GGA1 interacts with the adaptor protein AP-1 through a WNSF sequence in its hinge region.

The Golgi-associated gamma-adaptin-related ADP-ribosylation factor-binding proteins (GGAs) are critical components of the transport machinery that mediates the trafficking of the mannose 6-phosphate receptors and associated cargo from the trans-Golgi network to the endosomes. The GGAs colocalize in vivo with the clathrin adaptor protein AP-1 and bind to AP-1 in vitro, suggesting that the two proteins may cooperate in packaging the mannose 6-phosphate receptors into clathrin-coated vesicles at the trans-Golgi network. Here, we demonstrate that the sequence, (382)WNSF(385), in the hinge region of GGA1 mediates its interaction with the AP-1 gamma-ear. The Trp and Phe constitute critical amino acids in this interaction. The binding of Rabaptin5 to the AP-1 gamma-ear, which occurs through a FXXPhi motif, is inhibited by a peptide encoding the GGA1 (382)WNSF(385) sequence. Moreover, mutations in the AP-1 gamma-ear that abolish its interaction with Rabaptin5 also preclude its association with GGA1. These results suggest that the GGA1 WXXF-type and Rabaptin5 FXXPhi-type motifs bind to the same or highly overlapping sites in the AP-1 gamma-ear. This binding is modulated by residues adjacent to the core motifs.

In eukaryotic cells, the budding and fusion of clathrin-coated vesicles (CCVs) 1 mediate the transport of proteins and lipids from the trans-Golgi network (TGN) and the plasma membrane to the intracellular endosomal membrane system (1). In addition to clathrin, the Golgi-and plasma membrane-derived CCVs contain the adaptor protein (AP) complexes AP-1 and AP-2, respectively, as the principal coat proteins. The heterotetrameric AP complexes are comprised of two large adaptin subunits (␥ and ␤1 for AP-1, ␣, and ␤2 for AP-2), which can be subdivided into an N-terminal trunk domain and a C-terminal appendage or ear domain linked by an extended flexible hinge, a medium adaptin subunit (1 or 2), and a small adaptin subunit (1 or 2). Electron microscopic images of purified AP-2 heterotetramers reveal two separate globular appendages, which correspond to the C-terminal portions of the large ␣ and ␤2 chains, projecting from the central core of the com-plex. The AP-1 ␥ and ␤1-appendages are believed to have an analogous gross structure (2).
In addition to the AP complexes, the two subtypes of coated vesicles contain an array of distinct and, to a small degree, overlapping protein components that contribute to cargo selection and facilitate CCV formation by combining clathrin-binding and membrane association domains within an oligomeric protein complex. Among these are the accessory proteins eps15, epsin, amphiphysin, AP-180, auxillin, numb, and disabled-2, which function at the plasma membrane (3), and the GGAs (Golgi-associated ␥-adaptin-related ADP-ribosylation factorbinding proteins), which are involved in vesicle budding from the TGN. The three mammalian GGAs are multidomain proteins that were identified on the basis of sequence similarity between the AP-1 ␥-appendage domain and the C-terminal 150 amino acids of these proteins, also known as the ␥-adaptin ear (GAE) domain (4 -8). This domain is linked to the GAT domain by a flexible hinge region that interacts with clathrin (9,10). The GAT domain binds ARF⅐GTP and mediates the membrane association of the GGAs with the TGN (9). The N-terminal region contains a VHS domain that has been shown to be important in the sorting of transmembrane proteins having acidic cluster-dileucine (AC-LL) signals, such as the mannose 6-phosphate receptors (MPRs) and sortilin, from the TGN to endosomes (10 -13). In the case of the MPRs, interaction with GGAs is critical for the efficient delivery of acid hydrolases to lysosomes (14).
