Pleiotropic Effects of Ubp6 Loss on Drug Sensitivities and Yeast Prion Are Due to Depletion of the Free Ubiquitin Pool*

Mutation of the mouse Usp14 gene, encoding the homolog of yeast deubiquitinating enzyme Ubp6, causes ataxia. Here we show that deletion of the UBP6 gene in Saccharomyces cerevisiae causes sensitivity to a broad range of toxic compounds and antagonizes phenotypic expression and de novo induction of the yeast prion [PSI+], a functionally defective self-perpetuating isoform of the translation termination factor Sup35. Conversely, overexpression of ubiquitin (Ub) increases phenotypic expression and induction of [PSI+] in the wild type cells and suppresses all tested ubp6Δ defects, indicating that they are primarily due to depletion of cellular Ub levels. Several lines of evidence suggest that Ubp6 functions on the proteasome. First, Ub levels in the ubp6Δ cells can be partly restored by proteasome inhibitors, suggesting that deletion of Ubp6 decreases Ub levels by increasing proteasome-dependent degradation of Ub. Second, fluorescence microscopy analysis shows that Ubp6-GFP fusion protein is localized to the nucleus of yeast cell, as are most proteasomes. Third, the N-terminal Ub-like domain, although it is not required for nuclear localization of Ubp6, targets Ubp6 to the proteasome and cannot be functionally replaced by Ub. The human ortholog of Ubp6, USP14, probably plays a similar role in higher eukaryotes, since it fully compensates for ubp6Δ defects and binds to the yeast proteasome. These data link the Ub system to prion expression and propagation and have broad implications for other neuronal inclusion body diseases.

The degradation of abnormal or damaged proteins occurs, at least in part, via the ubiquitin (Ub) 1 -proteasome pathway (1). Ub conjugates are found associated with intracellular deposits of aggregated protein in neurons in several amyloidoses and other aggregation-associated neurodegenerative diseases such as Alzheimer's, Parkinson's, and prion diseases (2,3). This, as well as a growing body of new evidence, suggests a linkage between Ub-proteasome system malfunction and amyloid-associated pathogenesis (4). Furthermore, it appears that protein aggregates can directly impair the function of the Ub-proteasome system (5).
The Ub-proteasome pathway of protein degradation regulates a wide variety of biological processes (1). Targeting of a protein for degradation to the proteasome requires conjugation of Ub through an isopeptide bond joining the C-terminal glycine of Ub to a side chain amino group of lysine in the target protein. A polymeric chain of Ub monomers is then assembled by linkage of the lysine residue at position 48 (Lys 48 ) of the proximal Ub to the C-terminal glycine of another Ub. This reaction requires the sequential action of three groups of enzymes: Ub-activating enzyme (E1), Ub-conjugating enzyme (E2), and/or Ub-ligase (E3) (1). Other factors like E4 may participate in elongation of the poly(Ub) chains (6). Ubiquitination is a reversible process. Deubiquitinating (DUB) enzymes are thiol proteases that specifically disassemble Ub conjugates releasing free Ub (7). Classically, DUB enzymes can be divided in two protein families: Ub-C-terminal hydrolases (UCH) and Ubspecific processing proteases (UBP) (8). Human UBP enzymes are also named Ub-specific proteases (USP).
Ubiquitinated substrates are bound to the 19 S regulatory particle of the 26 S proteasome, unfolded, and translocated into the lumen of the 20 S proteasome core, where the protein is degraded by peptidases (9). At some point in this process, the poly(Ub) targeting signal has to be detached from the substrate. Several DUB activities are known to be associated with the proteasome (10 -14). One of them, UCH37, disassembles poly(Ub) chains from the distal end, shortening it such that the attached protein can be released from the proteasome if there is a delay in efficient degradation. Two recent independent studies (15,16) have shown that Rpn11/Poh1, which is a component of the proteasome lid, is a metalloprotease possessing the deubiquitinating activity required for protein degradation (10). Another DUB enzyme associated with the proteasome in mammal is USP14 and its yeast homolog Ubp6. The observation that labeling of USP14 with active site-directed probes is increased upon proteasome inhibition and that proteasome binding increases activity of Ubp6 at about 300-fold suggests functional coupling between USP14/Ubp6 activities and the proteasome (11,12).
Although little is known about the biological roles or substrates of USP14, depletion of its homolog in mice causes defects in synaptic transmission (17). In this work, we characterized the consequences of the depletion of yeast homolog of USP14, Ubp6. To better define the role of Ub-proteasome system in the pathogenesis of neurodegenerative disease, we have tested the influence of Ubp6 depletion on yeast prion [PSI ϩ ] (an aggregated amyloidogenic isoform of the translation termination factor Sup35), which serves as a model for mammalian amyloidoses (18). We show that ubp6 deletion phenotypes parallel a subset of those caused by disruption of another Ubpencoding gene, doa4, and are largely due to depletion of cellular pools of Ub. We also show that binding of Ubp6 to the proteasome via the Ubl domain is required for biological function in vivo and for maintaining Ub pools but not for the localization of Ubp6 to the nucleus of yeast cell. Human USP14 fully complements yeast Ubp6 deletion phenotypes, demonstrating that USP14 and Ubp6 are not only structural but also functional homologs.

EXPERIMENTAL PROCEDURES
Strains, Plasmids, Media, and Genetic Techniques-Saccharomyces cerevisiae strains used in this study are listed in Table I. Strains of the MHY series (gift from Dr. M. Hochstrasser) are isogenic derivatives of MHY501 with UBP6, DOA4, or UBP14 deleted (19). GT81-1C and GT81-1D, isogenic haploid strains that derived from self-homozygous diploid GT81 (20), contain the [PSI ϩ ]-suppressible marker ade1-14 (UGA) (21). GT159 (22) and GT282 (this study) are derivatives of GT81-1C and GT81-1D, respectively, obtained via curing of [ (23) of D1142-1A series. Yeast rich organic (YPD) and synthetic (SD) media and standard genetic analysis procedures were as described (24). The standard Escherichia coli strains DH5␣ and BL-21 (DE3) and standard procedures for recombinant DNA construction and analysis were used (25). Yeast expression vectors (gift from Dr. J. Fridovich-Keil) pMM2 (CEN TRP1) and pYEpGAP (CEN URA3), bearing the general amino acid permease (GAP) promoter, were used for expressing the Ubp6 and USP14 proteins. The plasmid pUB39 and derivatives carrying the alleles of the Ub-encoding gene under control of the CUP promoter with Lys residues at different positions mutated to Arg (gift from Dr. D. Finley) (26) as well as the plasmid YEp24-UBI4 carrying the Ub gene under the control of the indigenous promoter (gift from Dr. S. Lemmon) were used for Ub expression in yeast. The plasmid pRSETB (Invitrogen) was used for protein expression in E. coli. The CEN URA3 plasmids CEN-GAL-SUP35 (27) and pRS316GAL-SUP35N (28) bear the complete SUP35 gene and N-prox-imal prion-forming domain of SUP35 (SUP35N), respectively, under the control of the galactose-inducible (GAL) promoter. HIS3 plasmids pLA1-SUP35N was constructed by G. Newnam and M. DelMazo in the Chernoff laboratory via insertion of a 0.4-kb SmaI-EcoRI fragment of CEN-GAL-SUP35 into pLA1 (29), cut with BamHI and blunted.
Gene Disruption and Tagging-Disruptions and tagged derivatives of the yeast chromosomal genes were created via direct PCR-mediated gene transplacement, as described by Longtine et al. (30). The following modules were used: pFA6a-GFP(S65T)-HIS3MX6 with primers P1F and P1R (Table II) for the C-terminal tagging with GFP (S65T); pFA6a-His3MX6 for UBP6 (primers P2F and P1R) and DOA4 (primers P3F and P3R) disruptions with the Schizosaccharomyces pombe homolog of the HIS3 gene (his5 ϩ ); and pFA6a-kanMX6 for UBP14 (primers P4F and P4R) and PDR5 (primers P5F and P5R) disruptions with the Kan R gene causing resistance to G418. In each case, the entire coding portion of the corresponding gene was eliminated in the process of transplacement. His ϩ or G418 R transformants were selected, and disruptions were confirmed by PCR.
