Rad23 provides a link between the Png1 deglycosylating enzyme and the 26 S proteasome in yeast.

In addition to a role in DNA repair events in yeast, several lines of evidence indicate that the Rad23 protein (Rad23p) may regulate the activity of the 26 S proteasome. We report evidence that a de-N-glycosylating enzyme, Png1p, may be involved in the proteasomal degradation pathway via its binding to Rad23p. Interaction of Rad23p and Png1p was first detected by two-hybrid screening, and this interaction in vivo was confirmed by biochemical analyses. The Png1p-Rad23p complex was shown to be distinct from the well established DNA repair complex, Rad4p-Rad23p. We propose a model in which Rad23p functions as an escort protein to link the 26 S proteasome with proteins such as Rad4p or Png1p to regulate their cellular activities.

Proteins that transit through the secretory pathway are subjected to a quality control system (1) in the endoplasmic reticulum (ER) 1 that recognizes aberrantly folded proteins/glycoproteins. It has been shown that in some cases these misfolded and/or unfolded proteins are degraded by ER-associated degradation mechanisms, which involves retrograde transfer of proteins from the ER to the cytosol followed by degradation by the proteasome (2-7). Previously, we described PNG1, a gene encoding a cytoplasmic deglycosylating enzyme, peptide:N-glycanase (PNGase), that is evolutionarily conserved throughout eukaryotes (8). It has been suggested that this enzyme activity is linked to a proteasomal degradation pathway and has a role in efficient degradation of glycoproteins by the proteasome (8 -13). This would be achieved by removing bulky N-linked glycans from misfolded glycoproteins that are translocated from the lumen of the ER into the cytosol for degradation. However, a physical link between the proteasome and this deglycosylating enzyme has not yet been described.
Rad23p is known to have a pivotal role in nucleotide excision repair (14 -16). Yeast Rad23p stoichiometrically forms a complex with Rad4p to form nucleotide excision repair factor 2 (NEF2). Unlike other NER proteins, the biochemical functions of NEF2 still remain largely unknown. However, NEF2 was recently shown to bind specifically to damaged DNA in an ATP-independent manner (17)(18)(19). While the absence of Rad4p (rad4⌬) causes extreme sensitivity to UV light in yeast, rad23⌬ mutants exhibited only moderate UV sensitivity, indicating that Rad23p may affect the efficiency of the excision repair process rather than directly mediating the repair of damaged DNA (20 -23).
One structural feature of Rad23p is that it contains a ubiquitin-like domain (UbL) at the N terminus that can bind to the 26 S proteasome (24). This association appears to be important for DNA repair (24 -26). In addition, Rad23p was shown to have an overlapping function with Rpn10p, a 19 S proteasome subunit that is known to be a multiubiquitin chain-binding receptor for the proteasome (27). It has been suggested that the Rad23p is a negative regulator of multiubiquitin chain assembly (28). However, thus far, the link between DNA repair and proteasome degradation with respect to Rad23p function remains elusive.
We report here a finding that the deglycosylation enzyme, Png1p, exists as a high molecular weight complex with Rad23p. The Png1p-Rad23p complex was found to be distinct from the well established DNA repair complex, Rad4p-Rad23p (NEF2). In addition, we report that the Png1p-Rad23p complex interacts with the 26 S proteasome. These findings led us to hypothesize that Rad23p may function to link the 26 S proteasome with other proteins, such as Rad4p or Png1p. This "escort" property of Rad23p may explain the complex effect of Rad23pproteasome associations in a variety of cellular processes, including deglycosylation of glycoproteins slated for degradation. EXPERIMENTAL (32)(33)(34).
