The Ear of -Adaptin Interacts with the COOH-terminal Domain of the Eps15 Protein

The role of Eps15 in clathrin-mediated endocytosis is supported by two observations. First, it interacts specifically and constitutively with the plasma membrane adaptor AP-2. Second, its NH terminus shows significant homology to the NH terminus of yeast End3p, necessary for endocytosis of α-factor. To gain further insight into the role of Eps15-AP-2 association, we have now delineated their sites of interactions. AP-2 binds to a domain of 72 amino acids (767-739) present in the COOH terminus of Eps15. This domain contains 4 of the 15 DPF repeats characteristic of the COOH-terminal domain of Eps15 and shares no homology with known proteins, including the related Eps15r protein. Precipitation of proteolytic fragments of AP-2 with Eps15-derived fusion proteins containing the binding site for AP-2 showed that Eps15 binds specifically to a 40-kDa fragment corresponding to the ear of α-adaptin, a result confirmed by precipitation of Eps15 by α-adaptin-derived fusion proteins. Our data indicate that this specific part of AP-2 binds to a cellular component and provide the tools for investigating the function(s) of the association between AP-2 and Eps15.

The role of Eps15 in clathrin-mediated endocytosis is supported by two observations. First, it interacts specifically and constitutively with the plasma membrane adaptor AP-2. Second, its NH 2 terminus shows significant homology to the NH 2 terminus of yeast End3p, necessary for endocytosis of ␣-factor. To gain further insight into the role of Eps15-AP-2 association, we have now delineated their sites of interactions. AP-2 binds to a domain of 72 amino acids (767-739) present in the COOH terminus of Eps15. This domain contains 4 of the 15 DPF repeats characteristic of the COOH-terminal domain of Eps15 and shares no homology with known proteins, including the related Eps15r protein. Precipitation of proteolytic fragments of AP-2 with Eps15-derived fusion proteins containing the binding site for AP-2 showed that Eps15 binds specifically to a 40-kDa fragment corresponding to the ear of ␣-adaptin, a result confirmed by precipitation of Eps15 by ␣-adaptin-derived fusion proteins. Our data indicate that this specific part of AP-2 binds to a cellular component and provide the tools for investigating the function(s) of the association between AP-2 and Eps15 .
Eps15 is the prototype of a new family of signal transducers characterized by their ability to interact with a large number of proteins (1). It was initially described as a substrate of the epidermal growth factor (EGF) 1 and platelet-derived growth factor tyrosine kinase receptors endowed with transforming properties (2,3). A novel protein, Eps15r, with 47% identity to Eps15 has recently been cloned using a probe derived from the region encoding the NH 2 -terminal domain of Eps15 (1). Both Eps15 and Eps15r are organized into three distinct structural domains. The amino terminus domain of Eps15 (amino acids 1-300) displays 70% identity to the amino terminus of Eps15r and is composed of three imperfect repeats of approximately 100 amino acids with candidate tyrosine phosphorylation sites and two EF-hand type calcium binding sites. Each repeat contains a domain of 70 amino acids, which is conserved not only in Eps15 and Eps15r, but also in several proteins in yeast and nematodes and is therefore designated EH for Eps15-Homology domain. The first domain of Eps15 interacts with several unidentified cytosolic proteins (1). The homology between the two proteins drops to 45% in the second domain, but the heptads required for coiled-coil structure are conserved. A possible function of these heptads in homo-or heterodimerization has been hypothesized (2,3). Finally, there is little conservation between the COOH-terminal domains (amino acids 520 -896) of Eps15 and Eps15r, with two notable exceptions. First, multiple DPF motifs are present in both proteins. Second, the two proteins contain a proline-rich domain, PALPPK, which binds the Src homology 3-domain of the crk protooncogene (4).
