The Interaction of Epsin and Eps15 with the Clathrin Adaptor AP-2 Is Inhibited by Mitotic Phosphorylation and Enhanced by Stimulation-dependent Dephosphorylation in Nerve Terminals*

Clathrin-mediated endocytosis was shown to be arrested in mitosis due to a block in the invagination of clathrin-coated pits. A Xenopus mitotic phosphoprotein, MP90, is very similar to an abundant mammalian nerve terminal protein, epsin, which binds the Eps15 homology (EH) domain of Eps15 and the α-adaptin subunit of the clathrin adaptor AP-2. We show here that both rat epsin and Eps15 are mitotic phosphoproteins and that their mitotic phosphorylation inhibits binding to the appendage domain of α-adaptin. Both epsin and Eps15, like other cytosolic components of the synaptic vesicle endocytic machinery, undergo constitutive phosphorylation and depolarization-dependent dephosphorylation in nerve terminals. Furthermore, their binding to AP-2 in brain extracts is enhanced by dephosphorylation. Epsin together with Eps15 was proposed to assist the clathrin coat in its dynamic rearrangements during the invagination/fission reactions. Their mitotic phosphorylation may be one of the mechanisms by which the invagination of clathrin-coated pits is blocked in mitosis and their stimulation-dependent dephosphorylation at synapses may contribute to the compensatory burst of endocytosis after a secretory stimulus.

mediated endocytosis, including the endocytosis of synaptic vesicles in nerve terminals (1)(2)(3)(4)(5). Two such proteins are Eps15 and epsin (6,7). Eps15 was first identified as an endogenous substrate for EGF 1 receptor kinase (8) and was subsequently found to be an interacting partner for the "appendage domain" of the AP-2 subunit ␣-adaptin. Binding of Eps15 to AP-2 is mediated by its COOH-terminal region (9,10), whereas the NH 2 -terminal region of Eps15 includes three Eps15 homology (EH) domains (11). Via these modules, Eps15 binds proteins with the consensus amino sequence NPF (12). Epsin, which contains three NPF motifs in its COOH-terminal region (NPF domain), is a major binding partner for Eps15 (7). Its NH 2terminal portion comprises an evolutionary conserved domain of unknown function, the ENTH domain (epsin NH 2 terminal homology domain), whereas its central region, which contains eight DPW repeats (DPW domain), binds the appendage domain of ␣-adaptin at a site that overlaps with the Eps15binding site (7). Perturbation of the interactions of both Eps15 and epsin with AP-2, as well as disruption of the function of both proteins by antibody injection, block clathrin-mediated endocytosis (7,(13)(14)(15)(16). It was proposed that Eps15 and epsin play an important role in clathrin-mediated endocytosis, possibly by participating in dynamic rearrangements of the clathrin coat during bud invagination and fission (7,17,18).
Clathrin-mediated endocytosis is blocked during mitosis. In mitotic cells clathrin coats assemble, but their invagination is impaired (19,20). The identification of substrates of mitotic kinases responsible for this effect may therefore shed new light on the still elusive mechanisms underlying the invagination reaction. Epsin is highly homologous to the Xenopus mitotic phosphoprotein MP90, which was identified in a screen for substrates of mitotic kinases (21), and contains a single putative consensus site for Cdc2 kinase, which is conserved in mammalian epsin. These considerations prompted us to investigate whether epsin undergoes mitotic phosphorylation. We report here that both epsin and Eps15 are phosphorylated in mitosis and that their phosphorylation inhibits binding to the clathrin adaptor AP-2. We also report that both epsin and Eps15, like other accessory proteins of clathrin-mediated endocytosis, undergo stimulation-dependent dephosphorylation in nerve terminals (22)(23)(24)(25), with a resulting increase in their binding to each other and to AP-2. Their dephosphorylation may facilitate endocytosis of synaptic vesicle membranes following an exocytotic burst.

MATERIALS AND METHODS
Cells and Reagents-B82 mouse fibroblasts were cultured in Dulbecco's modified Eagle's medium (Life Technologies, Inc.) containing 20% calf serum and 5 mM methotrexate. Antibodies directed against epsin, Eps15, amphiphysin 1, and GST fusion proteins of ␣-adaptin and the DPW domain of epsin were previously described (7). Antibody against ␣-adaptin was purchased from Sigma.
