Dap160, a Neural-specific Eps15 Homology and Multiple SH3 Domain-containing Protein That Interacts with DrosophilaDynamin*

The discovery of overlapping hot spots of dynamin (Estes, P. S., Roos, J., van der Bliek, A., Kelly, R. B., Krishnan, K. S., and Ramaswami, M. (1996) J. Neurosci. 16, 5443–5456) and the heterotetrameric adaptor 2 complex (Gonzalez-Gaitan, M., and Jäckle, H. (1997)Cell 88, 767–776) in Drosophila nerve terminals led to the concept of zones of active endocytosis close to sites of active exocytosis. The proline-rich domain of Drosophiladynamin was used to identify and purify a third component of the endocytosis zones. Dap160 (dynamin-associated protein 160 kDa) is a membrane-associated, dynamin-binding protein of 160 kDa that has four putative src homology 3 domains and an Eps15 homology domain, motifs frequently found in proteins associated with endocytosis. The binding capacities of the four putative src homology 3 domains were examined individually and in combination and shown to bind known proteins that contained proline-rich domains. Each binding site, however, was different in its preference for binding partners. We suggest that Dap160 is a scaffolding protein that helps anchor proteins required for endocytosis at sites where they are needed in the Drosophilanerve terminal.

Endocytosis of receptors such as the transferrin receptor from the plasma membrane is constitutive, whereas internalization of receptors such as the ␤-adrenergic receptor is stimulated by ligand binding (3). A third type of endocytosis, compensatory endocytosis, recovers the extra membrane added to the plasma membrane when exocytosis is stimulated in a regulated secretory cell. All three types of endocytosis appear to use a similar coating mechanism that involves the heterotetrameric adaptor 2 complex (AP2), 1 clathrin and dynamin (4). One apparent difference between the three is that compensatory endocytosis is faster than the other two (5). A possible explanation of the efficiency of compensatory endocytosis is that the machinery of endocytosis is not free in the cytoplasm but is concentrated in the active zones, the sites of exocytosis of synaptic vesicles. The first indication of this was the discovery in Drosophila nerve terminals that dynamin, a large GTPase known to be necessary for synaptic vesicle biogenesis, is restricted to presynaptic sites that coincide with clusters of synaptic vesicles around active zones (1). Dynamin concentrations at the plasma membrane are even more obvious (1) when Drosophila shibire ts1 mutants, temperature-sensitive in dynamin function, are depleted of synaptic vesicles by stimulating exocytosis at nonpermissive temperatures (6). The idea that these hot spots are indeed loci of intense endocytosis received strong support when it was found that AP2 co-localized with dynamin at these spots (2). These authors reported that AP2 distribution was even more tightly localized to the hot spots than dynamin.
It thus seems plausible that Drosophila neurons have specialized sites on the plasma membrane at or close to the sites of exocytosis, where the endocytotic machinery accumulates. One way to understand how such an endocytotic machine might operate is to identify the other components of the complex, particularly those that might anchor it to the plasma membrane. In mammalian cells the C-terminal, proline-rich domain (PRD) of dynamin is involved in membrane anchoring since removing it inhibits the association of dynamin with the plasma membrane (7,8). Consistent with a role for the dynamin PRD in anchoring is the observation that the src homology 3 (SH3) domain of amphiphysin helps anchor dynamin via its PRD to neuronal plasma membranes in lamprey nerve terminals (9). To find how Drosophila dynamin might be anchored at endocytotic hot spots, it seemed reasonable to look for a protein on Drosophila membranes to which the proline-rich domain of Drosophila dynamin might bind.
By using this approach, a new peripheral membrane protein has been identified that can be released from the membranes in a complex with Drosophila dynamin. Sequencing revealed that the protein is multimodular with four contiguous SH3 domains and an Eps15 homology (EH) domain. The SH3 domains were shown to be capable of binding dynamin and other PRD-containing proteins. Each SH3 domain, however, had a different repertoire of binding partners. EH domain-containing proteins have been implicated in receptor and fluid phase endocytosis in yeast (10). Furthermore, the progenitor of the family, Eps15, binds AP2 (11,12) and clusters near the neck of clathrin-coated pits (12). A ligand for the EH domain in the new protein has not yet been identified. One of only two proteins thus far demonstrated to bind to a GST-EH fusion protein is the Drosophila Numb protein (13). Numb, in turn, is membrane-associated, possesses a phosphotyrosine-binding (PTB) domain, and associates with Numb-associated kinase (14). Thus, EH domain-and SH3 domain-containing proteins may function as either scaffolding or anchoring proteins. The multimodular protein we describe has, therefore, the capacity to bind simultaneously to several different proteins at once, and to keep them on a membrane.
We refer to the protein we have isolated as dynamin-associated protein 160 kDa (Dap160). By immunofluorescence, Dap160 was restricted to the synaptic nerve terminal, or boutons, of Drosophila neuromuscular junctions. In addition it was found in spots that were associated with active zones in the resting nerve terminal and in membrane-associated "hot spots" after depletion of vesicle content by stimulating Drosophila shibire ts1 mutants at the nonpermissive temperature. The observed distribution of Dap160 is indistinguishable from that reported for dynamin (1) and AP2 (2). Thus, Dap160 has the properties expected of a membrane-associated scaffolding protein that helps hold proteins required for endocytosis close to where they are needed both before and during the recovery of synaptic vesicle membrane proteins. In so doing it could enhance the efficiency or the specificity of compensatory endocytosis.
