A Large PEST-like Sequence Directs the Ubiquitination, Endocytosis, and Vacuolar Degradation of the Yeast a-Factor Receptor

Three URA3/2μ plasmids were constructed for CUP1-driven overexpression of wild-type or myc-tagged ubiquitin (Ellison and Hochstrasser, 1991). pND186 (CUP1-ubiquitin), pND747 (CUP1-myc-ubiquitin), and pND187 (CUP1 promoter only) were constructed through the replacement of the ClaI–BamHI interval of YEp24 with the CUP1-ubiquitin– containing ClaI–BamHI fragment from the corresponding URA3/CEN/ ARS plasmids pND164, pND165, and pND167, respectively (Roth and Davis, 1996).

The other plasmids constructed and used in this work divide into series of equivalent constructs carrying either wild-type STE3 or STE3 mutant versions. The strategy used in the construction of each series is reported in the following three sections.

For constructing in-frame deletions within the Ste3p CTD, the general strategy involved introduction of XhoI restriction sites by oligonucleotide-directed mutagenesis at various positions within the STE3 CTD-encoding sequences. Restriction and then ligation of upstream to downstream sites deletes the interval in between. Each XhoI site mutation replaced two adjacent STE3 codons with the sequence CTCGAG. Each replacement encoded the dipeptide leucine-glutamate and ligation of any two sites results in in-frame translation across the deletion.

Eight XhoI site mutations were constructed, at each of the following dipeptide codons: Leu₃₉₈Lys₃₉₉, Phe₄₁₃Asp₄₁₄, Ser₄₂₃Lys₄₂₄, Leu₄₃₃His₄₃₄, Phe₄₄₁Glu₄₄₂, Leu₄₄₆Cys₄₄₇, Pro₄₅₀Ala₄₅₁, and Ser₄₅₈Ser₄₅₉. In addition, a SalI site compatible with the XhoI sites was introduced at the Leu₃₂₀Leu₃₂₁ dicodon changing it to Val₃₂₀Asp₃₂₁.

Oligonucleotide-directed mutagenesis was by the method of Kunkel et al. (1987). The ssDNA template was derived from pSL1839, a 5.5-kb BamHI–SalI fragment from the original STE3 genomic library isolate (Hagen et al., 1986) carried on pRS306 (Sikorski and Hieter, 1989). Mutant plasmids for each restriction site were subjected to DNA sequencing in the vicinity of the site to confirm the fidelity of the mutagenesis.

Two additional SalI restriction sites, compatible with the XhoI site reading frames, were also used: the natural SalI site at STE3 codons 364 and 365, as well as a SalI site introduced via linker ligation to the PstI site located at STE3 codons 466–468 (see pSL1922 in Davis et al., 1993). This linker-derived SalI site was used in combination with the different XhoI site mutants to construct the series of “COOH-terminally”-truncated STE3 mutants. In fact, these truly are in-frame deletions, as they all retain the natural COOH-terminal STE3 dipeptide Gly₄₆₉Pro₄₇₀.

Assessment of receptor ubiquitination levels required that each of the receptor mutants be overexpressed from the GAL1 promoter. The plasmid pSL552 (Bender and Sprague, 1986) has a GAL1-controlled wild-type STE3 gene carried on the yeast URA3/CEN/ARS plasmid YCp50 (Rose et al., 1987). In-frame STE3 deletions constructed above in the pSL1839 plasmid context, were introduced into pSL552 via replacement of 380-bp SalI–SacI restriction fragment covering the COOH-terminal coding portion of STE3, with the equivalent fragment containing the desired deletion mutation.

The 12 triple alanine mutations created within the 414–449 interval of STE3 (see Fig. 8) also were constructed via oligonucleotide-directed mutagenesis (Kunkel et al., 1987). For each, three adjacent STE3 codons are replaced by the sequence GCTGCAGCC, encoding three consecutive alanines. This substitution also introduces a PstI site, allowing identification of the mutant clone by restriction analysis. The ssDNA template for mutagenesis was from pND210, the Δ450–468 mutant version of pSL1839. Fidelity of mutagenesis was confirmed for each via DNA sequencing in the vicinity of mutation site.

Two of the triple alanine mutations were also constructed within the context of wild-type (i.e., full-length) STE3. Reconstruction of STE3 with either the 417–419 or 426–428 triple alanine mutations (see Fig. 9) used the HindIII site at codons 433, 434, allowing the wild-type COOH-terminal STE3 sequences to be restored to the STE3Δ450–468 versions of these mutations.

The plasmid pXZ28 (Harris et al., 1994) has a GAL1-driven HA epitope– tagged PMA1 gene carried on a yeast URA3/CEN/ARS plasmid. The single HA epitope tag inserted near the NH₂-terminal coding sequences, following the second codon of PMA1 does not affect PMA1 function. The 750-bp EcoRI–XhoI pXZ28 GAL1 promoter fragment plus the 3.6-kb XhoI–PstI HA-PMA1–containing pXZ28 fragment were inserted into the polylinker of the yeast LEU2/CEN/ARS vector pRS315 (Sikorski and Hieter, 1989) between EcoRI and SacI to give pND533. This required modification of the PstI end of the 3.6-kb fragment by addition of a SacI linker. The unique XhoI site of pND533, at the fusion boundary between GAL1 and PMA1 sequences was destroyed via end-filling and re-ligation to give pND542. To create pND563, the immediate progenitor of PMA1-STE3 fusion protein plasmids, a new XhoI site was introduced into PMA1 by oligonucleotide-directed mutagenesis at COOH-terminal position two residues from the end of the coding sequences (the XhoI site substitutes for the codons Lys₉₁₆Glu₉₁₇). The PMA1 XhoI site is designed to accommodate in-frame fusions with the STE3 XhoI sites. For most of these fusions the XhoI site at STE3 codons 398, 399 was used (see above) with the contributed STE3 sequences extending from codon 400 to different COOH-terminal endpoints. For instance, to construct the PMA1- STE3(400–440) fusion protein, the 398, 399 XhoI site was introduced via oligonucleotide-directed mutagenesis into the STE3Δ441–468 deletion mutant context. From this plasmid, a 250-bp XhoI–SacI fragment could be excised (the SacI site is located 3′ of the STE3 open reading frame [ORF]) and used to replace the XhoI–SacI fragment from the GAL1-HA-PMA1 plasmid pND563.

