Isolation and characterization of SUG2. A novel ATPase family component of the yeast 26 S proteasome.

Using a genetic strategy designed to find proteins involved in the function of the Saccharomyces cerevisiae transcriptional activator GAL4, we isolated mutants in two genes which rescue a class of gal4 activation domain mutants. One of these genes, SUG1, encodes a member of a large family of putative ATPases, the Conserved ATPase containing Domain (CAD) proteins (also known as AAA proteins) that are involved in a wide variety of cellular functions. Subsequently, SUG1 was identified as a subunit of the 26 S proteasome. We have now cloned the gene defined by the second complementation group. SUG2 encodes an essential 49-kDa protein that is also a member of the CAD family and is 43% identical to SUG1. The mutation in sug2-1, like that in sug1-1, is found in the CAD near the highly conserved ATPase motif. We present biochemical and genetic evidence that SUG2 is associated in vivo with SUG1 and is a novel CAD protein subunit of the 26 S proteasome. With its highly conserved mammalian homologs, human p42 and ground squirrel CADp44, SUG2 defines a new class of proteasomal CAD proteins.

Using a genetic strategy designed to find proteins involved in the function of the Saccharomyces cerevisiae transcriptional activator GAL4, we isolated mutants in two genes which rescue a class of gal4 activation domain mutants. One of these genes, SUG1, encodes a member of a large family of putative ATPases, the Conserved ATPase containing Domain (CAD) proteins (also known as AAA proteins) that are involved in a wide variety of cellular functions. Subsequently, SUG1 was identified as a subunit of the 26 S proteasome. We have now cloned the gene defined by the second complementation group. SUG2 encodes an essential 49-kDa protein that is also a member of the CAD family and is 43% identical to SUG1. The mutation in sug2-1, like that in sug1-1, is found in the CAD near the highly conserved ATPase motif. We present biochemical and genetic evidence that SUG2 is associated in vivo with SUG1 and is a novel CAD protein subunit of the 26 S proteasome. With its highly conserved mammalian homologs, human p42 and ground squirrel CADp44, SUG2 defines a new class of proteasomal CAD proteins.
The GAL4 protein of Saccharomyces cerevisiae is responsible for a very large stimulation (ϳ1000-fold) in the transcription of genes required for galactose metabolism under inducing conditions. The N-terminal 97 amino acids of GAL4 encode the DNA binding domain, while activation function and interaction with the negative regulator GAL80 map to 34 amino acids near the C terminus (1). Partial deletion of this activation domain in the gal4D and gal4-62 alleles leads to a defect in activation by GAL4 and an inability to grow on galactose as the sole carbon source (2,3). A frameshift mutation of gal4-62 that changes the C-terminal amino acids of the truncated protein from FGITT to FMNV restores the ability to grow on galactose. While the truncated gal4D activation domain retains only 3% of wild type activity, the frameshift mutation restores activity to 40% (4,5). We reasoned that if a small change in the truncated activation domain could restore function, we might genetically identify proteins contacted by GAL4 in the activation process by selecting extragenic suppressors of gal4D.
This strategy yielded mutants in two complementation groups (2). 1 One of these, sug1, had been isolated independently by Matsumoto et al. (3) as a suppressor for galactose (sug) of the gal4-62 allele and was presumed to be an information suppressor. However, sug1-1 suppresses gal4D, which terminates at the same amino acid as the gal4-62 allele with a frameshift and multiple stop codons, arguing that sug1-1 does not act as an information suppressor.
We have previously characterized the sug1-1 allele. It is able to suppress a variety of mutations in the activation domain of GAL4, short of its complete deletion, but is unable to suppress mutations in other regions of the protein. The SUG1 gene was cloned and sequenced and was one of the first identified members of a large family of proteins that contain a highly conserved 220 amino acid motif. We have referred to this as the CAD (Conserved ATPase-containing Domain) (2) because one of its most highly conserved features is a Walker-type nucleotide binding motif. Proteins that share this motif are also collectively known as the AAA family (6). They are involved in a remarkably wide range of cellular processes, including vesicle fusion, proteolysis, peroxisomal, and mitochondrial biogenesis and transcription. The basic function of the CAD is still unknown.
Although sug1 was originally identified as a suppressor of transcriptional defects, subsequent studies suggested that SUG1 is a component of the 26 S proteasome (7-9), a large multiprotein complex that degrades proteins targeted for degradation by the ubiquitin pathway in an ATP-dependent fashion. It is composed of at least two functionally interdependent parts; the 20 S, or core catalytic subunit of the proteasome that can act as a peptidase in isolation, and a 19 S regulatory subunit (also called the PA700 (10)) that is required for degradation of proteins (reviewed in Refs. [11][12][13]. A screen for suppressors of the cdc28N-1 mutation in yeast yielded a sug1 mutation (cim3) along with a mutation in a known component of the 26 S proteasome, CIM5 (14). Both the cim5 mutation and the cim3 mutations affect the half-life of some proteasomal substrates in vivo (14,15). More convincingly, a protein nearly identical to the human homolog of SUG1 (16) was biochemically isolated as a component of PA700, a complex that stimulates the activity of 20 S proteasome purified from bovine red blood cells (17). Subsequently, it has been physically demonstrated that yeast SUG1 is a component of the 26 S proteasome (7,9), a localization it shares with four other CAD proteins (listed as homologous pairs, yeast/human): YTA5/S4 (18,19), CIM5/MSS1 (14,20), YTA1/TBP1 (18,21), and YTA2/TBP7 (18,22). These proteins are postulated to act as "reverse chaperones," unwinding protein substrates in an ATP-dependent manner for digestion by the 26 S proteasome (13,23). The mechanism for rescue of mutant gal4 proteins by sug1-1 is not yet understood.