Generation of Antibodies and Affinity Purification-Polyclonal antibodies to Drosophila dynamin and Dap160 were raised against purified recombinant fusion protein. pJR2 was expressed and sent to Immuno-Dynamics, Inc. (La Jolla, CA) for antibody production, which generated rabbit polyclonal antibodies Ab2074 and Ab2075. pJR73 was expressed and sent to Alpha Diagnostic Inc. (San Antonio, TX) for antibody production, which generated rabbit polyclonal antibodies Ab1703 and Ab1704. Antibodies were affinity purified by the method of Smith and Fisher (16). To affinity purify anti-dynamin antibodies from Ab2074 sera, MBP-Ddyn was overexpressed in bacteria and recovered as a sarkosyl extract (17). To affinity purify anti-Dap160 antibodies from Ab1703 sera, GST-p160(JR120) was overexpressed in bacteria and purified as described above. Sarkosyl extract containing MBP-Ddyn or purified recombinant GST-p160(JR120) fusion protein was resolved on a 5-15% SDS-PAGE preparative mini-gel and transferred to nitrocellulose using standard methods (18). The nitrocellulose bearing the fusion protein was excised, blocked in PBS, 0.05% Tween 20, 1% BSA, washed with PBS, 0.05% Tween 20, 0.1% BSA, and then incubated with sera overnight at 4°C. The antibody-bound nitrocellulose was washed three times with PBS, 0.05% Tween 20, 0.1% BSA and then eluted with 300-l aliquots of 5 mM glycine, pH 2.3, 500 mM NaCl, 0.5% Tween 20, 100 mg/ml BSA. The eluates were immediately neutralized with 15.75 l of 1 M Na 2 HPO 4 . To affinity purify anti-GST antibodies, GST was overexpressed in bacteria. Soluble recombinant protein was purified as described for GST-p160(JR120) above. Affinity purification proceeded as described above, using Ab1703 as the sera containing the anti-GST antibodies.
Anti Overlay Analysis-Samples for overlay analysis were prepared by resolving 20 g of protein on 5-15% SDS-PAGE gels. The gels were electroblotted onto nitrocellulose and briefly stained with 0.2% Ponceau S. For single sample overlay analysis, each lane of the blot was excised, blocked with 5% nonfat dry milk powder in PBS, 0.05% Tween 20 (BLOTTO), and incubated with 40 g/ml fusion protein overnight at 4°C. Fusion protein binding was determined by incubating the overlays with affinity purified anti-GST antibodies and detected with a goat anti-rabbit horseradish peroxidase-conjugated secondary antibody (Cappel/ICN, Aurora, OH) and developed with ECL detection system (Amersham Pharmacia Biotech) according to the manufacturer's directions.
Biochemical Fractionation-All procedures were conducted on ice or at 4°C. Heads from 10 ml packed wild-type (OR) adult flies, flashfrozen in liquid nitrogen and stored at Ϫ80°C, were isolated using a molecular sieve (Fisher) and kept cold with liquid nitrogen. The heads were then ground using a mortar and pestle and subsequently homogenized in 2 ml of either Hepes buffer (10 mM Hepes, pH 7.4, 1 mM EGTA/0.1 mM MgCl 2 ) or Buffer A (10 mM Hepes, pH 7.4, 150 mM NaCl, 1 mM EGTA, 0.1 mM MgCl 2 ) ϩ protease inhibitors (leupeptin, 5 mg/ml, antipain, 2 mg/ml, benzamidine, 10 mg/ml, aprotinin, 10 mg/ml, pepstatin, 5 mg/ml, chymostatin, 1 mg/ml, and 1 mM phenylmethylsulfonyl fluoride) with 20 strokes using a Teflon Dounce homogenizer. The homogenate was centrifuged at 1000 ϫ g in an Eppendorf microcentrifuge for 20 min, resulting in an S1 and P1 fraction. The S1 fraction was then centrifuged at 25,000 ϫ g in a Beckman TL100 tabletop ultracentrifuge for 30 min, resulting in an S2 and a P2 fractions. The S2 fraction was centrifuged at 208,000 ϫ g in a Beckman TL100 tabletop ultracentrifuge for 60 min, resulting in an S3 and a P3 fraction.
To generate rat brain cytosol, rat brains (PelFreeze, Rogers, AR) were diced and homogenized in Buffer A. The resulting homogenate was then fractionated as described above. Further fractionation of the cytosol was achieved by charging 50 mg of cytosol on a MonoQ 5/5 anion exchange FPLC column (Amersham Pharmacia Biotech) and eluting the bound proteins with a 20-ml gradient (Buffer 1, Hepes buffer; Buffer 2, Hepes buffer, 1 M NaCl).
