Atg19p Ubiquitination and the Cytoplasm to Vacuole Trafficking Pathway in Yeast*

The cytoplasm to vacuole (Cvt) trafficking pathway in S. cerevisiae is a constitutive biosynthetic pathway required for the transport of two vacuolar enzymes, aminopeptidase I (Ape1p) and α-mannosidase (Ams1p), to the vacuole. Ape1p and Ams1p bind to their receptor, Atg19p, in the cytosol to form a Cvt complex, which then associates with a membrane structure that envelops the complex before fusing with the vacuolar membrane. Ubiquitin-like modifications are required for both Cvt and macroautophagy, but no role for ubiquitin itself has been described. Here, we show that the deubiquitinating enzyme Ubp3p interacts with Atg19p. Moreover, Atg19p is ubiquitinated in vivo, and Atg19p-ubiquitin conjugates accumulate in cells lacking either Ubp3p or its cofactor, Bre5p. Deletion of UBP3 also leads to decreased targeting of Ape1p to the vacuole. Atg19p is ubiquitinated on two lysine residues, Lys213 and Lys216, which, when mutated, reduce the interaction of Atg19p with Ape1p. These results suggest that both ubiquitination and deubiquitination of Atg19p are required for its full function.

Biosynthetic trafficking of many hydrolases to the yeast vacuole, the equivalent of the mammalian lysosome, involves transit through the secretory pathway (reviewed in Ref. 1). A unique pathway exists for the two vacuolar enzymes aminopeptidase I (Ape1p) 2 and ␣-mannosidase (Ams1p) (reviewed in Ref. 2). These proteins are synthesized in the cytosol, assembled into a large oligomeric structure called the cytoplasm to vacuole trafficking (Cvt) complex, and enwrapped by a newly formed double membrane to form a Cvt vesicle. The outer Cvt membrane docks and fuses with the vacuole membrane to release the inner membrane vesicle into the vacuole lumen. The inner membrane is then degraded, and the contents are released (3)(4)(5)(6)(7)(8). Most of the molecular machinery required for Cvt is shared by macroautophagy, a starvationinduced process that delivers bulk cytosol and organelles to the vacuole for degradation and recycling (9 -12).
Atg19p is one of the components that are uniquely employed in the Cvt pathway. Atg19p binds specifically to both cargo proteins Ape1p and Ams1p, is part of the Cvt complex, and mediates the interaction of this complex with components required for formation of the double membrane that will enclose the Cvt vesicle. Atg19p is enclosed in the Cvt vesicle along with the cargo proteins, and when the vesicle fuses with the vacuolar membrane and the inner membrane is degraded, Atg19p is released into the lumen and destroyed (4,13,14).
Most cellular proteins are ultimately degraded either by macroautophagy or the ubiquitin-proteasome pathway. In contrast to the bulk degradation of cytoplasm by macroautophagy, ubiquitin modification selectively targets protein substrates for destruction by the proteasome. Ubiquitin is activated by an ATP-dependent activating enzyme (E1) and transferred to a protein substrate by a ubiquitin-conjugating enzyme (E2), usually with the assistance of a ubiquitin ligase (E3). The conjugating enzyme attaches the C-terminal glycine of ubiquitin to a lysine residue on the substrate, forming an isopeptide bond with the free amino group of the lysine. Ubiquitin can also be conjugated to itself in this manner, forming polyubiquitin chains on substrate molecules. These chains typically serve as a degradation tag, targeting the substrate for destruction by the proteasome, a large multienzyme complex (15)(16)(17)(18).
The ubiquitination of proteins does not always lead to their degradation by the proteasome (19,20). Monoubiquitination of membrane proteins can trigger their endocytosis and nonautophagic degradation in the vacuole (21,22). A single lysine residue of proliferating cell nuclear antigen in yeast can be modified by a ubiquitin monomer, a polyubiquitin chain, or the ubiquitin-like modifier SUMO. None of these modifications triggers the degradation of proliferating cell nuclear antigen. Instead, monoubiquitination stimulates error-prone DNA repair, polyubiquitination triggers an error-free repair pathway, and SUMOylation recruits the helicase protein Srs2p and suppresses homologous recombination during S phase (23)(24)(25)(26)(27). In human cells, Fanconi's anemia protein FANCD2 is activated and targeted to DNA damage foci by monoubiquitination (28). This ubiquitin monomer can be removed by the deubiquitinating enzyme Usp1p, which is itself cell cycle-regulated and may serve to prevent FANCD2 hyperactivity (29).
The cellular roles of deubiquitinating enzymes (Dubs) such as Usp1p are poorly understood. More than 80 Dubs are encoded in the human genome (30) and 17 in the yeast genome, most bearing little homology to the others beyond the residues that comprise their active sites. Physiological roles for Dubs exist at several points in the ubiquitin pathway. These roles include the generation of free ubiquitin, which is always synthesized as a protein fusion to itself or to ribosomal proteins; the removal of ubiquitin from substrate fragments after substrate degradation by the proteasome; the processing of polyubiquitin chains, either for remodeling their branched structure or for regeneration of free ubiquitin; and, as in the case of Usp1p, the reversal of ubiquitinations that serve to modify substrate activity rather than target substrates for degradation (17,18).