Genotype and source for some of the yeast strains used in this work are reported in Table I. The two new strains listed, NDY343 and NDY344, isogenic ste3Δ::LEU2 versions of the strains NDY334 and NDY335 (Roth and Davis, 1996), were constructed via the two-step gene replacement strategy (Boeke et al., 1987) as described previously (Roth and Davis, 1996), wherein a ste3Δ::LEU2 allele was used to replace STE3. The ste3Δ:: LEU2 allele, carried on the integrating plasmid pSL2165, a pSL1839 derivative (see “Plasmids”), in which the entire STE3 ORF is replaced by the 2.85-kb BglII LEU2 fragment. The ste3 deletion extends from a point 417 bp upstream of the STE3 ORF to a point 111 bp downstream. pSL2165 was uniquely restricted at a SfiI site located upstream of STE3 to direct chromosome insertion.

In addition to the strains reported in Table I, numerous other strains were constructed and used in this work. The turnover of each Ste3p mutant was measured in strains in which the mutant STE3 allele replaces the ste3Δ:: LEU2 allele of NDY343. This strain construction used the two-step gene replacement strategy (Boeke et al., 1987) as previously described (Roth and Davis, 1996). For each, the transplacing allele, derivatives of the integrating plasmid pSL1839 with the mutant STE3 in place of wild-type STE3, were restricted with HpaI to direct chromosomal insertion. This eliminated one potential source of experimental variability—differences in receptor expression levels resulting from clonal fluctuations in plasmid copy number.

Rabbit antisera was raised against the STE3 CTD, as part of one of two bacterial fusion proteins, TrpE-STE3 (Clark et al., 1988) or MalE-STE3. All Ste3p-specific antisera were affinity purified to the MalE-STE3 antigen coupled to CNBr-activated Sepharose 4B (Sigma Chemical Co., St. Louis, MO). The HA.11 mAb (Berkeley Antibody Co., Berkeley, CA) was used for detection of HA epitope–tagged PMA1-STE3 fusions. The myc mAb (Oncogene Research, Uniondale, NY) was used for detection of the myc 9E10 epitope–tagged ubiquitin (Ellison and Hochstrasser, 1991). Pgk1p-specific antiserum was generously provided by T. Stevens (University of Oregon, Eugene, OR).

Pulse-chase analysis (Roth and Davis, 1996) entailed in vivo labeling of yeast cultures via a 10-min pulse of [³⁵S]methionine, which was followed by addition of the non-radioactive chase: methionine added to a concentration of 750 μg/ml, cysteine to 150 μg/ml, and yeast extract to 0.2%. At various times after chase addition, culture aliquots were removed and extracts were prepared by the glass bead method described previously (Roth and Davis, 1996). Wild-type or mutant receptors were then isolated via two iterative rounds of immunoprecipitation (Roth and Davis, 1996). Each round used different affinity-purified anti-Ste3p antibodies derived from the polyclonal antisera of different rabbits. Immunoprecipitated samples were subjected to SDS-PAGE, and then both to autoradiography and to Storm PhosphorImaging (Molecular Dynamics, Inc., Sunnyvale, CA) for the quantitative analysis.

For immunoprecipitation of the HA-tagged PMA1-STE3 fusion proteins, 20 μl of extract corresponding to 50 × 10⁶ cells, prepared as described (Davis et al., 1993), was subjected to a single round of immunoprecipitation (Roth and Davis, 1996) using 30 μl of a slurry of the HA.11 mAb conjugated to agarose beads (Berkeley Antibody Co.). Incubation with the mAb in this case was for 16 h at 4°C.

To assess the susceptibility of newly synthesized receptor to digestion by extracellular proteases, cells were first subjected to in vivo labeling via the pulse-chase protocol described above. At various times after chase addition, culture aliquots were removed and intact cells were treated with Pronase (Calbiochem-Novabiochem Corp., La Jolla, CA) and protein extracts were prepared (Davis et al., 1993). Immunoprecipitation and further analysis was as described above.

For Western analysis, the relevant STE3 mutant or PMA1-STE3 fusion protein was expressed from the GAL1 promoter. For these experiments, overnight log-phase cultures in rich YP medium containing 2% raffinose were inoculated from minimal plates on which plasmid retention was selected. Induction of the GAL1 promoter involved the addition of galactose to 2%. The duration of the galactose induction was as indicated for each figure and was terminated when synthesis repression was desired with the addition of glucose to 3%. Protein extract preparation, SDS-PAGE, and Western analysis were as described previously (Roth and Davis, 1993).

Treatment of protein extracts with potato acid phosphatase (Boehringer Mannheim Corp., Indianapolis, IN) at a final concentration of 0.06 U/ml was as described previously (Roth and Davis, 1996).

