Differential expression of endophilin 1 and 2 dimers at central nervous system synapses.

Endophilin 1 is proposed to participate in synaptic vesicle biogenesis through SH3 domain-mediated interactions with the polyphosphoinositide phosphatase synaptojanin and the GTPase dynamin. Endophilin family members have also been identified as binding partners for a number of diverse cellular proteins. We define here the endophilin 1-binding site within synaptojanin 1 and show that this sequence independently and selectively purifies from brain extracts endophilin 1 and a closely related protein, endophilin 2. Endophilin 2, like endophilin 1, is highly expressed in brain, concentrated in nerve terminals, and found in complexes with synaptojanin and dynamin. Although a fraction of endophilins 1 and 2 coexist in the same complex, the distribution of these endophilin isoforms among central synapses only partially overlaps. Endophilins 1 and 2 are found predominantly as stable dimers through a predicted coiled-coil domain in their conserved NH2-terminal moiety. Dimerization may allow endophilins to link a number of different cellular targets to the endocytic machinery.

Neurotransmitter release requires the fusion of small neurotransmitter-containing vesicles with the presynaptic plasma membrane. Nerve terminals, which are spatially segregated from the biosynthetic machinery in the neuronal cell body, must locally recycle synaptic vesicle membrane components to maintain a functional pool of secretory vesicles (1,2). The physiological importance of this recycling pathway has been demonstrated in a variety of systems in which inhibition of synaptic vesicle recycling results in synaptic failure (3,4).
One mechanism for the retrieval of synaptic vesicle membranes following exocytosis is clathrin-mediated endocytosis. Many protein factors that participate in clathrin-mediated synaptic vesicle recycling have been identified as direct or indirect interactors of either components of the clathrin coat or the GTPase dynamin, a protein required for the maturation and scission of nascent coated pits (reviewed in Ref. 5). The polyphoinositide phosphatase synaptojanin 1 is one such accessory factor. Synaptojanin 1 contains two separable enzymatic activities that dephosphorylate phosphoinositides. A central inositol 5Ј-phosphatase domain mediates cleavage of the 5Ј phosphate and an NH 2 -terminal domain homologous to the yeast protein Sac1p that hydrolyzes phosphates at the 3Ј, 4Ј, or 5Ј positions from a number of phosphoinositide species (6,7). As a phosphatidylinositol (4,5)-bisphosphate catabolizing enzyme, synaptojanin has been proposed to regulate interactions between membranes and the clathrin coat or membranes and the actin cytoskeleton (8 -10).
A major binding partner for synaptojanin 1 is the SH3 1 domain-containing protein endophilin 1 (11,12). Endophilin 1 is a 40-kDa protein containing a COOH-terminal SH3 domain and an NH 2 -terminal lipid-binding domain (13). It has also been reported to be a lipid-modifying enzyme (14). Endophilin 1 and two closely related gene products (endophilins 2 and 3) have been identified as ligands for diverse cellular targets (15)(16)(17)(18) suggesting that endophilin family members participate in many cellular pathways, perhaps coupling them to the endocytic machinery.
We now show that the two major endophilin isoforms expressed in brain exist as dimers and can therefore act as bivalent SH3 domain-containing adaptors between different binding partners. Endophilins 1 and 2 share many properties, including highly specific binding to a proline-rich peptide motif PPXRP previously reported in a preliminary form (19) and synaptic localization. Endophilin 2, however, is expressed in a restricted set of synapses.

EXPERIMENTAL PROCEDURES
Antibodies-Affinity purified antibodies against endophilin 1, endophilin 2, and anti-synaptojanin 1 monoclonal antibodies have been described previously (19 -21). Anti-dynamin 1 monoclonal antibodies were purchased from Upstate Biochemicals. Anti-synaptophysin monoclonal antibodies were a kind gift of Dr. R. Jahn, Gottingen, Germany. Polyclonal anti-HA antibodies were purchased from Santa Cruz Biotechnology and monoclonal anti-FLAG antibodies were purchased from Sigma.
Immunoprecipitations-Triton X-100 extracts of total brain homogenate or transfected cells were prepared from frozen rat brains (Pel Freeze) as previously described (12). 5 g of affinity purified antibodies were incubated for 1 h at 4°C in 500 g of either total brain cytosol (see below), Triton X-100-extracted total brain homogenate, or cell lysate. Immune complexes were precipitated for 1 h at 4°C with 50 l of a 50% slurry of protein G-Sepharose (Amersham Pharmacia Biotech) equilibrated in 20 mM HEPES-KOH, pH 7.4, 100 mM KCl (buffer A) or buffer A plus 1% Triton X-100. Pellets were washed 3 times in buffer A or buffer A plus 1% Triton X-100 and eluted into SDS-PAGE sample buffer.
