The yeast Epsin Ent1 is recruited to membranes through multiple independent interactions.

In addition to its well known role in targeting proteins for proteasomal degradation, ubiquitin (Ub) is also involved in promoting internalization of cell surface proteins into the endocytic pathway. Moreover, putative Ub interaction motifs (UIMs) as well as Ub-associated (UBA) domains have been identified in key yeast endocytic proteins (the epsins Ent1 and Ent2, and the Eps15 homolog Ede1). In this study, we characterized the interaction of Ub with the Ede1 UBA domain and with the UIMs of Ent1. Our data suggest that the UIMs and the UBA are involved in binding these proteins to biological membranes. We also show that the Ent1 ENTH domain binds to phosphoinositides in vitro and that Ent1 NPF motifs interact with the EH domain-containing proteins Ede1 and Pan1. Our findings indicate that the ENTH domain interaction with membrane lipids cooperates with the binding of membrane-associated Ub moieties. These events may in turn favor the occurrence of other interactions, for instance EH-NPF recognition, thus stabilizing networks of low affinity binding partners at endocytic sites.

Endocytosis is a multistep process in which cells selectively internalize plasma membrane proteins and lipids, as well as extracellular macromolecules such as nutrients and peptide hormones. It is an effective way to control the composition of the plasma membrane, and thus the physiological responses of the cell, so it is a tightly regulated pathway. Many cytosolic proteins are required for the early stages of clathrin-dependent endocytosis, including the adaptor protein complex AP2, the coat protein clathrin, and more recently, a class of proteins called accessory factors (1,2). Because accessory factors interact with AP2 and are required for endocytosis, but are not enriched in purified clathrin-coated vesicles, they have been suggested to play a regulatory rather than a structural role in clathrin-coated vesicle formation (3). However, the precise functions, interactions, and order of operation of most accessory factors have not yet been determined.
One accessory factor of interest to many investigators is the protein epsin, which is conserved from yeast to humans. Rat epsin 1 (Eps15 interactor 1) was first identified in a yeast two-hybrid screen as a protein that binds to another endocytic accessory factor, Eps15 (4). Epsin and eps15 are localized to endocytic sites at the plasma membrane and interact directly through the Asn-Pro-Phe (NPF) tripeptide motifs of epsin binding to the Eps15 homology (EH) 1 domains of Eps15 (4). The yeast epsins Ent1 and Ent2 were similarly identified as binding partners of the EH domain-containing endocytic protein Pan1 (5,6). Several independent lines of evidence have indicated an important role for epsin in the internalization step of endocytosis. First, overexpressing fragments of epsin in cultured cells inhibits internalization of ligands such as transferrin and epidermal growth factor (4,7,8). Second, mutation of the yeast epsins results in growth defects and reduced internalization of plasma membrane lipids and proteins (6,9,10). Third, Drosophila melanogaster epsin (liquid facets) exhibits genetic interactions with genes encoding other endocytic machinery components (11). Fourth, in vitro studies show that epsin can stimulate clathrin polymerization into the spherical basket that encases endocytic vesicles (12). Finally, recent work suggests that epsin binding to phospholipids initiates the membrane curvature necessary to form an endocytic vesicle (13).
Epsins from all species share several features (Fig. 1A), including a conserved epsin amino-terminal homology (ENTH) domain that binds the phospholipid phosphatidylinositol-4,5bisphosphate (PI(4,5)P 2 ), two or three copies of the ubiquitin interaction motif (UIM, Fig. 1A) that binds ubiquitin (Ub), and a carboxyl-terminal region containing numerous short linear peptide sequences that serve as ligands for binding to components of the endocytic machinery such as AP2 (mammalian epsins), EH domains, and clathrin ( Fig. 1A; Refs. 4,6,7,10,and 14). To understand the function of such multivalent proteins, it is necessary to characterize the nature of each interaction in isolation as well as in the context of the entire protein, where cooperativity may influence some of its properties. In this study, we have defined the binding partners for each of the characterized domains of the yeast epsin Ent1, using in vitro recombinant protein binding, yeast two hybrid, and immobilized phospholipid binding experiments.