Quantitating the proportion of receptor proteins with attached ubiquitin was by Western analysis of phosphatase-treated protein extracts for the various STE3 mutants. Quantitative AMBIS densitometry of films (AMBIS Systems, Inc., San Diego, CA) resulting from ECL development of the Western blot (Amersham Corp., Arlington Heights, IL) was used to estimate the relative amount of the receptor protein present in the two ubiquitin-modified electrophoretic positions (the presumptive mono- and di-ubiquitinated forms; Roth and Davis, 1996), and at the unmodified position. Many aspects of this analysis rely on techniques wherein the measured output is produced non-linearly. To circumvent this potential source of error, we have compared signals only of similar intensity levels. This was achieved by applying the Western analysis to a dilution series for each sample. For instance, with phosphatase-treated extracts derived from wild-type Ste3p, the intensity level of the signal associated with the mono-ubiquitinated band was found to be intermediate between the level of the unmodified receptor band present from the 1:5 and 1:25 dilutions of the same extract. Quantitative densitometry of the band signals from the same compared samples led to the determination that 10% of the wild-type receptor was present in the mono-ubiquitinated species band. As 10% was also present in the di-ubiquitinated band, the summed estimate of 20% was reported in Fig. 3 as proportion of receptor ubiquitinated.

Our previous work showed that the rapid constitutive endocytosis of Ste3p requires sequence elements within the COOH-terminal 72–amino acid residues of the receptor (Davis et al., 1993). To better delineate this requirement we have constructed a series of in-frame deletion mutants within the receptor's dispensable CTD. First we have first investigated if CTD sequences in addition to the 72-residue COOH-terminal segment are required for rapid constitutive endocytosis. The 187-residue-long Ste3p CTD extends from Asp₂₈₄ (the first charged residue after the seventh transmembrane segment) to a COOH-terminal Pro₄₇₀. Three large deletion mutants, deleting residues 320– 363 (STE3Δ320–363), residues 365–414 (STE3Δ365–414), and the COOH-terminal residues 413–468 (STE3Δ413–468) were used. As a first assessment of endocytosis, the rate of degradative turnover of each of the different receptor mutants was determined (Fig. 1 A). While wild-type Ste3p turns over rapidly with a t_1/2 estimated to be ∼15 min, the Δ413–468 mutant receptor is completely stable, indicating that the COOH-terminal 58 residues are required for this ligand-independent receptor turnover. Like wild-type, both the Δ365–414 and Δ320–363 receptors show rapid turnover consistent with endocytosis that is largely unimpaired. Thus, of the STE3 CTD sequences, the COOH-terminal 58 residues play the primary role in specifying rapid turnover. We have therefore concentrated our analyses on this interval.

The turnover defect of Ste3Δ365–468p, a mutant missing its COOH-terminal 105 residues, is a consequence of blocked endocytic transport from surface to vacuole (Davis et al., 1993). With endocytosis blocked the receptor accumulates at the cell surface. If the turnover defect of the Δ413–468 mutant receptor also results from impaired endocytosis, then this receptor should also accumulate at the cell surface. To assess surface localization of the Δ413– 468 receptor, we have used protease susceptibility as a probe, testing the accessibility of the receptor within whole cells to digestion by added, external protease. This assay distinguishes surface-localized receptor (susceptible to external proteases), from receptor that localizes to compartments inside the cell (resistant to external proteases) (Davis et al., 1993). As for Fig. 1,A, cells were subjected to pulse-chase analysis. At various times after the initiation of the non-radioactive chase, aliquots of the labeled, intact cells were subjected to protease treatment (Fig. 1,B). Digestion of surface-localized receptor is evident both in terms of loss of receptor protein and by the production of a protease digestion product corresponding to the protected, intracellular CTD plus final transmembrane domain (Davis et al., 1993; the antibody used for the immunoprecipitation is STE3 CTD specific). Wild-type Ste3p in such a regimen is only transiently susceptible to the added proteases, reflective of its short surface residency time (Davis et al., 1993). Here, we find that the Δ413–468 mutant receptor remains susceptible to the proteases even 90 min after synthesis (Fig. 1,B). Quantitation via PhosphorImaging indicates that 75% of Δ413–468 receptor protein is susceptible to protease 45 min after synthesis and 60% remains susceptible at the 90-min time point. Cell integrity is maintained throughout the digestion protocol as evidenced by the complete protection from digestion of a cytoplasmic marker protein, phospho-glycerol kinase (Pgk1p; Fig. 1 B, bottom panel). We conclude from its prolonged surface residency time that turnover defect found for Ste3Δ413–468p does indeed reflect impaired endocytosis.

To identify the sequence element(s) that direct the constitutive endocytosis of Ste3p we have constructed a finer series of deletions within the 413–468 interval. First we tested the effects of three deletions extending progressively in from the receptor COOH terminus. The in vivo rate of turnover was assessed for wild-type Ste3p and for each mutant receptor via a pulse-chase regimen similar to that described for Fig. 1,A. Immunoprecipitated receptors were subjected to SDS-PAGE (Fig. 2,A) and then to PhosphorImaging analysis to quantitate their rate of loss to degradation (Fig. 2,B). As seen previously (Fig. 1,A), turnover is rapid for wild-type Ste3p and wholly blocked for Ste3Δ413–468p. The two intermediate deletion mutants, Δ450–468 and Δ441–450 show intermediate rates of turnover. Quantitation of the receptor-associated radioactivity at each time-point is shown in Fig. 2,B. Most of the data points for each receptor conform to a linear plot, allowing the calculation of turnover half-life for each mutant (Fig. 2 B). For wild-type Ste3p, a t_1/2 of 16 min is calculated. For Ste3Δ450–468p, turnover remains rapid, though somewhat less so than wild-type: a t_1/2 of 21 min is calculated. With the progressive extension of these deletions, a progressive impairment is evident. A t_1/2 of 65 min is calculated for the Δ441–468 receptor and the turnover block appears to be complete for the Δ413–468 receptor.

