Ubiquitination of the PEST-like Endocytosis Signal of the Yeast a-Factor Receptor*

A 58-residue-long, PEST-like sequence within the yeast a-factor receptor (Ste3p) specifies the ubiquitination, endocytosis, and consequent vacuolar degradation of the receptor protein (Roth, A. F., Sullivan, D. M., and Davis, N. G. (1998) J. Cell Biol. 142, 949–961). The present work investigates three lysyl residues that map within this sequence as the potential ubiquitin acceptor sites. Lys 3 Arg substitution mutants were tested for effects on both ubiquitination and endocytosis. Results indicate that the three lysines function redundantly; a severe blockade to both ubiquitination and endocytosis is seen only for receptors having all three lysines replaced. Of the three, Lys 432 plays the predominant role; ubiquitination and turnover are significantly impaired for receptors having just the K432R mutation. CNBr fragmentation of the receptor protein, used for the physical mapping of the ubiquitin attachment sites, showed PEST-like sequence lysines to be modified both with single ubiquitin moieties as well with short multi-ubiq-uitin chains, two or three ubiquitins long. Thus, in addition to being the signal for ubiquitination, the Ste3p PEST-like sequence also provides the site for ubiquitin attachment. To test if this endocytosis signal functions solely for ubiquitination, we have asked if the requirement for the PEST-like sequence in endocytosis might be bypassed through pre-attachment of ubiquitin to the receptor protein. Indeed, Ste3-ubiquitin Ubiquitination— turnover mechanisms or to digestion by added proteases was measured using quantitative AMBIS densitometry (AMBIS Systems, Inc., San Diego, CA) of films developed from ECL chemiluminescent Western blots (Amersham Pharmacia Biotech). Estimation of the fractional loss of Ste3 antigen entailed comparing remaining Ste3 antigen (following turnover or protease shaving) to a standard curve generated by dilution of the initial sample ( e.g. receptor present at the 0-h time point of a turnover experiment or present in the control sample not subjected to protease shaving).

The mechanisms that select plasma membrane substrates for clathrin-mediated endocytosis in mammalian cells are reasonably well understood (3,4). Substrates subject to constitu-tive uptake display short peptidyl endocytosis signals centered around either an aromatic residue, usually a tyrosine, or a Leu-Leu dipeptide (5). These signals are decoded through the binding of the AP-2 adaptin complex which then serves to coordinate the assembly of the clathrin coat protein complex, thereby initiating the invagination of a patch of the plasma membrane into the cytoplasm. A variation on this theme has recently been described for the agonist-dependent uptake of the ␤ 2 -adrenergic receptor (6). In this case, the adaptin role in clathrin-coated pit assembly is replaced by the binding of ␤-arrestin to the phosphorylated cytoplasmic tail domain of the liganded receptor. ␤-Arrestin also binds strongly to clathrin, thus allowing the assembly of the receptor into the clathrincoated pit.
Endocytosis in the yeast Saccharomyces cerevisiae differs in several regards from the classic clathrin-dependent endocytosis of mammalian cells. While yeast express obvious homologues of the key players for clathrin-mediated uptake, i.e. clathrin heavy and light chains and the adaptin subunits, a clear demonstration of the requirement of these proteins in yeast uptake has remained elusive. Exhaustive deletion of the genes encoding the adaptin subunits is without discernible effect on yeast uptake processes (7). For clathrin, while mutations that disable clathrin function show partial endocytosis defects, the interpretation of this partial defect remains uncertain (8,9). Clathrin may either function redundantly as one of several coat proteins functioning in yeast uptake, or alternatively, as these clathrin mutants also severely impair viability, the effect on endocytosis could be indirect.
In yeast, it is now clear for a variety of endocytic substrates that the selectivity of uptake depends at least in part on selective ubiquitination (10,11). Plasma membrane proteins destined for uptake are marked with added ubiquitin. Nonetheless, as with the clathrin-mediated uptake, the selectivity of the process ultimately depends upon the recognition of determinants or signals encoded within the endocytic substrate's amino acid sequence. Sequences involved in directing uptake have been identified for several yeast endocytic substrates (1,(12)(13)(14)(15). Given the differences between clathrin-and ubiquitindependent uptake processes, it is not surprising that these yeast sequences bear little resemblance to the signals directing the clathrin-mediated endocytosis of mammalian cells. The two yeast pheromone receptors, the ␣-factor receptor (Ste2p) and the a-factor receptor (Ste3p), both have been well studied in terms of endocytosis. These two G protein-coupled receptors mediate the hormonal cross-talk that precedes the sexual conjugation of the MATa cell with the MAT␣ cell. The two receptors are subject both to a ligand-dependent internalization mechanism as well as to a ligand-independent (i.e. constitutive) uptake mechanism. For Ste2p, the sequence SINNDAKSS, which maps to the regulatory C-terminal cytoplasmic tail do-main (CTD), 1 has been shown to be required for ␣-factor-induced uptake (12,16). Mutations in this sequence block both ubiquitination and endocytosis. Indeed, the Lys of this sequence has been suggested to be the major site for ubiquitination as its replacement by Arg blocks both the ubiquitination and internalization of the receptor (16). The sequence NPF-STD, located within the Ste3p CTD, has been shown to be required for its a-factor-induced endocytosis (13). While required for uptake, these sequences within Ste2p and Ste3p clearly do not constitute sufficient signals for endocytosis. Additional receptor domains also are required. For instance, liganddependent uptake obviously also depends upon the ability of the receptor to bind ligand. Indeed, for Ste2p, in addition to the CTD sequence SINNDAKSS, mutations that map to predicted extracellular domains of the receptor also are found to impair ligand-dependent uptake (17).
One well characterized yeast endocytosis signal is the signal that directs the rapid constitutive, ligand-independent internalization of the a-factor receptor (1). For Ste3p, constitutive endocytosis is the more prominent of the two uptake modes. In the absence of a-factor ligand, Ste3p is rapidly internalized from the cell surface and delivered to the vacuole for degradation (18). Ste3p consequently is a short-lived protein with a t 1/2 of only 15 min in cells growing at 30°C. This rapid endocytosis depends upon a 58 residue-long sequence mapping to the C terminus of the receptor CTD (1). Receptor mutants deleted for this sequence fail to undergo both constitutive ubiquitination and endocytosis and instead stably accumulate at the cell surface. Furthermore, this sequence also is a sufficient signal; when transplanted to the cytoplasmically disposed C terminus of the plasma membrane ATPase Pma1p, normally a stable cell surface resident, the Ste3p sequence directs both ubiquitination and rapid internalization of the ATPase to the vacuole for degradation. With its preponderance of both the acidic residues Asp and Glu and hydroxylated residues Ser and Thr, the Ste3p signal most closely resembles a class of ubiquitination signals known as PEST sequences (19). Similar sequences rich in acidic and hydroxylated residues have also been implicated in the endocytosis of both the a-factor export protein Ste6p and the uracil permease Fur4p (14,15).
The present work extends the analysis of the Ste3p PESTlike endocytosis signal, particularly with regards to its role in ubiquitination. First, we ask if, in addition to specifying ubiquitination, the PEST-like sequence also provides the locus for ubiquitin attachment. Lys 3 Arg replacement mutagenesis of the three lysine residues located within the PEST-like sequence indicates that the three likely function redundantly as the sites for ubiquitin attachment: receptors having these three lysines replaced, fail to be ubiquitinated and fail to be internalized. Direct mapping of the ubiquitination site confirms the PEST-like signal as the primary locus for ubiquitin attachment. In addition, this analysis identifies the presence of short multi-ubiquitin chains attached therein. To test if these multiubiquitin chains might be required for directing uptake, we have applied the ubiquitin fusion methodology of Terrell et al. (2). Results with Ste3-ubiquitin fusions confirm results made previously with the Ste2-ubiquitin fusions (2), namely that recognition of mono-ubiquitin, and not the multi-ubiquitin chain, is key for endocytosis. In addition, the conclusions of Terrell et al. (2) are also extended with the demonstration that ubiquitin addition to a plasma membrane protein is not only necessary for initiating uptake, but also is sufficient; ubiquitin functions alone to specify uptake, not in conjunction with other receptor sequences or signals.

