Specific Interaction between SNAREs and Epsin N-terminal Homology (ENTH) Domains of Epsin-related Proteins in trans -Golgi Network to Endosome Transport*

SNARE proteins on transport vesicles and target membranes have important roles in vesicle targeting and fusion. Therefore, localization and activity of SNAREs have to be tightly controlled. Regulatory proteins bind to N-terminal domains of some SNAREs. vti1b is a mammalian SNARE that functions in late endosomal fusion. To investigate the role of the N terminus of vti1b we performed a yeast two-hybrid screen. The N terminus of vti1b interacted specifically with the epsin N-terminal homology (ENTH) domain of enthoprotin/ CLINT/epsinR. The interaction was confirmed using in vitro binding assays. This complex formation between a SNARE and an ENTH domain was conserved between mammals and yeast. Yeast Vti1p interacted with the ENTH domain of Ent3p. ENTH proteins are involved in the formation of clathrin-coated vesicles. Both epsinR and Ent3p bind adaptor proteins at the trans -Golgi network. Vti1p is required for multiple transport steps in the endosomal system. Genetic interactions between VTI1 and ENT3 were investigated. Synthetic defects sug-gested that Vti1p and Ent3p cooperate in transport from the trans -Golgi network to the prevacuolar endosome. Our experiments identified the first cytoplasmic protein binding

SNARE 1 proteins on transport vesicles and target membranes mediate recognition and fusion between membranes by complex formation with other SNAREs via SNARE motifs (1). SNAREs can be subdivided into four different groups (R-, Qa-, Qb-, and Qc-SNAREs, with R and Q being the single-letter abbreviations of arginine and glutamine, respectively) according to similarities in their amino acid sequences. All well characterized SNARE complexes are composed of four different SNARE helices, one from each group. The conserved SNARE motif is located next to a C-terminal transmembrane anchor in most SNAREs. Many SNAREs contain N-terminal domains that fold independently and are highly divergent in their amino acid sequences. The N-terminal domains of SNAREs belonging to syntaxins or to Qa-SNAREs can bind proteins that regulate SNARE complex formation. Other binding partners recruit Qaor R-SNAREs into budding vesicles. The N-terminal domains of the R-SNAREs Sec22p (2) and Ykt6p (3) form a mixed ␣-helical/ ␤-sheet profilin-like fold. By contrast, the N-terminal domains of Qa-SNAREs consist of a three-helix bundle.
We set out to identify interaction partners for Qb-SNAREs, which function in endosomal traffic. The yeast Qb-SNARE Vti1p is required for the following: (i) transport from the TGN to the prevacuolar endosome; (ii) traffic to the vacuole (equivalent to the mammalian lysosome); (iii) retrograde traffic to the cis-Golgi; and (iv) homotypic TGN fusion as part of four different SNARE complexes (4 -6). Mammalian cells contain two homologs of yeast Vti1p. vti1a and vti1b share 30% of their amino acid residues with each other as well as with Vti1p. vti1a and vti1b have distinct subcellular localization, distinct SNARE partners, and more specialized roles than their yeast counterpart. The mammalian SNAREs vti1a, syntaxin 6, syntaxin 16, and VAMP-4 form a complex required for retrograde transport from the endosome to the TGN and for early endosome fusion (7)(8)(9). A complex consisting of vti1b, syntaxin 8, syntaxin 7, and endobrevin/VAMP-8 is involved in fusion of late endosomes (7). Syntaxin 7, probably as part of the same SNARE complex, is also required for transport from the late endosome to the lysosome (10,11). The N-terminal domains of vti1b, syntaxin 7, and syntaxin 8 have high ␣-helical contents. This domain consists of a three-helix bundle in the Qa-SNARE syntaxin 7 as well as in the Qb-SNARE vti1b. However, the N-terminal domain seems to have a different function in syntaxin 7 than in the other two proteins. The N terminus of syntaxin 7 interacts with its own SNARE domain, resulting in inhibition of SNARE complex formation. By contrast, vti1b and syntaxin 8 do not adopt this closed conformation (12). The N-terminal domain of syntaxin 8 most likely also forms a threehelix bundle, because this structure is formed in the related Qc-SNARE syntaxin 6 (13).
Vti1a and vti1b are related but functionally distinct proteins. Here, we have used the yeast two-hybrid screen to identify a specific binding partner for the N-terminal domain of vti1b.