We have previously demonstrated that the GGAs and AP-1, along with MPRs, colocalize in clathrin-coated buds at the TGN of mouse L cells and human HeLa cells (15). We further showed a direct interaction between the hinge domains of the GGAs and the ␥-ear domain of AP-1. In the case of GGA1, our in vitro binding data implicated residues 370 -429 in the hinge segment as being important for its interaction with the AP-1 ␥-ear. Moreover, a mutant MPR that was defective in binding the GGAs was poorly incorporated into AP-1 CCVs, suggesting that the GGAs and AP-1 cooperate to package MPRs into coated vesicles at the TGN. Thus, the in vitro interactions that we observed between the GGAs and AP-1 may be physiologically important in ensuring proper transfer of the MPRs from the GGAs to AP-1. In the current study, we sought to identify the sequence in the GGA1 hinge that is responsible for its association with AP-1. We show that the interaction between the AP-1 ␥-ear and GGA1 is mediated by a WNSF sequence (residues 382-385) within the hinge region of GGA1 and that the 2 anchor aromatic amino acids constitute key residues. A functionally analogous WXX(F/W) sequence found in the proteins NECAP1 and amphiphysin II also mediates their interactions with the AP-1 ␥-ear.
Human NECAP1 was cloned from a brain cDNA library (Clontech) using primers corresponding to the 5Ј and 3Ј ends of the published hNECAP1 sequence (GenBank TM accession number AK074880) into the TA-cloning vector (Invitrogen). The cloned cDNA was sequenced in its entirety to ascertain the identity of the cloned product. GST-hN-ECAP1 was generated by PCR and ligation into EcoRI-digested pGEX-5X-3. The various GST-AP-1 ␥-ear, myc-GGA1pCR3.1, GST-hNECAP1, and GST-SIPWDLWTTS mutant constructs were made using primers incorporating the desired mutations with the QuikChange system (Stratagene, La Jolla, CA). All constructs and mutations were confirmed to be correct by dideoxynucleotide sequencing.
Peptides-The amino acid sequences derived from the GGA1 hinge region (376 -391) corresponding to the peptides used in this study are as follows: SLDGTGWNSFQSSDAT, termed GGA1WnsF peptide; SLDGTGANSAQSSDAT, termed GGA1AnsA peptide; SLDGTGFNS-FQSSDAT, termed GGA1FnsF peptide. All peptides were synthesized at the Protein Chemistry Laboratory at Washington University in St. Louis, MO and purified by reverse phase high-performance liquid chromatography.
Protein Expression and Purification-myc-GGA1 (WT and mutants) were expressed in SF9 insect cells, and cell lysates were prepared as described previously (18). COS-7 cells were transfected with plasmid DNA using LipofectAMINE Plus according to the manufacturer's instructions (Invitrogen). Cells were harvested 48 h after transfection and lysed into cold assay buffer A (25 mM Hepes-KOH, pH 7.2, 125 mM potassium acetate, 2.5 mM magnesium acetate, 1 mM dithiothreitol, and 0.4% Triton X-100) by sonication. Following centrifugation at 20,000 ϫ g, the supernatant containing the GGA protein was stored at Ϫ80°C for use in the binding assays. All lysates were clarified by centrifugation at 20,000 ϫ g immediately prior to use in pull-down experiments. Bovine adrenal cytosol was prepared essentially as described (19) except that frozen glands, instead of fresh tissue, were used.
The various glutathione S-transferase (GST) fusion proteins were expressed in Escherichia coli BL21(RIL) cells and purified on a glutathione-Sepharose affinity column as described (16). His-myc-GGA1 and His-Xpress-␥-ear-(703-822) were expressed in SF9 insect cells and E. coli BL21(RIL) cells, respectively, and purified on a Ni-NTA column (Qiagen) as per the manufacturers protocol. Briefly, the Ni-NTA beads were washed and equilibrated with binding buffer (50 mM NaH 2 PO 4 , 300 mM NaCl, and 10 mM imidazole, pH 7.4) before incubating the His-myc-GGA1-containing insect cell lysate or the His-Xpress-␥-ear-(703-822) containing bacterial lysate in the same buffer containing 0.4% Triton X-100 overnight at 4°C with constant tumbling. Following the overnight incubation, the beads were gently spun down and washed repeatedly with wash buffer (binding buffer with 20 mM imidazole, pH 7.4) until there was no detectable protein being released and then incubated for 15 min at room temperature with elution buffer (binding buffer with 300 mM imidazole, pH 7.4). Elutions were repeated three times, and the fractions were pooled, concentrated using a Centricon-10 apparatus, and frozen in aliquots at Ϫ80°C for further use. The concentration of the purified protein was determined using the Bradford assay (Bio-Rad) with bovine serum albumin as a standard.
