate the role of the sorting motifs identified in ClC-6, we compared the subcellular localization of full-length WT with mutant ClC-6 in transfected HeLa cells. As reported previously (37), heterologously expressed ClC-6 colocalized with the recycling-endosome marker transferrin receptor (TfR) (Fig. 4A) rather than with markers of late endosomes (not shown). This contrasts with the late endosomal localization of native ClC-6 (36). When the tyrosines of the confirmed AP3-binding 866 YQTI motif and of the 48 YESL and 61 YLEV consensus sequences were changed to alanine either alone or in combination, the localization of the resultant mutant was not altered (not shown). We, therefore, mutated all potential "classical" sorting motifs in the N and C termini of ClC-6 ( Fig.  2A) in combination. However, even this protein (ClC-6 Y48A,Y61A,LL705/ 706AA,Y717A,Y774A,Y784A,LL794/795AA,

Y866A
) was targeted to TfR-positive endosomes of transfected cells like WT ClC-6 ( Fig. 4A). We then introduced into this heavily mutated construct two other point mutations (W590A,E591A) to disrupt a potential unconventional sorting signal reported to mediate lysosomal targeting (49). These two residues are located at the end of the last intramembrane helix R and had not been included in the fusion proteins used for pulldown experiments ( Fig.  2A). However, even these additional mutations did not change the co-localization of ClC-6 with the TfR (not shown).
We finally generated chimeric proteins in which portions of ClC-6 were replaced by equivalent segments of the plasma membrane Cl Ϫ channel ClC-0 from Torpedo marmorata (42). Cytoplasmic N-terminal and C-terminal domains and the central transmembrane domain were assembled in different combinations, and their subcellular localization was determined in transfected HeLa cells (Fig. 4B). When both the N and C termini of ClC-6 were replaced in chimera 0-6-0 by those of ClC-0, predominant ER-like staining was observed in less than 50% of transfected cells. In the majority of cells the 0-6-0 chimera was strongly plasma membrane-localized just like ClC-0 itself (Fig.  4B), demonstrating the importance of the cytosolic domains for for ClC-6 (green in merge) and the TfR (red in merge). Both ClC-6 proteins colocalize strongly with the TfR (yellow). B, subcellular localization of chimeras between ClC-6 and ClC-0 in transiently transfected HeLa cells is shown. In chimeras the N terminus (first number of the name), the transmembrane region (second number), and the C-terminal domain (third number) carry the respective parts of ClC-6 (6) or ClC-0 (0). Immunostaining used antibodies directed against the N terminus of ClC-6 (6N2) or the C terminus of ClC-6 (6C3) or of ClC-0 (antibodies indicated in brackets). NOVEMBER 5, 2010 • VOLUME 285 • NUMBER 45 endosomal sorting. Unfortunately, chimeras possessing the ClC-6 C terminus and the N terminus of ClC-0 (0-6-6 and 0-0-6) did not leave the ER (Fig. 4B). The two chimeras with an N terminus of ClC-6 and the C terminus of ClC-0 (6-6-0 and 6-0-0) localized to intracellular punctate structures (Fig. 4B) where they colocalized with the TfR (not shown) just as heterologously expressed ClC-6. Hence, the N terminus of ClC-6 is sufficient for endosomal targeting. Unexpectedly, the chimera 6-0-6 displayed a perinuclear localization pattern (Fig. 4B) and colocalized with the Golgi protein GM130 (supplemental Fig.  3A). Disrupting the C-terminal 866 YQTI AP-3 binding site of ClC-6 by the Y866A mutation did not affect the apparent ER localization of 0-6-6 and 0-0-6 (not shown) nor the perinuclear localization of 6-0-6 (supplemental Fig. 3B) even when combined with mutations Y48A and Y61A, which together virtually abolished binding of AP-2 and AP-3 to the ClC-6 N terminus (Fig. 2D).

Endosomal Sorting of CLCs
Sorting Motifs Responsible for the Subcellular Localization of ClC-7/Ostm1-To investigate the role in lysosomal sorting of ClC-7 of identified AP and GGA binding motifs, we transfected HeLa cells with rat ClC-7 (rClC-7) or with chimeras between rClC-7 and ClC-0. We did not cotransfect the ␤-subunit Ostm1 because Ostm1 bound APs in our pulldown experiments and, therefore, might have confounded our results. Transfected fulllength rClC-7 nearly perfectly colocalized with the late endosomal/lysosomal marker protein LAMP-1 (Fig. 5A) as observed previously in native cells (40). When we replaced the cytosolic N-and C-terminal regions of rClC-7 by those of ClC-0, the resulting chimera 0-7-0 yielded a predominantly reticular staining pattern indicative of ER retention (Fig. 5A). However, a small proportion of 0-7-0 reached the plasma membrane. Unlike WT ClC-7, the chimera did not colocalize with LAMP-1 (Fig. 5A).
