A Novel Subtype of AP-1-binding Motif within the Palmitoylated trans-Golgi Network/Endosomal Accessory Protein Gadkin/γ-BAR*

Membrane traffic between the trans-Golgi network (TGN) and endosomes is mediated in part by the assembly of clathrin-AP-1 adaptor complex-coated vesicles. This process involves multiple accessory proteins that directly bind to the ear domain of AP-1γ via degenerate peptide motifs that conform to the consensus sequence ØG(P/D/E)(Ø/L/M) (with Ø being a large hydrophobic amino acid). Recently, γ-BAR (hereafter referred to as Gadkin for reasons explained below) has been identified as a novel AP-1 recruitment factor involved in AP-1-dependent endosomal trafficking of lysosomal enzymes. How precisely Gadkin interacts with membranes and with AP-1γ has remained unclear. Here we show that Gadkin is an S-palmitoylated peripheral membrane protein that lacks stable tertiary structure. S-Palmitoylation is required for the recruitment of Gadkin to TGN/endosomal membranes but not for binding to AP-1. Furthermore, we identify a novel subtype of AP-1-binding motif within Gadkin that specifically associates with the γ1-adaptin ear domain. Mutational inactivation of this novel type of motif, either alone or in combination with three more conventional AP-1γ binding peptides, causes Gadkin to mislocalize to the plasma membrane and interferes with its ability to render AP-1 brefeldin A-resistant, indicating its physiological importance. Our studies thus unravel the molecular basis for Gadkin-mediated AP-1 recruitment to TGN/endosomal membranes and identify a novel subtype of the AP-1-binding motif.


Membrane traffic between the trans-Golgi network (TGN) and endosomes is mediated in part by the assembly of clathrin-AP-1 adaptor complex-coated vesicles. This process involves multiple accessory proteins that directly bind to the ear domain of AP-1␥ via degenerate peptide motifs that conform to the consensus sequence ØG(P/D/E)(Ø/L/M) (with Ø being a large hydrophobic amino acid). Recently, ␥-BAR (hereafter referred to as Gadkin for reasons explained below) has been identified as a novel AP-1 recruitment factor involved in AP-1-dependent endosomal trafficking of lysosomal enzymes. How precisely Gadkin interacts with membranes and with AP-1␥ has remained unclear. Here we show that Gadkin is an S-palmitoylated
peripheral membrane protein that lacks stable tertiary structure. S-Palmitoylation is required for the recruitment of Gadkin to TGN/endosomal membranes but not for binding to AP-1. Furthermore, we identify a novel subtype of AP-1-binding motif within Gadkin that specifically associates with the ␥1-adaptin ear domain. Mutational inactivation of this novel type of motif, either alone or in combination with three more conventional AP-1␥ binding peptides, causes Gadkin to mislocalize to the plasma membrane and interferes with its ability to render AP-1 brefeldin A-resistant, indicating its physiological importance. Our studies thus unravel the molecular basis for Gadkin-mediated AP-1 recruitment to TGN/endosomal membranes and identify a novel subtype of the AP-1-binding motif.
Morphological as well as live cell-imaging studies have shown that AP-1 partitions between endosomal membranes and the TGN (6, 7), but how precisely such partitioning is accomplished has remained unclear. Membrane recruitment of AP-1 requires nucleotide exchange on the small GTPase Arf1, a process blocked by the fungal metabolite brefeldin A (BFA). Furthermore, assembly of AP-1 coats at the TGN is facilitated by phosphatidylinositol 4-kinase-mediated production of phosphoinositides (18), which may act as membrane-embedded cofactors together with Arf1-GTP and cargo proteins. Based on their molecular properties and intracellular localization, it seems likely that accessory proteins contribute to the specific targeting of AP-1 to the TGN or to endosomal membranes (4,14,16,17). In support of this view, it has been shown that ␥-BAR, an AP-1-binding accessory protein localized predominantly to perinuclear endosomes, can stabilize AP-1, but not the closely related GGA proteins at membranes even in the presence of BFA (8). The precise mechanism underlying this activity of ␥-BAR has not been revealed. We have recently shown that ␥-BAR directly interacts with the light chains of kinesin KIF5, thereby contributing to the peripheral movement of TGN-derived endosomal vesicles (19). Because ␥-BAR does not harbor a curvature-sensing BAR (BIN/amphiphysin/ Rvs167) domain, we propose to refer to this protein as Gadkin, for ␥1-adaptin and kinesin interactor, to avoid possible confusion with BAR domain proteins.
Here we have addressed the mechanism by which Gadkin associates with AP-1 and with TGN/endosomal membranes. We show that Gadkin is a peripheral S-palmitoylated membrane protein and that this modification is absolutely required for its recruitment to TGN/endosomal membranes in living cells. Further biochemical and biophysical studies led to the discovery of a novel subtype of AP-1-binding motif within the largely unstructured Gadkin that specifically associates with the ␥1-adaptin ear domain. Mutational inactivation of this novel type of motif either alone or in combination with three more conventional AP-1␥-binding peptides causes Gadkin to mislocalize to the cell surface and interferes with its ability to render AP-1 BFA-resistant. Our studies thus unravel the molecular basis for Gadkin-mediated AP-1 recruitment to TGN/endosomal membranes and identify a novel subtype of AP-1-binding motif.