Construction of Plasmids-DNA manipulations were performed ac-cording to Sambrook et al. (37). Plasmids used in this study are listed in Table I. pCS13 was a kind gift from Dr. Kiran Madura. The yeast genes used in this study were isolated from the genomic DNA of W303-1a by polymerase chain reaction using Vent DNA polymerase (New England Biolabs). For isolation of the PNG1 gene, the following primers were used: 5Ј-AAAAAGAATTC-ATGGGAGAGGTATACGAAAAAA-3Ј (5Јprimer) and 5Ј-AAAAACTCGAG-CTATTTACCATCCTCCCCACGC-3Ј (3Ј-primer). The amplified fragments were digested with EcoRI/XhoI and cloned into EcoRI/SalI sites of pBTM116 (29) and EcoRI/XhoI sites of pRD53 (38). The pBTM116-PNG1 was subsequently digested with MluI/PstI and cloned into MluI/PstI sites of pBTM116-ADE2 (39) to give rise to pBTM116-ADE2-PNG1. pRD53-PNG1His 6 was prepared by amplifying the PNG1 gene using another 3Ј-primer, 5Ј-AAAAACTCGAGT-CAGTGGTGGTGGTGGTGGTGTTTACCATCCTCCCCACG-3Ј, and the EcoRI/XhoI fragments were cloned into pRD53. pRS314-GAL1 PNG1His 6 was prepared by digestion of pRD53-PNG1His 6 with NotI/XhoI and cloning the PNG1-containing fragment into NotI/XhoI sites of pRS314 (40). pESC-TRP-RAD4(Myc) was constructed by co-transformation of ApaI/SalI-digested pESC-TRP (Stratagene) and polymerase chain reaction-amplified RAD4 gene using the following primers: 5Ј-AGAAAA-AACCCCGGATCCGTAATACGACTCACTATAGGGCGAATTCATGA-ATGAAGACCTGCCCAAGG-3Ј (5Ј-primer) and 5Ј-AAGCTTACTCGAG-GTCTTCTTCGGAAATCAACTTCTGTTCGTCGACGTCTGATTCCTC-TGACATCTC-3Ј (3Ј-primer). The Trp ϩ colonies were isolated, and the transformants bearing plasmids with the correct insert were identified by colony polymerase chain reaction. The expression of the Rad4Mycp was further confirmed by Western blotting. The RAD23 gene was isolated by polymerase chain reaction using the primers 5Ј-AAAAAGAA-TTCATGGTTAGCTTAACCTTTAAAAATTTC-3Ј (5Ј-primer) and 5Ј-AAAAAGTCGACTCAGTCGGCATGATCGCTGAATAG-3Ј (3Ј-primer), and the amplified DNA was digested with EcoRI/SalI and cloned into pGAD424 (41). Other truncated versions of RAD23 (amino acids 78 -398, 186 -398, 253-398, 1-185, 1-252, 1-317, and 1-354) were amplified in a similar manner using primers that contain EcoRI and SalI sites and cloned into pGAD424. The sequences of the resulting constructs were confirmed. Yeast Two-hybrid Library Screening-The two-hybrid experiments were carried out as described previously (39). Strain L4O was transformed with pBTM116-ADE2-PNG1 (target plasmid) and a two-hybrid genomic library (pGAD library (Ref. 42; kindly provided by Dr. Phillip James). After screening 2.2 ϫ 10 6 transformants in a yeast genomic library, two distinct GAD library plasmids that showed a reproducible His ϩ ␤-galactosidase ϩ phenotype in PNG1-specific manner were recovered. These candidate GAD plasmids were then transformed into strain L4O and mated with AMR70 cells with each of several plasmids encoding various lexA-plasmids. These candidate plasmids did not exhibit an interaction with the lexA-lamin or lexA-SIR4 C terminus, both of which have been used to test for potential false positives in two-hybrid library screening (39). However, isolated plasmids showed the His ϩ ␤-galactosidase ϩ phenotype with the original target plasmids.
Gel Filtration Analysis-Yeast cytosol was prepared using glass beads as described previously (43) except that the extraction of cytosol was carried out in the presence of a protease inhibitor mixture (final concentrations: leupeptin, 1 g/ml; antipain, 2 g/ml; benzamide, 10 g/ml; chymostatin, 1 g/ml; pepstatin, 1 g/ml; phenylmethanesulfo-nyl fluoride, 1 mM). Preparation of a crude extract of bacterially expressed Png1p was reported elsewhere (8). 0.5 ml (ϳ10 mg of protein) of cytosol were loaded on a Sephacryl S-300 column (Amersham Pharmacia Biotech; 1.5 ϫ 50 cm) equilibrated with elution buffer (20 mM Hepes buffer (pH 6.8), 5 mM magnesium acetate, 1 mM dithiothreitol, 2 mM ATP, 150 mM NaCl, and 0.4 M sorbitol with protease inhibitor mixture as described above), and fractions of 0.9 ml were collected. Fractions were assayed for PNGase activity. For protein determination, 0.3 ml of fractions were precipitated with 10% trichloroacetic acid, and the tagged protein was visualized by Western blotting.
