The Rsp5 E3 Ligase Mediates Turnover of Low Affinity Phosphate Transporters in Saccharomyces cerevisiae*

In an effort to identify novel components of the PHO regulon in Saccharomyces cerevisiae, we have isolated and characterized suppressors of the Pho– phenotype associated with deletion of the Pho4 transcriptional activator. Here we report that either a defective form of the Rsp5 E3 ubiquitin ligase or deletion of the End3 component of the endocytic pathway restores growth of the pho4Δ mutant in the presence of limiting inorganic phosphate (Pi). The spa1-1 suppressor allele of RSP5 encodes a phenylalanine-to-valine replacement at position 748 (F748V) within the catalytic HECT domain of Rsp5. Consistent with suppression due to impaired ubiquitin ligase activity, the heat-sensitive growth defect of the spa1-1 mutant is suppressed either by overexpression of ubiquitin or by osmotic stabilization. Western blot analyses revealed that the cellular levels of the Pho87 and Pho91 low affinity Pi are markedly increased in the spa1-1 mutant, yet Pho84 high affinity Pi transporter levels are unaffected. Furthermore, Pho87 and Pho91 are ubiquitinated in vivo in an Rsp5-dependent manner, and the Pho+ phenotype of the spa1-1 suppressor is dependent upon Pho87 and Pho91. We conclude that turnover of the low affinity Pi transporters is initiated by Rsp5-mediated ubiquitination followed by internalization and degradation by the endocytic pathway.

In an effort to identify novel components of the PHO regulon in Saccharomyces cerevisiae, we have isolated and characterized suppressors of the Pho ؊ phenotype associated with deletion of the Pho4 transcriptional activator. Here we report that either a defective form of the Rsp5 E3 ubiquitin ligase or deletion of the End3 component of the endocytic pathway restores growth of the pho4⌬ mutant in the presence of limiting inorganic phosphate (P i ). The spa1-1 suppressor allele of RSP5 encodes a phenylalanine-to-valine replacement at position 748 (F748V) within the catalytic HECT domain of Rsp5. Consistent with suppression due to impaired ubiquitin ligase activity, the heat-sensitive growth defect of the spa1-1 mutant is suppressed either by overexpression of ubiquitin or by osmotic stabilization. Western blot analyses revealed that the cellular levels of the Pho87 and Pho91 low affinity P i are markedly increased in the spa1-1 mutant, yet Pho84 high affinity P i transporter levels are unaffected. Furthermore, Pho87 and Pho91 are ubiquitinated in vivo in an Rsp5-dependent manner, and the Pho ؉ phenotype of the spa1-1 suppressor is dependent upon Pho87 and Pho91. We conclude that turnover of the low affinity P i transporters is initiated by Rsp5-mediated ubiquitination followed by internalization and degradation by the endocytic pathway.
The yeast Saccharomyces cerevisiae has evolved an elaborate system to sense, acquire, and store inorganic phosphate (P i ) in response to its availability in the extracellular environment (for review, see Refs. 1 and 2). The PHO system consists of (i) the Pho3, Pho5, Pho11, and Pho12 acid phosphatases that are localized to the periplasmic space, (ii) the Pho8 and Pho13 alkaline phosphatases that are localized to the vacuole and periplasm, respectively, (iii) the high affinity plasma membrane P i transporters Pho84 and Pho89, which are regulated in response to P i availability, and the low affinity, constitutively expressed P i transporters Pho87, Pho90, and Pho91, (iv) Git1, a transporter that scavenges glycerophosphoinositol derived from secreted phosphatidylinositol, thereby replenishing inositol, P i , and glycerol (3), and (v) the PHM proteins that are involved in the synthesis and breakdown of polyphosphate, a storage form of P i in the vacuole (for review, see Ref. 4).
