Multiple Roles for Rsp5p-dependent Ubiquitination at the Internalization Step of Endocytosis* integral plasma membrane proteins

Ubiquitination of triggers their rapid internalization into the endocytic pathway. The yeast ubiquitin ligase Rsp5p, a homologue of mammalian Nedd4 and Itch, is required for the ubiquitination and subsequent internalization of multiple plasma membrane proteins, including the a -factor receptor (Ste2p). Here we demonstrate that Rsp5p plays multiple roles at the internalization step of endocytosis. Temperature-sensitive rsp5 mutant cells were defective in the internalization of a -factor by a Ste2p-ubiquitin chimera, a receptor that does not require post-transla-tional ubiquitination. Similarly, a modified version of Ste2p bearing a NPF X D linear peptide sequence as its only internalization signal was not internalized in rsp5 cells. Internalization of these variant receptors was dependent on the catalytic cysteine residue of Rsp5p and on ubiquitin-conjugating enzymes that bind Rsp5p. Thus, a Rsp5p-dependent ubiquitination event is required for internalization mediated by ubiquitin-dependent and -independent endocytosis signals. Constitutive Ste2p-ubiquitin internalization and fluid-phase endocytosis also required active ubiquitination machinery, including Rsp5p. These observations indicate that Rsp5p-dependent ubiquitination of a trans-acting protein component of the endocytosis machinery is required for the internalization step of endocytosis.

Ubiquitination of integral plasma membrane proteins triggers their rapid internalization into the endocytic pathway. The yeast ubiquitin ligase Rsp5p, a homologue of mammalian Nedd4 and Itch, is required for the ubiquitination and subsequent internalization of multiple plasma membrane proteins, including the ␣-factor receptor (Ste2p). Here we demonstrate that Rsp5p plays multiple roles at the internalization step of endocytosis. Temperature-sensitive rsp5 mutant cells were defective in the internalization of ␣-factor by a Ste2p-ubiquitin chimera, a receptor that does not require post-translational ubiquitination. Similarly, a modified version of Ste2p bearing a NPFXD linear peptide sequence as its only internalization signal was not internalized in rsp5 cells. Internalization of these variant receptors was dependent on the catalytic cysteine residue of Rsp5p and on ubiquitin-conjugating enzymes that bind Rsp5p. Thus, a Rsp5p-dependent ubiquitination event is required for internalization mediated by ubiquitindependent and -independent endocytosis signals. Constitutive Ste2p-ubiquitin internalization and fluidphase endocytosis also required active ubiquitination machinery, including Rsp5p. These observations indicate that Rsp5p-dependent ubiquitination of a transacting protein component of the endocytosis machinery is required for the internalization step of endocytosis.
Ubiquitin is a highly conserved 76-amino acid polypeptide that becomes covalently linked to substrate proteins by an isopeptide bond. Two characterized functions of protein ubiquitination are to target proteins for degradation by the cytosolic 26 S proteasome or to promote the internalization of cell surface proteins into the endocytic pathway. Recognition of cytosolic proteins by the proteasome generally requires modification with a polyubiquitin chain of at least four ubiquitin subunits (1,2). In contrast, modification of plasma membrane proteins with monoubiquitin or Lys 63 -linked di-ubiquitin chains triggers internalization into the endocytic pathway (3)(4)(5)(6)(7)(8), ultimately leading to degradation in the lumen of the lysosome.
Protein ubiquitination is catalyzed by a cascade of three enzymes (reviewed in Refs. 9 and 10). Ubiquitin-activating enzymes (E1s) 1 activate ubiquitin in an ATP-dependent reaction, forming a high-energy thiolester bond with ubiquitin. The activated ubiquitin is then passed to a cysteine residue in a ubiquitin-conjugating enzyme (E2). Normally, E2s function together with ubiquitin ligases (E3s) to catalyze isopeptide bond formation between the carboxyl terminus of ubiquitin and ⑀-amino groups in lysine side chains of specific substrates. E3s bind directly to substrates and comprise the major specificity determinants of the ubiquitination machinery. There are two major classes of E3s. One class carries a RING finger domain, and the other carries a hect (homologous to E6-AP carboxyl terminus) domain. RING finger E3s function as adapter proteins, bringing the substrate to the ubiquitin-charged E2, whereas E3s of the hect domain family participate directly in catalysis by forming a thiolester with ubiquitin during the ubiquitination reaction. Ubiquitin-dependent endocytosis regulates the cell surface activities of diverse plasma membrane proteins (reviewed in Refs. 11 and 12). In yeast, ubiquitin is a widely used endocytosis signal that promotes the internalization of plasma membrane proteins such as peptide pheromone receptors, transporters, and nutrient permeases. In mammalian cells, ubiquitin and the ubiquitination machinery regulate the endocytosis of the growth hormone receptor, epithelial sodium channel, epidermal growth factor receptor, and colony-stimulating factor receptor (13)(14)(15)(16). In addition, many more mammalian signaltransducing receptors are known to undergo ligand-stimulated ubiquitination (reviewed in Ref. 17).