The N terminus of rClC-7 suffices to direct the plasma membrane Cl Ϫ channel to late endosomes and lysosomes, as revealed by the co-localization of 7-0-0 with LAMP-1 in transfected cells (Fig. 5B). When we disrupted in this construct the N-terminal EGAPLL 24 and DDELL 67 motifs (homologous to EAAPLL 24 and DDELL 69 in human (supplemental Fig. 4), which bind APs and GGAs, respectively (Fig. 2F)) either alone (supplemental Fig. 5) or in combination (Fig. 5B), the mutant 7-0-0 chimeras were still sorted to late endosomes/lysosomes. Additionally mutating tyrosine Tyr 92 (homologous to human Tyr 94 in the YESL 94 motif that was not involved in AP binding (Fig. 2F)) did not alter this localization (supplemental Fig. 5). Compared with human ClC-7, rClC-7 exhibits an additional (DE)XXXL(LI) consensus motif (EETPLL 37 ) that is a candidate site for AP protein binding (supplemental Fig. 4). Replacing both leucines of this motif by alanines did not affect the subcellular localization of 7-0-0, whereas the combined disruption of all three N-terminal dileucine motifs brought the 7-0-0 chimera to the plasma membrane (supplemental Fig.  5). Surprisingly, disruption of the GGA binding motif DDELL 67 was not required for this effect, as also the mutant in which only both consensus sites for AP binding were disrupted (7 LL23/24AA,LL36/37AA -0-0) reached the cell surface instead of colocalizing with LAMP-1 (Fig. 5B).  -7 N terminus. A, subcellular localization of rClC-7 (top), a chimera of rClC-7 with N-and C-terminal domains replaced by those of ClC-0 (0-7-0, below), and WT ClC-0 (bottom) after transient transfection of HeLa cells, in comparison to LAMP-1 as marker for late endosomes and lysosomes. ClC-7 colocalizes with LAMP-1, whereas ClC-0 shows plasma membrane expression. The 0-7-0 chimera shows weak plasma membrane expression in addition to strong ER-like staining. B, sorting determinants in the ClC-7 N terminus investigated in HeLa cells transfected with a chimeric protein (7-0-0) in which the N terminus of the plasma membrane channel ClC-0 was replaced by that of rClC-7 (top panel) or by ClC-7 N termini carrying two combinations of mutations in the EGAPLL 24 , EETPLL 37 , and DDELL 67 dileucine motifs (EETPLL 37 present in rat, but not humans). The N terminus of ClC-7 sufficed to target ClC-0 to lysosomes, and combined disruption of the first two motifs in 7 LL23/24AA,LL36/37AA -0 -0 resulted in cell surface localization.
The remaining partial colocalization of these mutants with LAMP-1 cannot be attributed to signals remaining in the mutated ClC-7 N terminus because these mutations completely abolished the late endosomal/lysosomal localization of the 7-0-0 chimera (Fig. 5B). Because the 0-7-0 chimera (Fig. 5A) suggests that this localization is not owed to the transmembrane part, it is probably the C terminus, which bound AP adaptors in our pulldown experiments (Fig. 1B), that provides additional cues for endosomal/lysosomal sorting. Although disruption of the only tyrosine-based consensus motif ( 715 YPRF 718 ) in the C-terminal GST fusion protein of human ClC-7 did not interfere with binding of APs (Fig. 2G), we mutated the homologous Tyr 713 and Phe 716 in full-length rClC-7. This mutant (rClC-7 YF713/716AA ) remained localized to late endosomes/lysosomes (supplemental Fig. 6). When this mutation was added on top of those combinations that already partially shifted the constructs to the cell surface, the resulting rClC-7 LL23/24AA,LL36/37AA,YF713/716AA (not shown) and rClC-7 LL23/24AA,LL36/37AA,LL66/67AA,YF713/716AA (supplemental Fig. 6) still displayed the partial colocalization with LAMP-1 in addi-tion to their presence at the plasma membrane. Obviously the YPRF motif of ClC-7 is not responsible for the apparent ability of the ClC-7 C terminus to partially direct ClC-7 to lysosomes when N-terminal lysosomal trafficking signals have been disrupted.