Besides its possible function in signal transduction, Eps15 may play a role in endocytosis. First, there is 62% homology between Eps15 EH domains and the NH 2 terminus of End3p, a protein required for clathrin-mediated internalization of ␣-factor in Saccharomyces cerevisiae. In addition, the temperaturesensitive internalization defect of end3 mutants can be complemented with wild type End3p but not with mutated End3p bearing small deletions in the EH domain (5). Furthermore, we have recently observed that, in all cell types tested and in several species, Eps15 is constitutively associated with the plasma membrane adaptor complex AP-2 (6), which serves several functions in endocytosis. On the one hand, AP-2 favors assembly of the clathrin triskelion and association of the clathrin lattice to the plasma membrane. On the other hand, AP-2 participates in the recruitment of endocytosed receptors in clathrin-coated pits (7,8). To gain insight into the mechanisms of interaction between AP-2 and Eps15, and thereby into the function of Eps15 and related proteins in endocytosis, a precise knowledge of the sites of interaction of the two proteins is required. With this goal in mind, we constructed a large series of GST fusion proteins derived from Eps15 and tested them for their ability to precipitate the AP-2 complex from cell lysates. Eps15-derived fusion proteins containing the binding site for AP-2 were then used to precipitate fragments derived from the limited proteolysis of AP-2 to define the component of AP-2 interacting with Eps15. The binding site of Eps15 was localized to the COOH-terminal appendage (ear) of ␣-adaptin, a result confirmed using fusion proteins derived from ␣-adaptin.

EXPERIMENTAL PROCEDURES
Cells and Antibodies-All studies were performed using the human MOLT16 leukemic T cell line (gift of Dr. Minowada, Fujisaki Cell Center). Production and characterization of the 6G4 anti-Eps15 monoclonal antibody (mAb) has been previously reported (6). Another anti-Eps15 mAb produced against a COOH-terminal fragment (residues 717-896) of murine Eps15 was obtained from Affiniti Research Products Ltd (Nottingham, UK). Mouse mAbs, AC2-M15, against ␣-adaptin A, and AC1-M11 against the NH 2 -terminal domains of ␣-adaptins A and C, were kind gifts of Dr. M. S. Robinson (9). mAb 100/2, against the COOH termini (ears) of ␣-adaptins A and C; mAb 100/1, against the NH 2 terminus of ␤-adaptin; and mAb 100/3, against ␥-adaptin, were * This work was supported by INSERM and a grant from the Association pour la Recherche contre le Cancer (ARC Contract 6836). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
§ Supported by a fellowship from the Association pour la Recherche contre le Cancer (ARC, Villejuif, France).
Construction of Glutathione S-transferase (GST) Fusion Proteins-Different fusion proteins were derived from eps15 using the GST gene fusion system and the PGEX5.1 vector (Pharmacia-LKB, Les Ulis, France). The cDNA of human eps15 subcloned in pBluescript II KS (Stratagene) was obtained in the laboratory 2 and used as a template to generate different cDNA fragments encoding domains DI, DII, and DIII of eps15 and truncated forms of DIII. A BamHI and a XhoI site were introduced in the upper and lower primers, respectively, to allow subcloning of the PCR products in the PGEX5.1 vector in frame with the GST moiety. The constructs were checked by nucleotide sequencing (Thermosequenase, Amersham Corp., Les Ulis, France). GST fusion proteins encoding the COOH domain of ␣-adaptin or parts of this domain were similarly generated by PCR using the cDNA of mouse ␣-adaptin C, the nonalternatively spliced ubiquitous form of ␣-adaptin (12), subcloned in pBluescript II SK as a template (a kind gift of Dr. M. Robinson). Sequences of the used primers are available on request. Production of fusion proteins in DH5␣ bacteria and purification were performed as described elsewhere (6).
For limited proteolysis of the AP-2 complex, MOLT16 cells were lysed in 50 mM Tris-HCl, pH 8, 150 mM NaCl, 0.5% Triton X-100, in the absence of protease inhibitors. Cell lysates were then digested with trypsin (Life Technologies, Inc., Eragny, France) at a 1:500 protein ratio at 37°C for various periods of time. Digestion was stopped by addition of a mix of protease inhibitors (see above) and 10% fetal calf serum (Life Technologies). Digested lysates and control undigested lysates were precipitated by GST fusion proteins.

