Effects of Ubiquitin System Alterations on the Formation and Loss of a Yeast Prion*♦

The yeast prion [PSI+] is a self-propagating amyloidogenic isoform of the translation termination factor Sup35. Overproduction of the chaperone protein Hsp104 results in loss of [PSI+]. Here we demonstrate that this effect is decreased by deletion of either the gene coding for one of the major yeast ubiquitin-conjugating enzymes, Ubc4, or the gene coding for the ubiquitin-recycling enzyme, Ubp6. The effect of ubc4Δ on [PSI+] loss was increased by depletion of the Hsp70 chaperone Ssb but was not influenced by depletion of Ubp6. This indicates that Ubc4 affects [PSI+] loss via a pathway that is the same as the one affected by Ubp6 but not by Ssb. In the presence of Rnq1 protein, ubc4Δ also facilitates spontaneous de novo formation of [PSI+]. This stimulation is independent of [PIN+], the prion isoform of Rnq1. Numerous attempts failed to detect ubiquitinated Sup35 in the yeast extracts. While ubc4Δ and other alterations of ubiquitin system used in this work cause slight induction of some Hsps, these changes are insufficient to explain their effect on [PSI+]. However, ubc4Δ increases the proportion of the Hsp70 chaperone Ssa bound to Sup35, suggesting that misfolded Sup35 is either more abundant or more accessible to the chaperones in the absence of Ubc4. The proportion of [PSI+] cells containing large aggregated Sup35 structures is also increased by ubc4Δ. We propose that UPS alterations induce an adaptive response, resulting in accumulation of the large “aggresome”-like aggregates that promote de novo prion generation and prion recovery from the chaperone treatment.

Prions are infectious aggregated protein isoforms that cause fatal and incurable neurodegenerative diseases such as sheep scrapie and "mad cow" diseases in mammals and Creutzfeldt-Jacob disease in humans (1). The pathology of prion diseases is reminiscent of the other noninfectious amyloidoses and protein assembly disorders associated with amyloid-like protein aggregation, such as Alzheimer, Huntington, and Parkinson diseases (2,3).
Degradation of abnormal or damaged proteins occurs, at least in part, via the ubiquitin-proteasome system (UPS) 4 (4). Targeting of a protein for degradation to the proteasome requires its conjugation to a 76-amino acid protein named ubiquitin (Ub) through the sequential action of several enzymes, including the Ub-conjugating enzymes (Ubc). This process can be reversed by the action of various deubiquitinating enzymes, including Ub-specific processing proteases (Ubp) that disassemble Ub conjugates and release free Ub (5). Mutations in different UPS components have been found to be associated with Alzheimer and Parkinson diseases (6). UPS failure can lead to aggregation of impaired proteins; in turn, aggregate formation is known to inhibit UPS function (7). Although proteasome inhibitors affect the turnover of mammalian prion proteins, and Ub is found in intracellular deposits of prion aggregates (8 -10), the role of Ub-mediated proteolysis in prion formation is not defined.
Several yeast proteins possess prion properties (11)(12)(13). All known yeast prions contain QN-rich stretches, somewhat resembling poly(Q) stretches involved in aggregation disorders such as Huntington disease. Unlike the mammalian prion protein PrP, yeast prions do not kill cells, although some of them appear to be pathogenic in a longer run (14 -16). Prion formation can antagonize the normal cellular function of a yeast protein, thus producing changes in phenotype that mimic conventional loss-of-function mutations. The yeast non-Mendelian element [PSI ϩ ] is a prion isoform of the yeast translation termination factor Sup35. [PSI ϩ ] cells are partially defective in termination of translation. [PIN ϩ ], a prion isoform of another protein of yet unknown function, Rnq1 (17), is usually needed for the de novo formation of [PSI ϩ ] but not for propagation of pre-existing [PSI ϩ ] (18 -21).
The propagation of yeast prions is modulated by chaperone proteins of the Hsp100 and Hsp70 families (11,12). The chaperone protein Hsp104, an ATPase known to promote solubilization of aggregated heat-damaged proteins in cooperation with other chaperones (22), is required for maintenance of all known yeast prions, possibly due to its ability to break prion aggregates into the oligomeric seeds, initiating new rounds of prion propagation. Transient overproduction of Hsp104 also cures yeast cells of [PSI ϩ ], although not of the other prions. Ssa proteins of the Hsp70 family also play an important role in prion propagation. Ssa overproduction antagonizes [PSI ϩ ] curing by excess Hsp104 and promotes de novo [PSI ϩ ] formation in the [PIN ϩ ] strains (23,24), while Ssa mutations destabilize [PSI ϩ ] (25). Another chaperone of the Hsp70 family, Ssb, consistently manifests itself as a [PSI ϩ ] antagonist (26). Hsp40 chaperones that act as co-chaperones of Hsp70s were shown to control propagation of [PIN ϩ ] (27) and modulate aggregation of heterologous poly(Q) proteins in yeast (28,29). Mechanisms of the chaperone effects on prions are not yet completely understood. As some of the chaperones of Hsp70 and Hsp40 families were implicated in targeting misfolded proteins for Ubdependent degradation (30,31), it is possible that some of the effects of chaperones on prions could be mediated by UPS.
We employed [PSI ϩ ] as a model to analyze the possible involvement of the UPS in modulation of prion formation, propagation, and clearance. Previously, we have shown that depletion of cellular Ub levels (by deleting the deubiquitinating enzyme Ubp6) decreases both the phenotypic manifestation of [PSI ϩ ] and de novo induction of [PSI ϩ ] by overproduced Sup35 protein in the [PIN ϩ ] background (32). Here, we demonstrate that both ubp6⌬ and deletion of a gene coding for the major yeast Ub-conjugating enzyme, Ubc4, antagonize [PSI ϩ ] curing by excess Hsp104. Moreover, ubc4⌬ also significantly increased spontaneous formation of [PSI ϩ ]. Our data implicate the UPS as one of the major modulators of prion formation and clearance in the yeast cells.