Analytical Size Fractionation-5 mg of Drosophila S3 extract was resolved on a 24-ml Superdex 200 FPLC column (Amersham Pharmacia Biotech) using either a Bio-Rad BioLogic workstation or Amersham Pharmacia Biotech LC501 FPLC. Protein samples were eluted from the column with Buffer A at a flow rate of 0.35 ml/min, and 0.5-ml fractions were collected. 30-l samples from each fraction beginning with the void volume fraction were then analyzed by SDS-PAGE, visualized with SYPRO Orange protein stain (Bio-Rad), and either Western blotted or processed for overlay analysis as described above. Immunoprecipitation Methods-Superdex 200 gel filtration fractions immunopositive for dynamin and Dap160 were diluted to 1 ml with IP Buffer (50 mM Tris, pH 7.4, 0.1 mM EDTA, 0.5% Tween 20). 25 g of affinity purified anti-dynamin-specific polyclonal antibodies or equivalent preimmune antibodies were added. Immunoprecipitation reactions were incubated overnight at 4°C with tumbling end over end. Immunoprecipitations were processed by addition of 30 l of a 50% protein A-Sepharose slurry, incubation at 4°C with tumbling end over end for 2 h, and collected by centrifugation at 1000 ϫ g for 5 min. Beads were washed three times with 1 ml of IP Buffer and eluted with SDS-PAGE sample buffer. Eluates were analyzed by resolving on a 5-15% gradient SDS-PAGE gel and immunoblotted using anti-Dap160 and anti-dynamin-specific antibodies.
Purification of Dap160 -Heads from 2.5 liters of adult Drosophila Oregon-R flies were isolated. This resulted in 40 g of packed fly heads. The heads were ground in a mortar and pestle and homogenized in 500 ml of Hepes buffer. The extract was then centrifuged at 100,000 ϫ g for 1 h. The pellet was extracted with 500 ml of Buffer A. Triton X-100 was added to 1%, and the extract was incubated for 1 h at 4°C and then centrifuged at 100,000 ϫ g. The resulting supernatant was precipitated with 30% ammonium sulfate. The 30% ammonium sulfate pellet was resuspended in 15 ml of Buffer A and charged on a 250-ml Superose 6 (Amersham Pharmacia Biotech) column and eluted with Buffer A. Three-ml fractions were collected and analyzed for the presence of Dap160 by the overlay assay. Dap160-positive fractions were then charged onto a DEAE-Sepharose column in Buffer A, washed extensively, and eluted with a Buffer A, 400 mM NaCl isocratic buffer flow. The DEAE eluate was concentrated, and 5-mg aliquots were charged onto a Superdex 200 FPLC column and eluted with Buffer A. Dap160positive fractions, as determined by the overlay assay, were then charged onto a MonoQ FPLC column and eluted with a Buffer A 3 Buffer A, 1 M NaCl gradient. Dap160-positive fractions were pooled and resolved on a 5-15% SDS-PAGE gel and stained with SYPRO Orange, and the Dap160 band was excised from the gel. This band was then sent to the University of Michigan PCSF facility for protein microsequencing.
Cloning of Dap160 -mRNA from adult Drosophila (OR) heads was isolated using TRI-zol reagent (MBP, San Diego). First-strand cDNA was generated (45) using oligo(dT) primer and SuperScriptII RT (Life Technologies, Inc.) as per the manufacturer's suggestions. 1 ng of cDNA template was then used for PCR. From the peptide microsequencing data, the following degenerate oligos were designed. Peptide KIPVTLPQEW generated 5Ј-AAR ATH CCN GTN ACN YTN CCN CAR GAR TGG-3Ј and was labeled 25F. The reverse complement corresponding to peptide IKEQNAKLPQ generated 5Ј-YTG NGG NAR YTT NGC RTT YTG YTC YTT DAT-3Ј and was labeled 4050 RC. Polymerase chain reaction generated a 768-bp PCR product whose translation product contained both peptide sequences from the protein microsequencing project. Additional sequencing data were generated by PCR using a gt11 expression library constructed by Ito et al. (19) as template and oligo JR116 (5Ј-CTC TGC TTG ACC CTT AA-3Ј) and the 5Ј gt11 sequencing primer. The full-length sequence was generated by using the 3Ј-and 5Ј-RACE kit (Life Technologies, Inc.). For 5Ј-RACE, oligo JR118 (5Ј-TAA GAG ACG ACA GCA GGC T-3Ј) was used first in conjunction with the abridged anchor primer in PCR round 1 and then with the AUAP primer in PCR round 2. For 3Ј-RACE, oligo JR108 (5Ј-AAA GCA GAA GGC ACA CA-3Ј) was used with the 3Ј anchor primer in PCR round 1 and then oligo JR123 (5Ј-GCC CAC AAG CAG TTA AT-3Ј) was used with the AUAP primer in PCR round 2. All PCR was performed with the EXPAND High Fidelity PCR kit (BMB, Indianapolis, IN), and PCR products were subsequently subcloned into the pT-Adv cloning vector (CLONTECH) and sequenced using an ABI model 373A DNA sequencing machine by Yi Zhang in the UCSF/HRI sequencing facility. A full-length clone was generated by PCR from Drosophila cDNA using oligo JR131 (5Ј-noncoding region) and oligo JR139 (3Ј-noncoding region) and sequenced to confirm the sequence of Dap160.
