|Home | About | Journals | Submit | Contact Us | Français|
Intracellular sorting of the general amino acid permease (Gap1p) in Saccharomyces cerevisiae depends on availability of amino acids such that at low amino acid concentrations Gap1p is sorted to the plasma membrane, whereas at high concentrations Gap1p is sorted to the vacuole. In a genome-wide screen for mutations that affect Gap1p sorting we identified deletions in a subset of components of the ESCRT (endosomal sorting complex required for transport) complex, which is required for formation of the multivesicular endosome (MVE). Gap1p-GFP is delivered to the vacuolar interior by the MVE pathway in wild-type cells, but when formation of the MVE is blocked by mutation, Gap1p-GFP efficiently cycles from this compartment to the plasma membrane, resulting in unusually high permease activity at the cell surface. Importantly, cycling of Gap1p-GFP to the plasma membrane is blocked by high amino acid concentrations, defining recycling from the endosome as a major step in Gap1p trafficking under physiological control. Mutations in LST4 and LST7 genes, previously identified for their role in Gap1p sorting, similarly block MVE to plasma membrane trafficking of Gap1p. However, mutations in other recycling complexes such as the retromer had no significant effect on the intracellular sorting of Gap1p, suggesting that Gap1p follows a genetically distinct pathway for recycling. We previously found that Gap1p sorting from the Golgi to the endosome requires ubiquitination of Gap1p by an Rsp5p ubiquitin ligase complex, but amino acid abundance does not appear to significantly alter the accumulation of polyubiquitinated Gap1p. Thus the role of ubiquitination appears to be a signal for delivery of Gap1p to the MVE, whereas amino acid abundance appears to control the cycling of Gap1p from the MVE to the plasma membrane.
The family of amino acid permeases expressed in Saccharomyces cerevisiae fall into two different classes with respect to their regulation. Most of the permeases are expressed constitutively and import specific amino acids or chemically related amino acids. A second class of permeases are most highly expressed under conditions of nitrogen limitation and are thought to scavenge amino acids for their use as a source of nitrogen (Magasanik and Kaiser, 2002 ). The major nitrogen-scavenging permeases are Gap1p, which transports all naturally occurring amino acids with a high capacity (Grenson et al., 1970 ; Jauniaux and Grenson, 1990 ), and Put4p, which is specific for proline (Vandenbol et al., 1989 ). In part, the activity of Gap1p and Put4p is determined by an intracellular sorting decision, which depends on the nitrogen source in the growth medium. When yeast cells are grown on medium that lacks amino acids, Gap1p and Put4p are delivered to the plasma membrane where they are active for amino acid uptake. However, when cells are grown on a medium rich in an amino acid such as glutamate, Gap1p and Put4p are sorted to the vacuole for degradation (Roberg et al., 1997a ; Chen and Kaiser, 2002 ).
Mutations that affect the intracellular sorting of Gap1p can be classified into two general types. The first type includes mutations that cause constitutive sorting of high levels of Gap1p to the plasma membrane. This category includes in the genes RSP5, BUL1, BUL2, and DOA4, all of which either partially or completely block the intracellular ubiquitination of Gap1p, which serves as a tag for sorting to the vacuole (Helliwell et al., 2001 ; Soetens et al., 2001 ; Springael et al., 1999 ). Ubiquitination constitutes a common signal for endocytic internalization of a variety of plasma membrane proteins (reviewed by Hicke, 1997 ; Horák, 2003 ). The second category includes mutations that cause constitutive sorting of Gap1p to the vacuole, regardless of the nitrogen source in the growth medium. Among the genes belonging to this category are the genes LST4, LST7, and LST8, which were initially identified because of their lethality in combination with a thermosensitive allele of SEC13 (Roberg et al., 1997b ). LST8 encodes a positively acting component of the TOR pathway that affects Gap1p sorting by negatively regulating the transcription factors Rtg1/3p and Gln3p, thereby limiting the synthesis of α-ketoglutarate, glutamate, and glutamine (Chen and Kaiser, 2003 ). The roles of LST4 and LST7 have not been elucidated. A fundamental relationship between mutations of the two types is that, when a mutation that blocks ubiquitination and causes constitutive sorting to the plasma membrane is combined with a mutation that causes constitutive sorting to the vacuole, the double mutants invariably show constitutive sorting to the plasma membrane. This finding has led to the hypothesis that ubiquitination of Gap1p precedes sorting of Gap1p to a compartment in which the mutations of the second type can exert their effect (Helliwell et al., 2001 ).
A variety of studies have revealed that sorting of most plasma membrane proteins for degradation in the vacuolar lumen occurs through the maturation of the late endosome or prevacuolar compartment into multivesicular endosomes (MVEs; reviewed by Katzmann et al., 2002 ; Raiborg et al., 2003 ; Babst, 2005 ). The protein machinery required for MVE formation was discovered by identification of class E vps mutants. These mutants share a common phenotype characterized by accumulation of proteins destined for the vacuole in an enlarged prevacuolar (class E) compartment (Raymond et al., 1992 ). Most of the class E VPS (vacuolar protein sorting) genes encode components of three protein complexes (endosomal sorting complex required for transport [ESCRT]), designated ESCRTI, ESCRTII, and ESCRTIII, which are required for the formation of inwardly budding luminal vesicles that fill the interior of MVEs. The luminal vesicles typically contain membrane proteins that arrive in the prevacuolar compartment either by vesicular trafficking from the Golgi or by endocytosis and are eventually degraded completely in the vacuole lumen. Mutations in ESCRT complex prevent formation of inwardly budding vesicles, leading to formation of a class E compartment rather than a MVE. Impaired formation of inwardly budding vesicles can block recycling of proteins such as Vps10p from the prevacuolar compartment to the Golgi, thus leading to their accumulation in the resulting class E compartment (Babst et al., 1998 , 2000 , 2002a , 2002b ; Babst, 2005 ; Katzmann et al., 2001 , 2003 ; Bilodeau et al., 2003 ; Odorizzi et al., 2003 ; Luhtala and Odorizzi, 2004 ).
Most membrane proteins that are normally delivered to the lumen of the vacuole require modification by ubiquitination as a signal for being packaged into luminal vesicles of the MVE. Recent work has demonstrated that in many cases Rsp5p is directly required at the MVEs for modification and adequate sorting of membrane proteins (Blondel et al., 2004 ; Dunn et al., 2004 ; Katzmann et al., 2004 ; Morvan et al., 2004 ). Instead of causing an increase in plasma membrane sorting, as observed for Gap1p (Helliwell et al., 2001 ), in these cases an rsp5 mutation causes the cargo to accumulate in the delimiting membranes of the endosomal and vacuolar compartments.
Deubiquitination of the MVE cargo before its internalization into luminal vesicles is also important for the proper sorting of MVE cargo proteins. For example, The ubiquitin (Ub) C-terminal hydrolase encoded by DOA4 plays a major role at this step in the deubiquitination of different MVE cargoes. A doa4Δ mutation causes mistargeting of MVE cargo proteins and a failure to recycle ubiquitin, which results in depletion of intracellular pools of free ubiquitin (Swaminathan et al., 1999 ; Amerik et al., 2000 ; Dupré and Haguenauer-Tsapis, 2001 ).
Although defects in cargo ubiquitination, recognition by ESCRT machinery, and deubiquitination at the MVEs may result in the accumulation of proteins in endosomal and perivacuolar membranes, recent studies indicate the existence of alternative pathways that allow recycling of plasma membrane proteins from the latest stages of lysosomal/vacuolar sorting (Babst et al., 2000 ; Nikko et al., 2003 ; Bugnicourt et al., 2004 ; Pizzirusso and Chang, 2004 ).