EXPERIMENTAL PROCEDURES
Yeast Strains-The Saccharomyces cerevisiae strains used in this study are listed in Table 1. GT81-1C and GT81-1D, isogenic haploid strains that derived from self-homozygous diploid GT81 (14), contain the [PSI ϩ ]-suppressible marker ade1-14 (UGA) (33). The [psi Ϫ PIN ϩ ] and [psi Ϫ pin Ϫ ] derivatives resulted from curing the isogenic [PSI ϩ ] strains by overproduction of Hsp104 or guanidine hydrochloride (GuHCl) treatment, respectively (33). All of the single deletion strains used were constructed for this study by one step PCR-mediated direct gene replacement with the pFA6a-His3MX6 module (34). Primers sequences are available by request. Double and triple knock-out strains were obtained by mating the appropriate deletion strains and then sporulating and dissecting the resulting diploid strain. Presence of double and multiple disruptions in the spore clones was confirmed by PCR.
Media and Growth Conditions-Standard yeast media, cultivation conditions, procedures for yeast growth, transformation, sporulation, and tetrad analysis were used (43). Yeast cultures were grown at 30°C unless otherwise specified. Gal medium contained 2% galactose instead of glucose. GalϩRaf  (33). Briefly, to measure [PSI ϩ ] curing by overproduced Hsp104, yeast cultures were transformed with the Hsp104 expression vectors or empty control vectors and analyzed as outlined below. For plate assays with the plasmids expressing Hsp104 from the endogenous (P HSP104 ) or constitutively active (P GPD ) promoters, plasmidcontaining cultures were grown on the medium selective for the plasmid and then velveteen replica plated onto the medium not selective for the plasmid and lacking adenine or onto YPD medium. For plate assays with the galactose-inducible (P GAL ) promoter, cultures were grown on selective medium containing Gal and then velveteen replica-plated onto selective media containing glucose but lacking adenine or onto YPD medium. For all plasmids, the efficiency of [PSI ϩ ] curing was assessed by decreased growth on ϪAde medium and increased intensity of red color on YPD medium compared with the control cultures. For quantitative assays, yeast cultures containing the plasmids bearing HSP104 under P GAL promoter were pregrown in selective medium containing glucose and then inoculated into the GalϩRaf medium of the same composition and grown with shaking (200 -250 rpm) at 30°C. Aliquots were taken after specified periods of time and plated onto selective medium containing glucose. The colonies grown after 4 -5 days were velveteen replica-plated onto YPD and ϪAde medium. [PSI ϩ ] retention or loss was scored by color on YPD and growth on ϪAde.
Assays for [PSI ϩ ] Formation-Plate assays and quantitative assays for spontaneous [PSI ϩ ] formation were performed as described previously (26). Rates of spontaneous [PSI ϩ ] formation and confidence limits were calculated by using the formulas used previously for measuring the rates of spontaneous mutations (44).
Protein Analysis and Antibodies-For measuring the levels of Sup35, chaperone proteins, and Ub, yeast cells were lysed by vortexing with glass beads (Sup35 and chaperones) or by boiling in SDS loading buffer (Ub) according to protocols described previously (24,32,41). Co-imunoprecipitation assays for detection of Sup35 and Ssa interaction were performed using immobilized protein A (Invitrogen) according to the published protocol (23). A ubiquitinated protein enrichment kit (Calbiochem EMD Biosciences, Inc., San Diego, CA) was used to search for ubiquitinated Sup35 according to the manufacturer protocol. Total protein extracts and immunoprecipitates were examined by Western analysis using specific antibodies. Antibodies to Sup35NM were obtained as described previously (41), and antibodies to Sup35C were the generous gift of D. Bedwell. Antibodies to Hsp104 were kindly provided by S. Lindquist. Antibodies to total Ssa, Ssa3/4, and Ssb were generously provided by E. A. Craig. Antibodies to Ydj1 and Sis1 were kindly provided by D. Cyr. We used anti-Ub antibody P4D1 (Santa Cruz Biotechnology, Inc., Santa Cruz, CA), HA-specific antibody 12CA5 from Maine Biotechnology Services, anti-Pgk antibody (Molecular Probes, Inc., Eugene, OR), and anti-c-Myc antibody (Ab-1) from Calbiochem EMD Biosciences, Inc. In all experiments appropriate secondary antibodies from GE Healthcare Ltd. (Buckinhamshire, UK) were used. Western blots were developed according to the ECL detection system protocol from Pierce. Densitometry was performed on HyBlotCL autoradiography films (Denville Scientific, Inc., Metuchen, NJ) using the LabWorks 4.0 program on a UVP BioImaging system (UVP Inc., Upland, CA).