Confocal Microscopy Studies-Third instar larval neuromuscular preparations were generated as described (1) for antibody staining and protein localization studies using a Leica TCS NT laser confocal micro-scope. Briefly, wild-type (Oregon-R) or shibire ts1 crawling third instar larvae were pinned dorsal side up and filleted in calcium-free saline (130 mM NaCl, 36 mM sucrose, 5 mM KCl, 5 mM Hepes, pH 7.3, 4 mM MgCl 2 , 0.5 mM EGTA). Viscera were removed, allowing visualization of type 1b synaptic boutons innervating the ventral longitudinal muscles 6 and 7. shibire ts1 -depleted nerve terminals were stimulated in high K ϩ saline (75 mM NaCl, 36 mM sucrose, 60 mM KCl, 5 mM Hepes, pH 7.3, 2 mM MgCl 2 , 2 mM CaCl 2 ) at 30°C for 10 min. After stimulation, the high K ϩ saline was replaced briefly with prewarmed calcium-free saline prior to fixation. Double-labeling experiments were performed using either affinity purified anti-dynamin or anti-Dap160 polyclonal antibodies in conjunction with anti-csp or anti-SAP-47 monoclonal antibodies. Secondary antibodies used in this study were goat anti-rabbit Texas Red (Cappel/ICN) and goat anti-mouse-fluorescein isothiocyanate (Cappel/ICN). Images were acquired using the Leica TCS software package. All images were viewed with a 100 ϫ 1.4NA lens and the zoom set to 4. Z series were taken where the focal plane was stepped by 0.25 m between images.

Characterization of GST-Ddyn(PRD) in Rat Brain Cytosol
Extracts-When a protein such as dynamin is known to have a role in synaptic vesicle recycling, the fusion protein overlay (or far-Western) technique can be used to identify additional interacting proteins that may mediate recycling. The identification of binding partners for proteins enriched in the mammalian nerve terminal has focused on SH3-containing proteins, such as Grb2 and amphiphysin, that bind proline-rich domains present in synaptojanin and dynamin (20 -26). A subset of the SH3-containing proteins identified by a polyproline peptide phage display screen (27) has been shown to bind to synaptojanin and is implicated in binding to dynamin (23,24).
We have used the overlay technique to identify dynaminbinding proteins in Drosophila, hoping to take advantage of its favorable genetics. To detect proteins that contain SH3 domains, the proline-rich domain of Drosophila dynamin (Ddyn-(PRD)) was fused to GST. Initial characterization of the GST-Ddyn(PRD) fusion is shown in Fig. 1A. As a positive control, GST-Amph(SH3) (28) overlays on rat brain cytosol identified two proteins, a 145-kDa protein and a 100-kDa protein (lane 2). These two proteins are predicted to be synaptojanin and dynamin, as reported by David et al. (22). GST-Ddyn(PRD) interacted with three major proteins in rat brain cytosol of 120, 55, and 40 kDa (lane 3). A minor band was also seen at 80 kDa. The mobilities of the four dynamin PRD interacting proteins closely resemble the migration pattern of four proteins recently characterized by McPherson and colleagues (23) as synaptojaninbinding proteins. To confirm the identity of these four proteins, rat brain cytosol was fractionated on a MonoQ anion exchange column, and fractions were assayed by fusion protein overlays and by Western blotting using antibodies directed against three of the synaptojanin-binding proteins. The 120-and 80-kDa proteins co-migrate with amphiphysin I and amphiphysin II, respectively (Fig. 1B, lanes 1-4). Assays on the MonoQ fraction containing the 55-and 40-kDa dynamin-interacting proteins support the conclusion that endophilin (SH3p4), a 40-kDa protein shown to interact with synaptojanin and dynamin (23)(24)(25), is one of the four dynamin-interacting proteins identified by the GST-Ddyn(PRD) reagent (Fig. 1B, lanes 5 and  6). The identity of the 55-kDa protein is currently being characterized and will be reported on in a separate study.
The interaction between the SH3 domain of amphiphysin I and the proline-rich domain of mammalian dynamin has already been described (9,22,29). Drosophila dynamin was not predicted to bind to amphiphysin I since the amphiphysinbinding site in the proline-rich domain of mammalian dynamin I, PSRPNR (29), is not conserved in the sequence of Drosophila dynamin (Fig. 1C, thick arrow, class 1 site). However, a less conserved class 1 site upstream from the C terminus of Dro-sophila dynamin, PAIPNR, may mediate the binding of Drosophila dynamin to rat amphiphysin in vitro.