Binding Assays-GST pull-down experiments were performed with COS cell or insect cell lysates as described (18). Binding assays using bovine adrenal cytosol were performed in assay buffer B (assay buffer A with 0.1% Triton X-100). The binding reactions were allowed to proceed for 3 h at 4°C with constant tumbling followed by centrifugation at 750 ϫ g for 1 min. An aliquot of the supernatant was saved, and the pellets were subjected to three washes in 1 ml of cold assay buffer B each time. The pellets were resuspended in SDS sample buffer and heated at 100°C for 5 min before loading. For the peptide inhibition studies, GST pull-down assays were performed similarly except that reactions were carried out in a final volume of 300 l containing the indicated concentrations of the various peptides.
Electrophoresis and Immunoblotting-Pellet and supernatant fractions (as indicated by P and S, respectively, in the figure legends) were resolved on 10% SDS-polyacrylamide gels and transferred to nitrocellulose membrane. Blots were blocked with TBST (100 mM Tris⅐Cl, pH 7.5, 150 mM NaCl, 0.1% Triton X-100) containing 5% non-fat milk for 1 h at room temperature. The blots were then probed with primary antibodies as indicated in the individual figure legends followed by horseradish peroxidase conjugated anti-mouse IgG. The immunoreactive bands were visualized on x-ray films using enhanced chemiluminescence (ECL) (Amersham Biosciences). Where necessary, the bands were quantified with the Kodak digital imaging system (Eastman Kodak Co.).

GGA1 WXXF Motif Mediates Its Interaction with AP-1-We
have previously demonstrated that the hinge region of GGA1 between residues 370 and 429 mediates its interaction with the AP-1 ␥-ear (15). To delineate the precise sequence requirement for AP-1 binding, a number of truncation mutants of myctagged GGA1 were expressed in COS-7 cells and tested for binding to GST fused to the AP-1 ␥-ear (residues 703-822) in GST pull-down assays. As shown in Fig. 1B, left, WT myc-GGA1 and truncations to 410 and 392 bound to GST-␥-ear, but further truncation to residue 387 resulted in the loss of this interaction. This indicated that residues 387 SSDAT 391 in the hinge region (Fig. 1A) are necessary for interaction in the context of this truncated mutant. Moreover, a mutant with residues 387-391 deleted (⌬387-391) in the full-length GGA1 also failed to bind to the AP-1 ␥-ear (Fig. 1B, right). However, when these 5 residues were mutated to Ala in the full-length GGA1 (myc-GGA1 387-391A), normal binding was observed (Fig. 1B, right). Based on these findings, we hypothesized that the actual binding motif was likely to be situated proximal to the 387-391 sequence and that truncations or deletions in this region impaired the ability of the binding motif to interact with the AP-1 ␥-ear. To test this, a series of alanine mutants was constructed and expressed in COS-7 cells as indicated in Fig.  1B, right. Only the 381-386A mutant failed to bind the AP-1 ␥-ear. Mutation of residues 376 -380 to Ala had a marginal effect on ␥-ear binding, whereas mutation of residues 371-375 to Ala had no detectable effect.
Analyses of single alanine substitutions within the GGA1 381-386 segment showed that Trp-382 and Phe-385 are critical for the interaction of GGA1 with the ␥-ear domain of AP-1 (Fig.  2). Further, replacement of Trp-382 with Leu, Val, or Phe abolished binding, indicating an absolute requirement for Trp at this position. The Phe at position 385 could be replaced with Trp with only a modest loss of binding, but substitutions with Leu or Val resulted in a complete lack of binding (Fig. 2).