The mammalian hect domain E3s Nedd4 and Itch recognize and ubiquitinate the epithelial sodium channel and Notch, respectively (14,18,20). Rsp5p, a homologue of Nedd4 and Itch, promotes the regulated ubiquitination of plasma membrane proteins in yeast (reviewed in Ref. 21). Rsp5p is an essential protein that is implicated in a number of cellular processes in addition to its role in endocytosis. Rsp5p carries an amino-terminal C2 domain, three WW protein-protein interaction domains, and a carboxyl-terminal hect catalytic domain. Structure-function analyses have indicated that the WW domains of Rsp5p play an important role in its endocytic function (22,23). All three WW domains are required for normal rates of stimulated internalization of the ␣-factor receptor Ste2p and the uracil permease Fur4p, implicating these domains in the recognition of endocytic substrates (22,23). The role of the C2 domain in endocytosis is less clear. Deletion of this domain has no effect on Ste2p internalization, but causes an increase in Gap1p and Fur4p activities at the cell surface and delayed fluid-phase endocytic transport to the vacuole (22,24,25). Thus, the C2 domain may be required for the internalization of a subset of plasma membrane proteins and/or may be involved in sorting within the endocytic pathway.
Observations in both yeast and mammalian cells have suggested that the role of ubiquitin in endocytosis may extend beyond its function as an appended internalization signal. The UbE motif, an endocytosis signal in the growth hormone receptor (GHR), is required for GHR internalization by a ubiquitin/ proteasome-dependent pathway (26,27). Because UbE-dependent internalization does not require ubiquitination sites in the receptor, Govers et al. (26) suggested that the binding of the ubiquitination machinery to the UbE motif, not ubiquitination of the GHR itself, is the critical event for internalization. Association of the ubiquitination machinery with the receptor may be required to recruit endocytic proteins to the GHR, or the ubiquitination machinery may modify a receptor-associated regulatory protein (26). Endocytic proteins are known to be substrates of the ubiquitin system. Eps15, an essential component of the clathrin-based endocytic machinery, undergoes epidermal growth factor (EGF)-stimulated monoubiquitination in fibroblasts (28). Genetic studies in Drosophila have suggested that cell fate specification in the compound eye requires specific deubiquitination of Liquid facets, the Drosophila orthologue of the endocytic protein epsin (29). In yeast, a rsp5 mutant lacking an intact C2 domain could ubiquitinate the Gap1 amino acid permease normally but was defective in permease downregulation (24). This observation suggests that the requirement for Rsp5p in endocytosis cannot be fully explained by its role in modifying specific cargo proteins.
As a model for the function of ubiquitin and ubiquitin ligases in the down-regulation of plasma membrane proteins, we have studied the role of Rsp5p in the internalization of the yeast ␣-factor receptor, Ste2p. Ste2p is a MATa cell-specific G-protein-coupled receptor that initiates an intracellular signal required for yeast mating in response to ␣-factor binding (reviewed in Ref. 30). Ligand binding also induces the sequential hyperphosphorylation and ubiquitination of the cytoplasmic tail of the receptor, modifications that trigger rapid receptor internalization (31,32). After internalization, receptors are delivered to the lysosome-like vacuole, where they are permanently inactivated by degradation (33). Mutation of lysines in the receptor cytoplasmic domain or mutations that impair Rsp5p or the Ubc1p/Ubc4p/Ubc5p E2s disrupt the ubiquitination and internalization of Ste2p (22,31). Monoubiquitination by Rsp5p is sufficient to induce Ste2p internalization, since fusion of a single ubiquitin moiety to the cytoplasmic domain rescues the defective internalization of a receptor lacking posttranslational ubiquitination sites (4). The internalization information in ubiquitin has been mapped to two distinct hydrophobic patches on the surface of the ubiquitin molecule (34); however, the mechanism by which ubiquitin induces internalization is undefined.