The PHO regulon is responsible for scavenging P i and has been most extensively studied with regard to regulation of PHO5 expression. The Pho84 transporter and the Pho80 -Pho85 cyclin/cyclin-dependent kinase are negative regulators of PHO5 transcription, whereas the Pho81 inhibitor of Pho80 -Pho85 and the Pho4 transcriptional activator are positive regulators. In the presence of high external concentrations of P i , Pho85 phosphorylates Pho4, resulting in Pho4 nuclear export via the Msn5 exportin and cytoplasmic retention such that PHO5 is repressed (5). Conversely, when P i concentrations are low, Pho81 inhibits Pho85 activity. Pho4 remains unphosphorylated and has high affinity for the Pse1 importin, resulting in its nuclear retention where the Pho2-Pho4 complex induces PHO5 transcription (5). The high affinity transporter genes, PHO84 and PHO89, are also components of the PHO regulon and are controlled in a manner similar to PHO5. In contrast, expression of the low affinity transporter genes, PHO87, PHO90, and PHO91, are independent of Pho4.
Although the downstream events in the PHO signal transduction pathway have been well characterized, the upstream components that sense P i concentrations and transduce that information are not well understood. Indeed, the direct sensor of P i has not been identified. Recently, it was proposed that the PHO regulon is controlled by two different regulatory signals, one sensing internal P i concentrations by an unidentified protein, the other sensing external P i concentrations by the low affinity P i transporters (6). In contrast to pho2⌬, pho4⌬, or pho81⌬ mutants that fail to activate PHO genes in response to P i depletion, pho87⌬, pho90⌬, or pho91⌬ deletion mutations result in constitutive derepression of PHO genes (6, 7). Furthermore, pho87⌬, pho90⌬, and pho91⌬ are hypostatic to pho2⌬, pho4⌬, and pho81⌬ (6). These results place the low affinity P i transporters in the PHO signal transduction pathway, upstream of Pho2, Pho4, and Pho81. As such, Pho87, Pho90, and Pho91 appear to function not only as P i transporters but as signal transducers that link the extracellular environment with the regulation of PHO gene expression (6).
Recently, it was reported that the protein kinase A signal transduction pathway is required for degradation of Pho84 (8), although there are no reports describing turnover of the low affinity P i transporters. A number of cell surface proteins, including transporters, permeases, and signaling receptors, are known to be down-regulated by ubiquitin-mediated endocytosis. In these cases the cytoplasmic C-terminal region of the plasma membrane protein is tagged with ubiquitin, internalized by endocytosis, and directed to the yeast vacuole or mammalian lysosome for degradation (for review, see Refs. 9 -11). This ubiquitin-mediated protein degradation pathway is distinct from the ubiquitin proteasome pathway (11). As examples of the ubiquitin-endosome pathway, the yeast Rsp5 E3 ubiquitin ligase is involved in turnover of the Gap1 general amino acid permease (12,13), the Fur4 uracil permease (14 -16), the Tat2 tryptophan permease (17), the Gal2 galactose transporter (18, 19), and the Mal61 maltose transporter (20,21). Accordingly, ubiquitin-mediated endocytosis appears to be a general mechanism to respond to changes in nutrient availability.
In an effort to identify proteins that regulate phosphate metabolism, we sought suppressors that would bypass the Pho4 requirement for cell growth in limiting P i medium. We anticipated finding (i) proteins that act downstream of Pho4 either to repress Pho4-regulated genes or to mediate Pho4 activation and (ii) factors that affect P i uptake independent of Pho4. Suppressors of the former class, which includes components of the RNA polymerase II mediator complex were found and will be described elsewhere. Here we describe suppressors of the latter class that includes the Rsp5 E3 ubiquitin ligase and the End3 component of the endocytic pathway. Our results identify the low affinity phosphate transporters Pho87 and Pho91 as novel targets of the Rsp5 endocytic pathway and define an important role for these proteins in sensing P i availability.
Plasmids   (25) using plasmid pDp207, and plasmid pRS425 was used as the vector control. Growth Media-Rich (YPD), 5 synthetic complete (SC), and omission media (ϪUra and ϪLeu) were prepared according to standard procedures (26). Media of defined P i concentrations were prepared either from YPD medium by precipitation of inorganic phosphate followed by add-back of KH 2 PO 4 , as described previously (27), or from SC medium using yeast nitrogen base without phosphate (BIO101) to which KH 2 PO 4 was added at the indicated final P i concentration. Where indicated, ϪLeu medium was supplemented with 0.05 mM CuSO 4 , and SC medium was supplemented with 1 M sorbitol.