To further examine the role of the GGA1 WXXF motif in binding the AP-1 ␥-ear, we performed competition studies with a peptide consisting of GGA1 residues 376 -391 or a similar peptide with Trp-382 and Phe-385 mutated to Ala. As shown in Fig. 3A, the binding of myc-GGA1 to the GST-␥-ear protein was strongly inhibited by the WT peptide but not the mutated peptide. The inhibition curve shown in Fig. 3B demonstrates that the inhibition is concentration-dependent and occurred with micromolar concentrations of free peptide. Furthermore, a GST fusion protein containing the 16 amino acid residues corresponding to the GGA1 peptide (residues 376 -391) was sufficient to pull-down AP-1 from bovine adrenal cytosol, and this interaction was totally abrogated by the addition of the WT peptide (Fig. 3C). This indicates that the WXXF motif of GGA1 is necessary and sufficient to mediate its interaction with the AP-1 ␥-ear.
WXXF-and FXX⌽-containing Proteins Bind to a Common or Overlapping Site on the AP-1 ␥-Ear-Recently, a number of accessory proteins that interact with the appendage domains of AP-1 and the GGA proteins were shown to contain sequences that fit a consensus motif, (D/E)FXX⌽ (⌽ represents leucine, phenylalanine, tryptophan, or methionine) (20 -22). These include p56 and Rabaptin5 (Fig. 4A), as well as ␥-synergin and EpsinR. The crystal structures of peptides with sequences conforming to this motif, in complex with the GAE domains of GGAs 1 and 3, have revealed the critical nature of the phenylalanine residue in the first position and ϩ3 residue for this interaction (20,21). On the basis of mutagenesis data coupled with the structural homology between the AP-1 ␥-ear and the GGA GAE domains, it appears likely that the accessory protein-binding site on both adaptors is very similar (20). To assess whether these proteins bind to the same site on the AP-1 ␥-ear as does GGA1, we tested the effect of the GGA1 WXXF peptide on the binding of Rabaptin5 to GST-␥-ear. Rabaptin5 contains an FGPLV sequence (Fig. 4A) that interacts with the GGA GAE domain (23). As shown in Fig. 4B, Rabaptin5 binding was strongly inhibited by 0.2 mM WT GGA1 peptide but not by 0.5 mM AXXA mutant peptide. This suggests that the (D/ E)FXX⌽ and the WXXF motifs bind to a common or closely overlapping site on the AP-1 ␥-ear.
We next tested whether the FXX⌽ peptide (FGGF), derived from the p56 protein GAE-binding sequence, inhibits binding of myc-GGA1 to the AP-1 ␥-ear. In competition assays (Fig. 5A), both the GGA and the p56 peptides displayed a similar inhibition profile except that the p56 peptide stimulated binding at the lowest concentration tested (30 M). This was consistent with both peptides binding to a common or overlapping site on the AP-1 ␥-ear. In contrast to this finding with the p56 peptide, the GGA1 mutant FXXF peptide is at least 8-fold less potent than the WT peptide in inhibiting binding of myc-GGA1 to GST-␥-ear (Fig. 5B). The most likely explanation for this difference is that residues in the proximity of the anchor aromatic amino acid influence the predilection for either a Phe or a Trp residue in the first position of the motif. These neighboring residues could possibly impact the specificity of the interactions of different accessory proteins with the GAE domains, as suggested previously (20). Curiously, the lowest concentration FIG. 1. Identification of the AP-1 interaction sequence within the hinge region of GGA1. A, amino acid sequence of the GGA1 hinge region between residues 371 and 392. B, WT and mutant myc-tagged GGA1 were expressed in COS-7 cells, and whole cell extracts containing the various GGA1 proteins were used directly in GST pull-down experiments. 25% of the pellet (P) and 2% of the supernatant (S) fractions were resolved on 10% SDS gels, transferred to nitrocellulose, and probed with the anti-myc mAb as described under "Experimental Procedures."