The yeast epsins are almost exclusively localized in cortical patches at the plasma membrane (6,9), but the cis-and transacting components necessary for their membrane association have not been determined. Here we show that Ub covalently attached to transmembrane proteins (often a requirement for internalization; reviewed in Ref. 15) or to peripheral membrane proteins mediates the recruitment of Ent1 to membranes. By examining the requirements for recruitment of Ent1 to biological membranes, we have uncovered evidence for cooperativity between the ENTH domain and the UIMs for efficient membrane binding. Other data indicate that the NPF-containing regions of Ent1 also contribute to additional layers of cooperative interactions. Together, our data are consistent with a model in which epsin might function as an endocytic adaptor protein. We suggest that epsin binds PI(4,5)P 2 -enriched membrane domains where it recruits components of the endocytic machinery and perhaps selects certain ubiquitinated transmembrane proteins for inclusion into endocytic vesicles.

EXPERIMENTAL PROCEDURES
Recombinant DNA Constructs-The Gal4AD and BD fusion constructs were prepared by ligating the appropriate PCR fragments (digested with BamHI-SalI) into the BamHI-XhoI and BamHI-SalI sites of the pGADT7 (LEU2, Clontech) and pGBT9 (TRP1, Clontech) vectors, respectively. To prepare His 6 -tagged and GST fusion proteins, cDNAs were subcloned into the pET28a (Novagen) or pGEX-4t-1 (Amersham Biosciences) vectors, respectively. Amino acid substitutions were made using the QuikChange XL site-directed mutagenesis kit (Stratagene, La Jolla, CA) according to the manufacturer's instructions. Murine p53 in pAS2-1 and SV40 large T-antigen in pACT2 were obtained from Clontech.
Ubiquitin and the terminal domain of clathrin cloned in pGEX-2t-1 were kindly provided by Dr. Scott Emr's lab (UCSD) and Dr. Peter McPherson (McGill University), respectively.

Recombinant His 6 -tagged and GST Fusion Protein Expression and Purification-Protein expression was induced in
Rosetta cells (Novagen) transformed with pET28a-or pGEX-derived constructs by the addition of isopropyl-␤-D-thiogalactopyranoside (1 mM final concentration) followed by a 5-h incubation at 30°C. The cells were harvested, resuspended in TBST (10 mM Tris, 140 mM NaCl, 0.1% Tween 20, pH 8.0), and incubated with 1 mg/ml lysozyme (Sigma) for 30 min at 4°C and sonicated. The lysates were centrifuged for 30 min at 15,000 rpm at 4°C in a Sorvall centrifuge, and the supernatants were collected.
For His 6 -tagged protein purification, the lysates were incubated in batch with nickel-nitrilotriacetic acid Superflow resin (Qiagen, Valencia, CA) in the presence of 5 mM imidazole (Sigma) for 1 h at 4°C. The unbound fraction was separated by centrifugation, and the resin containing the bound His 6 -tagged protein was packed into a disposable polypropylene column (Pierce). The column was washed with TBST/15 mM imidazole, and the protein was then eluted with TBST/150 mM imidazole, and fractions containing the His 6 -tagged protein were pooled. The imidazole from the His 6 -tagged protein and the glutathione from the GST fusion protein fractions were eliminated by using a PD-10 desalting column (Amersham Biosciences).
For GST fusion protein purification, the lysates were incubated in batch with agarose beads coupled to glutathione for 2 h at 4°C and then washed 4 times with TBST. The beads were then incubated with 20 mM glutathione (Sigma) in TBST for 30 min at room temperature, and the supernatant containing the purified fusion protein was collected.
GST and Ub-Agarose Pull-down Assays-All binding assays were incubated for 1 h at 4°C in the presence or absence of 10 mM free ubiquitin (Sigma). For GST pull-downs, glutathione-agarose beads (Sigma) loaded with either GST or GST fusion proteins were incubated with purified His 6 -tagged recombinant protein. For Ub binding, Ub coupled to agarose beads (Sigma) was incubated with either purified His 6tagged recombinant protein or 35 S-labeled in vitro transcription/translation products. After washing with TBST, bound radiolabeled samples were quantified in a scintillation counter, whereas all other samples were boiled in Laemmli's protein sample buffer and resolved by SDS-PAGE followed by immunoblotting using either a rabbit polyclonal antibody raised against the Ent1 protein (previously described in Ref. 9) or an anti-His 6 tag mouse monoclonal antibody (Clontech).