For two of the receptors, namely wild-type and Δ450– 468, the 90-min time-points fail to conform to the line defined by the other time-points (Fig. 2 B). This is seen consistently and indicates that a minority of these receptors, 3% for wild-type and 13% for Ste3Δ450–468p, do not participate in the normal endocytic turnover mechanism. Hypothetically, these could be receptors in the process of undergoing recycling or receptors that are trapped to a locale that prohibits their endocytosis.

Fig. 3 summarizes the results from a total of eight COOH-terminal deletion mutants. Again, it is apparent that endocytic function is lost gradually with progressive removal of sequences from the COOH-terminal end. A small, but significant slowing of the turnover rate is apparent with removal of the COOH-terminal 12 amino acids (Fig. 3; t_1/2 of 25 min for Δ458–468 vs. 15 min for wild-type). A second transition is broached with the Δ447–468 deletion; the t_1/2 for turnover slows from 25 to 40 min. Further impairment is seen when the deletion is extended to residue 441, with the t_1/2 for Δ441–468 being 70 min. Finally, with the deletion extended to residue 423, endocytosis appears to be wholly blocked. This gradual loss of function for deletions extending over the COOH-terminal 48 amino acids indicates the “signal” for constitutive endocytosis may be large and complex, potentially composed of redundant functional elements.

To delimit the upstream boundary of this endocytosis signal, small internal deletions were introduced into the Δ450–468 mutant receptor. The Δ450–468 deletion removes the COOH-terminal 20 amino acid residues and was used as the context for these other small deletions to eliminate the potential contribution of redundant elements within the 450–470 interval. The experiment of Fig. 1 indicated that a large portion of the receptor CTD (the 320–414 interval) is not required for rapid endocytic turnover. Again here within the Δ450–468 context, introduction of a second deletion of residues 398–414 caused no additional defect; the t_1/2 for turnover of the double deletion mutant Δ398–414, 450–468 was unchanged from that of the parental Δ450–468 construct (Fig. 3). However, introduction of either the Δ413–424 or the Δ423–434 deletions into the Δ450–468 receptor wholly blocked turnover. This indicates that sequences required for rapid endocytosis extend to the NH₂-terminal side of residue 423, potentially as far as residue 413. Thus, the 36-residue-long 414– 449 sequence interval appears to be the most critical for directing rapid endocytosis. Consistent with this, deletion of just these amino acids (STE3Δ413–451) abolishes turnover (Fig. 3). However, the COOH-terminal 20 residues also may contribute to turnover: both the Δ450–468 and Δ458–468 deletions have small, but significant effects on the turnover rate (Figs. 2 and 3). Support for the participation of the COOH-terminal 20 residues in endocytosis is offered below (see Fig. 9) with experiments indicating a functional redundancy between sequence elements in the 414–449 interval and those in the 450–470 interval.

We next tested if the STE3 mutants defective for constitutive endocytosis are also defective for the associated constitutive ubiquitination. Receptor ubiquitination levels of the various STE3 mutants were assessed via Western analysis: ubiquitinated receptor species manifest an 8-kD shift with each added ubiquitin moiety. Detection of the ubiquitinated forms is improved with the use of several key experimental conditions, including receptor overproduction (∼10-fold) from the GAL1 promoter. This enhances detection of the ubiquitinated forms, yet neither alters receptor turnover rate nor the proportion of the receptor subject to ubiquitination (data not shown). Secondly, to avoid loss of the ubiquitinated forms to vacuolar degradation, pep4Δ mutants deficient in vacuolar protease activity were used. Finally, treatment of protein extracts with phosphatase before electrophoresis eliminates the substantial heterogeneity in gel mobility because of the heterogenous phosphorylation of Ste3p (Fig. 7; also Roth and Davis, 1996).

In Fig. 4,A, the ubiquitination of three receptor mutants is compared with wild-type. For wild-type, the majority species (∼80%) has no attached ubiquitin (Roth and Davis, 1996). The remaining 20% distributes almost equally between two, more slowly-migrating, ubiquitinated forms (indicated in Fig. 4,A, arrows at left), the mono- and di-ubiquitinated receptor species (Roth and Davis, 1996). Mutants that block endocytosis, namely Δ413–468 and Δ413–451 (Fig. 3), also blocked receptor ubiquitination (Fig. 4,A). Furthermore, the Δ450–468 mutant receptor that showed diminished turnover (Fig. 2) also shows diminished ubiquitination (Fig. 4,A). Quantitation of these results along with similar analyses applied to the other mutant receptors is reported in Fig. 3. Comparison of the turnover half-life with ubiquitination levels shows a consistent correlation between the two—further compelling evidence for a functional connection linking ubiquitination and endocytosis.

Previously, we have shown that in end4-1 cells, constitutive Ste3p endocytosis is blocked and the receptor is consequently trapped at the cell surface in a ubiquitinated state (Roth and Davis, 1996). Here, we again make use of the end4 mutant strain background to eliminate endocytosis as a possible contributing cause to the differences seen in ubiquitination levels for the different receptor mutants. If, for instance, ubiquitin addition occurred at some post-surface, endosomal step, then mutations like Δ413–468, which block uptake, might block ubiquitination by preventing access of the receptor to downstream steps. In such an experiment (Fig. 4,B), essentially the same result is seen as was apparent in the END4⁺ background (Fig. 4 A). While both the wild-type and Δ450–468 receptors are ubiquitinated, the Δ413–468 receptor is not, and the Δ441– 468 receptor shows greatly reduced levels of ubiquitination. Ste3p ubiquitination, therefore, likely occurs while receptor is surface localized, serving potentially as the initiating trigger for uptake.