Analysis of Interphase and Mitotic Cells-B82 mouse fibroblasts were grown to 60 -80% confluency and synchronized with 50 ng/ml nocodazole for 4 h. Mitotic cells were collected by detaching rounded up cells. Interphase (G 1 phase) cells were obtained by washing away the nocodazole from the mitotic cells, replating cells in fresh medium, and allowing cells to grow for 4 h before harvesting. Particulate and soluble fractions from these cells were obtained as described (7). Cells extracts * This study was supported in part by Grants CA46128 and NS36251 from the National Institutes of Health, by the Human Frontier Science Program and the U. S. Army Medical Research and Development Command (to P. D. C.), and by grants from the Associazione Italiana per la Ricerca sul Cancro, European Community (Biomed-2), the Armenise-Horrad Foundation, and the Ferrero Foundation (to P. P. D. F.). 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.
Generation of a Mutant DPW Domain-The mutant DPW domain was obtained by polymerase chain reaction-based site-directed mutagenesis. A pair of primers harboring the serine 328 to aspartate mutation were generated: 5Ј-GACCCTTGGGGAGGTGATCCT-3Ј and 5Ј-AGGATCACCTCCCCAAGGGTC-3Ј. Using 5Ј-GACCCTTGGGGAG-GTGATCCT-3Ј together with 5Ј-AAACGCGTCGACGTCGAAGTCT-GAGAACTCATC-3Ј and 5Ј-AAACCGGAATTCCGGATCCGTCGTGGG-GAT-3Ј together with 5Ј-AGGATCACCTCCCCAAGGGTC-3Ј, two cDNA fragments were obtained and purified by a QIAGEN kit (QIA-GEN). These two DNA fragments were used as both primers and templates in a standard polymerase chain reaction to generate a fulllength mutant DPW domain, which was subsequently cloned into the PGEX6 -1 vector (Amersham Pharmacia Biotech). The sequence of the mutant DPW domain was confirmed by standard double-strand sequencing. Phosphorylation of Brain Cytosol-Brain cytosol was prepared by homogenizing rat brains in 2 volumes of 10 mM Hepes (pH 7.4), 1 mM EDTA, and a protease inhibitor mixture (3 g/ml each of aprotinin, antipain, leupeptin, and pepstatin). The lysate was centrifuged at 100,000 ϫ g for 2 h, and the resulting supernatant (cytosol) was desalted on Sephadex G-25 (Amersham Pharmacia Biotech) into 20 mM Hepes (pH 7.4), 120 mM KCl, and 1 mM MgCl 2 at room temperature. The desalted cytosol was incubated at 30°C for 30 min either with a general protein kinase inhibitor (K252a) (dephospho-cytosol) or with 5 mM ATP, 5 mM MgCl 2, 1 mM CaCl 2, 1 mM sodium orthovanadate (Sigma), 2 M cyclosporin (Calbiochem), and 100 nM okadaic acid (Calbiochem) (phospho-cytosol). The incubation was terminated by the addition of 1% SDS, and SDS was then "neutralized" by the addition of 2% Triton X-100 in 10 mM Hepes (pH 7.4), 150 mM NaCl to achieve a Triton X-100:SDS ratio Ͼ 4 (w/w).
Miscellaneous Procedures-Affinity purification of brain extracts and experiments on intact synaptosomes were performed as described previously (27).

RESULTS
We investigated whether mammalian epsin undergoes phosphorylation in mitotic cells as predicted by the property of Xenopus MP90 to act as a substrate for mitotic kinases (21). The mitotic phosphorylation of MP90 was shown to lower its mobility in SDS-PAGE (21). Cell extracts from mitotic and interphase (G 1 phase) B82 cells (a mouse fibroblastic cell line) were separated by SDS-PAGE, and the mobility of epsin was analyzed by Western blotting. As shown by Fig. 1A, epsin from mitotic extracts had a slower mobility than epsin from interphase cells. A similar shift was observed for Eps15. A putative Cdc2 phosphorylation site (28) is present in the COOH-terminal region of Eps15 (threonine 779 of mouse Eps15 (8)), which is the ␣-adaptin-binding region.
To confirm that the electrophoretic shifts were due to phosphorylation, epsin and Eps15 were immunoprecipitated from interphase and mitotic cell extracts. The immunoprecipitates obtained from mitotic extracts were then incubated with or without alkaline phosphatase or with both alkaline phosphatase and a phosphatase inhibitor mixture. As shown by Fig. 1B, incubation with alkaline phosphatase completely reversed the mobility shift due to mitosis, and this effect was blocked by the presence of phosphatase inhibitors.