EXPERIMENTAL PROCEDURES
Plasmids-The plasmid pSL1904 serves as the starting point for many of STE3 mutations constructed in this work. pSL1904 is GAL1-STE3 carried on the URA3 integrating plasmid pRS306 (20). The presence of chromosomal flanking sequences upstream and downstream of the GAL1-STE3, 1.1 and 2.3 kilobase pairs long, respectively, allow both wild-type and mutant GAL1-STE3 constructs to be introduced back into the yeast chromosome by homologous recombination in place of ste3⌬::LEU2 alleles. pSL1904 is identical to pSL1839 (1), except in having a 760-base pair GAL1,10 promoter fragment substituting for the STE3 promoter (21), replacing the sequences 417 to 110 base pairs upstream of the STE3 ORF.
The plasmid pND751 is identical to the CUP1-driven, myc-ubiquitin 2/URA3 expression plasmid pND747 (1), except that the Met 15 codon of the myc-ubiquitin fusion (corresponding to ubiquitin Met 1 ) has been replaced by a Leu codon. The M15L mutation removes a CNBr cleavage site that would otherwise result in the loss of the identifying myc epitope extension from the tagged ubiquitin. The M15L mutation was introduced by oligonucleotide mutagenesis (Table I) of single-stranded DNA derived from pND153, a BamHI to KpnI subclone of the 0.6kilobase pair CUP1-myc-Ubi portion of YEp105 (22) into pRS306 (20). Finally, pND751 was derived by swapping the BamHI-KpnI fragment carrying the M15L mutation back into pND747.
Construction of Ste3-Ub Fusions-The ubiquitin sequences were joined to STE3 sequences at one of two XhoI sites previously introduced into STE3, one at codons 398/399 and the other at codons 466/467 (1). The Ste3(⌬399)-Ub class of fusions lacks the 72 C-terminal residues of Ste3p (residues 399 -470), including the PEST-like endocytosis signal (residues 413-470). The Ste3(wt)-Ub fusions lack just the C-terminal five amino acid residues. Interpolated between the STE3 and ubiquitin sequences in all fusions is an artificial, 15-residue-long, glycine-rich linker sequence described by Robinson and Sauer (23) and utilized by Terrell et al. (2) in the construction of the Ste2-Ub fusions. The linker, which encodes the peptide GGGSGGGTGGGSGGG, was constructed through hybridization of two complementary 52-mer oligonucleotides: TCGAGGGAGGTGGAAGTGGAGGCGGAACAGGCGGAGGGTCAGG-AGGGGGGCA and GATCTGCCCCCCTCCTGACCCTCCGCCTGTTC-CGCCTCCACTTCCACCTCCC. The resulting double-stranded DNA has an XhoI sticky end for fusion with upstream STE3 sequences and a BglII sticky end for ligation to the BglII site located at the second and third codons of the ubiquitin gene from YEp96 (22). Fusions to the 7K3 R ubiquitin used the BglII site of pTER62 (Mike Ellison, University of Alberta, Edmonton, Alberta, Canada), a plasmid identical to YEp96, except with AGA arginine codons in place of all seven lysine codons. The Ste3-Ub fusions were ligated into the GAL1-STE3 context of pSL1904 where the upstream and downstream chromosomal flanking sequences allow for the homologous integration of the GAL1-STE3-Ub constructs at the STE3 locus. The pSL1904-derived plasmids for GAL1-STE3(⌬399)-Ub(wt), GAL1-STE3(⌬399)-Ub(7K3 R), GAL1-STE3(wt)-Ub(7K3 R), and GAL1-STE3(3K3 R)-Ub(7K3 R) are pND774, pND885, pND942, and pND901, respectively. In addition, two C-terminal ubiquitin mutations, ⌬75,76 and G76A were introduced into the GAL1-STE3(⌬399)-Ub(7K3 R) context via oligonucleotide mutagenesis (Table I).
Site-directed Mutagenesis-A number of site-directed mutations (24) were introduced both into STE3 and into various ubiquitin constructs (Table I). In addition to the introduced coding change, mutations also were designed to either add or remove restriction enzyme digestion sites, simplifying the identification of the mutant plasmids by restriction analysis (Table I). Fidelity of the mutations was checked by DNA sequencing.
Strains-Strains were constructed for each of the different STE3 mutations and Ste3-Ub fusions characterized in this work, wherein the mutant construct is integrated at the chromosomal STE3 locus. Constructs were integrated in place of chromosomal ste3⌬::LEU2 allele via a two-step gene replacement strategy (25). Plasmid constructs linearized at a unique upstream HpaI site (for the Lys 3 Arg and methionyl substitution mutants) or at a unique upstream SfiI site (for the Ste3-Ub fusions) were introduced by transformation into one of three strains, either the wild-type MAT␣ strain NDY343 (ste3⌬::LEU2 ura3 leu2 his4), or the isogenic end4 -1 or pep4⌬ versions, NDY344 and NDY372, respectively (1). The resulting Ura ϩ integrants have the introduced mutant allele in tandem array with the ste3⌬::LEU2 allele. As a final step, 5-fluoro-orotic acid counterselection was used as described previously (25) to isolate Leu Ϫ , mating-competent segregants that had lost via homologous recombination, the ste3⌬::LEU2 together with the integrated pRS306 plasmid sequences.
Assessment of Turnover and Ubiquitination-Ste3p and Ste3-Ub turnover was assessed via a non-radioactive pulse-chase protocol (18). A 2-h "pulse" of receptor synthesis was induced with the addition of 2% galactose to exponential cultures of cells carrying the GAL1-driven constructs growing in YP-raffinose (2%) medium. Following this, the "chase" period was instigated with the addition of 3% glucose. The time-dependent loss of Ste3 antigen was followed by Western blotting of protein extracts from culture aliquots removed at various times during the glucose chase. For the first time point (taken immediately subsequent to the addition of glucose), culture volumes corresponding to 5 ϫ 10 7 cells were collected by centrifugation and protein extracts were prepared for SDS-PAGE and Western blotting as described previously (18). For subsequent time points, the same culture volume collected for the initial time point was again harvested. Although this translates into increased number of cells being collected at later time points, the amount of Ste3 antigen present in this aliquot should remain the same for conditions in which Ste3p is not turning over and should diminish appropriately under conditions in which Ste3p turns over. The affinitypurified rabbit polyclonal antibodies used for detecting Ste3p were prepared as described (1).
The analysis of Ste3-Ub protein turnover in some instances utilized a protocol in which cells were treated with energy poisons prior to the preparation of extracts. The conditions and the buffers utilized are identical to those used in the protease-shaving protocol, except that no protease is added.
Visualization of the ubiquitinated forms of the receptors was enhanced with treatment of the protein extracts with potato acid phosphatase (Roche Molecular Biochemicals) prior to analysis by SDS-PAGE and Western blotting (25).
Protease Digestion of Intact Cells-The treatment of whole cells with Pronase (Calbiochem-Novabiochem Corp., La Jolla, CA) and the subsequent extract preparation was as described previously (18). To assess the maintenance of cell integrity during the course of the protease treatments, extracts were routinely monitored for the resistance of the cytoplasmic protein, phosphoglycerate kinase (Pgk1p) to digestion.
For the a-factor treatment of the cell cultures used for following ligand-dependent endocytosis, a 50% volume of a cell-free filtrate from a saturated culture of EG123 cells (26) transformed with the the a-factor overproduction plasmid pKK16 (27) was added. Mock pheromone treatments involved the addition of an equivalent filtrate from the isogenic mfa1::LEU2 mfa2::URA3 strain (18).