Here we present the results of a genome-wide screen used to identify new functions involved in the control of Gap1p intracellular sorting. We find that mutations in all of the class E VPS genes involved in the MVE pathway cause missorting of intracellular Gap1p to the plasma membrane. Our results show both that Gap1p must follow the MVE pathway in order to be delivered to the vacuole and that a deficiency in the MVE machinery allows efficient retrieval of Gap1p from this intracellular compartment (probably via the trans-Golgi) to the plasma membrane rather than causing its accumulation in the class E compartment. Evaluation of the Gap1p sorting in cells grown on different nitrogen sources shows that cycling of Gap1p from the MVE to the plasma membrane is the main step in the intracellular trafficking itinerary of Gap1p that is regulated by amino acids.
A genome-wide screen was performed using the collection of kanMX-marked deletion mutants in nonessential genes of S. cerevisiae, from EUROSCARF (http://www.uni-frankfurt.de/fb15/mikro/euroscarf/data/by.html; Brachmann et al., 1998 ). All the null mutants utilized in our screen are in the BY4741 (Y00000) and BY4742 (Y10000) genetic backgrounds (MATa his3Δ1 leu2Δ2 met15Δ0 ura3Δ0 and MATα his3Δ1 leu2Δ2 lys2Δ0 ura3Δ0, respectively). Characteristically, strains of this genetic background (derived from S288C) produce high Gap1p and Put4p activity when ammonia is used as nitrogen source (Courchesne and Magasanik, 1983 ). All of the mutants assayed for Gap1p activity and localization were reconstructed in our laboratory genetic background, also derived from S288C, and are listed in Table 1. All complete gene deletions described here were obtained by replacement of the functional ORF of the corresponding gene by homologous recombination with either a kanMX4/6 or a natMX4 cassette (Longtine et al., 1998 ; Goldstein and McCusker, 1999 ) in the wild-type strain CKY835.
Plasmids used in this study are summarized in Table 2. Plasmids previously available in the Kaiser collection used for this work are as follows: pRS423, HIS3 2 μ (Christianson et al., 1992 ); pRS415, a LEU2-CEN vector (Sikorski and Hieter, 1989 ); pPGAP1-lacZ (pMS29), a centromeric plasmid carrying a PGAP1-LacZ fusion at codon 53 of GAP1 in the URA3-CEN vector pBL101 (Stanbrough and Magasanik, 1995 ); pGAP1-GFP (pCK230), a URA-CEN vector carrying the GAP1-sGFP fusion under the GAP1 promoter (Helliwell et al., 2001 ); and pBUL1 (pCK323), an HIS3 2 μ pRS423 plasmid containing BUL1 ORF plus 5′ and 3′ regions. To make the centromeric plasmid covering for leucine, histidine, and methionine auxotrophies, pCEN-HIS3-LEU2-MET15 (pCK283), a DNA fragment containing the HIS3 marker from pRS423 (Christianson et al., 1992 ), flanked by the restriction sites SacI and SalI, was obtained by PCR and introduced in the plasmid pRS415 (Sikorski and Hieter, 1989 ). A PCR fragment containing MET15 was obtained by genomic PCR from CKY835. Both the fragment and the intermediate vector (containing LEU2 and HIS3 markers), previously digested with ApaI, were treated with T4-DNA polymerase before the ligation. To construct the plasmids pPCUP1-myc-UBI (pCK322) and pPCUP1-UBI (pCK331), a BamHI-ClaI fragment containing the copper-inducible ubiquitin cassette PCUP1-myc-UBI amplified from pCK231 plasmid (Helliwell et al., 2001 ) or PCUP1-UBI amplified from YEP96 plasmid (Ecker et al., 1987 ; Ellison and Hochstrasser, 1991 ) were, respectively, ligated to BamHI/ClaI-digested pRS306 2 μ, for URA3 selection (Sikorski and Hieter, 1989 ; Helliwell et al., 2001 ). Escherichia coli DH5α was used for each cloning step. The rest of the genetic and DNA manipulation general procedures were performed according to the protocols described in Sambrook et al. (1997) and Adams et al. (1996) .
The defined growth media are based on Yeast Nitrogen Base (Difco, Detroit, MI) and designated according to the nitrogen source added (urea, ammonia, or glutamate), and the composition and preparation are as described by Roberg et al. (1997a) . All growth experiments were carried out at 24°C. Plates containing the toxic proline analog l-azetidine-2-carboxylic acid (ADCB; Sigma-Aldrich, St. Louis, MO) were prepared by using minimal medium with ammonia or urea as the only nitrogen source. When auxotrophic EUROSCARF strains were used, minimal medium was supplemented appropriately with amino acids, purines, or pyrimidines, added at concentrations given in Adams et al. (1996) .
Assays of the rate of uptake of radiolabeled amino acids were performed as described by Roberg et al. (1997b) . β-Galactosidase activity was measured using the permeabilized cell method (Adams et al., 1996 ).
A primary screen was performed using master 96-well microtiter dishes containing the entire collection of deletions in nonessential genes of the EUROSCARF collection. Approximately 4859 BY strains of each haploid mating type were grown on solid YPD medium overnight and transferred to 96-well plates. Strains were spotted onto plates of minimal urea medium supplemented with amino acids and containing five different concentrations (0, 8, 30, 60, 100 mg/l) of ADCB. Sensitivity or resistance to ADCB was scored after growth for 3–4 d. As controls on each screening plate, we used the wild-type standard strain, BY4741 (Y00000) and the mutant strains lst4Δ (Y05026), which causes reduced levels of Gap1p activity (Roberg et al., 1997b ), and gln3Δ (Y00173), which causes increased levels of Gap1p activity (Stanbrough and Magasanik, 1995 ; Chen and Kaiser, 2002 ). Candidate strains with either increased sensitivity or resistance to ADCB were reconfirmed in both mating types (BY4741 and BY4742 backgrounds). Subsequent to the initial screening of the deletion strains, Gap1p activity and transcription were determined after transformation of strains with pCK283 to render the strains prototrophic (so that amino acid supplements would not be necessary in the growth medium) and with pMS29 as a reporter to monitor GAP1 expression.
Labeling with FM4-64 (Molecular Probes, Eugene, OR) was performed according to the procedure described in Vida and Emr (1995) . For GFP localization studies, Gap1p-GFP–expressing cells were grown to exponential phase in ammonia liquid cultures. Glutamate was added to a final concentration of 3 mM and the cells were incubated for 30 min to 1 h at 24°C. Cells were then collected by centrifugation, washed, and suspended in 1 M Tris, pH 8.0, 5% NaN3 as described by Urbanowski and Piper (1999) . This treatment stops the membrane traffic and provides alkaline conditions optimal for GFP fluorescence imaging. Images were captured with a Nikon E800 microscope (Melville, NY) equipped with a Hamamatsu digital camera (Bridgewater, NJ). The FITC filter set was used for imaging of FM4-64 detection and the chroma filter set at 41012 was used for GFP detection. Improvision OpenLabs 2.0 software (Lexington, MA) was used to process images.
To measure recycling of endocytosed membranes back to the cell surface, we carried out FM4–64 recycling assays as described in Wiederkehr et al. (2000) . Fluorescence was recorded using a Fluorolog spectrofluorometer (Jobin/Yvon, Horiba, Irvine, CA).
Detection of secreted carboxy peptidase Y (CPY) was carried out as described by Lafourcade et al. (2004) .