Detection of Sup35NM-GFP Aggregation in the Yeast Cells-GFP detection in live yeast cells was performed as described previously (33,45). Yeast cultures containing the plasmid pmCUPNMsGFP were pregrown in selective medium lacking uracil for a period of 2 days and then inoculated into the same medium with 50 or 100 M CuSO 4 , to induce the P CUP1 promoter (the starting concentration of cells was 2-3 ϫ 10 6 cells/ ml). After the specified periods of time, aliquots were taken, and proportions of transformed cells containing detectable aggregated structures were determined by using the microscope BX41 (Olympus) with the Endow GFP Bandpass Emission filter. Cultures containing the pmCUPsGFP construct that was used as a control never produced cytologically detectable aggregates in these conditions (data not shown), confirming that aggregation is specific to the Sup35-GFP fused protein. Deletion strains and wild-type control strain were transformed with the plasmids producing Hsp104 from either the galactose-inducible (P GAL ) or a strong constitutive (P GPD ) promoter, as well as by empty control vectors. In the case of ubc4⌬, we also moderately overexpressed HSP104 by providing a single plasmid-borne copy under the control of its endogenous promoter. Levels of overexpressed Hsp104 were similar in wildtype and UPS mutants as confirmed by Western blot analysis ( Fig. 1A and data not shown). The efficiency of [PSI ϩ ] curing in the presence of either P GAL -HSP104 (Fig. 1A) or P GPD -HSP104 (data not shown) constructs was detected by color on YPD medium and growth on ϪAde medium. Efficiency of curing was decreased in ubc4⌬ and ubp6⌬ single deletion strains, compared with the wild-type strain. [PSI ϩ ] loss in the presence of an extra copy of P HSP104 -HSP104 was also decreased in ubc4⌬ (data not shown). Quantitative assays employing the P GAL -HSP104 construct demonstrated that the effect of double ubc4⌬ ubp6⌬ deletion on [PSI ϩ ] curing by excess Hsp104 is the same as the effect of single ubc4⌬ deletion (Fig. 1B). This result suggests that both proteins influence [PSI ϩ ] curing via one and the same pathway and that Ubc4 acts before Ubp6. Such a conclusion is in agreement with the known roles of both enzymes in the Ub-proteasome degradation pathway. Interestingly, the efficiency of [PSI ϩ ] curing by excess Hsp104 was not altered in the ubc5⌬ strain (data not shown), despite the fact that Ubc5 is known to have a role that is very similar to and partly redundant with that of Ubc4. A double ubc4⌬ ubc5⌬ deletion strain was not viable in the genetic background of our strains.

Alterations of the UPS Decrease [PSI
The only other known mutation decreasing [PSI ϩ ] curing by excess Hsp104 is the double deletion of the genes SSB1 and SSB2, coding for the Ssb protein of the Hsp70 family (26 Fig. 2B. One should note that the negative result for ubp6⌬ is somewhat ambiguous, as ubp6⌬ is known to weaken growth of the [PSI ϩ ] strains on ϪAde medium (32), so that some newly arisen Ade ϩ derivatives might remain unnoticed in this strain.
Increased spontaneous papillation of the ubc4⌬ strain on ϪAde media was eliminated by reintroduction of the wild-type ] and isogenic ubc4⌬ (GT349) and ubp6⌬ (WTY105) strains were transformed with the URA3 plasmid pGAL104, expressing Hsp104 from the galactose-inducible (P GAL ) promoter (1Hsp104), and with the control plasmid pRS316GAL (Control). Transformants were selected on ϪUra medium, replica-plated onto ϪUra/Gal medium to induce the P GAL promoter, incubated for 4 -5 days, and then velveteen replica-plated onto either complete YPD medium or ϪUraϪAde medium. [PSI ϩ ] loss was scored by more reddish (dark on figure) color on YPD or decreased growth on ϪUraϪAde. Levels of overexpressed Hsp104 were compared by Western blot using anti-Hsp104 antibody. B, epistatic interaction between ubc4⌬ and ubp6⌬ in [PSI ϩ ] curing. Wild-type [PSI ϩ ] strain GT81-1C and isogenic [PSI ϩ ] strains with ubc4⌬ (GT349), ubp6⌬ (WTY105), and ubc4⌬ ubp6⌬ (GT684-8B), bearing the URA3 plasmid pGAL104 with the P GAL -HSP104 cassette, were incubated in the liquid Ϫ-Ura/GalϩRaf medium to induce the P GAL promoter. After specified periods of time, aliquots were plated onto ϪUra medium. The grown colonies were replica plated onto ϪAde and YPD media to test for [PSI ϩ ] loss. C, additive effects of ubc4⌬ and ssb1/2⌬ on [PSI ϩ ] curing. Wild-type [PSI ϩ ] strain GT81-1C and isogenic [PSI ϩ ] strains ubc4⌬ (GT349), ssb1/2⌬ (GT147), and ubc4⌬ ssb1/2⌬ (GT389-14A), bearing the TRP1 plasmid pFL39-GALHSP104 with the P GAL -HSP104 cassette, were studied for [PSI ϩ ] loss in the same way as in part B, with the only difference that ϪTrp media were used for the incubation and plating. Despite some variations in the efficiencies of [PSI ϩ ] curing between B and C, which are apparently due to different P GAL -HSP104 plasmids used, differences between the wild-type and ubc4⌬ strains are maintained in these two variants of experiment.

Ubiquitin System Alterations Influence Yeast Prions
UBC4 ϩ gene on a plasmid (Fig. 2C). When the ubc4⌬ [psi Ϫ PIN ϩ ] strain was mated to the isogenic [psi Ϫ pin Ϫ ] UBC4 ϩ strain of the opposite mating type, and the resulting diploid was sporulated and dissected, the increased papillation on ϪAde co-segregated with the ubc4⌬ marker (HIS3) in tetrad analysis ( Fig. 4D and Table 2). These data confirm that increased papillation is due to lack of UBC4 gene rather than any independent alteration in the yeast genome.