CNBr Fragmentation-Protein extracts for CNBr digestion differed in two ways from those utilized for Western analysis (18); extracts were prepared from 1.5 ϫ 10 8 cells instead of 5 ϫ 10 7 , and urea was omitted from the SDS-PAGE sample buffer (40 mM Tris/Cl, pH 6.8, 5% SDS, 0.1 mM EDTA, 1% ␤-mercaptoethanol) used in the glass bead disruption of the cells. Twenty microliters of protein extract in a 200-l reaction volume was incubated for 1 h in the dark at 25°C in 2.5 M CNBr, 70% formic acid. One milliliter of 13% trichloroacetic acid was added, and, following 1 h at 0°C, the precipitated protein was pelleted with a 15-min microcentrifuge spin at 4°C. The pellet was washed once with acetone, desiccated, dissolved at 65°C for 5 min with 20 l of 40 mM Tris/Cl, pH 6.8, 8 M urea, 0.5% SDS, 0.1 mM EDTA, 1% ␤-mercaptoethanol. Fifteen microliters of the CNBr-fragmented extract was then subjected to digestion with potato acid phosphatase in a 1-ml reaction volume, essentially as described previously (25). Following the phosphatase treatment, protein was trichloroacetic acid-precipitated and redissolved in sample buffer as described (25).
Quantitative Methods-The loss of Ste3 antigen either to endogenous turnover mechanisms or to digestion by added proteases was measured using quantitative AMBIS densitometry (AMBIS Systems, Inc., San Diego, CA) of films developed from ECL chemiluminescent Western blots (Amersham Pharmacia Biotech). Estimation of the fractional loss of Ste3 antigen entailed comparing remaining Ste3 antigen (following turnover or protease shaving) to a standard curve generated by dilution of the initial sample (e.g. receptor present at the 0-h time point of a turnover experiment or present in the control sample not subjected to protease shaving).

RESULTS
Lysine Residues within the PEST-like Ubiquitination/Endocytosis Signal-Twenty-four lysine residues map to the predicted cytoplasmic portions of the a-factor receptor. Three of these, Lys 424 , Lys 432 , and Lys 453 , fall within the C-terminal, 58-residue PEST-like ubiquitination/endocytosis signal (Fig.  1). We have first focused on these three lysines as potential acceptor sites for the ubiquitination associated with constitutive endocytosis. Mutant receptors were constructed with conservative arginine substitutions at one, two, or all three lysines. Mutants were compared with wild-type Ste3p for effects on both turnover and ubiquitination.
Mutant receptors were expressed from the GAL1 promoter, and turnover was assessed by following the rate of receptor loss subsequent to glucose-mediated repression of new receptor synthesis. Turnover rates measured in this way are roughly equivalent to those measured via in vivo labeling, pulse-chase analysis of cells expressing Ste3p from its natural promoter (1,18,25). Analysis of the single lysyl replacements showed no detectable effect on turnover of either the K424R or K453R mutations ( Fig. 2A). For K432R, however, turnover was clearly impaired, but not fully blocked. For receptors with two of the three lysines replaced (Fig. 2B), the K424R,K453R receptor displayed wild-type turnover kinetics. The other two double mutants, both of which included the K432R mutation, showed partial impairments similar to that observed for the single K432R mutant. The most complete blockade to turnover was seen for the triple 3K3 R mutant (K424R,K432R,K453R), its turnover defect being substantially greater than any single or double mutants. Thus, while the K424R and K453R mutations alone or together showed essentially no perturbation of Ste3p turnover, it is evident from the 3K3 R receptor results that these two residues clearly are capable of making at least a redundant contribution; complete blockade of Ste3p turnover requires that all three of the lysines be substituted (Fig. 2B).
To assess receptor ubiquitination, mutant receptors were expressed in a pep4⌬ cell context. As Ste3p turnover is blocked in pep4⌬ cells, ubiquitinated receptor forms are not lost to Removes Removes HaeII a Each of the mutations either adds or removes a restriction enzyme digestion site, which was used diagnostically to identify plasmids with the mutations. vacuolar turnover. In addition, protein extracts were treated with phosphatase to remove heterogeneity in gel mobility due to phosphorylation of the receptor, thereby improving the Western blot visualization of minor ubiquitinated species (25). For the wild-type receptor, we find approximately 20% of isolated receptor is modified by ubiquitin, distributing equally between mono-and di-ubiquitinated forms (25). For the single Lys 3 Arg mutants (Fig. 2C), only K432R noticeably reduces ubiquitination levels. Likewise for the double mutants ( Fig.  2D), only the two mutants that involve substitution of Lys 432 , i.e. K424R,K432R and K432R,K453R, reduce ubiquitination. The most severe reduction in receptor ubiquitination is seen for the 3K3 R receptor (Fig. 2, C and D). As previously seen with deletions into the PEST-like sequence (1), receptor ubiquitination levels correlate well with turnover ( Fig. 2, A and B).
While the 3K3 R receptor shows the most complete ubiquitination impairment, a faint band is apparent at the monoubiquitinated receptor position, suggesting that the 3K3 R receptor may remain subject to residual ubiquitination. Subsequent analyses confirm this low level mono-ubiquitination (Fig. 4).
The requirement for the PEST-like sequence lysines in receptor ubiquitination and turnover suggests that they serve as redundant acceptor sites for ubiquitin attachment. If so, a corollary conclusion would be that the 21 lysyl residues that map to other cytoplasmic portions of the receptor protein do not normally serve as acceptor sites. In addition, the significant impairments to ubiquitination and turnover caused solely by the K432R mutation, suggest a prominent role for this lysine; Lys 432 may function as the primary ubiquitination site with Lys 424 and Lys 453 serving as alternative, secondary attachment sites.
The 3K3 R Receptor Is Defective for the Internalization Step of Endocytosis-Evidence for a variety of yeast plasma membrane proteins indicates that ubiquitination acts at the cell surface internalization step of endocytosis (11). We expect, therefore, that the 3K3 R receptor that lacks the presumptive ubiquitination sites is blocked for turnover because of failed internalization. If so, this receptor should accumulate at the plasma membrane. To assess receptor localization, we have used a protease-shaving protocol, wherein intact yeast cells are treated with proteases (18). Surface-localized receptor is degraded by the added proteases, while receptor residing intracellularly is unavailable for digestion. We have compared 3K3 R receptor localization both to wild-type Ste3p and to the ⌬413 truncation mutant, which is deleted for the PEST-like endocytosis/ubiquitination signal.
Ninety minutes after synthesis, wild-type Ste3p was found to be wholly resistant to the added, extracellular proteases ( Fig.  3A), indicative of an intracellular localization and consistent with the vacuole as the known end point for Ste3p constitutive endocytosis (18). In contrast, Ste3⌬413p remains available to protease digestion, indicating a surface localization. The 3K3 R mutant behaves essentially like Ste3⌬413p, remaining largely susceptible to the added proteases (Ͼ80% digested), indicating that the 3K3 R accumulates at the plasma membrane, presumably a result of failed uptake. Mutants were compared with wild-type Ste3p both for receptor turnover and ubiquitination. A, turnover of the three single Lys 3 Arg mutants. NDY343-derived MAT␣ strains were subjected to a 2-h period of galactose-induced receptor expression. Glucose was added to block further receptor synthesis, and, at the indicated times thereafter, culture aliquots were removed, extracts prepared, and the loss of Ste3 antigen was followed via SDS-PAGE and then Western blotting with Ste3p-specific antibodies. B, turnover of the three double Lys 3 Arg mutants. This experiment followed the protocol described for panel A. C, ubiquitination of the three single Lys 3 Arg mutants. Receptor expression was induced from pep4⌬ versions (isogenic to NDY372) of the five strains utilized in panel A with a 2-h period of growth in galactose medium. Protein extracts prepared from these cells were treated with potato acid phosphatase and then subjected to SDS-PAGE and Western blotting with Ste3p-specific antibodies. D, ubiquitination of the three double Lys 3 Arg mutants. pep4⌬ versions of the five strains utilized in panel B were cultured and treated as described above (panel C). Arrows at the left of panels C and D indicate the positions of the mono-and di-ubiquitinated forms of the receptor. The arrows at the right of panels C and D indicate the position of a protein that crossreacts with the anti-Ste3p antibodies.