EXPERIMENTAL PROCEDURES
Plasmids, Mutagenesis, and Antibodies-The full-length Gadkin coding sequence or truncated versions thereof were cloned into pEGFP-N (Clontech) for mammalian expression of C-terminally eGFP-tagged proteins. Vectors for the mammalian expression of N-terminally FLAG-tagged proteins were custom-made based on the pcDNA3 backbone (Invitrogen). For bacterial expression of N-terminally GST-tagged proteins we used the pGEX 4T-1 vector (Amersham Biosciences), for expression of His 6 -tagged proteins we used the pET28a(ϩ) vector. Point mutants were generated using the QuikChange II Kit (Stratagene). The presence of the mutations was verified by double strand DNA sequencing. The following antibodies were used for detection of proteins in immunoblots, in immunofluorescence, and for immunoprecipitations: monoclonal antibodies against ␥-adaptin (BD Biosciences), TfR (clone H68.4, Zymed Laboratories Inc.), clathrin heavy chain (clone TD.1), FLAG tag (M2, Sigma), glyceraldehyde-3-phosphate dehydrogenase (clone 71.1, Sigma), actin (clone AC-15, Sigma), GFP (clone 1E4, Stressgen), His tag (Novagen), and COPI (kind gift from F. T. Wieland, Heidelberg, Germany). Polyclonal antibodies against Gadkin were raised in rabbits by injecting a purified His-tagged Gadkin (amino acids 52-302). The serum was affinity-purified on cross-linked GST-tagged Gadkin (amino acids 52-302) and tested for specificity. Fluorescent dye-conjugated secondary antibodies (Alexa 488 , Alexa 568 , and Alexa 594 ) were purchased from Molecular Probes. HRP-coupled secondary antibodies were from Dianova.
Cell Culture and Transfections-HeLa, COS7, and HEK293 cells were cultured in Dulbecco's modified Eagle's medium (Invitrogen) supplemented with fetal calf serum and antibiot-ics. Cell lines were transfected with plasmid DNA using Lipo-fectamine2000 (Invitrogen).
Immunofluorescence Microscopy-Cells seeded on glass coverslips were fixed with 4% paraformaldehyde, 4% sucrose in PBS, pH 7.4, for 20 min. After washing 3 times in PBS and permeabilizing for 20 min in blocking solution (30% goat serum, 0.3% Triton X-100 in PBS), primary antibody was applied in the same buffer for 1 h at room temperature. PBSwashed coverslips were incubated with secondary antibodies in blocking solution for 1 h at room temperature. Afterward coverslips were washed again three times in PBS, mounted onto glass slides using Immumount solution (Thermo Electron) supplemented with 1 g/ml DAPI. Image acquisition was performed with a motorized Axiovert 200M inverted microscope (Carl Zeiss) equipped with a Stallion System (Intelligent Imaging Innovations). Data analysis was performed with Slidebook Software (Intelligent Imaging Innovations).
BFA Treatment-HeLa cells seeded on glass coverslips were incubated with 5 g/ml BFA for 15 min at 37°C in serumcontaining medium prior to paraformaldehyde fixation and processing for indirect immunofluorescence microscopy.
2-Bromopalmitate Treatment-HeLa cells were preincubated with 50 M 2-bromopalmitate (dissolved in EtOH) or the same volume of EtOH for 2 h before transient transfection. Afterward the 2-bromopalmitate treatment was continued for 20 h before fixation and processing for indirect immunofluorescence microscopy.
Cell Fractionation-Cells were homogenized in 20 mM HEPES, pH 7.4, 2 mM MgCl 2 , 100 mM NaCl supplemented with 1 mM phenylmethylsulfonyl fluoride and mammalian protease inhibitor mixture (Sigma) by passing 15 times through a 27-gauge needle. The homogenate was centrifuged at 700 ϫ g for 3 min to remove debris. The supernatant was centrifuged at 180,000 ϫ g. The soluble fraction was solubilized with Triton X-100 and centrifuged again at 180,000 ϫ g. The membrane pellet was rinsed twice with buffer, resuspended in buffer containing 1% Triton X-100, incubated for 15 min on ice for solubilization, and centrifuged at 20,000 ϫ g for 15 min. The supernatants of the soluble and membrane protein fractions were analyzed by SDS-PAGE.