Isolation of spa1-1 by Gap-repair-The spa1-1 allele was cloned from strain YMH655 by gap-repair (22). The RSP5-URA3-CEN plasmid pN1688 was digested with BstEII, creating a gap in the RSP5 open reading frame, and introduced into YMH655 (spa1-1). Ura ϩ transformants were selected and screened for retention of the Pho ϩ and Tsm Ϫ suppressor phenotypes. Plasmid DNA was retrieved, screened by restriction analysis for recovery of gapped DNA, and reintroduced into YMH655. Whereas undigested pN1688 complemented YMH655, restoring the Pho Ϫ and Tsm ϩ phenotypes, transformants derived from the gap-repaired plasmid (pN1704), retained the Pho ϩ and Tsm Ϫ suppressor phenotypes.
Extract Preparation and Western Blot Analyses-Yeast whole cell extracts were prepared as described previously (28). Briefly, 2 A 600 units of yeast cells were collected by centrifugation and lysed by resuspension in 100 l of 1.85 M NaOH plus 7% 2-mercaptoethanol. Proteins were precipitated by the addition of 100 l of 50% trichloroacetic acid and collected by centrifugation at 13,000 rpm (Sorvall Biofuge) for 5 min. The pellet was rinsed in 0.5 ml of 1 M Tris base without resuspension, dissolved in 100 l of 2ϫ sample buffer (4% SDS, 0.1 M Tris HCl, pH 6.8, 4 mM EDTA, 20% glycerol, 2% 2-mercaptoethanol), and dissociated by heating for 10 min at 37°C. Protein samples corresponding to 0.1-0.2 A 600 units were separated in 8% SDSpolyacrylamide gels, and Western blot analyses were performed as described previously (29). Pho87-3xHA and Pho91-3xHA were assayed by incubation using a 1:3000 dilution of anti-HA antibody followed by incubation with horseradish peroxidase-conjugated anti-mouse antibody. Control anti-Ess1 antibody (a gift from S. Hanes) was used at 1:10,000 dilution. Immunoprecipitated Pho87-3HA and Pho91-3HA proteins were resolved by SDS-PAGE (8%), transferred to nitrocellulose, and assayed by Western blot using 1:3000 dilution of anti-ubiquitin antibody (Sigma). Control anti-Rpa1 antibody (a gift from S. Brill) was used at a 1:10,000 dilution. Antigen-antibody complexes were detected using the western Lightning Chemiluminescence Reagent (PerkinElmer Life Sciences).
Immunoprecipitation of Membrane Proteins-Proteins were extracted and immunoprecipitated as described previously (30) with minor modifications. Briefly, 1 A 600 unit of cells was lysed by incubation for 10 min on ice with 40 l of 1.85 M NaOH plus 2% 2-mercaptoethanol. Proteins were precipitated by the addition of 40 1 of 50% trichloroacetic acid, collected by centrifugation for 5 min at 13,000 rpm. Samples were resuspended in 30 l of 2ϫ sample buffer (without 2-mercaptoethanol or bromphenol blue) plus 20 l of 1 M Tris base followed by heating for 10 min at 37°C. TNET buffer (0.6 ml, 50 mM Tris-HCl, pH 7.4, 150 mM NaCl, 5 mM EDTA, 1% Triton X-100) was added, and the insoluble material was removed by centrifugation at 13,000 rpm for 30 min. The supernatant was incubated overnight at 4°C with 10 l of anti-HA antibody (Sigma) plus 40 l of protein A-Sepharose beads (Sigma). Immunoprecipitates were washed five times with TNET buffer, and proteins were eluted in 30 l of 2ϫ sample buffer (4% SDS, 0.1 M Tris-HCl, pH 6.8, 4 mM EDTA, 20% glycerol, 2% 2-mercaptoethanol, 0.02% bromphenol blue) for 10 min at room temperature. Proteins were immunoprecipitated using anti-ubiquitin (1:50) and anti-HA (1:50) antibodies.
The spa1-1 Suppressor Is Allelic to RSP5-The Tsm Ϫ phenotype associated with the spa1-1 suppressor was used to clone the wild type gene. A YCp50 genomic library (31) was introduced into strain YMH655, and transformants were selected on ϪUra medium at 37°C. A single Tsm ϩ Pho Ϫ transformant was recovered from 23,500 transformants. Plasmid DNA was isolated, amplified, and reintroduced into YMH655, confirming plasmid dependence of the Tsm ϩ and Pho Ϫ phenotypes. DNA sequence analysis from either end of the vector insert identified a 15-kilobase fragment of chromosome V encompassing five open reading frames, including RSP5, which encodes an essential ubiquitin E3 ligases that has been implicated in diverse biological processes in S. cerevisiae. Plasmid DNA (pN1688) carrying RSP5 and none of the flanking open reading frames fully complemented spa1-1 (data not shown). Genetic linkage analysis confirmed that spa1-1 is allelic to RSP5 (see "Experimental Procedures").