FIG. 3. The 3K3 R receptor is defective for the cell surface uptake step of endocytosis. Receptor was synthesized from MAT␣ pep4⌬ strains having GAL1-driven alleles of either wild-type STE3, STE3(3K3 R), or STE3(⌬413) with a 2-h galactose expression period, which was followed by a 90-min glucose chase period. Culture aliquots were collected, and the intact cells were treated by incubation with (ϩ) or without (Ϫ) proteases in the presence of energy poisons (10 mM sodium azide, 10 mM potassium fluoride) as described (see "Experimental Procedures"). Extracts prepared from these cells were subjected to SDS-PAGE and Western blotting with Ste3p-specific antibodies. The position of the protease-protected digestion products from both the 3K3 R and ⌬413 receptors is indicated at right.
In addition to showing the different localizations, the wildtype and the 3K3 R receptors also manifest strikingly different electrophoretic presentations (Fig. 3); while the wild-type receptor appears as a heterogeneous cluster of bands, the 3K3 R receptor migrates largely as a single, slightly lower molecular weight species (compare minus protease samples in Fig. 3). This difference in receptor modification status is explored more fully below (Fig. 6).
Determining the Site of Ubiquitination-For analysis of the Ste3p ubiquitination sites, we have made use of the K424R,K453R (2K3 R) receptor. This mutant, which retains Lys 432 , shows wild-type ubiquitination and turnover (Fig. 2D). If, as predicted by our mutational data, the PEST-like sequence lysines do serve as the primary ubiquitination sites, then 2K3 R receptor ubiquitination should be largely limited to Lys 432 ; this provides a simplification for our analyses.
The 185-residue-long Ste3p CTD contains only two methionines, Met 288 and Met 304 , which map just to the C-terminal side of the seventh transmembrane domain (Fig. 5A). Cleavage of Ste3p with CNBr cleavage should release a 166-residue-long fragment extending from residue 305 to the C terminus at residue 470. As our Ste3p antibody was raised against the CTD, only this 305-470 fragment should be detected. If the sites for ubiquitin attachment are contained within this fragment, then ubiquitinated forms of the fragment also should be detected. For identification of the ubiquitinated receptor fragments, we have compared CNBr digests of the 2K3 R receptor both to the 3K3 R receptor (reduction in ubiquitinated species expected for the 3K3 R receptor), and to 2K3 R receptor isolated from cells overexpressing a myc-tagged ubiquitin (22). The epitope tag adds 14 residues to the N terminus of ubiquitin and Ste3p species modified with tagged ubiquitin rather than wild-type ubiquitin show gel migrations that are correspondingly slowed (25).
2K3 R and 3K3 R receptors were isolated from cells with wild-type ubiquitin expression levels or from cells that overexpress myc-ubiquitin. Intact receptors were analyzed first (Fig.  4A). 2K3 R receptor isolated from the myc-ubiquitin-expressing cells showed modified receptor forms migrating at slightly higher molecular weights than the analogous receptor species modified with wild-type ubiquitin. In addition, receptor ubiquitination was greatly increased in cells expressing the mycubiquitin (the myc-ubiquitin construct is carried on a multicopy 2 plasmid vector with expression being driven from the CUP1 promoter). Mono-, di-, and possibly tri-ubiquitinated species were apparent. 3K3 R receptor ubiquitination remains impaired relative to 2K3 R receptor even when isolated from the myc-ubiquitin overexpressing cells (Fig. 4A). Nonetheless, a species corresponding to mono-ubiquitinated 3K3 R receptor is clearly evident, although di-and tri-ubiquitinated forms are not. Thus, the 3K3 R receptor appears to be subject to a low level of mono-ubiquitination, which increases when ubiquitin is overproduced.
The protein extracts of Fig. 4A were subjected to CNBr cleavage and phosphatase digestion (Fig. 4B). For the 2K3 R receptor extracted from the myc-ubiquitin cells, three prominent species show both increased intensity and retarded gel mobility relative to fragments from the control cell context (Fig.  4B). The three presumptive ubiquitinated species migrate near molecular weights calculated for the 305-470 CTD fragment with one, two, and three attached ubiquitin moieties, i.e. 28, 38, and 48 kDa. We conclude that the 305-470 receptor interval provides a locus for mono-, di-, and tri-ubiquitination. For the di-and tri-ubiquitinated fragments, this analysis does not distinguish if ubiquitins are attached as multi-ubiquitin chains or as mono-ubiquitins to separate lysine residues.
For 3K3 R receptor, only the mono-ubiquitinated 305-470 fragment is apparent (Fig. 4B). As the 3K3 R receptor lacks potential lysyl acceptor sites within the PEST-like sequence, this mono-ubiquitination must be attached to lysyl sites located elsewhere in the CTD.
Our expectation from the Lys 3 Arg mutational analysis was that most of 2K3 R receptor ubiquitination should occur at Lys 432 . To test this, we have used site-directed mutagenesis to introduce methionyl sites for CNBr cleavage at various positions surrounding Lys 432 (Fig. 5A). If Lys 432 is a major site for ubiquitin attachment, then CNBr fragments that retain Lys 432 should retain the ubiquitin modification, while fragments that do not retain Lys 432 , should not retain this modification (Fig.  5A). Methionines were introduced at sites within the PEST-like sequence expected to impact receptor ubiquitination and endocytosis minimally, based on previous mutagenic analysis of this sequence (1). Four receptor mutants were constructed with methionines introduced at residues 413, 421, 441, and 452 of the 2K3 R receptor. The methionyl mutations were found to result in no obvious impairment to receptor endocytic function; turnover rates (data not shown) and ubiquitination levels (Fig.  5B) of the four were found to be roughly equivalent to that of the 2K3 R parent.
Extracts from cells expressing the 2K3 R receptor or the four mutant derivatives were subjected to CNBr digestion and then phosphatase treatment. In preliminary experiments, CNBrgenerated CTD fragments from the five receptors showed different reactivities with the polyclonal anti-CTD antibody, presumably a reflection of the loss of epitopes as the CTD fragment is shortened from the C-terminal end through cleavage at the introduced methionines (Fig. 5A). To compensate for this, samples were normalized to give comparable detection of the major, non-ubiquitinated CTD fragment (see Fig. 5C legend for details).
In Fig. 5C, we see that cleavage of the parental 2K3 R receptor gives the major 305-470 fragment and the three higher molecular weight bands corresponding to this fragment with one, two, or three attached ubiquitins. Introduction of methionines into the 2K3 R receptor results in predictable changes to the pattern of the CNBr peptide fragments. First, the major, non-ubiquitinated fragments (Fig. 5C, white dots) from the different methionyl mutants migrate predictably to positions consistent with the C-terminal shortening of the 305-470 fragment. (The one exception to this is the somewhat anomalous migration of the 305-421 fragment.) In addition, the corresponding ubiquitinated species also show the same downward gel mobility shift. This second effect is clear for receptors with cleavage sites at residues 441 and 452; the three ubiquitinated species show the predicted increases in gel mobility consistent with the C-terminal shortening of the 305-470 fragment (Fig. 5C, ubiquitinated species designated as 1, 2, and  3). Thus, the multi-ubiquitination seen for the 305-470 fragment is retained for both the 305-441 and 305-452 fragments. This contrasts sharply with results for receptors with methionines introduced at residues 413 and at 421. The 305-413 and 305-421 fragments clearly lack the the di-and tri-ubiquitinated species (the faint high molecular weight bands present in the Met 413 and Met 421 samples result from partial CNBr cleavage; described below). These fragments lack Lys 432 and as they are clearly under-ubiquitinated, we conclude that Lys 432 is indeed a major site for ubiquitin attachment. Furthermore, as the di-and tri-ubiquitinated species are lost simultaneously, we can conclude that Lys 432 also provides a locus for the attachment of short multi-ubiquitin chains, two or three ubiquitin moieties long. For each of the methionine mutants, the identity of the ubiquitinated CTD fragments was confirmed as with the 2K3 R receptor (Fig. 4B) through CNBr cleavage of extracts from cells overproducing the myc-tagged ubiquitin (data not shown).