A number of Ade ϩ colonies that had independently arisen from UBC4 ϩ and ubc4⌬ strains were tested for the curability of Ade ϩ phenotype by growth in the presence of GuHCl, an agent known to eliminate yeast prions (33,47). While the majority of colonies obtained in the [PIN ϩ ] strains were curable by GuHCl, about 77% Ade ϩ colonies obtained in the UBC4 ϩ [psi Ϫ pin Ϫ ] strain were incurable, indicating that they originated from mechanisms other than prion formation, such as reversion of the ade1-14 mutation or suppressor mutation. In contrast, 94% of Ade ϩ colonies obtained in the ubc4⌬ [psi Ϫ pin Ϫ ] strain were curable by GuHCl. All GuHCl-curable Ade ϩ colonies isolated from this strain were also curable by the P GPD -HSP104 plasmid, known to eliminate [PSI ϩ ] as described above. This confirms that majority of spontaneous Ade ϩ colonies detected in the ubc4⌬ [psi Ϫ pin Ϫ ] strain result from [PSI ϩ ] formation rather than from suppressor or reverse mutation. Quantitative assay confirmed that the differences in rates of spontaneous [PSI ϩ ] formation between the ubc4⌬ and wild-type strains in both [psi Ϫ PIN ϩ ] (data not shown) and [psi Ϫ pin Ϫ ] ( To check whether Rnq1 protein in its non-prion form influences formation of the [PSI ϩ ] prion in the ubc4⌬ background, we deleted the RNQ1 gene in both wild-type and ubc4⌬ [psi Ϫ ] strains and found that rnq1⌬ dramatically reduced frequency of the spontaneous [PSI ϩ ] papillae in the ubc4⌬ strain, making it indistinguishable from the wild-type strain (Fig. 2B). These data show that the presence of the Rnq1 protein is required for the increased de novo prion formation observed in the ubc4⌬ cells. To our knowledge, this is the first evidence demonstrating a pin Ϫ ] strains GT409 (UBC4 ϩ ), GT387 (ubc4⌬), and GT 385-14A (ubc5⌬), and the [psi Ϫ ] rnq1⌬ strains GT564 (UBC4 ϩ rnq1⌬) and GT820 (ubc4⌬ rnq1⌬). All strains were isogenic except for the markers indicated. Independent colonies of each strain were patched on YPD medium and replica plated onto ϪAde medium. The ϪAde plates were photographed after 14 days of incubation. A sample of independent Ade ϩ derivatives was confirmed to contain [PSI ϩ ] by using an assay based on the loss of Ade ϩ phenotype following GuHCl treatment (see "Results"). C, increased spontaneous formation of Ade ϩ papillae in the ubc4⌬ strain is compensated by the plasmid pUBC4 containing the wildtype UBC4 ϩ gene. The [psi Ϫ pin Ϫ ] ubc4⌬ strain GT385-14C was transformed with either control vector pFL44 (ubc4⌬) or pUBC4 (ubc4⌬ ϩ pUBC4). Transformants were patched on ϪUra medium and replica plated onto ϪUra-Ade medium. Papillation was scored after 14 days. One typical transformant is shown in each case. D, increased formation of Ade ϩ papillae co-segregates with ubc4⌬ in tetrad analysis. The [psi Ϫ PIN ϩ ] ubc4⌬ strain GT386 was crossed to the isogenic wild-type strain GT234. The resulting diploid strain was sporulated and dissected. One typical tetrad is shown. His ϩ corresponds to ubc4⌬ (see Table 2 for numbers).

TABLE 2
Meiotic co-segregation of the Pap ؉ phenotype with ubc4⌬::HIS3 The ubc4⌬::HIS3 ͓psi Ϫ PIN ϩ ͔ strain GT386 was crossed to the isogenic UBC4 ͓psi Ϫ pin Ϫ ͔ strain GT234. Resulting diploids were sporulated and dissected. Each clone was assayed for the frequency of spontaneous ͓PSI ϩ ͔ formation. High frequency is designated as high papillation (Pap ϩ ) phenotype, while low frequency is designated as low papillation (Pap Ϫ ) phenotype. In majority of the spores, ubc4⌬ (His ϩ ) and UBC4 ϩ (His Ϫ ) co-segregate with Pap ϩ and Pap Ϫ , respectively. Rare exceptional spores (His Ϫ Pap ϩ and His ϩ Pap Ϫ ) occur at the frequency similar to the expected frequency of meiotic conversion of the HIS3 marker.

Phenotype
No. of spores  (19,20). (32,48) have previously observed that the ubp6⌬ deletion results in depletion of free Ub and increased sensitivity to the translation inhibitor anisomycin and that most of the detectable phenotypes of ubp6⌬ are due to Ub depletion. The proposed (30) involvement of Ssb in UPS-dependent protein degradation makes it tempting to speculate that some effects of ssb1/2⌬ on [PSI ϩ ] could be through alteration of UPS. To examine potential mechanisms by which defects of UPS influence prion formation, we first characterized the levels of Ub and sensitivity to anisomycin in the strains with ubc4⌬, ubp6⌬, and ssb1/2⌬ deletions. As expected, ubp6⌬ reduced levels of free Ub in comparison with the isogenic wildtype strain (Fig. 3A). Interestingly, levels of free Ub were also reduced in the double ssb1/2⌬ mutant (Fig. 3A). The ubc4⌬ strain exhibited a slight increase in the levels of free Ub (Fig. 3A) apparently due to impaired conjugation. Deleting UBC4 in the ubp6⌬ or ssb1/2⌬ deletion strains (Fig. 3A) partially restored free Ub levels suggesting that the effects of UPS alterations on [PSI ϩ ] curing are not simply due to decreased Ub levels.