Affinity Chromatography and Immunoprecipitation Experiments-GST fusion proteins were expressed in Escherichia coli and purified from benzonase-treated bacterial lysates (to remove possible nucleic acid contaminants) using GST-bind TM resin (Novagen) according to standard protocols. Adult rat brain tissue was homogenized in 320 mM sucrose, 4 mM HEPES, pH 7.4, supplemented with 1 mM phenylmethylsulfonyl fluoride and mammalian protease inhibitor mixture (Sigma), and the postnuclear supernatant was extracted with 1% Triton X-100. The buffer concentration in the extract was adjusted to 20 mM HEPES, pH 7.4, 2 mM MgCl 2 , and 80 -100 mM KCl or NaCl. Cleared protein extracts were prepared by ultracentrifugation at 180,000 ϫ g. Protein extracts from cultured mammalian cells were prepared in lysis buffer containing 20 mM HEPES, pH 7.4, 2 mM MgCl 2 , 100 mM NaCl, and 1% Triton X-100 supplemented with 1 mM phenylmethylsulfonyl fluoride and mammalian protease inhibitor mixture (Sigma). Cell extracts were cleared by ultracentrifugation. For affinity purifi-cations, GST fusion proteins were incubated with rat brain extracts with a protein concentration of 4 -6 mg/ml for 2 h at 4°C. Following extensive washes bound proteins were eluted in sample buffer. For immunoprecipitation experiments, antibodies were immobilized on Protein A/G-agarose beads (Santa Cruz Biotechnology) and incubated with either rat brain extracts as above or with cleared cell extracts (total protein concentration: 0.5-1 mg/ml) for 4 h at 4°C under gentle agitation. Beads were washed extensively and eluted with sample buffer. Samples were analyzed by SDS-PAGE and immunoblotting. For direct binding assays immobilized GST fusion proteins (25 g) were incubated for 1 h at 4°C with 25 g of recombinant His 6 -tagged Gadkin proteins and washed extensively. Complexes were analyzed by SDS-PAGE with Coomassie Blue staining.
Detection of Protein Palmitoylation Using the Biotin-switch Method-Protein palmitoylation was analyzed according to a previous study (20) using the biotin-switch method. FLAG-Gadkin-(1-140) was immunoprecipitated from transfected COS7 cells (see above) using anti-FLAG antibodies. Beads were extensively washed in 50 mM Tris buffer, pH 7.4, containing 150 mM NaCl, 2 mM MgCl 2 , and 0.1% Triton X-100 and treated with 25 mM N-ethylmaleimide (diluted from a freshly prepared 1 M stock solution in EtOH) in buffer for 30 min at room temperature to quench free sulfhydryl (SH)-groups. After washing, beads were treated with 1 M hydroxylamine, pH 7.4 (freshly prepared), to liberate cysteine SH-groups from palmitate moieties or 1 M Tris, pH 7.4, as control for 1 h at room temperature. Beads were washed again, and liberated cysteine SH-groups were biotinylated using 1-biotinamido-4-(4Ј-[maleimidoethylcyclohexane]carboxamido)butane (Pierce). Beads were extensively washed and eluted in sample buffer. Samples were analyzed by SDS-PAGE and immunoblotting. Biotinylated proteins were detected using streptavidin-HRP.
CD Spectroscopy-CD spectra of 2.4 M His 6 -Gadkin ⌬51 in PBS containing 10 mM dithiothreitol were recorded at increasing temperatures from 25°C to 55°C on a JASCO J-600 spectropolarimeter (wavelength 245-190 nm, speed 20 nm/min, and resolution 0.2 nm) using a quartz cell of path length 2 mm. Three individual spectra were accumulated. The baseline was corrected by subtracting buffer runs. Selcon3 and Contin programs were applied for calculation of secondary structure content.
Limited Proteolysis-Limited proteolysis was performed with different concentrations of proteinase K and trypsin (1, 5, and 25 g/ml) at 37°C for 10 min. Protein bands were analyzed by in-gel digestion followed by mass fingerprinting. All fragments could be unequivocally assigned to the N-terminal portion of Gadkin (⌬51) (not shown).
Analytical Ultracentrifugation-Analytical ultracentrifugation experiments were performed in a Beckman Coulter XL-I TM analytical ultracentrifuge using the interference optics of the instrument. All experiments were performed in a buffer containing 20 mM K 2 PO 4 , pH 7.4, 150 mM NaCl, and 10 mM dithiothreitol at 20°C. Conversion of the signal in fringe units to molar quantities used a value for the specific refractive index increment of 3.29 fringes⅐mg Ϫ1 ⅐ml Ϫ1 . Data were acquired every 30 s until the major fraction of material had sedimented. Sedimentation traces were analyzed with the c(S) method implemented in Sedfit8.9 (National Institutes of Health). In the case of pure components, data were also analyzed by using a discrete-species model to obtain s, D, and Mw directly. Auxiliary parameters and hydrodynamic values were calculated and corrected to standard conditions using Sednterp. Preparations of pure proteins were examined at different concentrations, spanning at least one order of magnitude on the concentration scale. In the case of mixtures of AP-1␥-ear with Gadkin (⌬51) and the respective mutants, the concentration of Gadkin (⌬51) was held constant at ϳ30 M, and increasing amounts of ␥-ear were added so that molar ratios (AP-1␥-ear/Gadkin (⌬51)) ranged between 0.25 and 4. One cell was filled with each of the pure components at ϳ30 M, respectively. The resulting c(S) distributions were integrated to obtain the loading concentration of protein and the average s value. The concentration of AP-1␥ear was obtained from the concentration of total protein by subtracting the concentration of Gadkin (⌬51) that was independently determined in the same experiment. The concentration of pure AP-1␥-ear served as a control for the expected concentration of AP-1␥-ear. Average s values were plotted as a function of [AP-1␥-ear] and fitted to a model of two independent binding sites using Sedphat4.06 (NIH).