While the 305-413 and 305-421 fragments lack the di-and tri-ubiquitinated species, they do show evidence for some mono-ubiquitination. The amount of mono-ubiquitin present on these two species likely is overestimated from Fig. 5C. The partial 289 -470 cleavage product co-migrates at this position. Detectable levels of this partial fragment as well as the 305-470 partial fragment are present for each of four methionine mutants (Fig. 5C). Furthermore, to achieve equivalent antibody reactivity (see above), increased amounts of the Met 413 FIG. 5. Lys 432 is a major site for ubiquitination. A, experimental design. Four mutant receptors genes were constructed each having a different methionyl substitution mutation (see Table I Fig. 4B. As the length of the CTD fragment is progressively truncated with CNBr digestion at the introduced methionines, reactivity with the polyclonal Ste3p antibody, raised against the entire Ste3p CTD peptide, is progressively lost. To compensate for the reduced reactivity of the shortened CTD fragments, increased amounts of sample were loaded onto the SDS-polyacrylamide for the methionine mutants so as to normalize antibody reaction with the primary (unmodified) CTD cleavage product of each (indicated by white dot). For Met 441 and Met 453 , 2-fold more sample relative to 2K3 R was analyzed. For Met 413 and Met 421 , 4-fold more sample was utilized. The presumptive mono-, di-, and tri-ubiquitinated forms of the primary cleavage products derived from each of the methionine mutants are indicated (1, 2, and 3, respectively). In addition at right, the positions of the 305-470 cleavage product as well as of the 289 -470 partial cleavage product are indicated. Finally, the pre-existing CNBr-independent Ste3p fragment described for Fig. 4B is shortened with CNBr digestion at the introduced methionines; the position of these fragments are indicated as well (white squares). and Met 421 samples were analyzed relative to the 2K3 R parental sample. The 289 -470 partial fragment that co-migrates with the mono-ubiquitinated 305-413 and 305-421 fragments is consequently over-represented in these samples, leading to an overestimate of ubiquitination. CNBr analysis of the receptors extracted from cells expressing the myc-tagged ubiquitin (data not shown) supports this view; attachment of the tagged ubiquitin to the 305-412 and 305-421 fragments retards the gel mobility, displacing it away from the 289 -470 partial fragment. With this separation, it is clear that the level of monoubiquitinated 305-412 and 305-421 species present is significantly less than for the ubiquitinated forms of the CTD fragments, which retain Lys 432 , indicating that mono-ubiquitination at lysyl sites mapping outside of the PEST-like sequence is minor. This conclusion is consistent with results for the 3K3 R receptor, where the presence of a faint mono-ubiquitinated receptor band (Fig. 2) indicated that a low level of monoubiquitination was occurring to receptor sequences outside the PEST-like sequence.
In addition to the 305-470 partial product being present for all of the methionyl mutants, so too are the ubiquitinated forms of this partial fragment. These co-migrate with the identical species generated through complete digestion of the parental 2K3 R receptor (Fig. 5C). The mono-and di-ubiquitinated forms of the 305-470 partial product are faintly apparent in the samples from Met 412 and Met 421 mutants, a reflection again of the increased amount of sample used for these two mutants (see above). Fig. 3, we noted that the wild-type and 3K3 R receptors were modified to different extents. For the protease-shaving protocol (utilized for Fig. 3), the protease treatment of intact cells takes place in the presence of energy poisons, added to block the possibility of further membrane traffic. In Fig. 6, we have compared the modification status of wild-type Ste3p and Ste3(3K3 R)p from extracts prepared from cells treated with or without energy poisons. When isolated from unpoisoned cells (lanes 2 of Fig. 6), wild-type Ste3p and Ste3(3K3 R)p gave similar presentations, both migrated as a heterogeneous cluster of bands. For both also, phosphatase treatment collapses the heterogeneous clusters to a single, faster migrating species (lanes 3 of Fig. 6), indicating phosphorylation is the responsible modification. Treatment of cells with energy poisons prior to extract preparation results in some loss of the phosphoryl modification for both receptors. For the 3K3 R receptor, this loss is more pronounced; it now co-migrates with phosphatase-treated receptor (Fig. 6, lane 1). In contrast, some portion of the wild-type receptor is able to resist the loss of phosphoryl modification during the energy poisoning incubations (Fig. 6, lane 1). The ⌬413 receptor behaves like 3K3 R; in the absence of poisons, the truncated receptor shows a heterogeneous phosphorylation, which is fully lost when receptor is isolated from poisoned cells (Fig. 6, lane 1). These different outcomes of the energy poisoning regimen appears to correlate with the different receptor localizations; the 3K3 R and ⌬413 receptors both localize to the plasma membrane (Fig. 3), while wild-type Ste3p localizes to the vacuole (18). Consistent with localization being the primary determinant, wild-type Ste3p trapped at the plasma membrane in end4 cells gives a similar energy poisoning result to the 3K3 R receptor, i.e. complete loss of phosphoryl modification (data not shown). Energy poisons also affect receptor ubiquitination. Ste3p ubiquitination is easily observed for receptor accumulating either in the vacuole of the pep4⌬ cells or at the cell surface in end4 cells (25). However, treatment of such cells with energy poisons results in a more complete loss of ubiquitin modification from the surface-localized receptor than from the vacuole-localized receptor; for the vacuole-localized receptor, some of the ubiquitinated receptor appears to be protected from the de-ubiquitinating effects of the energy poisons (data not shown).

Differential Effects of Energy Poisons on the Modification of Surface-and Vacuole-localized Receptor-For
One explanation for these phenomena relies on the differential energy requirements of the processes that add these modifications, versus the processes that remove them. Phosphorylation and ubiquitination both consume ATP, while removal of these modifications by endogenous phosphatases or ubiquitin proteases does not. Under energy poisoning conditions, the balance is expected to be shifted toward removal. When localized to the pep4⌬ vacuole, the receptor may be less available to the cytoplasmic activity of phosphatases and ubiquitin proteases. Indeed, recent findings suggest that endocytosed substrates in yeast may be delivered wholly to the lumen of the vacuole, with the cytoplasmic domains of the endocytosed proteins being fully sequestered to the interior of the vacuole (28,29).
Ubiquitin Can Functionally Substitute for the Ste3p PESTlike Endocytosis Signal-The analysis above demonstrated the presence of short multi-ubiquitin chains attached to Lys 432 of Ste3(2K3 R)p (Fig. 5C). However, previous experiments with Ste2p indicated that single ubiquitin moieties, rather than multi-ubiquitin chains, provide the key recognition elements that specify Ste2p uptake (2). To assess the role of mono-versus multi-ubiquitination in Ste3p endocytosis, we have followed the lead of Terrell et al. (2) and have constructed a fusion protein between Ste3p and ubiquitin. The Ste3-Ub fusion has the 7K3 R ubiquitin (all lysines mutated to arginine) fused to the C terminus of a truncated Ste3p (⌬399), lacking its Cterminal 72 residues, including the PEST-like signal. The attached 7K3 R ubiquitin lacks potential lysyl acceptor sites and thus is incapable of nucleating multi-ubiquitin chain formation. Furthermore, as the Ste3p constitutive endocytosis signal has been wholly deleted from the Ste3-Ub fusion, we can address questions regarding the sufficiency of the ubiquitin role that were not accessible by the original Ste2-Ub fusion studies (2). Does the fused ubiquitin functionally replace the PEST-like sequence for endocytosis? Restated, does the Ste3p PEST-like signal function solely for ubiquitination? Further, does the attached ubiquitin provide a sufficient signal to trigger uptake, or is ubiquitin recognized in conjunction with other peptidyl signals within the endocytic substrate?