In this study, we used ␣-factor receptor variants that do not require post-translational ubiquitin modification to uncover a novel function of Rsp5p in endocytosis that is distinct from its role in ubiquitination of plasma membrane cargo. We found that the Rsp5p catalytic cysteine, ubiquitin-conjugating enzymes, and free cellular ubiquitin were required for internalization mediated by ubiquitin-dependent and -independent endocytosis signals. Our results indicate that Rsp5p has two distinct roles at the internalization step of endocytosis: the regulated ubiquitination of cargo proteins and the ubiquitination of a component of the constitutive endocytic machinery.
This study demonstrates a requirement for trans-acting ubiquitination in endocytosis and identifies Rsp5p and Ubc1p/ Ubc4p/Ubc5p as the enzymatic machinery required for this event.

EXPERIMENTAL PROCEDURES
Strains, Media, and Reagents-The genotypes of strains used in this study are listed in Table I. Cells were propagated in synthetic minimal (SD) medium (35), or rich (YPUAD) medium (2% bactopeptone, 1% yeast extract, 2% glucose, supplemented with 20 mg/ml adenine, methionine, and tryptophan). The purification of 35 S-labeled ␣-factor and Ste2p antiserum was performed as described previously (31,36,37). Hemagglutinin (HA 12CA5) monoclonal antibodies were provided by Robert Lamb (Northwestern University, Evanston, IL). The rsp5-2 mutation was provided by Jentsch and colleagues (38). The rsp5-1 allele was isolated by Winston and colleagues (Harvard Medical School, Boston, MA) and was characterized by Wang et al. (39).
Plasmid Construction-All mutations in the STE2 sequence were constructed by site-directed mutagenesis using the two-step polymerase chain reaction (PCR) procedure (40). To generate the ste2-NPFXD mutant, the mutation G392N was introduced into a mutant ste2 (ste2-7xR) with Lys to Arg mutations at amino acids 337, 352, 358, 374, 387, 400, and 422 (4). To generate ste2-All Lys, the F394A mutation was introduced into pJR3 (41) by PCR mutagenesis. Sequences encoding Ste2p-NPFXD and Ste2p-All Lys were subsequently subcloned into the centromeric plasmid YCplac33 (42) resulting in LHP426 and LHP427, respectively. Construction of ste2-ubi (LHP361) and ste2-378Stop (LHP147) has been described previously (4). The ste2-ubi construct used in this study contains a single ubi K48R ubiquitin fused after amino acid 376 of the receptor. ste2-ubi and ste2-378Stop were introduced at the ura3 locus of yeast strains by single-step gene transplacement. PCRderived sequences were verified by automated and/or manual DNA sequencing. All ste2 variants were able to complement the mating defect of ste2⌬ cells. RSP5 plasmids were based on the yeast-Escherichia coli shuttle vector pRS414 (43). Construction of the HA-tagged RSP5 (LHP478) and rsp5-C777A (LHP510) plasmids has been described (22).
␣-Factor Receptor and Lucifer Yellow Endocytosis Assays-␣-Factor internalization assays were performed as described (22,37). Specific variations in growth conditions and assays are indicated in the figure legends. Cells were grown to early logarithmic phase, harvested by centrifugation, and suspended in YPUAD medium at 5 ϫ 10 8 cells/ml. For continuous presence assays, cells were shifted to the nonpermissive temperature for 15 min prior to the addition of 35 S-labeled ␣-factor. For pulse-chase internalization assays, 35 S-labeled ␣-factor was bound to cells on ice for 45-60 min. Unbound radioactivity was removed by centrifugation at 4°C, and internalization was initiated by the addition of media warmed to 30°C or 37°C as indicated in the figure legends. Percentage of internalization is expressed as the ratio of internalized to total cell-associated radioactivity. For unknown reasons, ubc1⌬ ubc4⌬ cells expressed a low level of ␣-factor binding sites. Loss of Ubc1p and Ubc4p has also been reported to cause a decrease in maltose permease expression (44). For this reason, internalization assays depicted in Fig.  5A were corrected by subtracting nonspecific cell-associated radioactivity that bound to ste2⌬ cells under the same conditions. This correction did not significantly alter the results of the assay but allowed a more accurate measurement of initial internalization rates.
Receptor clearance assays were performed as described (32) with the following changes: Cells were shifted to 37°C at the same time as the addition of cycloheximide to 10 g/ml. The 0-min time point was taken 30 min after cycloheximide treatment was initiated to allow Ste2p en route through the biosynthetic pathway to reach the cell surface. The number of ␣-factor receptor binding sites present on the cell surface at each time point was determined by binding with a mix of 35 S-labeled ␣-factor and ϳ3 ϫ 10 8 M unlabeled ␣-factor to ensure that ␣-factor binding sites were saturated in each sample.