No direct connection between the ubiquitin system and the Cvt pathway has yet been established. Ubiquitin-like modifications are essential events in the Cvt and macroautophagic pathways (reviewed in Ref. 31). The ubiquitin-like proteins Atg8p and Atg12p are both activated in an ATP-dependent manner, transferred to E2-like conjugation enzymes, and attached to their respective targets via C-terminal glycine residues. In the case of Atg8p, which is attached to phophatidylethanolamine, there is even a Dub-like enzyme, Atg4p. Atg4p activates Atg8p after synthesis by removing a C-terminal peptide and can also remove Atg8p from its phosphatidylethanolamine substrate. However, these modifications involve neither ubiquitin itself nor components of the ubiquitin system.
In this study, we show that ubiquitination of the Cvt cargo receptor Atg19p occurs in vivo and that the Ubp3p-Bre5p Dub complex is most likely responsible for removing this ubiquitin. Intriguingly, our data suggest that both attachment and removal of ubiquitin are required for full activity of Atg19p. These findings implicate ubiquitin as a functional component of the Cvt pathway.

EXPERIMENTAL PROCEDURES
Yeast Strain Construction and Growth Media-Yeast strains used in this study are listed in TABLE ONE. To create strains BB510 (wild type) and BB511 (ubp3-⌬), S288C-derived MAT␣ strain BY4742-ubp3-⌬::kanMX4 (EuroScarf) was crossed with S288C-derived MATa strain PSY1838 (32) to create a diploid heterozygous for trp1 and ubp3-⌬. This diploid was sporulated to give haploids BB510 and BB511. Strain BB520 (bre5-⌬) was similarly derived from a cross between BY4742-bre5-⌬::kanMX4 (Euro-Scarf) and PSY1838. Strains BB532 and BB529 carry URA3-marked insertions of ATG19 (with or without an N-terminal protein A tag, respectively) at the ATG19 locus. These strains were created by amplifying the ATG19 and URA3 coding regions of plasmid pTS477 (a generous gift from Dan Klionsky (4)) by PCR using primers whose 5Ј-ends would direct integration to the ATG19 locus and whose 3Ј-ends annealed to the URA3 promoter (for the downstream primer) and either the beginning of the ATG19 open reading frame (for BB529) or the beginning of the protein A tag (for BB532). The resulting linear DNA fragment was used for integrative transformation of BB510, resulting in strains in which the inserted ATG19 (without or with a protein A tag) is under control of its own 5Ј-untranslated region, and its 3Ј-untranslated region has been replaced by pTS477 vector sequence including the URA3 marker gene. Uracil prototrophs were selected, and integration at the ATG19 locus was confirmed by PCR. Strains BB530, BB531, BB533, and BB534 were constructed in the same manner, using pTS477-derived plasmid templates carrying point mutations in ATG19 that replace Lys codons with Arg codons (see "Site-directed Mutagenesis").
For all six strains, the entire ATG19 open reading frame was amplified and sequenced after integration to confirm the absence of any unintended nonsilent mutations.
Standard growth media (YPD, SD, SC, and LB) were as described (33,34). Yeast were routinely grown at 30°C, and Escherichia coli were grown at 37°C.
Site-directed Mutagenesis-The ATG19 2ϫmut and ATG19 11ϫmut alleles were created by site-directed mutagenesis using the QuikChange TM protocol (Stratagene). For the 2ϫ mutant, one primer pair was used to change ATG19 codon 216 from AAA (Lys) to AGA (Arg) in plasmid pTS477 (4). This mutation was confirmed by sequencing, and then a second pair of primers was used to change codon 213, also from AAA to AGA. The entire ATG19 open reading frame was then sequenced to confirm the absence of any unintended mutations. For the 11ϫ mutant, codons 165, 169, 176, 182, 186, 238, 247, 248, and 251 of the 2ϫ mutant were all changed from Lys to Arg by successive rounds of site-directed mutagenesis. Again, the entire open reading frame was sequenced, and the absence of any unintended mutations was confirmed.
Plasmid Constructions-UBP3-containing two-hybrid bait plasmids were constructed by amplifying UBP3 by PCR and ligating the PCR products into pGBD-C1 (a generous gift of Phil James (35)) using PCRintroduced EcoRI and PstI sites. The resulting constructs encode the Gal4 DNA binding domain fused to amino acids 1-459 (Nterm), 449 -912 (Cterm), or 1-912 (full length) of Ubp3. The "Cys box" and "His box" regions of the active site of Ubp3 (36) are thus included in the Cterm and full length, but not the Nterm bait plasmid. GST fusion plasmid pGEX-UBP3-N was constructed by amplifying the UBP3 portion of pGBD-UBP3-Nterm by PCR and ligating the PCR product into pGEX-6P-1 (Amersham Biosciences) using PCR-introduced EcoRI and XhoI sites. The resulting plasmid encodes a fusion of GST to amino acids 1-459 of Ubp3 under control of the isopropyl 1-thio-␤-D-galactopyranoside-inducible ptac promoter. Plasmid pET29-ATG19, encoding full-length Atg19 with an N-terminal S epitope tag, was created by amplifying the ATG19 coding region by PCR and ligating the PCR product into pET29a (Novagen), using PCR-introduced BamHI and XhoI sites.