Only two complementation groups were recovered in our selection for suppressors of gal4D. There were 14 independent isolates in each group, 1 suggesting that the selection was saturated. Matsumoto et al. (3) found, and we have confirmed, that a mutation in the second complementation group, sug2-1, is synthetically lethal with sug1-1, suggesting that the two genes are functionally related and may be physically associated with each other. Therefore, we reasoned that characterization of SUG2 could be important to understanding proteasomal function and might provide insight into the mechanism of gal4D rescue.
We describe here cloning and characterization of the SUG2 gene and the identification of its protein product as a novel CAD family subunit of the yeast 26 S proteasome. Like SUG1, SUG2 has highly conserved mammalian homologs. With its human homolog p42 (24) and its functionally interchangeable ground squirrel homolog CADp44 (25), SUG2 defines a new, sixth class of proteasomal CAD proteins.  (29). Sc527 (␣ sug2-1 gal4-69) was made by changing the mating type of Sc392 (30). Sc528 is a derivative of Sc392 carrying an integrated wild type SUG2 along with a URA3 marker on integrating vector pRS306 (31). The genomic SUG1 gene of Sc342 was tagged at the N terminus with the 11-amino acid T7 (S10) epitope tag (Novagen) to produce Sc507. Three tandem copies of the epitope tag are appended to the N terminus of SUG2 in Sc562, also derived from Sc342. Yeast transformations were done by the lithium acetate methods (32,33). Yeast media (34) contained as carbon sources either 2% galactose, 2% glucose, or 3% glycerol plus 2% lactic acid as indicated. The galactose anti-metabolite 2-deoxygalactose was used at 0.5% w/v where noted.

Strains and
DNA Manipulations and Sequencing-The yeast centromeric (single copy) vector YCP50 (35) was used to express physiological amounts of proteins, and the 2-m (multicopy) pYEP24 (36) allowed moderate overexpression. pMTL1 (a gift of C. Giroux) was used to express proteins under the GAL1-10 promoter. Plasmid pRS306 (31) was used as integrating vector for SUG2. Vent polymerase (New England BioLabs) was used for all PCR 2 . DNA manipulations were done according to standard protocols (37).
For mapping the putative SUG2 clone to the sug2-1 locus, a PCR fragment (1.58 kb) from the genome containing the entire SUG2 gene as well as 5Ј-and 3Ј-untranslated regions was blunt-end cloned into the SmaI site of pUC118. By using ClaI sites in the oligonucleotides, the SUG2 containing fragment (from 210 base pairs 5Ј of the ATG to 41 base pairs 3Ј of the stop codon) was subcloned to integrating vector pRS306 carrying a URA3 marker. The recombinant plasmid was linearized within the gene at the single BglII site and transformed into Sc392. Genomic DNA was prepared from 2 Ura ϩ isolates. The locus of integration was confirmed by diagnostic PCR using one oligonucleotide in the genome and another in the URA3 gene.
The SUG2 disruption construct was generated by replacing the 1.05-kb StuI-BclI fragment (leaving 54 amino acids at the N terminus and 29 amino acids at the C terminus) of SUG2 in pUC118 with a 1.1-kb SmaI-BamHI fragment containing the URA3 gene by blunt cloning. The SUG2::URA3 fragment was removed by ClaI digestion and transformed into Sc342 carrying pMTL1-SUG2 (SUG2 under GAL1/10 control). Successful integration events were selected on galactose plates lacking uracil and confirmed by diagnostic PCR. Site-directed mutagenesis was performed with the Sculptor kit (Amersham Corp.).
pH6-SUG2 was constructed by amplifying the open reading frame of the SUG2 gene by PCR with a 5Ј primer that produced an NcoI site at the start codon. The product was subcloned into the prokaryotic expression plasmid H6-pQE60, a derivative of pQE60 (Qiagen), to produce pH6-SUG2 which encodes SUG2 tagged at the N terminus with six histidines (Met-Ala-(His) 6 -Ala-SUG2). The sequence of SUG2 was determined by automated cycle sequencing using fluorescently tagged terminator base analogs (Applied Biosystems Inc.). Comparison of the SUG2 sequence with other CAD proteins in the GenEMBL combined data base was done at the National Center for Biotechnology Information using the BLAST network service. Percent identity values for pairwise comparisons of CAD proteins were calculated using the GAP program, and the dendrogram of yeast CAD proteins was generated using the Pileup program in the Wisconsin Sequence Analysis Package (Genetics Computer Group).