Anti-Ste3p Western blots of extracts from cells expressing the Ste3-Ub fusion protein show the presence of a major new band migrating near the 50-kDa molecular mass expected for the Ste3-Ub fusion (Fig. 7A). Unexpectedly, however, in addition to the 50-kDa species, Ste3 antigen was also found to distribute heterogeneously throughout the gel lane, at molec-FIG. 6. Effects of energy poisons on receptor phosphorylation status. MAT␣ pep4⌬ strains (NDY372-derived) having GAL1-driven alleles of either wild-type STE3, STE3(3K3 R), or STE3(⌬413), chromosomally integrated in place of STE3 were cultured as described for Fig. 3. Culture aliquots were collected and protein extracts were either immediately prepared or cells were subjected to a further incubation at 37°C for 90 min in the presence of energy poisons (10 mM sodium azide/10 mM potassium fluoride) prior to protein extract preparation (lanes designated as 1). Extracts prepared from cells not treated with poisons were treated with phosphatase (lanes designated as 3) or were mock-treated (lanes designated as 2) prior to SDS-PAGE and Western blot analysis with Ste3p-specific antibodies. ular masses both below the 50-kDa band and extending well above to the gel origin (Fig. 7A, compare with ste3⌬ lane). This unusual presentation, we believe, results not from modification of the Ste3-Ub fusion protein (for instance, by ubiquitination or phosphorylation), but rather from the ubiquitin-mimetic conjugation of C-terminal portions of this fusion protein to the various cellular substrates for ubiquitination. In a sense, the Ste3-Ub fusion resembles those ubiquitin constructs that have N-terminal epitope extensions. Such tagged ubiquitins efficiently substitute for wild-type ubiquitin in a variety of ubiquitin conjugation reactions (16,22,25,30). The N-terminal extensions can be quite large; a glutathione S-transferase-Ub fusion with ubiquitin fused to the complete 26-kDa glutathione S-transferase enzyme may be efficiently conjugated to substrate (31). Thus, C-terminal fragments of the Ste3-Ub fusion (essentially, ubiquitin with N-terminally attached Ste3p portions) may be substituting for ubiquitin in the modification of other cellular substrates of ubiquitination. The heterogeneous gel smear then would reflect the huge variety of cellular proteins that normally serve as substrates for ubiquitination.
The covalent linkage of ubiquitin to substrate joins the Cterminal Gly 76 of ubiquitin to the ⑀-amino group of a substrate lysyl side chain. Unlike alterations at the ubiquitin N terminus, which generally do not affect conjugation of ubiquitin to substrate (see above discussion), alteration of the Gly 75 and Gly 76 C-terminal residues might be expected to be more disabling. We have constructed two mutations at the Ste3-Ub C terminus: Ste3-Ub(⌬75, 76) having the two C-terminal ubiquitin glycines removed and Ste3-Ub(G76A) for which Gly 76 of ubiquitin is changed to alanine. Both mutations were found to both profoundly reduce the heterogeneous gel mobility associated with the Ste3-Ub fusion and concomitantly, increase the proportion of the Ste3 antigen migrating at the 50-kDa position (Fig. 7A). The effectiveness of these two C-terminal ubiquitin mutations in blocking the heterogeneous re-distribution of Ste3 antigen through the gel lane supports the explanation that this heterogeneity is a consequence of the ubiquitin-mimetic conjugation of Ste3-Ub fragments to cellular ubiquitination substrates.
We have also tested the energy poisoning regimen used for Fig.  6 for effects on the heterogeneous gel distribution of the Ste3-Ub fusion protein (Fig. 7A). This treatment, which can result in the loss of Ste3p modifications, might also reverse the attachment of Ste3-Ub to other cellular proteins. Indeed, we find that this is the case: energy poisons fully reverse the heterogeneous distribution of Ste3 antigen throughout the gel lane (Fig. 7A).
As a test of the functionality of these fusions in endocytosis, we have first examined effects on constitutive turnover. The truncated ⌬399 receptor (equivalent to the Ste3 portion of the Ste3-Ub fusion) is not subject to constitutive turnover (Table II) and stably accumulates at the cell surface (1). In contrast, this FIG. 7. Expression and turnover of Ste3-Ub fusions. Strains having GAL1-driven STE3(⌬399)-Ub alleles replacing STE3 of isogenic wild-type (NDY343-derived), end4 -1 ts (NDY344-derived), and pep4⌬ (NDY372-derived) strains were subjected to the non-radioactive galactose-to-glucose pulse-chase regimen described for Fig. 2A, except that the cells were initially cultured at 25°C and then shifted to 37°C, 15 min prior to the completion of the galactose pulse. Following the 2-h period of galactose-induced Ste3-Ub expression, glucose was added and, at the indicated times, culture aliquots were collected. Protein extracts were either immediately prepared (Ϫ), or when indicated (ϩ), cells received an additional incubation in the presence of energy poisons as described for Fig. 6. Protein extracts were subjected to SDS-PAGE and then Western blotting with Ste3p-specific antibodies. A, effects of energy poisons and of C-terminal ubiquitin mutations on the gel distribution of the Ste3-Ub fusions. Four isogenic strains (NDY343-derived) were constructed with one of the following replacements at the STE3 locus: the ste3⌬::LEU2 allele (designated ⌬), a GAL1-STE3(⌬399)-Ub(7K3 R) construct (designated Ste3-Ub), as well as the identical GAL1-STE3(⌬399)-Ub(7K3 R) construct having, in addition, either deletion of the two C-terminal ubiquitin glycine codons (designated Ste3-Ub(⌬75,76)) or substitution of the C-terminal Gly 76 codon of ubiquitin by alanine (designated Ste3-Ub(G76A)). Protein extracts were prepared from cells either immediately following the 2-h galactose induction period (Ϫ) or following an additional incubation with energy poisons (ϩ) as described for Fig. 6. The position of the major Ste3-Ub protein is indicated at left (arrow). Hash marks at right indicate the positions of protein species that cross-react with the Ste3p-specific antibodies. B, the Ste3(⌬399)-Ub(7K3 R) fusion shows rapid turnover. Turnover kinetics were examined for three of the four strains described in panel A. At the indicated times following glucose addition, protein extracts were prepared either immediately (Ϫ) or following incubation with the energy poisons (ϩ). C, turnover of the Ste3(⌬399)-Ub(7K3 R) fusion protein requires both endocytosis and vacuolar protease activity. Isogenic wild-type, end4, and pep4⌬ strains were constructed to have either GAL1-STE3 (top panel) or GAL1-STE3(⌬399)-Ub(7K3 R) (bottom panel) chromosomally substituting for STE3. At the indicated times following glucose addition, culture aliquots were collected, treated with energy poisons, and protein extracts were prepared and Western-blotted. 20 min a MAT␣ strains isogenic to NDY343, except for having the above GAL1-driven STE3 mutants at the STE3 locus in place of ste3⌬ϻLEU2 were subjected to the non-radioactive pulse-chase described for Fig. 7C. Extracts were prepared from cells harvested at the time of glucose addition (0 min), and 30, 60, and 90 min afterward. The rate of Ste3 antigen loss was assessed via Western blotting with Ste3p-specific antibodies, followed by quantitative densitometry of the resulting films (as described under "Experimental Procedures"). truncation having the 7K3 R ubiquitin fused to its C terminus clearly does turn over (Fig. 7B). Thus, the added ubiquitin restores turnover to the signal-deleted receptor. Furthermore, rapid turnover kinetics are unaffected when the C-terminal G76A mutation is introduced (Fig. 7B). Likewise, the addition of the energy poisoning regimen to the experimental protocol also does not affect the outcome; rapid turnover is still seen (Fig. 7B). Given the lack of effect of the energy poisoning treatment on the experimental result, this regimen has been routinely incorporated into the following experiments that test different Ste3-Ub fusions.
Rapid constitutive turnover of the wild-type a-factor receptor depends on the endocytic delivery of surface receptor to the vacuole for degradation. Ste3p turnover is blocked in mutant cell backgrounds that are defective for either endocytosis (e.g. end4 cells) or for vacuolar protease activity (e.g. pep4⌬ cells) (Fig. 7C). We find that Ste3-Ub turnover shows the same requirements, being blocked in both end4 and pep4 cells (Fig. 7C). In addition, we have used the protease-shaving protocol to examine the localization of the Ste3-Ub fusion in these two cell backgrounds (Fig. 8). In end4 cells, Ste3-Ub accumulates at a locale where it is largely sensitive to digestion by the added proteases, probably the plasma membrane. In pep4⌬ cells, Ste3-Ub accumulates within some intracellular compartment (probably the vacuole) where it resists digestion by the extracellular proteases (Fig. 8).