Characterization of the spa1-1 Suppressor Mutation-Rsp5 is a member of the Nedd4 class of E3 ubiquitin ligases (for review, see Refs. 11, 32, and 33). The structure of Rsp5 includes a C2 domain, implicated in membrane association, two WW domains that mediate enzyme-substrate recognition, and a catalytic HECT domain that includes the active site cysteine. To address how the spa1-1 allele of RSP5 suppresses pho4⌬, we isolated and sequenced the spa1-1 allele. The spa1-1 allele was cloned from strain YMH655 by gaprepair. DNA sequence analysis of the entire open reading frame from plasmid pN1704 identified a single base pair substitution encoding a phenylalanine-to-valine replacement at position 748 (F748V), located within the HECT domain of the Rsp5 C terminus (Fig. 2A).
The HECT domain alone is sufficient for E3 ligase catalytic activity (24). A crystal structure of the human E6AP HECT domain is available, revealing a larger N-terminal lobe and a smaller C-terminal lobe (34). Phe-748 is located within the smaller C-terminal lobe, in the same hydrophobic environment as the catalytically impaired L733S replacement encoded by the rsp5-1 allele (24). The smaller C-terminal lobe also includes the active site cysteine residue (Cys-777) that is essential for E3 ligase activity and cell viability (35,36). To test whether loss of Rsp5 E3 ligase activity is responsible for suppression of pho4⌬, plasmid-borne rsp5 alleles encoding the L733S and C777A replacements were introduced into the spa1-1 mutant and scored for complementation of the Tsm Ϫ phenotype. In contrast to the wild type RSP5 plasmid, neither of the catalytically defective rsp5 plasmids complemented the Tsm Ϫ phenotype of spa1-1 (Fig.  2B). Thus, diminished Rsp5 E3 ligase activity compensates for the Pho Ϫ phenotype associated with loss of the Pho4 transcriptional activator.
Suppression of spa1-1 by Overexpression of Ubiquitin or by Osmotic Stabilization-The growth phenotypes of several rsp5 ubiquitin ligase-defective alleles are suppressed by elevated expression of UBI4, the gene encoding ubiquitin (37). If the spa1-1 suppressor affects a ubiquitin-mediated process, then the Tsm Ϫ phenotype of spa1-1 might be suppressed by UBI4 overexpression. To test this possibility, we introduced plasmid DNA (pM1761) carrying the UBI4 gene under control of the copper-inducible CUP1 promoter (CUP1p-UBI4) into wild type, pho4⌬, and pho4⌬ spa1-1 strains. The LEU2 vector pRS425 was included as a control. Transformants were streaked on ϪLeu medium and ϪLeu plus 0.05 mM CuSO 4 and incubated at 37°C. The Tsm Ϫ phenotype of the spa1-1 mutant was weakly suppressed by CUP1p-UBI4 in the absence of copper and markedly suppressed in the presence of copper, whereas no effect of CUP1p-UBI4 was observed in the wild type or pho4⌬ strains (Fig. 3A). Weak suppression in the absence of copper is presumably a consequence of leaky expression of UBI4 from the CUP1p promoter (38). Thus, overexpression of ubiquitin effectively compensates for diminished Rsp5 E3 ligase activity.
The Tsm Ϫ phenotype of ligase-defective rsp5 mutants is also suppressed by osmotic stabilization in the presence of 1 M sorbitol (39). The Tsm Ϫ phenotype of the spa1-1 suppressor is likewise suppressed by 1 M sorbitol (Fig. 3B). Taken together, these results indicate that loss of Rsp5 E3 ligase activity bypasses the normal Pho4 requirement for growth in low phosphate medium. We conclude that Rsp5-mediated protein ubiquitination plays a negative role in regulating the response to P i availability.