Eps15 Binds AP-2 via Its Third COOH-terminal Domain-
Previous studies have shown that the anti-Eps15 mAb 6G4 as well as a fusion protein encompassing the full-length of Eps15 precipitate the various polypeptides of the AP-2 complex (6). To determine which domain of Eps15 interacts with AP-2, three GST fusion proteins, each one comprising one of the three structural domains of Eps15, GST-DI (amino acids 1-315), GST-DII (amino acids 305-538), and GST-DIII (amino acids 529 -896), were prepared (Fig. 1A). Correct translation of the fusion proteins was checked by Coomassie Blue staining and/or immunoblotting. The GST-DI protein had the predicted size of 61 kDa and reacted with mAb 6G4 (Fig. 1B, lane 1), indicating that this antibody recognizes the NH 2 terminus of Eps15. The epitope was further mapped to the first 97 NH 2 -terminal amino acids (not shown). The GST-DII fusion protein, visualized by Coomassie Blue staining, had the expected size of 54 kDa (not shown). The GST-DIII protein reacted with a mAb produced against the C terminus of Eps15 (Fig. 1B, lane 3). Its apparent molecular mass of 97 kDa contrasted with its expected size of 67 kDa. However, the delayed migration of GST-DIII is reminiscent of the behavior of the full-length fusion protein which has an expected size of 130 kDa and migrates as a 160 -170-kDa polypeptide in SDS-PAGE even in the absence of posttranslational modification (2). The delayed migration of Eps15 and its third domain is thus likely due to the amino acid composition of the latter domain and particularly to its high number of prolines, 15 of which are adjacent to an acidic residue in the 15 DPF repeats.
GST-DI, -DII, and -DIII fusion proteins and the anti-Eps15 antibody 6G4 were used to precipitate lysates of 35 S-labeled MOLT16 cells ( Fig. 2A). As described previously (6), the anti-Eps15 antibody precipitated two major bands of 140 and 102 kDa (lane 2). The 140-kDa band corresponds to Eps15 since it reacts with 6G4 (6) and can be cleaved by endoprotease-Lys C into the same peptides as the in vitro translation product of eps15 cDNA. 2 The 102-kDa band contains the ␣and ␤ 2adaptins, as previously demonstrated by microsequencing and immunoblotting (6). A large band with the same molecular mass of 102 kDa was also precipitated by GST-DIII (lane 5) but not by GST-DI, GST-DII, or the control GST (lanes 3, 4, and 6). The identity of the 102-kDa band precipitated by GST-DIII with the adaptins of the AP-2 complex was demonstrated by immunoblotting experiments, which showed that it reacts with anti-␣ (Fig. 2B, lanes 1 and 2) and anti-␤ (Fig. 2B, lane 4) but not with an anti-␥ adaptin antibody (Fig. 2B, lane 3). In addition, GST-DIII precipitated bands of 50 and 17 kDa (lane 5). Bands of comparable molecular mass were previously observed in 6G4 immunoprecipitates (6), although only the 50-kDa band is visible in the 6G4 immunoprecipitate shown in this experiment (lane 2). These two bands have a molecular mass compatible with that of the two small components of AP-2, 2 and 2 .
Besides the components of AP-2, GST-DIII coprecipitated polypeptides with molecular masses of 180 and 30 -35 kDa ( Fig. 2A, lane 5). The molecular mass of these polypeptides were comparable to those of the heavy and light chains of clathrin, which interact with the AP-2 complex (13). The presence of clathrin in the precipitate of GST-DIII was confirmed by immunoblotting with the TD.1 antibody specific for the clathrin heavy chain, which detected a 180-kDa band in the GST-DIII precipitate but not in a control GST precipitate (Fig. 2C,   terminal Domain of Eps15-The results described above indicate that the COOH-terminal domain of Eps15 is sufficient for mediating the association between Eps15 and AP-2. To further characterize the region of Eps15 responsible for the association, a series of truncated forms of the GST-DIII proteins was used to precipitate cold cell lysates, and the precipitation of AP-2 was detected by immunoblotting using the 100/2 mAb specific for ␣-adaptin. As summarized in Fig. 3B, a segment comprising amino acids 667-739 and including four of the 15 DPF repeats was required for AP-2 binding. Further