Affinity Purification from Brain Extracts-Fragments of the prolinerich tail of synaptojanin 1 were generated by polymerase chain reaction, subcloned into the pGEX-4T1 vector for expression as GST fusion pro-teins and sequenced (see Fig. 1). Fusion proteins were expressed in the bacterial strain DH5␣ and purified according to the manufacturer's instructions. 50 g of fusion protein immobilized on glutathione-Sepharose (Amersham Pharmacia Biotech) was added to 5 mg of Triton X-100 extracted total brain homogenate and incubated at 4°C for 3 h. The affinity matrix was collected by gentle centrifugation and washed 3 times with buffer A plus 1% Triton X-100. Fusion proteins and bound proteins were eluted with SDS-PAGE sample buffer.
Cross-linking of Detergent Extracts of Brain Homogenate and Recombinant Endophilins-Full-length endophilin 1 and endophilin 2 cDNAs or various deletion mutants were subcloned into the pGex6p1 expression vector (Amersham Pharmacia Biotech), expressed in the bacterial strain BL21 and purified on glutathione-Sepharose according to the manufacturer's instructions. Purified proteins on the beads were incubated overnight with Precision protease in 100 mM KCl, 20 mM HEPES-KOH, pH 7.4, 1 mM dithiothreitol, and 0.1% Triton X-100 to remove the GST tag. 3 g of each purified protein was incubated in 20 l of final volume of buffer A plus varying amounts of the heterobifunctional cross-linker 1-ethyl-3-[dimethylaminopropryl]carbodiimide⅐HCl (EDC) (Pierce) for 20 min at 30°C. The cross-linking reaction was stopped by the addition of 5 ϫ SDS-PAGE sample buffer to a final concentration of 1 and the resulting products were separated on a 10, 12, or 15% polyacrylamide SDS reducing gel. After electrophoresis, the gel was stained with Coomassie Brilliant Blue. For cross-linking of brain extracts, 50 g of Triton X-100 extracted total brain homogenate was incubated in the presence of increasing concentrations of EDC for 30 min at room temperature. As above, the reaction was stopped by the addition of 5 ϫ SDS-PAGE sample buffer and the reaction products were separated on a 5-12% polyacrylamide reducing gel and processed for Western blotting.
Miscellaneous-The affinities of proline-rich peptides for the recombinant SH3 domain of endophilin 1 (12) were measured using changes in intrinsic tryptophan fluorescence according to Ref. 22 in a Hitachi F3010 fluorimeter. GST was cleaved from the fusion protein with biotinylated thrombin (Novagen). Gel filtration chromatographies of total brain cytosol and recombinant proteins were performed on a Sephadex G100 column and a Superdex 200 column equilibrated with buffer A, respectively. Columns were calibrated using gel filtration molecular weight markers (Amersham Pharmacia Biotech). Double immunofluorescence experiments with rat brain sections and cultured hippocampal neurons were performed as described (23). Cell transfections were performed with LipofectAMINE cationic lipids (Life Technologies, Inc.) according to the manufacturer's instructions.
Monodimensional SDS-PAGE was performed according to standard procedures. Western blotting was performed using horseradish peroxi-

FIG. 1. Identification of endophilin 1-binding sites in the proline-rich tail of rat synaptojanin 1. A panel of deletion mutants (A)
derived from the proline-rich domain of rat synaptojanin 1 (amino acids 985-1307) was expressed as GST fusion proteins and used to affinity purify interacting proteins from detergent extracts of rat brain homogenates. Dark boxes indicate PxxP motifs implicated in endophilin binding and white boxes denote sites implicated in amphiphysin binding. B, bound proteins were eluted with SDS-PAGE sample buffer and processed for Western blot analysis using anti-endophilin 1 or antiamphiphysin 1 antibodies.

FIG. 2. Endophilin 1-binding peptides also bind a second endophilin isoform of 45 kDa, endophilin 2.