Plasmid Constructions-The UBP6 coding sequence was PCR-amplified with primers P6F and P6R (Table II; start and stop codons are underlined) from S. cerevisiae genomic DNA and cloned into pCR®2.1 (Invitrogen) according to the manufacturer's protocol, generating pWTD28. The C118S mutation was introduced into UBP6 open reading frame in pWTD28 by using the Transformer site-directed mutagenesis kit (Clontech) with oligodeoxynucleotide P7M, corresponding to the antisense strand of UBP6, as a mutagenic primer (base pairs introducing C118S change are underlined). P7S, used as a selection primer, corresponds to the sense strand of pWTD28 and introduced a C to T (underlined) base pair change in the NcoI site of pWTD28. The resulting plasmid was named pWTD60. The UBP6 open reading frames from pWTD28 and pWTD60 were subcloned into EcoRI-SalI-digested pMM2 and pYEpGAP under the control of the GAP promoter. To clone the Ubl domain, the N-terminal 78 codons of UBP6 were amplified by PCR using primers P8F and P8R (start and stop codons underlined). The PCR product was digested with EcoRI and SalI and subcloned into pMM2 and pYEpGAP. Deletion of the Ubl domain of UBP6 was accomplished by using PCR with the primer P9F, corresponding to the 5Ј-end of the UBP6 coding region, in which the nucleotides coding for the first (underlined) and second amino acids of the UBP6 open reading frame are followed by nucleotides coding for residues 76 -83. The reverse primer was P6R that is complementary to the sense strand of the C-terminal end of UBP6 (stop codon is underlined). Plasmids pWTD28 and pWTD60 were used as templates. PCR products were subcloned into EcoRI and SalI restriction sites of pMM2, generating pMM2ubp6 [ubl⌬] and pMM2C118Subp6 [ubl⌬] To replace the Ubl domain of Ubp6 by Ub, 71 codons of Ub-encoding sequence from the yeast genome were amplified with P20F as a forward primer (start codon underlined) and P20R as a reverse primer. Both primers introduce EcoRI site. The PCR product was digested with EcoRI and cloned in pMM2ubp6 [ubl⌬] , generating pMM2ub-ubp6 [ubl⌬] . To create the C-terminal GFP fusion, UBP6 and ubp6 [ubl⌬] were amplified by PCR with respective forward primers P10F and P11F (start codon underlined) and the same reverse primer P10R, containing an XhoI site (underlined) substituting for the stop codon of UBP6. Plasmids pWTD28 and pMM2 ubp6 [ubl⌬] were used as a template. PCR products were fused in frame with sGFP (S65T, V163A) in pAC45 (31) using PstI and XhoI sites. For all of the USP14 constructs, plasmid pRSETBUSP14 was used as a source (32). To create the USP14 truncated mutants, the silent mutation generating a second NdeI site in pRSETBUSP14 was introduced at one of the following nucleotide positions beginning from the ATG codon in the USP14 open reading frame, 221, 280, 356, or 383, by using the site-directed mutagenesis kit from Clontech. Corresponding mutagenic primers P12M, P13M, P14M, and P15M and the selection primer P12S, eliminating the unique XmnI site in pRSETB, were used. Nucleotides introduced by mutagenesis in USP14 and pRSETB sequences are underlined. The mutagenized plasmids were subjected to the NdeI digestion, which produced a fragment between the NdeI site at position Ϫ1 and the newly introduced NdeI site for each mutant. The plasmids were religated, generating the following truncations: pRSETBusp14 [ubl⌬46] , pRSETBusp14 [ubl⌬66] , pRSETBusp14 [ubl⌬90] , and pRSETBusp14 [ubl⌬100] , respectively. To replace the N-terminal 66 amino acids of USP14 by Ub, 68 codons of the Ub-encoding gene, cloned in pRSETB, were amplified by PCR with P16F as a forward primer (start codon underlined) and P16R primers 3Ј as a reverse primer introducing a second NdeI site (underlined). PCR product was cloned in the NdeI site of pRSETBusp14 [ubl⌬66] , generating the plasmid pRSETBub-usp14 [ubl⌬66] . The correct orientation of ubiquitin sequence in this plasmid was confirmed by using NdeI/NcoI digestion. For expression in yeast, the USP14 gene was amplified by PCR and cloned into EcoRI and SalI sites of pMM2 and pYEpGAP. P17F and P17R were used as a forward and reverse primer (start and stop codons of USP14 are underlined). For cloning USP14 deletion mutants, the series of pRSETB-USP14 with a second NdeI site introduced at position 221, 280, 356, or 383 were used as templates. Forward primer P18F, introducing an EcoRI site right before the first NdeI site (underlined start codon is a part of the NdeI site), and P17R as a reverse primer (stop codon is underlined) were used for PCR. PCR products were cloned in pCR®2.1. The resulting plasmids were digested with NdeI to remove the USP14 sequence between the two NdeI sites, religated, and digested again with EcoRI and SalI to subclone into pMM2 and pYEpGAP. For cloning ub-usp14 [ubl⌬66] , pRSETBub-usp14 was used as a template for PCR with P19F and P19R primers (start and stop codons are underlined). PCR product was cloned into SacI and XhoI sites of pMM2 and pYEpGAP. All of the new constructs made by using the PCR approach were verified by dideoxy DNA sequencing.
Assays for [PSI ϩ ]-The presence of the yeast prion [PSI ϩ ], an aggregated partially inactive isoform of the translation termination factor Sup35, was monitored by its ability to cause a read-through of the nonsense alleles, such as ade1-14 (UGA), resulting in growth on the corresponding omission synthetic medium (in the case of ade1-14, on the medium lacking adenine) (21). Suppression of ade1-14 also inhibits accumulation of the red pigment by yeast cells, resulting in lighter color on the complete (YPD) medium. Other nonsense-suppressible markers, such as trp5-48 (UAA), lys2-187 (UGA), and leu2-1 (UAA) were also used. De novo formation of [PSI ϩ ] in the [psi Ϫ ] cells containing another prion [RNQ ϩ ] was induced by transforming these cells with the plasmids bearing the SUP35 gene or its N-proximal fragment, coding for the prion-forming domain (SUP35N), under the control of the GAL promoter. Resulting transformants were replica-plated onto galactose medium selective for the plasmid in order to induce the GAL promoter. After 2-3 days of incubation, transformants were replica-plated onto the synthetic glucose medium lacking adenine (where GAL promoter is turned off) in order to detect [PSI ϩ ] induction leading to suppression of ade1-14 (21). To analyze the de novo [PSI ϩ ] induction, yeast cultures bearing the GAL-SUP35 (or GAL-SUP35N) constructs were grown in the liquid medium, containing galactose and raffinose (instead of glucose) in order to induce GAL promoter. After various periods of time, aliquots were taken and plated onto glucose ϪAde medium, in order to check for Ade ϩ (i.e. [PSI ϩ ]) colonies. Dilutions of the same culture were also plated onto glucose adenine-containing medium selective for the plasmid(s) in order to determine the numbers of plasmid-containing cells. To prove that de novo induced Ade ϩ colonies indeed contain [PSI ϩ ], a sample of them have been checked for their ability to lose the Ade ϩ phenotype upon growth on the medium containing 5 mM guanidine HCl, a chemical known to eliminate yeast prions (21).