FIG. 2. Trp-382 and Phe-385 constitute critical amino acids for the interaction of GGA1 with the AP-1
␥-ear. GGA1 Trp-382 and Phe-385 mutants expressed in COS-7 cells were tested for binding to the GST-␥-ear in pull-down experiments as described under "Experimental Procedures." 25% of the pellet (P) and 2% of the supernatant (S) fractions were resolved on 10% SDS gels, transferred to nitrocellulose, and probed with the anti-myc mAb.
of the GGA1 mutant FXXF peptide stimulated GGA1 binding to the AP-1 ␥-ear, just as observed with the p56 peptide. An explanation for this anomaly is not clear at this point.
GGA1 WXXF Motif Engages the Same Key Residues in AP-1 ␥-Ear as FXX⌽ Motif-X-ray crystallography together with structure-based mutational analysis have revealed the binding site for accessory proteins within the AP-1 ␥-ear (24,25). The crystal structure showed the ␥-ear domain to consist solely of an immunoglobulin-like fold with several basic and hydrophobic residues that are highly conserved between the ␥-ear and GGA GAE domains forming a cluster on the surface of the ␤-sandwich structure. It has been shown that mutations K756Q, R793Q, R795Q, K797Q, E812K, and L762E in the ␥-ear domain abolished its binding to the accessory proteins, ␥-synergin and Rabaptin5 (24,25). Since the GGAs also bind to the ␥-ear, we wanted to determine whether the GGA-␥-ear interaction required the same structural determinants as did the other accessory proteins. To address this issue, the six mutations described above were introduced into GST-␥-ear, and the recombinant fusion were proteins tested for their ability to pull-down Rabaptin5 and GGA1. The results of the binding assays show that point mutations in 5 of the 6 residues within the ␥-ear abolished its binding to GGA1 (Fig. 6). All of these mutants except for K797Q also impaired Rabaptin5 binding. The E812K mutation did not affect the ability of the ␥-ear to interact with either GGA1 or Rabaptin5. Thus, with one exception, the same basic and hydrophobic residues on the ␥-ear are required for interaction with both Rabaptin5 and GGA1.
NECAP and Amphiphysin Bind AP-1 in Vitro through WXX(F/W) Motifs-Two recent reports have identified a novel AP-2 ␣-appendage-binding sequence, WXX(F/W)X(D/E), that is similar to the WXXF motif described here (26,27). In the protein NECAP (26), the WXXF sequence is positioned at the extreme C terminus of the protein. Hence, it was proposed that for the NECAP AP-2 ␣-appendage-interacting motif, the terminal carboxyl group may replace the distal acidic side chain normally seen in the consensus sequence (27). Ritter et al. (26) also found that the AP-1 ␥-ear interacted with NECAP proteins, but this interaction did not require the extreme C-terminal WXXF. Upon examining the NECAP amino acid sequence (Fig. 7A), we observed a second WXXF motif, 252 WGDF 255 , positioned 16 residues proximal to the C-terminal ␣-appendage-interacting WXXF motif. In view of our finding that the internal WXXF sequence of GGA1 interacted with the AP-1 FIG. 3. The GGA1 WXXF WT peptide competes with full-length myc-GGA1 for binding to the AP-1 ␥-ear. A, GST pull-down assays with COS-7 cell lysates containing myc-GGA1 were performed in the presence of 1 mM of either the WT WXXF peptide derived from the GGA1 hinge region or a mutant AXXA peptide. Binding of myc-GGA1 to GST-␥-ear was determined by immunoblotting and probing with the anti-myc mAb. 40% of the pellet fraction was loaded. B, various concentrations of the GGA1 WXXF peptide were included in the binding assays with GST-␥-ear and purified His-myc-GGA1 to generate the peptide inhibition curve. 40% of the pellet fraction was resolved on 10% SDS-gel and transferred to nitrocellulose, and the immunoblot was probed with the anti-myc mAb. The bands were quantified using the Kodak digital imaging system. C, GST pull-down assay was performed with GST-GGA1-(376 -391) using bovine adrenal cytosol (5 mg/ml) as the source of cytosolic AP-1 and 1 mM WXXF peptide as noted. 40% of the pellet (P) and 5% of the supernatant (S) fractions were loaded. The immunoblot was probed with the anti-AP-1 ␥-subunit-specific mAb (100/3).