We have identified a domain mapping to the COOH-terminal portion of the receptor's CTD that directs both the ubiquitination and the rapid constitutive endocytosis of the receptor. In this section we test the sufficiency of this sequence to direct the ubiquitination and endocytosis of a normally long-lived constituent of the plasma membrane, the plasma membrane ATPase Pma1p. Like Ste3p, Pma1p is a polytopic plasma membrane protein with a cytoplasmically exposed, COOH-terminal tail. Unlike Ste3p, Pma1p is quite stable with a t_1/2 estimated at >11 h (Benito et al., 1991). By site-directed mutagenesis, an XhoI site was created at the penultimate PMA1 codon to provide a locus for the in-frame fusion of the various STE3 sequences to be tested. Fusions are expressed from the GAL1 promoter, and have embedded near the NH₂-terminal–encoding end of the PMA1 ORF, a single HA epitope tag that does not impair PMA1 function (Harris et al., 1994). Thus, turnover of the fusion proteins may be easily assessed as follows: (a) PMA1-STE3 proteins are expressed via growth on galactose-containing medium, (b) synthesis of the construct protein is then blocked with the addition of glucose, and (c) the time-dependent loss is followed via immunoblotting with anti-HA mAb. This same experimental system was used previously to show that a 108-residue portion of Ste6p sufficed to direct endocytosis of a PMA1-STE6 fusion protein (Kolling and Losko, 1997).

Five different PMA1-STE3 fusion proteins were constructed, each with a different segment of the STE3 CTD fused at the COOH terminus of PMA1. The parental HA-tagged Pma1p showed no evidence of turnover during the 2-h course of this experiment (Fig. 5). Attachment of the 347–398 STE3 sequence interval was without effect; this PMA1-STE3 fusion protein shows a stability equivalent to that of the starting PMA1 construct (Fig. 5). In contrast, attachment of the 400–470 interval, encoding the COOH-terminal 71 amino acids of Ste3p, does destabilize the resulting PMA1-STE3 fusion protein (Fig. 5). This fusion protein is degraded with a t_1/2 of 40 min (Table II). Likewise, the fusion carrying the 400–449 STE3 sequence interval also turns over (Fig. 5); a t_1/2 of 50 min is calculated for this protein (Table II). Turnover slows dramatically when the critical 414–449 interval is violated as it is for PMA1-STE3(400–440) (Fig. 5); a t_1/2 of 90 min is calculated (Table II). Finally, removal of the entire 414–449 interval, as for PMA1-STE3(400–412), effectively abolishes turnover (Fig. 5; Table II).

In addition to the five PMA1-STE3 constructs described above, four additional constructs were made with the STE3 contribution extending various distances into the critical 414–449 interval. These constructs were subjected to analyses identical to those in Fig. 5 and the resulting calculated turnover t_1/2 is reported for each (Table II). In general, these results correlate quite well with the role of these sequences in Ste3p turnover (Fig. 3). Sequences that showed an intermediate rate of turnover in the Ste3p context also show intermediate effects when tested in the PMA1-STE3 context. In addition, we find that the minimal endocytic signal, the 414–449 STE3 interval, suffices to promote turnover of the resulting PMA1-STE3 fusion (Table II).

Degradation of Ste3p depends on transport from the cell surface to the vacuole (Davis et al., 1993; Roth and Davis, 1996). Is the turnover of the PMA1-STE3 constructs instigated by a similar route of transport? To test involvement of vacuolar proteases in PMA1-STE3 fusion turnover, constructs with the STE3 400–449 peptide sequence interval were introduced into isogenic PEP4⁺ and pep4Δ strains and turnover assessed. While turnover proceeds in the PEP4⁺ background, in the pep4Δ context, turnover is completely blocked (Fig. 6, left panel). This is also true for the longer PMA1-STE3(400–470) fusion protein (data not shown). Thus, like wild-type Ste3p, turnover of these PMA1-STE3 fusions also depends on vacuolar proteases and therefore also likely depends on transport to the vacuole.

To test the dependence of turnover on endocytic transport, an end4 mutant was used. The PMA1-STE3(400– 449) fusion was introduced into isogenic END4⁺ and end4-1 cells. Turnover is blocked in the end4 background (Fig. 6, right panel), indicating that as for Ste3p itself, turnover of the PMA1-STE3(400–449) fusion depends on its endocytosis from cell surface to vacuole. Again, this is also true for the PMA1-STE3(400–470) fusion as well (data not shown).

Also apparent in the end4 background is the presence of modified forms of the 400–449 PMA1-STE3 fusion that migrate more slowly than the mature fusion protein band (Fig. 6, species indicated by arrowhead). Although accentuated by end4, a similar modification of the PMA1-STE3(400–449) fusions was apparent both in the wild-type strain backgrounds (Fig. 5) and in the pep4Δ background (Fig. 6, left panel). Such modified species also were seen for the longer PMA1-STE3(400–470) fusion, but were not seen either for the other fusions or for the HA-tagged Pma1p alone (Fig. 7 B).

To test if the modification is ubiquitination, we have made use of a myc epitope–tagged ubiquitin (Ellison and Hochstrasser, 1991). Multi-copy yeast plasmids expressing either wild-type ubiquitin, the myc-tagged ubiquitin, or no ubiquitin were introduced into end4 cells with one of three PMA1-STE3 constructs, the 400–470 fusion, the 400–449 fusion, or the long-lived 400–412 fusion. The PMA1-STE3 fusion proteins were purified from extracts of these cells via immunoprecipitation with the HA mAb. Immunoblotting of the purified fusion proteins with the HA mAb, allowed visualization of the fusion constructs and showed that overexpression of either the wild-type or epitope-tagged ubiquitin results in no obvious change in the PMA1-STE3 fusion protein electrophoretic mobility (Fig. 7,A). Parallel immunoblots, probed instead with the myc mAb, allowed assessment of the incorporation of the myc-tagged ubiquitin into the fusion constructs (Fig. 7,A). The stable 400-412 fusion, did not react with the anti-myc monoclonal (Fig. 7,A). In contrast, both the 400–449 and 400– 470 fusions showed a clear reaction. Reaction was specific for cells coexpressing the myc epitope–tagged ubiquitin. The 400–470 fusion showed a somewhat stronger reaction than did the 400–449 fusion (Fig. 7 A), consistent with a greater proportion of this fusion protein being ubiquitin modified.