To create a centromeric plasmid encoding C-terminally tagged Atg19 under its own promoter, the triple HA epitope was PCR-amplified from plasmid GTEPI (a generous gift of Rita Miller (37)) using primers designed to add an XbaI site to the N-terminal end of the epitope and a stop codon followed by a SacII site to the C-terminal end. This product was ligated to XbaI/SacII-digested pRS316 (38) to create p316-Cfus-HA. ATG19 sequence, including 1192 bp of upstream flank and the entire open reading frame without the stop codon, was PCR-amplified

Atg19p Ubiquitination and the Cvt Pathway in Yeast
from strains BB529, BB530, and BB531 (for WT, 2ϫ mutant, and 11ϫ mutant sequence; see "Yeast Strain Construction") with primers designed to introduce an XhoI site at position Ϫ1192 and an XbaI site at the end of the open reading frame. These products were ligated to XhoI/ XbaI-digested p316-Cfus-HA to create p316-ATG19-WT-HA, p316-ATG19 -2ϫ-HA, and p316-ATG19 -11ϫ-HA. The absence of nonsilent mutations was confirmed by DNA sequencing. HIS3-marked plasmids encoding ubiquitin driven by the CUP1 promoter (with or without a Myc epitope tag) were created by subcloning the CUP1 promoter, ubiquitin open reading frame, and CYC1 terminator from YEp105 or YEp96 (generous gifts of Mark Hochstrasser (39)) as a BamHI/ClaI fragment into pRS413 (38). The new plasmids are p413-Ub and p413-MycUb.
Two-hybrid Screen-Two-hybrid strain PJ69-4a (35) was transformed with each of the three UBP3 bait plasmids and then with yeast two-hybrid libraries Y2HL-C1, Y2HL-C2, and Y2HL-C3, provided by Phil James (35). Transformations were performed using standard methods (40). More than 1 ϫ 10 6 transformants/library were obtained with the C-terminal and full-length bait plasmids, and more than 3 ϫ 10 6 transformants/library were obtained for the N-terminal bait plasmid, giving Ͼ95 and Ͼ99.9% confidence of full genomic coverage, respectively. Transformation mixtures were spread on SC plates lacking Trp, Leu, and His and allowed to incubate for 18 -21 days before being replica-plated to SC plates lacking Trp, Leu, and Ade and incubated approximately 1 more week. Colonies were then picked from these plates and streaked to fresh SC plates lacking Trp, Leu, and Ade. Candidates able to form colonies in these fresh streaks were scored as Adeϩ. These candidates were then tested for dependence of adenine and histidine prototrophy on retention of the bait plasmid. Library plasmids were rescued from candidates showing bait-dependent adenine and histidine prototrophy, retransformed into fresh bait-carrying PJ69-4a as a final confirmation, and sequenced.
GST Pull-down Assay-For induction of GST-Ubp3 or GST, a single colony of BL21 cells carrying pGEX-UBP3-N or pGEX-6P-1 was used to inoculate 100 ml of LB ϩ 100 g/ml ampicillin and grown at 37°C to an A 600 of ϳ0.85. Isopropyl 1-thio-␤-D-galactopyranoside was added to a final concentration of 1 mM, and cells were grown for another 2 h before being harvested and frozen at Ϫ80°C. S-tagged Atg19 was induced similarly from BL21 (DE3) cells carrying pET29a-ATG19, except that cells were grown in LB ϩ 100 g/ml G418 to an A 600 of ϳ0.4 before induction with 0.4 mM isopropyl 1-thio-␤-D-galactopyranoside. Lysis was performed at room temperature for 15-20 min in BugBuster reagent with rLysozyme and Benzonase (Novagen) with the addition of 1.5 mM phenylmethylsulfonyl fluoride and protease inhibitor mixture (Roche Complete, EDTA-free). Lysates were clarified in a microcentrifuge for 10 min at 4°C. For each binding assay, 400 l of GST Bind Resin suspension (Novagen) was washed with bind/wash buffer (Novagen) and added to an aliquot of GST or GST-Ubp3 lysate representing 0.25 or 6.0 A 600 units of induced cell culture, respectively. Less of the GST lysate than the GST-Ubp3 lysate was used in each binding assay, because GST was expressed at least 50 times more efficiently than GST-Ubp3, as assessed by Coomassie staining of the extracts (not shown). An aliquot of uninduced BL21 lysate representing ϳ5.75 A 600 units of cells was added to the GST binding assays to normalize total protein added. Beads and lysates were incubated together at room temperature for 30 min before the beads were spun and the supernatant was removed. An aliquot of S-tagged Atg19 lysate representing ϳ0.25 A 600 units of cells was then added together with an aliquot of uninduced BL21 lysate representing ϳ2.25 A 600 units of cells as a nonspecific competitor. Beads and lysates were again incubated together at room temperature for 30 min before the beads were spun, washed three times with 1 ml of bind/wash buffer (Novagen), and eluted with 50 l of elution buffer for 20 min at room temperature. Eluates were separated in 10% acrylamide gels and transferred to a polyvinylidene difluoride membrane, and the S tag was detected using S-protein horseradish peroxidase conjugate (Novagen) at a 1:5000 dilution.
IgG Affinity Purifications/Anti-HA Immunoprecipitations-For the experiment in Fig. 2B, overnight cultures were diluted to an A 600 of ϳ0.1 and allowed to grow to an A 600 of ϳ0.3 before a 3-h induction with 100 M CuSO 4 . Cells were then harvested and lysed by vortexing with glass beads in Triton lysis buffer (1% Triton X-100, 150 mM NaCl, 50 mM HEPES, pH 7.5, 5 mM EDTA) with 1.5 mM phenylmethylsulfonyl fluoride and protease inhibitor mixture (Roche Applied Science). Lysates were clarified by centrifugation at 4°C for 15 min. Normal rabbit IgGagarose (Santa Cruz Biotechnology, Inc. (Santa Cruz, CA) or Sigma) was added to the clarified lysates, and samples were incubated overnight, rotating, at 4°C. The agarose beads were then pelleted and washed three times with IP wash buffer (Triton lysis buffer plus 0.1% SDS) before resuspension in one part IP wash buffer plus one part 2ϫ SDS sample buffer (125 mM Tris, pH 6.8, 4% SDS, 5% ␤-mercaptoethanol, 20% glycerol, 0.01% bromphenol blue). Resuspended samples were boiled for 5 min, chilled on ice, and loaded on a 10% acrylamide gel, with the equivalent of 2.5 A 600 units of cells/lane. Proteins were separated by SDS-PAGE, transferred, and detected using anti-Myc monoclonal antibody (Santa Cruz Biotechnology, Inc., Santa Cruz, CA). The experiment in Fig. 2C was conducted similarly, except that no CuSO 4 was added to the cultures, and the equivalent of 50 A 600 units of cells was loaded per lane.