Yeast Proteasome Fractionation-Yeast cells, grown in 500 ml of minimal medium to an A 600 of 1.0, were collected by centrifugation, washed, resuspended in the lysis buffer (50 mM Tris-HCl, pH 7.5, 2 mM MgCl 2 , 10% glycerol, 2 mM ATP, and 1 mM dithiothreitol) containing protease inhibitors (Roche Applied Science) and broken by vortexing with glass beads. The lysate was clarified by spinning at 8,000 ϫ g for 30 min and fractionated on the Amersham Biosciences Superose 6HR column, Bio-Rad BioLogic Duo-Flow Chromatography System. Protein fractions were assayed for proteasome activity by using succinyl-leuleu-val-tyr-amidomethyl coumarin as described previously (33) and subjected to Western blot analysis with anti-Ubp6 Ab (12) or anti-Usp14 Ab (11).
Localization of Ubp6-GFP-The Ubp6-GFP fusion proteins were localized by directly viewing the GFP signal in living cells through a GFP-optimized filter (Chroma Technology, Brattleboro, VT) using the Olympus (Tokyo, Japan) BX60 epifluorescence microscope equipped with a Photometrics (Tucson, AZ) Quantrix digital camera. The nuclei of living cells were stained with 4Ј,6-diamidino-2-phenylindole and visualized in the UV channel.
Measurement of Ubiquitin Half-life and Immunoblot Analysis-Measurement of Ub half-life was performed as described previously (34) with a few modifications. Cycloheximide (50 g/ml) was added to cells grown in minimal medium to an A 600 of 1.0. Cells were split into two flasks, and MG132 (100 m) in Me 2 SO was added to one flask while an equal volume of Me 2 SO was added to another. Extracts were made from equal aliquots of cells taken at the desired time points by heating for 10 min at 100°C in SDS gel-loading buffer. After centrifugation, samples were run on the 16% Tricine gels and transferred to nitrocellulose membrane (Bio-Rad). Blots were boiled in water for 15 min and incubated with a 1:500 dilution of anti-Ub antibody P4D1 (Santa Cruz Biotechnology, Inc., Santa Cruz, CA). Antibody binding was detected by using the ECL reagents from Amersham Biosciences. The membrane were subsequently stripped and probed with an anti-3-phosphoglycerate kinase (Pgk1) antibody as a loading control (Molecular Probes, Inc., Eugene, OR).
Purification of USP14 from E. coli Cells-To induce expression of USP14 in E. coli BL-21 (DE3), bacterial cells transformed with pRSETB-USP14 were grown to an A 600 of 0.8 in LB plus 100 g/ml ampicillin at 30°C. The cells (1.5 liters) were collected by centrifugation, resuspended in the same amount of fresh medium, and heatshocked at 42°C for 30 min. After the addition of 10 M isopropyl-␤-Dthiogalactopyranoside, followed by incubation at 15°C for 48 h, cells were pelleted, resuspended in 150 ml of lysis buffer (50 mM Tris-Cl, pH 8.0, 25 mM EDTA, 10 mM ␤-mercaptoethanol, 100 g/ml lysozyme, and protease inhibitor mixture from Roche Applied Science) and sonicated. Cell debris was removed by centrifugation at 14,000 ϫ g for 30 min. The supernatant was applied to the FPLC Fast Flow Q anionic exchange column and eluted in a linear gradient of 0 -500 mM NaCl. Fractions containing USP14 as judged by SDS-PAGE were pooled, concentrated using the YM30 Ultrafiltration membrane to a final volume of 10 ml, applied to a Sepharyl-S100 gel filtration column, and eluted with 1 M NaCl. The pooled fractions were dialyzed against 10 mM Tris-Cl, pH 7.6, loaded on a Mono Q anionic exchange column, and eluted in a linear gradient of 0 -500 mM NaCl. Fractions containing homogeneous USP14 as judged by Coomassie-stained SDS-PAGE were pooled. Truncated proteins were purified in the same manner as full-length USP14.
Assay for USP14 Enzymatic Activity-Assays for Ub carboxyl-terminal hydrolase activity were performed using the C-terminal ethyl ester of Ub as substrate (35,36). Assays were done in triplicate. The peak areas were integrated and normalized with respect to a Ub standard.

Deletion of UBP6 Reduces the Expression and Formation of the Yeast Prion [PSI
ϩ ]-Mutation in the gene encoding a deubiquitinating enzyme, Usp14, results in ataxia in mice. In humans, a number of ataxias are caused by mutations leading to abnormal protein deposition in neurons (37). In addition, other aggregation-related diseases in mammals are associated with the defects of the Ub system (4). We have used the model system, yeast S. cerevisiae, to ask whether deletion of UBP6, the yeast homolog of USP14, has any effect on a similar system of protein aggregation. The yeast prion [PSI ϩ ] is a self-perpetuating aggregated isoform of the yeast translation termination factor Sup35 (eRF3) and has been widely used as a model to study amyloidoses and aggregation-related disorders (18).
To study the possible effects of the Ub-proteasome pathway on [PSI ϩ ], we deleted the UBP6 gene in both [PSI ϩ ] (GT81-1C) and isogenic [psi Ϫ ] (GT159) strains and characterized the phenotypes of the deletion derivatives. The [PSI ϩ ] strains are defective in termination of translation due to partial inactivation of the translation termination factor by aggregation (18). This results in read-through of the stop codons (nonsense suppression). This read-through is detected, for example, by partial restoration of the Ade ϩ phenotype in the strain containing a UGA mutation (ade1-14) in the reporter gene ADE1 (21) and could be used as a measure of phenotypic expression of [PSI ϩ ]. Fig. 1 shows that nonsense suppression of ade1-14 is decreased in the [PSI ϩ ] strain lacking Ubp6, compared with the isogenic wild type strain GT81-1C (Fig. 1A). Reintroduction of the wild type UBP6 gene restored suppression (not shown). One possibility is that the effects of ubp6⌬ on nonsense suppression are explained by a decrease of Sup35 conversion into the prion state. If this is true, we might also expect to see a decreased rate of spontaneous formation of the [ Deletion of UBP6 Alters Susceptibility to a Broad Range of Drugs-To examine the mechanisms by which Ubp6 affects prion formation, we screened deletion mutants for a variety of phenotypes. It was shown previously that ubp6⌬ causes sensitivity to arginine analog canavanine (38) and translation inhibitors anisomycin, chloramphenicol, and cycloheximide (39). Deletion of UBP6 in both MHY501 and GT81 strains gave similar results. Each deletion exhibited an increased sensitivity to translation inhibitor trichodermin and anti-tumor agent methotrexate (Table III). There was a slight but reproducible increase in sensitivity to cadmium, 4-nitroquinoline-N-oxide, and methylmethanesulfonate. We also noticed that the ubp6⌬ mutant grew slightly faster on YPD plates at 14°C and at a slightly slower rate at 37°C than wild type cells (not shown).
The ubp6⌬ Phenotypes Are Caused by Decreased Levels of Free Ub-The ubp6⌬ mutation in our strains exhibited reduced levels of both Ub-protein conjugates and free Ub ( Fig. 2A), in agreement with previous observations made in different genetic backgrounds (19). To check whether the effects of ubp6⌬ on [PSI ϩ ] and antibiotic sensitivities are due to depletion of free Ub, we transformed ubp6⌬ strains with a multicopy plasmid encoding the UBI4 gene that restores levels of free Ub (Fig.  2B). Restoration of free Ub levels complemented all drug sensitivities ( Fig. 2C and Table III), and restored nonsense suppression in the ubp6⌬ derivative of the [PSI ϩ ] strain GT81-1C (Fig. 2D). This demonstrates that the majority of ubp6⌬ phenotypes, including the defect in [PSI ϩ ]-mediated suppression, are due to decrease in levels of free Ub.