Effects of UPS Alterations and Ssb Deficiency on Ub Levels and Antibiotic Sensitivity-We and others
As shown previously for ubp6⌬ (32), decreased Ub levels correlate with increased sensitivity to anisomycin. We found that like ubp6⌬, both ubc4⌬ and ssb1/2⌬ mutants were more sensitive to anisomycin than the wild type strain. The ubp6⌬ strain exhibited the highest level of sensitivity, the ubc4⌬ strain exhibited an intermediate sensitivity, and ssb1/2⌬ was the least sensitive. The double ubc4⌬ ubp6⌬ deletion strain was more sensitive to anisomycin then either single mutant, while the sensitivity of the triple ubc4⌬ ssb1/2⌬ deletion strain was closer to that of the ubc4⌬ than to the ssb1/2⌬ strain (Fig. 3B). These data indicate that decreased Ub levels and decreased Ub conjugation affect resistance to anisomycin in a partially independent fashion. Interactions between Ubc4, Ubp6, and Ssb affect both Ub levels and anisomycin sensitivity differently than interactions between the same players affect the [PSI ϩ ] curing assays. This suggests that the effects of UPS alterations on [PSI ϩ ] cannot be explained as a simple consequence of the alterations in Ub levels.
Ubiquitinated Sup35 Is Not Detectable in Vivo-The simplest explanation for the effects of UPS alterations on [PSI ϩ ] is that misfolded Sup35, generated either spontaneously or as a result of the Hsp104-mediated prion disaggregation, is normally eliminated via the Ubc4-mediated ubiquitination and subsequent proteasome-mediated degradation. If this process is defective in the UPS mutants, accumulation of misfolded Sup35 would lead to increased spontaneous prion formation and decreased prion curing. To address the possibility that misfolded Sup35 is targeted for UPS-mediated degradation via ubiquitination, we have searched for evidence of ubiquitinated Sup35 using two different approaches. First, we employed immunoprecipitation of Sup35 from cell lysates followed by immunoblotting with a Ub-specific antibody. These experiments were performed with both epitope-tagged Sup35-HA (Fig. 3C) and endogenous Sup35 (data not shown). Some experiments also employed the strains containing tagged derivatives of Ub (Ub-Myc in case of Sup35-HA (Fig. 3C) or Ub-HA in case of endogenous Sup35). Second (Fig. 3D), we purified Ub-protein conjugates from cell extracts using either a Ub conjugates enrichment kit, based on the Rad23 UBA domain fused to agarose beads, or protein A-and Myc-agarose for the HA-tagged or Myc-tagged Ub. The eluted Ub conjugates were then detected by Sup35-specific antibody. In both assays, wild-type strains were compared with the isogenic mutant strains in which Ub levels are depleted (that is, strains carrying doa4⌬ or ubp6⌬ mutations) or to the ubc4⌬ strains where Sup35 ubiquitination was not expected to occur. As it is likely that ubiquitinated Sup35 is short-lived, some experiments were performed in the presence of an inhibitor of proteasomal proteolysis (MG132), using the specially constructed mutant strains (erg6⌬ or pdr5⌬) with increased permeability of the yeast membrane to MG132. In the other experiments, we employed the rpn4⌬ mutant that exhibits decreased proteasomal function due to a defect in transcription of proteasome genes (49). All major experiments were repeated with a series of the isogenic strains that differed by prion composition (that is, presence or absence of [PSI ϩ ] and [PIN ϩ ]), as well as by levels of the Sup35 production (normal or high). Immunoprecipitation experiments were also performed with the [PSI ϩ ] strains overproducing Hsp104.
Despite numerous tries, none of the experiments produced any evidence for the existence of ubiquitinated Sup35. Neither were steady state levels of Sup35 influenced by the proteasome inhibitors (Fig. 3E). Likewise, a large scale analysis of the ubiquitinated proteins in yeast (50) failed to detect Sup35, although ubiquitinated forms of another prion protein (Rnq1) and some members of the Hsp70 family were detected. These negative results do not completely exclude the possibility that there is a small and short-lived ubiquitinated fraction of Sup35 that cannot be identified by the approaches used. However, at least in the case of the excess Hsp104-mediated [PSI ϩ ] curing, the ubiquitinated fraction of Sup35 should be sizeable for ubc4⌬ to influence the outcome, and our results make such an explanation unlikely. Therefore, we propose that UPS alterations influence [PSI ϩ ] via a more complex mechanism than simply by altering Sup35 ubiquitination. a Proportion of ͓PSI ϩ ͔ colonies was determined in the random sample of the independently arisen Ade ϩ colonies on the basis of GuHCl curability assay (see "Results" for details). FEBRUARY 2, 2007 • VOLUME 282 • NUMBER 5

Levels of Sup35 and Hsps in the Strains with UPS Alterations-
In the strains where it is viable, the ubc4⌬ ubc5⌬ double deletion increases background levels of some stress-inducible chaperones, specifically Hsp70s (51). As both Sup35 levels and Hsp levels control [PSI ϩ ] propagation (12), one possibility could be that UPS alterations influence [PSI ϩ ] via alterations of either Sup35 or Hsp levels. To address this possibility, we checked the effects of UPS alterations on the levels of Sup35 and levels of Hsps known to influence [PSI ϩ ].
Our experiments detected no differences in Sup35 levels between the wild-type strain and isogenic strains bearing the single ubc4⌬ and ubp6⌬ or double ubc4⌬ ubp6⌬ deletions. This observation was confirmed for both [PSI ϩ ] (Fig. 4A) and [psi Ϫ ] (data not shown) backgrounds. There was also no detectable change in the levels of Ssb, Hsp40-Ydj, or Hsp40-Sis1 chaperones in any of these strains (Fig. 4A). Level of Hsp104 and total Ssa protein were slightly but reproducibly increased by ubc4⌬ in both [PSI ϩ ] (Fig. 4, A and B) and [psi Ϫ ] (data not shown) backgrounds. This increase in Ssa levels was primarily due to induction of the normally silent SSA3 and SSA4 genes, as confirmed by using an Ssa3/ 4-specific antibody (data not shown). A slight but not statistically significant increase in Hsp104 levels was also detected in the ubp6⌬ strain. However, neither Hsp104 nor Ssa levels were increased in the double ubc4⌬ ubp6⌬ mutant. The defect of [PSI ϩ ] curing by excess Hsp104 detected in this strain (see above), cannot be explained by alterations of the Hsp104 or Ssa levels.