For the experiment in Fig. 4C, wild-type BB510 cells were transformed with pTS477 (CUP1-ProtA-ATG19 (4)) and YEp96, YEp105 (encoding wild-type ubiquitin without or with a Myc epitope tag, respectively), or YEp110 or YEp111 (encoding K48R mutant ubiquitin without or with a Myc tag; generous gifts of Mark Hochstrasser) (41). Cells were grown to early log phase in selective medium, induced with 100 M CuSO 4 for 3.5 h, and harvested in midlog phase. Cells were lysed, and protein A-Atg19 was isolated as described above. Because expression of K48R mutant ubiquitin was consistently less than that of wild-type ubiquitin under these conditions, loadings were normalized to total Myc-ubiquitin staining of cell lysates. For the final gel shown, 3.75 A 600 units of cells were loaded per lane for wild-type ubiquitin samples, and 15 A 600 units were loaded for the K48R samples.
For the experiments shown in Fig. 5, A and D, cells were grown in selective medium with induction by 100 M CuSO 4 (Fig. 5, A (left) and D) or without it (Fig. 5A (right)). Cells were lysed, and protein A-Atg19 was isolated as described above. Two A 600 units of cells were used per lane in Fig. 5A (left), 30 A 600 units were used in Fig. 5A (right), and 5-7.5 A 600 units were used in Fig. 5D. For Fig. 5D, loadings were first normalized to protein A-Atg19 staining, and the blot was probed with Ape1p antibody to detect cargo interaction.
For the experiments shown in Fig. 5B, atg19-⌬ cells (Euroscarf) were transformed with p413-Ub or p413-MycUb together with p316-ATG19-WT-HA, p316-ATG19 -2ϫ-HA, and p316-ATG19 -11ϫ-HA (see "Plasmid Constructions"). Cells were grown in selective medium, and ubiquitin expression was induced with 100 M CuSO 4 for the final 4 h of growth. Cells were lysed as described above, and Atg19-HA was immunoprecipitated with anti-HA antibody conjugated to agarose beads (Santa Cruz Biotechnology). Immunoprecipitates were washed as described above. Loadings were normalized to total Myc-Ub staining of lysates, and 8 -17 A 600 units of cells (for anti-HA immunoblot) or 32-70 A 600 units (for anti-Myc) were loaded per lane in the final gels. NOVEMBER 25, 2005 • VOLUME 280 • NUMBER 47

Atg19p Ubiquitination and the Cvt Pathway in Yeast
Gel Electrophoresis and Mass Spectrometry-Strain BB520 (bre5-⌬) was transformed with plasmids pTS477 (CUP1-ProtA-ATG19 (4)) and YEp105 (CUP1-Myc-Ub (39)) and grown to an A 600 of 3.0 in SC lacking uracil and tryptophan and supplemented with 100 M CuSO 4 . Approximately 1000 A 600 units of cells were harvested and lysed by vortexing with glass beads in 10 ml of Triton lysis buffer supplemented with 1.5 mM phenylmethylsulfonyl fluoride and protease inhibitor mixture (Roche Applied Science). The clarified lysate was precipitated with 100 l of IgG-agarose suspension (Santa Cruz Biotechnology) by rotating overnight at 4°C. Beads were washed four times with 1 ml of immunoprecipitation wash buffer, resuspended in 25 l of immunoprecipitation wash buffer plus 25 l of 2ϫ SDS sample buffer, and boiled for 10 min. The sample was cooled on ice before being loaded to an 8% acrylamide SDS gel. After electrophoresis, the gel was stained with GelCode Blue stain (Pierce).
All visibly stained bands were excised and subjected to in-gel digestion (see, on the World Wide Web, donatello.ucsf.edu/ingel.html) (42) with trypsin (porcine, side-chain protected; Promega). The digests were analyzed by capillary HPLC-tandem mass spectrometry using a C18 PepMap 75 m ϫ 150-mm column on an Eksigent nano-HPLC pump (Eksigent, Livermore, CA) linked with a FAMOS autosampler (LC Packings, San Francisco, CA). Solvent A was 0.1% formic acid in water, and B was 0.1% formic acid in acetonitrile, and a flow rate of ϳ300 nl/min was employed. One l of the digests was injected at 5% B, and then the organic content of the mobile phase was increased linearly to 50% B over 30 min. The column effluent was directed to a QSTAR Pulsar tandem mass spectrometer (Applied Biosystems/MDS Sciex, Toronto, CA). Throughout the chromatographic separation, 1-s MS acquisitions were followed by a 3-s collision-induced dissociation acquisition for computer-selected precursor ions in information-dependent acquisition mode. The collision energy was set according to the m/z value and charge state of the precursor ion. The collision-induced dissociation spectra were submitted for data base searching using in-house ProteinProspector (available on the World Wide Web at prospector.ucsf.edu), allowing for ubiquitination as a variable modification. In addition, the collision-induced dissociation spectra of the modified peptides were manually inspected.