FIG. 4. WXXF and FXX⌽ motifs bind to a common or overlapping site on the AP-1 ␥-ear.
A, alignment of the AP-1-interacting sequence and surrounding residues of p56, Rabaptin5, and GGA1. B, GST-␥-ear binding to Rabaptin5 from bovine adrenal cytosol (5 mg/ml) was assayed in the presence of 0.2 or 0.5 mM of either the GGA1 WT WXXF peptide or a mutant AXXA peptide. 30% of the pellet fraction was loaded. The immunoblot was probed with the anti-Rabaptin5 mAB.
␥-ear, we reasoned that the proximal WXXF sequence of NE-CAP1 may represent the ␥-ear-interacting motif. To test this, various constructs of GST-NECAP1 were expressed and purified from bacteria for use in pull-down assays with bovine adrenal cytosol as a source of cytosolic AP-1. The results shown in Fig. 7, B and C, indicate that the proximal 252 WGDF 255 sequence of NECAP is the ␥-ear-interacting motif and that it behaves in many respects similar to the GGA1 WXXF sequence. For instance, Trp-252 could not be substituted by either Phe or Ala at this position, but Phe-255 could be substituted by a Trp (Fig. 7C). Moreover, if the protein terminated with the 252 WXXF 255 sequence (Ser-256 stop), it was unable to bind to the ␥-ear, similar to the observation with the GGA1 387 stop mutant. This is in contrast to the findings with the AP-2 ␣-appendage-interacting WXXF motif, which requires an Asp or Glu in the ϩ5 position (where Trp is the 0 position) or the motif be positioned at the extreme C terminus of the protein (27).
In the course of characterizing the endocytic interactions between amphiphysin and clathrin, Drake and Traub (17) observed that a GST-SIPWDLWEPT peptide fusion harboring the distal amphiphysin II clathrin-binding sequence was able to associate with AP-1 from cytosol. Although these investigators did not ascertain whether the AP-1 binding to GST-SIP-WDLWEPT was direct or the indirect consequence of its interaction with clathrin, it was shown that mutation of the proximal Trp to Ala, Phe, Tyr, or His resulted in the loss of AP-1 association. Mutation of the distal Trp to Phe, however, had no effect on AP-1 binding. These results are reminiscent of our data with the GGA1 and NECAP ␥-ear-interacting WXXF motifs. Hence, we sought to determine whether the GST-SIP-WDLWEPT peptide fusion could associate directly with the FIG. 5. p56 FXXF peptide and GGA1 FXXF mutant peptide have significantly different inhibitory property. A, the binding of GST-␥-ear to myc GGA1 was assessed in the presence of different concentrations of either the p56 FXXF peptide or the GGA1 WXXF peptide. B, the critical nature of the first Phe residue of the GGA1 WXXF motif was determined in peptide inhibition using either the WT GGA1 peptide or a mutant peptide with a Trp to Phe substitution. 40% of the pellet fraction was loaded in both competition assays. Immunoblots were probed with the anti-myc mAb.
FIG. 6. The same residues in AP-1 ␥-ear are important for binding both the GGA1 WXXF motif and the Rabaptin5 FXX⌽ motif. GST-␥-ear and point mutants of the appendage domain were tested for their ability to bind to either myc-GGA1 from insect cell lysate or Rabaptin5 from bovine adrenal cytosol (5 mg/ml). 40% of the pellet fractions and either 5% of the bovine adrenal cytosol or 4% of the insect cell lysate input were loaded. The immunoblots were probed with either the anti-myc or anti-Rabaptin5 mAb.

FIG. 7. NECAP1 and amphiphysin II associate with AP-1 through WXX(F/W) motifs.