For the detection of Gap1 protein levels, two OD600 of cells were collected at the required times after shifting nitrogen sources and protein extracts carried out by following a protocol from Adams et al. (1996) . Proteins were resolved by 8% SDS-PAGE and detected by immunoblot using rabbit polyclonal anti-Gap1p antibody (made as explained in Risinger and Kaiser, unpublished results) at 1:2000 dilution, mouse monoclonal anti-3-phosphoglycerate kinase (1:1000, Molecular Probes) used for simultaneous detection of Pgk1p as a loading control, and HRP-conjugated donkey anti-rabbit IgG or HRP-coupled sheep anti-mouse IgG (1:10,000 dilution, Amersham, Indianapolis, IN).
For the detection of Gap1p-HA and its ubiquitin conjugates, Gap1p was immunoprecipitated and then detected by immunoblotting following an adaptation of the protocol described by Laney and Hochstrasser (2002) . Strains were transformed with pPCUP1-myc-UBI (pCK322) and then cultured in minimal urea medium to an initial OD600 of 0.01/ml. myc-Ubi expression was induced for 16 h with 1 μM CuSO4. An equivalent to 10 OD600 units of cells were collected on 0.45-μm nitrocellulose filters once cultures reached 0.4 OD600/ml. Cells were washed twice in 10 mM NaN3, suspended in 200 μl of SDS buffer containing NEM and protease inhibitors (1% (wt/vol) SDS; 45 mM Na-HEPES, pH 7.5; 50 mM NEM; 1 mM phenylmethylsulfonyl fluoride (PMSF); 0.5 μg/ml leupeptin, 2 μg/ml aprotinin, and 0.7 μg/ml pepstatin), and lysed with glass beads by vortexing 5 min at room temperature. Lysates were diluted in 700 μl of Triton lysis buffer with NEM and protease inhibitors (1% [vol/vol] Triton X-100; 150 mM NaCl; 50 mM Na-HEPES, pH 7.5; 5 mM Na-EDTA; 10 mM NEM; 1 mM PMSF; 0.5 μg/ml leupeptin; 2 μg/ml aprotinin, and 0.7 μg/ml pepstatin), and any remaining cell debris was removed by centrifugation at 4°C, 14,000 × g. Samples were preadsorbed for 1 h at room temperature with 40 μl of a 20% suspension of protein G-Sepharose 4 fast flow (Amersham Pharmacia Biotech, Piscataway, NJ) followed by brief centrifugation to remove the beads. This procedure was repeated twice. Immunoprecipitation was carried out by overnight incubation at 4°C with 10 μl of monoclonal antibody preparation (rat anti-HA [3F10]; Roche, Indianapolis, IN), followed by overnight incubation at 4°C with 60 μl of protein G-Sepharose suspension. The beads were washed five times with 1% (vol/vol) Triton in phosphate-buffered saline buffer containing NEM 10 mM, 1 mM PMSF, 0.5 μg/ml leupeptin, 2 μg/ml aprotinin, and 0.7 μg/ml pepstatin. Immunoprecipitates were solubilized by incubation in sample buffer for 1 h at 37°C and resolved by 8% SDS-PAGE. Antibodies used for immunoblotting included mouse anti-HA monoclonal 16B12 (BAbCO, Richmond, CA) at 1: 1000 dilution; mouse anti-myc, monoclonal 9E10 (Santa Cruz Biotechnology, Santa Cruz, CA) at 1: 500 dilution; and HRP-coupled sheep anti-mouse serum at 1: 10,000 dilution (Amersham Pharmacia).
To assess the inhibition of bulk translation caused by cycloheximide, cells growing exponentially in minimal glutamate medium lacking methionine were suspended in fresh medium at 5 OD600/ml and pulse-labeled for 4 min by addition of 30 μCi of [35S]methionine and [35S]cysteine (EXPRESS, NEN, Boston, MA) per OD600. Cycloheximide was then added and cells were incubated for 30 min at room temperature. Metabolic labeling of proteins was stopped by the addition of unlabeled 10 mM methionine and 10 mM cysteine before washing twice with ice-cold 20 mM NaN3. A 0.4-ml aliquot of culture was suspended in 200 μl of sample buffer containing protease inhibitors as above. Cells were lysed by vortexing with glass beads for 5 min at 4°C, and boiled for 1 min, 20-μl samples were resolved by SDS-PAGE (8% gel), and labeled proteins were detected with a 445si PhosphorImager (Molecular Dynamics, Sunnyvale, CA).
To identify new genes that govern proper sorting of nitrogen-regulated permeases, we screened 4859 haploid deletion mutants in nonessential genes of S. cerevisiae from the EUROSCARF collection. For the first step of selection in this screen we used the toxic proline analog ADCB, which is taken up primarily by Gap1p and Put4p permeases in cells grown on ammonia or urea as a nitrogen source (Roberg et al., 1997a ). Mutations causing increased Gap1p and Put4p activities display enhanced sensitivity to ADCB, whereas mutations that reduce these activities are more resistant to this compound (Roberg et al., 1997b ). By spotting each strain from the collection onto plates with minimal urea medium containing different concentrations of ADCB, a total of 162 mutants showing sensitivity to a normally sub-LC (<12.5 mg/l) of ADCB were identified. (An additional 118 mutants that displayed resistance to ADCB were identified, and these mutants will be described elsewhere.)
Enhanced sensitivity to ADCB, indicating increased activity of Gap1p and Put4p could be due to an effect on either the intracellular sorting of nitrogen-regulated permeases or their level of expression. To distinguish these possibilities, we assayed the effect of each mutant on Gap1p activity and expression. Gap1p activity was measured by the rate of uptake of [14C]citrulline, which is transported only by Gap1p. For these experiments, uptake of [14C]arginine was used to control for nonspecific effects on permease activity. GAP1 transcription was evaluated by measuring the β-galactosidase activity expressed from a PGAP1-lacZ reporter. These assays showed that 73 of the mutants had significantly increased Gap1p activity, and that for 70 of these mutants the increase in Gap1p activity was not due to an increase in GAP1 transcription (our unpublished data). Among the mutants that displayed increased Gap1p activity but normal GAP1 transcription, we almost the entire set of class E of VPS, which code for members of the ESCRT machinery acting at the MVE (Babst, 2005 ). The phenotypes obtained in the ADCB screen for mutations in all known VPS genes are summarized in the Table 3.
As shown in Table 4, most class E vps mutants showed 10- to 30-fold increases in Gap1p activity on urea medium, suggesting a greatly increased probability of sorting of the permease to the cell surface. Moreover, the deletion of DOA4, which causes a depletion of the pools of free ubiquitin, also caused increased levels of Gap1p activity. None of the mutations in this subset of genes caused a corresponding increase in the uptake of arginine, suggesting that their effect was specific for Gap1p. Similar results were observed when the same VPS genes were deleted in an S288C strain prototrophic for amino acids (CKY strains; Table 1). These mutants showed an inability to grow on plates of minimal urea medium containing 7 mg/l of ADCB, as shown for did4Δ, vps4Δ, vps27Δ, and doa4Δ (Figure 1A). Uptake assays revealed from 10- to 20-fold increases in Gap1p activity compared with the wild-type strain in the same medium (Figure 1B). This increase in Gap1p activity was about the same as that for a bul1Δ bul2Δ double mutant, which causes a 30-fold increase in Gap1p activity on urea medium (Figure 1B).
To examine the mutational effects on Gap1p trafficking, exclusive of any influence on GAP1 gene transcription, we placed the GAP1 coding sequence under control of the constitutive ADH1 gene promoter, which is unaffected by the amino acids regulatory signal (Chen and Kaiser, 2002 ; Figure 1C). All of the class E vps mutants exhibited 10- to 20-fold increased levels of activity expressed from PADH1-GAP1, showing that the increase in Gap1p activity was not due to changes in transcriptional regulation. Together, these results show that mutations in components of the MVE machinery can cause a dramatic redistribution of Gap1p to the plasma membrane.