Gel bands with molecular mass corresponding to protein A-Atg19p with zero, one, two, three, and four ubiquitin molecules attached all yielded Atg19p peptide sequence, with 17-24 peptides per gel band, representing 42-55% coverage of the protein. Two of these bands (representing one and four ubiquitin modifications, respectively) yielded ubiquitinated Atg19p peptides, and all four of the larger bands yielded peptides of ubiquitin itself.
In Vivo Labeling and Immunoprecipitation-Yeast cultures were grown to A 600 of 0.5 in SD supplemented with the required amino acids. Cells (2 A 600 units/time point) were harvested by centrifugation at 1800 ϫ g and resuspended in 450 l of SD medium with amino acids. A 10-min pulse with 20 mCi of [ 35 S]cysteine/methionine per time point was followed by the addition of 4 ml of chase solution (5 mM methionine and 1 mM cysteine final concentration in YPD). Samples (1 ml) were taken at the indicated time points. The samples were precipitated with 10% trichloroacetic acid and washed twice with cold acetone, dried, and resuspended in 100 l of cracking buffer (50 mM Tris, pH 6.8, 1% SDS, 6 M urea, 1 mM EDTA) plus an equal volume of acid-washed glass beads. Following a 5-min vortex, samples were diluted 10-fold in immunoprecipitation dilution buffer (50 mM Tris, pH 7.5, 150 mM NaCl, 0.5% Tween 20, 5 mM EDTA, and 100 mg/ml bovine serum albumin) and centrifuged for 5 min at 13,000 ϫ g. The supernatant was precipitated overnight with anti-Ape1p antibody (0.5 l/2 A 600 units of cells) fol-lowed by a 2-h incubation with protein A-conjugated Sepharose (2% (v/v) final). Immune complexes were then washed and processed for SDS-PAGE. Dried gels were exposed to bio-imager plates and read on a Fuji BAS bio-imager.

RESULTS
Atg19p Interacts Genetically and Biochemically with Ubp3p-To identify proteins that interact with Ubp3p, we used three different Ubp3p bait constructs to screen a set of yeast two-hybrid libraries (35). Full-length ATG19 was isolated three times in these screens, once by interaction with the N-terminal half of Ubp3p (amino acids 1-459) and twice with the full-length bait construct. Interestingly, in none of the three plasmids was the ATG19 open reading frame fused to the GAL4 gene fragment, suggesting that Atg19p has a cryptic transcriptional activation domain. Upon retesting, all three ATG19 plasmids were positive for interaction with both the N-terminal and full-length bait constructs (e.g. Fig. 1A).
Specific interaction between the N-terminal half of Ubp3p and Atg19p was confirmed biochemically. A bacterially expressed fusion protein between GST and the N-terminal 459 amino acids of Ubp3p was bound to glutathione beads and then exposed to a bacterial lysate containing epitope-tagged Atg19p. Atg19p bound specifically to beads carrying GST-Ubp3p (Fig. 1B). In addition to confirming the interaction between Ubp3p and Atg19p, this experiment demonstrates that the interaction requires neither ubiquitin itself nor any other component of the ubiquitin system, which would not be present in the E. coli extracts.
Atg19p Is Ubiquitinated in Vivo-Interaction of a deubiquitinating enzyme with Atg19p was surprising, because ubiquitin had not been implicated in the Cvt pathway. One explanation for the interaction would be that Atg19p is ubiquitinated in vivo and that Ubp3p serves to remove this ubiquitin. If so, the cofactor of Ubp3p, Bre5p, which is required for deubiquitination by Ubp3p of some substrates (43,44), might also be involved. We tested the role of Ubp3p and Bre5p in the deubiquitination of Atg19 by overexpressing protein A-tagged Atg19p in wild-type, ubp3-⌬, and bre5-⌬ cells together with either an untagged or a Myc epitope-tagged version of ubiquitin (39). This protein A-Atg19p construct was functional, as shown by its ability to complement an atg19-⌬ strain's defect in Ape1p processing (Fig. 2A, lane 4  versus lane 3), and its overexpression did not compromise endogenous Atg19p protein function (Fig. 2A, lane 2 versus lane 1).
Protein A-Atg19p was isolated by affinity for IgG-conjugated agarose beads, and the bead-bound proteins were probed by anti-Myc immunoblot. The results show that Atg19p can in fact be ubiquitinated in vivo (Fig. 2B). Moreover, ubiquitin conjugates of Atg19p were consistently FIGURE 1. Ubp3p interacts with Atg19p. A, two-hybrid results. PJ69-4a cells carrying pGAD library plasmid N3-10, which contains full-length ATG19, were transformed with pGBD plasmids containing no insert (vect.), full-length UBP3, or the N-or C-terminal half of UBP3. Transformants were streaked on a plate selective for both plasmids and lacking histidine. Similar results were seen on plates lacking adenine. B, epitope-tagged Atg19p interacts specifically with GST-Ubp3p. Agarose beads carrying no protein, GST alone, or GST-Ubp3p were exposed to a bacterial extract containing S-tagged Atg19p. Bound proteins were separated by size in a polyacrylamide gel, and the S tag was detected by its affinity for an S-protein horseradish peroxidase conjugate. more abundant in ubp3-⌬ cells and bre5-⌬ cells than in wild type, although this effect was subtle (Fig. 2B, compare lanes 4 and 6 with lane  2). For this experiment, both Atg19p and ubiquitin were under control of the CUP1 promoter, which was strongly induced by the addition of copper sulfate to the growth medium. We reasoned that the overexpression of both components might drive an abnormally high accumulation of Atg19p-ubiquitin conjugates even in wild-type cells, minimizing the effect of UBP3 or BRE5 deletion. This prediction was borne out; when cells were grown without added copper, the effect of UBP3 deletion on Atg19p ubiquitination was dramatic (Fig. 2C, lanes 2 and 4). This effect was completely suppressed by the addition of a plasmid-borne copy of UBP3 (Fig. 2C, lane 6). Together, these data indicate that Ubp3p is a deubiquitinating enzyme for Atg19p and that Bre5p is a required cofactor for this activity.