Rsp5-mediated Ubiquitination of Pho87 and Pho91
The Endocytic Pathway Regulates Phosphate Metabolism-The End3-Pan1-Sla1 complex is essential for the internalization step of endocytosis (40), and End3 has been implicated in Rsp5-mediated endocytosis and vacuolar degradation of plasma membrane proteins (20,41). To test whether the endosomal pathway is involved in phosphate metabolism, we asked whether an end3⌬ deletion would suppress pho4⌬ in a manner similar to the spa1-1 suppressor. We deleted the END3 gene in the WT, pho4⌬, and pho4⌬ spa1-1 backgrounds and scored the resulting strains for growth in the presence of high and low concentrations of P i . Consistent with earlier results (40), the end3⌬ deletion was lethal at 30°C but viable at 24°C, albeit exhibiting a slow-growth (Slg Ϫ ) phenotype in the presence of 1 mM P i . Accordingly, all growth phenotypes were scored at 24°C. Results are shown in Fig. 4. First, the Pho Ϫ phenotype of the pho4⌬ mutant is suppressed by end3⌬ (cf. sectors 3 and 4 on 25 M P i ). Presumably, end3⌬ suppression of pho4⌬ is even more pronounced than it appears due to the Slg Ϫ phenotype associated with end3⌬ even in the presence of high P i at 24°C (cf. sectors 2, 4, and 6 with sectors 1, 3, and 5 on 1 mM P i ). Second, suppression of pho4⌬ by spa1-1 and end3⌬ are not additive, as the pho4⌬ spa1-1 end3⌬ triple mutant exhibited the same growth phenotype on low P i medium as either the pho4⌬ spa1-1 or pho4⌬ end3⌬ double mutants (cf. sectors 4, 5, and 6 with sector 3 on 25 M P i ), consistent with the idea that Rsp5 and End3 function in the same pathway. Interestingly, the Slg Ϫ phenotype of the end3⌬ mutant on high P i medium was suppressed by P i limitation (cf. sectors 2 on 1 mM and 25 M P i ). This observation suggests that the Slg Ϫ phenotype of all end3⌬ mutants in the presence of 1 mM P i might be an adverse consequence of the inability of end3⌬ mutants to down-regulate the low affinity P i transporters in the presence of excess P i (see below). Taken together, these results define end3⌬ as a suppressor of the pho4⌬ deletion and implicate the End3 endosomal pathway in regulation of P i metabolism.
Loss of Rps5 Stabilizes the Low Affinity P i Transporters-What is the proteolytic substrate(s) of the Rsp5 E3 ligase responsible for pho4⌬ suppression? Given that Rsp5 ubiquitinates and stimulates the turnover of several plasma membrane proteins, including permeases, transporters, and receptors (for review, see Ref. 11), we asked whether diminished Rsp5 activity might overcome the Pho Ϫ phenotype of the pho4⌬ mutant by stabilizing plasma membrane P i transporters.
To determine whether the Pho ϩ phenotype of the pho4⌬ spa1-1 revertant is due to stabilization of P i transporters, we assayed the levels of the low affinity Pho87 and Pho91 transporters and the high affinity Pho84 transporter from wild type, pho4⌬, and pho4⌬ spa1-1 strains. Transporter levels were assayed by Western blot using cell extracts from isogenic strains that had been 3xHA epitope-tagged at the normal chromosomal loci of PHO84, PHO87, and PHO91 (repeated efforts to generate 3xHA-tagged derivatives of the Pho89 high affinity and Pho90 low affinity P i transporters were unsuccessful). Results are shown in Fig. 5. The Pho87 and Pho91 transporters were barely detectable by Western blot in either the wild type or pho4⌬ strains, whereas the steady-state levels of both proteins were clearly elevated in the spa1-1 strains (panel A). By contrast, Pho84 levels were detectable in the wild type strain and induced upon phosphate