Previously, we have shown that Sup35 and Ssa proteins physically interact in vivo (23). Interestingly, co-immunoprecipitation experiments demonstrated that the fraction of Ssa bound to Sup35 is increased about 3-fold in the ubc4⌬ strain compared with the isogenic wild-type strain (Fig. 4, C and D), while total amount of Ssa in the cell is increased at only about 1.5-fold (Fig. 4B). These data indicate that lack of Ubc4 increases either ability of Ssa to recognize Sup35 or proportion of the Sup35 protein that can be recognized by Ssa or both. The possible relevance of this observation to the ubc4⌬ effect on [PSI ϩ ] is discussed below (see "Discussion").
Aggregation Status of the Sup35 Protein in Strains with UPS Mutations-Next, we checked whether the aggregation status of the Sup35 protein is altered in strains with UPS defects. For this purpose, we have employed the Sup35NM-GFP construct that bears the N-proximal portion of Sup35 (encompassing the prion-forming and middle domains) fused to the green fluorescent protein. Expression of this construct from the copper-inducible (P CUP1 ) promoter results in accumulation of "clumped" or "dot"-like fluorescent aggregates in a fraction of the [PSI ϩ ] cells, while diffused fluorescence is detected in the [psi Ϫ ] strains (52). We have observed that the [PSI ϩ ] ubc4⌬ culture expressing Sup35NM-GFP contained larger proportion of cells with cytologically detectable aggregates, compared with the isogenic wild-type strain (Fig. 4E and data not  shown). These results indicate that lack of Ubc4 facilitates formation of the large cytologically detectable Sup35 aggregates.

DISCUSSION
Our data show that mutational alterations of the UPS, specifically deletions of the genes UBC4, coding for one of the . Ub levels and lack of detectable ubiquitinated Sup35 in yeast cells. A, defects of the Ub system alter the levels of free Ub and Ub-conjugates. The following isogenic [PSI ϩ PIN ϩ ] strains were used: GT81-1C (wild-type, or WT), GT349 (ubc4⌬), WTY105 (ubp6⌬), GT832-7B (ubc4⌬ ubp6⌬), GT147 (ssb1/2⌬), and GT389-14A (ubc4⌬ ssb1/2⌬). Aliquots were normalized by optical density and lysed. Proteins were separated on a Tricine/polyacrylamide gel followed by Western blotting using anti-Ub antibody. B, sensitivity to anisomycin reflects changes in Ub levels in UPS mutants. 10-Fold serial dilutions of the same strains as in A were spotted on the minimal medium with or without drug and grown for 3-5 days at 30°C. C, no ubiquitinated Sup35 was detected by immunoprecipitation. The [PSI ϩ PIN ϩ ] strains GT81-1D (wild-type (WT)) and GT349 (ubc4⌬ (or ⌬)), the [psi Ϫ PIN ϩ ] strains GT159(WT) and GT386(ubc4⌬), and the [psi Ϫ pin Ϫ ] strains GT174(WT) and GT387(ubc4⌬), carrying a plasmid that expresses Myc-tagged Ub in combination with either HA-tagged Sup35 (ϩ) or empty vector (Ϫ), were grown on the ϪUraϪTrp medium containing 100 M CuSO 4 to induce P CUP1 , which controls expression of both Sup35-HA and Ub-Myc. Proteins were immunoprecipitated by using anti-HA antibody and protein A-agarose. The immunoprecipitates (IP) and total lysates (TL) were analyzed by Western blotting with anti-Myc and anti-Sup35 antibodies. A typical Western blot is shown. D, Sup35 was not found among the ubiquitinated proteins. Protein extracts from the isogenic wild-type (WT) and ubc4⌬ (⌬) strains listed in A were applied to the poly-Ub affinity beads and control beads as shown. Polyubiquitinated proteins bound to the beads were collected, fractionated by SDS-PAGE, and probed sequentially with anti-Sup35 and anti-Ub antibody. E, steady state levels of Sup35 are not affected by proteasome inhibitors. Strains GT156(erg6⌬) and WTY271(pdr5⌬), bearing mutations that increase permeability of yeast cells to the proteasome inhibitor MG132, were grown in the liquid synthetic medium to the mid-exponential phase at 30°C. Then, half of the culture (ϩ) was incubated with 100 M MG132, dissolved in 0.1% Me 2 SO, while another half (Ϫ) was incubated with only 0.1% Me 2 SO for 2 h. Aliquots, normalized by OD, were taken to prepare cell extracts. Equal amounts of protein were analyzed by Western blotting, using anti-Sup35 antibody. Reaction to anti-Pgk antibody was used as a loading control.
major Ub-conjugating enzymes, and UBP6, coding for the deubiquitinating enzyme responsible for the normal recycling of Ub, increase resistance of the yeast prion [PSI ϩ ] to curing by excess Hsp104 chaperone. The effect of the ubc4⌬ ubp6⌬ double mutant resembled that of the ubc4⌬ single mutant. This indicates that both deletions likely influence prion curing via one and the same pathway. Most of the known phenotypic effects of ubp6⌬ are explained by depletion of the free Ub pool (32,48). This results in a decrease of Ub available for con-jugation and consequently in a decreased amount of Ub-protein conjugates. Ubc4 is responsible for conjugating free Ub to the protein substrates, therefore ubc4⌬ causes a more severe decrease in the amount of Ub-protein conjugates than ubp6⌬ does, and the decrease of the Ub pool in a ubp6⌬ strain cannot significantly add to this effect. Therefore, observation that ubp6⌬ does not further exacerbate the effect of ubc4⌬ on prion curing suggests that UPS effects on [PSI ϩ ] are mediated by the Ub-protein conjugation step. As Ubc4 is playing a major role in this step, we primarily focused on studying the effects of ubc4⌬.