Ape1p Maturation Is Delayed in Cells Lacking Ubp3p-One interpretation of the results described above is that the expression of nonnative Atg19p, for example as a two-hybrid or protein A fusion, leads to its ubiquitination. We therefore set out to test whether ubiquitination of Atg19p is physiologically relevant.
Atg19p is required for trafficking of Ape1p and Ams1p to the vacuole via the Cvt pathway. Since Ape1p is activated by proteolytic cleavage upon its release into the vacuole lumen, its trafficking, and therefore the function of Atg19, can be monitored by pulse-chase analysis. We asked whether Ape1p maturation was affected by deletion of UBP3. Wild-type and ubp3-⌬ cells were pulse-labeled with [ 35 S]methionine, and Ape1p was isolated by immunoprecipitation. Ape1p maturation was reproducibly delayed in ubp3-⌬ cells (Fig. 3, A and B), a defect that was complemented by the introduction of plasmid-borne UBP3 (Fig. 3, C and D). These findings support the conclusion that ubiquitination of Atg19p is physiologically relevant and that deubiquitination of Atg19p by Ubp3p is required for full Atg19p activity.

Mutation of Atg19p Ubiquitination Sites Impairs Cargo Interaction and Delays Ape1p
Processing-Deletion of UBP3 is likely to be pleiotropic, since Ubp3p is known to be involved in a variety of cellular processes (43,45,46). Therefore, to test the role of Atg19p ubiquitination more directly, we employed mass spectrometric peptide sequencing to identify the ubiquitin acceptor sites in Atg19p.
Protein A-Atg19p-ubiquitin conjugates and their associated proteins were isolated by affinity for IgG-conjugated agarose, separated by SDS-PAGE, digested with trypsin, and sequenced. In addition to protein A, Atg19p, and ubiquitin, the peptide sequences identified included tryptic fragments of Ams1p, Atg11p, and Ape1p. Ams1p and Ape1p are cargo proteins of the Cvt pathway and, together with Atg19, form the Cvt complex. Atg11p/Cvt9p associates with the complex and facilitates its interaction with the preautophagosomal structure (4). Identification of Ams1p, Ape1p, and Atg11p in the IgG-bound fraction in this experiment demonstrates that the protein A-Atg19p reporter was assembled into native Cvt complexes in vivo, consistent with the ability of this construct to complement an atg19 deletion ( Fig. 2A).
Ubiquitinated peptides were identified in the mass spectrometric analysis by the presence of the diglycine residue that remains attached to a conjugated lysine after tryptic digestion (47). Two ubiquitinated peptides were identified from Atg19p, SELVNFFTELK 213 (GG)TVK and TVK 216 (GG)QLEDVFQR, with Lys 213 and Lys 216 as ubiquitin acceptor sites, respectively (Fig. 4, A and B). No double modified peptides with both Lys 213 and Lys 216 ubiquitinated were observed, although their presence cannot be ruled out. No Atg19p residues other than Lys 213 and Lys 216 were found to be ubiquitinated, although their existence also cannot be ruled out. Reversal of ubiquitination may have occurred during protein isolation, and the 42-55% peptide coverage of the protein per band means that additional ubiquitination sites may have been missed.
Ubiquitin itself also showed ubiquitination, at Lys 48 (data not shown), indicating that the ubiquitin ladder we observed in Atg19p pull-downs (Fig. 2, B and C) represents polyubiquitin chains attached to Atg19p. We confirmed that these polyubiquitin chains are linked at Lys 48 of ubiquitin by overexpressing a Myc-tagged version of ubiquitin with a lysine to arginine substitution at position 48. In the presence of this construct, monoubiquitinated Atg19p accumulated, and levels of polyubiquitinated Atg19p decreased (Fig. 4C). The remaining polyubiquitinated bands can be accounted for by the continued presence of endogenous, unlabeled ubiquitin in these strains.
To block ubiquitination of Atg19p at the sites we identified, we mutated both Lys 213 and Lys 216 to arginine by site-directed mutagene- ). Twenty times as much cell culture (as judged by A 600 ) was used in C as was used in B. The level of ubiquitin conjugates is lower in C than in B, because cells in B were induced for 3 h with 100 M CuSO 4 to increase expression of both plasmid-borne ubiquitin and Atg19p, whereas cells in C were grown with no additional copper. In both B and C, the nonspecific reactivity of protein A in the immunoblot is much lower than the specific reactivity of the Myc epitope, as assessed by parallel Coomassie staining (not shown). NOVEMBER 25, 2005 • VOLUME 280 • NUMBER 47 sis. Surprisingly, IgG-agarose isolation of this "2ϫ" mutant construct in the presence of Myc-ubiquitin showed no effect of the mutations on Atg19p ubiquitination (Fig. 5A, left). Because ubiquitination can be promiscuous, occurring at flanking lysine residues when the primary sites are blocked by mutation (48 -50), we systematically changed every lysine codon within 75 codons of the identified ubiquitination sites to arginine, creating an "11ϫ" mutant in which Lys 213 , Lys 216 , and nine additional flanking lysine residues are altered, and a stretch of over 150 amino acids centered on Lys 213 and Lys 216 is lysine-free. This construct was still ubiquitinated (Fig. 5A, right). The persistence of ubiquitination is not a by-product of either the protein A tag or of overexpression, because an HA-tagged Atg19p expressed from its own promoter on a centromeric plasmid was still ubiquitinated, whether the ATG19 sequence was wild-type, 2ϫ, or 11ϫ (Fig. 5B). In these strains, the HAtagged Atg19p was the only copy, and its expression was somewhat lower than endogenous Atg19p in a wild-type strain (data not shown).