We conclude that, as with wild-type Ste3p, turnover of the Ste3-Ub fusion protein also depends on its endocytic transport from the cell surface to the vacuole. A further conclusion is that the Ste3p constitutive endocytosis signal may be fully functionally replaced by the added mono-ubiquitin. With ubiquitin preattached to the receptor, the need for the endocytosis signal in endocytosis is bypassed; it has become dispensable, suggesting that this signal functions solely in endocytosis for directing ubiquitination. Furthermore, we can conclude that ubiquitin functions alone to trigger uptake. No other receptor sequence or signal is required in conjunction.
Our results also confirm those of Terrell et al. (2), which indicated that multi-ubiquitin chain formation is not required for endocytosis; rather, mono-ubiquitin provides the key elements recognized by the endocytic apparatus in initiating uptake. While mono-ubiquitination of Ste3p suffices, it has been suggested from studies of the uracil permease Fur4p and of the general amino acid permease Gap1p that short multi-ubiquitin chains having Lys 63 ubiquitin-ubiquitin linkages may play a role in stimulating the rate of uptake (32,33). Indeed, our analysis of Ste3p ubiquitination showed the presence of short multi-ubiquitin chains associated with some of the receptors (Fig. 5C). Furthermore, while turnover of the Ste3-Ub fusion is rapid (t 1/2 of 22 min; Table II), it is somewhat less rapid than the turnover of wild-type Ste3p (t 1/2 ϭ 15 min; Table II). Perhaps this rate difference reflects the potential for multiple ubiquitin attach-ment to wild-type Ste3p, but not to the Ste3-Ub (which lacks lysyl acceptor sites). As a test of this, we have compared the turnover of the Ste3-Ub fusion utilizing the 7K3 R ubiquitin, Ste3(⌬399)-Ub(7K3 R), to a new Ste3-Ub fusion protein retaining all seven ubiquitin lysines, Ste3(⌬399)-Ub(wt). As Ste3(⌬399)-Ub(wt) retains the ubiquitin lysines, it also retains the potential for multi-ubiquitination. We find that the Ste3(⌬399)-Ub(wt) fusion does turnover somewhat faster than the 7K3 R form (Fig.  9), with a 15-min half-life estimated (Table II).
We have also constructed fusions of the 7K3 R ubiquitin attached to full-length receptor: both to the wild-type receptor, which retains its PEST-like signal intact, and to the 3K3 R receptor mutant: Ste3(wt)-Ub(7K3 R) and Ste3(3K3 R)-Ub(7K3 R). In Fig. 10, we have compared the turnover of these two fusions to wild-type Ste3p. Both fusions turn over rapidly. The Ste3(wt)-Ub(7K3 R) fusion is particularly striking in that it turns over faster than wild-type Ste3p (t 1/2 ϭ 9 min; Table II), the only Ste3-Ub fusion to do so. This fusion, in addition to having an attached ubiquitin, also retains the potential for ubiquitination of the PEST-like sequence lysines. The equivalent 3K3 R fusion, i.e. Ste3(3K3 R)-Ub(7K3 R), lacks these PEST-like sequence ubiquitination sites and consequently does not show this augmented turnover rate (t 1/2 ϭ 20 min; Table II). Thus, while mono-ubiquitination suffices to initiate uptake, attachment of multiple ubiquitin moieties, either as a chain or as mono-ubiquitins added to separate lysines, may serve to augment the rate of uptake.
Ligand-dependent Endocytosis of the 3K3 R Receptor-A ligand-dependent endocytosis has been demonstrated for the a-factor receptor only under conditions in which rapid, constitutive endocytosis has been mutationally blocked, either with truncated forms of the receptor that lack the constitutive endocytosis signal (18) or with mutant strain backgrounds specifically defective for constitutive uptake (34). Most of the analysis of the a-factor-dependent uptake mode has made use of the ⌬365 Ste3p truncation mutant (9,13,18,25), which lacks the C-terminal 104 residues of the 184-residue-long CTD. In addition to removing the PEST-like signal for constitutive endocytosis, this deletion may remove additional sequences potentially important to the ligand-dependent uptake process. To test if additional receptor sequences contribute, we have compared the ligand-induced uptake rate of the ⌬365 receptor to that of the more minimally mutated 3K3 R receptor.
For this analysis, we have again used the protease-shaving protocol, monitoring the rate of internalization of the ⌬365 and 3K3 R receptors following challenge of the cells with a-factor. At the outset of the experiment, prior to a-factor addition, both receptors localize mainly to the cell surface, i.e. both are largely susceptible to digestion by the added, extracellular proteases ( Fig. 11A; 0-min time point). With time following a-factor addition, an increasing proportion of both receptors is seen to undergo internalization, evidenced by the increasing fraction of the receptor population found to resist digestion by the extracellular proteases. A quantitative depiction of the Fig. 11A experimental result shows that the two receptors are internalized at roughly equivalent rates. Thus, even while retaining the receptor CTD in its entirety, the 3K3 R receptor does not manifest an enhanced capacity for uptake relative to the ⌬365 receptor. This suggests that with both receptors we are likely measuring the inherent rate for this uptake process. It is noteworthy in this regard that Givan and Sprague (34) report that the kinetics of a-factor-induced uptake of wild-type Ste3p observed in akr1⌬ cells (the akr1 defect is specific for the constitutive mode of Ste3p endocytosis) are again similar to those observed for the ⌬365 receptor.

DISCUSSION
The Lysyl Acceptor Sites for Ubiquitin Addition-The three lysines that mapped within the PEST-like endocytosis/ubiq-uitination signal provided an obvious starting point in our efforts to identify the ubiquitin acceptor sites associated with Ste3p rapid constitutive endocytosis. All combinations of Lys 3 Arg replacements were constructed and tested for effects both on receptor turnover and ubiquitination. Results support a model in which the three lysines function redundantly as sites for ubiquitin conjugation. Receptor mutants having one or two of the three lysines replaced showed, at most, partial defects for ubiquitination and turnover. Complete impairment was seen only for receptors in which all three lysines had been removed. This redundancy implies a certain plasticity in terms of which of the three is used as the acceptor site. This plasticity is limited, however, as it clearly does not extend to the 21 lysine residues that map outside of the PEST-like sequence to other cytoplasmic portions of the receptor protein; these other lysines may not substitute for the loss of the three endocytosis signal lysines in the 3K3 R mutant receptor, indicating that they may not be able to serve as ubiquitin acceptors.
Of the three PEST-like sequence lysines, Lys 432 may normally play the most prominent role; K432R mutant receptors are partially impaired for both ubiquitination and turnover, and K424R,K453R receptors, which retain Lys 432 , show wildtype ubiquitination and turnover. It is likely that Lys 432 normally serves as the primary ubiquitination site with Lys 424 and Lys 453 serving as alternative, secondary attachment sites.
Results from CNBr fragmentation of ubiquitinated Ste3p largely confirm results from Lys 3 Arg mutagenesis, indicating that the PEST-like sequence does indeed provide the site for the much of the ubiquitination that is associated with constitutive endocytosis. In addition, this analysis provides the first direct biochemical evidence for multi-ubiquitin chains associated with endocytosis. Our analysis of Ste3p ubiquitination was simplified by concentrating on the K424R,K453R mutant receptor, which retained just the Lys 432 acceptor site. CNBr fragments that retained Lys 432 , i.e. the 305-441, 305-452, and 305-470 CTD fragments, were found to be modified with 0, 1, 2, or 3 ubiquitin moieties (Fig. 5C). Fragments lacking Lys 432 , i.e. 305-412 and 305-421, showed a commensurate loss of the di-and tri-ubiquitinated species. From this we conclude that Lys 432 is modified with a di-ubiquitinated chain and a subset may also be modified with a tri-ubiquitin chain. The uncertainty regarding the tri-ubiquitin chain relates to the low level mono-ubiquitination of non-PEST-like sequence lysines, apparent for both the 3K3 R receptor (Fig. 2, C and D, and Fig. 4B) and for the K424R,K453R receptors (Fig. 5C). The tri-ubiquitinated CNBr fragments could consist of two moieties linked in series at Lys 432 with the third ubiquitin attached to a separate lysyl residue, elsewhere in the receptor CTD.