In addition to increasing [PSI ϩ ] resistance to curing by excess Hsp104, ubc4⌬ also increased spontaneous [PSI ϩ ] formation. This is the first mutation shown to have such a "protein mutator" effect in the absence of the other pre-existing prion, [PIN ϩ ]. In principle, both effects of ubc4⌬ on [PSI ϩ ] could be explained by a model, suggesting that misfolded Sup35 protein, occasionally generated spontaneously in the cells lacking prion or produced in result of disaggregating activity of Hsp104 on prion aggregates, is normally targeted by Ubc4-dependent ubiquitination for degradation via the proteasome. In the absence of Ubc4, this misfolded protein is not eliminated, increasing the probability of turning it into a prion. This model predicts that a fraction of Sup35 is ubiquitinated. Moreover, in order to explain the anti-curing effects of ubc4⌬, the ubiquitinated fraction of Sup35 should be sizeable, at least in the prion-containing cells that overproduce Hsp104. However, neither direct approaches employed in our work (see above, Fig. 3, C and D) nor large scale analysis performed by another group (50) detected Sup35 among ubiquitinated proteins. Moreover, the prion domain of Sup35 contains a QN-rich stretch resembling poly(Q) proteins, and it is known that proteasomal degradation of proteins with poly(Q) stretches is inefficient (53). Although we cannot completely exclude a possibility that the small short-lived fraction of misfolded ubiquitinated Sup35 may exist in the yeast cells, it seems unlikely that elimination of such a fraction could explain all the observed effects of ubc4⌬ on [PSI ϩ ].  Fig. 3A, grown at 25°C on minimal medium, separated by SDS-PAGE, and Western blotted with the antibodies specific to indicated proteins. Pgk1 protein was used as a loading control. A photograph of a typical Western blot (A) is shown. Densitometry was used to determine the relative level of Hsp protein (B). For each immunoblot, the integrated area of densitometry peak corresponding to the protein band was calculated and adjusted according to the measurement of Pgk1 protein band. Average wild-type measurements for each blot were set to a value of 1.0 and compared with the measurement of the same protein in other strains. Deviations from the mean value are shown. C, ubc4⌬ increases the proportion of Ssa bound to Sup35. The [psi Ϫ PIN ϩ ] strains GT159(WT) and GT386(ubc4⌬) were transformed with pmCUP1-SUP35-HA, encoding HA-tagged Sup35. Proteins were immunoprecipitated from lysates with anti-HA antibody, transferred to nitrocellulose, and reacted with antibody to Sup35 (upper panel) or Ssa (lower panel). D, densitometry was used to determine the levels of Sup35 and Ssa proteins in the experiment shown in C. The ratio of Ssa to Sup35 in the wild-type strain was set to the value of 1.0. The error bar represents standard deviation of the mean for three independent experiments. E, accumulation of cells containing cytologically detectable aggregates is increased in the ubc4⌬ [PSI ϩ PIN ϩ ] culture (ubc4⌬), in comparison with the isogenic wild-type [PSI ϩ PIN ϩ ] culture (WT). An example of the cell containing aggregates is shown; cell boundaries are indicated. Chimeric protein bearing the N-terminal (prion-forming) and middle domains of Sup35 fused in frame to GFP was produced from the copper-inducible (P CUP1 ) promoter, induced by addition of 50 M CuSO 4 to the growth medium. Cells were monitored at 30 h after addition of CuSO 4 . Standard deviations are shown. Differences between the wild-type and ubc4⌬ cultures are statistically significant. Similar results were obtained at the other periods of incubation, as well as when 100 M of CuSO 4 was added (data not shown).
Some UPS alterations are known to have broad pleiotropic effects, for example by inducing stress response. As some stress-induced proteins are shown to influence prions, we have checked whether levels of these proteins are altered in the cells with UPS alterations. Indeed, we have detected slightly increased levels of Hsp104 and Ssa in ubc4⌬ and slightly increased level of Hsp104 in ubp6⌬. However, neither of these chaperones was induced in the double ubc4⌬ ubp6⌬ mutant, where [PSI ϩ ] resistance to excess Hsp104 remained high. Moreover, neither overproduction of Hsp104 nor overproduction of Ssa leads to a detectable increase in spontaneous [PSI ϩ ] formation in the [psi Ϫ pin Ϫ ] strain (data not shown). Therefore, it is highly unlikely that effects of UPS alterations on [PSI ϩ ] are explained by slight alterations in the levels of these chaperones. Remarkably, additive effects of ubc4⌬ and ssb1/2⌬ on [PSI ϩ ] curing by excess Hsp104 indicate that Hsp70-Ssb also influences prion resistance to excess Hsp104 via a pathway that does not involve Ubc4 and vice versa.