Western Blot Analysis-Western blot analysis was carried out as described (8) using 1:10 dilution of mouse anti-HA (tissue culture supernatant; 12CA5) or 1:1000 dilution of rabbit or mouse anti-Myc (Santa Cruz Biotechnology, Inc., Santa Cruz, CA) rabbit or mouse anti-His 6 (Santa Cruz Biotechnology), or mouse anti-FLAG antibody (M2; Sigma) followed by a 1:2000 dilution with the anti-rabbit or mouse IgG horseradish peroxidase-conjugated secondary antibody (Roche Molecular Biochemicals). 10% SDS-PAGE gels were used, and gels were visualized using chemiluminescence (KPL) after exposure to medical x-ray film (Fuji).
Co-immunoprecipitation-Immunoprecipitation experiments were carried out as previously described (27). Briefly, cell extracts were prepared in lysis buffer (20 mM Hepes-KOH, pH 7.5, 100 mM potassium acetate, 5 mM EDTA, 10% glycerol) including various protease inhibitors as described above, and equal amounts of extract (2 mg of total protein) were incubated with 20 l of protein G-agarose with and without respective anti-Tag antibodies (rabbit anti-Myc (Santa Cruz Biotechnology); rabbit anti-HA (Santa Cruz Biotechnology); rabbit anti-His 6 (Santa Cruz Biotechnology), 1:50 in dilution; or mouse anti-FLAG antibody (Sigma), 1:250 in dilution) and incubated overnight at 4°C. The immunoprecipitates were washed twice with buffer A (50 mM Hepes-NaOH, pH 7.5, 5 mM EDTA, 150 mM NaCl, 1% Triton X-100), dissolved in 20 l of sample buffer, and analyzed by 10% SDS-PAGE. Subsequently, Western blotting analysis was carried out as described above.
UV Sensitivity Analysis-For qualitative UV sensitivity analysis, cells were grown in synthetic dextrose (SD) ϪLeu ϪUra ϩgalactose or ϪLeu ϪUra ϩglucose medium to saturation. Cells were further cultured in the presence of 0.1 mM CuSO 4 for another 2 h, and the cell density was normalized to A 600 ϭ 1.0 (ϳ5.0 ϫ 10 6 cells/ml). Cells were taken from the aliquot with Q-tips and streaked as a single line onto the respective plate (ϪLeu ϪUra ϩgalactose ϩ0.1 mM CuSO 4 or ϪLeu ϪUra ϩglucose ϩ0.1 mM CuSO 4 ). After these plates were covered with a glass plate, they were placed 20 cm distant from the germicidal UV light (254 nm; Sylvania). Then the glass plate was slid parallel to the line of cells so that the cells were exposed to UV for different times from 0 to 8 s. The plate was then incubated at 25°C in the dark for 4 days. If a cell line had UV sensitivity, cells would not be expected to grow all the way across the line when exposed to UV. In contrast, wild-type cells would be expected to exhibit growth across the entire line even when they were exposed to the maximum dose of UV.
For quantitative assay for UV treatment, cells were grown in the respective medium (SD ϪLeu ϪUra ϩglucose or SD ϪLeu ϪUra ϩgalactose) overnight, and after 0.1 mM CuSO 4 was added, they were cultured for 2 h, plated on YPAD at appropriate dilutions, and then Rad23p Binds to De-N-glycosylating Enzyme exposed to 254-nm UV light using a UV cross-linker (UV Statalinker model 1800; Stratagene) at given doses. Cells were plated in triplicate and incubated at 25°C in the dark for 3 days, and the number of surviving colonies were counted.