trimming of this segment by removing either amino acids 667-675 (Fig.  3A, upper panel, lane 7) or amino acids 712-739 (upper panel, lanes 9 and 16) prevented AP-2 binding. However, the amount of ␣-adaptin precipitated by the fusion protein comprising only amino acids 667-763 (lanes 6 and 15) was significantly less than the amount of ␣-adaptins precipitated by fusion proteins containing a larger NH 2 -terminal fragment of 43 amino acids (lanes 5 and 14). The NH 2 -terminal fragment does not appear to provide a second binding site for AP-2, since GST-529/682 failed to precipitate ␣-adaptin (lane 3). More likely, the presence of this fragment influences the conformation and/or the accessibility of the binding site present on segment 667-739. However, it was not possible to define more precisely the minimal size of the NH 2 -terminal fragment allowing optimal binding of AP-2, since the presence of 6 DPF repeats, very close to each other between amino acids 624 -667, precluded the design of specific primers for fusion proteins of intermediate size. The Eps15-derived Fusion Protein Binds to the Proteolyzed Ear Domain of ␣-Adaptin-Precipitation studies did not allow us to identify the component of AP-2 that binds directly to Eps15, since the four components of AP-2 are dissociated only under strong denaturing conditions and therefore coimmunoprecipitate with Eps15 (6,14,15). Previous studies have shown that AP-2 consists of a brick corelike structure or head with two small appendages or ears linked to the head by hinges containing sites for proteolytic cleavage. The head and ears can thus be separated by limited proteolysis of AP-2. The head consists of the NH 2 -terminal domains of ␣and ␤-adaptins associated with the medium and small chains, whereas the ears correspond to the COOH-terminal domains of the two adaptins. Proteolysis does not affect the interactions between the truncated ␣ and ␤ subunits and the protease resistant 50-and 17-kDa subunits (14,15). Therefore, to define the AP-2 domain that binds to  Eps15, the head and the ears were prepared by limited proteolysis of AP-2 with trypsin. As shown in Fig. 4, lysates treated with trypsin for 20 min at 37°C contained fragments of 60 -65 kDa that were reactive with AC1-M11 and 100/1 antibodies against the head of ␣and ␤-adaptins, respectively (15) (Fig. 4,  b and d, lanes 5), and 40-kDa fragments reactive with 100/2 and Ab32 antibodies against the ears of ␣and ␤-adaptins, respectively (10, 15) (Fig. 4, a and c, lanes 5). When proteolysis was further prolonged, there was a decrease in the amount of immunoreactive 60 -65-kDa fragments, indicating that these fragments were proteolyzed to smaller peptides (not shown). Therefore, control undigested lysates or lysates treated with trypsin for 5, 10, or 20 min were precipitated with GST-DIII (lane 1, 2, 3, and 4, respectively). As shown in Fig. 4, a and b, GST-DIII precipitated undigested ␣-adaptin visible in control precipitates (lanes 1) and in precipitates obtained after a digestion for 5 min (lanes 2) or 10 min (Fig. 4b, lane 3). In addition, GST-DIII precipitated a fragment with a molecular mass of 40 kDa, which was reactive with mAb 100/2 (Fig. 4a,  lanes 2-4) and therefore corresponded to the ear of ␣-adaptin. In contrast, GST-DIII failed to precipitate the other proteolyzed fragments of AP-2. Thus, it precipitated neither the NH 2terminal domain of ␣-adaptin recognized by mAb AC1M11 (Fig.  4b, lanes 2-4), nor that of ␤-adaptin, recognized by mAb 100/1 (Fig. 4d, lanes 3 and 4). Some intact ␤-adaptin and a light band of 60 kDa that reacted with the 100/1 antibody were observed in the GST-DIII precipitate after a 5-min digestion (Fig. 4d, lane 2) but were not detectable after longer times of digestion (Fig. 4d, lanes 3 and 4). Since intact ␣-adaptin remained detectable at 5 and 10 min of digestion (Fig. 4b, lanes 2 and 3), these bands most likely correspond to digested or undigested ␤-adaptin that is still associated with undigested ␣-subunit. Finally, GST-DIII did not precipitate the ear of ␤-adaptin detected by the Ab32 antiserum (Fig. 4c). These results indicate that Eps15 specifically binds the ear of ␣-adaptin.