␣and ␤-Galactosidase Assays-Yeast cells were harvested at A 600 ϭ 0.6 -0.8. Extracts were prepared and assayed for ␤-galactosidase (38) or ␣-galactosidase activity (39) as described. Activity was converted to ⌬A 400 •min Ϫ1 mg Ϫ1 protein. One unit of ␤-galactosidase is equal to a 0.0045 ⌬A 400 •min Ϫ1 mg Ϫ1 protein. One unit of ␣-galactosidase is equal to a 0.0182 ⌬A 400 min Ϫ1 mg Ϫ1 protein.
Two-hybrid Interaction Assay-Two-hybrid reagents developed by Elledge and co-workers (28) were used in this study. The full-length SUG1 gene was fused at its N terminus with the GAL4 DNA binding domain on pASCYH2 and used as bait in the two-hybrid assay. Strain Y190 carrying the bait was transformed with a cDNA library of fusions with the GAL4 activation domain on pACTII. Approximately 100,000 transformants were selected on His Ϫ Leu Ϫ Trp Ϫ medium containing 25 mM 3-aminotriazole. Blue colonies were identified using the 5-bromo-4-chloro-3-indoyl ␤-D-galactoside filter lift assay for ␤-galactosidase activity as described (28). Positive colonies failing to activate nonspecific fusions were chosen for further study.
Generation of Antibodies-SUG2 tagged at the N terminus was produced in bacterial strain BL21(DE3) carrying pREP4 and pH6-SUG2 and was isolated as inclusion bodies. This protein was approximately 50% pure as estimated after SDS-PAGE and Coomassie Blue staining (data not shown) and was further purified by preparative SDS-PAGE. The H6-SUG2 band was cut out of the gel, and a suspension of finely macerated acrylamide in phosphate-buffered saline containing purified H6-SUG2 along with an equal mass of Adjuprime (Pierce) was used for all inoculations. After obtaining preimmune serum, antibodies were raised in mice by standard methods (40). Extracts for testing the specificity of the SUG2 antibody were prepared by directly boiling yeast grown in yeast extract peptone glucose medium and harvested at an A 600 nm of 1.0 in 2 ϫ SDS sample buffer.
Preparation of Yeast Extract and Gel Filtration Chromatography-Two liters of yeast strain Sc507 (wild type for GAL4, SUG1, and SUG2) were grown at 30°C to an A 600 nm of 1.0 in yeast extract peptone medium with 2% glucose as the carbon source. The cells were harvested by centrifugation, washed once with cold water and re-centrifuged, then resuspended in 3 ml/g wet weight SCED (1.2 M sorbitol, 0.1 M sodium citrate, pH 5.8, 60 mM EDTA, 5 mM dithiothreitol). 50 l of ␤-glucuronidase (Sigma) and 1 mg of zymolyase 20T (Seikagaku Corp.) was added per g wet weight of cells. The suspension of cells was shaken at 150 rpm in an air incubator at 30°C until completely spheroplasted (30 -45 min). The spheroplasts were centrifuged at approximately 700 ϫ g at 4°C for 5 min and then gently resuspended in cold 1.2 M sorbitol and recentrifuged. This washing step was repeated. The pelleted spheroplasts were then resuspended in an equal volume of column buffer A (20 mM Tris, pH 8, 2 mM dithiothreitol, 20 mM potassium acetate, 1 g/ml leupeptin, 1 g/ml pepstatin, 0.5 mM phenylmethylsulfonyl fluoride, 0.5 mM ATP, and 20% glycerol) made without the glycerol and incubated on ice for 5 min resulting in lysis of the spheroplasts. One spheroplast pellet volume of column buffer A containing 60% glycerol was then added to give a final concentration of glycerol of approximately 20%. The viscous extract was sonicated for seven bursts of 1 s using the microprobe of a Branson Sonifier 450 and then centrifuged at 35,000 ϫ g for 45 min at 4°C. The supernatant was transfered to a fresh tube and respun. The cleared extract was then diluted to 10 mg/ml protein with column buffer A. 10 ml (100 mg) was fractionated on a 2.5 ϫ 100 cm Sephacryl S400 HR (Pharmacia Biotech Inc.) gel filtration column running at 1 ml/min at 4°C, and 15-ml fractions were taken. The remainder of the extract and all fractions were snap-frozen in liquid nitrogen in aliquots.
10 ml from 26 S peak fraction 8 of the crude extract fractionation was depleted of ATP by treatment with 4 units of apyrase type 5 (Sigma) for 1 h on ice. An appropriate amount of 2 M KCl was added to bring the fraction to 300 mM KCl, and it was then refractionated on the S400 HR column in column buffer B (20 mM Tris, pH 8, 2 mM dithiothreitol, 20 mM potassium acetate, 1 g/ml leupeptin, 1 g/ml pepstatin, and 300 mM KCl).
Proteasome peptidase activity was assayed, in the absence or presence of 0.05% SDS, against Suc-Leu-Leu-Val-Tyr-AMC (Bachem) as described (41). Protein concentrations were determined with the Bio-Rad protein assay using bovine serum albumin as the standard.