Png1p Forms a High Molecular Weight Complex in Yeast
Cytosol-Our initial assumption that Png1p might be a part of a multiprotein complex in yeast was based on the results of gel filtration analysis of a cell-free yeast extract. As shown in Fig.  1, PNGase activity measurement of the gel filtration fractions showed distinct differences between the elution position of Png1p in the yeast cytosol and that of Png1p expressed in Escherichia coli. While the bacterially expressed protein showed the expected molecular mass of a monomer form (ϳ45 kDa), the yeast cytosol protein had a much higher molecular mass (estimated to be ϳ200 kDa). This result suggested that in the yeast cytosol Png1p may bind to other proteins to form a high molecular weight protein complex.
Rad23p Specifically Binds to Png1p-The observation described above led us to carry out two-hybrid library screening with PNG1 as a bait to search for possible Png1p-binding proteins. Upon screening 2.2 ϫ 10 6 transformants with a yeast genomic library, we recovered two distinct plasmids that showed a reproducible His ϩ ␤-galactosidase ϩ phenotype in a target plasmid (lexA-PNG1)-specific manner. Sequencing of inserts recovered from these two plasmids showed that they consisted of two different Rad23p fragments (one containing amino acid residues 218 -398 and the other containing residues 229 -398) fused in frame to the GAL4 activation domain sequence (Fig. 2). These plasmids did not exhibit an interaction with either lexA-lamin or lexA-SIR4 C terminus, which have been used to test for potential false positives in two-hybrid library screening (39), suggesting that the interaction observed was specific (data not shown).
Rad23p Binds to Png1p in Yeast Cytosol-To confirm the interaction of Png1p with Rad23p in vivo, gel filtration analysis was carried out to examine the migration of these two proteins. First, we determined the elution position of Png1p in rad23⌬ cells. As shown in Fig. 3, a drastic shift in elution position of this enzyme to a lower apparent mass was observed in rad23⌬ cells, and the elution position was similar to that for bacterially expressed Png1p. This result suggested that the high molecular complex containing Png1p might also contain Rad23p.
To provide direct evidence for interaction of Rad23p and Png1p, FLAG-tagged Rad23p (FLAG-Rad23p; Ref. 24) was expressed under an inducible CUP1 promoter. This protein construct was used to show interaction of Rad23p with the 26 S proteasome as well as with Rad4p (24,27). Png1p is known to be present in extremely low abundance in cells under normal experimental conditions (8), while Rad23p is relatively abundant (17). Therefore, only a minute fraction of Rad23p was expected to be bound to Png1p; this would make it difficult to observe co-migration of Png1p and Rad23p. For this reason, Png1p was overexpressed using a inducible GAL1 promoter in a png1⌬ rad23⌬ strain for this experiment (TSY190; for details of genotypes of strains, see "Experimental Procedures"). When both proteins were expressed, Png1p exhibited two peaks of activity, one corresponding to the higher molecular mass peak and one similar to bacterially expressed Png1p (Fig. 4A). FLAG-Rad23p was shown to co-migrate with the first peak and not with the second peak (Fig. 4, A and B). The proteasome was shown to be present in fractions 34 -40 by activity assay as well as Western blotting using anti-Rpt1p antibody, showing that the first PNGase activity peak did not contain the proteasome complex (data not shown). When only Png1p was overexpressed in the absence of Rad23p (rad23⌬), the higher molecular weight peak was not detectable (data not shown; see also Png1p (png1⌬), the elution position of Rad23p was shifted to a lower molecular weight (Fig. 4C), suggesting that when Png1p was expressed, all of Rad23p detected co-migrated with Png1p.
The C Terminus of Rad23p Is Critical for Its Binding to Png1p-Having biochemical evidence that Png1p binds to Rad23p in yeast cytosol, we determined which domain of Rad23p was involved in interaction with Png1p using the twohybrid assay. Deletion constructs of GAL4 activation domain-RAD23 were tested for binding against lexA-PNG1. As shown in Table II, deletion of N-terminal portions of Rad23p did not have a significant effect on its binding to Png1p. In sharp contrast, when the C-terminal region of Rad23p was truncated, no interaction with Png1p was observed. This result most likely suggested that the C-terminal ubiquitin-associated domain (UBA) of Rad23p (see Fig. 2) was important for its binding to Png1p.