A, peptides corresponding to endophilin 1-binding sequences in the proline-rich tail of rat synaptojanin 1 (PP19), to an amino acid sequence in the proline-rich tail of rat dynamin 1 which contain the PPXRP motif (DPP19), or to a region of the synaptojanin 1 proline-rich domain upstream of the identified endophilin 1-binding sites (control peptide PP10) were synthesized and coupled to Sepharose beads. B, immobilized peptides were used to affinity purify interacting proteins from detergent extracts of rat brain homogenate. Bound proteins were eluted with an excess of free peptide and silver stained to analyze total protein composition or processed for Western blot analysis using both anti-endophilin 1 and anti-endophilin 2 antibodies. Asterisks denote protein species identified as endophilin isoforms by Western blot. C, the affinity of the endophilin 1 SH3 domain for various peptides was determined by measuring changes in intrinsic fluorescence observed upon ligand binding by the endophilin 1 SH3 domain in solution. Binding of the PP10 peptide (closed diamonds) was undetectable in this assay while the PP19 (closed circles) and DPP19 (open squares) peptides bound the endophilin 1 SH3 domain with affinity constants of 14 and 79 M, respectively. D, immobilized endophilinbinding peptide PP19 was used to affinity purify endophilin 2 from detergent extracts of BHK21 cells as described in B. The asterisk denotes a species specifically bound by the PP19 matrix identified as endophilin 2 by Western blot.

RESULTS
Identification of an Endophilin 1-binding Sequence-It has been shown by us and others that interactions between endophilins and synaptojanin 1 are mediated by the binding of the SH3 domain of endophilins to the COOH-terminal prolinerich domain of rat synaptojanin (11,12). To identify specific endophilin-binding elements in the proline-rich domain of synaptojanin we generated a panel of GST fusion proteins derived from the COOH terminus of synaptojanin (Fig. 1A). These fusion proteins were immobilized on glutathione-Sepharose beads and used to affinity purify interactors from extracts of total rat brain homogenate. Purified material was analyzed by Western blot for the presence of endophilin 1 and, as a control, for the presence of amphiphysin 1, another SH3 domain-containing interactor of synaptojanin (6) (Fig. 1B). Initially, deletional analysis indicated that amino acids 1112-1130 were required for endophilin 1 binding. This region contains two PPXRP motifs, where x ϭ A or Q, which conform to the core PXXP SH3 binding consensus (24). Deletion of these motifs in either direction revealed that either one of them was sufficient to mediate endophilin binding. We note that synaptojanin orthologs from diverse taxa have at least one PPXRP motif and that the two PPXRP motifs of rat synaptojanin 1 are conserved between mammals and a distant vertebrate, lamprey (9). Furthermore, one such motif is also present in a short fragment of the dynamin 1 proline-rich domain previously shown to bind endophilin 1 (25). In contrast, amphiphysin 1 binding to the proline-rich domain of synaptojanin required either amino acids 1054 -1069 which conform to a previously identified amphiphysin binding consensus (PXRPXR) (26) or the very COOH terminus of the proline-rich domain, demonstrating the existence of two distinct binding sites for this protein.
To test whether PPXRP motifs are sufficient to bind endophilin 1, and to assess the specificity of this interaction, peptides containing either the two conserved PPXRP motifs of synaptojanin (PP19) or the single PPXRP motif present in dynamin (DPP19) were synthesized, immobilized on a solid support, and used to affinity purify interactors from extracts of total brain homogenate. As a negative control, another proline-rich peptide (PP10) corresponding to sequences NH 2 -terminal to the first PPXRP motif in synaptojanin and predicted not to bind endophilin by our deletional analysis was used ( Fig. 2A). Bound material was eluted with an excess of free peptide and analyzed either by silver stain or Western blot using anti-endophilin antibodies. Both PPXRP-containing peptides specifically bound a major protein species of 40 kDa that was identified as endophilin 1 by Western blot (Fig. 2B).
We also measured the affinity of the interaction between PPXRP motifs and the SH3 domain of endophilin 1. The SH3 domain of endophilin 1 contains a single tryptophan residue whose intrinsic fluorescence changes specifically in the presence of PPXRP containing peptides. Peptide-induced changes in the intrinsic fluorescence of the endophilin 1 SH3 domain were dose-dependent and saturable and reflected an affinity constant of ϳ14 M for the interaction between endophilin 1 and the synaptojanin peptide and ϳ79 M for the interaction between endophilin 1 and the dynamin peptide.
Endophilin 2 Is Also Bound by PPXRP-containing Sequences and Is Abundantly Expressed in Brain-In addition to endophilin 1, a 45-kDa protein species was found in eluates from both PPXRP columns (Fig. 2B). We postulated that this band could be the endophilin family member of 45 kDa, endophilin 2. Endophilin 2 was previously identified as an endophilin isoform whose message was present in all tissues analyzed (12). Using antibodies specific to endophilin 2 we detected endophilin 2 protein expression predominantly in brain and testis although low levels were detected in many tissues (Fig. 3). The same antibodies confirmed that the 45-kDa protein retained on our peptide columns was endophilin 2 (Fig. 2B). To show that the binding of endophilin 2 to PPXRP sequences is independent of endophilin 1, we repeated the affinity chromatography experiments using detergent extracts of the cell line BHK21 which only expresses endophilin 2 (20) (Fig. 2D). In these experiments, we detected only a single protein of 45 kDa specifically bound to the PPXRP column and Western blot analysis identified this protein as endophilin 2. Since endophilin 2 has not been characterized, we compared its properties to those of endophilin 1.