ClC-7 is required for ER export of its ␤-subunit, Ostm1, but ClC-7 is targeted to lysosomes even in the absence of Ostm1 (38). As we found weak binding of APs to the ClC-7 ␤-subunit Ostm1 (Figs. 1C and 2H), we wondered whether Ostm1 could support lysosomal sorting of ClC-7 mutants whose dominant lysosomal targeting sequences had been disrupted. To this end, we transiently cotransfected Ostm1 bearing a C-terminal GFP tag (Ostm1-GFP) with either WT or sorting mutants of rClC-7. Coexpression of both WT rClC-7 and rClC-7 LL23/24AA,LL36/37AA was sufficient to ensure ER export of Ostm1. In both cases Ostm1-GFP colocalized with the ClC-7 construct (Fig. 7). With WT ClC-7, Ostm1-GFP was sorted to late endosomes/lysosomes, whereas it strongly labeled the cell surface in addition to a partial lysosomal localization when cotransfected with rClC-7 LL23/24AA,LL36/37AA . Thus, ClC-7 determines the localization of Ostm1.

DISCUSSION
Despite the pivotal role of CLC Cl Ϫ /H ϩ exchangers in endosomal/lysosomal function, it has remained enigmatic how their differential localization to the various endosomal/lysosomal compartments is achieved. We used GST fusion proteins of the N-and C-terminal cytosolic domains of all intracellular CLCs to systematically test and compare their interactions with clathrin and its adaptors, AP-1-4 and GGA proteins. The resulting interaction pattern did not depend on the source of cell lysates used for the pulldown assay (HeLa cells, mouse brain, or kidney) and agreed well with the subcellular localization of the various CLC proteins (Fig. 1D). For example, AP-3, which mediates cargo sorting for transport to late endosomes, interacted specifically with late endosomal ClC-6 and lysosomal ClC-7/Ostm1, whereas it was not bound by the other CLCs, which localize to earlier endosomal compartments or the plasma membrane. On the other hand, the AP-2 adaptor complex involved in endocytosis from the plasma membrane was strongly bound by ClC-5, which cycles between endosomes and the cell surface.
The amino acid sequences of cytoplasmic CLC domains suggested the presence of consensus binding motifs. In several cases, N-or C-terminal domains displayed more than one candidate binding site. For instance, the N terminus of ClC-7 contains sites for both AP and GGA binding, and the N terminus of ClC-6 displays several sites for AP binding that may be functionally redundant to some degree. Candidate binding sites were validated experimentally by disrupting them through mutagenesis, either individually or in combination. When introduced into the respective fusion protein, these mutations often abolished or reduced binding of adaptor proteins or clathrin, thereby confirming these motifs as being functionally relevant (an overview of these motifs is given in supplemental Table  1). In some cases, however, such mutations failed to affect binding. This situation is not unusual because the candidate binding site might be sterically inaccessible or may require more amino acids than those specified in the consensus sequence. Moreover, even if a "real" binding site had been disrupted by mutagenesis, the functional consequence might be masked by an additional, unidentified binding site in the fusion protein.
Indeed, unconventional binding sites do exist, and we were unable to identify the site(s) by which the C termini of either ClC-7 or its ␤-subunit Ostm1 bound AP-2 and AP-3.
Whereas the binding of specific adaptor proteins to the various CLC transporters agreed well with their intracellular localization (Fig. 1D), it often proved difficult to demonstrate their involvement in the intracellular trafficking of CLC proteins. Even when all confirmed adaptor binding sites in the N and C termini of ClC-6 were disrupted by mutagenesis, the heavily mutated ClC-6 was still trafficked to TfR-positive recycling endosomes just like transfected WT ClC-6. One has to realize, however, that this localization is abnormal. Native ClC-6 is found in late endosomes of neurons, the only cells significantly expressing this CLC protein (36). Because the ClC-6 mRNA is rather ubiquitously expressed (50), ClC-6 might require a neuron-specific ␤-subunit for its stability similar to ClC-K Cl Ϫ channels, which are unstable without their ␤-subunit barttin (51), or like ClC-7, which needs Ostm1 (38). In contrast to Ostm1, which is not needed for the lysosomal localization of ClC-7 (38), barttin plays a crucial role in targeting ClC-K channels to the plasma membrane (18). Likewise, a so far unknown ␤-subunit for ClC-6 might traffic ClC-6 to late endosomes. If so, our study addressed a situation that is not found in vivo. Nonetheless, the present 6-0-0 chimera showed that the N terminus of ClC-6 contains endosomal targeting signals. Unfortunately, we obtained ambiguous localization results (not shown) when we disrupted the confirmed N-terminal AP binding sites in this chimeric construct. We are, therefore, unable to state with confidence that those motifs play a role in ClC-6 sorting.