Fractions were separated by SDS-PAGE and blotted to polyvinylidene difluoride. Tagged SUG1 was detected with a monoclonal antibody to the T7 S10 epitope (Novagen). SUG2 was detected with mouse antiserum. 20 S proteasome subunits were detected with antiserum against the yeast enzyme generously provided by K. Tanaka. Immunoblots were visualized using horseradish peroxidase-conjugated goat anti-mouse and anti-rabbit IgGs (Pierce), the Rennaissance chemiluminescent detection system (Amersham Corp.), and x-ray film (Amersham Corp.).
Immunoprecipitations-Fixed protein Aϩ S. aureus cells were purchased from Boehringer Mannheim and prepared as suggested by Mark Biggin. A 10% suspension of the cells was washed three times with 1 volume of TES (50 mM Tris, pH 8, 2 mM EDTA, 0.2% Sarcosyl). They were then resuspended in 2 volumes of phosphate-buffered saline supplemented with 3% SDS and 10% ␤-mercaptoethanol and incubated in boiling water for 30 min. After washing two times in TES, cells were frozen aliquoted at Ϫ80°C as a 20% (w/v) suspension in TES until use.
To load cells with antibody, 100 l of cell suspension was washed three times with 1 ml of NET buffer (150 mM NaCl, 1 mM EDTA, 50 mM Tris, pH 8) and then incubated with 40 l of anti-yeast 20 S proteasome or preimmune serum in a total volume of 1 ml of NET for 2 h at 4°C with inversion. Cells were washed three times with 1 ml of NET buffer, taken up to a final concentration of 10% w/v in immunoprecipitation buffer, and stored at Ϫ80°C.
Samples were pre-cleared before use in immunoprecipitation reactions. Extract or column fractions were diluted to a final volume of 400 l with IP buffer (the appropriate column buffer supplemented with 0.05% Nonidet P-40, 5 mM MgCl 2 , 2 mM ATP, and 0.5 mM EDTA). 20 l of 20% unloaded protein Aϩ S. aureus cells were added, and the reaction was nutated at 4°C for 1 h and then spun at 12,000 rpm in a microcentrifuge. The supernatant was used as the input for immunoprecipitations.
For immunoprecipitations, 20 l of loaded S. aureus cells were added to the pre-cleared extract and nutated at 4°C for 1 h. After gently pelleting the cells (7,500 rpm in a microcentrifuge), the supernatant was removed and concentrated by precipitation with 10% trichloroacetic acid using 150 g/ml deoxycholate as carrier. The pellet was then subjected to three 5-min washes with 1 ml of IP buffer. Both the concentrated supernatant proteins and the pellet were resuspended in equal volumes of 1 ϫ SDS-PAGE loading buffer supplemented with 6 M urea and boiled for 5 min. Equal aliquots of the samples were separated on a 10% Tricine SDS-polyacrylamide gel, transferred to polyvinylidene difluoride membrane, and detected by Western blotting. The 20 S proteasome was detected in immunoprecipitations using the anti-20 S antibody that had been biotinylated using Biotin (Long Arm) NHS (Vector Laboratories) following the manufacturer's instructions. Avidin-conjugated horseradish peroxidase (Molecular Probes, Inc.) was used for visualization.
Image Processing-Images on x-ray film were scanned using the Ofoto 1.1 software and an Apple One Scanner at 8 bit/300 dpi and stored as PICT files. Annotations and labels were added in Adobe Photoshop, and the images were printed on a Kodak XLS 8600 PS dye sublimation printer.

Cloning of SUG2 by Complementation-A strain relying on
gal4-69 (which, like gal4-62, encodes a protein truncated at amino acid 853) for activation of genes in the GAL regulon is unable to grow on galactose as the sole carbon source (Gal Ϫ ) and is resistant to the galactose anti-metabolite 2-deoxygalactose (2-DG ϩ ). In an sug2-1 background, however, growth on galactose is restored (Gal ϩ ) and growth on 2-DG is inhibited (2-DG Ϫ ). In order to isolate the dominant, wild type SUG2 gene, we transformed Sc392 (a sug2-1 gal4-69) strain with multi-copy and single-copy genomic libraries.
For each library approximately 10,000 transformants se-lected on glucose medium were replica-plated to both galactose and glycerol/lactic acid/2-DG (gly/lac/2-DG) medium. Colonies that could not utilize galactose (Gal Ϫ ) and grew on gly/lac/2-DG (2-DG ϩ ) were selected for further characterization. The single copy library did not yield any candidate clones. Plasmids were isolated from the five multi-copy transformants obtained, colony purified in Escherichia coli, and re-transformed into Sc392. Three of these plasmids complemented the sug2-1 mutation, i.e. they prevented the rescue of gal4-69 by the recessive sug2-1 allele. Restriction enzyme maps of these plasmids were identical. One plasmid, p49-7, was chosen for further study.