depletion (panel B). Consistent with the Pho4 dependence of PHO84 expression (42), the Pho84 protein was undetectable in the pho4⌬ mutant under either inducing or repressing conditions, and this defect was not suppressed by spa1-1. Thus, the Pho87 and Pho91 low affinity P i transporters accumulate in the absence of normal Rsp5 E3 ligase activity. In contrast, the high affinity Pho84 transporter is uninducible in the absence of Pho4 and is unaffected by loss of Rsp5. These results implicate Rsp5-mediated ubiquitin conju- , and pho4⌬ spa1-1 (YMH655) strains were transformed with the high copy number CUP1p-UBI4 LEU2 plasmid (pM1761) or vector alone (pRS425). 10-Fold serial dilutions of each strain were spotted onto either ϪLeu ϪCuSO 4 medium or onto ϪLeu ϩCuSO 4 medium to induce UBI4 expression. Plates were photographed after 2 (30°C) or 3 (37°C) days of incubation. B, the Tsm Ϫ phenotype of the pho4⌬ spa1-1 mutant is suppressed by osmotic stabilization in the presence of 1 M sorbitol. 10-Fold serial dilutions of wild type (W303-1B), pho4⌬ (EY131), and pho4⌬ spa1-1 (YMH655) strains were spotted onto SC medium or SC medium containing 1 M sorbitol. Plates were photographed after 2 (30°C) or 3 (37°C) days of incubation. The synthetic media depicted in panels A and B contain normal (high) P i levels; accordingly, none of the growth phenotypes can be attributed to P i auxotrophy.  pho4⌬ end3⌬ (YMH843), pho4⌬ spa1-1 (YMH655), and pho4⌬ spa1-1 end3⌬ (YMH844) strains were streaked onto SC medium containing the indicated concentrations of P i . SC medium at 24°C was used rather than YPD medium at 30°C because end3⌬ mutants have reduced fitness in rich medium and exhibit a severe Slg Ϫ defect at 30°C (40,50). gation in turnover of the Pho87 and Pho91 low affinity P i transporters in yeast.
The Pho87 and Pho91 Transporters Are Ubiquitinated in an Rsp5-dependent Manner-To determine whether the Pho87 and Pho91 transporters are directly ubiquitinated, we immunoprecipitated (IP) the Pho87-3xHA and Pho91-3xHA proteins from cell extracts of the WT, pho4⌬, and pho4⌬ spa1-1 strains using ␣-HA antibody followed by immunoblot (IB) analysis using ␣-ubiquitin (␣-Ub) antibody. As shown in Fig. 6, immunoprecipitated Pho87-3xHA and Pho91-3xHA from the wild type strains were readily detected using ␣-Ub antibody (lanes 4  and 7). The ubiquitinated levels of both proteins were lower in the pho4⌬ mutants ( lanes 5 and 8), implicating Pho4 in ubiquitination of the low affinity P i transporters, although the mechanism of this effect is unknown. Importantly, the levels of ubiquitinated Pho87-3xHA and Pho91-3xHA were also diminished in the pho4⌬ spa1-1 strains (lanes 6 and 9) even though the levels of these two proteins are higher in this background (Fig. 5A, lanes 3 and 6). Immunoblot analysis (␣-HA) of ␣-HA-immunoprecipitated protein confirmed that the recombinant proteins are intact (Fig. 6, second panel). Analysis of the same strains that had not been 3x-HA tagged confirmed that the observed bands do indeed correspond to Pho87 and Pho91 (Fig. 6, lanes 1-3). We also performed the reciprocal experiment, first immunoprecipitating ubiquitinated proteins (␣-Ub) from cells extracts, followed by immunodetection (␣-HA) of Pho87-3xHA and Pho91-3xHA. The results confirmed that Pho87 and Pho91 are indeed ubiquitinated (Fig. 6, third panel).
Taken together, the results in Figs. 5 and 6 demonstrate that the Pho87 and Pho91 low affinity P i transporters are ubiquitinated in an Rsp5-dependent manner that affects the steady-state levels of both proteins.