GST Fusion Proteins Derived from the COOH-terminal Domain of ␣-Adaptin Bind Eps15-To confirm the results of the proteolysis studies, a GST-fusion protein encompassing the entire COOH terminus of ␣-adaptin and including the hinge region was derived from mouse ␣-adaptin (Fig. 5A). We observed that a large fraction of the ␣-adaptin ear was cleaved from the GST moiety during purification of the GST fusion protein and released in the bacterial lysate (not shown). This proteolytic cleavage was probably due to the presence of bacterial proteases and could not be prevented by the use of exogenous protease inhibitors. Therefore, to demonstrate the association of Eps15 with the ␣-adaptin ear, bacterial lysates containing the ␣-adaptin ear were precipitated with Eps15derived GST fusion proteins. After three clearing cycles with glutathione-Sepharose 4B beads to remove free GST, the supernatant was analyzed by Western blotting using the anti-␣adaptin ear antibody 100/2. This antibody detected a large band with a molecular mass of 40 kDa, corresponding to the ␣-adaptin ear, and lighter bands of 37 and 67 kDa, corresponding, respectively, to a proteolyzed fragment of the ␣-adaptin ear (15) and to the intact GST ␣-adaptin ear fusion protein (GST-␣ ear) which had not been entirely removed by the clearing procedure (Fig. 5B, lane 1). The same supernatant was then precipitated with GST-DIII (lane 3), or GST-DI as a control (lane 2). The complete GST-␣ ear was precipitated nonspecifically by glutathione-Sepharose 4B beads and was present in both the GST-DIII precipitate and the control GST-DI precipitate. In contrast, the 40-kDa ␣-adaptin ear fragment bound to GST-DIII (lane 3) but not to GST-DI (lane 2), confirming the specific interaction between the COOH-terminal domain of Eps15 and the ear of ␣-adaptin. GST-DIII did not precipitate the 37-kDa fragment derived from the ear, suggesting that this smaller fragment did not contain the binding site for Eps15. To avoid proteolytic cleavage of GST-␣ ear, smaller GST fusion proteins were designed. In the absence of the hinge region, the GST fusion proteins were not sensitive to protein degradation and could therefore be used to precipitate Eps15 from cell lysates. Eps15 could be precipitated by a GST fusion protein containing the ␣-adaptin ear without the hinge region (amino acids 706 -938) (Fig. 5C, lane 2). In contrast, it was not precipitated by fusion proteins with an additional deletion of 50 amino acids or more on the COOH-terminal side (Fig. 5C, lane  1 and not shown). Together with the lack of precipitation of the 37-kDa fragment derived from the GST-␣ ear, this result indicates that Eps15 binding requires the proximal part of the ␣-adaptin ear.  DISCUSSION We have recently demonstrated a specific and constitutive interaction between Eps15 and the plasma membrane adaptor, AP-2 (6). In the present study, the binding site of Eps15 for AP-2 was localized to a domain of 72 amino acids in the COOH terminus of Eps15. In addition, we have identified ␣-adaptin as the component of AP-2 that binds Eps15 and have localized the binding site within its ear domain. In addition, we have provided preliminary evidence that Eps15 can associate with a fraction of AP-2 bound to clathrin.
The predicted primary sequence of Eps15 identifies a modular protein with three domains, and each may be involved in specific protein interactions (1-3) (Fig. 6A). An EH domain comparable to those observed in the NH 2 terminus of Eps15 is observed in End3p (1), a yeast protein involved in internalization of the ␣-factor (5), and the study of deletion mutants suggests that the EH domain of End3p is required for normal endocytosis (5). To define which domain of Eps15 interacts with AP-2, GST fusion proteins derived from each of the three domains of Eps15 were tested for their ability to precipitate AP-2 from cell lysates. No interaction could be demonstrated between AP-2 and the central domain of Eps15 or its NH 2 terminus. This result does not exclude a role for the EH domains of Eps15 in endocytosis, but indicates that they do not bind AP-2. In contrast, the GST fusion protein encoding the COOH-terminal domain of Eps15 precipitated ␣and ␤2-adaptins very efficiently, indicating that this domain contains the binding site for AP-2. The minimal region required for this interaction was defined using a series of GST fusion proteins derived from the COOH terminus of Eps15. The smallest fusion protein able to precipitate AP-2 comprised amino acids 667-739 and included only four of the DPF repeats among the 15 present in human Eps15. Its sequence is shown in Fig. 6A. The binding site of AP-2 was thus close but distinct from the binding site of Crk (residues 765-771) (1), indicating that these proteins bind Eps15 independently. A search of the data banks using the Blast program revealed 90% identity with the corresponding sequence in the murine Eps15 protein, which also interacts with AP-2 (6). In contrast, the sequence was poorly conserved in Eps15r, the homology being largely related to the presence of DPF repeats. Furthermore, there was no significant homology with other known proteins, suggesting that the association of Eps15 with AP-2 has a very specific function.