A, amino acid sequence of the human NECAP1 protein C terminus from residues 248 to 260, showing the AP-1-interacting WXXF motif. B and C, pull-down experiments were performed with either GST-hNECAP1 or C-terminal truncations or point mutants of the NECAP1 protein using bovine adrenal cytosol as a source of AP-1. 30% of the pellet (P) and 5% of the supernatant (S) fractions were loaded. Immunoblots were probed with the anti-AP-1 ␥-subunit-specific mAb (100/3). D, GST-SIPWDLWEPT (amphiphysin II clathrin-binding sequence) and peptide fusions with point mutations of the anchor aromatic amino acids were assayed for their ability to interact with purified AP-1 ␥-ear in pull-down experiments. 2% each of the pellet (P) and supernatant (S) fractions was resolved on 15% SDS gel, transferred to nitrocellulose, and probed with the anti-Xpress mAb.
AP-1 ␥-ear and whether this interaction was dependent on the 2 Trp residues. As shown in Fig. 7D, the Ni-NTA-purified AP-1 ␥-ear domain was able to bind directly to the WT GST-SIP-WDLWEPT peptide fusion, but binding was severely compromised by mutating the proximal Trp to either an Ala or a Phe. Mutation of the distal Trp to Phe, however, had no detectable effect on the ability of the purified ␥-ear to associate with the GST-amphiphysin II peptide fusion, but a Trp to Ala substitution at this position abolished binding. These results lend further support to the concept that the WXX(F/W) motif is a bona fide GAE-binding motif that has distinct characteristics from the (D/E)FXX⌽ motif. DISCUSSION The data presented in this study establish that the 382 WNSF 385 sequence within the GGA1 hinge region is responsible for its interaction with the AP-1 ␥-ear. Within this sequence, the 2 anchoring aromatic residues are critical determinants of the interaction. Recently, a peptide motif conforming to the consensus sequence, (D/E)FXX⌽ (⌽ represents leucine, phenylalanine, tryptophan or methionine), has been shown to be a GAE-binding motif (20 -22). This motif is found in several accessory proteins that interact with both the AP-1 ␥-ear and the GGA GAE domains, including ␥-synergin, Rabaptin 5, and EpsinR. Structural studies of peptides containing this motif in complex with the GAE domains of GGAs 1 and 3 show that the peptides bind in an extended conformation to two complementary hydrophobic pockets on the surface between strands ␤4 and ␤5 of the GAE domains (20,21). The invariant anchoring phenylalanine residue (position 0), in particular, was observed to make specific hydrophobic contacts within the first pocket with 2 highly conserved arginine residues in the GAE domain. The flatness of both the aromatic ring of the phenylalanine and the guanidino group of the arginine made the resultant stacking interactions highly favorable. The crystal structures also revealed that the more general hydrophobic nature of the second pocket would permit accommodation of bulky hydrophobic residues in position ϩ3. This was consistent with the fact that various bulky hydrophobic residues occupy this position in the GAE-binding motifs of the different accessory proteins. Since the GGA1 WNSF sequence has the aromatic amino acid, Trp, in its first position, we envisaged that Trp-382 of GGA1 participates in a very similar interaction with the AP-1 ␥-ear. In agreement with this model, we found that the analogous arginine residues in the AP-1 ␥-ear, Arg-793 and Arg-795, when mutated to glutamine, rendered the GST ␥-ear mutants incapable of interacting with myc-GGA1. Thus, the structural data, when taken together with this binding experiment, suggest that both Phe and Trp should be functional and interchangeable in the first position of the motif and that several bulky hydrophobic amino acids could occupy the fourth position of the motif.