Using our panel of PMA1-STE3 fusions, we next tested the STE3 sequence requirement for the ubiquitin modification. Extracts prepared from end4 cells expressing each of the different PMA1-STE3 fusions were subjected to Western analysis with the HA mAb (Fig. 7,B). A prominent accumulation of modified forms is apparent for the two fusions retaining the critical 414–449 interval, i.e., the 400–449 and 400–470 fusions (Fig. 7,B). Smaller fusions violating the integrity of 414–449 STE3 interval do not show this massive accumulation of modified species (Fig. 7,B). (The minor species migrating slightly above the major fusion protein band for each of these other fusions is not likely to represent ubiquitinated protein: the 400–412 fusion that manifests this minor species in Fig. 7,B, fails to react with myc mAb in Fig. 7 A.) Thus, just as truncation of the 414–449 interval impairs Ste3p ubiquitination, intrusion into this interval also abolishes ubiquitination of the corresponding PMA1-STE3 fusion proteins. We conclude that the 414–449 STE3 sequence interval is the minimal signal that suffices to specify both ubiquitination and endocytosis.

The 36-residue-long 414–449 peptide sequence comprises a ubiquitination/endocytosis signal for Ste3p. To gain insight into the nature of this sequence, we have applied alanine-scanning mutagenesis across this entire interval. Again, to avoid the potential effects of redundant signal elements within the 450–470 interval, mutations were made within the STE3Δ450-468 context. 12 mutants in total spanning the 414–450 interval were constructed (Fig. 8). For each, three adjacent residues were substituted by three alanines. Each mutant was subjected to pulse-chase analysis identical to that previously applied to the various STE3 deletion mutants (Figs. 1 and 2). The calculated t_1/2 are reported in Fig. 8.

Results from the alanine scanning show a good concordance with the deletion analysis (Fig. 3). While the deletion analysis indicated a requirement of the entire 414–449 interval, the 414–433 interval was found to be particularly crucial. For instance, though grossly impaired, the Δ441– 468 and Δ434–468 receptor mutants still showed a slow rate of turnover (t_1/2 = 70 min; Fig. 3). Deletions that violated the more critical 414–433 interval (e.g., Δ423–468 or Δ413–468) completely abolished turnover. Likewise for the triple alanine mutations, mutations that totally disabled endocytosis were concentrated within the crucial 414–433 interval; of the five mutants with t_1/2 of >2 h, four locate within the 414–433 interval (Fig. 8). Alanine mutations in the 434–450 interval generally showed intermediate levels of impairment (Fig. 8). The one exception to this is the replacement of the Ser-Ser-Asn 438–440 tripeptide, which appeared to fully disable endocytosis. We are presently without an explanation of how this triple alanine replacement could have a more severe phenotype than the deletion interval encompassing it.

The number of these triple alanine mutations having pronounced effects on turnover is striking (Fig. 8). Residues important for endocytosis distribute throughout the 414–450 interval. No singular sequence motif is discernible. Instead, the entire 36-residue-long sequence is required in its totality. The most prominent feature of this interval, which we later discuss, is the high proportion of acidic (glutamates and aspartates) and hydroxylated residues (serines and threonines).

The 414–449 sequence is the minimal ubiquitination/endocytosis signal. This, however, does not exclude the possible participation of other STE3 sequences. A small, but reproducible effect on turnover was seen, for instance, with removal of the COOH-terminal 20 amino acids: the t_1/2 for Ste3Δ450–468p is 25 min compared with 15 min for wild-type (Fig. 2). Furthermore, like the 414–449 interval, the 450–470 interval is also rich in both acidic and hydroxylated amino acids. To test the contribution of these sequences to endocytosis, we have examined several mutations that were known to disable turnover in the Δ450–468 receptor, within the context of the full-length receptor.

The triple alanine replacements were constructed and tested originally within the context of the Δ450–468 deleted receptor. Two of these that had disabling effects in the Δ450–468 receptor, were chosen and reconstructed in the full-length receptor context. The effects on the rate of receptor turnover were tested via pulse-chase analysis in both contexts (Fig. 9). In the Δ450–468 context, both mutations severely retard endocytic turnover. Where the Δ450–468 receptor turns over with a t_1/2 of 25 min, the same receptor with the triple alanine substitution of the Thr-Leu-Asp tripeptide at 426–428 turns over with a t_1/2 of 90 min (Fig. 9). The effect of the alanine substitution of the 417–419 tripeptide Ile-Ser-Leu is even more severe, showing a t_1/2 of >2 h (Fig. 9). In the full-length receptor context, while the two triple alanine mutations do cause some slowing of turnover, the effects are far less severe. While the wild-type receptor turns over with a t_1/2 of 12 min in this experiment, for triple alanine substitutions at 417–419 and at 426–428, turnover is slowed only to a t_1/2 of 21 and 17 min, respectively (Fig. 9). We conclude therefore that elements within the 450–470 sequence may compensate the loss of key elements in the 414–449 interval.