Lucifer Yellow (LY) endocytosis assays were performed as described previously (22,37). As indicated in the figure legends, cells were shifted to 37°C for 15 min prior to the addition of 10 l of 40 mg/ml Lucifer Yellow CH (Fluka Chemika-Biochemika, Switzerland), or LY was immediately added at 30°C. After 60 min of incubation, the assay was stopped by the addition of 1 ml of ice-cold phosphate/stop buffer (50 mM sodium phosphate, pH 7.5, 10 mM sodium azide, 10 mM sodium fluoride). Cells were viewed using a Zeiss LSM410 confocal microscope equipped with fluorescein isothiocyanate filter sets.
Cell Lysates and Immunoblots-Lysates for Ste2p immunoblotting were prepared as described previously (31) with minor modifications.
Cells were grown in SD medium to early logarithmic phase, harvested by centrifugation, and transferred to YPUAD medium. Cells were incubated 12.5 min at 37°C before treatment with 1 ϫ 10 Ϫ6 M ␣-factor for 8 min. Cells were not treated with cycloheximide. Lysis was achieved by mechanical agitation with glass beads in urea/SDS buffer containing 5 mM N-ethylmaleimide to inhibit isopeptidase activities in the extract. Lysates were supplemented with ␤-mercaptoethanol (2%) and bromphenol blue (0.002%) and incubated at 37°C for 10 -15 min prior to loading. Extracts used for detecting HA-Rsp5p in Fig. 4A were prepared by an alkaline lysis protocol as described (45). Immunoblotting was performed as described previously (22).

Rsp5p Is Required for Internalization of Receptors That Do
Not Require Post-translational Ubiquitination-Ste2p-Ub is a chimeric, functional ␣-factor receptor that lacks post-translational ubiquitination sites but carries a monoubiquitin moiety fused in-frame to the carboxyl-terminal cytoplasmic tail (Fig.  1). We have shown previously that the fused ubiquitin moiety is sufficient to signal internalization (4,34). Because the internalization of Ste2p-Ub does not require post-translational ubiquitin conjugation, we hypothesized that Ste2p-Ub internalization would not depend on Rsp5p. To test this idea, we compared the internalization of Ste2p-Ub with a receptor that carries the same tail sequence but lacks a fused ubiquitin and requires post-translational ubiquitination for internalization (Ste2p-378Stop, see Fig. 1). The ability of these receptors to internalize ␣-factor was examined in the rsp5-2 temperature-sensitive mutant and in wild-type cells (Fig. 2). Both Ste2p-Ub and Ste2p-378Stop were internalized rapidly in wild-type cells. Consistent with previous observations, rsp5-2 cells were unable to internalize Ste2p-378Stop. The internalization of Ste2p-Ub in rsp5-2 cells was also defective, although it occurred more rapidly than the internalization of Ste2p-378Stop. These data suggest that fusion of ubiquitin to the receptor partially relieves the requirement for Rsp5p in receptor internalization, consistent with its role in appending ubiquitin to activated receptors, but that Rsp5p is also required downstream of receptor modification.
Two distinct classes of internalization signals function in yeast. In addition to ubiquitin, a linear aromatic residue-based sequence, NPFXD, that is present in the cytosolic domain of the a-factor receptor, Ste3p, can act as an efficient internalization signal (46). A weak version of the NPFXD signal, GPFAD, is present in the cytoplasmic tail of wild-type Ste2p. 2 To analyze receptor endocytosis that is mediated solely by the NPFXD signal, we generated a mutant receptor with all tail ubiquitination sites mutated to Arg and with the mutation G392N to convert the GPFAD sequence to a strong NPFXD signal (Ste2p-NPFXD, Fig. 1). Ste2p-NPFXD is internalized with kinetics similar to wild-type Ste2p (see Fig. 3B). 2 To confirm that Ste2p-   (47), and activated receptors accumulate as phosphorylated and ubiquitinated forms at the plasma membrane (31). Ste2p-NPFXD was efficiently hyperphosphorylated, but did not undergo detectable levels of ubiquitination, whereas wild-type Ste2p was ubiquitinated in the presence of ligand (Fig. 3A). As a control for internalization assays, we constructed a Ste2p variant that bears only ubiquitin-dependent internalization signals by mutating Phe-394 to Ala to inactivate the endogenous GPFAD signal (Ste2p-All Lys, see Fig. 1). When we measured ␣-factor internalization in rsp5-2 and wild-type strains expressing these receptors, we found that Ste2p-NPFXD and Ste2p-All Lys were equally affected by the rsp5 mutation (Fig.  3B). These observations indicate that Rsp5p is required for internalization mediated by both known classes of yeast internalization signals, even when the ubiquitination of endocytic cargo is not necessary. Thus, Rsp5p is required for a novel function in endocytosis downstream of cargo modification.