Sequencing the entire insert of p49-7 revealed three fulllength open reading frames, the previously identified GCD1 (42) and CDC31 (43) genes, and between them a novel CAD family gene with 43% identity to SUG1 (Fig. 1). Given the similarity to SUG1, we tested whether this CAD gene could complement the suppression of gal4-69 by sug2-1. A 1.58-kb fragment between GCD1 and CDC31, containing the entire predicted open reading frame of the CAD gene, as well as 5Јand 3Ј-untranslated regions, was PCR-amplified and cloned. This fragment alone, when carried on a multicopy plasmid, was able to suppress the sug2-1 rescue of gal4-69 (Table I) while fragments containing GCD1 and CDC31 did not (data not shown).
The 1.58-kb fragment between GCD1 and CDC31 only partially suppressed the rescue of gal4-69 when carried on a single copy plasmid (Table I). To determine if the cloned CAD gene was SUG2 or if it was an extragenic multicopy suppressor, it was integrated into an sug2-1 strain. A clone of the novel CAD gene on the URA3 integration vector pRS306 was cut within the CAD open reading frame and transformed into Sc392 (a sug2-1 gal4-69). All 12 integrants selected for uracil prototrophy were Gal Ϫ . Integration of this fragment diminished gal4-69 activity to a level equal to that found in a SUG2 wild type background (Table I), consistent with integration at the sug2-1 locus and generation of one wild type copy of SUG2 by recombination. To further demonstrate that the novel CAD gene maps to the sug2-1 locus, we performed crosses and random spore analysis. A strain in which the URA3 gene was targeted to the novel CAD protein locus by integration was crossed with a mating type switched derivative of the parental strain, Sc527 (␣ sug2-1 gal4-69) carrying a LEU2 marker plasmid (to allow selection of diploids). Diploids selected on glucose medium lacking uracil and leucine were used for random spore analysis. A total of 49 spores were assayed for Ura, Gal, and 2-DG phenotypes. 21 of the 49 spores were Ura ϩ . All of the Ura ϩ spores were Gal Ϫ /2-DG ϩ (no complementation of gal4-69), and all Ura Ϫ spores were Gal ϩ /2-DG Ϫ , verifying that the URA3 marker integrated at or very close to the sug2-1 locus. We conclude that sug2-1 is an allele of the CAD gene isolated, hereafter called SUG2.
SUG2 Is an Essential Gene-SUG1 (2) and other tested yeast proteasomal CAD proteins (YTA1, YTA2 (18), and CIM5 (14)) are encoded by essential genes. To test whether SUG2 is also an essential gene, we generated a disruption construct in which all of the predicted SUG2 open reading frame except the 54 N-terminal and 29 C-terminal amino acids was replaced with the URA3 gene. This construct was used in a one-step disruption protocol by digesting at restriction sites flanking SUG2 and then transforming the sug2::URA3 fragment into Sc342 (a SUG2 ura3-1 leu2-3, 112) carrying the pMTL-SUG2 plasmid. This plasmid carries a LEU2 marker and expresses SUG2 under the control of the GAL1/10 promoter, which is tightly repressed on glucose medium.
Transformants were selected on galactose plates lacking uracil and leucine, which allowed plasmid MTL-SUG2 to supply SUG2. Individual transformants were then streaked to selective plates with either galactose or glucose as the carbon source. Growth occurred on galactose but not on glucose plates as expected if SUG2 is essential for vegetative growth (Fig. 2). The disruption construct was also transformed into the diploid strain Sc344 (a/␣ SUG2/SUG2 ura3-1/ura3-1) which was then sporulated. Two of four spores were viable in each of five tetrads, and all viable spores were Ura Ϫ (data not shown). As in the SUG1 deletion, SUG2 deleted spores germinated, but grew for only 3-5 divisions. Therefore SUG2, like SUG1, is essential for growth.
Characterization of the sug2-1 Mutation-The mutation responsible for the rescue of gal4D by sug1-1 is the change of conserved glycine 214 in the CAD to aspartate. This finding suggests that an alteration in ATPase function may be respon-sible for the sug1-1 phenotype (2). A similar mutation in sug2-1 would suggest that sug1-1 and sug2-1 might share a common mechanism for suppression of truncated GAL4.
In order to identify the sug2-1 mutation, we amplified the mutant (from Sc392) and wild type (from Sc311) SUG2 alleles in two pieces from genomic DNA by PCR and entirely sequenced them. Two differences between the wild type clone and sug2-1 were found. One coded for an alanine to valine substitution at amino acid 56, the other for a glutamic acid to lysine substitution at amino acid 300. Glutamate 300 is near the Walker nucleotide binding motif of the CAD and is conserved in most members of this family of proteins, whereas alanine 56 is outside of the CAD region.
Site-directed mutagenesis was used to insert the E300K mutation into the wild type sequence. The synthetic E300K allele was used in a one-step gene replacement of SUG2 in YJOZ (⌬GAL4) carrying gal4D on a plasmid. Transformants were screened for the ability to grow on galactose. Sequencing of the SUG2 allele from one of these GAL ϩ isolates revealed only the E300K mutation, demonstrating that this mutation is sufficient for suppression of the gal4D phenotype. We conclude that the SUG2 mutation responsible for the rescue of gal4-69 is a glutamate to lysine alteration within the CAD region (Fig. 1).
