A Proteasome Cap Subunit Required for Spindle Pole Body Duplication in Yeast

Standard yeast genetic methods were used (66) and yeast media were made as described (37). For temperature shift experiments requiring cell synchronization, cells were arrested with 10 μM α-factor (Sigma Chemical Co., St. Louis, MO) and released as described (101). Yeast strains are listed in Table I. Bacterial strain DH5α was used for plasmid construction and bacterial media were made as described (84).

PCS1 was fortuitously identified during studies of CDC31 (4). A plasmid (pHM27) encoding a lexA DNA-binding domain–Cdc31 protein fusion was used as “bait” in a two-hybrid screen (12). The L40 reporter strain (98) was cotransformed with pHM27 and the pGAD libraries (12). Library plasmids were recovered from transformants giving positive signals in the screen. A fragment of PCS1 was found immediately downstream from the GAL4 activation domain sequence in one plasmid, YL2-29, but subcloning showed that a gene elsewhere in the genomic DNA insert (CDC31) was responsible for the signal. Independent library plasmids containing CDC31 were identified, mapped, and shown to contain PCS1 in the same relative position. One plasmid, YL1-19, contained the complete PCS1 sequence.

The PCS1 gene was sequenced on both strands using YL2-29, YL1-19, and constructs derived from these plasmids as templates. Sequencing reactions were done using the Taq DyeDeoxy terminator cycle sequencing kit (Applied Biosystems, Foster City, CA) and were analyzed by the University of Washington Molecular Pharmacology Facility using an Applied Biosystems 373A automated sequencer. Oligonucleotides designed to prime the sequencing reactions at appropriate positions in PCS1 were synthesized by GIBCO BRL (Gaithersburg, MD). Sequence assembly and analysis were performed using the Genetics Computer Group package (36), and comparisons with sequence databases were performed using the BLAST program (2).

As a first step toward disrupting PCS1, the yeast integrating plasmids pHM35 and pHM47 were constructed. Standard recombinant DNA techniques were used (84), and restriction enzymes, DNA polymerases, and T4 DNA ligase were purchased from New England Biolabs (Beverly, MA). To introduce a unique EcoRI site in the region of the gene encoding the ATPase domain, oligonucleotide primers hm15-BamHI (5′GAAGATCTCGGATCCACTGGTG3′) and hm16-EcoRI (5′GGAATTCTATAACCTCCCTTAA3′) (underlined nucleotides represent restriction sites) were used together with Vent™ DNA polymerase (New England Biolabs) to amplify YL2-29 template DNA between nucleotides 583 and 689 (see Fig. 2 B) in a PCR. The product was digested with BamHI and EcoRI and ligated into BamHI/EcoRI-digested pRS306 (88) to create plasmid pHM34. Primers hm17-EcoRI (5′CGAATTCCCATTGAAGAACCCAG3′) and hm18-KpnI (5′CCGGTACCAGGGGGAC3′) were used to amplify the YL2-29 template between nucleotides 685 and 775. This product was digested with EcoRI and KpnI and ligated into EcoRI/KpnI- digested pHM34 to create pHM35. The insert in pHM35 was excised with BamHI and KpnI and ligated into BamHI/KpnI-digested pRS304 (88) to create pHM47.

To disrupt one genomic copy of PCS1, plasmid pHM47, which carries the TRP1 gene, was digested with EcoRI and used to transform the diploid yeast strain D8bx5cA (which is congenic with S288C) to Trp⁺. Southern blot analysis (not shown) confirmed that integration of pHM47 into one copy of PCS1 had occurred as expected in the transformant used in further experiments (YHM9.1).

To create a temperature-sensitive PCS1 mutant, we constructed plasmid pHM55, which encodes a degron–Pcs1p fusion. The EcoRI–HindIII fragment encoding the degron unit (Ub-DHFR^ts-ha) (21) was previously cloned into the EcoRI–HindIII sites in the pRS316 polylinker by Dr. Frank Russo (University of Washington, Seattle, WA). This fragment was excised with EcoRI and ClaI and treated with Klenow. pHM53 (see below) was digested with NcoI and treated with Klenow, and the degron- encoding fragment was ligated into this site. These manipulations resulted in the degron fragment being inserted at the start codon of PCS1, in frame with the open reading frame. We refer to this allele as pcs1^td (temperature-inducible degron allele). Next, the yeast integrating plasmid pHM56 was constructed. The insert from pHM55 was excised with PvuII and inserted into PvuII-digested pRS306 (88) to create pHM56.