Membrane proteins targeted to the vacuole through the MVE pathway are sorted into the membrane of inwardly budding vesicles generated at the MVE, which are ultimately delivered into the vacuolar lumen when the MVE fuses with the vacuole (Reggiori and Pelham, 2001 ; Katzmann et al., 2002 ; Babst, 2005 ). Accordingly, Gap1p-GFP in a wild-type strain growing in a high concentration of the amino acid glutamate, which induces Gap1p vacuolar sorting, localizes within the vacuolar lumen (Figure 2).
These results suggest that functional MVE machinery is necessary for the vacuolar sorting of the nitrogen-regulated amino acid permeases and that defects in the MVE pathway can lead to efficient delivery of Gap1p to the cell surface.
We asked whether the increased levels of Gap1p activity in class E vps mutants might be due to a defect in responding to amino acids. Mutations in MKS1 display high, unregulated expression of TCA pathway enzymes responsible for α-ketoglutarate formation and thus produce high levels of glutamate and glutamine (Butow and Avadhani, 2004 ). We have found that high amino acid content of mks1Δ mutants cause Gap1p to be sorted to the vacuole (Chen and Kaiser, 2002 ). Mutants defective in Gap1p polyubiquitination, when combined with mks1Δ, showed highly increased levels of active Gap1p localized to the cell surface (as observed in Figure 3, A and B, for the triple mutant mks1Δ bul1Δ bul2Δ). This result establishes that if Gap1p cannot be ubiquitinated, that Gap1p sorting no longer responds to high levels of amino acids. By contrast, a double mutant mks1Δ vps27Δ exhibited low levels of Gap1p activity (Figure 3A). Examination of Gap1p-GFP in mks1Δ vps27Δ strains revealed that Gap1p-GFP accumulated in endosomal/perivacuolar membranes, suggesting that Gap1p rerouting to the vacuolar sorting pathway in response to high internal amino acid concentrations is not impaired in this double mutant (Figure 3B).
Because the class E vps mutants exhibited high Gap1p activity and localization of Gap1p-GFP to the plasma membrane, we also considered the possibility that class E vps mutants might be defective for Gap1p endocytosis from the plasma membrane. We tested three representative class E mutants (vps4Δ, vps27Δ, and did4Δ) for a rate of decline in Gap1p activity, indicative of functional endocytosis, right after the addition of 3 mM glutamate to cells growing in ammonia medium (Figure 3C). Although cells continuously growing in the absence of amino acids maintained relatively high levels of Gap1p activity, addition of glutamate caused a decline in Gap1p activity of more than 20-fold after 20 min that was equivalent in wild-type and in the class E vps mutants. As a control to demonstrate that the decline in Gap1p activity was at least in part due to endocytosis, we showed that the endocytic mutant end3Δ did not exhibit a significant reduction in Gap1p activity after the addition of glutamate (Figure 3C).
These data indicate that, although class E vps mutants caused increased traffic of Gap1p to the plasma membrane, these mutants can respond normally to high external and internal amino acid concentrations by redirecting Gap1p away from the plasma membrane to the vacuolar targeting pathway. This behavior clearly distinguishes class E vps mutants from mutations that affect Gap1p ubiquitination, such as a bul1Δ bul2Δ double mutant, because the latter not only cause an increase in Gap1p activity, but also render Gap1p sorting insensitive to the effect of amino acids.
Mutations in the gene LST4 dramatically reduce the activity of Gap1p because they cause constitutive sorting of Gap1p to the vacuole, regardless of the nitrogen source in the growth medium (Roberg et al., 1997b ). Nevertheless, the lst4Δ phenotype can be completely suppressed by a bul1Δ bul2Δ double mutant, suggesting that Gap1p encounters the sorting step specified by Bul1p and Bul2p before the step that depends on Lst4p (Helliwell et al., 2001 ). In contrast to the behavior shown by the triple mutant bul1Δ bul2Δ lst4Δ, a simultaneous mutation of the gene LST4 completely rescued growth of the bro1Δ, vps4Δ, vps27Δ, and did4Δ mutants on plates of minimal ammonia (Figure 4A) or urea (our unpublished data) medium containing 7 mg/l of ADCB. Uptake assays to measure Gap1p activity levels were consistent with the sensitivity to ADCB. Double mutants, lst4Δ vps4Δ, lst4Δ vps27Δ, and lst4Δ did4Δ, all had ~5% of the Gap1p activity as the corresponding single class E mutant (Figure 4B). Null double mutants with lst4Δ for the remaining class E genes showed similar reductions of Gap1p activity except for an lst4Δ bro1Δ double mutant, which exhibited 20% of the activity as a bro1Δ single mutant.
The subcellular localization of Gap1p-GFP in wild-type cells grown in ammonia medium typically showed the majority of Gap1p-GFP located at the cell surface and a minor intracellular signal corresponding to internally stored pools (Figures 2 and and5A).5A). By contrast, Gap1p-GFP in the lst4Δ mutant was localized within the vacuolar lumen and to punctate structures surrounding the vacuole. Null mutations in the class E genes, VPS4, VPS27, DID4, and BRO1, showed Gap1p-GFP predominantly located at the plasma membrane. Instead, the double mutants, lst4Δ vps4Δ, lst4Δ vps27Δ, and lst4Δ did4Δ, showed most of Gap1p-GFP contained in structures adjacent to the vacuole, which was visualized using DIC optics. Similar results were obtained for deletions in VPS28, SNF8, VPS25, VPS36, VPS20, SNF7, and VPS24 (our unpublished data).
These data clearly distinguish the effect of Gap1p ubiquitination mutants (exemplified by bul1Δ bul2Δ) from class E vps mutants. Although both types of mutants exhibit a redistribution of Gap1p to the plasma membrane accompanied by greatly increased Gap1p permease activity, the class E vps mutants will respond to regulation by amino acids and lst4Δ mutations, whereas ubiquitination-defective mutants are not influenced by either. Overall, these results along with the previously demonstrated existence of an internal pool of Gap1p, even when cells are grown in poor nitrogen sources (Roberg et al., 1997a ; Helliwell et al., 2001 ), suggest that Gap1p is continuously sorted through the MVE, but in mutants defective for MVE formation, the Gap1p that accumulates in this compartment is available to be returned to the plasma membrane by a recycling pathway that depends on LST4 function and is regulated by amino acids.
Our results above indicate that a specific subset of class E vps mutants growing in the absence of high amino acid concentrations fail to retain Gap1p in the MVE. Although all class E vps mutations showed increased levels of Gap1p activity under these conditions, some of the mutants had more subtle effects on Gap1p. Mutants such as vps60Δ and hse1Δ were slightly less sensitive to low concentrations of ADCB (our unpublished data). Similarly, did2Δ, hse1Δ, vps23Δ, and vps37Δ, had only 5- to 7-fold increased levels of Gap1p activity when grown in urea as a nitrogen source (Table 4), which is about half of the effect shown by the mutations described above.