Ubiquitination of a Non-receptor Substrate Is Required for Rapid Receptor Internalization-The internalization defect of
Ste2p-Ub and Ste2p-NPFXD in rsp5 cells could be explained in two different ways. First, Rsp5p could function to ubiquitinate a trans-acting protein that functions in endocytosis. Alternatively, the amino-terminal domains of Rsp5p could participate in a non-catalytic function that is essential for endocytosis. To distinguish between these possibilities, we investigated whether catalytically inactive Rsp5p carrying a mutation in the hect domain could promote internalization of Ste2p-Ub and Ste2p-NPFXD. The C777A mutation abolishes thiolester formation with ubiquitin (48), and this mutant cannot serve as the sole source of Rsp5p in the cell (24,39). We constructed an (HA) epitope-tagged version of Rsp5p that is fully functional (22) and a similarly tagged mutant version with the C777A mutation.
The C777A mutation did not affect Rsp5p expression (39, Fig. 4A). To compare the function of wild-type Rsp5p and Rsp5p C777A in endocytosis, we expressed Rsp5p and Rsp5p C777A in rsp5-1 cells and assayed ␣-factor internalization after inactivating the endogenous Rsp5-1p by incubation at 37°C. We used the rsp5-1 allele for this set of experiments because expression of Rsp5p C777A in rsp5-2 cells caused a dramatic growth defect even at the normally permissive temperature of 24°C. 3 The endocytosis defect of rsp5-1 cells was only partially rescued by plasmid-borne wild-type Rsp5p (compare Fig. 4 (B-D) with Figs. 2 and 3B), because rsp5-1 has a semidominant effect on ␣-factor internalization, perhaps due to the formation of mixed Rsp5-1p/Rsp5p multimers (22). Rsp5p C777A did not rescue the internalization of ␣-factor by Ste2p-All Lys, as expected since Ste2p-All Lys requires ubiquitination to be internalized (Fig. 4B). Rsp5p C777A was also unable to rescue Ste2p-Ub and Ste2p-NPFXD internalization in rsp5-1 cells (Fig. 4, C and D), indicating that the catalytic function of Rsp5p is essential for internalization of receptors that do not require post-translational ubiquitin modification.
To confirm that a Rsp5p-dependent ubiquitination event was required for Ste2p-Ub and Ste2p-NPFXD internalization, we performed two additional experiments. First, we tested the role of E2 enzymes in the internalization of Ste2p-Ub. The E2 enzymes Ubc1p, Ubc4p, and Ubc5p are homologous and form an essential gene family (49). Wild-type Ste2p internalization is impaired in ubc1⌬ ubc4⌬ and ubc4⌬ ubc5⌬ mutants (31). To test whether the internalization of Ste2p-Ub depends on the function of these enzymes, we generated ubc1⌬ ubc4⌬ and congenic UBC1 UBC4 strains expressing Ste2p-Ub and Ste2p-378Stop. The internalization of both Ste2p-Ub and Ste2p-378Stop was slower in ubc1⌬ ubc4⌬ than in wild-type cells; 3 R. Dunn and L. Hicke, unpublished results.

FIG. 2. Rsp5p is required for the internalization of a Ste2pubiquitin chimera. Plasmids encoding Ste2p-Ub and Ste2p-378Stop
were introduced into rsp5-2 ste2⌬ and RSP5 ste2⌬ cells. Continuous presence ␣-factor internalization assays were performed at 37°C. Cells were grown in YPUAD medium at 24°C and were pre-incubated at 37°C for 15 min prior to the addition of radiolabeled ␣-factor.

FIG. 3. Rsp5p is required for internalization mediated by Lysand NPFXD-dependent signals.