Png1p Interacts with the 26 S Proteasome Subunit in a Rad23p-dependent Manner-Since Rad23p has been shown to require the N-terminal UbL for its interaction with the 26 S proteasome (24), we assumed that Png1p might associate with the 26 S proteasome through Rad23p. To test this hypothesis, co-immunoprecipitation analysis with His 6 -tagged Png1p (Png1-His 6 p) and the GFP/HA-tagged 26 S proteasome subunit, Rpt1p, was carried out. The His 6 -tagged Png1p was expressed under the GAL1 promoter. The His 6 -tagged Png1p was shown to express enzyme activity (data not shown). The GFP/ HA-tagged Rpt1p was previously shown to be integrated into a functional 26 S proteasome complex (31). Cell lysates were prepared from cells expressing Png1-His 6 p with or without FLAG-Rad23p in png1⌬ rad23⌬ cells, and co-immunoprecipitation was carried out using anti-His 6 or anti-HA antibody. As shown in Fig. 5, Png1p and Rpt1p co-immunoprecipitated considerably more than background level, which might represent nonspecific interaction of proteins with resin (compare lanes 1   and 3 with lanes 2 and 4). This interaction was observed only in the presence of Rad23p (Fig. 5; lanes 5 and 7). The fraction of Png1p-Rad23p bound to the 26 S proteasome under these experimental conditions was estimated to be 2-5% based on a comparison of intensity of a band detected by Western blotting of whole extract with the intensity of immunoprecipitate pulled down by Rpt1-GFP-HAp (data not shown).
Png1p Associates with Rad23p in a Complex Distinct from the Rad23p-Rad4p Complex (NEF2)-The C-terminal domain of Rad23p, identified above as a critical binding region to Png1p, was previously shown to be critical for the Rad23p-Rad4p interactions (44). Therefore, it was possible that Png1p and Rad4p might compete for the same binding site on Rad23p. If so, one might expect Png1p to associate with Rad23p as a complex distinct from the Rad4p-Rad23p (NEF2 complex). To test this hypothesis, co-immunoprecipitation experiments were carried out using three different tags: His 6 tag on Png1p, FLAG tag on Rad23p, and Myc tag on Rad4p. As shown in Fig. 6, Rad23p could be detected when either Png1p His 6 or Rad4p (Myc) was precipitated (lanes 1 and 3). This interaction was also confirmed by reciprocal precipitation (lanes 5 and 9). However, no interaction (above background level) was observed between Png1p and Rad4p (compare lane 7 with lane 8, and compare lane 11 with lane 12), supporting the idea that the Png1p-Rad23p complex is distinct from the Rad4p-Rad23p complex (NEF2).
Overexpression of Png1p Causes Moderate UV Sensitivity-Having immunochemical evidence for the existence of two distinct complexes, Rad4p-Rad23p and Png1p-Rad23p, we carried out genetic experiments to confirm the formation of two complexes. The assumption made was that overexpression of Png1p would prevent the formation of NEF2 (Rad4p-Rad23p complex), thereby causing UV sensitivity of yeast cells even in the presence of Rad23p. For this experiment, we used png1⌬ rad23⌬ cells. Png1p was expressed overnight using an inducible GAL1 promoter, and then FLAG-Rad23p was transiently expressed for 2 h using the CUP1 promoter. First, a qualitative test was performed to check the UV sensitivity of these cells. As shown in Fig. 7A, under the experimental conditions described above, cells overexpressing Png1p exhibited moderate UV sensitivity (compare sample 2 with sample 1). This level of sensitivity was not seen when the same experiment was carried out without induction of Png1p (compare sample 2C and sample 1C), indicating that this effect was Png1p-dependent. That a comparable amount of Rad23p was expressed in samples 1 and 2 was confirmed by Western blotting, excluding the possibility that the observed UV sensitivity was due to the amount of Rad23p expressed in cells (data not shown). In contrast to the above experiment, when we expressed FLAG-Rad23p overnight, we observed no apparent UV sensitivity in PNG1-overexpressing cells, suggesting that the observed UV sensitivity is a transient effect and cells can make enough Rad4p-Rad23p complex at stationary stage even in the presence of large amount of Png1p (data not shown). With this assay, png1⌬ cells exhibited no apparent UV sensitivity, implying that PNG1 itself does not have a role in nucleotide excision repair (data not shown).