We examined the subcellular localization of Gap1p-GFP in this different subgroup using did2Δ, hse1Δ, vps23Δ, vps37Δ, or vps60Δ null mutant strains. These single mutant strains growing in ammonia displayed a pattern of Gap1p-GFP localization similar to that of a vps27Δ lst4Δ double mutant (compare Figure 5, A and B), with most of the Gap1p-GFP being located intracellularly. Confocal microscopy of these cells revealed that the diffuse pattern of localization of Gap1p-GFP was the result of some Gap1p-GFP in punctate structures adjacent to the vacuole as well as in the vacuolar membrane (our unpublished data). Apparently mutations in this subset of ESCRT genes do not allow Gap1p to exit the MVE efficiently. The differences among ESCRT mutants with respect to their effect on Gap1p do not strictly correlate with the known subdivisions of these proteins into different complexes; however, most of the mutants that allow efficient cycling of Gap1p to the plasma membrane are in ESCRT complex II and III.
To demonstrate directly the existence of a pathway for cycling of existing Gap1p from the MVE to the plasma membrane, we accumulated Gap1p in the class E compartment in a vps4Δ mutant grown on glutamate and then transferred the cells to amino acid–free medium. As expected, the vps4Δ mutant grown on glutamate exhibited low levels of Gap1p activity, but when these cells were transferred to urea medium Gap1p activity increased rapidly (Figure 6A). The initial rise in activity experienced by the vps4Δ mutant did not depend on newly synthesized permease because the early phase of induction of activity was not blocked by addition of 1.5 μg/ml cycloheximide. This concentration of cycloheximide was chosen as the minimum necessary to inhibit translation completely as assayed by incorporation of radiolabeled methionine into newly translated proteins (Figure 6B; Roberg et al., 1997a ). In contrast, activity did not increase substantially in an lst4Δ vps4Δ double mutant (that was not treated with cycloheximide), showing that Gap1p cycling to the plasma membrane does depend on Lst4p activity (Figure 6A). An immunoblot control (Figure 6C) shows that each of the cultures contained similar amounts of Gap1p protein, indicating that the differences in activity were not due to Gap1p degradation. In a parallel localization experiment, Gap1p-GFP could be seen to partially redistribute from the endosome to the plasma membrane in a vps4Δ mutant after transfer from glutamate to urea medium. However, most of the Gap1p-GFP remained in the class E compartment in an lst4Δ vps4Δ double mutant under the same conditions (Figure 6D).
Interestingly, a different pattern of modified species of Gap1p can be detected by Western blot between the mutants vps4Δ and lst4Δ vps4Δ (Figure 6C). The mutant lst4Δ vps4Δ shows accumulation of several high-molecular-weight species containing Gap1p that are absent in protein extracts from vps4Δ. A wild-type strain shows a pattern similar to that observed in the lst4Δ vps4Δ, and these species are absent in a mutant strain carrying an allele of GAP1 with the two lysine acceptor sites for ubiquitination mutated to arginine (gap1K9R,K16R; Soetens et al., 2001 ), suggesting that these species correspond to ubiquitinated forms of Gap1p (Risinger and Kaiser, personal communication). A lower level of polyubiquitinated species of Gap1p has also been observed in other class E vps mutant strains grown in similar conditions (our unpublished data). This phenomenon may be related to a particular form of the endosomally localized Gap1p that accumulates when the MVE pathway is blocked.
Three different protein complexes have been identified for their role in the recycling of different subsets of proteins from the endosome to the Golgi. The vps fifty three (VFT)/Golgi-associated retrograde protein (GARP) complex consists of four subunits (Vps51p–Vps54p) and is required in association with the Rab Ypt6p for the retrograde transport of Golgi resident proteins from the endosome to the Golgi (Conibear and Stevens, 2000 ; Siniossoglou and Pelham, 2002 ). A second complex, known as the retromer complex, is necessary for recycling of Vps10p and is composed of Vps35p, Vps29p, and Vps26p and the Vps5p–Vps17p sorting nexin dimer (Reddy and Seaman, 2001 ; and reviewed by Seaman, 2005 ). A third multisubunit tethering complex with a recently discovered role in the recycling of the uracil permease Fur4p (Bugnicourt et al., 2004 ) that was first identified as being involved in the docking and fusion of vacuolar and/or endosomal membranes is the class C Vps/homotypic fusion and vacuole protein sorting (HOPS) complex (reviewed by Wickner, 2002 ; Whyte and Munro, 2002 ). Proteins forming this complex include Pep3/Vps18p, Pep5/Vps11p, Vps16p, Vps33p, Vam3p, and Ypt7 (Raymond et al., 1992 , Whyte and Munro, 2002 ). Notably, a requirement for Vam3p in the recycling of Gap1p has been suggested (Nikko et al., 2003 ).
If any of these complexes were necessary for Gap1p trafficking from the MVE to the plasma membrane, the corresponding mutants would be expected to exhibit an effect on Gap1p trafficking similar to lst4Δ. However, none of the mutations in the three known recycling complexes exhibited a significant decrease in Gap1p activity (Tables 3 and and4;4; Figure 7). Moreover, some of the class C Vps mutant strains, even showed increased levels of Gap1p activity that closely resembled to those observed for class E Vps mutations.
As an explicit test of the role of known recycling complexes in Gap1p recycling, we constructed double mutants with vps4Δ and determined the effect on Gap1p activity (Figure 7). Retromer mutations did not cause a significant reduction of the increased levels of Gap1p activity observed in a vps4Δ mutant. Some mutations such as vps26Δ actually increased the level of Gap1p activity. VFT/GARP complex mutants combined with vps4Δ caused a modest decrease in Gap1p activity relative to a vps4Δ single mutation, but the effect was much less than for that of an lst4Δ mutation and likely does not indicate a direct effect on Gap1p trafficking to the plasma membrane. Vps/HOPS complex mutants had a heterogeneous effect in the levels of Gap1p activity in a vps4Δ genetic background: compared with a vps4Δ single mutant, double mutations with vps16Δ or vps33Δ modestly decreased Gap1p activity, whereas double mutants with vam3Δ or pep3/vps18Δ either did not alter Gap1p activity or caused increased activity.
We were interested in whether the genes we have identified that do have a significant effect on Gap1p trafficking from the MVE to the plasma membrane affected other recycling processes. We therefore examined LST4 and LST7, two of the genes we have isolated that have the greatest effect on Gap1p trafficking to the plasma membrane (Roberg et al., 1997b ), for their effects on a variety of trafficking events. Both lst4Δ and lst7Δ caused a greater than 20-fold decrease in Gap1p activity in the context of a vps4Δ genetic background (Figure 8A). This decreased Gap1p activity did not correspond to a decrease in GAP1 transcription (Figure 8B) and most likely resulted from a failure of Gap1p to be transported efficiently from the class E compartment to the plasma membrane (Figure 8C). However, neither lst4Δ nor lst7Δ appeared to have an effect on Vps10p cycling between the Golgi and endosome because these mutants did not cause an increase in secretion of pro-CPY into the extracellular medium (Figure 8D). Moreover, neither lst4Δ nor lst7Δ had an effect on FM4-64 recycling from an endosomal compartment back to the plasma membrane (Figure 8, E and F). Taken together these results indicate that Gap1p has a discrete set of genetic requirements for trafficking from the MVE to the plasma membrane that does not appear to overlap with known endosomal or MVE recycling pathways.