In Vitro Transcription/Translation and Binding of the in Vitro-transcribed/translated Products to Yeast Membranes-35 S-Labeled proteins were obtained by in vitro transcription/translation of pET28a constructs using the TNT T7 Quick coupled transcription/translation system (Promega, Madison, WI) and Easytag expression protein labeling mixture (PerkinElmer Life Sciences) according to the manufacturer's instructions. The transcription/translation products were diluted 1:100 in TBS, centrifuged (136,000 ϫ g, 1 h, 4°C), and their specific activity was determined by a conventional trichloroacetic acid precipitation assay (only tracers showing at least 10% incorporation were used for further experiments). The predominance of a single radiolabeled species was verified by SDS-PAGE followed by autoradiography. The concentration of radioactive material for different mutants was normalized according to 35 S incorporation and the number of methionines present in the corresponding Ent1 truncations.
3-10 ϫ 10 6 cpm of 35 S-labeled proteins were incubated with membrane suspensions prepared from 1 A 600 of cells in a total volume of 300 l, for 1 h at either room temperature or 4°C, and in the presence or FIG. 1. The bacterially produced yeast epsin Ent1 interacts with ubiquitin through its UIMs in a cooperative manner. A, Ent1 domain organization. The scheme depicts the multimodular domain architecture of the yeast epsin Ent1 (from amino to carboxyl terminus): ENTH domain, UIMs, NPF tripeptides, and a CBD. B, Ent1 binds ubiquitin. Agarose beads were incubated with purified His 6 -Ent1 for 1 h at 4°C. After washing, bound His 6 -tagged protein was visualized by chemiluminescent Western blotting with an anti-His 6 antibody. Starting material (lane 1) and fractions bound to Ub-agarose beads in the presence or absence of 10 mM Ub (lanes 3 and 2, respectively). C, Ent1 interaction with Ub is mediated by its UIMs in a cooperative fashion. Ub-agarose beads were incubated with 35 S-labeled Ent1 WT and UIM-double and single S 3 D mutants (see text for details) for 1 h at 4°C and washed, and the bound radioactivity (in cpm) was measured in a scintillation counter. Values are the means Ϯ S.D. of triplicates from one of three independent experiments. Statistical significance in comparison with binding showed by WT Ent1 is indicated by ** (p Ͻ 0.005).

Ent1 Recruitment to Membranes
absence of 10 mM purified ubiquitin. After washing, membrane associated radioactivity was measured in a scintillation counter. Experiments were conducted in triplicate, and their S.D. was calculated. Slightly different washing procedures resulted in some variability in background between experiments. Another source of variability may be caused by differences in the initial status and in the preservation (from batch to batch and time of storage) of the two major reagents used in these assays: (i) the membrane suspensions (composition of phosphoinositides and ubiquitinated proteins); (ii) the 35 S-labeled proteins (concentration of binding-competent species). Each experiment was performed two to four times (as indicated in the corresponding figure legend) and always showed similar trends.
Binding of Recombinant Proteins to Phosphoinositides Immobilized onto Nitrocellulose-Nitrocellulose membranes containing immobilized phospholipids (PIP-Strips membranes, Echelon, Salt Lake City, UT), blocked with TBST with 3% fatty acid-free bovine serum albumin (Sigma), were incubated with purified His 6 -tagged recombinant protein for 1 h at room temperature. The bound proteins were detected by immunoblotting using an anti-His 6 tag mouse monoclonal antibody (Clontech).