Clathrin-coated pits and vesicles bud from two membrane compartments, the plasma membrane and the trans-Golgi network. Two distinct adaptor complexes link the clathrin lattice to the appropriate membrane: AP-2, associated with the plasma membrane coated vesicles, and AP-1, associated with trans-Golgi coated vesicles (7,8). Both adaptors are heterotetramers consisting of two 90 -110-kDa adaptins (␣ and ␤ 2 for the plasma membrane, ␥ and ␤ 1 for the trans-Golgi) complexed with two smaller proteins of 48 -50 and 16 -17 kDa (1, 2 and 1, 2 respectively) (7,8). In agreement with our previous observations (6), Eps15-derived fusion proteins precipitated four proteins with molecular masses consistent with the four components of the adaptor complexes. The ␣and ␤-adaptins, but not ␥-adaptin, were found in the GST-Eps15 precipitate, a result that confirms the specific association of Eps15 with AP-2 but does not allow us to determine which subunit of AP-2 interacts with Eps15. Several protein associations with components of the adaptor complexes have already been described. The ␤1 and␤2 subunits, which are 85% identical (14,16), mediate binding of both AP-1 and AP-2 adaptors to clathrin and promote clathrin coat assembly (17)(18)(19). The 1 and 2 subunits, which share 40% identity (20,21), interact with tyrosinebased signals of several integral membrane proteins (22). The ␣and ␥-adaptins have an overall identity of only 25%, mainly restricted to the NH 2 domain (23). This domain contains sequences which simultaneously determine coassembly with the correct and subunits and targeting to the appropriate membrane (24,25). In addition, the NH 2 -terminal domain of ␣-adaptin binds to clathrin cages (26) and contains a binding site for polyphosphoinositols which, in vitro, inhibit AP-2 selfassociation, binding of AP-2 to clathrin, and clathrin coat assembly (27,28). In contrast to the other components of adaptor complexes, the ears of ␣and ␥-adaptins show no homology, suggesting that they have distinct functions (23). Precipitation of the fragments of AP-2 released by limited proteolysis indicates that the Eps15-derived fusion proteins bound to the ear of ␣-adaptin. This result was confirmed by precipitation of Eps15 with a fusion protein encompassing the ear of ␣-adaptin C. Furthermore, the ear of ␣-adaptin C released in the lysate of transformed bacteria could be precipitated by a GST-Eps15 fusion protein, demonstrating that Eps15 and ␣-adaptin interact directly. These data are consistent with the specificity of Eps15 for AP-2 and supports the hypothesis that the ear of ␣-adaptin is endowed with a specific function. A model summarizing the interaction of Eps15 with AP-2 is shown in Fig.  6B.
The association of Eps15 with AP-2 and its homology with End3p strongly suggest that this protein has a function in clathrin-mediated endocytosis. Furthermore, the presence of the clathrin heavy chain in the GST-DIII precipitate indicates that Eps15 may be a component of clathrin-coated pits and vesicles where AP-2 and clathrin may interact, a hypothesis supported by preliminary electron microscopic data. 3 Eps15 might thus be related to the unidentified 150-kDa protein observed by Beck and Keen (18) in AP-2 aggregates and in coated vesicles and/or to the 140-kDa protein observed by Lindner and Ungewickell (29) among the components of bovine clathrin-coated vesicles. The putative function of Eps15 in endocytosis remains to be defined. Our study, which delineates the domain of Eps15 involved in AP-2 binding, provides the basis for the design of mutated proteins that could elucidate the in vivo function of Eps15.