A, ␣-factor stimulated modification of Ste2p-NPFXD and wild-type Ste2p. end4 -1 STE2-NPFXD (LHY1955) and end4 -1 STE2 (LHY1562) cells were grown in SD medium at 24°C. Cells were transferred to YPUAD medium, incubated at 37°C for 12.5 min, and treated (ϩ) or not (Ϫ) with 10 Ϫ6 M ␣-factor for 8 min. Cell lysates were resolved by SDS-polyacrylamide gel electrophoresis, transferred to nitrocellulose, and probed with anti-Ste2p antibodies. Labeled brackets indicate the migration of hyperphosphorylated and ubiquitinated species. B, ␣-factor internalization assays performed by the continuous presence protocol at 37°C. Cells were grown in SD medium at 24°C and assayed after a 15-min pre-incubation at 37°C. Panel shows rsp5-2 Ste2p-NPFXD (LHY757, Ⅺ), RSP5 Ste2p-NPFXD (LHY758, f), rsp5-2 Ste2p-All Lys (LHY759, E), RSP5 Ste2p-All Lys (LHY760, q). Curves represent the average of at least three independent assays, and error bars indicate the standard deviation at each time point. however, the internalization of Ste2p-Ub was slightly more rapid than that of the receptor requiring post-translational ubiquitination (Fig. 5A), similar to the internalization of Ste2p-Ub in rsp5 cells. Ubiquitin-conjugating enzymes were also required for Ste2p-NPFXD internalization. 3 Second, we tested whether the internalization of Ste2p-Ub was dependent on the levels of free ubiquitin in the cell. To do this we assayed Ste2p-Ub and Ste2p-378Stop internalization in a doa4⌬ mutant. DOA4 encodes a deubiquitinating enzyme, and the pool of free ubiquitin is severely reduced in doa4⌬ cells because conjugated ubiquitin is not recycled efficiently (50). Many of the phenotypes of this mutant can be reversed by the overproduction of ubiquitin from a multicopy plasmid (50). The rate of Ste2p-Ub internalization was decreased in doa4⌬ cells, although it was somewhat faster than the internalization of Ste2p-378Stop (Fig. 5B). Internalization of both Ste2p-378Stop and Ste2p-Ub in doa4⌬ cells was restored by the overexpression of ubiquitin. 4 These data indicate that Ste2p-Ub internalization depends on ubiquitin-conjugating enzymes and normal levels of free ubiquitin in the cell.
Ubiquitination Is Required for Constitutive Endocytosis-Components of the endocytic machinery can be modified in a signal-dependent manner. One example of this is the EGFinduced tyrosine phosphorylation of clathrin heavy chain, an event that influences clathrin localization and EGF receptor endocytosis (51). It has also been suggested that Ca 2ϩ -triggered dephosphorylation of endocytic proteins may represent a general mechanism to stimulate the assembly of endocytic coats after nerve terminal depolarization (52). Ste2p is constitutively ubiquitinated at a low level, and receptor ubiquitination can mediate the slow constitutive internalization that occurs in the absence of ligand (32). If the novel function of Plasmids encoding HA-Rsp5p and HA-Rsp5p C777A were introduced into rsp5-1 cells expressing Ste2p-Ub, Ste2p-NPFXD, and Ste2p-All Lys. A, expression of HA-Rsp5p and HA-Rsp5p C777A . Cells were grown in SD medium at 24°C. Cell extracts were prepared by alkaline lysis and trichloroacetic acid precipitation. Cellular proteins were resolved by SDS-polyacrylamide gel electrophoresis, transferred to nitrocellulose, and probed with anti-HA antibodies. Panel shows no HA tag (LHY1122), HA-Rsp5p (LHY1634), and HA-Rsp5p C777A (LHY1635). B-D, ␣-factor internalization assays performed as described in Fig. 3B.
Curves represent the average of at least three independent experiments, and error bars represent the standard deviation at each time point. pared with wild-type cells (Fig. 6). Therefore, Rsp5p-dependent ubiquitination of a trans-acting endocytic protein is not exclusively coupled to receptor activation.
Because the involvement of Rsp5p seemed to be a constitutive, signal-independent requirement, we investigated the role of ubiquitination in fluid phase endocytosis. LY is a soluble fluorescent molecule that is internalized by fluid-phase endocytosis and delivered to the vacuole (37). Mutants that block the internalization step of endocytosis cannot localize LY to the vacuole (37). Previous studies showed that Rsp5p and a subset of its amino-terminal domains are required for LY localization to the vacuole (22,23,53). We tested whether the role of Rsp5p in fluid phase endocytosis involves protein ubiquitination by analyzing fluid-phase endocytosis in mutants deficient in ubiquitin-conjugating enzymes that cooperate with Rsp5p. We performed LY endocytosis assays in ubc1⌬ ubc4⌬ and ubc4⌬ ubc5⌬ cells and compared these mutants with wild-type cells and cells deficient in only ubc1⌬, which transport LY normally. ubc1⌬ ubc4⌬ cells showed a decrease in vacuolar fluorescence intensity, and ubc4⌬ ubc5⌬ cells showed an even stronger defect compared with ubc1⌬ cells (Fig. 7). Consistent with previous studies, rsp5-1 cells were also unable to internalize LY efficiently (Fig. 7). These data support the conclusion that Rsp5pdependent ubiquitin ligation to a trans-acting protein is required for efficient constitutive internalization from the plasma membrane.