Rabbit Antisera-Anti-Ent1/2 serum was described (9). Anti-Ub serum was a gift of Cecile Pickart (The Johns Hopkins Bloomberg School of Public Health). Anti-Ede1 serum was produced in our lab against a peptide and preincubated with extract from ede1⌬ cells. (10). We confirmed and further characterized the specificity of this interaction. In an in vitro pull-down assay, we found that recombinant His 6 -Ent1 protein bound Ub-agarose beads as assessed by immunodetection of starting and bound material with an anti-His 6 antibody (Fig. 1B, lanes 1 and 2). Importantly, this interaction was specifically inhibited by the presence of an excess (10 mM) of free Ub (Fig. 1B, lane 3).

Ent1 Binds Ub via Its UIMs in a Cooperative Fashion-It has been shown previously that Ent1 binds Ub via its UIMs
We next investigated whether both UIMs are necessary for the recognition of Ub. Both conserved serines within the Ent1 putative UIMs (Ser 177 and Ser 201 ) were mutated to aspartate, glutamate, lysine, arginine, or cysteine residues, which were selected based on random permutation probabilities according to Dayhoff's PAM250 matrix (i.e. the residue substitutions least likely to cause structural perturbation in the protein). 35 S-labeled His 6 -Ent1 wild type, single, and double mutants were produced by in vitro transcription/translation, incubated with Ub-conjugated agarose beads, and bound radioactivity was quantified. Compared with Ent1 WT, all five of the Ent1 UIM double mutants exhibited significantly impaired ubiquitin binding (data not shown), and as demonstrated by the 34% reduction for the S 3 D double mutant (Fig. 1C). Also, mutation of a single conserved serine within either UIM decreased the binding of the mutant proteins for Ub-agarose beads by at least 50% (Fig. 1C). These results suggest that the two UIMs cooperate for the interaction of Ent1 with Ub.

Ent1 Binds Yeast Membranes in a Ub-and UIM-dependent and Cooperative
Fashion-It has been demonstrated that certain cell-surface transmembrane proteins, such as Ste2 and Ste6, are ubiquitinated as a prerequisite to their internalization (17,18). Thus, it is possible that these and/or other Ubmodified proteins such as peripheral membrane proteins may act as docking sites for Ub-binding proteins such as Ent1. To test this possibility, we prepared a crude membrane fraction from yeast that contains plasma membrane, which is the major localization site of many endocytic proteins in vivo ( Fig. 2A), including the yeast epsins (6). We confirmed by Western blotting that these membranes contained an abundance of Ubconjugated proteins ( Fig. 2A), and tested whether these membranes could recruit recombinant His 6 -Ent1 in a Ub-dependent manner.
Purified His 6 -Ent1 was incubated with a suspension of membranes for 1 h at 4°C in the presence or absence of 10 mM free Ub. After washing, the presence of the recombinant protein associated with the membrane pellet was assessed by SDS-PAGE, followed by Western blotting with an anti-His 6 antibody. Our results showed that His 6 -Ent1 interacts with membranes prepared from either MATa (Fig. 2B, lane 1) or MAT␣ cells (Fig. 2B, lane 3). 10 mM free Ub greatly reduced this membrane recruitment (Fig. 2B, lanes 2 and 4). We also observed that even in cases where our His 6 -Ent1 preparations contained some protein degradation products (some as abundant as the intact protein), only full-length Ent1 was recruited to the yeast membranes. Additionally, an excess of yeast cytosolic proteins did not compete the interaction (data not shown).
To characterize further this membrane association process, we studied the recruitment of Ent1 UIM mutants. For these experiments, 35 S-labeled in vitro transcription/translation products were incubated with membrane suspensions for 1 h at 4°C, and after washing, the radioactivity associated with the membrane pellet was quantified. As shown in Fig. 2C, and

Ent1 Recruitment to Membranes
consistent with their reduced binding to Ub-agarose, the S 3 D single and double mutants exhibited at least 50% reduced recruitment to membranes as compared with WT Ent1. We observed reduced membrane binding with all five UIM double mutants (S 3 D, S 3 E, S 3 K, S 3 R, and S 3 C), and the reduction was comparable with the binding of WT Ent1 to membranes in the presence of free Ub (ranging from 40% to 60% inhibition, depending on the experiment). The remaining bound material may be nonspecific binding or, alternatively, could represent a Ub-independent component of membrane recruitment (see below). The finding that a single mutation introduced into either the first or second UIM of Ent1 was sufficient to drastically decrease the membrane binding capability of the protein is again consistent with cooperative binding of the two UIMs to Ub.