The Strep-tag system was obtained from IBA (Göttingen, Germany). All other reagents were purchased from Sigma.
Plasmid manipulations were performed in the Escherichia coli strains XL1Blue or BL21(DE3) using standard media. Yeast strains (Table I) were grown in rich YEPD medium (1% yeast extract, 1% peptone, and 2% dextrose) or standard minimal medium (SD) with appropriate supplements.
Yeast Two-hybrid Assays-The yeast two-hybrid screen was performed as described (15) in L40 cells using pLexN with the N-terminal domain of mouse vti1b (amino acids 1-128) as bait vector and a library derived from rat E18 embryonic brain RNA in pVP16 -3 as prey vector (16). Positive cells were selected on plates with minimal medium lacking uracil, lysine, tryptophan, leucine, and histidine complemented with 2.5 mM 3-aminotriazole. Yeast cells containing pairs of LexA DNA-binding domain and VP16 activation domain fusions were streaked out on selective plates to study specific two-hybrid interactions.
In Vitro Binding Assays-ENTH domains of epsinR and Ent3p with a C-terminal Strep-tag were expressed in E. coli using the plasmid pASK-IBA3 and purified using Strep-Tactin affinity columns according to the manufacturer's instruction (IBA). The cytosolic domains of vti1a, vti1b, and Vti1p were expressed with an N-terminal His 6 tag in E. coli and purified using Ni-NTA-agarose (Qiagen). The purified recombinant proteins were analyzed by SDS-PAGE and Coomassie Blue staining. Purity was estimated to be Ͼ90%.
To study in vitro binding of bacterially expressed purified proteins, 4.4 g of epsinR-Strep and 6 g of His 6 -vti1b or His 6 -vti1a were mixed in 100, 200, 500, or 1000 l of PBS and 1% Triton X-100, resulting in final concentrations of 2, 1, 0.2, or 0.1 M, respectively, for each protein. The mixtures were incubated for 30 min at 4°C, Ni-NTA beads were added, and the incubation was continued for another 30 min. After quick washes, the pellets were separated by SDS-PAGE and transferred to nitrocellulose, and Strep-tagged proteins were detected by Strep-Tactinhorseradish peroxidase (IBA) and enhanced chemiluminescence.
Liposomes were generated, and liposome binding was performed as described (17). 100 g of liposomes (70% phosphatidylcholine, 20% phosphatidylethanolamine, and 10% variable lipids) were incubated with 4 g of the Strep-tagged ENTH domain of Ent3p in 150 l of 25 mM Hepes, pH 8, and 50 mM NaCl at 4°C for 15 min. Liposomes were pelleted at 27,000 ϫ g max for 20 min. After a rinse with buffer, liposome pellets were analyzed by SDS-PAGE and stained with Coomassie Blue.
Transport Assays-Yeast cells were grown in log phase at 24°C and 0.5 OD and preincubated at the indicated temperature for 15 min. Cells were labeled with [ 35 S]methionine (100 Ci/0.5 OD) for 10 min, chased for 30 min unless indicated otherwise, and carboxypeptidase Y (CPY) was immunoprecipitated from cellular extracts and medium, or alkaline phosphatase (ALP) was immunoprecipitated from cellular extracts as described (5,18,19). CPY and ALP antisera were kindly provided by T.H. Stevens. Immunoprecipitates were analyzed by SDS-PAGE and autoradiography. A BAS1000 (Fuji) was used for quantification.

RESULTS
To identify proteins that bind to the N terminus of vti1b, we performed a yeast two-hybrid screen. The bait protein consisted of the DNA-binding domain of LexA and amino acid residues 1-128 of vti1b. This sequence was chosen because it is highly divergent between different SNAREs. The SNARE motif starts at amino acid residue 146, and the third helix of the N-terminal helix bundle ends at amino acid residue 98 (12). An interacting plasmid was isolated multiple times from a library encoding fusions of the VP16 activation domain with rat embryonic brain cDNAs (Fig. 1a). This plasmid encoded a fulllength ENTH domain protein that was recently described as enthoprotin, CLINT, and epsinR (20 -23). epsinR has a Nterminal ENTH domain with an affinity for several phosphoinositides, especially for phosphatidylinositol 4-phosphate, and binding sites for clathrin as well as for the TGN adaptor proteins AP1 and GGA2 in the rest of the molecule. Both domains were tested separately for two-hybrid interaction with vti1b. vti1b interacted with the ENTH domain of epsinR but not with the C terminus. This interaction was specific, as vti1b did not exhibit a two-hybrid interaction with ENTH domains of epsin1 or the related ANTH domains of AP180 or Hip1-R (24). The N terminus of vti1a did not interact with the ENTH domains tested, further emphasizing the specificity. To confirm the twohybrid interaction, a bacterially expressed ENTH domain of epsinR with a C-terminal Strep-tag was incubated with the cytosolic domains of vti1b or vti1a as His 6 fusion proteins (Fig.  1b). The ENTH domain of epsinR bound to His 6 -vti1b but not to His 6 -vti1a.