Using a more quantitative UV sensitivity assay, we confirmed that overexpression of Png1p under the experimental conditions described above caused moderate sensitivity (10-fold decrease in survival) (Fig. 7B). This relatively moderate effect may be caused by a residual amount of NEF2 (Rad4p-Rad23p) formation. As expected when Png1p alone (without Rad23p) was induced, the sensitivity was very great (1,000-fold decrease in survival).

DISCUSSION
The de-N-glycosylation process catalyzed by PNGase has been proposed to be involved in proteasomal degradation of misfolded glycoproteins following their transfer from the ER to the cytosol (9 -13). Earlier we reported that S. cerevisiae has a soluble PNGase activity that is very similar to the soluble PNGase found in higher eukaryotes (36). Subsequently, we isolated the gene encoding this enzyme, PNG1, by isolating a mutant that is defective in PNGase activity and then mapping of the locus responsible for the loss of this activity (8). Comparison of the protein sequence of yeast Png1p with a number of sequences in other eukaryotic data bases revealed that this enzyme is highly conserved, suggesting that it was functionally important in all eukaryotes.
Following the finding that the apparent molecular mass of PNGase detected in the cytosol by an enzyme assay was much greater than its calculated mass of 42.5 kDa, we carried out a two-hybrid analysis to identify yeast proteins that interact with Png1p. By this screening, two distinct candidates were isolated, both of which encoded a part of Rad23p. In a rad23⌬ strain, we observed a dramatic decrease in elution position of the yeast Png1p to a position similar to that of bacterially expressed Png1p. These results supported the idea that the formation of a higher molecular weight complex of Png1p involved Rad23p. This was further confirmed by showing that epitope-tagged Rad23p co-migrated with Png1p during gel filtration analysis of the yeast cytosol.
In this experiment, using Png1p-overexpressed cells, the finding of two distinct peaks of PNGase activity may be due to the fact that a fraction of the large amount of Png1p produced under these conditions was bound to Rad23p. The calculated apparent molecular mass of this complex is ϳ200 kDa, which is approximately twice as large as the expected molecular mass of a 1:1 stoichiometric Rad23p-Png1p complex. This large molecular mass could be due to the presence of more than one molecule of each of the subunits or a change in conformation of the proteins in a complex. Alternatively, other proteins may also be present in the complex; currently, we are investigating this possibility.
Interaction studies of Png1p using deletion constructs of Rad23p provided two important conclusions. First, the N-terminal UbL, which is important for Rad23p to interact with the proteasome (24), was found not to be required for Rad23p binding to Png1p. This finding led us to hypothesize that a physical interaction of Png1p with the 26 S proteasome might be mediated by Rad23p. This hypothesis was confirmed by a co-immunoprecipitation analysis that utilized an antibody to an epitope-tagged subunit of the 26 S proteasome (Rpt1p) and His 6 -tagged Png1p. This interaction between Png1p with the 26 S proteasome was Rad23p-dependent, which is consistent with the idea of Rad23p being an "escort protein" to connect Png1p and the 26 S proteasome. It has been reported that there is a so-called "membrane-associated form" of PNGase, which can be precipitated by ultracentrifugation and has enzymatic Ϫ Ϫ a No color change on ␤-galactosidase assay; number of plus signs indicates the intensity of color on ␤-galactosidase assay. b This plasmid was originally isolated from the pGAD library.
The second important observation regarding domain interaction of Rad23p was that the C-terminal ubiquitin-associated domain of Rad23p was critical for its binding to Png1p. This result is reminiscent of the previously reported domain interaction of Rad4p-Rad23p (NEF2) (44). Our results, together with the previous observation, suggest that Png1p and Rad4p might both independently bind to the same domain of Rad23p. In fact, the formation of a Png1p-Rad23p complex that is distinct from NEF2 (Fig. 8) was confirmed by co-immunoprecipitation analysis and further supported by genetic analysis. We have also confirmed that the mouse homologue of Png1p can bind to mouse Rad23p homologues. 2 In this connection, it should be noted that Rad4p is the only structurally related protein among Saccharomyces cerevisiae proteins that was found to contain a domain similar to one in Png1p (based on BLAST analysis; E value ϭ 0.023; Png1p amino acids 220 -321 versus Rad4p amino acids 316 -434; identities, 30/119 (25%); positives, 53/119 (44%); gaps, 17/119 (14%)). This fact may mean that this region has in common the same folding struc-ture, which could be important for binding to the C terminus region of Rad23p.