Identification of p160 as a Major Dynamin Binding Protein in Drosophila-The proline-rich domain of Drosophila dynamin bound at least three proteins in rat brain that have already been implicated in endocytosis. Thus, the protein interactions needed for synaptic vesicle endocytosis have been highly conserved during evolution. We expected to use the GST-Ddyn(PRD) fusion to identify the Drosophila homologues of the amphiphysins, p55 and endophilin. Drosophila head extracts were resolved on SDS-PAGE gels and assayed for binding to GST-Ddyn(PRD), using GST alone as a negative control ( Fig. 2A) and GST-grb2 as a positive control (Fig. 2B). As expected, grb2 bound dynamin but also an unknown 50-kDa protein present in membrane fractions. Overlays on Drosophila head extracts using GST-Ddyn(PRD) showed that the dynamin PRD did not recognize homologues of the rat proteins but bound specifically to a 160-kDa protein (Fig. 2C) that was present in soluble and membrane fractions, under these particular extraction conditions. The protein has been named Adult Drosophila heads were homogenized in Buffer A and fractionated by differential speed centrifugation. S1 and P1 represent 1000 ϫ g supernatant and pellet, respectively. S2 and P2 represent 25,000 ϫ g supernatant and pellet, respectively, and S3 and P3 represent 208,000 ϫ g supernatant and pellet, respectively. 20 g from each fraction was resolved on 5-15% SDS-PAGE gels, transferred to nitrocellulose, and incubated with 1 mg of indicated fusion protein. gel filtration chromatography, MonoQ anion exchange chromatography, and Superdex 200 FPLC gel filtration chromatography resulted in the isolation of 75 g of Dap160 which was submitted for internal peptide sequencing. Amino acid sequence was obtained from three CnBr digestion-generated peptides. Degenerate oligos corresponding to a 10-amino acid segment from two of the peptides were synthesized and used to isolate a partial cDNA by PCR from Drosophila adult head cDNA. A 768-bp PCR product encoded two of the peptides obtained from protein sequencing. Sequence-specific oligos from this PCR product were used to generate additional sequence by nested PCR from a gt11 cDNA library. Finally, a full-length clone was obtained by 5Ј-and 3Ј-RACE. Two independent full-length clones were isolated and are depicted in Fig. 3A. The sequence of clone 1 is shown in Fig. 3B. Sequence was confirmed by using polymerase chain reaction and sequence-specific primers to the 5Ј-and 3Ј-noncoding regions to obtain a full-length cDNA from an adult fly head cDNA.
A schematic of the predicted primary structure of Dap160 is shown in Fig. 3C. Dap160 possesses two predicted N-terminal EF-hand, calcium-binding domains. The second EF-hand domain is situated within an EH domain (Eps15 homology domain). The middle third of the Dap160 amino acid sequence has a high percentage of acidic and basic amino acid residues (acidic/basic domain, or A/B domain), yet is overall neutral. In fact, the pI of the first 650 amino acids is slightly basic (pI ϭ 8.8). The C terminus of Dap160 has four consecutive SH3 domains.
The SH3 domains of Dap160 are abundant in acidic amino acid residues, giving Dap160 an overall pI ϭ 4.9. The predicted molecular mass of Dap160 is 120 kDa, yet the mobility of Dap160 on SDS-PAGE is 160 kDa. This aberrant mobility on SDS-PAGE may be due to the abundance of negatively charged amino acids in the SH3 domains. Similar aberrant mobilities of negatively charged proteins have been reported previously for amphiphysin and stoned A (30,31). The consecutive SH3 domains of Dap160 share homology to SH3p17 and SH3p18 (Fig.  3D), two partial human cDNA clones isolated from a polyproline peptide phage-display screen for SH3-containing proteins (27). SH3p17 has four consecutive C-terminal SH3 domains, whereas SH3p18 has three consecutive SH3 domains. The Cterminal regions are the most highly conserved feature of all three proteins; homologous SH3C and SH3D domains are separated by 20 amino acids. Moreover, the separation of SH3B and SH3C is also maintained when the shorter splice variant of Dap160 is included.
Recently, the sequence for Xenopus intersectin has been reported (accession no. AF032118). The putative coding sequence of intersectin has an EH domain and five C-terminal SH3 domains. The homology between the EH domain of Dap160 and intersectin (60% overall) and between the SH3 domains of Dap160, SH3p17, SH3p18, and intersectin suggest that the four proteins represent a family of multiple SH3-containing proteins, highly conserved through evolution.
Dap160 and Dynamin Form a Membrane-associated Complex-To study the cellular distribution of Dap160, a polyclonal antibody was generated. From the partial cDNA clone of Dap160, a fusion protein to the acidic and basic domain was constructed (GST-p160(JR120)), expressed, and used as an antigen to generated polyclonal antisera to Dap160. Dap160 antisera was affinity purified and used to immunoblot subcellular  (Fig. 4B). The antibodies identified a protein doublet that is predicted to be the two isoforms of Dap160. The sera recognized a protein with the same subcellular distribution as was already seen in the GST-Ddyn(PRD) overlay assay (Fig. 2). When Drosophila heads were homogenized in low ionic strength Hepes buffer (Fig. 4A), most of dynamin and Dap160 are associated with membranes, with little recovered in the soluble fraction (S3). Thus, dynamin and Dap160 co-distributed under both medium and low salt conditions, and both were membrane-bound at low ionic strength.
To determine if Dap160 was in a complex with dynamin, membranes from Drosophila heads were extracted using medium ionic strength buffer. Five mg of the resulting extract was resolved on a Superdex 200 FPLC column. Fractions were collected and immunoblotted for both Dap160 and dynamin (Fig. 5). By gel filtration analysis, dynamin existed in heterogeneously sized high molecular weight complexes. Dap160, on the other hand, exists only in one high molecular weight fraction. In this latter high molecular weight fraction, both dynamin and Dap160 were present. The predicted molecular mass of this fraction is approximately 500 kDa. The presence of these two proteins in the high molecular weight fraction was salt-sensitive, for when membranes were extracted in a 300 mM salt buffer, both Dap160 and dynamin migrated at their predicted molecular weights (data not shown).