Next, we investigated whether this interaction between a SNARE and an ENTH domain was conserved between mammals and yeast. The yeast genome encodes eight ENTH and ANTH domain proteins, four of which form pairs of closely related proteins (Ent1p with Ent2p and yAP1801 with yAP1802). The N-terminal domain of yeast Vti1p was expressed as a fusion protein with the LexA DNA binding domain. Different ENTH domains were fused to the VP16 activation domain. The N-terminal domain of Vti1p interacted specifically with the ENTH domain of Ent3p but not with Ent1p, Ent4p, Ent5p, Sla2p, or yAP1801 in the yeast twohybrid assay (Fig. 2a). Additionally, Vti1p did not interact with epsinR, nor did vti1b interact with Ent3p in the two-hybrid assay (data not shown). In a pull-down assay, the ENTH domain of Ent3p with a Strep-tag-bound His 6 -Vti1p at high pro-  (Fig. 2b), confirming the significance of the two-hybrid data. Ent3p binds the clathrin adaptor Gga2p and has been implicated in clathrin-coated vesicle formation at the Golgi or endosome (25). Different phosphoinositides are regulators of trafficking in these organelles in yeast (26). Phosphatidylinositol 4-phosphate is implicated in secretory traffic from the Golgi to the plasma membrane, and phosphatidylinositol 3-phosphate is implicated in traffic to the endosome and autophagocytosis. Phosphatidylinositol 3,5-biphosphate is required for sorting into the internal vesicles in prevacuolar endosomes to form multivesicular bodies. We tested whether the Strep-tagged ENTH domain of Ent3p pelleted with liposomes containing these phosphoinositides. Ent3p bound specifically to liposomes with phosphatidylinositol 3-phosphate and phosphatidylinositol 3,5-biphosphate (Fig. 3).
Several deletion strains were generated to investigate the function of ENT genes in membrane traffic. The newly synthesized endoplasmic reticulum-modified form of the vacuolar carboxypeptidase Y (p1CPY) is modified to p2CPY in the Golgi. p2CPY is transported from the TGN to the prevacuolar endosome and, in a second step, to the vacuole where it is cleaved to mature CPY (27). p2CPY is secreted from the cells if transport   3. Ent3p binding to phosphoinositides. The ENTH domain of Ent3p with a C-terminal Strep-tag was incubated with liposomes containing 10% of the indicated lipids. PC, phosphatidylcholine; PI, phosphatidylinositol; PI3P, phosphatidylinositol 3-phosphate; PI4P, phosphatidylinositol 4-phosphate; PI3,5P 2 , phosphatidylinositol 3,5-biphosphate. Liposomes were pelleted, and recruitment of Ent3p was followed by separation of the pellet fraction by SDS-PAGE and Coomassie blue staining. 30% is the percentage of the protein used in the assay that was loaded for comparison. from the TGN to the vacuole is defective. CPY transport was analyzed by pulse-chase labeling followed by immunoprecipitation. No defects in CPY maturation were observed in ent4⌬ and ent5⌬ cells (Fig. 4). Slightly elevated levels of p2CPY were secreted into the medium (shown as E on Fig. 4) by ent3⌬ and ent3⌬ent4⌬ cells, even though most CPYs reached the vacuole (mature CPY in the intracellular fraction). ent3⌬ ent5⌬ cells secreted even more p2CPY. This slight transport defect indicates that Ent3p and Ent5p have a partially redundant function in CPY transport, but other proteins can almost compensate for their loss. We used different vti1 mutant strains to investigate whether Ent3p or Ent5p have a functional connection with Vti1p and in which trafficking steps. vti1-1 cells are only defective in transport from the Golgi to the prevacuole (4). The absence of Ent3p reduced mature CPY levels in vti1-1 cells from 64.3% (S.D. 12) to 11.9% (S.D. 5.8; n ϭ 3) at a semipermissive temperature (Fig. 5a). Lack of Ent5p did not lead to a synthetic defect. vti1-2 cells are defective in transport from the Golgi to the prevacuole as well as in transport to the vacuole (5). Vacuolar maturation of the pro form of ALP was used as an assay for fusion with the