Peptide Synthetic Peptide Array on Nitrocellulose (SPOT)
Synthesis-SPOT syntheses were performed on amino-functionalized Whatman 50 cellulose membranes (Whatman, Maidstone, UK) using a semi-automatic SPOT synthesizer (INTAVIS AG, Köln, Germany). The pepscan arrays were synthesized on an ester-type cellulose membrane, amino-modified by two ␤-alanine units (22). The substitutional and length analyses were prepared on a cellulose-(3-amino-2-hydroxypropyl)ether (CAPE) membrane for a better signal-to-noise ratio. SPOT synthesis was predominantly performed as described in the standard SPOT synthesis protocol (23). For design and arrangement of the substitutional analysis, length analysis, and the pepscan array our in-house software (LISA) was applied. The peptide (His) 6 and the peptide LASDLIVPRR were used as control spots at the upper left, upper right, and lower right corners of the synthesized arrays.
AP-1␥-ear Binding Studies of Cellulose-bound Peptides-All incubations and washing steps were carried out under gentle shaking. After washing the membrane for 10 min with EtOH and three times for 10 min with Tris-buffered saline (TBS, 50 mM Tris(hydroxymethyl)aminomethane, 137 mM NaCl, 2.7 mM KCl, adjusted to pH 8 with HCl), the membrane-bound peptide arrays were blocked for 3 h with blocking buffer (10% blocking reagent (CRB, Norwich, Great Britain) and 1% sucrose in 1:10 TBS buffer). The Gadkin-deduced cellulose-bound pepscan array was incubated with the His 6 tag-labeled AP-1␥-ear (30 g/ml), dissolved in the same blocking buffer at 4°C for 14 h. After washing three times for 10 min with TBS, the HRP-labeled anti-His 6 IgG antibody (Roche 1965085) was added (1:500). Subsequently, the membrane was washed three times for 10 min with TBS. Analysis of peptide-bound AP-1␥-ear was carried out using a chemiluminescence substrate and the Lumi-Imager TM instrument (Roche Diagnostics, Basel, Switzerland). The substitutional and length analyses were probed for AP-1␥-ear binding with a GST-tagged moiety. Peptidebound ␥-ear was detected by adding first the anti-GST IgG antibody (G1160, Sigma) at a concentration of 1 g/ml in blocking buffer (2 h) and subsequently after washing (3 times as described above) by adding the HRP-labeled antimouse IgG antibody (1 g/ml in blocking buffer). As described above chemiluminescence was used for visualization.

Association of Gadkin with Membranes Is Independent of Its Ability
to Bind to the AP-1 Complex-Gadkin has been identified as an AP-1interacting peripheral membrane protein. Indeed, when HeLa cells were subjected to subcellular fractionation, endogenous Gadkin was associated with the membrane pellet, similar to the integral membrane protein TfR, whereas the soluble enzyme glyceraldehyde-3-phosphate dehydrogenase was present in the cytosolic fraction (Fig. 1A). Previous work has suggested that membrane targeting of Gadkin is mediated by the AP-1 complex (8). We tested this model by analyzing truncation mutants of Gadkin either lacking its N-terminal 51 residues (Gadkin ⌬51) or its C-terminal 202 amino acids (Gadkin-(1-100)). GST-fused WT Gadkin or Gadkin ⌬51 but not Gadkin-(1-100) effectively associated with AP-1 from Triton X-100-solubilized rat brain extracts. None of the fusion proteins tested bound to clathrin (Fig. 1B). To investigate the localization of these truncation mutants in living cells we transiently expressed Gadkin mutants tagged at their C-terminal end with eGFP in HeLa cells. Overexpressed WT Gadkin localized to peripheral puncta (Fig. 1C) corresponding to Tfand AP-1-positive TGN-derived endosomal vesicles (EVs) that accumulate in the periphery due to the tight association of Gadkin with the plus-end-directed microtubule motor protein kinesin KIF5 (19). A similar dispersion of AP-1-and Tf-positive endosomes was also observed upon overexpression of c-myc-or FLAG-tagged Gadkin (data not shown). Surprisingly, we observed that Gadkin ⌬51 displayed a cytoplasmic distribution in HeLa cells (Fig. 1C) despite retaining the ability to associate with AP-1 (compare Fig. 1B), whereas Gadkin-(1-100) was concentrated in the perinuclear area and on punctate, presumably endosomal structures (Fig. 1C) but failed to stably bind to AP-1 (Fig. 1C). These data therefore suggested that Gadkin might associate with membranes independently of its interaction with the AP-1 complex. Based on the truncation mutants FIGURE 1. The association of Gadkin with membranes is independent of its binding to AP-1. A, Gadkin is a membrane-associated protein. Total, cytosolic, and membrane fractions were prepared from HeLa cells. 50 g of protein extracts was used for SDS-PAGE and immunoblotting using antibodies against Gadkin, the transmembrane protein TfR, and the soluble enzyme glyceraldehyde-3-phosphate dehydrogenase (GAPDH). B and C, membrane association of Gadkin does not depend on its AP-1 binding ability. B, 20 g of GST (control), GST-tagged Gadkin WT, or its truncation mutants ⌬51 and 1-100 (bound to glutathione beads) were incubated with 1.5 mg of Triton X-100-solubilized rat brain extract. Bound proteins were analyzed by SDS-PAGE and immunoblotting using antibodies against AP-1 ␥-adaptin or clathrin heavy chain (HC). Std, standard. C, subcellular distribution of Gadkin-eGFP truncation mutants. The localization of eGFP-Gadkin (full-length WT, ⌬51, or amino acids 1-100) transiently expressed in HeLa cells was analyzed by fluorescence microscopy (green). DAPI-stained nuclei are shown in blue. Scale bar, 10 m.