We have seen previously that extension of a COOH-terminal deletion from residue 450 to residue 441 severely impairs endocytic turnover (Figs. 2, 3, and 9; compare Δ450–468 with Δ441–468). This implies a critical role for the 441–449 sequence. To test the importance of these sequences in an otherwise full-length receptor, we have constructed a receptor mutant in which only the 441–451 sequences are deleted. Like the two triple alanine mutants tested above, the Δ441–451 deletion has only minor effects in the full-length receptor context (Fig. 9). Again this suggests a functional redundancy between elements in the 414–449 interval with elements in the COOH-terminal 450–470 interval. It may be that the entire 57-residue-long segment from 414 to the COOH terminus normally functions to specify Ste3p ubiquitination and endocytosis. The 414–449 interval, highlighted by our deletion analysis may correspond simply to that minimal portion still capable of promoting rapid endocytosis.

We have characterized the sequences within the yeast a-factor receptor that direct its rapid, ligand-independent endocytosis. The essential sequence is striking, both in terms of its large size, minimally defined as 36 residues long, and its high density of both acidic amino acids and hydroxylated amino acids, particularly serine. Not only does this sequence specify uptake, but also ubiquitination. Receptor mutants that lack this sequence are not ubiquitinated and mutants partially deleted for this interval show a reduced level of ubiquitination that correlates well with their residual capacity for endocytosis. This sequence also is a sufficient, self-contained, transportable unit. When transplanted into the normally stable cell surface protein Pma1p, the STE3 sequence directs both ubiquitination and endocytosis: the result being delivery of the PMA1-STE3 fusion protein to the vacuole for degradation.

Examination of the STE3 signal by visual inspection and by alanine-scanning mutagenesis reveals no obvious similarities to the short L- or Y-based peptidyl signals that mediate clathrin-dependent uptake in mammalian cells (Mellman, 1996; Schmid, 1997). Furthermore, no example of the two sequences that have been previously associated with the ligand-dependent uptake modes of the two yeast pheromone receptors are apparent (Rohrer et al., 1993; Tan et al., 1996). The large size of the STE3 ubiquitination/endocytosis signal could reflect a folding requirement to a particular three-dimensional shape for functionality. Alternatively, the functional signal could consist of an extended string or cluster of certain classes of residues. In this regard, the high proportion of both acidic and hydroxylated residues within both the 36-residue minimal signal and the more inclusive 57-residue sequence is striking. These are two of the three key sequence elements of PEST sequences, a set of sequences that act as signals for ubiquitination and consequent proteosomal turnover for various short-lived cytoplasmic and nuclear proteins. The third element of PEST sequences are prolines; there are two prolines within the COOH-terminal 57 residues of Ste3p, however, neither locate within the 414–449 interval.

PEST sequences are responsible for directing the ubiquitin-dependent proteosomal degradation of a variety of short-lived cytoplasmic and nuclear proteins (Rechsteiner and Rogers, 1996). Though PEST sequences share no primary sequence identity to one another, they do share overall character—protein sequences, 10–50-residues-long containing unusually high densities of prolines, the acidic residues glutamate and aspartate, and the hydroxylated amino acids serine and threonine. The resemblance of the STE3 signal to PEST sequences is consistent with the initiating role that ubiquitin has been proposed to play in Ste3p endocytosis (Roth and Davis, 1996). The STE3 signal may be first and foremost a signal for ubiquitination.

Recent work on two other yeast plasma membrane proteins, the a-factor export protein Ste6p and the uracil permease Fur4p, indicate that sequences similar to the STE3 signal likely participate in the uptake of these proteins as well (Kolling and Losko, 1997; Marchal et al., 1998). Like Ste3p, endocytosis and turnover of Ste6p and Fur4p also appear to be ubiquitin dependent (Kolling and Hollenberg, 1994; Galan et al., 1996; Loaza and Michaelis, 1998). For Ste6p, deletion of an acidic 61-residue sequence rich in serines and threonines, abolished both Ste6p ubiquitination and its rapid turnover. Though this 61-residue-long STE6 sequence was not sufficient to direct endocytosis of a PMA1-STE6 fusion construct (equivalent in design to the PMA1-STE3 constructs used herein), a larger, more inclusive 108-residue sequence did suffice (Kolling and Losko, 1997). Mutagenic studies on Fur4p, also demonstrate the involvement of a sequence rich in acidic and hydroxylated residues in uptake and vacuolar degradation of this permease (Marchal et al., 1998). These two sequences, together with the STE3 sequence appear to constitute a new class of endocytosis signal—a class where the primary function of the signal may be to instigate ubiquitination.

Though resembling PEST sequences, these three ubiquitination/endocytosis signals do not score strongly as PEST sequences. This algorithm (Rogers et al., 1986) has proved to be a powerful predictor of proteins subject to rapid ubiquitin-dependent proteosomal turnover. Nonetheless, as there is little direct molecular data concerning the essential features of PEST sequences, the strictures of this algorithm are necessarily somewhat arbitrary. The algorithm examines only those sequences that are bounded by basic residues (K, R, or H), and that include at least one Pro, at least one acidic residue (D or E), and at least one hydroxylated amino acid (S or T). Of these, the strength of the resulting PEST score mostly depends on the proportion of amino acids within the interval that are either P, D, E, S, and T. The failure of the three ubiquitination/endocytosis signals to score strongly as PEST sequences reflects the paucity of proline residues within these sequences. This could indicate either that the PEST algorithm is overly stringent with regards to its proline requirement, or more interestingly, a point of functional divergence of these signals with PESTs. Although there appears to be significant overlap in the enzymatic machinery catalyzing the ubiquitination of both proteosomal substrates and cell surface proteins targeted for endocytosis (see below), the ubiquitination requirements for these two processes may be distinct. Where formation of a multi-ubiquitin chain on the substrate is generally required for proteosomal recognition (Hochstrasser, 1996), recent work on the α-factor receptor indicates that the addition of a single ubiquitin moiety suffices to direct endocytosis (Terrell et al., 1998). One possibility, therefore, is that the prolines present in the PEST sequences of proteosomal substrates play a role in specifying the construction of the multi-ubiquitin chain. In any case, as the relationship of these new endocytosis signals to PEST sequences remains uncertain, we perpetuate the term “PEST-like,” coined by Marchal et al. (1998) in their description of the Fur4p signal.