However, our mutagenesis analysis of the GGA1 382 WNSF 385 sequence and the surrounding residues revealed that this was not the case. We found that phenylalanine cannot substitute for tryptophan within the GGA1, NECAP1, or amphiphysin II WXX(W/F) motifs. We also observed that position ϩ3 of the motif had to be occupied by either phenylalanine or tryptophan to maintain binding to the AP-1 ␥-ear. Mutants of GGA1 with the bulky hydrophobic residues Val and Leu at the ϩ3 position were unable to bind to the ␥-ear. Furthermore, we found that the nature of the amino acids surrounding the core motif influenced binding to the GAE domain. For instance, when myc-GGA1 was truncated immediately following the 382 WNSF 385 sequence, binding to GST-␥-ear was abrogated, suggesting that the GGA1 WXXF motif cannot be located at the extreme C terminus of the protein. Similarly, truncation of the NECAP1 protein following the 252 WGDF 255 sequence abolished its abil-ity to bind AP-1. In addition, we observed that binding of myc-GGA1-(⌬387-391) to GST-␥-ear was impaired and myc-GGA1-(376 -380A) exhibited a partial decrease in binding. These findings show that the WXXF motif differs from the (D/E)FXX⌽ GAE-binding motif in several respects. On the other hand, our peptide inhibition studies, together with the behavior of the GST-␥-ear mutants, suggest that the two motifs bind to a common site on the AP-1 ␥-ear. One explanation consistent with all these findings is that the WXXF and the (D/E)FXX⌽ motifs bind to the same site on the AP-1 ␥-ear, but this binding is modulated by the residues neighboring the 0 and ϩ3 positions of the motifs. For example, the FXX⌽ motif is preceded in every case by an acidic residue that is not present in the WXXF motif (21). We are currently exploring this possibility using additional site-directed mutagenesis to change the residues surrounding the WXXF core motif. Another possibility is that the two motifs bind to overlapping sites on the AP-1 ␥-ear.
Very recently, a novel AP-2 ␣-appendage-binding motif conforming to the consensus WXX(F/W)X(D/E) was identified (26,27). Although this AP-2-binding motif has a similar requirement to the AP-1-binding WXXF sequence for tryptophan and phenylalanine residues, a key difference is that the AP-2-interacting motif must be located at the extreme C terminus of the protein, as is the case with NECAP1 (26), or have a negatively charged residue at position ϩ5, as seen in the long splice isoform of synaptojanin 1, AAK1, GAK, and stonin (27). This difference can explain the failure of GGA1 to bind to AP-2 and illustrates once again the important role that residues surrounding the core motif may play in modulating binding.
We have previously demonstrated that GGAs 2 and 3, as well as GGA1, interact with the AP-1 ␥-ear (15). We found that the ability of GST-␥-ear to bind GGA2 and GGA3 was completely blocked in the presence of 1 mM of the GGA1 peptide (data not shown). Hence, we would predict the occurrence of analogous binding motifs (WXXF or FXX⌽) in the hinge regions of GGAs 2 and 3. An alignment of the three GGA hinge segments revealed no obvious candidate sequence in GGA2, but the 417 WHLL 420 sequence within the GGA3 hinge (long form) aligned with the 382 WNSF 385 sequence of GGA1 and could potentially be the ␥-ear-binding motif. However, when we mutated both Trp-417 and Leu-420 to alanine, we observed no impairment in the binding of the GGA3 mutants to GST-␥-ear (data not shown). Thus, the GGA2 and GGA3 AP-1-binding motifs may represent other novel sequences that fit into the same hydrophobic pockets of the AP-1 ␥-ear domain.
While this manuscript was in preparation, Mattera et al. (28) reported that the NECAP1 GAE-binding motif resides in the sequence 252 WGDF 255 , which is the same proximal WXXF motif that we have identified in the present study. In their study, the investigators performed a combinatorial analysis of GAEbinding motifs by phage display peptide library screening and found that the majority of the binding peptides had a tryptophan at position 0 and ϩ3. Moreover, when they substituted a tryptophan for the phenylalanine in the Rabaptin5 FXX⌽ GAE-interacting motif, they observed strong binding, suggesting that tryptophan may replace phenylalanine in FXX⌽-type motifs. Clearly, the reverse does not hold true, raising the possibility that the FXX⌽-type motif may have evolved from the WXX(F/W)-type motif with concomitant changes in the surrounding residues, thereby determining the predilection for either tryptophan or phenylalanine at position 0.