Gadkin Association with Membranes and with AP-1 FEBRUARY 5, 2010 • VOLUME 285 • NUMBER 6 analyzed so far, membrane binding of Gadkin involves its 100 N-terminal residues.
Membrane Targeting of Gadkin Requires Its S-Palmitoylation-Protein association with membranes frequently is mediated by post-translational lipid modifications, including prenylation, N-myristoylation, and S-palmitoylation. Because membrane association of Gadkin does not depend on C-terminal determinants and because Gadkin lacks any obvious prenylation sites, we focused on N-myristoylation and S-palmitoylation as possible mechanisms. Gadkin contains a single glycine residue in position 2, representing a putative N-myristoylation site, as well as three cysteine residues that could potentially form acceptor sites for S-linked palmitates ( Fig. 2A). To analyze the potential contribution of either type of modification to membrane targeting of Gadkin-(1-100), we created a series of mutant proteins in which either its 9 N-terminal residues had been deleted (10 -100) or potential lipid modification sites had been inactivated by mutation. Gadkin-(10 -100), indeed, displayed a completely cytoplasmic distribution, whereas Gadkin-(1-100, G2A) associated with peripheral membrane puncta ( Fig. 2A), similar to its WT counterpart. Mutation of all three potentially palmitoylated cysteines to serines abrogated membrane targeting of Gadkin-(1-100) completely, suggesting that a cysteinebased mechanism, most likely S-palmitoylation is required for membrane association of Gadkin-(1-100). By contrast, mutation of either cysteine alone had no effect (data not shown). To analyze whether cysteines 4, 5, and 9 are also required for targeting of full-length Gadkin to membranes, we created a triple point mutant, in which these cysteines had been replaced by serines. As shown in Fig. 2B Gadkin (C4/5/9S) had largely lost its association with membranes in subcellular fractionation experiments. Moreover, transfected Gadkin (C4/5/9S)-eGFP displayed a cytoplasmic distribution in transfected HeLa cells (Fig. 2C).
Two approaches were then taken to more directly address whether Gadkin is subject to S-palmitoylation. First, we incubated transfected HeLa cells with 2-bromopalmitate, a known inhibitor of S-palmitoylation. As expected, such treatment caused Gadkin-(1-100) to dissociate from intracellular membrane puncta, whereas treatment with the solvent alone (EtOH) had no effect (Fig. 2D). Second, palmitoylation of Gadkin-(1-140) was analyzed using the biotin-switch method (20) in which free cysteines are first protected by N-ethylmaleimide, then palmitate is dissociated from palmitoylated cysteine residues by treatment with hydroxylamine (NH 2 OH) liberating selectively palmitoylated cysteines that are then cross-linked to biotin and thus can be decorated with streptavidin-HRP. In agreement with the data above, immunoprecipitated FLAG-Gadkin-(1-140) was detected with streptavidin-HRP following hydroxylamine cleavage but not after treatment of samples with Tris (Fig.  2E), indicating that Gadkin-(1-140) indeed is S-palmitoylated. Finally, we assessed whether S-palmitoylation and, hence, membrane targeting is a prerequisite for AP-1 binding of Gadkin. To test this, transfected HeLa cells expressing FLAGtagged WT or palmitoylation-defective Gadkin (C4/5/9S) were subjected to immunoprecipitation using anti-FLAG monoclonal antibodies. As seen in Fig. 2F, WT and mutant Gadkin were found to associate with endogenous AP-1 in cells with equal efficiencies. Collectively, our data indicate that Gadkin targeting to membranes requires its S-palmitoylation and occurs independently of its complex formation with AP-1. This conclusion is also in agreement with the observation that overexpressed Gadkin is capable of retaining AP-1 at membranes in the presence of BFA (8) (compare also with Fig. 7).