Two-hybrid Interaction between SUG2 and SUG1-The sug1-1 and sug2-1 mutations have similar phenotypes, and the combination of these two alleles is lethal (3). 1 This synthetic lethality suggests that these two proteins may physically interact with each other in a protein complex that is essential for viability. Two-hybrid analysis (28) provided additional evidence of SUG1-SUG2 interaction. SUG1 fused to the DNA binding domain of GAL4 was used as bait in a screen for cDNAs encoding SUG1 interacting proteins. Approximately 100,000 transformants were screened yielding 8 clones that displayed significant reporter gene activity. Plasmids were isolated from these isolates and colony-purified in E. coli. Upon re-transformation, 5 of these remained positive. Each of these was further tested and eliminated from consideration if they could activate transcription in collaboration with nonspecific GAL4 DNA binding domain fusions. Three bona fide interacting clones were isolated. Upon sequencing, one of these was found to be identical to SUG2, reinforcing the conclusion that SUG1 and SUG2 associate with each other in vivo (Table II). The other two clones isolated in this screen encoded YTA2 (18) (data not shown), the yeast homolog of the 26 S proteasomal subunit TBP7 (22).
Localization of SUG2 to the 26 S Proteasome-The finding that SUG2 interacts with SUG1, a component of the 26 S proteasome, in a two-hybrid assay suggested that SUG2 might also be part of this protein complex. To verify this, we physi- FIG. 2. SUG2 is essential for the viability of yeast. SUG2 was deleted in strain W303 a carrying a single copy of SUG2 under the control of the glucose repressible GAL1/10 promoter (pMTL1-SUG2). W303 a sug1::URA3 (pMTL1-SUG1) and W303 a (pMTL1) were used as controls. Single colonies from each transformant were streaked on galactose (left) and glucose (right) containing plates. W303 a sug2::URA3 (pMTL1-SUG2) grew on galactose but not on glucose medium.
cally localized SUG2. Antibodies were generated to recombinant SUG2 produced in E. coli. These antibodies detected a single protein with an apparent molecular mass of approximately 60 kDa in Western blots of crude yeast extract. A larger immunoreactive band was seen in a strain in which SUG2 is increased in size by the addition of a peptide tag, demonstrating the specificity of the antibody which was used for detection of SUG2 in the following experiments (Fig. 3).
Immunoprecipitations using antibodies raised against the purified 20 S proteasome were performed on crude yeast extract. Both SUG2 and SUG1 co-immunoprecipitated with subunits of the 20 S proteasome but were not precipitated by pre-immune serum (Fig. 4A). To determine the size of the complex containing SUG2, SUG1, and 20 S proteins, an aliquot of the extract was separated on a Sephacryl S400 HR gel filtration column. SUG2 co-fractionated with SUG1, subunits of the catalytic 20 S core of the 26 S proteasome and 20 S peptidase activity, with the peak at a molecular mass of approximately 2000 kDa, consistent with the size of the 26 S proteasome (8,9,13) (Fig. 4B). To confirm that SUG2 was associated with SUG1 and 20 S subunits in these fractions, we performed immunoprecipitations. SUG1 and SUG2 co-immunoprecipitate with antibodies against the 20 S proteasome in both peak fractions (Fig. 4C). These results imply that, like SUG1 (7,9,17), SUG2 is a component of the 26 S proteasome. Increased dissociation of SUG1 and SUG2 from 20 S proteins during immunoprecipitations in fraction 8 versus fraction 7 may reflect heterogeneity in the 26 S peak.
The 19S/PA700 regulator of the proteasome can be dissociated from the 20 S core by depletion of ATP and/or high salt treatment (44). 3 26 S peak fraction 8 was treated with apyrase to deplete it of ATP and brought to 300 mM KCl. The treated fraction was then refractionated on the S400 HR gel filtration column equilibrated with buffer B which contains no ATP, no glycerol, and 300 mM KCl. SUG1, SUG2, and 20 S subunits still co-fractionated but at a much smaller apparent molecular mass. Since both the 20 S proteasome and the 19S/PA700 are estimated to have very similar molecular masses of approximately 700 kDa (13,45), this is consistent with the splitting of 3 G. DeMartino, personal communication.

FIG. 3. Antibodies generated against recombinant H6-SUG2
produced in E. coli are specific for SUG2. Equal amounts of protein extracts from WT and S10 3 SUG2 (triple-tagged SUG2) strains were separated by SDS-PAGE and detected by Western blot with anti-SUG2 antiserum or anti-epitope antibody. The weaker signal seen in the Sc562 extract is due to increased degradation of triple-tagged SUG2.