The biological significance of the interaction of the 26 S proteasome with Rad23p still remains to be understood. Now it is evident that there are at least two distinct complexes involving the proteasome and Rad23p (Fig. 8). This finding may to some extent explain the complex role of Rad23p-proteasome interactions. We hope that in the future it will be possible to dissect the specific functions of proteasome or Rad23p in each distinct complex. In this regard, it is of interest that a number of domain-specific Rad23p-binding proteins, other than the nucleotide repair proteins, have been identified in mammalian cells (46 -48). These studies, combined with our results, lead us to speculate that such domain-specific interactions may allow Rad23p to connect a proteasome protein through its N terminus to other proteins at its C terminus. This link might be necessary to modulate functions of various binding proteins. Most intriguing is the finding that a human homologue of Dsk2p, a Rad23p-like protein, has recently been shown to provide a link between the proteasome and ubiquitin ligase (49). Thus, a general feature of UbL-containing proteins would be to link proteins to the proteasome in order to control or modulate the function of these proteins.
While the precise biological function of Png1p is still unclear, it is important to note that this enzyme has been proposed to be involved in proteasome degradation of misfolded glycoproteins in mammalian cells (9,10). We have previously shown that glycopeptides exported from the ER to the cytosol in yeast are subsequently deglycosylated by the activity of Png1p (36). However, because of the inefficient uptake of proteasome inhibitors (50), it has not been possible to observe an accumulation of de-N-glycosylated protein intermediates in yeast. Interestingly, biochemically purified PNGase from mammalian cells, as well as the purified bacterially expressed yeast Png1p, do not act on intact glycoprotein substrates in vitro (8,35). As discussed earlier (8), these observations may be relevant to the fact that the 20 S catalytic proteasome subcomplex can only act efficiently on small peptide substrates in vitro, and it has been proposed that the function of the 19 S ATPases is to unwind protein substrates prior to their degradation by the 20 S proteolytic complex (51,52). Such "protein-unfolding" molecular chaperone activities of 19 S subunits have been recently reported (53,54). Therefore, the association of Png1p with the 26 2 H. Park, T. Suzuki, and W. J. Lennarz, unpublished results. FIG. 7. Sequential overexpression of Png1p followed by Rad23p causes moderate UV sensitivity. A, cells used were as follows. 1, TSY190 (W303-1a rad23⌬::URA3 png1⌬::his5 ϩ (pombe) FOA R ) bearing pCS13 (P CUP1 ::FLAG-RAD23) and pRD53 (no Png1p control); 2, TSY190 bearing pCS13 and pRD53-PNG1 (P GAL1 ::PNG1); 3, TSY190 bearing YEp351 (no Rad23p control) and pRD53-PNG1. The absence of cell growth at the right end of the streak is indicative of UV sensitivity of these cells. For control, the same experiment was carried out using glucose medium so that expression of Png1p could not occur (samples 1C, 2C, and 3C). B, quantitative analysis of UV sensitivity was carried out. Yeast cells were exposed to 254-nm UV light, and the survival was determined. Sample 2 (E; ϩRad23p, ϩPng1p) displayed intermediate sensitivity between sample 1 (q; ϩRad23p, ϪPng1p) and sample 3 (ϫ; ϪRad23p, ϩPng1p). The Rad4p-Rad23p (left) is a well characterized complex required for DNA repair (17)(18)(19) that is known to associate with the proteasome (24). This interaction has been shown to be important for the DNA repair functions (24,25) in a proteolytic activity-independent manner (26). A postulated second complex contains Png1p and Rad23p (right). This complex is also associated with the 26 S proteasome and may have a function as a specialized protein degradation complex used for degradation of misfolded glycoproteins (see "Discussion").
S proteasome would produce a complex in which de-N-glycosylation and proteolysis of unwound glycoprotein substrates could be accomplished in an efficient manner. Our findings indicate that the highly conserved enzyme, Png1p, is a part of the proteasomal degradation machinery. Further studies should reveal its precise role in the degradation of misfolded glycoproteins by this pathway.