The 36-residue-long STE3 signal is the minimal sequence sufficient for ubiquitination and endocytosis. PEST-like sequences in STE3, however, extend to the COOH terminus and include in addition to the 36 residues (414–449), the contiguous 21 COOH-terminal residues (450–470). We have presented evidence that this COOH-terminal sequence (450–470) may participate together in redundant fashion with the 414–449 signal. Similar redundancy may also apply within the minimal 414–449 interval. Loss of function was gradual with extension of the deletions into this interval. With extension of the deletion from residue 450 to residue 447, the t_1/2 increased from 25 to 40 min (Fig. 3, compare Δ450–468 and Δ447–468). With further extension to residue 441 or to residue 434, the t_1/2 increased to 70 min (Fig. 3, Δ441–468 and Δ434–468). Only deletions extending to residue 423 and residue 413 effectively abolished endocytosis (Fig. 3, Δ423–468 and Δ413– 468). Thus it appears that both the rate of endocytosis, and also the degree of ubiquitin modification (Fig. 3) reflects the length of the PEST-like domain remaining in each mutant. Functionality of the signal perhaps depends on its overall size and on overall density of acidic and hydroxylated residues. Larger, more negatively charged sequences may provide better substrates for ubiquitination.

Negative charge within a potential ubiquitination/endocytosis signal could be further increased through the introduction of phosphates. Such a model has been suggested both for Fur4p endocytosis (Marchal et al., 1998) and for the ligand-dependent endocytosis of Ste2p (Hicke and Riezman, 1996; Hicke et al., 1998). Addition of acidic phosphate moieties within an already acidic domain perhaps elevates a sequence through some threshold barrier of required negative charge density, allowing recognition by the E3 ubiquitination machinery. For a number of diverse proteosome substrates, phosphorylation within PEST sequences often precedes and serves to instigate subsequent ubiquitination (Hochstrasser, 1996; Rechsteiner and Rogers, 1996). This may be true for Ste3p as well, as it clearly is subject to phosphorylation (Roth and Davis, 1996).

In terms of potential ubiquitin acceptor sites, Ste3p has a total 23 cytoplasmically available lysine residues. Three map within the COOH-terminal 57-residue PEST-like sequence and of these, two within the 36-residue minimal signal. The disabling effect of the triple alanine mutation of residues 432–434 (Fig. 8) suggested the possibility that Lys432 might serve as the sole ubiquitin acceptor site within the Δ450–468 receptor. However, this is not supported by the effects of a K432R point mutation. Within the context of the Δ450–468 receptor, this mutation failed to block receptor ubiquitination and showed only an intermediate turnover disability (t_1/2 = 60 min; data not shown). Possibilities currently being tested are either that the ubiquitination occurs at lysines mapping outside of the PEST-like sequence and/or that multiple lysines are used redundantly.

The endocytosis signal characterized herein is also a ubiquitination signal. While decoding this signal could involve the binding of the adaptins and coat proteins, it seems more likely that initial interactions are with enzymes that catalyze ubiquitin addition. Generally for ubiquitination, although the precise biochemical or genetic requirements are incompletely understood, it is clear that the E2 and E3 classes of enzymes participate both in recognizing the substrate protein and in the subsequent ubiquitin transfer (Hochstrasser, 1996). For Ste3p ubiquitination, there are several E2 and E3 candidates. A component of the E3 complex could be the hect-domain protein Rsp5p. Though involvement in pheromone receptor endocytosis remains to be determined, RSP5 is required for the ubiquitination and endocytic turnover of several different yeast plasma membrane proteins (Hein et al., 1995; Lucero and Lagunas, 1997). In terms of participating E2 functions, the redundant enzymes Ubc4p and Ubc5p, as well as Ubc1p are likely involved as they have been implicated in the ubiquitination and endocytosis for a variety yeast plasma membrane proteins (Kolling and Hollenberg, 1994; Egner and Kuchler, 1996; Hicke and Riezman, 1996; Roth and Davis, 1996). These E2 and E3 activities, however, can not be exclusively devoted to plasma membrane protein turnover. Rsp5p, Ubc4p, and Ubc5p also are required for the ubiquitination and proteosomal turnover of a variety of short-lived cytoplasmic and nuclear proteins (Hochstrasser, 1996). In addition, Rsp5p recently has been implicated as a participant in the mitochondrial import process (Zolladek et al., 1997). Thus, these enzymes participate in a number of quite distinct cellular processes occurring at a number of distinct cellular sites. Specificity for membrane proteins will likely also involve the inclusion of other, presently unknown factors into these E2 and E3 complexes to direct both cellular localization and proper substrate recognition. A goal for the future, then, is to identify the specific factors that bind to and decode this new class of ubiquitination/endocytosis signal.

We thank T. Stevens for the PGK-specific antiserum and J. Haber for the HA-tagged GAL1-PMA1 plasmid.

This work was supported by a research grant from the National Science Foundation (MCB 95-06839). D.M. Sullivan was supported by a postdoctoral research fellowship from the National Institutes of Health (T32- GM08420).

CTD

cytoplasmically disposed, COOH-terminal domain

ORF

open reading frame