Gadkin Associates with a Conserved Site within the AP-1␥ear Used Also by Other Accessory Proteins-Although S-palmitoylation of Gadkin is clearly required for membrane targeting, its binding to AP-1 might still control the subcellular distribution and physiological function of Gadkin. To address this we studied the association of Gadkin with AP-1 in biochemical and biophysical detail. As expected from previous work, Gadkin and AP-1 efficiently co-immunoprecipitated from rat brain detergent extracts, suggesting that both proteins form a complex in vivo (Fig. 3A). Furthermore, in vitro binding experiments using the purified ear domains of AP-3␤A, AP-2␣, AP-2␤, AP-1␤, and AP-1␥ indicated that Gadkin specifically associates with AP-1␥ but not with other AP appendages (Fig. 3B). Because deletion of its N-terminal 51 residues did not impair the ability of recombinant Gadkin to bind to AP-1 (compare with Fig. 1), but greatly improved protein expression and solubility, we used this Gadkin (⌬51) for further experiments. First, we carried out analytical ultracentrifugation experiments using purified recombinant Gadkin (⌬51) and His 6 -AP-1␥-ear. Both ␥-ear and Gadkin (⌬51) were monomeric in solution. If ␥-ear and Gadkin (⌬51) were mixed, the c(S) distributions indicated the formation of a new species with an s value different from both ␥-ear and Gadkin (⌬51) (supplemental Fig. 3). Integration of these distributions yielded an isotherm indicating the presence of two ␥-ear binding sites within Gadkin of high (42 nM) and low affinity (high M K D ) (Table 1, Fig. 3F). To identify the cognate recognition site for Gadkin within the AP-1␥-ear, we used NMR spectroscopy. A conserved surface area on the ␥-ear domain was identified by NMR chemical shift mapping. It involves amino acids Leu 763 , Arg 793 , Arg 795 , and Lys 797 (Fig. 3, C and D, and supplemental Figs. 1 and 2) as the main binding site for Gadkin. The same site is also used by other AP-1-associated accessory proteins, including ␥-synergin, aftiphilin, and enthoprotin/epsinR. This conclusion was further confirmed by site-directed mutagenesis and direct binding experiments using purified components (Fig. 3E). L763E or R793D/R795D mutants of AP-1 ␥-ear had lost the ability to associate with Gadkin (⌬51), and an R795D/K797E mutant displayed reduced binding.
Identification of a Novel Subtype of AP-1␥-ear Binding Motif within Gadkin-The ability to associate with the ear domains of heterotetrameric clathrin adaptor complexes such as AP-1 is encoded within short linear peptide motifs, often found within unstructured regions of accessory proteins. In the case of Gadkin several putative AP-1-binding motifs had been proposed that were distributed almost over the entire protein sequence. To study the structure of Gadkin we made use of limited proteolysis, CD, and NMR spectroscopy. Treatment of purified Gadkin (⌬51) with either trypsin or proteinase K resulted in near complete breakdown of the protein (Fig. 4A). CD measurements (carried out at different temperatures; Fig. 4B and Table 2) indicated that Gadkin mainly adopts random coil was immunoprecipitated from Triton X-100-solubilized extracts of transfected COS7 cells (500 g of total protein), free sulfhydryl (SH)-groups were quenched with N-ethylmaleimide, and samples were treated with either 1 M hydroxylamine (NH 2 OH) to remove palmitate moieties or 1 M Tris buffer as a control. Liberated SH-groups were labeled with 300 M 1-biotinamido-4-(4Ј-[maleimidoethylcyclohexane]carboxamido)butane. After elution, samples were analyzed by SDS-PAGE and immunoblotting using either an anti-FLAG tag antibody or streptavidin-HRP for detection. Std., 45 g (9%) of cell lysate used for immunoprecipitation and biotin labeling. F, Gadkin WT or a C4/5/9S mutant bind to AP-1. COS7 cells were transiently transfected with plasmids encoding FLAG-Gadkin WT or its C4/5/9S mutant. Triton X-100-solubilized cellular extracts (350 g of total protein) were used for co-immunoprecipitation using anti-FLAG antibodies immobilized on beads (␣-FLAG IP). Bound proteins were analyzed by SDS-PAGE and immunoblotting using antibodies against FLAG and ␥-adaptin (AP-1). Input, 30 g (8%) of the cell lysates used for immunoprecipitation.
conformations and only contains ϳ10 -20% helical structure. NMR spectroscopy (Fig. 4C) shows that the protein predominantly gives rise to random coil chemical shifts and peaks with narrow line shapes characteristic for proteins lacking stable ter-tiary structure. These features largely remain upon addition of AP-1␥-ear (Fig. 4C), supporting the hypothesis that the function of Gadkin is independent of stable tertiary structure. Aggregation of Gadkin can be excluded on the basis of analytical ultracentrifugation experiments, which confirm that the protein remains monodisperse at the concentration used for NMR spectroscopy (data not shown). Because we failed to detect stable secondary or tertiary structure within Gadkin following addition of AP-1␥-ear (Fig. 4C), it is likely that its AP-1 binding ability might be encoded within short linear peptide segments.
To identify such short AP-1 binding determinants within Gadkin wethusturnedtocellulose-boundpeptideSPOT(pepscan)arrays.To this aim the complete amino acid sequence of human Gadkin was covered by an array of cellulose-bound 15-mer peptides overlapping in sequence by three amino acids each. This matrix was incubated with GST-tagged AP-1␥-ear, washed extensively, and bound protein was detected by HRP-labeled anti-GST antibodies. Pepscan overlay assays identified two main interacting peptides (#12-14 and #84-87) containing the sequences 45 FENL 48 (residues critical for binding to AP-1 are highlighted in bold and italics, see also below) and 260 WENDF 264 (Fig. 5A). Whereas 45 FENL 48 resembles (although it does not strictly match) the AP-1␥-ear binding consensus ØG(P/D/ E)(Ø/L/M) (16), 260 WENDF 264 does not. In addition to 45 FENL 48 Neubrandt et al. (8) had proposed two further peptides with the core sequences 10 FGLL 13 and 277 YSGF 280 asputativeAP-1-bindingmotifs that did not yield positive signals in our assay. These motifs might moderately contribute to the overall AP-1 binding affinity of Gadkin, although they are clearly not essential (compare Fig. 5C; see also further below).