Clathrin-coated pits are present in mitotic cells, and the main difference from interphase cells is the predominance of pits with shallow curvature (19,20). A block of the invagination reaction may result from a dissociation from the pits of factors that normally assist the coat during its progressive rearrangement. We therefore compared the partitioning of clathrin, epsin, and Eps15 between the soluble and particulate fraction of interphase (G 1 ) and mitotic (M) cells. As shown by Fig. 2A, the fraction of clathrin recovered in the two fractions was similar in both conditions. In contrast, the mitotic phosphorylation of epsin and Eps15 (revealed by their upper mobility shift), correlated with a drastic decrease of their recovery in the particulate fraction. One of the factors responsible for this effect may be a decreased affinity of mitotic epsin and Eps15 for the clathrin adaptor AP-2. To address this question, Triton X-100 extracts of interphase and mitotic B82 cells were affinity-puri- fied onto a GST fusion protein comprising the appendage domain of ␣-adaptin (Fig. 2B). Analysis of the bound fraction by Western blotting revealed a major decrease in binding of the mitotic forms of both epsin and Eps15 compared with the binding of the corresponding interphase proteins (Fig. 2B).
Epsin and MP90 contain a single putative phosphorylation site for Cdc2 kinase in the DPW domain (serine 328 of rat epsin) (7,21). To determine whether this site acts as a substrate for the Cdc2 kinase and mediates the inhibition of AP-2 binding in mitosis, a mutant DPW domain of rat epsin was generated harboring a S328D mutation. Purified wild type and mutant DPW domains were incubated with [␥-32 P]ATP in the presence of the purified Xenopus p34 cdc2 -cyclin B kinase complex (26). As shown in Fig. 3A, a very strong difference was observed in the 32 P incorporation of the two proteins, indicating that serine 328 is a key substrate site for the kinase in vitro. Furthermore, affinity purification on the immobilized ␣-adaptin appendage domain of the 32 P-labeled wild type DPW domain revealed no binding of the 32 P-labeled protein, proving that its phosphorylation by Cdc2 kinase blocks the interaction (Fig. 3B). Finally, as shown by Fig. 3C, affinity purification of a brain cytosolic extract on wild type and mutant DPW domains demonstrated a significant decrease in the binding of ␣-adaptin to the mutant domain, confirming that the introduction of an acidic charge at position 328 affects the epsin-AP-2 interaction.
Stimulation of neurotransmitter release from nerve terminals correlates with an increased state of phosphorylation of several proteins with a putative role in synaptic vesicles exocytosis (29) and a decreased state of phosphorylation of several proteins with a putative role in their endocytosis including dynamin 1, synaptojanin 1, and amphiphysin (22,23,25,30,31). We investigated whether epsin and Eps15 as well undergo dephosphorylation. Fig. 4A shows immunoblots of rat synaptosomes incubated for 1 min in either control buffer or high K ϩ buffer. A downward shift of epsin in depolarized synaptosomes can be seen. This shift correlates with the concomitant downwards shift of amphiphysin 1 previously shown to reflect its Ca 2ϩ -dependent dephosphorylation (23). In control synaptosomes, Eps15 migrated as a doublet. Depolarization resulted in a decreased immunoreactivity on the upper bands with a corresponding increase of the immunoreactivity in the lower band, suggesting that Eps15 as well undergoes depolarization-dependent dephosphorylation.
The phosphorylation of amphiphysin, dynamin 1, and synaptojanin 1 in brain extracts inhibits their property to interact with partners proteins implicated in endocytosis (25). The effect of epsin and Eps15 phosphorylation in brain extracts on their interaction with the AP-2 subunit ␣-adaptin was therefore investigated. Brain cytosolic extracts were incubated in either phosphorylating or dephosphorylating conditions as described (25). Western blots of anti-epsin immunoprecipitates generated at the end of these incubations demonstrated an upper mobility shift of epsin in the phosphorylating conditions that could be reversed by treatment of the immunoprecipitates FIG. 3. Epsin is an in vitro substrate for the Cdc2 kinase, and its phosphorylation inhibits binding to ␣-adaptin. A, wild type (wt) and S328D DPW domains of epsin were incubated with purified Xenopus p34 cdc2 -cyclin B complex and separated by SDS-PAGE. The figure shows autoradiography and immunolabeling by an alkaline phosphatase method of the two proteins. B, the wild type DPW of epsin was phosphorylated in vitro by the p34 cdc2 -cyclin B complex and subsequently affinity-purified on the bead-immobilized appendage domain of ␣-adaptin. The figure shows 32 P autoradiograph and anti-epsin Western blotting of the phosphorylated starting material (SM) and of the material bound (B) and not bound (NB) by the beads. C, a Triton X-100 rat brain extract (SM, starting material) was affinity-purified on three different amounts (3, 6, and 9 g) of immobilized GST fusion proteins of wild type and S328D DPW domains of rat epsin, and bound ␣-adaptin was revealed by Western blotting.

FIG. 4. Epsin and Eps15 undergo stimulation-dependent dephosphorylation in brain, and their phosphorylation in brain extracts decreases the interaction with AP-2.