Dynamin had previously been shown to form high molecular weight oligomers in a salt-sensitive manner (32). To determine if Dap160 and dynamin were present together in the same 500-kDa complex or were in separate high molecular weight complexes, Superdex 200 fractions positive for Dap160 and dynamin were incubated with either preimmune control antisera or anti-dynamin-specific antibodies (Ab2074). Immunoprecipitates were then blotted for dynamin and Dap160. Both proteins were precipitated by dynamin-specific antibodies and were not present in the preimmune control lane (Fig. 6). Thus, dynamin and Dap160 are present together in a high molecular weight complex. They are membrane-associated in low ionic strength media, extract together as a complex from membranes in medium ionic strength buffer, and dissociate in high salt buffer.
Localization of Dap160 in Drosophila Third Instar Neuromuscular Junctions-The localization of Dap160 at neuromuscular junctions was visualized in Drosophila third instar neuromuscular junctions. To determine if Dap160 was enriched in neuronal tissue, the distribution of Dap160 was observed by high resolution laser confocal microscopy in resting synapses immunostained for both dynamin and Dap160. Antibodies to Dap160 exclusively labeled strings of synaptic terminals, or boutons, as identified by nerve terminal-specific antibodies such as anti-SAP-47 (Fig. 7A) or anti-cysteine string protein (csp) (33). Dap160 immunoreactivity was not detected in either muscle or the axon. Dap160 was found to be restricted to presynaptic sites associated with the plasma membrane. These sites co-distributed with, but did not have the same morphology as, the clusters of synaptic vesicles detected by both synaptic antigens. In each case Dap160 immunoreactivity was constrained in its distribution to domains within the synaptic vesicle clusters and seemed to be more tightly associated with the plasma membrane.
Similar distributions have also been seen in Drosophila resting nerve terminals stained with antibodies to dynamin (1) and to the heterotetrameric adaptor protein, AP2 (2). Comparison of the localization of dynamin (Fig. 7B) to that of Dap160 showed no discernible difference. Thus, Dap160 and dynamin have similar distributions in resting nerve terminals, consistent with the biochemical data presented in this paper.
To determine if the localization pattern observed for Dap160 and dynamin changed in different phases of the synaptic vesicle cycle, the distribution of Dap160 was examined in nerve terminals of shibire ts1 flies, which were depleted of their synaptic vesicle content by stimulation in high K ϩ media at the nonpermissive temperature for 10 min. Under these conditions, synaptic vesicles disappeared, and their proteins were recovered on the plasma membrane (34). By confocal microscopic analysis using antibodies to the synaptic vesicle cysteine string protein, the nerve terminals appear to have expanded in diameter. The csp immunoreactivity was transferred from synaptic vesicle clusters to the plasma membranes where it could diffuse out of the bouton (Fig. 7C), consistent with earlier findings (1,35). Both dynamin and Dap160 are now more clearly seen to be in spots, termed hot spots (1), around the perimeter of the enlarged nerve terminal (Fig. 7C). There was little lateral diffusion of either Dap160 or dynamin immunoreactivity along the membrane; the Dap160 and dynamin distribution was significantly different from plasma membrane and synaptic vesicle antigens. This suggests that Dap160 and dynamin are  Fig. 5) were diluted to 1 mg/ml with IP buffer. Antibodies were added, and the reactions were incubated with tumbling at 4°C overnight. Co-precipitating proteins were captured by addition of 30 l of 50% protein A-Sepharose slurry. Immunoprecipitation complexes were eluted from the protein A-Sepharose beads with 50 l of SDS-PAGE sample buffer and resolved on 5-15% SDS-PAGE gel, transferred to nitrocellulose, and blotted with anti-dynamin and anti-Dap160-specific antibodies.
closely associated with each other, both prior to synaptic vesicle recovery in stimulated shibire flies as well as in resting terminals.