We also analyzed sensitivity to these drugs in two other Ubp mutants that alter Ub pools, doa4⌬ (MHY623, WTY191, WTY192) and ubp14⌬ (MHY839). DOA4 deletion strains show reduced levels of free Ub (see Fig. 2A and Ref. 40) and demonstrate sensitivity to anisomycin, chloramphenicol, cycloheximide, trichodermin, and methotrexate, although to a slightly lower extent than ubp6⌬, in addition to the earlier reported (40) sensitivity to canavanine and cadmium (data not shown). UBP14 deletion strains show increased levels of ubiquitin conjugates but normal levels of free ubiquitin (41). Of all of the tested drugs, ubp14⌬ was sensitive only to canavanine (data not shown).
Compensation of ubp6⌬ Phenotypes by Mutant Ub-Ub is often found in poly(Ub) chains. Targeting of misfolded proteins for degradation by the proteasome usually requires polyubiquitination, whereas nondegradative modification of proteins is often limited to monoubiquitination (42) or formation of non-Lys 48 -linked chains (43). All seven lysine residues of the Ub have been shown to be capable of forming chains in vitro, and chains linked through lysines 11, 29, 48, and 63 have been detected in vivo (44). Mutation of Lys 48 to Arg is lethal in the absence of wild type Ub, whereas the corresponding mutations of Lys 29 , Lys 63 , and Lys 11 are not (26). There are no known phenotypic defects associated with Ub K29R and K11R mutation, but Lys 63 -linked Ub polymers have been implicated in tolerance to stress, DNA damage, and inhibition of translation (45)(46)(47)(48).
To define potentially important chain linkages, we examined the ability of overexpressed Ub Lys to Arg mutants to suppress the phenotypic defects of ubp6⌬. Since the mutants are expressed in cells that also make wild type Ub, these derivatives probably act as chain terminators. For instance, incorporation of K63R Ub into a poly(Ub) chain would prevent the extension of this chain through Lys 63 . All Ub mutants were found to restore the level of free Ub, and all but K48R increased the level of Ub-protein conjugates (Fig. 2B). Overexpression of the K29R and K11R mutant complemented all of the ubp6⌬ phenotypes tested (Fig. 2C), whereas the K63R mutant failed to suppress sensitivities to canavanine and anisomycin. In contrast, the K48R mutant did not compensate for the defects in [PSI ϩ ]-mediated suppression and de novo [PSI ϩ ] induction (Fig. 2D). These data suggest that at least some ubp6⌬ phenotypes are mediated by defects in assembly of the poly(Ub) chains, and those effects on [PSI ϩ ] may be mediated via Lys 48linked chains.
Free Ub Levels Modulate [PSI ϩ ]-dependent Phenotypes in the UBP6 ϩ Cells-If the observed ubp6⌬ effects on [PSI ϩ ] are due to decreases in free Ub levels, then increased levels of free Ub in the wild type cells might have an opposite effect on these [PSI ϩ ]-associated phenotypes. Indeed, this was the case. Overexpression of Ub under control of the CUP1 promoter (Fig. 3A) or from its own promoter on a multicopy plasmid (not shown) increased suppression of ade1-14 (UGA) in the strain GT81-1C. The same effect was observed in the other genotypic backgrounds, both with the same ade1-14 reporter (strain OT55) and with the other nonsense reporters such as lys2-187 (UGA), leu2-1 (UAA), and trp5-48 (UAA) in the strains GT25 and GT26 (data not shown).
To check whether polyubiquitination is important for the effects of excess Ub on [PSI ϩ ], we tested the K48R and K63R Ub mutants for their ability to increase [PSI ϩ ]-mediated suppression. Whereas the K63R mutant increased suppression to the same extent as wild type Ub, the K48R mutant did not (Fig.  3A). This suggests that effects of excess Ub on [PSI ϩ ] involve Lys 48 -dependent polyubiquitination.
Excess Ub also increased de novo [PSI ϩ ] induction by transient overproduction of Sup35 (not shown) or Sup35N (Fig. 3B Among all ubp deletions tested, only doa4⌬ (ubp4⌬) decreases levels of free Ub to a level comparable with, and even lower than, that of ubp6⌬ (19). We have checked whether doa4⌬ affects [PSI ϩ ]-dependent phenotype in the same manner, as does ubp6⌬. The DOA4 gene was deleted in the [PSI ϩ RNQ ϩ ] strain GT81-1C and isogenic [psi Ϫ RNQ ϩ ] strain GT159. doa4⌬ inhibited growth, most severely at 37°C, and exhibited drug sensitivity phenotypes similar to those observed for ubp6⌬. doa4⌬ cells were sensitive to canavanine, cadmium, anisomycin, cycloheximide, trichodermin, and methotrexate (not shown) but to a lesser extent than ubp6⌬. doa4⌬ also inhibited [PSI ϩ ]-mediated suppression of ade1-14 (Fig. 3C) and de novo [PSI ϩ ] induction by transient overproduction of Sup35 or Sup35N (Fig. 3D) even more severely than did ubp6⌬. This confirms that observed ubp6⌬ phenotypes most likely result from depletion of the free Ub pool in the yeast cell.
Both    (Fig. 4). The constructs were transformed into the ubp6⌬ strain MHY821 and tested for their ability to complement the sensitivity to canavanine and anisomycin. Only wild type UBP6 construct was able to restore the ability to grow in the presence of these drugs (Fig. 4). These data demonstrate that the presence of both Ubl and Ubp domains is required for biological function of Ubp6 and that the Ubl domain is unique and cannot be replaced by Ub. The same UBP6 constructs were transformed into wild type strain MHY501 and tested for their effect on the function of wild type Ubp6 by checking growth in the presence of canavanine or anisomycin. Overexpression of wild type UBP6 in the wild type strain did not have any detectable phenotypic effect on wild type strain. In contrast, the inactive allele with a C118S substitution in the active site strongly inhibited growth in the presence of either drug. This dominant negative effect was completely eliminated by deletion of the Ubl domain. Expression of the Ubl domain alone exhibited only a slight inhibitory effect on anisomycin and canavanine sensitivity. This dominant negative phenotype, together with the fact that inhibition was completely eliminated by deletion of the Ubl domain, suggests that Ubp6 functions in the context of a protein complex, with Ubl domain being a targeting domain responsi-ble for localization of Ubp6. Inactive Ubp6 is able to compete with the wild type Ubp6 for this binding mediated by the Ubl domain.