DISCUSSION
In this study, we demonstrate that the Rsp5 ubiquitin ligase is required to regulate an unknown component of the endocytic machinery by ubiquitination. Previously, Rsp5p has been shown to modify endocytic cargo with a three-dimensional ubiquitin internalization signal (reviewed in Ref. 21). A second function for Rsp5p-dependent ubiquitination in endocytosis was revealed in our experiments using receptors that do not require cis-acting, post-translational modification by ubiquitin for internalization. We found that the internalization of a chimeric ␣-factor receptor carrying a functional ubiquitin internalization signal was dependent on Rsp5p catalytic activity, ubiquitin-conjugating enzymes, and on ubiquitin itself. These data argue strongly that receptor internalization requires Rsp5p-mediated ubiquitination of an unidentified protein. Rsp5pdependent ubiquitination was also required for internalization mediated by a Ste2p variant bearing the linear peptide internalization signal NPFXD, for fluid-phase endocytosis, and for constitutive internalization of the Ste2p-ubiquitin chimera.
Therefore, the novel Rsp5p-dependent function is not exclusively coupled to ligand-stimulation or to internalization mediated by the ubiquitin internalization signal, but instead is a constitutive requirement for the internalization step of endocytosis. We propose that Rsp5p regulates one or more constitutive components of the endocytosis machinery by ubiquitination.
An alternative explanation for our observations is that Rsp5p is required for the internalization of Ste2p-Ub and Ste2p-NPFXD because a threshold level of cargo ubiquitination by Rsp5p is essential for the productive formation of primary endocytic vesicles. However, we consider this explanation unlikely for the following reasons. First, we found that the expression of an abundant, internalized plasma membrane protein carrying the ubiquitin signal, Pma1p-Ub (34), did not influence the internalization of Ste2p-Ub in rsp5 mutant cells. Even though Pma1p-Ub was expressed at a high level, so that it represented ϳ10 -20% of the total plasma membrane protein in yeast (54), rsp5 cells were still unable to internalize Ste2p-Ub. 3 Second, the cytosolic domains of Slg1p, Sro4/Bud10p, and many other predicted plasma membrane proteins contain a perfect consensus NPFXD sequence. Since multiple plasma membrane proteins probably use the NPFXD internalization signal, a significant level of cargo competent for internalization may be available even in the absence of ubiquitin-modified cargo.
Recently, Jentsch and colleagues (38) described an essential role for Rsp5p in a novel activation pathway required for the synthesis of unsaturated fatty acids. The lethality of a rsp5⌬ mutation can be suppressed by overexpression of OLE1, a gene required for synthesis of oleic acid, or by exogenous addition of oleic acid to the growth medium. Oleic acid does not rescue the internalization of wild-type Ste2p in rsp5 cells. 3 This result was expected because Rsp5p is required to ubiquitinate Ste2p prior to internalization. We considered the possibility that the defective internalization of the Ste2p-Ub and receptors carrying the NPFXD signal could be due to a general oleic acid requirement for efficient endocytosis. However, we found that the exogenous addition of oleic acid to rsp5 mutant cells carrying Ste2p-Ub Ubiquitin-conjugating enzymes are required for Lucifer Yellow endocytosis. LY localization assays were performed on wild-type (LHY886), rsp5-1 (LHY883), ubc1⌬ (LHY180), ubc1⌬ ubc4⌬ (LHY1394), and ubc4⌬ ubc5⌬ (LHY196) cells. Wild-type and rsp5-1 cells were grown to early logarithmic phase in YPUAD medium at 24°C, shifted to 37°C for 15 min, then incubated with LY for 1 h. ubc1⌬, ubc1⌬ ubc4⌬, and ubc4⌬ ubc5⌬ cells were grown to early logarithmic phase in YPUAD medium at 24°C and then incubated with LY for 1 h at 30°C.
could not restore ␣-factor internalization, although the growth defect of rsp5 cells was substantially rescued by oleic acid. 3 Thus, our results cannot be explained by a deficiency in unsaturated fatty acids in rsp5 mutant cells.