Atg19p Ubiquitination and the Cvt Pathway in Yeast
Although not unprecedented (48 -50), the resistance of Atg19p ubiquitination to mutation was surprising. The ubiquitin conjugation machinery must be able to use one or more of the remaining 17 lysine residues in the 11ϫ mutant as ubiquitin attachment sites. It is even possible that Lys 213 and Lys 216 are preferred ubiquitin attachment sites only in the context of Atg19p overexpression and that other sites normally predominate.
We speculated that if Lys 213 and Lys 216 are functionally important ubiquitin attachment sites, their mutation to arginine, although not causing a detectable change in the level of Atg19p ubiquitination, might still impair Atg19p function. This speculation was correct. As detected by steady-state Western blot, maturation of Ape1p was slightly delayed in the 2ϫ mutant and dramatically delayed in the 11ϫ mutant (Fig. 5C). This defect is probably a result of impaired interaction with Ape1p, since IgG-agarose isolation of protein A-Atg19p resulted in dramatically less co-purification of Ape1p in the 2ϫ mutant than in wild-type, a defect that was even stronger in the 11ϫ mutant (Fig. 5D)

DISCUSSION
We have shown that the cargo receptor of the Cvt trafficking pathway in yeast, Atg19p, is both ubiquitinated and deubiquitinated in vivo, and our data suggest that both modifications are required for fully efficient Ape1p processing. Two models are consistent with these findings. First, ubiquitin could serve as a degradation signal to either down-regulate Cvt trafficking or remove excess Atg19p by targeting it to the proteasome. Alternatively, ubiquitination and deubiquitination of Atg19p could play a structural or mechanistic role in the normal progression of the Cvt pathway.
Does Ubiquitination of Atg19p Serve as a Signal for Degradation by the Proteasome?-Ubiquitination of Atg19p occurs on two lysine residues, Lys 213 and Lys 216 . No doubly-modified peptides were isolated, suggesting that modification of these residues is mutually exclusive. Lys 48 -based polyubiquitin chains were present. These results indicate that the series of bands we observed in Atg19p pull-down experiments represents one, two, three, four, or more ubiquitin molecules attached to a single lysine residue of Atg19p and conjugated to each other via Lys 48 of ubiquitin.
Lys 48 -based polyubiquitin chains typically constitute a degradation signal, targeting a substrate to the proteasome for proteolysis (51,52). Two pieces of evidence suggest that this is not the case for Atg19p. First, Atg19p degradation is known to occur in the vacuole and to be dependent on the Cvt/macroautophagy pathways (4,13). Second, mutation of the relevant lysine residues to arginine, which retains the positive charge of lysine but is unable to serve as a ubiquitin attachment site, decreases the interaction of Atg19p with Ape1p and thus, at least in principle, the ability of Atg19p to function in the Cvt pathway. It is possible that the mutations have unintended effects not related to ubiquitination, but Lys 213 and Lys 216 are not in a region of Atg19p known to be involved in any of its protein-protein interactions (4), and the mutations to arginine are conservative substitutions.
There is precedent for a Lys 48 -linked polyubiquitin tag modifying the activity of a protein without targeting it for degradation. Under some conditions, transcriptional activator Met4p in yeast is inacti- , with (ϩ) or without (Ϫ) a Myc epitope tag on ubiquitin. Cells were grown in selective medium to early log phase, and plasmidborne ubiquitin and Atg19p expression were induced with 100 M CuSO 4 for an additional 3.5 h before harvesting in midlog phase. Protein A-Atg19p was isolated by affinity for IgG-agarose, and Myc-ubiquitin conjugates were detected by anti-Myc immunoblot. Expression of the K48R mutant ubiquitin was lower than its parallel WT construct, so loadings were normalized for total Myc-ubiquitin staining (not shown).
vated, but not destabilized, by Lys 48 -linked polyubiquitination (53,54). The mechanism by which ubiquitinated Met4p is protected from proteasomal degradation remains unclear, but the data presented here suggest that Atg19p may provide a second example for study.
The possibility remains that a small fraction of Atg19p, for example a cytosolic pool not associated with cargo proteins, could be destroyed by the proteasome and that this is the purpose of Atg19p polyubiquitination. To be consistent with our finding that mutation of the ubiquitin acceptor sites of Atg19p impairs interaction with Ape1p, this model would require that excess Atg19p has the effect of reducing the efficiency of the Cvt pathway, and its normal clearance by the proteasome is beneficial. However, we found that overexpression of protein A-tagged Atg19p does not impair Ape1p maturation ( Fig. 2A). Clarification of whether ubiquitinated Atg19p is degraded by the proteasome will require further study, but an alternative model exists that is consistent with the data so far.