UBP6 Must Be Bound to the Proteasome to Function in Vivo-The most likely site of Ubp6 functioning in the cell is the proteasome. To analyze how the deletion or replacement of Ubl domain by Ub as well as mutation of active site Cys in the Ubp domain affect Ubp6 association with the proteasome, we fractionated yeast proteasome from ubp6⌬ cells expressing the UBP6 mutant constructs described above. Both wild type Ubp6 and C118S mutant were found in proteasome-containing fractions (Fig. 5A, fractions 14 -20), demonstrating that mutational alteration of the active site does not affect binding to the proteasome. At the same time, deletion of Ubl domain or substitution of the Ubl domain with Ub abolished the ability of Ubp6 to bind to the proteasome (Fig. 5A, fractions 30 -38). These data are in an agreement with drug sensitivity data and demonstrate that the Ubp6 active site C118S mutant with the Ubl domain deleted does not affect the phenotype of the wild type strain because it is unable to bind to the proteasome (Fig.  4). Our data confirm that the Ubl domain is responsible for the Ubp6 binding to the proteasome and show that Ub cannot replace the Ubl domain. The majority of the Ubl domain was found not associated with proteasomes (Fig. 5A, fractions 28 -40). This is in agreement with the previous observation that the Ubl domain alone does not bind to the proteasome as

FIG. 2. Overexpression of wild type or mutant ubiquitin with different Lys to Arg substitutions affects the level of free Ub, drug sensitivity, and defect in [PSI ؉ ]-mediated nonsense suppression caused by ubp6⌬. A, aliquots of the ubp⌬ mutants (genotype is indicated)
were normalized by optical density and lysed, and proteins were separated on a Tricine gel followed by Western blotting using a monoclonal antibody against ubiquitin. Free ubiquitin (Ub) is indicated, as are the positions of unanchored multiubiquitin chains. WT, wild type. B, ubp6⌬ (WTY105) cells were transformed with pUB70 (vector) and pUB39 expressing wild type ubiquitin or mutant derivative with the indicated lysine substituted by arginine. Protein lysates from these cells were subjected to anti-ubiquitin immunoblot analysis as described for A. C, 10-fold serial dilutions of the same ubp6⌬ transformants as in B, overexpressing wild type or mutant ubiquitin as indicated, were spotted on the minimal medium with or without drug and grown for 3-5 days at 30°C. D, equal amounts of yeast cultures (genotype is indicated) were spotted onto ϪAde medium (to check for suppression) also lacking lysine (to select for the plasmid LYS2 marker). Growth was documented after 4 days of incubation at 30°C. Strains were as follows: ubp6⌬, WTY105; UBP6 ϩ , GT81-1C. The same result as with pUb39 was observed with YEp24-UBI4, pUb39K29R, and pUb39K63R (not shown). efficiently as a full-length protein (12) and explains why expression of the Ubl domain exhibited only a slight inhibitory effect on the growth of the wild type strain in the presence of drugs (Fig. 4).
The Ubl Domain Is Not Required for Nuclear Localization of Ubp6 in the Yeast Cell-To gather more information about the functional significance of the Ubl domain in vivo, we determined the intracellular localization of the Ubp6 protein. The chromosomal copy of UBP6 in the strain MHY501 was replaced with a copy of UBP6 fused to the N terminus of GFP under control of the endogenous UBP6 promoter, generating the strain WTY43. WTY43 was able to grow on synthetic medium in the presence of canavanine and anisomycin (in contrast to ubp6⌬ derivative of MHY501, which is unable to grow at these conditions; data not shown), indicating that Ubp6-GFP protein is functional. In living cells, the Ubp6-GFP protein was localized in the nucleus (Fig. 5B). Since the mutant phenotype and inability to bind proteasome of ubp6 [ubl⌬] protein could arise from its failure to localize to the nucleus, we examined the cellular location of these protein derivatives. Ubp6 protein with the Ubl domain deleted and wild type Ubp6, to use as a control, were fused to the N terminus of GFP and expressed from the GAL promoter on the episomal plasmid. Remarkably, both full-length Ubp6 and mutant Ubp6 with the Ubl domain deleted localized to the nucleus, demonstrating that nuclear localization of Ubp6 in S. cerevisiae does not require the Ubl domain.
Ub Is Degraded by the Proteasome in the ubp6⌬ Strain-Ub was previously shown to be unstable in ubp6⌬ mutants (12). Levels of free Ub decrease with time in ubp6⌬ cells treated with cycloheximide to inhibit the synthesis of new Ub. To investigate whether ubiquitin was being degraded by the proteasome in ubp6⌬ cells, we analyzed the stability of Ub in cells treated with proteasome inhibitors. Proteasome inhibitors are usually not effective in the wild type yeast cells, apparently because they fail to penetrate the yeast cell wall. However, strains carrying a deletion of the gene encoding major drug efflux pump protein, Pdr5, show significantly increased sensitivity to proteasome inhibitors (49). Therefore, we deleted the PDR5 gene in the ubp6⌬ mutant MHY821, generating strain WTY270. Treatment WTY270 with 50 M MG132 did not have a significant effect on cell growth but resulted in inhibition of

FIG. 3. Ubiquitin levels modulate [PSI ؉ ]-mediated nonsense suppression and de novo [PSI ؉ ] induction. A, excess Ub increases
[PSI ϩ ]-mediated nonsense suppression in the wild type strain GT81-1C. This effect requires Lys 48 linkage, since it is not observed with the K48R Ub mutant (in contrast to K63R). Plasmids are as follows: control, pUb70 (vector), pUb39 (WT Ub), pUbK48R, and pUbK63R. Transformants were replica-plated onto ϪAde (to check for suppression) also lacking lysine (to select for the plasmid), and growth was documented after 5 days at 30°C. The same increase of suppression was observed with the URA3 plasmid YEp24-UBI4 (not shown). B, excess Ub increases the efficiency of [PSI ϩ ] induction by overproduced Sup35N. Transformants of the [psi Ϫ RNQ ϩ ] strain GT159, containing HIS3 plasmid pLA1-SUP35N in combination with either URA3 plasmid YEp24-UBI4, coding for Ub (1Sup35Nϩ1Ub), or control URA3 plasmid pFL44 (1Sup35N), were incubated in the liquid ϪUraϪHis medium containing galactose and raffinose (instead of glucose) in order to induce the GAL promoter. After specified periods of time, aliquots were taken and plated onto ϪUraϪHisϪAde medium to check for Ade ϩ colonies. It has been confirmed by the guanidine HCl curability assay that most of the Ade ϩ colonies originating from induced cultures indeed resulted from de novo formation of [PSI ϩ ]. Dilutions were also plated onto ϪUraϪHis medium to determine total number of plasmid-containing colonies. No induction was seen in transformants bearing the empty plasmid controls, pLA1 and pFL44 (not shown). proteasome activity and accumulation of Ub conjugates. To analyze the effect of proteasome inhibition on the stability of Ub in the ubp6⌬ strain, we compared the level of free Ub in WTY270 cells treated with cycloheximide and MG132 or Me 2 SO as a control. Treatment with MG132 partially prevented loss of Ub and Ub conjugates, suggesting that protea-some activity is responsible for the decreased levels of Ub (Fig. 6A).
Genetic Interactions between Ubp6 and Ubp14 -Ubp6 is a proteasome-associated DUB enzyme, and defects of this enzyme are associated with proteasome-dependent decreases in Ub levels. We hypothesize that Ubp6 is responsible for remov- To find the proteasome-enriched fractions, all fractions were assayed for peptidase activity against succinyl-leu-leuval-tyr-amidomethyl coumarin. Samples from each fraction were subjected to SDS-PAGE followed by immunoblot analysis using anti-Ubp6 antibodies. B, ubp6⌬ cells were transformed with pAC45 encoding Ubp6-GFP or Ubp6 [ubl⌬] -GFP under control of the GAL promoter. Transformants were grown overnight in minimal medium containing raffinose. Expression of fusion proteins was induced by the addition of galactose for 4 h. Ubp6-GFP was viewed directly in living cells.
ing either Ub or poly(Ub) from Ub-protein conjugates before their translocation into the proteolytic core particle of a proteasome. If unanchored poly(Ub) chains are generated during such a process, they could become substrates for Ubp14, another DUB enzyme that has been shown to disassemble unanchored polyubiquitin chains (41), and it would be expected that the Ub conjugate profile of ubp14⌬ mutant would be affected by the addition of ubp6⌬. To analyze possible interaction between Ubp6 and Ubp14, we generated the ubp6⌬ ubp14⌬ double mutant WTY269 and compared the levels of free Ub, Ub conjugates, and resistance to drugs in the single and double mutants (Fig. 6, B and C). As has been shown before (41), deletion of the UBP14 gene results in accumulation of unanchored Ub chains but does not affect the level of free Ub (Fig. 6B). We find that ubp14⌬ does not show any of the drug sensitivities shown by ubp6⌬, with the exception of sensitivity to canavanine (Fig. 6C and data not shown). Consistent with the lack of effect on free Ub pools, ubp14⌬ does not affect suppression by [PSI ϩ ]. 2 The ubp6⌬ ubp14⌬ mutant was slightly more sensitive to canavanine then either single mutant and almost as sensitive to anisomycin as the single ubp6⌬ mutant. Anti-Ub immunoblots of ubp6⌬ ubp14⌬ cell extracts revealed the same pattern of ubiquitinated species as the ubp14⌬ single mutant, but the amount of all conjugates, as well as the amount of free Ub, was reduced. The fact that the phenotype of the double mutant resembles a combined profile of the two single mutants indicates that neither mutation is fully epistatic to the other. Thus, the major substrates for Ubp14 are still generated even in the absence of Ubp6. These results suggest that Ubp6 does not directly generate significant amounts of free poly(Ub) chains.