Functional Replacement of the Endocytosis Signal by Ubiquitin-Given our finding of a short multi-ubiquitin chain attached to Lys 432 , we were interested to test if such multiubiquitin chains are required for Ste3p endocytosis. Previous experiments for Ste2p had indicated that endocytosis differs from proteosomal targeting in that single ubiquitin moieties, and not multi-ubiquitin chains, provide the targeting information for internalization (2). Repeating these experiments with Ste3p, with the construction of translational fusions between Ste3p and a 7K3 R ubiquitin, confirms the Ste2p findings (2). Like Ste2p, multi-ubiquitination is not required for endocytosis. Mono-ubiquitin provides the necessary targeting information.
In addition to confirming conclusions made for Ste2p (2), our results with the Ste3-Ub fusions also extend these conclusions. The Ste3(⌬399)-Ub fusions lack all Ste3p signals and sequences required for constitutive endocytosis (1). The ability of the attached ubiquitin to functionally replace these sequences indicates the sufficiency of the mono-ubiquitin in specifying FIG. 10. Turnover of Ste3-Ub fusions retaining either the wildtype Ste3p PEST-like endocytosis signal or the 3K3 R version of this sequence. Three strains (NDY343-derived) were constructed having either GAL1-STE3, GAL1-STE3(wt)-Ub(7K3 R), or GAL1-STE3(3K3 R)-Ub(7K3 R) at the chromosomal STE3 locus. Cells were cultured and treated as described for Fig. 7C. Arrows at the left and at the right indicate the positions of Ste3p and of the Ste3(wt)-Ub(7K3 R) fusion protein, respectively. The position of protein species that crossreact with the Ste3p-specific antibodies are indicated by the hash marks to the right.
FIG. 11. Equivalent rates of ligand-dependent endocytosis for the 3K3 R and ⌬365 truncated receptors. Two MAT␣ pep4⌬ strains (NDY372-derived) were constructed having GAL1-STE3⌬365 or GAL1-STE3(3K3 R) replacing chromosomal STE3. Receptor expression was induced for 90 min in galactose medium and then repressed with the addition of glucose for an additional 45 min. Cultures were then treated with a-factor (ϩ) or were mock-treated (Ϫ) in parallel. A, time course for a-factor-dependent internalization of the Ste3(3K3 R)p. At the times indicated following pheromone addition, culture aliquots were collected and cells were subjected to the protease shaving protocol described for Fig. 3. The arrow to the right of the lower panel indicates the position of a protein species that cross-reacts with Ste3p-specific antibodies. B, quantitation of results from panel A. The fraction of receptor protein surviving the protease digestion was determined relative to amount of receptor present in the untreated aliquot from the same time point. This percent was subtracted from 100 and expressed as the "percentage of receptor at cell surface." Quantitation was by film densitometry as described under "Experimental Procedures." uptake. This leads us to the important conclusion that ubiquitin functions alone to trigger uptake, not in conjunction with additional receptor sequences or signals. Attachment of a single ubiquitin moiety to a yeast plasma membrane protein is all that is required to specify uptake. A further, more definitive test of this conclusion might involve construction of 7K3 R ubiquitin translational fusions with a plasma membrane protein such as Pma1p which normally does not undergo endocytosis. If monoubiquitin indeed functions alone to trigger uptake, then such a Pma1-Ub(7K3 R) fusion should undergo endocytosis.
Ubiquitination Signals: Proteosomal Targeting Versus Endocytosis-Our finding that ubiquitin acceptor sites map within the confines of the PEST-like signal perhaps seems unremarkable. For phosphorylation signals, the phosphate acceptor site (i.e. Ser, Thr, or Tyr) generally is a landmark feature within the consensus used by the kinase for recognition. This, however, generally is not true for ubiquitination signals that have been characterized to date, these being the signals associated with proteosomal targeting. The idealized ubiquitination signal has been described as being bipartite, with the recognition element and the ubiquitin attachment site, being separate and separable (35). The two elements may be distant from each other in the primary sequence and extreme cases, may even localize to distinct polypeptide subunits (36). Although formulated within the context of N-end Rule substrates (35), this bipartite model also appears to apply to a variety of ubiquitination signals; the recognition element, be it a PEST sequence or a destruction box, generally does not include the lysyl attachment site (37). Indeed, for a number of proteosomal substrates, investigation of the lysyl acceptor site use has uncovered a shocking level of plasticity; in several cases, any lysine within the substrate protein is capable of functioning as the ubiquitin acceptor site (38 -40).
The separation of recognition element and acceptor site seen for proteosomal substrates may relate to the need for multiubiquitin chain assembly. Attachment of an initial ubiquitin to a site embedded within the recognition element could have the consequence of disrupting the recognition element, potentially inhibiting subsequent ubiquitin addition. The addition of the multiple ubiquitin moieties, as required for the assembly of a multi-ubiquitin chain, may require a prolonged recognition of the signal by the E2 and/or E3 enzymes catalyzing this addition. For endocytosis, the multi-ubiquitin chain is not required (this work; Ref. 2). Furthermore, it has been suggested that recognition of substrate mono-versus multi-ubiquitination may be the key for deciding which of the two outcomes is invoked: endocytosis or proteosomal degradation (2). The strategy employed by the a-factor receptor wherein the acceptor sites are embedded within the recognition element may help to assure the endocytic outcome. If attachment of the first ubiquitin moiety disrupts the PEST-like recognition element and impairs its use for subsequent rounds of recognition, then ubiquitin addition would be self-limiting, with only one or several ubiquitin moieties being added.
Attachment of Multiple Ubiquitins Increases Uptake Rate-In addition to being subject to mono-ubiquitination, analysis of Ste3p ubiquitination shows that a subset of the receptor population is subject to bi-ubiquitination and perhaps to tri-ubiquitination as well. While the Ste3-Ub fusion results indicate that mono-ubiquitin suffices for initiating uptake, these same results also suggest a possible role for multi-ubiquitination in augmenting the rate of uptake. First, the Ste3(⌬399)-Ub(wt) fusion showed slightly faster turnover than did the Ste3(⌬399)-Ub(7K3 R) fusion ( Fig. 9 and Table II). As the Ste3(⌬399)-Ub(wt) fusion retains the ubiquitin lysines, its increased turnover may reflect secondary ubiquitination of the translationally attached ubiquitin moiety. A more striking example of en-hanced endocytosis due to multi-ubiquitination is evident with the Ste3(wt)-Ub(7K3 R) fusion; this fusion turns over with a t 1/2 of 9 min (Table II), significantly faster than wild-type Ste3p (t 1/2 ϭ 15 min; Table II). For this fusion, there apparently is a synergy between the natural ubiquitination of the receptor (specified by the PEST-like signal) and the artificial ubiquitination (the translationally fused 7K3 R ubiquitin). The added ubiquitin moiety fused to the receptor C terminus stimulates uptake beyond the rate seen for the wild-type receptor.
Based upon the effects of K63R ubiquitin mutants on Gap1p and Fur4p endocytosis, it has been suggested that short multiubiquitin chains with Lys 63 ubiquitin-ubiquitin linkages may serve to enhance the rate of uptake (32,33). For Ste3p, we now supply additional evidence for a stimulatory effect of multiubiquitination. However, given the functionality of the receptor fusions, which are obligately mono-ubiquitinated (this work; Ref. 2), we feel that the key recognized feature for endocytosis must be the mono-ubiquitin itself, not a linked chain. From this perspective, we would expect that multiple individual ubiquitins attached to separate substrate lysyl residues would prove just as stimulatory for endocytosis as linking these ubiquitins together in the form of a multi-ubiquitin chain.