We also carried out saturation experiments by incubating increasing concentrations of 35 S-labeled Ent1 with a constant amount of membrane in the presence or absence of free Ub (data not shown). Although we were not able to confirm that binding equilibrium was achieved after 1 h at 4°C, we roughly estimated an avidity dissociation constant of ϳ20 M for each UIM.
The ENTH Domain Cooperates with the UIMs for Membrane Recruitment-The ENTH domains of mammalian epsins bind phosphoinositides (7); therefore, we tested whether the Ent1 ENTH domain also interacts with lipids, thus possibly contributing to Ent1 membrane association. Purified His 6 -tagged proteins plus 3% fatty acid-free bovine serum albumin were incubated with nitrocellulose strips containing a variety of immobilized phospholipids. After washing, the bound proteins were detected using an anti-His 6 antibody and visualized by chemiluminescence. Both His 6 -Ent1 1-454 (full-length) and His 6 -Ent1 1-149 (ENTH domain) bound phosphoinositides immobilized onto nitrocellulose (Fig. 3A). As expected, an Ent1 amino-terminal truncation lacking the ENTH domain (Ent1 149 -454 ) did not exhibit phospholipid binding activity (Fig.  3A). It should be noted that, because of the qualitative nature of this binding assay, slight differences observed between fulllength and the ENTH domain binding were considered insignificant.
Given that the Ent1 ENTH domain binds phosphoinositides and its UIMs bind ubiquitin, we next explored a possible interplay between the UIMs and the ENTH domain for membrane recruitment. To this end, we prepared two Ent1 truncations (see Fig. 1A), one lacking the first 148 amino acids that comprise the ENTH domain (Ent1 149 -454 , containing both UIMs plus the carboxyl-terminal part of the protein) and another consisting of amino-acids 1-240 (Ent1 1-240 , containing the ENTH domain plus both UIMs and a putative coiled coil region).
We assayed the binding of the Ent1 1-240 and Ent1 149 -454 truncations at room temperature rather than 4°C to preserve physiological membrane characteristics such as membrane lipid fluidity. As before, 35 S-labeled in vitro transcription/translation products were incubated with membranes, and the membrane-associated radioactivity was quantified. Each Ent1 truncation bound the membranes, consistent with Ub-dependent recruitment. However, we found that Ent1 1-240 bound membranes to a greater extent than Ent1 149 -454 (Fig. 3B). This suggested that, in addition to the UIMs, the ENTH domain also contributed to the membrane recruitment of Ent1. Moreover, we also found that the highly phosphorylated soluble inositol phosphate IP 6 (phytic acid) that binds ENTH domains (13,19) was able to reduce significantly (by 20 -40%) some of the membrane recruitment of Ent1 1-240 (Fig. 3B). In contrast, efficient membrane recruitment of Ent1 1-240 was still observed on in-cubation with IS 6 (myoinositol hexasulfate) (Fig. 3B). Membrane recruitment of the Ent1 149 -454 truncation was the same in the presence of either IP 6 or IS 6 (Fig. 3B). These results further support the idea that the interaction of the Ent1 ENTH domain with phosphoinositides plays a role in its membrane association. It should be noted that the precise phosphoinositide content of these membranes was not characterized.
When studied at 4°C, both Ent1 1-240 and Ent1 149 -454 truncated proteins showed levels of membrane recruitment similar to full-length Ent1 (data not shown), which suggests a predominant role for the UIMs (common to both truncations, see Fig.  1A) in membrane recruitment at low temperature. It is possible that when the binding to membranes is instead conducted at 4°C, the ENTH domain interaction with lipids may be disfavored because of membrane rigidity and weakened entropic contributions of the hydrophobic contacts. Taken together, these data suggest that, whereas the contribution of the UIM-Ub interaction to the overall membrane binding is favored at 4°C, the ENTH domain provides a significant contribution to membrane binding at room temperature, which is more similar to physiological conditions, and may be more akin to what occurs in vivo.