USP14, the Human Ortholog of Ubp6, Is Fully Functional in Yeast-Human USP14 is 32% identical to yeast Ubp6 over the entire length of the protein. The sequence similarity of the yeast Ubp6 and human USP14 suggests that they may have similar functions. To investigate this possibility, a cDNA copy of the human USP14 transcript was cloned into a yeast expression vector. A mutant of USP14 lacking the Ubl domain and a chimeric USP14 gene in which Ubl domain sequence was replaced by the human Ub sequence were also constructed. The resulting plasmids were transformed into ubp6⌬ strains WTY105, WTY106, and MHY821, and phenotypes of the transformants were characterized (Fig. 7). Based on Western blot analysis, the expression levels of all constructs appeared to be similar to each other and to wild type USP14 (Fig. 7A). Expression of wild type USP14 suppressed the sensitivities to canavanine, anisomycin, methotrexate, and cycloheximide in the ubp6⌬ strain (Fig. 7B and data not shown) and almost completely restored wild type levels of free Ub and Ub conjugates 2 A. Zink and Y. Chernoff, unpublished data. 6. Effects of ubp14⌬ and proteasome inhibitors on ubp6⌬ phenotypes. A, ubp6⌬ pdr5⌬ (WTY270) cells were grown in liquid minimal medium to log phase at 30°C. Cycloheximide was added to inhibit the synthesis of new ubiquitin, and cells were split. Half of the cells were incubated with 0.1% Me 2 SO (DMSO), and the half were incubated with 50 M MG132 (dissolved in 0.1% Me 2 SO). Aliquots, normalized by optical density, were taken at the indicated intervals to prepare cell extract. Proteins were separated on a Tricine gel followed by Western blotting using anti-ubiquitin antibodies. Anti-Pgk antibodies were used as a loading control. B, proteins from double ubp6⌬ ubp14⌬ (WTY269) and single ubp6⌬ (MHY821) and ubp14⌬ (MHY839) mutants were separated on a Tricine gel and subjected to Western blot analysis using a monoclonal anti-ubiquitin antibody and anti-phosphoglycerate kinase (PGK) Ab for the control of protein loading. C, drugs sensitivity of double mutant was compared with single mutants and wild type. 10-Fold serial dilutions of cells of the indicated genotype were spotted on minimal medium with or without drug. (Fig. 7C). USP14 also compensated for the defect in [PSI ϩ ]mediated suppression caused by the ubp6⌬ deletion (not shown). Neither the Ubl deletion nor the chimeric protein was able to restore these functions of ubp6⌬ cells. USP14 protein expressed in ubp6⌬ cells was found in proteasome-enriched fractions (Fig. 7D), whereas USP14 proteins with the Ubl domain deleted or replaced by Ub were not associated with proteasomes (data not shown). These data demonstrate that mammalian USP14 is a functional homolog of Ubp6, able to bind to the yeast proteasome, and that the Ubl domain of mammalian USP14 is also absolutely required for the function.
Deletion of Ubl Domain Does Not Increase Catalytic Activity of USP14 -It has been shown that Ubp6 deubiquitinating activity increases upon its binding to the proteasome (12). Activation of mammalian USP14 by proteasome binding also was demonstrated by the fact that only proteasome-bound USP14 can be labeled with active site-directed probe UbVS (11). One possible explanation is that the Ubl domain acts as a pseudosubstrate occupying the active site of unbound Ubp6 or USP14 and that binding to the proteasome competes for the Ubl domain, making the active site available for the substrate. If this is the case, deletion of the Ubl domain should increase the enzyme's affinity for substrate. It has been shown that Ubp6 with the Ubl domain deleted is catalytically active as a Ub-hydrolase, although the activity was not quantitatively characterized (38). To characterize the enzymatic activity of USP14, we cloned the human USP14 cDNA in a bacterial expression vector and constructed a series of mutants with 66, 90, or 100 N-terminal amino acids deleted (Fig. 8A). The full-length and truncated USP14 proteins were expressed in E. coli and purified to homogeneity. The catalytic activity was measured by using the generic substrate Ub ethyl ester. The purified recombinant proteins all catalyzed the hydrolysis of Ub ethyl ester, albeit with little evidence of saturation by substrate. We estimate that the K m values for wild type USP14, as well as the  7. Human USP14 complements the S. cerevisiae ubp6⌬ phenotype. A, human USP14 wild type and mutant proteins expressed at the same level in yeast. ubp6⌬ (WTY105) cells were transformed with pMM2 plasmid encoding USP14 wild type or mutant protein with the Ubl domain deleted or substituted by ubiquitin as indicated (see "Experimental Procedures" for details). An equal amount of cells were lysed, resolved on 10% SDS gel, and subjected to Western blot analysis using polyclonal anti-USP14 antibody. B, an equal amount of Ubp6 deletion cells transformed with empty vector as a control or with vector expressing wild type or mutant protein indicated were spotted on the minimal medium with anisomycin and grown at 30°C for 3-5 days. C, aliquots of the same transformed cells as in B were normalized by optical density and lysed, and proteins were separated on a Tricine gel followed by Western blotting using anti-ubiquitin antibodies. The position of free ubiquitin is shown. D, protein extracts were prepared from ubp6⌬ cells expressing USP14, separated on a column, and analyzed for the proteasome activity as described in the legend to Fig. 5B. Samples from each fraction were subjected to Western blot analysis using anti-USP14 antibodies. PGK, phosphoglycerate kinase. truncation mutants, are over 200 M. Catalytic efficiency was mildly affected, with deletion mutants retaining 10 -25% of activity (Fig. 8B). This suggests that USP14 and mutants with the Ubl domain deleted exhibit very low affinity for Ub. The fact that removing the Ubl domain did not increase USP14 affinity for Ub or its catalytic activity makes it unlikely that the Ubl domain is acting as a pseudosubstrate. DISCUSSION The work described here was inspired by the observation that a defect of Usp14, a deubiquitinating enzyme bound to the proteasome, leads to ataxia in mice (17). It has recently become clear that disturbances in the Ub-proteasome pathway can have profound effects on protein aggregation (37), degradation of abnormal proteins (4), and apoptotic cell death (50). The presence of ubiquitinated proteins in a variety of neuronal inclusion body diseases is an intriguing observation, although the precise role of the Ub-proteasome system in the etiology of these diseases is still unclear. We have used budding yeast to demonstrate that Ubp6, the yeast homolog of Usp14, has effects on self-perpetuating protein aggregates (prion). We find that the proximal defect caused by deletion of Ubp6 is a reduction in free Ub levels and that this is caused by the degradation of Ub by the proteasome.

Ubp6⌬ Affects Phenotypic Expression and de Novo Formation of the Yeast Prion [PSI ϩ ]-[PSI ϩ
] is a prion isoform of the yeast translational termination factor Sup35 (eRF3). [PSI ϩ ] cells are defective in termination of translation, since most of cellular Sup35 is sequestered by prion aggregates in these cells and is not capable of properly functioning in termination (18). This results in read-through of nonsense codons (nonsense suppression) that is used as a measure of [PSI ϩ ] phenotypic expression. Overexpression of Sup35 or the N-terminal fragment of Sup35 in cells that lack the prion can de novo induce [PSI ϩ ], apparently due to increased aggregation of Sup35, resulting in the conversion of the native conformation to the stably propagated aggregated state.