Biochemical Characterization of the SH3 Domains of Dap160 -The primary structure of Dap160 predicted two EFhand, Ca 2ϩ binding domains, an EH domain, and four consec-FIG. 7. Co-localization of dynamin and Dap160 in Drosophila third instar larval nerve terminals. Third instar larvae from both wild-type (OR) flies and shibire ts1 flies were dissected as described. Larval neuromuscular junctions were double-labeled with the indicated primary antibodies. A, wild-type, resting nerve terminals were labeled with anti-Dap160 antibodies and anti-SAP-47 antibodies. B, wild-type, resting nerve terminals were labeled with anti-dynamin antibodies and anti-SAP-47 antibodies. A and B, Dap160 and dynamin localize to spots on the plasma membrane as well as the interterminal space. The red (Dap160 or dynamin) and green (SAP-47) overlay of the two images show that Dap160 and dynamin immunoreactivity overlaps with synaptic vesicles in Drosophila nerve terminals. Additionally, Dap160 and dynamin can be seen to localize preipherally at the membrane of the nerve terminal. C, third instar larvae from shibire ts1 flies were dissected and incubated at 30°C for 2 min in prewarmed calcium-free saline and then incubated in pre-warmed 60 mM K ϩ -supplemented saline at 30°C for 10 min. shibire ts1 -depleted preparations were fixed and double-labeled with anti-csp antibodies and either anti-Dap160 or anti-dynamin primary antibodies. Synaptic vesicle proteins are redistributed to the plasma membrane. Dap160 and dynamin immunoreactivity is restricted to hot spots on the plasma membrane, in contrast to the more diffuse staining pattern of csp. The red and green overlay emphasizes this difference. Scale bar, 2 m. utive SH3 domains. To investigate the binding capabilities of the predicted SH3 domains of Dap160, each individual SH3 domain, as well as contiguous subsets of the four, were fused to GST (Fig. 8A). Overlays on both rat brain cytosol and Drosophila head extracts showed that each individual SH3 domain possessed unique specificities (Fig. 8B). The binding capabilities of the individual SH3 domains were additive when combined together; there was no evidence that additional sequence ever resulted in an inhibition of the binding capacity of an individual SH3 domain. By co-migration, the SH3 constructs detected three rat brain proteins that had been previously described, namely synaptojanin, dynamin, and synapsin I (data not shown), and one additional protein at about 65 kDa. Drosophila brain extracts contained what are likely to be Drosophila equivalents of the same three, proline-rich domain-containing proteins, and an additional band at about 180 kDa that co-migrated with clathrin (data not shown). Because of these findings, it was possible to define the minimum domains that bind strongly to each of the four identified proteins (Table I). From this summary, we can see that rat brain dynamin could be bound to three domains, whereas synapsin strongly bound only the B domain and the combinations that contained it. The binding of Dap160 to Drosophila extracts was different. The SH3A and SH3B domains were the only ones that bound dynamin. Domain C did not bind to any proteins in rat brain cytosol but bound to a protein that co-migrated with clathrin in Drosophila head extracts. Binding to both synaptojanin and synapsin I required three contiguous domains, SH3-BCD, presumably as a result of an avidity increase. The overall conclu-sion, therefore, is that the different SH3 domains do not have identical binding specificities.

DISCUSSION
Characterization of Dap160 -By using Ddyn(PRD) as a probe of Drosophila head extracts a new multimodular protein, Dap160, has been identified that can associate with membranes while complexed to dynamin. The properties of the protein suggest that it may be a scaffold that anchors some of the machinery required for endocytosis at the characteristic endocytosis zones or hot spots previously described at Drosophila neuromuscular junctions (1,2).
Dap160 is a peripheral membrane protein that is complexed with dynamin. At physiological ionic strengths, all of the Dap160 is recovered in a 500-kDa complex that can be immunoprecipitated with antibodies to dynamin. Other components of the complex and the stoichiometry of dynamin to Dap160 are not known. During subcellular fractionation at reduced ionic strengths, both dynamin and Dap160 co-distribute together in the membrane fractions. Association of dynamin with Drosophila membranes has already been reported (36). Since both proteins can be eluted by high salt or by detergent, both proteins behave like peripheral membrane proteins. They also are associated with the plasma membrane by immunofluorescence. It thus is possible that Dap160 plays a role in anchoring dynamin to the plasma membrane via its PRD domain.
Dap160 is found to be a multimodular protein with four predicted SH3 domains and one EH domain. The SH3 domains were examined individually for their ability to bind PRD-containing proteins. Individually or in combination (Table I) they bound the three major rat brain PRD proteins, synaptojanin, dynamin, and synapsin Ia already identified using SH3 domain fusion proteins (20,22). In addition to binding Drosophila dynamin, Dap160 can bind two other proteins that, by their electrophoretic mobility, are likely to be Drosophila synaptojanin and synapsin. We conclude that the PRD-containing proteins in rat brains and Drosophila heads are highly conserved during evolution and are recognized by the SH3 domains in Dap160. The SH3C domain, however, has an unexpected specificity in that it binds a protein in Drosophila heads that co-migrates as clathrin. Dap160 also has an EH domain. Eps15, an EH-containing protein in mammalian brain, binds AP2 and accumulates around the neck of coated pits (12). When EH domains were used to screen a random phage displayed peptide library, they were found to bind peptides containing the NPF sequence (13). The stoned B protein in Drosophila has four N-terminal repeats of the NPF motif (31) and so is a potential target. Although no mutations are yet known in stoned B, it contains sequence homology to the mu (AP50) chain of AP2. AP50 is the subunit of AP2 believed to be in-  Several puzzling observations were made by the overlay assay that remain to be fully explained. For example, although Ddyn(PRD) bound to several SH3 domain-containing proteins in rat brain, including amphiphysin I, it did not bind detectably to a rat brain homologue of Dap160 nor did it detect Drosophila homologues of amphiphysin I, II, or endophilin. Such negative results are hard to interpret with confidence since a failure to see binding could reflect difficulties in extraction or in refolding after SDS-PAGE as well as protein abundance. A second puzzling feature was the lack of complementarity. Thus, although Ddyn(PRD) bound rat amphiphysin I, rat amphiphysin I SH3 did not bind Drosophila dynamin (data not shown). One possible explanation is that the amino acid context in which an SH3 domain is placed can affect its affinity. Such an hypothesis could also explain why the combination of domains Dap160-(SH3C) and Dap160(SH3D) acquired binding capacity whereas neither alone demonstrated it in rat brain. These findings suggest caution when interpreting binding data using SH3 domains taken out of their usual protein context.