The UBA Domain of Ede1 Binds Yeast Membranes in a Ub-dependent Manner-Ede1 is an endocytic protein (the yeast homolog of eps15) with a multimodular domain architecture, including 3 EH domains and a UBA domain (14,20). Because the Ede1 UBA domain binds Ub (10), 2 and given that this

Ent1 Recruitment to Membranes
protein is also found almost exclusively associated with membranes (20), 2 we tested whether the Ede1 UBA domain interacts with our membrane preparations similarly to the Ent1 UIMs. As shown in Fig. 4B, the UBA domain of Ede1 (residues 1217-1382) was able to bind biological membranes, and in two independent experiments this recruitment was impaired (Ϸ30% inhibition in the experiments shown) by the presence of free Ub. These experiments required the use of high concentrations of 35 S-labeled UBA domain and membranes to detect significant binding, which could suggest a weaker interaction than that of the UIM/Ub complex. However, by analogy to the cooperative membrane binding observed among different domains of Ent1, it is possible that full-length Ede1 may bind membranes more efficiently than its UBA domain alone.
Ent1 Is Involved in a Network of Interactions with Other Endocytic Proteins That Localize to Membranes-Ede1 exhibits a cortical patch membrane localization pattern similar to that of yeast epsins (6,9,20,21). 2 It is possible that Ede1 and Ent1 colocalize through recognition of ubiquitinated membrane proteins. Alternatively, they could colocalize through an EH/NPF interaction, because the yeast epsins were originally identified in a yeast two-hybrid screen with the EH domains of Pan1 (5,6). We found that Ent1 is indeed able to interact in a twohybrid system with Ede1 EH domains, specifically with the third EH domain (EH3), and not the first or second EH domains (Fig. 5A). The carboxyl-terminal Ent1 149 -454 protein was sufficient to bind the Ede1 EH3 domain, consistent with the NPF motifs being the binding determinants involved in this interaction (Fig. 5A). In fact, we established that this interaction is indeed mediated by Ent1 NPF motifs, as mutation of both NPF motifs to NPM abolished the binding (Fig. 5B). Mutation of a single NPF motif decreased but did not abrogate the EH domain-dependent recognition, suggesting that each Ent1 NPF motif participated in the interaction with Ede1 (Fig. 5B). Interestingly, we also found that the Ent1 interaction with Pan1 is much weaker than with Ede1 (Fig. 5A), suggesting a preference for Ede1 over Pan1.
To confirm the Ent1/Ede1 interaction, we performed GST pull-down experiments using a GST-EH3 fusion protein. As shown in Fig. 5C, His 6 -Ent1 bound Ede1 EH3, whereas mutation of the first NPF motif to NPM significantly decreased the interaction. We previously showed that Ent1 binds clathrin through a carboxyl-terminal clathrin-binding domain (CBD) (6), and this experiment also confirmed that the Ent1 CBD binds the amino-terminal domain of the clathrin heavy chain (22) (Fig. 5C). Deleting the last four residues of Ent1 that constitute the CBD resulted in no binding of the clathrin amino-terminal domain, suggesting that the carboxyl-terminal CBD is the unique CBD present in Ent1. GST and GST-Ub fusion proteins were used as negative and positive controls, respectively (Fig. 5C).