We have shown that phenotypic expression of [PSI ϩ ] (i.e. nonsense suppression) is decreased in an ubp6⌬ background (Fig. 1A) and in a doa4⌬ background (Fig. 3C), pointing to the partial restoration of the translational termination activity. These phenotypes were compensated by excess Ub (see Fig.  2D), confirming that they are due to depletion of free Ub levels as described below. Moreover, overproduction of Ub in the wild type cells increased nonsense suppression by [PSI ϩ ] (Fig. 3A), confirming the inverse correlation between levels of free Ub and efficiency of translational termination in prion-containing cells.
Since ubiquitination of some ribosome components has been observed previously (48), one could suggest that Ub levels affect translational termination via modification of the translational machinery rather than prion formation and propagation. Although, we have not observed any detectable effect of Ub levels on suppression in the wild type [psi Ϫ ] cells, deletion of DOA4 appeared to decrease nonsense suppression caused by certain suppressor mutations, indicating that its effect on translation is not restricted to mechanisms solely involving [PSI ϩ ]. 2 However, ubp6⌬ (Fig. 1B) and doa4⌬ (Fig. 3D) also decreased de novo induction of [PSI ϩ ] by overproduced Sup35, whereas excess Ub increased it (Fig. 3B). Modification of the translational machinery could in principle explain the inhibitory effects of ubp6⌬ and doa4⌬ on detectable de novo [PSI ϩ ] induction by inhibition of nonsense suppression. In this model, some [PSI ϩ ] isolates with low levels of Sup35 aggregation induced de novo are simply not detectable in the ubp6⌬ (or doa4⌬) background due to inhibition of nonsense suppression. This may lead to the decrease in frequency of detectable [PSI ϩ ] isolates. However, such a suggestion cannot explain a stimulatory effect of excess Ub on [PSI ϩ ] induction, since all of the [PSI ϩ ] strains, induced de novo in the presence of excess Ub, remained [PSI ϩ ] and exhibited detectable nonsense suppressors after elimination of the UBI4 plasmid (see above). Therefore, Ub levels probably influence certain steps of the prion formation and/or propagation cycle, in addition to their effect on nonsense suppression. Our data provide the first evidence for the effect of the Ub system on yeast amyloidogenic proteins and parallel data obtained previously in mammalian cells (51,52).
The precise mechanism by which Ub levels affect the dynamics of the [PSI ϩ ] prion remains unclear. Thus far, we were not able to detect any sizeable fraction of ubiquitinated Sup35 in the [PSI ϩ ] cells. 3 Preliminary data show that ubp6⌬ and doa4⌬ cells contain more Sup35 protein in the soluble fraction compared with the isogenic wild type [PSI ϩ ] cells. 4 This suggests a decrease of the size of prion aggregates in the Ub-depleted cell. Soluble (monomeric or oligomeric) Sup35 protein is probably more functional as a termination factor than is the protein that is included in the insoluble aggregates. This may explain the partial relief of the translational termination defect of [PSI ϩ ] in the ubp6⌬ and doa4⌬ backgrounds. Further investigations are needed to decipher a mechanism by which Ub levels regulate size and/or solubility of the prion aggregates.
The Phenotypic Effects of ubp6⌬ Are Due to Depletion of Ub Levels-Mutations in genes encoding proteins that function in Ub-dependent proteolysis often cause cells to become sensitive to various stress conditions (53,54). In this paper, we report on detailed analysis of the functional consequences of depletion of Ubp6, a yeast proteasome-bound deubiquitinating enzyme. We confirmed the previous observations that deletion of either the UBP6, DOA4, or UBP14 gene leads to sensitivity to the amino acid analogue canavanine (38,40,41) in yeast strains of two different genetic backgrounds. In addition, ubp6⌬ and doa4⌬ mutants also exhibit sensitivities to several other drugs including translation inhibitors, such as anisomycin, chloramphenicol, cycloheximide, trichodermin, and methotrexate, although these sensitivities were more apparent in ubp6⌬ strains (Table  I). The fact that all ubp6⌬ and doa4⌬ drug sensitivity phenotypes tested are compensated by excess Ub demonstrates that they are due to a decrease in levels of free Ub. These phenotypes may be characteristic of all mutations that cause the depletion of Ub pool in yeast cells. Depletion of Ub levels can in principle affect directly or indirectly various cellular processes regulated by mono-or polyubiquitination. One possibility is that effects of ubp6⌬ and doa4⌬ on drug sensitivity are due to alterations of the ABC transporter system when Ub levels are low. Indeed, it has previously been shown that drug hypersensitivity phenotypes exhibited by the gain-of-function mutation in the gene PDR2 (a regulator of ABC transporters) requires functional Ubp6 (39). It is possible that when Ub levels are insufficient, misfolded proteins are accumulated, and this alters the transporter systems.
Different Chain Linkages Are Involved in Different Ubiquitin Functions-Poly(Ub) chains can be formed in vivo by linkages via at least three different lysine residues in the Ub molecule, Lys 29 , Lys 48 , and Lys 63 (45,55,56). Lys 63 -linked Ub polymers have been implicated in tolerance to stress and DNA damage (45,46). In yeast, the Ub K63R mutant failed to suppress canavanine and cadmium sensitivity of doa4⌬ (47), suggesting the involvement of Lys 63 -linked chains in this function. We showed that the Ub K63R mutant failed to suppress canava-nine and cadmium sensitivity of ubp6⌬ as well. Spence et al. (48) demonstrated that the K63R mutation in Ub affects function of the L28 protein in the peptidyl transferase region of the ribosome and creates sensitivity to translational inhibitor anisomycin and resistance to cycloheximide. In agreement with this finding, the Ub K63R mutant was not able to compensate for anisomycin sensitivity of both ubp6⌬ and doa4⌬ but did compensate for sensitivity to cycloheximide. Ub may act in this pathway to help regulate ribosome activity possibly in response to stress conditions. Quite remarkably, the compensatory effect of excess Ub on [PSI ϩ ] suppression in the ubp6⌬ (Fig. 2D) and doa4⌬ (not shown) mutants, as well as the [PSI ϩ ]-stimulating effect of excess Ub in the [PSI ϩ ] cells (Fig. 3A), requires Lys 48 linkage rather than Lys 63 linkage that is involved in the other effects of ubp6⌬. These findings show that different phenotypic defects of ubp6⌬ are mediated by different types of poly(Ub) chains; some functions require Lys 48 -linked chains and presumably proteosomal functions, whereas others may involve targeting pathways requiring Lys 63 -linked chains. The functions of both types of chains are altered in ubp6⌬ and doa4⌬, suggesting that depletion of the free pool of Ub is the proximal event.
Structural and Functional Analysis of Ubp6 -Ubp6 contains two distinctive domains: an N-terminal Ubl domain and a C-terminal Ubp domain with an active site including cysteine 118. Both deletion of the Ubl domain and mutation of Cys 118 exhibit the same phenotype as a ubp6⌬ mutant, indicating that both domains are important for biological function of Ubp6 (Fig. 4). Overexpression of the C118S mutant allele of ubp6 acts as a dominant negative mutation, strongly inhibiting growth of wild type yeast in the presence of canavanine and anisomycin. This inhibition was completely eliminated by deletion of the Ubl domain, suggesting that Ubp6 functions in the context of a protein complex, the assembly of which requires the Ubl domain. The Ubp6 C118S mutant is able to compete with wild type Ubp6 for binding to this complex. Indeed, Ubp6 associates with the Rpn1, subunit of the base of the proteasome 19 S regulatory complex (12), suggesting that the proteasome may be the site of localization required for Ubp6 function.