We first tested whether endophilin 2 is found in protein complexes with dynamin and synaptojanin 1 by immunoisolating endophilin complexes from detergent extracts of brain homogenate using isoform-specific antibodies. Anti-endophilin 1 immunoprecipitates contained three major protein species (asterisks in Fig. 4A) whose identities were assigned by Western blot (Fig. 4B): endophilin 1 (40 kDa), dynamin 1 (100 kDa), and synaptojanin 1 (145 kDa). In anti-endophilin 2 immunoprecipiates only endophilin 2 (45 kDa) could be detected by protein staining. Synaptojanin 1 and dynamin were detected in the isolated material by Western blot, however (Fig. 4B). The lower amounts of co-precipitated dynamin and synaptojanin with endophilin 2 are consistent with the lower abundance (ϳ8 fold) of endophilin 2 in the immunoprecipitates. Since these experiments we performed under antigen-depleting conditions, this difference in the amounts of the two endophilin isoforms reflects their relative abundance in brain. Note that anti-endophilin immunoprecipitates primarily contain the cognate endophilin isoform recognized by the antibody (Fig. 4A) although a minor cross-precipitating pool of endophilins does exist as revealed by Western blotting (Fig. 4C), possibly reflecting the existence of a small pool of heterodimers (see below).
Endophilins 1 and 2 Are Dimers-Within the NH 2 terminus of the endophilins are sequences predicted to adopt a coiled-coil conformation with the potential to mediate protein-protein interactions or dimerization. We therefore compared the oli- 5. Endophilin 1 and endophilin 2 form 80 -100-kDa multimers. A, rat brain cytosol was fractionated on a Sephadex S100 gel filtration column and individual fractions were assayed by immunoblot for the presence of endophilins 1 and 2. Arrowheads indicate the elution peaks of molecular weight standards. B, detergent extracts of rat brain homogenate were incubated at 30°C for 20 min in the presence of 0, 0.5, 1, 5, and 10 mM EDC, quenched with SDS-PAGE sample buffer, separated by SDS-PAGE, and blotted using anti-endophilin 1 and antiendophilin 2 antibodies. C, anti-HA immunoprecipitates from detergent extracts of Chinese hamster ovary cells co-transfected with FLAG and HA epitope-tagged endophilin 1 cDNAs were analyzed for the presence of FLAG-tagged endophilin 1 by Western blot. Anti-Myc immunoprecipitates were used as a negative control. 10% of starting material (SM) and supernatant (VOID) and 50% of bound material was loaded. gomerization state of endophilin 1 and 2. Both proteins were found to elute as 80 -90-kDa complexes from a number of gel filtration columns and when fractionated on a Sephadex G-100 column they eluted in overlapping fractions with distinct peaks (Fig. 5A). Chemical cross-linking of detergent extracts of rat brain homogenate with the heterobifunctional cross-linker EDC followed by Western blot using anti-endophilin antibodies similarly showed that both endophilins could be cross-linked into 80 -100-kDa complexes (Fig. 5B). Furthermore, when FLAG-tagged endophilin 1 and HA-tagged endophilin 1 were co-transfected into Chinese hamster ovary cells, FLAG-tagged endophilin 1 could be co-immunoprecipitated with HA-tagged endophilin 1 from detergent extracts of these cells using an anti-HA antibody (Fig. 5C). The self-association of endophilins, together with the apparent size of endophilin complexes observed in brain extracts, suggests that endophilins form homodimers. Accordingly, both purified recombinant endophilin 1 (Fig. 6A) and endophilin 2 (data not shown) could be crosslinked into several closely spaced bands in the same size range. The heterogeneous electrophoretic motility of these bands may be due to multiple conformers of the dimers.
To confirm the role of the predicted coiled-coil in endophilin dimerization, we generated a number of endophilin deletion mutants which were expressed, purified, and subjected to chemical cross-linking followed by SDS-PAGE (Fig. 6). Recombinant proteins containing amino acids 125-290 could be crosslinked into higher molecular weight complexes while deletion of this sequence resulted in a protein that remained monomeric after treatment with EDC.