DISCUSSION
The mechanisms mediating the recruitment of the endocytic molecular machinery at sites where internalization takes place are not completely understood. In this study, we present evidence suggesting that the yeast epsin Ent1, an integral part of the endocytic machinery, is recruited to biological membranes via multiple, cooperative interactions with lipids and proteins through its numerous binding domains. Importantly, we show that biological membranes containing plasma membrane proteins, endocytic machinery, and ubiquitinated proteins can recruit epsins and Ede1. It has not yet been clearly determined which specific ubiquitinated proteins are the targets in these membranes; however, an attractive possibility would be membrane receptors and/or ubiquitinated endocytic machinery. In fact, preliminary evidence from cross-linking His 6 -Ent1 to MAT␣ membranes indicated that ubiquitinated Ste3 is one of the interacting partners for Ent1. 2 Many studies have uncovered roles for ubiquitination in the endocytosis and lysosomal/vacuolar degradation of membrane receptors (reviewed in Ref. 15), which may be facilitated by Ub binding to some endocytic proteins (10,14,23). We have demonstrated that Ent1 binds Ub in a UIM-dependent manner, and that, similarly to Vps27 (10), the two Ent1 UIMs cooperate in this interaction. Whether each UIM binds to different Ub moieties to generate a stable multivalent-complex or instead both UIMs together bind a single Ub unit requires further investigation. A UIM is predicted to form an ␣-helix (14), and other Ub-binding domains such as UBA and CUE domains are composed of three ␣-helices. In fact, the CUE domain binds Ub via an interface of two ␣-helices, 3 so it is possible that two UIMs might similarly join to form a binding site for a single Ub.
We propose that at 4°C, Ent1 recruitment to membranes is primarily UIM-dependent, as binding is decreased (from 40 to 60%) either by addition of free ubiquitin or by mutation of the UIMs. Similarly to the cooperative Ub/UIM interaction we observed with recombinant protein, our studies showed that the individual UIMs cooperate for membrane association. Analysis of the membrane binding data, considering the individual contributions of each UIM bound to Ub ([UIM1 ⅐ Ub], [UIM2 ⅐ Ub], and [UIM1 ⅐ Ub ⅐ UIM2] complexes), gave an estimated apparent dissociation constant of ϳ20 M, which is similar to the binding affinity of Ub to the single UIM present in the mammalian protein Hrs (24). However, a remarkable consequence of the UIM-bivalency of Ent1 is that the effective overall avidity of the protein for membrane-linked Ub falls in the 10 Ϫ8 M range (data not shown). Additionally, the apparently high affinity of Ent1 for membranes could represent Ent1 cooperativity, as a Hill coefficient significantly greater than 1 (n H ϭ 1.78 Ϯ 0.04) was calculated from membrane binding experiments and Ent1/ Ent1 interactions were observed in a yeast two-hybrid study. 2 Because a wild-type ENTH domain of yeast epsin is required for endocytosis and viability in yeast (6), ENTH domain-binding activity may be most central to the endocytic function of epsin. Previous studies have shown that ENTH domains bind phosphoinositides (7), and we have confirmed this for the Ent1 ENTH domain. Thus, it was surprising to find that a major contribution to membrane binding is provided by the recognition of Ub by the Ent1 UIMs (at 4°C). These results suggest that alteration of membrane biophysical properties due to lipid phase transitions (likely to occur at temperatures below 20°C) and/or the weakening temperature-dependent entropic interactions (the primary driving force for hydrophobic contacts) would result in unfavorable conditions for ENTH domain binding to membranes at 4°C. Supporting this interpretation, our experiments conducted at room temperature indicated that the ENTH domain not only binds lipids, but also cooperates with the UIMs for binding to membranes. The order of events for ENTH domain-and UIM-mediated recruitment of Ent1 to sites of endocytosis remains to be determined. However, one possibility is that the ENTH domain localizes the epsins to phosphoinositide-rich plasma membrane domains, whereas the UIMs provide specific targeting to ubiquitinated transmem-brane and/or peripherally associated proteins.
It should be noted that our recruitment to membranes assay using radiolabeled protein resulted in a degree of background variability for each experiment, although all experiments showed similar statistically significant trends. We are currently working on optimizing reconstitution assays using liposomes and/or noninvasive in vivo techniques. However, these membrane binding experiments do have the advantage of using biological materials that provide an indication of in vivo interactions.
We have shown Ub-dependent membrane recruitment of the Ede1 UBA domain, suggesting that Ede1 may also bind ubiquitinated membrane-associated proteins. We also found that Ede1 and Ent1 bind one another through a synergistic EH/NPF interaction, which illustrates a recurring theme of cooperativity for Ent1 interactions. Because an Ent1 amino-terminal truncation (Ent1 149 -454 ) still bound membranes (Fig. 3B), it is possible that EH domain-containing proteins like Ede1 may also recruit Ent1 to membranes via EH/NPF interactions, independently of Ent1 binding to lipids.