vacuole, because the pALP travels from the Golgi to the vacuole without passage through the prevacuole (27). A deletion of ENT3 did not result in reduced appearance of vacuolar mature ALP in vti1-2 cells at a semipermissive temperature (Fig. 5b). Slightly less mature ALP was found in the absence of Ent5p in vti1-2 cells after both a 5and a 30-min chase period (percentage of mature ALP after 30 min: vti1-2, 75.5%, S.D. 1.5; vti1-2ent3⌬, 73.9%, S.D. 0.1; vti1-2ent5⌬, 60%, S.D. 1.0, n ϭ 2). These data implicate Ent3p in traffic from the Golgi to the prevacuole. vti1-2 cells grow very slowly at 37°C. Both vti1-2ent3⌬ and vti1-2ent5⌬ cells were unable to grow at 37°C. The synthetic growth defect in vti1-2ent3⌬ cells is probably due to reduced trafficking through the prevacuole, which was detected as synthetic defect in CPY maturation (data not shown). The small synthetic defects in ALP maturation may be the reason for the synthetic growth defect in vti1-2ent5⌬ cells. Alternatively, the growth defect may be due to the action of Ent5p in an unidentified transport step. DISCUSSION Only two proteins binding to ENTH domains have been identified previously. Epsin1 is a cytosolic protein required for endocytosis. An additional pool of epsin1 is present in the nucleus in a complex with the transcription factor PLZF (28). Tubulin interacts with all ENTH and ANTH domains tested (epsin1, epsinR, AP180, HIP1, and Hip1R) (29). Here, we described the specific interaction of the N terminus of the mammalian SNARE vti1b with the ENTH domain of epsinR and of yeast Vti1p with Ent3p. These data as well as the following lines of evidence indicate that Ent3p may be the yeast equivalent of epsinR, even though both proteins bind different phosphoinositides (see below). The ENTH domain of epsinR is most related to that of Ent3p in BLAST searches of yeast protein sequences. In addition, only Ent3p shares sequence homologies with epsinR that extend beyond the ENTH domain. Both proteins bind TGN adaptor proteins of the GGA family and AP-1 and colocalize with clathrin-coated vesicles in the Golgi area, and they have therefore been implicated in clathrin coatedvesicle formation at the TGN or endosomes (20 -23, 25). Synthetic defects in TGN-to-prevacuole traffic in vti1-1 cells indicate that VTI1 and ENT3 function together in this trafficking step. These data are in agreement with the interactions between Gga2p and Ent3p and the requirement for GGA proteins in traffic from the TGN to the prevacuole (30,31). vti1-2ent5⌬ cells displayed a synthetic growth defect at 37°C. Temperature-sensitive growth defects have been observed in the absence of proteins involved in different post Golgi transport steps. Growth at 37°C, for example, requires Pep12p, a syntaxin involved in fusion with the prevacuole (32), Vps33p, a sec1-related protein required for fusion with the vacuole (33), and the GARP or VFT complex, tether proteins in retrograde traffic from the prevacuole to the TGN (34). The synthetic growth defect in vti1-2ent5⌬ cells was not due to a synthetic defect in transport from the TGN to the prevacuolar endosome, because CPY transport did not show a synthetic defect. The growth defect may be due to a small synthetic defect in ALP transport to the vacuole. However, ALP transport to the vacuole does not require GGA proteins (31) and is only marginally delayed in ent3⌬ent5⌬ cells (25). ent3⌬ent5⌬ cells secrete ␣-factor precursor (25). This processing of ␣-factor precursor requires Kex2p in the TGN, pointing toward a loss of Kex2p from the TGN in ent3⌬ent5⌬ cells. As Kex2p cycles between TGN and endosomes, retrograde traffic to the TGN may be defective in ent3⌬ent5⌬ cells. Therefore it is possible that Ent5p is important for retrograde transport to the TGN, and this may be the reason for the synthetic growth defect in vti1-2ent5⌬ cells. Different synthetic transport defects in vti1 mutant backgrounds indicate that Ent3p and Ent5p have slightly different functions.