As shown above, a class E vps mutant requires relatively low amino acid levels and functional Lst4p to give rise to increased levels of Gap1p at the plasma membrane. Accordingly, we hypothesized that Lst4p is required for trafficking of Gap1p from the class E compartment to the plasma membrane (probably via the Golgi) under conditions of low intracellular amino acids. However, the possibility remained that Lst4p could negatively regulate Gap1p endocytosis and that amino acids such as glutamate could stimulate endocytosis, such that the rate of Gap1p endocytosis was greatly increased in an lst4Δ mutant. One way to resolve these possibilities is to examine the effect of a mutation known to block Gap1p endocytosis. END3 encodes a component of general endocytic vesicles and end3Δ mutants completely block endocytosis of a variety of membrane proteins, including Gap1p (Benedetti et al., 1994 ; Nikko et al., 2003 ; Figure 3C). If an lst4Δ mutation caused an increased rate of Gap1p endocytosis, then it should be possible to reverse the effect of an lst4Δ mutant by blocking endocytosis with end3Δ. We examined Gap1p-GFP localization in end3Δ and lst4Δ single mutants and end3Δ lst4Δ double mutant strains (Figure 9A). A block in Gap1p-GFP endocytosis was evident in end3Δ mutants since Gap1p-GFP remained at the plasma membrane for more than 30 min when grown on ammonia, after the addition of glutamate, whereas the wild-type strain showed complete redistribution of Gap1p-GFP to the vacuole under the same conditions. When end3Δ mutant cells were grown continuously on glutamate, Gap1p-GFP accumulated in the vacuole, and no signal appeared at the plasma membrane, indicating that upon continued exposure to glutamate vacuolar sorting of Gap1p occurs through a pathway that does not involve previous sorting to the plasma membrane (as previously demonstrated by Roberg et al., 1997a ). Similarly, in an lst4Δ mutant most Gap1p-GFP is transported to the vacuole by a pathway that bypasses the plasma membrane because most Gap1p-GFP was delivered to the vacuole in an end3Δ lst4Δ double mutant grown on ammonia. Some of the end3Δ lst4Δ mutant cells showed a small fraction of Gap1p-GFP at the cell surface consistent with the expectation that the small fraction of Gap1p-GFP delivered to the cell surface in an lst4Δ mutant would accumulate in that location in an end3Δ mutant. The amount of Gap1p-GFP at the cell surface in an end3Δ lst4Δ, quantified by Gap1p activity assay, was similar to the low activity in an lst4Δ mutant (Figure 9B), in accordance with previous observations made by Helliwell et al. (2001) . These observations rule out the possibility that the effect of lst4Δ or growth on glutamate act to decrease Gap1p at the plasma membrane by greatly increasing the rate of endocytosis.
GGA (Golgi-associated, γ-adaptin homologues, ARF-binding) proteins have a well-characterized role in the recycling of sorting receptors of vacuolar/lysosomal hydrolases from endosomes to the TGN (reviewed by Bonifacino, 2004 ). Recent work has shown that GGA proteins can bind ubiquitin directly through their GAT domains and that defects in GGA function result in defective vacuolar sorting of Gap1p (Bilodeau et al., 2004 ; Scott et al., 2004 ). These observations suggest that GGA proteins may be responsible for sorting ubiquitinated Gap1p from the TGN to endosomes, although a direct binding of GGA proteins to ubiquitinated Gap1p has not been demonstrated. We wanted to ascertain whether the constitutive vacuolar sorting occurring in a null mutant lst4Δ strain is impaired when both GGA genes from Saccharomyces, GGA1 and GGA2, are simultaneously deleted. With this aim we decided to monitor Gap1p-GFP localization in the double mutant gga1Δ gga2Δ and the triple mutant gga1Δ gga2Δ lst4Δ. As shown in Figure 10, when cells from the double mutant gga1Δ gga2Δ growing in ammonia are switched to medium with glutamate, cells show a significant defect in the ability to sort Gap1p to the vacuole. This is in accordance with the previous observations reported by Scott et al. (2004) . However, the triple mutant gga1Δ gga2Δ lst4Δ did not show any defect associated with the lack of GGA function and instead behaved as an lst4Δ mutant, showing constitutive sorting of Gap1p to the vacuole independent from the nitrogen source. In contrast, an lst4Δ pep12Δ double mutant showed Gap1p-GFP associated with small punctae that probably correspond with multiple small vesicles unable to fuse/form an endosome (Figure 10). This latter result served as a control to show that a known block in Golgi-to-MVE traffic impairs constitutive sorting of Gap1p to the vacuole in an lst4Δ (PEP12 gene encodes a t-SNARE of the PVC required for the fusion with the endosome of vesicles trafficking toward the vacuolar sorting pathway; Becherer et al., 1996 ) and established that whatever effect loss of GGA gene function may have on Gap1p sorting in the Golgi and endosomal compartments, it is neither quantitatively or qualitatively similar to the effect of PEP12 gene function.
Deletion of DOA4, which encodes a deubiquitinating enzyme (ubiquitin C-terminal hydrolase) produces an effect on Gap1p trafficking to the cell surface, much like that of mutations in the Rsp5p/Bul1p/Bul2p ubiquitin ligase complex (Springael et al., 1999 ). In a doa4Δ mutant Gap1p-GFP was localized to the plasma membrane but did not respond to the addition of 3 mM glutamate, suggesting a lack of proper ubiquitination-directed vacuolar sorting of the permease (Figure 11A). This finding is consistent with that of previous studies showing that Doa4p-deficient cells have a depleted pool of free ubiquitin because of a failure to recycle free ubiquitin monomers from ubiquitinated proteins (Swaminathan et al., 1999 ). However, Doa4p may not only affect ubiquitination of newly synthesized Gap1p in the secretory pathway for vacuolar targeting but also affect deubiquitination of this permease at the MVE because it has already been shown that Doa4p associates with this compartment in order to deubiquitinate other proteins such as the uracil permease (Amerik et al., 2000 ; Dupré and Haguenauer-Tsapis, 2001 ). To examine the effect of doa4Δ mutation on Gap1p deubiquitination, we used a strategy previously shown to compensate for the loss of free ubiquitin (caused by the defect in ubiquitin recycling) by overexpression of a functional N-terminal myc epitope-tagged ubiquitin (Ecker et al., 1987 ; Ellison and Hochstrasser, 1991 ). A doa4Δ mutant strain carrying a multicopy plasmid (pPCUP1-myc-UBI) was grown in the presence of 1 μM CuSO4 to induce the expression of myc-Ub. Overexpression of myc-Ub could compensate for the effect of doa4Δ on Gap1p sorting as shown by the restoration of Gap1p-HA wild-type levels of activity (Figure 11B), and the ability of Gap1p-GFP to be rapidly relocated to the vacuole lumen when cells grown on ammonia (Figure 11A) or urea (Figure 11B) were transferred to glutamate medium. By contrast, the overexpression of myc-Ub was unable to restore wild-type levels of Gap1p activity in a doa4Δ bul1Δ bul2Δ strain (Figure 11B).
Gap1p-GFP was sorted into the vacuole lumen of doa4Δ cells overexpressing myc-Ubi. However, luminal sorting occurred with a lower efficiency than in a wild-type strain, with some of Gap1p-GFP accumulating at the vacuolar membrane and at adjacent punctate structures that may represent endosomes (Figure 11A). This partial defect in luminal sorting could be attributed to an effect of myc-tagged ubiquitin interfering with efficient targeting of Gap1p into luminal vesicles of the MVE, because parallel experiments with untagged ubiquitin (pPCUP1-UBI) showed normal localization of Gap1p-GFP in a doa4Δ mutant (Figure 11A, bottom panel). Importantly, this result demonstrates that deubiquitination of Gap1p by Doa4p is not required for sorting of the permease in the intralumenal vesicles at the MVE.