A, Western blots of total proteins of rat synaptosomes incubated in vitro for 1 min in the control condition or in the presence of 55 mM K ϩ . B, epsin immunoprecipitates generated from rat brain dephospho-(D) and phospho-(P) cytosol (see "Materials and Methods") were processed by anti-epsin Western blotting after incubation in the absence or in the presence of alkaline phosphatase (Alk. P.) or alkaline phosphatase and protein phosphatase inhibitors (P.I.) as indicated. C, rat brain dephospho-(black bars) and phospho-cytosol (shaded bars) was affinity-purified on three different amounts of immobilized ␣-adaptin appendage domain, and the amount of bound epsin and Eps15 was determined by Western blotting followed by band densitometry. Values are expressed in arbitrary units. D, anti-Eps15 and anti-epsin immunoprecipitates were generated from dephospho-(D) and phospho-cytosol (P) (see "Materials and Methods") and then processed by Western blotting for the proteins indicated. SM, starting material used for the immunoprecipitations.
with alkaline phosphatase (Fig. 4B) consistent with the interpretation that the mobility shift of epsin and Eps15 observed in Fig. 4A is due to dephosphorylation. Aliquots of the phosphoand dephospho-extracts were then incubated with the appendage domain of ␣-adaptin, and the bound material was analyzed by Western blotting. Quantitation of the blots demonstrated that binding to ␣-adaptin is inhibited by phosphorylation (Fig. 4C).
We next performed immunoprecipitation experiments from these extracts to test the effect of phosphorylation on the endogenous interactions within the cytosolic extract of epsin and Eps15 with AP-2. In both anti-Eps15 and anti-epsin immunoprecipitates coprecipitation of AP-2 was clearly decreased by the previous phosphorylation of the extracts (Fig. 4D). This result is consistent with an inhibitory effect of the phosphorylation of epsin and Eps15 on their interaction with AP-2, although they do not rule out a contribution of phosphorylation of AP-2 itself to their reduced interaction. DISCUSSION The results of this study demonstrate that the interaction of Eps15 and epsin with the ␣-adaptin subunit of AP-2 is modulated by phosphorylation.
Both Eps15 and epsin undergo phosphorylation in mitosis, a stage of the cell cycle where clathrin-mediated endocytosis is blocked (19,20). In mitosis, clathrin-coated pits are still present, but early stages of clathrin-coated pits (shallow domes and wide necks) predominate over late stages (narrow neck). This observation led to the speculation that invagination of clathrincoated pits is a processive reaction and that some proteins that play a key role in this process may be the target of mitotic kinases (19,20). Epsin/MP90 and Eps15 may be two such proteins. Both Eps15 and epsin were detected at clathrincoated pits but were found not to be co-enriched, together with clathrin, in clathrin-coated vesicles (7,18), suggesting that they are accessory and not key intrinsic components of the clathrin coat. The phosphorylation-dependent block of the interaction of epsin and Eps15 with AP-2 may dissociate them from the clathrin coat and play a role in the inhibition of the invagination process. If further studies confirm a role of epsin and Eps15 in the mitotic block of endocytosis, the elucidation of the function of these proteins may be crucial to understand the still elusive mechanism of clathrin-coated pit invagination.
Both Eps15 and epsin are also shown here to be regulated phosphoproteins in nerve terminals. They join the list of several other proteins implicated in the clathrin-dependent endocytosis of synaptic vesicles that undergo stimulation-dependent dephosphorylation (22,23,25,30,31). Based on the recent study of Slepnev et al. (25), it would appear that triggered dephosphorylation, which favors the assembled state of cytosolic endocytic proteins, represents a general and characteristic feature of "endocytic" proteins of nerve terminals. A reduced state of phosphorylation of endocytic proteins may increase the efficiency of the synaptic vesicle membrane internalization reaction. This effect may account for the enhanced rate of endocytosis observed in cultured neurons after incubation with the protein kinase inhibitor staurosporine (32) and is consistent with the inhibition of synaptic vesicle endocytosis produced by inhibitors of the Ca 2ϩ -dependent phosphatase calcineurin (32). Finally, dephosphorylation by a Ca 2ϩ -dependent phosphatase may play a role in the block of synaptic vesicle endocytosis produced by depletion of internal Ca 2ϩ (34). Eps15 is a substrate for EGF receptor kinase (8). We have found that epsin as well undergoes phosphorylation in response to EGF stimulation of fibroblastic cells, although this phosphorylation does not occur on tyrosine residues, indicating an indirect effect of EGF receptor kinase. 2 Based on our preliminary experiments, however, the EGF-dependent phosphorylation of Eps15 and epsin does not appear to decrease their interaction with AP-2, suggesting that these phosphorylation reactions play a role in cell physiology different from those described in this study.