Additionally, and as expected, the Ent1 carboxyl-terminal CBD was found to be sufficient for binding to the terminal domain of clathrin. Because a yeast strain lacking the five protein complexes previously recognized as homologues of clathrin assembly factors and/or adaptors (deleted for genes encoding yAP180A, yAP180B, and the ␤-subunits of AP1, AP2, and AP3; Ref. 25) is viable, can perform endocytosis, and form clathrin-coated vesicles, it is clear that other factors remain to support these activities. Epsin is an obvious candidate, particularly in light of the recent demonstration that rat epsin can stimulate clathrin cage assembly (12).
Based on our results, we propose a model (Fig. 6) in which Ent1 is recruited to nascent endocytic sites as a combined result of its lipid-membrane binding property (through its ENTH domain) and its UIM-mediated specificity for ubiquitinated proteins, depicted here as cargo. This could be followed by the recruitment of Ede1 that may also interact with cargo and/or covalently linked Ub moieties via its UBA domain and cooperatively bind Ent1 NPFs through its EH domain. In turn, other accessory proteins could bind to both Ent1 and Ede1 and the endocytic coat would be assembled by interaction with the Ent1 CBD, among other clathrin-binding determinants. Therefore, we suggest that the epsins may play a key role as adaptors 3 J. Hurley, personal communication.
FIG. 6. Model of Ent1 interactions with lipids and proteins at the plasma membrane. Each domain of Ent1 interacts with one or more partners. The amino-terminal ENTH domain binds to phosphoinositides (such as PI(4,5)P 2 : PIP 2 ) present at the plasma membrane, possibly as an early step of the membrane recruitment process. The Ub moieties covalently attached to membrane proteins may serve as docking sites for Ent1 (via its UIM motifs) as well as Ede1 (through its UBA domain). Interaction between the NPF motifs of Ent1 and the EH domain of Ede1 could further stabilize the multivalent complex. Clathrin and accessory proteins may be also recruited to form part of the endocytic network.
linking ubiquitinated cargoes to the endocytic machinery, recruiting proteins with specific activities and/or exerting undiscovered functions by themselves that make endocytosis possible.
One interesting prediction of our model is that multiple mutations would be necessary to affect significantly the function/localization of the epsins. In agreement with this hypothesis, neither mutation nor deletion of the Ent1 UIMs is sufficient to render a phenotype in yeast cells expressing the mutant protein as the exclusive source of epsin (10). 2 Our model predicts that such UIM mutants would still be able to properly localize at endocytic sites via their ENTH and NPF interactions, consistent with the finding by Shih et al. (10) that both deletion of Ede1 and Ent1 UIM inactivation are required for a significant endocytic defect in vivo. In fact, mutating the ENTH domain is the only known case in which single-domain mutations produce a phenotype (6). This is consistent with the findings of Ford et al. (13) and supports our model in which epsin binds to membranes via its ENTH domain as an early and required step for membrane localization.
We also expect Ent1 recruitment to membranes to be highly regulated by reversible post-translational modifications such as phosphorylation (9,23,26) that may influence the stability of the endocytic network. For example, phosphorylation of Ent1 Thr 394 and Thr 416 , flanking the second NPF motif, has been shown to affect the endocytic function of Ent1 (9). Consistent with studies of phosphoregulation of endocytic complex function in nerve terminals (25), our preliminary data indicate that this modification impairs the Ent1/Ede1 interaction. 2 On the other hand, it was recently demonstrated that UIM-containing proteins can be ubiquitinated in an UIM-dependent manner (23). Although the functional consequences of covalent Ub modification of epsins are unknown, it is possible that an intramolecular Ub/UIMs interaction may compete for the recognition of Ub linked to cargo by the UIMs (27).
In summary, we provide evidence suggesting that the epsins, as well as other proteins such as Ede1, are recruited to membranes through a network stabilized by cooperative interactions throughout their multiple modular domain architecture. Future in vitro reconstitution studies will be necessary to de-termine the order of action of these and other proteins in initiating and assembling the endocytic machinery.