A role of Ent3p in traffic between TGN and the prevacuole is also supported by the observed binding to phosphatidylinositol 3-phosphate and phosphatidylinositol 3,5-biphosphate, which are implicated in TGN-to-endosome traffic and in the formation of multivesicular bodies, respectively (26). By contrast, epsinR interacted most strongly with phosphatidylinositol 4-phosphate (21)(22)(23). This different specificity in phosphoinositide FIG. 4. Role of Ent proteins in CPY transport to the vacuole. Transport of CPY to the vacuole was normal in ent4⌬ and ent5⌬ cells. ent3⌬, ent3⌬ent4⌬, and especially ent3⌬ent5⌬ cells secreted slightly elevated levels of CPY. CPY was immunoprecipitated from cellular extracts (I) and the medium (E) after pulse-chase labeling at 37°C. WT, wild type; p1CPY, endoplasmic reticulum proCPY; p2CPY, Golgi proCPY; mCPY, vacuolar mature CPY.
FIG . 5. Genetic interactions between VTI1 and ENT3 or ENT5. a, absence of Ent3p resulted in strong defects of CPY transport to the vacuole in vti1-1 cells. CPY was immunoprecipitated from cellular extracts (I) and the medium (E) after pulse-chase labeling at 33°C. p1CPY, endoplasmic reticulum proCPY; p2CPY, Golgi proCPY; mCPY, vacuolar mature CPY. b, transport of ALP to the vacuole was slightly slowed in vti1-2 cells in the absence of Ent5p but not Ent3p after a 5-or 30-min chase period at 30°C. pALP, proALP; mALP, vacuolar mature ALP. c, the absence of Ent3p or Ent5p led to a synthetic growth defect in vti1-2 cells at 37°C. binding has been predicted by modeling the tertiary structure of the ENTH domains of epsinR and Ent3p (24). As both phosphatidylinositol 3-phosphate and phosphatidylinositol 4-phosphate are present in the Golgi complex (35), either phosphoinositide may help to recruit epsinR or Ent3p to the same organelle.
GGAs function in TGN-to-endosome traffic in mammalian cells (36). It is unclear which mammalian SNAREs are involved in this pathway. The binding to epsinR indicates that vti1b may be required for traffic from the TGN to endosomes. To date, vti1b has been implicated only in late endosome fusion (7).
What are the functional implications of an interaction between ENTH proteins and SNAREs? One possibility is that Ent3p and epsinR function in the sorting and recruitment of Vti1p and vti1b, respectively, into budding vesicles. Interactions between adaptor proteins and cargo are often of low affinity and have to be very dynamic because the complexes have to dissociate after vesicle formation. Therefore, binding between adaptor protein complexes and tails of membrane receptors are rarely detectable by co-immunoprecipitations and have been studied almost exclusively using in vitro binding and yeast two-hybrid assays (37). In line with these observations, we have not been able to co-immunoprecipitate Ent3p with Vti1p or epsinR with vti1b (data not shown). A similar adaptorcargo interaction may explain why the loss of UNC-11, a homolog of the ANTH protein AP180, results in mislocalization of the R-SNARE synaptobrevin but not of other synaptic vesicle proteins to the plasma membrane in Caenorhabditis elegans (38). UNC-11 may directly recruit synaptobrevin into clathrincoated vesicles during endocytosis. It has been shown that SNAREs bind other components of the budding machinery like COPII coat proteins in endoplasmic reticulum-to-Golgi transport (39) and adaptor complexes that couple cargo selection to clathrin recruitment. The mammalian R-SNARE VAMP-4 interacts with AP-1 at the TGN (40). In yeast, the vacuolar syntaxin Vam3p binds AP-3 in transport from the TGN to the vacuole (41). A complex of mammalian AP-3 and synaptobrevin mediates formation of synaptic vesicles from endosomes (42), indicating that incorporation of a SNARE can regulate vesicle budding. In ENTH proteins, only the ubiquitin interaction motif has been implicated in cargo binding (43,44). This UIM motif is located outside of the ENTH domain and is absent from epsinR and Ent3p. Our data suggest that cargo sorting may be a novel function of the ENTH domain.