Because the sequence WENDF does not conform to the known AP-1␥-ear binding consensus, we carried out substitution analysis using pepscan arrays. This analysis identified Trp and Phe as being critical for the interaction with AP-1␥ (Fig.  5B), suggesting that 260 WENDF 264 might associate with corresponding hydrophobic grooves within AP-1␥-ear via a two-pin plug. Thus, WENDF and related sequences appear to constitute a novel subtype of AP-1␥-ear-binding motif. To test the importance of the WENDF motif for Gadkin (⌬51) binding to ␥-ear, we created point mutants, in which Trp 260 and Phe 264 had been mutated to leucines and analyzed these by affinity chromatography (Fig. 6A) and analytical ultracentrifugation (Fig. 3F and Table 1). Mutant His 6 -Gadkin (⌬51) failed to bind to GST-␥ear immobilized on beads (Fig. 6A) and displayed a weak, but still measurable affinity with a K D of ϳ100 M for AP-1␥-ear in analytical ultracentrifugation experiments (Fig. 3F and Table  1). Residual binding was completely abrogated by mutational inactivation of 277 YSGF 280 (Y277A/F280A) in a W260L/F264L mutant background (Fig. 3F and Table 1). Mutating Y277A/ F280A alone eliminated the second low affinity ␥-ear binding site within Gadkin (⌬51) (Fig. 3F and Table 1). This analysis thus identifies 260 WENDF 264 as the high and 277 YSGF 280 as the low affinity AP-1␥-ear binding site within Gadkin (⌬51). 45 FENL 48 and 10 FGLL 13 may provide additional AP-1-binding determinants within the N-terminal domain of Gadkin.
A Novel Subtype of AP-1␥-ear Binding Motif within Gadkin Regulates Its AP-1 Binding Ability-To assess the contribution of the newly identified AP-1␥-ear-binding motifs within Gadkin to complex formation with AP-1, we transfected HEK293 cells with expression plasmids encoding Gadkin-eGFP (WT) or various mutants thereof. These included variants, in which either 260 WENDF 264 (⌬WENDF), the remaining three conventional ␥-ear-binding motifs (⌬3 motifs), or all four motifs in combination (⌬4 motifs) had been inactivated by mutation of conserved hydrophobic residues (Fig. 5C). Detergent lysates of these cell extracts were then passed over a GST-␥-ear affinity matrix, and the amount of bound Gadkin was assessed by immunoblotting. As expected from our biochemical and biophysical analysis the ⌬WENDF mutant showed a greatly diminished ability to associate with the AP-1␥-ear, whereas binding of a Gadkin mutant lacking the three conventional ØXXØ motifs (⌬3 motifs) was nearly unperturbed. Mutation of all four motifs in combination (⌬4 motifs) completely eliminated AP-1␥-ear binding (Fig. 6B). Nearly identical results were obtained if complex formation between Gadkin-eGFP and endogenous AP-1 was analyzed by co-immunoprecipitation (data not shown). These data confirm that a non-conventional AP-1␥-ear-binding motif within Gadkin is a major determinant for its association with AP-1.
We then asked whether a single WENDF motif might suffice to bind to AP-1 in vitro. Indeed, a WENDF peptide fused to GST was capable of affinity-purifying native endogenous AP-1 from rat brain extracts, whereas actin was absent from this sam-   (Fig. 6C). Putative AP-1-binding motifs similar to the WENDF sequence identified in Gadkin were found in a number of other proteins with a putative role in membrane traffic (Fig.  6D). These include the known AP-1 interaction partner epsinR, SNX-18, a PX-BAR domain-containing protein acting as a membrane tubulator in AP-1-dependent endosomal traffic (24), the exocytosis regulator Munc13-1 (25), and Sec6, a component of the mammalian exocyst complex, which extensively co-localizes with the AP-1B isoform (5). We speculate that all of these proteins might associate with AP-1 at least in part via this novel subtype of ␥-ear-binding motif.