Several observations made by other investigators support the conclusion that the role of ubiquitin in endocytosis is not limited to tagging cargo proteins. Genetic evidence has suggested that cell fate specification in the Drosophila eye requires modulation of endocytosis through specific deubiquitination of the endocytic protein epsin (29). In mammalian cells, Eps15, an epsin-binding protein required for clathrin-mediated endocytosis, undergoes EGF-induced monoubiquitination (28). Finally, internalization of a mutant growth hormone receptor lacking ubiquitination sites is blocked by specific inhibitors of the proteasome or by temperature inactivation of a thermolabile E1 activity, suggesting the receptor itself does not require ubiquitination (26,27).
What endocytic proteins might be regulated by Rsp5p? Yeast homologues of the EH domain protein Eps15 and the epsins are candidates because there is evidence for ubiquitin modification of their counterparts in different organisms. Furthermore, mutations in RSP5 and the Eps15 homologue PAN1 interact genetically (53). It has been proposed that Rsp5p and Pan1p interact via the binding of Rsp5p WW domains to proline-rich motifs in the carboxyl terminus of Pan1p (55). Rsp5p also contains a NPF motif, the ligand for EH domains found in Pan1p (56,57). However, we have been unable to detect a physical interaction between Rsp5p and Pan1p, and we have not detected ubiquitinated forms of Pan1p or the yeast epsins Ent1p and Ent2p. In addition, disruption of the NPF sequence in Rsp5p has no affect on ␣-factor internalization. 3 Chang et al. (58) showed that Rsp5p WW domains are type I WW domains that bind preferentially to PPXY-containing ligands. The yeast amphiphysin homologue, Rvs167p, and Arc15p, a component of the Arp2/3p actin complex that is required for endocytosis, contain PPXY motifs that may interact with Rsp5p WW domains. Thus, Arc15p and Rvs167p are candidate endocytic proteins that may be regulated by Rsp5p-mediated ubiquitination. Since some WW domains have affinity for phosphoserine and phosphothreonine (59), the WW domains of Rsp5p may interact with and promote ubiquitination of endocytic phosphoproteins. Because interactions that require post-translational modification cannot be easily predicted by sequence gazing, the identification of Rsp5p-regulated endocytic proteins may require a combination of unbiased genetic and biochemical approaches.
A rsp5 mutant lacking its C2 domain is competent for nitrogen-regulated ubiquitination of the Gap1 permease but not for its subsequent degradation (24). This observation prompted the proposition that the C2 domain is required for interaction with or ubiquitination of an endocytic protein downstream of permease ubiquitination (21,60). We have observed that a mutant lacking the Rsp5p C2 domain exhibits wild-type internalization of ␣-factor by Ste2p-Ub or by Ste2p-NPFXD. 4 indicating that the C2 domain is not involved in the novel function of Rsp5p described in this study. Furthermore, our results indicate that the C2 domain is dispensable for internalization mediated by either of the defined endocytosis signals in yeast. We have, however, observed a defect in the transport of a fluid phase marker to the vacuole in a rsp5-⌬C2 mutant (22), consistent with a role for the C2 domain in a post-internalization step of endocytosis. Our observations do not rule out the possibility that the C2 domain is required specifically for permease internalization.
What is the molecular mechanism by which Rsp5p regulates the constitutive endocytic machinery? The internalization of Ste2p-378Stop and Ste2p-Ub was not significantly affected by mutations in the PRE1 and PRE2 genes encoding catalytic subunits of the proteasome. 4 Therefore, the novel function of Rsp5p-dependent ubiquitination is not likely to involve proteasomal degradation of a substrate protein. We propose that Rsp5p modulates the activity of the endocytic machinery by reversible modification with ubiquitin in a manner independent of the proteasome. There are a growing number of proteins whose modification by ubiquitin has been shown to regulate the activity of the protein without degradation. The core nuclear histones H2A and H2B are stably modified by ubiquitin in vivo (61), and loss of ubiquitination of H2B causes a meiosis defect in yeast (62). Monoubiquitination of the human protein FANCD2, a protein involved in ionizing radiation-induced DNA repair, causes its recruitment to nuclear foci containing other DNA repair factors including BRCA1 (63). Lys-63-linked ubiquitin chains modify the stable ribosomal protein L23 and activate the IB kinase TRAF6 (64,65). Finally, ubiquitination negatively regulates the transcription factor Met4p, but ubiquitin-dependent inactivation of Met4p is clearly not accompanied by its proteasome-mediated degradation (19). We suggest that reversible ubiquitination of endocytic proteins acts to control protein activity. Modification with ubiquitin may alter protein activity by inducing a conformational change, changing the protein's localization, or altering the affinity of the substrate for its protein or lipid binding partners. Molecular characterization of this function awaits the identification of the Rsp5p endocytic substrate(s).