FIG. 4. SUG2 is physically associated with SUG1 as a component of the 26 S proteasome. A, SUG2 and SUG1 are co-immunoprecipitated from crude yeast extracts with subunits of the 20 S proteasome. Immunoprecipitations were performed with either preimmune (PI) or 20 S subunit antiserum (20 S) as described under "Materials and Methods." The input (I), supernatants (S), and pellets (P) were separated by SDS-PAGE and probed for SUG2, SUG1, and 20 S proteins by Western blotting. Cross-reactivity of the secondary (streptavidin-HRP) to IgG light chain in the 20 S panel is designated with an asterisk. B, SUG2 co-elutes with SUG1 and subunits of the 20 S proteasome from an S400 HR gel filtration column. A crude yeast extract was fractionated by gel filtration using Sephacryl S400 HR gel filtration column as described under "Materials and Methods." Fractions were separated by SDS-PAGE and probed for SUG2, SUG1, and 20 S proteins by Western blot. The peptidase activity (filled circles) was assayed against Suc-Leu-Leu-Val-Tyr-AMC. Total protein (open circles) was determined with the Bio-Rad protein assay. C, SUG2 and SUG1 are co-immunoprecipitated with subunits of the 20 S proteasome from 26 S peak fractions. Aliquots of fractions 7 and 8 were immunoprecipitated and assayed as in A.

TABLE II SUG2 and SUG1 interact by two-hybrid assay
Plasmids were introduced into yeast strain Y190 (gal4 gal80) containing lacZ and HIS3 genes regulated by the GAL1 promoter. Constructs were assayed in duplicate. The values varied within 5%.
the 26 S proteasome into 20 S and 19S/PA700 components under these conditions (Fig. 5A). Addition of 0.05% SDS to peptidase assays of column fractions stimulated the activity of the ϳ500-kDa peak fraction approximately 14-fold. Stimulation by SDS is characteristic of 20 S proteasome not associated with activating proteins (44). We believe that an estimated molecular mass of approximately 500 kDa for proteins in this fraction by comparison to molecular weight standards is consistent with this interpretation within experimental error. Immunoprecipitation experiments show that in these fractions SUG2 and SUG1 no longer co-immunoprecipitate with 20 S proteasome subunits (Fig. 5B) consistent with co-migration of dissociated 19S/PA700 and 20 S proteasome because of their similar molecular masses. Considering the two hybrid and biochemical data together, we conclude that SUG2 is physically associated with SUG1 as a component of the 19S/PA700 regulatory complex of the yeast 26 S proteasome. DISCUSSION We have cloned and characterized SUG2, an essential gene of the CAD family in yeast. A mutation in SUG2 that rescues a partial deletion of the activation domain of the GAL4 maps to the CAD of SUG2, as does a mutation in SUG1 which has the same phenotype. Like the closely related SUG1, our data show that SUG2 is a component of the 19S/PA700 regulatory cap of the 26 S proteasome.
Comparison of the SUG2 sequence with all other CAD proteins reveals that, as expected, it is most related to the other yeast family members localized to the 19S/PA700 subunit of the 26 S proteasome (Fig. 6). However, generation of dendrogram for this subset of CAD proteins from all species (see Fig.  4 in Ref. 25) reveals that SUG2 can be classified as a member of a novel subgroup. In the same class are its recently cloned human and ground squirrel homologs p42 and CADp44, respectively. Although the human p42 (24) and ground squirrel CADp44 (25) are 48 and 49 amino acids shorter at the N terminus than SUG2, they are 65 and 67% identical, respectively, over the region they share. Despite its N-terminal truncation compared with yeast SUG2, CADp44 is able to complement the deletion of SUG2 in yeast (25). This is another example of the remarkable functional conservation of these proteins and an indication that the N terminus of SUG2 may not be required for its essential function(s). The identities between SUG2 and its mammalian homologs are similar to those found between yeast SUG1 and its human homolog, TRIP1 (76%), which can functionally replace SUG1 in yeast (16).
Besides the 200-amino acid region of homology that all proteins of this family share, all of the previously identified proteasomal CAD proteins except for YTA5/S4 are also notable for the presence in their N-terminal region of a strongly predicted coiled-coil region of approximately 42 amino acids, or 6 repeats of the 7 amino acid motif (Fig. 7) (13). SUG2 also has a strongly predicted coiled-coil motif of 43 amino acids. The significance of this motif is not clear, but it has been proposed to play a structural role in the assembly of the PA700 regulatory complex. It could also potentially play a role in the attachment of the PA700 to the ␣ subunits of the 20 S proteasome. Notably, FIG. 5. SUG2 and SUG1 co-fractionate with, but are not physically associated with, the 20 S proteasome after dissociation of the 26 S proteasome. A, SUG2 and SUG1 co-fractionate with the 20 S proteasome after ATP depletion and salt treatment. Peak 26 S fraction 8 from the separation shown in Fig. 4 was depleted of ATP with apyrase and then taken to 300 mM KCl and refractionated on the Sephacryl S400 HR gel filtration column in column buffer B (300 mM KCl) as described under "Materials and Methods." Fractions were separated by SDS-PAGE and probed for SUG2, SUG1, and 20 S proteins by Western blot. The peptidase activity was assayed against Suc-Leu-Leu-Val-Tyr-AMC in the presence (ϫ) or absence (filled circles) of 0.05% SDS. Total protein (open circles) was determined with the Bio-Rad protein assay. B, SUG2 and SUG1 are not physically associated with the 20 S proteasome after ATP depletion and salt treatment. Fraction 10 was immunoprecipitated with antibodies against subunits of the 20 S proteasome as in Fig. 4A. While most of the 20 S proteasome was precipitated, none of the SUG1 or SUG2 was co-precipitated. Cross-reactivities of the secondary antibody to IgG heavy chain in the SUG1 panel and of streptavidin-HRP to IgG light chains in the 20 S panel are marked with an asterisk.