The Ability of Gadkin to Stabilize AP-1 at Membranes Depends on an Intact AP-1␥-ear-binding WENDF Motif and S-Palmitoylation of Gadkin-Previous work has suggested that Gadkin is involved in the control of membrane association of the AP-1 adaptor complex. More specifically, overexpression of Gadkin has been shown to prevent the BFA-induced dissociation of AP-1 from TGN/endosomal membranes (8). We used this assay system to analyze the contribution of the various AP-1␥-ear-binding motifs within Gadkin identified here and to study the importance of the S-palmitoylation of Gadkin. As expected, brief incubation of HeLa cells with BFA led to a rapid and near complete dispersion of endogenous AP-1 into the cytoplasm (Fig. 7, A and D). Overexpression of WT Gadkin prevented AP-1 dissociation from membranes and instead caused the accumulation of AP-1-positive puncta, presumably endosomes, in the cell periphery where both proteins co-localized (Fig. 7, B and D). By contrast, Gadkin-eGFP was unable to protect the related COPI complex from BFA-induced cytosolic dispersion (Fig. 7C), in agreement with its specific association with AP-1. Mutational inactivation of either the WENDF motif or of the three ØXXØ-type ␥-ear-binding motifs (compare Fig. 5C) strongly diminished the ability of Gadkin to stabilize AP-1 at membranes in the presence of BFA (Fig. 7, B and  D). A mutant, in which all four motifs had been inactivated (⌬4 motifs) was completely unable to protect AP-1 from BFA-induced cytosolic dispersion (Fig. 6, B and D). Lack of AP-1 binding of Gadkin mutants correlated with their redistribution to the plasma membrane (Fig. 6B). S-Palmitoylation-defective C4/5/9S mutant Gadkin localized to the cytoplasm (Fig. 7A) and also failed to prevent BFA-induced dissociation of AP-1 from membranes. We conclude that stable membrane recruitment of Gadkin depends on its S-palmitoylation. Moreover, AP-1 association is not required for membrane targeting of Gadkin per se, but regulates its localization at TGN/endosomal membranes. . Gadkin appears to lack stable tertiary structure. A, limited proteolysis performed with different concentrations of proteinase K and trypsin (1, 5, and 25 g/ml) showed a pattern typical for a protein lacking stable tertiary structure. Fragments of different length were observed. Bands marked with numbers were in-gel-digested and analyzed by mass fingerprinting. All fragments could be assigned to the N-terminal portion of Gadkin (⌬51). B, CD spectra of 2.4 M His 6 -tagged Gadkin (⌬51) in PBS, 10 mM dithiothreitol, taken at increasing temperature from 25°C to 55°C, shows the loss of structure elements with temperature increase to 40°C. The process is irreversible; by cooling down to 25°C (p25b) Gadkin (⌬51) remains partially unstructured. C, comparison of 1 H, 15 N HSQC spectra of the 2 H-, 15 N-, and 13 C-labeled Gadkin (⌬51) in the absence (blue) or presence (red) of ␥-ear (1:2 ratio). Gadkin (⌬51) does not appear to adopt stable secondary structure upon addition of AP-1␥-ear. leged access to AP-1. Other sequences resembling conventional ␥-ear-binding motifs, such as the 45 FENL 48 sequence within the N-terminal domain of Gadkin, also contribute to complex formation with AP-1. However, even these motifs within Gadkin do not conform strictly to the known ␥-ear binding consensus sequence ØG(P/D/E)(Ø/L/M) found in most other AP-1-associated accessory proteins, including ␥-synergin, aftiphilin, eps15, epsinR, or rabaptin-5 (16). The use of AP-1-binding motifs significantly deviating from the known consensus thus appears to distinguish Gadkin from these other factors and fits well with the observation that Gadkin selectively associates with the ␥-ear domain of AP-1 but not with that of the related GGA proteins (8). 4 Selectivity for AP-1 versus GGA binding within accessory proteins may thus be encoded at least in part by variations from the consensus sequence as well as by the use of non-conventional subtypes of ␥-ear-binding motifs.
In agreement with this proposal, motifs conforming to the (WY)XX-(DE)(WF) consensus can be found within the known AP-1-binding proteins epsinR (10, 12) (alternatively termed enthoprotin (13) or Clint (11)) and SNX18 (24), the exocytosis regulator Munc13-1 (25), and the exocyst component Sec6. For epsinR and SNX18, complex formation with AP-1 has been demonstrated, whereas Munc13-1 and Sec6 have not yet been shown to bind to AP-1 in vitro. However, a putative direct interaction between Sec6 and AP-1 is likely, based on the fact that exocyst selectively associates with AP-1B-positive endosomal membranes and that expression of AP-1B facilitates exocyst recruitment during basolateral cargo sorting in epithelial cells (5). Future studies will need to address this point in more detail.
Little is known so far regarding the precise physiological function of the Gadkin-AP-1 complex characterized here in biochemical detail. Overexpression and knockdown studies have implicated Gadkin in mannose 6-phosphate receptordependent sorting of lysosomal enzymes (8), a pathway known to involve an endosomal pool of AP-1 (7). Whether or not Gadkin is also involved in other postulated functions of AP-1 such as the formation of Weibel-Palade bodies (26), maturation of secretory granules (27), or basolateral sorting of endosomally delivered cargo in polarized cells (5), remains to be investigated. The precise mapping of the AP-1 binding determinants described here should pave the way for a more thorough investigation of these possibilities. We have recently shown that Gadkin directly binds to the light chains of kinesin KIF5 via a site non-overlapping with any of the AP-1 binding determinants identified in the present work (19). This positions Gadkin at the interface of AP-1-mediated sorting of transmembrane cargo and transport of TGN-derived EVs along the microtubule-based cytoskeleton. Such a function is very compatible with the observed high affinity interaction of Gadkin with AP-1 via the motifs identified here and with its S-palmitoylation-dependent stable association with membranes. Future studies capitalizing on our results will need to address the physiological role of the Gadkin-AP-1 complex in kinesin-driven transport of TGN/endosomal membranes in more detail in vivo.