Rpn1 also binds two other proteins with N-terminal Ubl domains; Rad23 and Dsk2 (57,58). In each case, the Ubl domain was required for binding (12,59). The homology between the Ubl domain and Ub raises the question of whether Ub can substitute for the Ubl. Indeed, Ub was shown to functionally substitute for the Ubl domain of Rad23 in at least some of its functions in the cell (60, 61). Lambertson et al. (61) reported that this chimeric protein can functionally replace Rad23 because of its ability to bind to the proteasome, albeit through an alternative mechanism. This binding required the polyubiquitination of the Ub moiety in Ub-Rad23 and the presence of the poly(Ub) chain-binding proteasome subunit Rpn10. However, in the case of Ubp6, the chimeric protein with the Ubl domain substituted by Ub was unable to complement ubp6⌬ phenotypes or target Ubp6 to the proteasome (Figs. 4 and 5A). These data demonstrate that Ubl domain is unique for Ubp6 function and cannot be replaced by Ub.
Cellular Localization of Ubp6 -Although the subcellular localization of the yeast proteasome is not well established, there are several reports demonstrating that components of 19 S regulatory complex are localized primarily in or around the nucleus of yeast throughout the cell cycle (62)(63)(64). By using a Ubp6-GFP fusion protein, we were able to localize Ubp6 to the nucleus in the living yeast cells for the first time (Fig. 5B). Since the Ubl domain is required for the biological function of Ubp6 and binding to the proteasome, we considered the possi-bility that this portion of the protein controls the entry of Ubp6 into the nucleus. However, our data show that the Ubp6 [ubl⌬] mutant protein is also localized in the nucleus (Fig. 5B), indicating that nuclear transport localization of Ubp6 does not depend on Ubl domain-mediated proteasome binding. Indeed, we note that the catalytic core of Ubp6 has a potential nuclear localization sequence, located at position 361 (RKVEKE-KNEK). Ubp6 behavior parallels that of Rad23, another proteasome-binding protein, that is also localized in a nucleus and does not require a Ubl domain for its localization (60).
Role of Ubp6 in Controlling Ub Levels-Most ubp6⌬ phenotypes are complemented by the addition of excess Ub, and Ubp6 hydrolase activity is dramatically increased by proteasome binding (12). Further, there is a tight coupling between the proteasomal localization of Ubp6 and its function in maintaining ubiquitin levels. We hypothesize that Ubp6 regenerates Ub by removing Ub from substrate proteins and preventing Ub degradation by proteasome. In this case, inhibition of the proteasome function should prevent the decline in Ub levels observed in ubp6⌬ cells. Indeed, we have observed that Ub is unstable in ubp6⌬ cells but stabilized by treatment with the proteasome inhibitor MG132 (Fig. 6A).
There are several possible mechanisms by which Ubp6 may function in Ub regeneration at the proteasome. Ubp6 can progressively remove the Ub group from the distal end of poly(Ub) chains attached to protein. This so-called editing function in mammals is mediated by the integral subunit of 26 S proteasome, UCH37 (10), and by its ortholog in Drosophila melanogaster (65) and S. pombe (66). No homolog of UCH37 encoded in the S. cerevisiae genome has been found. Deletion of editing enzyme should increase degradation of some ubiquitinated proteins, but this was not observed in ubp6⌬ cells. Ubp6 might remove Ub chains from the chain-substrate junction. If this is the case, the ubp6⌬ cells should accumulate polyubiquitinated proteins, but this has not been observed. At the same time, the active site mutant of Rpn11, the proteasome lid subcomplex subunit with deubiquitinating activity (15,16), does accumulate polyubiquitinated proteins (67), suggesting that Rpn11 probably is a major DUB furnishing this function. This does not exclude the possibility that Ubp6 also contributes to this function, especially since yeast bearing single mutation in active site of either Ubp6 or Rpn11 are viable, whereas the double mutation in Ubp6 and Rpn11 active sites results in lethality (68). If Ubp6 can disassemble free poly(Ub) chains, then ubp6⌬ cells should accumulate unanchored Ub chains. Although it was not observed, it is possible that another DUB is compensating in the absence of Ubp6. It is known that Ubp14 specifically disassembles unanchored Ub chains in vitro, and deletion of the UBP14 gene results in accumulation of free Ub chains in vivo (41). We have generated the ubp6⌬ ubp14⌬ mutant and observed that anti-Ub immunoblots of double mutant cell extracts revealed a profile intermediate between those of the two single mutants (Fig. 6B). The pattern of accumulated ubiquitinated species was generally typical for ubp14⌬ but with a substantial reduction in the level of both unanchored Ub chains and free Ub. These results indicate that Ubp6 probably does not contribute to the release of long unanchored chains.
The model for Ubp6 function most consistent with all of the data suggests that once substrate translocation into the lumen of the proteasome is initiated, the bulk of the polyubiquitin chain is released, probably by the action of Rpn11, and a residual one or two ubiquitins remaining attached to the target protein as it is being degraded. These are probably removed by the action of Ubp6 at the throat of the proteasome. If Ubp6 is lacking, then the last one or two ubiquitins still attached to the protein are unfolded and co-transported into the lumen of the proteasome, where they are degraded. This results in a reduced Ub level and the consequent phenotypes observed here.
Evolutionary Conservation of Ubp6 -Proteasome subunits are well conserved in all eukaryotic organisms, and the overall composition of the holocomplex is nearly the same from yeast to humans (69). However, differences between individual human and yeast proteins were observed (70). The Ubp6 homolog found in mammalian cells, USP14, interacts with the 26 S proteasome, although it is not known whether it represents an abundant and functionally significant component of the mammalian proteasome complex (11). By using complementation analysis in yeast, we have directly confirmed that Ubp6 and USP14 are not only structurally, but also functionally, homologous to each other (Fig. 7). Expression of USP14 restored levels of free Ub and Ub conjugates in the ubp6⌬ cells and complemented the ubp6⌬-associated phenotypes in yeast (Fig.  7B). In agreement with the notion that the activity of yeast Ubp6 is dramatically increased upon binding to the proteasome (12), free USP14 exhibits an unusually high K m for Ub (Ͼ200 M), indicating that human protein also needs to be catalytically activated by other components of the proteasome complex. We considered the possibility that Ubl domain acts as a pseudosubstrate, occupying the active site of the enzyme and preventing efficient binding of ubiquitin. However, we showed that the K m of USP14 for Ub does not change after deletion of the Ubl domain (Fig. 8B). Our data, together with the observation that only proteasome-bound USP14 can be labeled by UbVS, an active site-directed probe specific for DUB enzymes (11), suggest that USP14 probably also undergoes the same activation process at the proteasome as its yeast ortholog, Ubp6.
Conclusions-The yeast deubiquitinating enzyme Ubp6 is required to regulate Ub levels in vivo and shares this function with at least one other DUB, Doa4. Ub depletion in ubp6⌬ cells results from enhanced Ub degradation, which depends on functional proteasomes. There is a tight coupling between proteasomal localization of Ubp6 and its ability to maintain ubiquitin levels. In contrast, Doa4 probably requires localization to the endocytic pathway and spares Ub by retrieving it before translocation of the ubiquitinated protein into the vesicular pathways. A decreased level of Ub is responsible for impairment of a subset of Ub-dependent processes. In ubp6⌬ and doa4⌬ mutants, this is manifested as sensitivity to several toxic compounds and defects in phenotypic expression and de novo formation of the yeast prion [PSI ϩ ]. Considering that [PSI ϩ ] is a model for mammalian amyloidoses and that yeast Ubp6 and its human ortholog USP14 are functional homologs, this study has implications for the role of the Ub-proteasome system in the pathogenesis of neurodegenerative diseases. For instance, we predict that the defects exhibited by the ataxic mouse may be completely alleviated by overexpressing Ub from a transgene.