To more thoroughly evaluate the possible role of Doa4p deubiquitination of Gap1p, we also detected ubiquitinated forms of the permease by immunoprecipitation followed by immunoblotting for myc-Ub. These experiments were carried out in strains expressing Gap1p-HA from the constitutive promoter ADH1 to avoid any possible variability in GAP1 expression and a deletion of PEP4 to minimize the possibility of degradation of ubiquitin conjugates by vacuolar proteases. The strains also carried pPCUP1-myc-UBI to maintain a pool of free ubiquitin and to provide an epitope tag for detection of ubiquitin conjugates. After Gap1p-HA immunoprecipitation with rat anti-HA, samples were resolved by SDS-PAGE and immunoblotted with either mouse anti-HA (to detect Gap1p-HA) or mouse anti-myc (to detect conjugates with myc-tagged ubiquitin). Control immunoblots revealed significantly higher amount of polyubiquitinated species of Gap1p-HA in a doa4Δ pep4Δ PADH1-GAP1-HA strain (lane 4, in Figure 12A) than in a pep4Δ PADH1-GAP1-HA strain (lane 3). Samples run in parallel from identical strains expressing untagged endogenous Gap1p demonstrate that only Gap1p-HA was immunoprecipitated by the rat anti-HA (because no signal appears in lanes 1 and 2). This result confirmed a role of the Doa4p enzyme in the deubiquitination of Gap1p at the MVE. It appears that the deubiquitination of Gap1p before its entry into the MVE is carried out with the aim to recycle ubiquitin rather than being a prerequisite for the delivery of Gap1p into the MVE. Moreover, this genetic background, deficient in Gap1p deubiquitination, provides us with much more sensitive conditions for detecting variations in the accumulation of polyubiquitinated Gap1p under different conditions.
We used this assay for the detection of ubiquitinated Gap1p to evaluate the extent to which Gap1p sorting to the vacuole during growth on glutamate or in an lst4Δ mutant might be a consequence of effects on Gap1p ubiquitination. We found previously that overexpression of BUL1 or BUL2 have a similar effect on targeting Gap1p to the vacuole as growth on glutamate or an lst4Δ mutant (Helliwell et al., 2001 ). Thus, the possibility remains that amino acids or a null mutation in LST4 could exert a positive effect on the Rsp5p/Bul1p/Bul2p ubiquitin ligase complex-dependent ubiquitination of Gap1p.
The amounts of ubiquitinated Gap1p-HA were compared in a genetic background with doa4Δ pep4Δ PADH1-GAP1-HA carrying the PCUP1-myc-UBI plasmid. Strains carrying the mutations bul1Δ bul2Δ or an allele of GAP1 with the two lysine acceptor sites for ubiquitination mutated to arginine (gap1K9R,K16R; Soetens et al., 2001 ) were used as negative controls to show that the dispersed high-molecular-weight species detected by Gap1p-HA immunoprecipitation followed by immunoblotting for myc-Ub in fact corresponded to polyubiquitinated Gap1p-HA (Figure 12A, and lanes 3 and 4 in Figure 12B). As a positive control to show the effect of hyperactivity of the Rsp5p/Bul1p/Bul2p ubiquitin ligase complex on Gap1p-HA ubiquitination, overexpression of Bul1p from a multicopy plasmid caused a corresponding increase in the amount of polyubiquitinated Gap1-HAp (Figure 12B, lane 7).
A doa4Δ lst4Δ mutant (lane 6) did not show increased levels of polyubiquitinated Gap1p-HA species when compared with a doa4Δ mutant alone (lane 2), showing that the effect of lst4Δ on Gap1p sorting is not an indirect consequence of hyper ubiquitination of Gap1p. When doa4Δ cells were transferred from urea medium to glutamate, only a modest increase in the amount of polyubiquitinated Gap1p-HA was observed (Figure 12B, lane 8). Although conditions of growth on glutamate medium cause Gap1p to be completely redirected from the plasma membrane to the vacuole, the corresponding increase in polyubiquitination of Gap1p-HA was much less than in the case of overexpression of BUL1. Thus a putative effect of glutamate on Gap1p ubiquitination does not suffice to explain the profound effect that amino acids have on Gap1p trafficking. Nevertheless we cannot yet rule out the possibility that intracellular amino acid concentrations exert, to some extent, their effect on Gap1p sorting by increasing the probability of Gap1p polyubiquitination.
Here we describe a comprehensive genome wide screen for genes required for the proper intracellular sorting of Gap1p in response to nitrogen source. Most of the mutations identified that increase the proportion of Gap1p sorted to the plasma membrane are in class E VPS genes that have been shown to be required for the formation of the MVE. These results establish that when Gap1p cannot enter the inwardly budding vesicles of the MVE efficient recycling of Gap1p from the endosome to the plasma membrane is possible. Genetic and physiological tests of the conditions under which cycling of Gap1p from endosome to the plasma membrane occurs show that high concentrations of amino acids can block cycling. Regulation of this cycling step appears to be the major physiological input of intracellular amino acid concentration in the overall process of regulating the amount of Gap1p delivered to the plasma membrane. The cycling step is also under genetic control by a collection of genes including LST4 and LST7, and these genes are potential targets for regulation by amino acids.
A second stage of the intracellular sorting itinerary of Gap1p with the potential to be regulated by intracellular amino acids is ubiquitination of Gap1p, which enables Gap1p to be targeted to the endosome in the first place. The ubiquitination state of Gap1p is set by both ubiquitinating and deubiquitinating processes, and we developed an assay for the effect of intracellular amino concentration on the accumulation of ubiquitinated Gap1p based on the finding that deubiquitination of Gap1p could largely be eliminated by deletion of DOA4. In a doa4Δ genetic background a large fraction of the total Gap1p is in a polyubiquinated state, but the fraction of Gap1p that is ubiquitinated does not change significantly in the presence or absence of amino acids. Moreover, mutations such as lst4Δ, which have an effect on Gap1p sorting similar to growth in high amino acid concentrations, also does not give rise to a significant increase in the fraction of Gap1p that is polyubiquitinated. These results lead us to conclude that poyubiquitination of Gap1p alone does not suffice to explain the dramatic effect that amino acids exert in the down-regulation of Gap1p. Regulation of Gap1p recycling must constitute an additional and very likely more important target for the regulation of Gap1p sorting by amino acid levels.
Figure 13 outlines the two proposed decision points in Gap1p sorting in the late secretory pathway, showing the steps affected by intracellular amino acids and each of the different classes of mutants known to affect the distribution of Gap1p in the cell. The first decision appears to take place in the trans-Golgi and depends on Gap1p ubiquitination by the Rsp5p/Bul1p/Bul2p ubiquitin ligase complex (Helliwell et al., 2001 ; Soetens et al., 2001 ). The second sorting decision is made in the MVE, and at this stage Gap1p can either be recycled to the plasma membrane by an LST4-dependent process or can be sorted into endosomal vesicles by ESCRT complex proteins.
We cannot, nor do we wish to, rule out the possibility that Gap1p recycling to the plasma membrane occurs by a direct MVE to plasma membrane transport step, as has been indicated for trafficking of misfolded Pma1p (Luo and Chang, 2000 ). However, when the partitioning probabilities for sorting decisions in both the Golgi and MVE are considered, we believe the present data indicates that Gap1p normally cycles between the TGN and MVE compartments. In the presence of high intracellular amino acid concentrations (or in an lst4Δ mutant) the amount of Gap1p delivered to the plasma membrane is only a few percent of the amount delivered in an ubiquitination defective mutant such as bul1Δ bul2Δ. This result shows that in a wild-type cell most Gap1p is initially transported to the MVE compartment where the amino acid–dependent sorting decision takes place. In the absence of high intracellular amino acids (or in a subset of ESCRT mutants) Gap1p is efficiently cycled from the MVE to the plasma membrane. Importantly, when either Bul1p or Bul2p is overproduced, most of Gap1p can be returned to the VPS pathway, showing that Gap1p that is recycled from the MVE enters a compartment that can still be influenced by the rate of Bul-dependent ubiquitination. This Bul-dependent compartment is most likely the TGN and not the plasma membrane because endocytosis defective mutants have little effect on the amount of Gap1p partitioned to the plasma membrane. It appears that Gap1p may cycle between the TGN and MVE multiple times—the sorting probabilities in each of these compartments thus determining the overall partitioning of Gap1p between the plasma membrane and vacuole.