The situation is much clearer with the late endosomal/lysosomal ClC-7/Ostm1 heteromer (38). Although AP-1 and AP-2 bound weakly to an unidentified binding site in its C terminus, Ostm1 did not significantly influence the localization of ClC-7 in transfected cells, agreeing with our previous work (38). Even when the disruption of lysosomal sorting signals in ClC-7 led to a partial mislocalization of the transporter to the plasma membrane, co-expression with Ostm1 did not increase the proportion of ClC-7/ Ostm1 found in late endosomes/lysosomes. Likewise, and also agreeing with our previous work (38), co-transfection with WT ClC-7 trafficked Ostm1 to late endosomes/lysosomes. Importantly, ClC-7 mutants that mislocalized to the plasma membrane carried Ostm1 to that domain as well. Hence, the subcellular localization of Ostm1 seems to depend entirely on sorting signals in ClC-7, and Ostm1 lacks an effect on ClC-7 trafficking.
Both N and C termini of ClC-7 strongly bound AP-3, an adaptor involved in trafficking to late endosomes. Subsequent transport to lysosomes does not require further sorting. The ClC-7 N terminus also bound GGA proteins, which might direct ClC-7 to early endosomes from where it would be sorted to late endosomes by AP-3. The prominent role in lysosomal sorting of the ClC-7 N terminus was revealed by a chimera in which it replaced the N terminus of the plasma membrane Cl Ϫ channel ClC-0. The resulting chimera 7-0-0 was targeted to late endosomes/lysosomes rather than to the plasma membrane. Strong lysosomal targeting signals are provided by the two dileucine AP-binding motifs present in the rat ClC-7 N terminus. When these two motifs were disrupted together, the mutated 7-0-0 chimera was found in the plasma membrane like ClC-0. Somewhat surprisingly, the GGA binding site did not seem important for lysosomal sorting. Full-length ClC-7 could be partially directed to the plasma membrane by disrupting just those two AP binding motifs. However, a large proportion of the mutant remained in late endosomes/lysosomes to which it was probably directed by its AP-3 binding C terminus. As disruption of the only conventional candidate AP binding site in the C terminus had no effect, we were unable to fully direct ClC-7 to the plasma membrane with a few point mutations. Nonetheless, the partially plasma membrane localized ClC-7 mutant that carries just four point mutations in the cytoplasmic FIGURE 7. ClC-7-dependent transport of Ostm1. Subcellular localization of LAMP-1 (magenta in merge), ClC-7 (yellow in merge) and Ostm1 (cyan in merge) in HeLa cells transiently cotransfected with Ostm1-GFP and WT rClC-7 (upper panel) or the rClC-7 LL23/24AA,LL36/37AA mutant that shows partial cell surface expression (lower panel). Although Ostm1-GFP stains an ER-like pattern in cells that do not overexpress ClC-7 strongly (asterisk), it colocalizes with ClC-7 to LAMP-1-positive late endosomes/lysosomes in rClC-7-overexpressing cells and with endosomes/lysosomes as well as the plasma membrane in cells expressing rClC-7 LL23/24AA,LL36/37AA . N terminus should prove useful for characterizing its biophysical properties.
With the notable exception of the N terminus of ClC-5, which bound AP-2 (and clathrin) to a site that we confirmed by mutagenesis, the N and C termini of ClC-3 through ClC-5 did not bind APs or GGAs in our pulldown experiments. Although AP-2 binding to ClC-5 would fit well with the assumed recycling of ClC-5 over the plasma membrane, the disruption of its binding motif did not increase its abundance in the plasma membrane. Hence, other mechanisms must operate in directing these endosomal CLCs to their respective compartments. One such mechanism may be binding to clathrin as described previously for ClC-3 (23). In addition, there might be binding sites in the cytoplasmic aspect of the membrane-spanning parts of CLC proteins, an issue we could not investigate with our pulldown experiments. Indeed, a tyrosine-based motif between intramembrane helices D and E has recently been implicated in the rapid recycling of the Cl Ϫ channel ClC-2 between the plasma membrane and an endosomal compartment (17). On the other hand, ClC-3, -4, and -5 may heterodimerize (4,52) akin to the previously described heteromer formation between plasma membrane CLC channels (44,53), and sorting signals present in one of the subunits may determine the trafficking of the heterodimer.
In summary, we newly identified several AP and GGA binding sites in the cytoplasmic parts of vesicular CLC anion/proton exchangers. The known roles of confirmed binding partners in facilitating specific sorting steps agreed well with the native subcellular localization of the CLCs they bound to. In several cases, however, these interactions are not the only ones that direct vesicular CLCs to their normal destination, because no change in localization was observed when the respective binding sites were disrupted. Those cases where vesicular CLC proteins could be directed to the plasma membrane with a few point mutations, however, should provide excellent opportunities to study their biophysical properties in detail.