Ubiquitination of Atg19p May Be a Functional
Step in the Cvt Pathway-Ubiquitin modification of Atg19p and its subsequent deubiquitination may be required for normal progression along the Cvt pathway. For example, attachment of ubiquitin may facilitate the interaction of Atg19p with cargo proteins in the assembling Cvt complex, as is suggested by the impaired interaction of the Lys to Arg Atg19p mutants with cargo protein Ape1p (Fig. 5D). Subsequent deubiquitination may in turn facilitate a later step in the pathway, such as association with the preautophagosomal structure. In this case, impairment of either ubiquitination or deubiquitination of Atg19p would be expected to decrease Cvt pathway function, as we observed. Reversible ubiquitination is an important regulatory mechanism for proteins such as histone H2B, proliferating cell nuclear antigen, and Fancomi's anemia protein FANCD2 (55,56).
A New Role for the Ubp3p-Bre5p Dub Complex-Some of the 17 Dubs encoded in the yeast genome have been implicated in specific roles. For example, Doa4p removes ubiquitin from substrate fragments after deg-  4 . Protein A-Atg19p was isolated by affinity for IgG-agarose beads, and Myc-ubiquitin was detected by anti-Myc immunoblot. B, ubiquitination is not an artifact of overexpression or of the protein A tag. atg19-⌬ cells expressed ubiquitin with or without a Myc epitope tag, as in A, together with Atg19p (with WT, 2ϫ mut, or 11ϫ mut sequence) tagged at its C terminus with a triple HA epitope and expressed from its own promoter on a centromeric plasmid. Atg19p-HA was immunoprecipitated with anti-HA antibodies conjugated to agarose (Santa Cruz Biotechnology). Ubiquitination of Atg19p is shown both by the increase in size of conjugates caused by the Myc epitope tag (left, lane 2 versus lane 1) and by the reactivity of conjugates in an anti-Myc immunoblot (right). A nonspecific background band (*) in the anti-Myc immunoblot is indicated. C, Ape1p maturation is delayed in the 2ϫ and 11ϫ mutants. Lysates of the cell cultures used in B, carrying HA-tagged Atg19p (with WT, 2ϫ mut, or 11ϫ mut sequence), together with Myc-tagged ubiquitin were separated by SDS-PAGE and probed by anti-Ape1p immunoblot. Precursor (prApe1p) and mature (mApe1p) forms of Ape1p are indicated. D, copurification of prApe1p with Atg19p is impaired in the 2ϫ and 11ϫ mutants. Wild-type or ape1-⌬ cells carrying an empty vector (none) or protein A-tagged Atg19p with wild-type (WT), 2ϫ mutant (2ϫ), or 11ϫ mutant (11ϫ) sequence were grown to early log phase in selective medium, and expression of protein A-Atg19p was induced with 100 M CuSO 4 for 3.5 h before cell harvest in midlog phase. Protein A-Atg19p was isolated by affinity for IgG-agarose beads, and co-purifying Ape1p was detected by anti-Ape1p immunoblot. Loadings were normalized for protein A-Atg19p. A nonspecific background band (*) is indicated, as are protein A-Atg19p and prApe1p. mApe1p is not detectable in this experiment because of its co-migration with the IgG used for purification. radation and from membrane proteins whose ubiquitination targets them for endocytosis and degradation in the vacuole. Deletion of DOA4 leads to a depletion of free ubiquitin pools and various pleiotropic effects (57)(58)(59)(60)(61). Ubp14p disassembles free polyubiquitin chains (62). Ubp8p, in association with histone acetylase complexes SAGA and SLIK, deubiquitinates histone H2B (63,64). Ubp6p associates with the proteasome, where it removes ubiquitin from substrates before their degradation, allowing ubiquitin to be recycled (65,66). Ubp1p is thought to target an as yet unidentified component of the endocytic pathway, since overexpression of a soluble Ubp1p isoform stabilizes at least two cargo proteins of this pathway, Ste6p and Ste2p (67).
A number of previous studies have focused on Ubp3p and its cofactor Bre5p. Ubp3p interacts physically with the chromatin silencing protein Sir4p, and ubp3-⌬ strains show enhanced silencing, but the molecular mechanism by which Ubp3p antagonizes silencing is unknown (68). Ubp3p and its cofactor Bre5p co-purify with transcription factor TFIID (69). Two TATA-binding protein-associated factors, Taf1p and Taf5p, are ubiquitinated, and the pattern of their ubiquitination is altered in a bre5-⌬ strain (and presumably in a ubp3-⌬ strain). The function of these modifications, and thus of Ubp3p and Bre5p in the TFIID complex, is not known (69). Others have suggested that Ubp3p may be a general "proofreader" for the ubiquitin-proteasome system, removing ubiquitin from damaged or misfolded proteins before they can be destroyed (45, 70). Deletion of UBP3 would thus increase the rate of degradation of proteasome substrates. Studies of the Ubp3p substrates Sec23p and ␤Ј-COP are consistent with this model; loss of Ubp3p or its cofactor Bre5p decreases both the level and the activity of these substrates (43,44). In contrast, the deletion of UBP3 stabilizes Ste7p, a kinase involved in the pheromone response pathway, leading to an increase in both the level of ubiquitinated Ste7p protein and Ste7p activity (46).
The work presented here adds a new role for the Ubp3p-Bre5p complex, the removal of a polyubiquitin modification from Atg19p. This activity appears to be unrelated to proteasomal degradation, instead serving to modify the activity or localization of Atg19p in the Cvt pathway.