Dap160 Is a Scaffolding Protein-Modular proteins such as Dap160 with more than one SH3 group have been identified as scaffolding proteins (38). Scaffolding proteins are commonly found in signaling pathways where they bind simultaneously several active enzymes and allow sequential, orderly action of the enzymes. Anchoring proteins, in contrast, keep enzymes in the vicinity of their site of action by attaching them to membrane or cytoskeletal elements. An attractive conjecture is that Dap160 acts as a scaffold, an anchor, or both.
A characteristic of a scaffolding protein is that it can bind an array of enzymes or proteins. Dap160 can bind dynamin, a GTPase (39), synaptojanin, an inositol-5-phosphatase (40), and synapsin, an ATP-binding protein (41). Although it is possible that Dap160 binds three or more dynamins simultaneously, the different binding specificities of the three active sites suggest that each might have different preferences in vivo (Table I). In addition Dap160 might well have the capacity to bind clathrin. It thus has the characteristic of a protein scaffold that might localize endocytotic machinery to zones of active endocytosis.
Dap160 may be a member of a growing family of scaffolding proteins. The first to be identified were those such as grb2 and IRS-1 that have two SH3 domains and which were shown to function as scaffolds for signaling machinery (38). The partial DNA sequences of human SH3p17 and SH3p18 (27) show remarkable homology at the C terminus even to the extent of the spacing between the contiguous SH3 domains (Fig. 3D). The sequence for a protein named intersectin, available in the data base, has five SH3 domains and an EH domain at its N terminus. The family of multimodular, SH3 domain-containing proteins might prove to be quite extensive. Currently, based on the grb2 and IRS-2 results, the most likely function of the family is to act as scaffolding proteins.
The hot spots for endocytosis appear to be present in resting nerve terminals, where they co-localize with the active zones (1,2). Since Dap160 is a component of those resting hot spots, it could also be acting as an anchor protein, concentrating the elements of the coating machinery in an inactive form at the sites where they will be needed. To test the validity of scaffolding and anchoring models it is essential to identify the other components of the endocytosis hot spots, both at rest and in recycling nerve terminals.
The Endocytotic Hot Spots-The synapse at the Drosophila neuromuscular junction can be surprisingly large, as much as 10 m in diameter, making it a favorable preparation for immunofluorescence studies. By using antibodies to synaptic ves-icles, clusters of synaptic vesicles can be seen within the synaptic boutons, surrounded by a continuous plasma membrane, detected with antibodies to horseradish peroxidase. Recycling of synaptic vesicle membranes in active nerve terminals can be readily monitored by the uptake of the fluorescent dye, FM1-43, into the synaptic vesicle clusters (35). When shibire ts1 flies are held at nonpermissive temperatures, the staining of the synaptic vesicle clusters disappears, and the nerve terminal swells due to the addition of the synaptic vesicle membranes (1). The synaptic vesicle proteins are not constrained to stay at the sites of exocytosis in shibire ts1 flies stimulated to exocytose at nonpermissive temperatures. The synaptic vesicle proteins can even diffuse into the plasma membrane regions between the boutons (1,35).
Dynamin was found to localize to hot spots that resided within the synaptic vesicle clusters leading to the notion that dynamin was anchored to some site in the vicinity of the active zone, where it could be available when endocytosis was required. Subsequently, AP2 (2) and now Dap160 appear to be similarly anchored, perhaps in large multiprotein complexes. After stimulating shibire ts1 nerve terminals at nonpermissive conditions, it becomes easier to resolve the zones for endocytosis because of the swelling of the nerve terminals. The number of such zones was about 8 per bouton, close to the average number of active zones (42). It appears as if the three components remain together in the hot spots. A simple explanation, but not the only one, is that the endocytotic machinery remains at what was the active zone at rest where it facilitates the regeneration of synaptic vesicles at the place they will be needed. Note that the vesicle proteins are unlikely to define the active site, for they diffuse in the plane of the plasma membrane (1) while the endocytosis zone remains intact.
At present, we cannot determine if the endocytosis zone is at the active site or nearby. Electron micrographs from Koenig and Ikeda (43) describe two pathways of endocytosis in nerve terminals, one near the active zone and another at some distance away. Uptake of FM1-43 into frog neuromuscular junctions indicates that there are bands of endocytosis that coincide with active zones determined by acetylcholine receptor clustering (44). It may now be possible to accurately localize the endocytosis zone and determine what controls that localization using the Drosophila nerve terminal and appropriate developmental mutants.
Clues to what the endocytosis zone does will be obtained by identifying more protein constituents. The advantage of using overlay techniques to identify the molecular interactions that take place between proteins at the active zone is that the peptide domains identify strong interactions and can subsequently be used as dominant negative inhibitors of the interaction. They can thus be used to dissect out the steps involved either using a transgenic fly, by microinjection, or by reconstituting endocytosis in vitro. Success has already been reported using the SH3 domain of amphiphysin to block synaptic vesicle recycling in lamprey axons (9). In addition, the formation of synaptic-like microvesicles from the plasma membranes of pheochromocytoma PC12 cells is inhibited by Ddyn(PRD). 2 It thus should be possible to link protein interactions to synaptic functions both by reverse genetics and in vitro reconstitution.