Interestingly, we have observed that while total levels of Ssa protein are only slightly increased in the ubc4⌬ strain, the amount of Ssa bound to Sup35 is increased significantly. This could occur by one of the following two mechanisms: either 1) an increase in the affinity of Ssa to Sup35 or 2) an increase in the fraction of Sup35 that can be bound by Ssa. It is worth noting that these mechanisms are not mutually exclusive. The first mechanism is supported by the observation that Ssa was detected among ubiquitinated proteins (50). Thus, it cannot be excluded that its activity is somehow influenced by ubiquitination, which is defective in ubc4⌬ strains. On the other hand, known patterns of Ssa interactions with other proteins speak in favor of the second mechanism. Indeed, Ssa with the help of the Hsp40 co-chaperones is known to bind partially unfolded proteins; therefore it is quite likely that it binds a partially misfolded fraction of Sup35. Increased Ssa binding to Sup35 in the absence of Ubc4 may therefore reflect either increased proportion of misfolded Sup35 in these conditions or its increased accessibility to Ssa.
Why, then, is misfolded Sup35 increased in abundance in the ubc4⌬ cells if not due to lack of its Ub-promoted degradation? One possible explanation could be drawn from our observation that the fraction of cells containing large Sup35 aggregates is increased in the [PSI ϩ ] ubc4⌬ culture (Fig. 4E). We have previously indicated (45) that large cytologically detectable aggregates of Sup35 share some (but not all) features of mammalian aggresomes, large cytoskeleton-associated structures generated by misfolded aggregation-prone proteins in the mammalian cells (54). It has previously been shown that aggresome formation is induced by UPS inhibition (55). It is therefore possible that defects of UPS, such as ubc4⌬, induce an adaptive response in the yeast cells, which results in increased formation of the large aggresome-like structures. This response could be triggered by accumulation of the misfolded proteins due to defective Ub-conjugation but not necessarily by accumulation of misfolded Sup35 per se. In case of prion proteins, such aggresome-like structures promote segregation of the prion state from the action of chaperones and therefore contribute to increased resistance of prion state to the prion curing agents. In the non-prion cells, occasional "aggresome" formation by Sup35 initiates de novo prion generation. Ssa protein has been implicated in disassembly of the large Sup35 aggregates into the smaller prion-propagating units (56), and our preliminary data suggest that Ssa can bind at least some cytologically detectable Sup35-GFP aggregates in the yeast cells. 5 Therefore, increased formation of large aggregates may be responsible for increased Ssa binding to Sup35, for example by making Sup35 more accessible to Ssa.
Certainly this model does not explain all the complexity of UPS interactions with prions. Indeed, while favoring the [PSI ϩ ] state in curing assays, ubp6⌬ decreased the phenotypic manifestations of [PSI ϩ ] and the de novo [PSI ϩ ] induction in the presence of excess Sup35 (32). These effects were even more pronounced in case of the deletion of DOA4, another gene whose product is involved in deubiquitination and maintenance of Ub pool. Interestingly, doa4⌬ appeared to inhibit [PSI ϩ ] curing by excess Hsp104 as well, although this effect was difficult to characterize quantitatively due to severe growth defects caused by doa4⌬ (data not shown). One possible explanation for the conflicting effects of ubp6⌬ and doa4⌬ on [PSI ϩ ] in different assays could be drawn from the necessity of ubiquitination for initiation of endocytosis (57), a process influencing formation of the yeast aggresome-like structures (45). Indeed, Doa4 is functionally linked to the endocytic pathway (58). Thus, ubp6⌬, and especially doa4⌬, may both induce Sup35 aggregation due to UPS defects as described above and inhibit it due to endocytosis defects caused by the depletion of free Ub pool. On the other hand, ubc4⌬ is not predicted to affect endocytosis, as it does not deplete the free Ub pool and is not known to be involved in endocytosis-associated Ub conjugation.
Unexpectedly, we have found that ubc5⌬ does not influence [PSI ϩ ] resistance to excess Hsp104 and de novo [PSI ϩ ] formation, despite the well known functional redundancy of Ubc4 and Ubc5 (51). One possible explanation is that modes of regulation of UBC4 and UBC5 are different from each other, so that Ubc4 is primarily expressed in the conditions that are crucially important for prion formation and curing. Indeed, some evidence suggests that Ubc4 and Ubc5 are preferentially produced at the different growth phases of the yeast culture (51). Such a hypothesis would indicate that processes of prion formation and elimination preferentially occur in certain physiological state(s), a notion that is worth investigating further.
At this moment, it is not known whether the Ub system specifically influences only [PSI ϩ ] or exhibits similar effects on the other prions. Although we have not detected any increase in de novo [PIN ϩ ] formation in the ubc4⌬ background, one should note that we lacked an unbiased genetic system for [PIN ϩ ] detection, and therefore all we can say is that [PIN ϩ ] is not being generated simultaneously with [PSI ϩ ]. While further experiments are needed to accurately address this question, our data have already provided one important insight into a role of the [PIN ϩ ] maintenance protein, Rnq1, in [PSI ϩ ] formation. Previously, it was shown that prion form of Rnq1 promotes de novo [PSI ϩ ] formation (18,20). However, we have found that the presence of Rnq1 protein per se, even in its non-prion form, facilitates spontaneous [PSI ϩ ] formation at least in the ubc4⌬ cells (Fig. 2B). This suggests that Rnq1 protein, independently of its aggregation, is playing an important role in the process of generation of other prions. Rnq1 was detected among ubiquitinated proteins in the large scale assays (50). It is therefore possible that some UPS effects on prions, e.g. increased spontaneous [PSI ϩ ] formation in ubc4⌬, could be at least in part due to variations in Rnq1 ubiquitination and, consequently, its proteolyitic stability. This possibility is currently under investigation.
Taken together, our results present strong evidence for the ability of the UPS to regulate de novo prion formation and prion resistance to curing treatments in the yeast model. As the efficiency of ubiquitination and of proteasomal degradation may be influenced by various environmental and physiological conditions, this provides a potential new mechanism by which the cell may modulate protein assembly disorders and proteinbased inheritance.