Endophilin 2 Is Expressed in a Subset of Synapses-Our characterization of endophilin 1 and endophilin 2 complexes showed that endophilin isoforms primarily exist in separate complexes while interacting with a common set of synaptic proteins. It was therefore of interest to determine whether endophilin isoforms are expressed in distinct populations of neurons. Cultures of hippocampal neurons were fixed, immunostained with antibodies that recognize selectively either endophilin 1 or endophilin 2, and counterstained for the presynaptic marker synaptophysin (Fig. 7). Both endophilin 1 (Fig. 7, A and C) and endophilin 2 (Fig. 7, E and G) immunoreactivities were concentrated in puncta which colocalized with synaptophysin ( Fig. 7, B, D, F, and H). While endophilin 1 staining was detected in the majority of synaptophysin-positive puncta, only a restricted population of nerve terminals was stained with anti-endophilin 2 antibodies (arrowheads in Fig. 7, G and H).
A co-localization of endophilin 1 and 2 with synaptophysin was also visible in a frozen section of rat brain (Fig. 8) where the differential expression of endophilins 1 and 2, and the more widespread expression of endophilin 1, was even more apparent (see, for example, the more restricted distribution of endophilin 2 in the CA1 region of the hippocampus). Some synapses, however, such as the mossy fiber synapses in the hippocampal area CA3, were predominantly endophilin 2 pos- itive (Fig. 8, A and C). In other regions of the brain, endophilin 1 and 2 had similar distributions such as in deep cerebellar nuclei where both proteins were enriched in axosomatic synapses (Fig. 8, G and J). DISCUSSION The endophilins have been proposed to play roles in membrane trafficking reactions through interactions with endocytic proteins and via physical or chemical modifications of membrane bilayers (11,12,14,19). Recently, new ligands for the endophilins have been identified that participate in a variety of cellular processes. These ligands include huntingtin (15), transmembrane proteins such as a disintegrin (17) and a Gprotein-coupled receptor (16), and a cytosolic serine/threonine kinase involved in JNK signaling (18). Our study demonstrates that endophilin isoforms exist as stable dimers that may bridge components of the endocytic machinery, cytoplasmic domains  of integral membrane proteins, and cytosolic signaling proteins. Sequences matching the endophilin binding motif described in this study are present in a number of signaling, trafficking, and integral membrane proteins not yet shown to bind endophilin family members (Table I) raising the possibility that endophilins participate as adaptors in a wide array of biological reactions. It is interesting to note that among SH3 domains, the SH3 domains of the endophilins are most highly related to the COOH-and NH 2 -terminal SH3 domains of the adaptor protein Grb2 which is known to couple signaling molecules to the endocytic machinery (27). Furthermore, other SH3 domain-containing proteins initially identified as components of the endocytic machinery are also multivalent (either through dimerization or the presence of multiple SH3 domains) and have been shown to function in signaling networks, most likely as adaptors (28 -31).
The endophilin SH3 domain-binding sequences identified in this study differ somewhat from the consensus PϩRPP (where ϩ ϭ K or R) previously identified by a phage display screen for endophilin-binding peptides (32). We note, however, that when inverted, the PPXRP motif identified in this study fits the PϩRPP consensus. SH3 domain ligands can, in fact, bind SH3 domains in opposing orientations (33) and it is possible that endophilin ligands exist in two classes as mirror images of one another. Some of the known interactors for endophilins contain sequences that only partially conform to these motifs suggesting that the spectrum endophilin interactors might be very broad. Our data indicate that tandem sequences of the form PPXRP between amino acids 1112 and 1125 are the major endophilin-binding sites in synaptojanin 1. Interestingly, a distinct mammalian synaptojanin protein, synaptojanin 2B, also contains a PPXRP sequence within an alternatively spliced exon in its COOH terminus and its presence or absence correlates with endophilin binding. 2 Although previous studies have shown that endophilin 2 mRNA is present in many tissues (12,34), we have now found that this protein is abundantly expressed in brain where, like endophilin 1, it is primarily concentrated in nerve terminals. The localization of endophilin 2 in nerve terminals and in complexes with synaptojanin and dynamin suggests that endophilin 2 participates in clathrin-mediated synaptic vesicle recycling. In non-neuronal cells, endophilin isoforms may also participate in clathrin-mediated endocytosis. In addition, endophilin 2 has been localized to specialized actin scaffolds along with a dynamin isoform (20). This localization may reflect a role for endophilins in actin function as well, possibly as components in a signaling pathway regulating the actin cytoskeleton. A dual role for endophilins in endocytosis and actin function is consistent with growing evidence that actin participates in many types of endocytosis including synaptic vesicle recycling (5,(35)(36)(37). The endophilins may therefore represent a set of adaptors that coordinate endocytosis, actin function, and signaling cascades at the synapse and in non-neuronal cells.