In agreement with our observations, Bugnicourt et al. (2004) have recently shown that ESCRT mutants cause recycling of the uracil permease, Fur4p, to the cell surface. In this case, they observed a block in Fur4p recycling in double mutants simultaneously affecting components of the ESCRT and HOPS complexes (the latter is involved in vacuolar fusion events; see Wickner, 2002 ). These observations were interpreted to indicate that Fur4p recycling is mediated by the HOPS-dependent pathway, which bypasses the Golgi. Similarly, Nikko et al. (2003) observed that in S. cerevisiae cells of the Σ1278b genetic background the cell surface accumulation of Gap1p in a bro1Δ mutant in the presence of ammonia is abolished by a simultaneous mutation in VAM3 or VAM7 genes, also suggesting the possible existence of a recycling pathway from the vacuolar and/or late endosomal membranes to the cell surface independent from the Golgi. Although we do not rule out that such a pathway may also exist for Gap1p, in our genetic background (S288C) we found that vam3Δ or vam7Δ mutations have no significant effect on Gap1p sorting (as shown in Tables 3 and and4,4, and Figure 7). We find that null mutations in genes encoding for components of the HOPS complex (such as vps16Δ or pep3Δ) caused significantly elevated levels of Gap1p activity rather than decreased levels that would be predicted if the HOPS complex were required for cycling of Gap1p to the plasma membrane. Moreover, these mutations did not have a significant effect on cycling of Gap1p to the plasma membrane in the presence of the class E vps mutation vps4Δ. These results indicate that mutations in the HOPS complex may have a similar effect on Gap1p trafficking as ESCRT mutants that block progression of Gap1p from the MVE to the vacuole.
In contrast, mutations such as lst4Δ and lst7Δ, which block Gap1p recycling to the plasma membrane, cause a dramatic reduction in Gap1p activity in cells grown on medium without amino acids. This simple assay has allowed us to test other mutants known to participate in endosome to Golgi trafficking for a specific effect on Gap1p recycling. As shown in Tables 3 and and44 and in Figure 7, we did not detect a significant effect on Gap1p activity by mutations in the retrograde complex or in retromer components. These mutations did not interfere with the efficient cycling of Gap1p to the plasma membrane in the presence of an ESCRT mutation such as vps4Δ.
Interestingly, although mutations such as lst4Δ and lst7Δ have a profound effect on Gap1p activity, this and other mutations identified in the screen that block Gap1p cycling to the cell surface do not seem to cause a more generalized defect on trafficking as shown by the normal CPY sorting and FM4-64 recycling patterns in the corresponding null mutant strains (Figure 8; and Gao and Kaiser, unpublished results). Taken together, the specificity of these different classes of cycling mutants for different cargo proteins implies the existence of at least two genetically distinct pathways for endosome-to-Golgi trafficking.
A recent report revealed a possible role of GGA proteins in facilitating transport of Gap1p from the TGN to a prevacuolar compartment (Scott et al., 2004 ). Although a possible role of these proteins in increasing the efficiency of Gap1p vacuolar sorting rate may exist, Gap1p sorting to the VPS pathway does not absolutely depend on these functions because an lst4Δ gga1Δ gga2Δ double mutant still shows constitutive sorting of Gap1p to the vacuole. Data shown by Scott et al. (2004) indicated that the defect in vacuolar sorting observed in a gga1Δ gga2Δ double mutant is only partial and that it may be related to defective endocytic trafficking. Because here we show that LST4 must have a role in Gap1p sorting that is independent from endocytosis of the permease, it is therefore possible that GGA-dependent sorting of Gap1p occurs only in specific steps of the vacuolar sorting of Gap1p (e.g., in the formation of endocytic vesicles).
Several recent studies have indicated a direct action of Rsp5p on different cargo at the MVE (Blondel et al., 2004 ; Dunn et al., 2004 ; Katzmann et al., 2004 ; Morvan et al., 2004 ). In all of these cases the corresponding cargo proteins accumulate at the MVE in an rsp5-deficient background. By contrast, mutations affecting the polyubiquitinating machinery Rsp5p/Bul1p/Bul2p cause Gap1p to be localized at the cell surface, not the MVE. Most importantly in a bul1Δ bul2Δ double mutant either the presence of high intracellular amino acids or a lst4Δ mutation have no effect on the trafficking of Gap1p to the cell surface, indicating that in the absence of ubiquitination Gap1p never reaches the sorting step controlled by amino acids and Lst4p. The most straightforward explanation for this result is that amino acids and Lst4p control sorting at the membrane of the MVE and that the sorting step governed by Rsp5/Bul1/Bul2p action on Gap1p takes place at an earlier stage of the pathway, most likely in the trans-Golgi.
Although sharing similar features in their vacuolar sorting through the MVE pathway, other permeases have already shown differences with Gap1p in the particular steps governing their ubiquitination. For instance, although Rsp5p plays an important role in Fur4p MVE sorting, the polyubiquitination of the uracil permease does not seem to depend on Bul proteins (Blondel et al., 2004 ).
The availability of Gap1p for its own recycling out of the MVE may be the result of a balance between the state of Gap1p polyubiquitination, its consequent recognition by ESCRT proteins, and the state activity status of both Lst4p and the components of the recycling sorting machinery for Gap1p in response to amino acid concentrations. Such regulation could determine the ability of any putative recycling machinery to compete for Gap1p interaction, with certain ESCRT machinery components specialized for recognition of Gap1p as MVE cargo. In this regard, our results show that different subsets of ESCRT proteins may have a more important role than others in the ability to retain Gap1p for its MVE sorting. Our results provide additional evidence that ESCRT proteins have differential roles in the sorting of specific MVE cargo, as suggested by previous works (Kranz et al., 2001 ; Köhler, 2003 ).
In agreement with previous observations made by Nikko et al. (2003) in our experimental conditions a bro1Δ mutant showed higher Gap1p activity than the rest of ESCRT mutations, reaching levels comparable to those displayed by mutations in the polyubiquitinating machinery or DOA4. Although the combination with an lst4Δ mutation rescued the ability of a bro1Δ mutant to grow in the presence of low levels of ADCB by reducing Gap1p activity, this double mutant exhibited an unusually high percentage of Gap1p-GFP located at the cell surface. The similarity of doa4Δ and bro1Δ mutations on their effect on Gap1p sorting can now be explained by the recent discovery that Bro1p functions to recruitment of Doa4p to endosomal membranes (Luhtala and Odorizzi, 2004 ).
Future analysis involving further characterization of the amino acids ability to modulate the activities of the steps controlled by Rsp5/Bul1/Bul2p, ESCRT proteins, Lst4p, Lst7p, and recycling machinery will help to elucidate how metabolic signals can change the final fate of Gap1p and other plasma membrane proteins.
We thank Barbara Karampalas and Hongjing Qu for technical assistance and members of the Kaiser lab for materials and advice. We also thank Dr. Susan L. Lindquist for her generosity and patient support during the latest stages of this work. This work was supported by a National Institute of General Medical Sciences Grant GM56933 to C.A.K.
This article was published online ahead of print in MBC in Press (http://www.molbiolcell.org/cgi/doi/10.1091/mbc.E05.07-0669) on April 26, 2006.