The spa1-1 Suppressor Phenotype Is Pho87-and Pho91dependent-To confirm that stabilization of the Pho87 and/or Pho91 low affinity P i transporters associated with diminished Rsp5 E3 ligase activity is responsible for reversion of the pho4⌬ mutant phenotype, we deleted the PHO87 and PHO91 genes. If the Pho ϩ suppressor phenotype of the pho4⌬ spa1-1 strain is dependent upon the low affinity P i transporters, then deletion of the encoding genes should at least partially restore the Pho Ϫ mutant phenotype. Indeed, deletion of either PHO87 or PHO91 resulted in loss of the Pho ϩ phenotype of the pho4⌬ spa1-1 suppressor strain (Fig. 7, cf. sectors 3, 5, and 6 for both the upper FIGURE 6. The Pho87 and Pho91 low affinity P i transporters are ubiquitinated in an Rsp5-dependent manner. Cell extracts from the strains described in Fig. 5A were immunoprecipitated (IP) with either ␣-HA or ␣-Ub antibody followed by immunoblot (IB) analysis using either ␣-Ub or ␣-HA antibody as indicated. The bottom panel depicts an immunoblot probed with antibody to Rpa1 (␣-Rpa1) that serves as a loading control. FIGURE 7. The spa1-1 Pho ؉ suppressor phenotype is dependent upon the Pho87 and Pho91 low affinity P i transporters. Isogenic wild type (W303-1B), pho87⌬ (YMH957), pho4⌬ (EY131), pho4⌬ pho87⌬ (YMH958), pho4⌬ spa1-1 (YMH655), pho4⌬ spa1-1 pho87⌬ (YMH959), pho91⌬ (YMH960), pho4⌬ pho91⌬ (YMH961), and pho4⌬ spa1-1 pho91⌬ (YMH962) strains were streaked onto SC medium containing the indicated concentrations of P i and incubated for 3 days at 30°C. (pho87⌬) and lower (pho91⌬) panels on 5 M P i medium). This effect is not due to impaired growth associated with either pho87⌬ or pho91⌬ alone as neither deletion impaired growth on low phosphate medium in the wild type background to the same extent as in the suppressor strain (cf. sectors 1, 2, and 6). These results confirm that suppression of the Pho Ϫ phenotype by defective Rsp5 requires the Pho87 and Pho91 low affinity P i transporters. Repeated attempts to construct double pho87⌬ pho91⌬ deletions in the pho4⌬ spa1-1 strain failed, even though both genes could readily be deleted in the wild type and pho4⌬ strains. It is, therefore, likely that loss of both low affinity P i transporters is lethal in the absence of functional Rsp5. These results underscore the functional relationship between the Rsp5 E3 ligase and the low affinity P i transporters and are consistent with our conclusion that turnover of the low affinity P i transporters is mediated by the Rsp5-endosome pathway.

DISCUSSION
In this study we have uncovered a role for Rsp5-mediated substrate ubiquitination in regulation of the Pho87 and Pho91 low affinity membrane P i transporters of S. cerevisiae. First, the Pho Ϫ phenotype associated with the absence of the Pho4 activator of the PHO regulon is suppressed by mutation in the Rsp5 E3 ubiquitin ligase. Suppression is presumably due to diminished Rsp5 ligase activity because the spa1-1 suppressor encodes a single amino acid replacement within the Rsp5 catalytic HECT domain. Also, two different rsp5 alleles encoding catalytically defective Rsp5 proteins (L733S and C777A) fail to complement the spa1-1 growth defect. Second, the spa1-1 Tsm Ϫ phenotype is suppressed by overexpression of ubiquitin, shown previously to suppress catalytically defective rsp5 mutants (39). In addition, spa1-1 is phenotypically suppressed by 1 M sorbitol, which is also known to suppress rsp5 catalytic mutants (37). Third, deletion of the End3 component of the endosome, which functions in Rsp5-mediated turnover of plasma membrane proteins in yeast (43), suppresses the pho4⌬ deletion. Fourth, the levels of the Pho87 and Pho91 low affinity P i transporters are clearly elevated in the spa1-1 mutants. Fifth, Pho87 and Pho91 are ubiquitinated, and ubiquitination of these proteins is diminished in the spa1-1 background. Finally, the Pho ϩ revertant phenotypes associated with loss of Rsp5 E3 ligase activity is dependent upon the Pho87 and Pho91 transporters. To our knowledge, this is the first study to implicate the ubiquitin-endosome pathway in turnover of plasma membrane P i transporters.
In contrast to its effects on the low affinity P i transporters, the spa1-1 allele of RSP5 is without effect on the high affinity Pho84 P i transporter. PHO84 is a component of the PHO regulon, whose components are regulated by the Pho4 transcription factor in response to P i availability. Accordingly, Pho84 is uninducible in response to low P i in the pho4⌬ deletion mutant, and this defect is not overcome by the spa1-1 suppressor (Fig. 5B).