FIG. 6. SUG2 is one of six closely related yeast proteasomal CAD proteins. A dendrogram showing relatedness between yeast CAD proteins, generated using the PILEUP program (GCG), suggests that there are only six members of the proteasomal subfamily of CAD proteins. The BLAST server of the SCG project at Stanford was used to search the entire yeast genome using the tblastn algorithm with SUG2 and SUG1 as input. For both proteins, the same set of 20 proteins showed the most similarity. Bcs1, the least related protein shown, was chosen as the cutoff point. It gave smallest sum probabilities of 2 ϫ 10 Ϫ10 and 2 ϫ 10 Ϫ9 for SUG2 and SUG1, respectively, while the next most related protein (on chromosome II between 600,000 and 605,000) gave corresponding scores of 0.039 and 0. 29. both human and ground squirrel homologs of SUG2 also have predicted coiled coil domains despite their N-terminal truncation relative to SUG2 (not shown).
DeMartino et al. (24) have shown that the apparent bovine homolog of SUG2, p42, is present in both the PA700/19S and a separate complex, the modulator, which is able to further stimulate the 20 S proteasome's proteolytic activity in a PA700-dependent fashion. The modulator is a 300-kDa complex that contains two additional proteins, TBP1, present in the PA700 as well, and a novel 27-kDa protein of unknown function that is not a component of the PA700. It is not clear from our data whether there is a yeast analog of the mammalian modulator complex. SUG2 elutes as a broad peak in our first gel filtration column (Fig. 4B), and it is possible that the early fractions of this peak contain a ternary complex of the 26 S proteasome associated with modulator. Likewise, a 300-kDa complex would be expected to migrate very near the 19S/PA700 on our column, so that the tail of the SUG2 peak in Fig. 5A could correspond to modulator-associated SUG2. In order to determine if there is a yeast modulator, it will be important to follow a modulator component that is not a subunit of the PA700/19S. SUG2 is the second CAD protein in the regulatory subunit of the 26 S proteasome to be identified by mutations that rescue the function of weak transcriptional activators. The discovery that SUG2, like SUG1, encodes a protein of the proteasomal regulatory subunit helps to explain the synthetic lethality of sug1-1 and sug2-1 but emphasizes the paradox of proteasomal proteins having strong effects on transcription. It has been proposed that proteasomal CAD proteins are responsible for unwinding substrates for transport into the interior of the 20 S proteasome for degradation (13,23). An obvious hypothesis to explain the rescue of gal4D by sug1-1 and sug2-1 is that altered proteolysis as a result of these mutations leads to an accumulation of the activators, compensating for their decreased potency with increased concentration. It is also possible that accumulation of another protein or proteins due to defective proteolysis may be responsible for amplifying the residual activity of gal4D. If this is true, the genetics of gal4D suppression may provide clues as to the substrate specificity of proteasomal CAD proteins. Multiple alleles of SUG1 and SUG2, and no mutants in other 19S/PA700 CAD proteins, were isolated by the selection for suppressors of gal4D, implying that proteasomal CAD proteins may have overlapping but non-identical protein substrate specificities. Therefore, the presence of six CAD proteins in the PA700 may reflect a requirement for recognizing a wide variety of cellular proteins with varying primary and secondary structure to be unfolded for degradation.
An alternative to mechanisms involving alterations in the proteolytic activity of the 26 S is that the rescue of gal4D may be due to changes in the proposed "reverse chaperone" activity of SUG1 and SUG2 on transcription factors or complexes. In this regard we note that while under our conditions almost all of the 19S/PA700 proteins can be found in association with the 20 S proteasome, an earlier report (46) found that unusual growth conditions were required to stabilize the 19S-20 S interaction. Given the dependence of the stability of 26 S to extract and growth conditions, and the possibility that other activators of the 20 S proteasome may compete with 19S/ PA700 for binding to the 20 S, we feel that the extent to which yeast 19S/PA700 is associated with the 20 S proteasome in vivo is still not clear. Thus it is possible that there is a significant population of free PA700 which might perform "reverse chaperone" activity independent of the 20 S proteasome.
Regardless of mechanism, the phenotypes of SUG2 and SUG1 mutants point to an interesting relationship between transcriptional regulation and protein degradation. The coiled coil probabilities for each of the yeast proteasomal CAD proteins were determined by the method of Lupas et al. (47) on the ISREC Coils server using the MTIDK matrix and a 28-amino acid window. Amino acid number is plotted on the horizontal axis, and coiled coil probability from 0 to 1 is plotted on the vertical axis for each protein.