Moreover, Pho84 appears to be degraded by a pathway different from that of Pho87 and Pho91. Add-back of P i to P i -starved cells normally results in rapid degradation of the Pho84 transporter. However, inhibition of protein kinase A was reported to stabilize Pho84, thereby defining a role for the protein kinase A signal transduction pathway in Pho84 turnover (8). Nonethe-less, Pho84 degradation occurs in the vacuole (44) rather than the proteasome, but the pathway by which Pho84 is targeted to the vacuole remains to be elucidated.
Similar to the effect of spa1-1 on Pho87 and Pho91, the catalytically impaired npi1 allele of RSP5 impairs Gap1 ubiquitination and turnover (12,13). Furthermore, Gap1 turnover is regulated by NH 4 ϩ availability in a manner dependent upon Rsp5 activity. In the case of Fur4, Rsp5-catalyzed ubiquitination promotes Fur4 internalization, entry into the endocytic pathway, and subsequent degradation in the vacuole (14 -16). Although this process occurs constitutively, Fur4 turnover is affected by changing metabolic conditions. Turnover of the Mal61 transporter, like Gap1, is dependent upon the Rps5 ligase and the End3 component of the endosome and is induced by nitrogen starvation in the presence of a fermentable carbon source (20,21). Pho87 and Pho91 can now be added to the growing list of membrane proteins that are targeted for endocytosis and vacuolar degradation by Rsp5-catalyzed ubiquitination in response to changes in nutrient availability.
It is not clear how Rsp5 identifies its target proteins within the plasma membrane. The WW domains of Rsp5 are involved in substrate recognition, but not all Rsp5 target proteins include consensus WW binding sites. In the case of Fur4, phosphorylation of a PEST-like sequence is required for ubiquitination and subsequent internalization (14,45,46). Ubiquitination and endocytosis of the Ste2 ␣-factor receptor is also dependent upon phosphorylation (47). Phosphorylation-dependent ubiquitination is not unique to Rsp5, as the WW domain of Nedd4, the mammalian homolog of Rsp5, also binds phosphorylated ligands (48). These results raise the interesting possibility that Rsp5 recognition of the low affinity P i transporters might be regulated by phosphorylation. Consistent with this idea, all three low affinity P i transporters display PEST-like sequences that include potential phosphorylation sites (Fig. 8). One possible scenario is that in the presence of relatively high P i concentrations, the low affinity P i transporters are phosphorylated, resulting in Rsp5-mediated ubiquitination and subsequent internalization. Conversely, when P i is limiting, the low affinity P i transporters would be hypophosphorylated and not subject to ubiquitination and endosomal turnover.
How does stabilization of the low affinity P i transporters suppress the Pho Ϫ phenotype of a pho4⌬ mutant? Given that the K m of the low affinity system (K m ϭ 770 M) is 2 orders of FIGURE 8. The low affinity P i transporters include PEST-like sequences. The indicated Pho87, Pho90, and Pho91 amino acid sequences (single letter code) were identified by PESTFIND. Potential phosphorylation sites (reverse font) conforming to protein kinase C and cAMP-dependent protein kinase (Pho87) or casein kinase II (Pho90, Pho91) consensus sequences were identified using MOTIFSCAN. PEST sequences have been identified as phosphorylation targets that are required for proper ubiquitination and subsequent turnover of yeast membrane proteins (see "Discussion"). FEBRUARY 29, 2008 • VOLUME 283 • NUMBER 9 magnitude greater than the K m of the high affinity system (K m ϭ 8 M) (49), it seems unlikely that the transporter activities of Pho87 and Pho91 account for growth on limiting P i medium. It seems more likely that the low affinity P i transporters function as signal transducers that detect external P i concentrations and transmit this information to the intracellular PHO pathway. This possibility is supported by the recent report that the low affinity P i transporters regulate PHO gene expression independently of internal P i concentrations (6). Our results are consistent with the suggestion that Pho87, Pho90, and Pho91 function as both P i transporters and sensors. Furthermore, this external P i signal transduction pathway can function independently of the Pho4 transcriptional activator, as our results demonstrate that stabilization of Pho87 and Pho91 bypass Pho4 for growth on limiting P i . It remains to be determined how the low affinity P i transporters function as signal transducers and how Rsp5 responds to changes in P i concentrations.