To replace the wild-type PCS1 gene with the pcs1^td allele in the chromosome, the two-step “pop-in/pop-out” gene replacement method was used (81, 85). pHM56 was digested with EcoRI and used to transform the haploid strain Wx257-5c (which is congenic with S288C) to Ura⁺. PCR analysis confirmed that integration had occurred at the PCS1 locus as predicted (data not shown), resulting in both PCS1 and pcs1^td flanking the plasmid sequences. Excision of the plasmid and either of the PCS1 alleles via homologous recombination was then selected for using 5-fluoroorotic acid (5-FOA; PCR Inc., Gainesville, FL), which kills Ura3⁺ cells (5). Two 5-FOA–resistant colonies were identified that had a temperature-sensitive phenotype, suggesting that they contained the pcs1^td allele. PCR analysis confirmed that this was true in the pcs1^td strain used in further experiments, YHM11.2 (data not shown).

To delete UBR1 from both YHM11.2 and Wx257-5c, plasmid pJDubrΔ4-B (a gift from Drs. Jürgen Dohmen and Alexander Varshavsky, California Institute of Technology, Pasadena) was digested with EcoRI and used to transform these yeast strains to His⁺. This manipulation resulted in the replacement of UBR1 with the ubr1Δ::HIS3 allele, creating strains YHM12.1 (pcs1^td) and YHM13.1 (PCS1).

Plasmids pHM48 and pHM49 were constructed to supply PCS1 function in a haploid yeast strain containing a disrupted genomic copy of PCS1. Primers hm27-XhoI (5′CCGCTCGAGAAGCTTGAAATACCATAGTG3′) and hm28-SmaI (5′TCCCCCGGGAGTTCAATAGCCATTTCGAGC3′) were used together with Vent^TM DNA polymerase to amplify the YL1-19 template between nucleotides 1 and 1,506 (Fig. 2 B). The product was digested with XhoI and SmaI and ligated into the yeast centromere (CEN) plasmids pRS315 and pRS316 (88), also digested with XhoI and SmaI, to create plasmids pHM48 and pHM49.

To facilitate later constructions, a unique NcoI site was engineered to overlap the PCS1 start codon in pHM48. Primer hm32-NcoI (5′CCTGTTCTTCACCCATGGTTAGTTATTAATATCTTTG3′) was used in a PCR-based site-directed mutagenesis technique (10). Primers hm32-NcoI and T3 (whose site is outside the polylinker upstream from the PCS1 gene in pHM48) were used to amplify a fragment of DNA from pHM48. This fragment was then used as a primer, together with the T7 primer (whose site is outside the polylinker downstream from the PCS1 gene in pHM48) in a second PCR. The second PCR product was digested with XhoI and SstI and ligated into XhoI/SstI-digested pRS315 (88) to create plasmid pHM53. As a result of the site-directed mutagenesis, the second codon in PCS1 was changed from AGT (Ser) to GGT (Gly).

Plasmid pHM54 was constructed to encode a green fluorescent protein (GFP)–Pcs1 protein fusion. Plasmid pZA66 was previously constructed by Dr. Michael Moser (University of Washington). It encodes S65T GFP (38) engineered to contain a COOH-terminal 10–amino acid spacer consisting of repeating Gly-Ala residues. The DNA fragment encoding this unit was excised from pZA66 with BspHI and ligated into NcoI-digested pHM53 to create pHM54. This manipulation resulted in the S65T GFP encoding DNA being inserted in frame with the PCS1 gene, at the PCS1 start codon.

Cells were prepared for immunofluorescence microscopy as described (49, 51). To detect microtubules, the rat monoclonal anti–α-tubulin antibody YOL1/34 (Accurate Chemical and Scientific Corp., Westbury, NY) and FITC-conjugated goat anti–rat antibodies (Amersham Corp., Arlington Heights, IL) were used at dilutions of 1:100 and 1:50, respectively. For hemagglutinin (HA) epitope tag localization, mAb 12CA5 (Boehringer Mannheim Biochemicals, Indianapolis, IN) was used at a concentration of 0.8 μg/ml followed by FITC-conjugated goat anti–mouse antibodies (Amersham Corp.) diluted 1:100. DNA was visualized with 4′,6-diamidino-2phenylindole (DAPI) (1 μg/ml; Sigma Chemical Co.). Fixed and stained cells were viewed using a Microphot FX fluorescent microscope (Nikon Inc., Garden City, NY) and photographed with TMAX 400 film (Eastman Kodak Co., Rochester, NY).

Cells containing the GFP–Pcs1p fusion were grown to a density of ∼10⁷ cells per ml and stained with DAPI by first resuspending the cells briefly in methanol, washing with PBS, and resuspending in a 1 μg/ml DAPI solution. These cells were viewed and photographed as described above.

Yeast cells were prepared for thin-section EM as described (9). Serial sections were viewed on an EM300 electron microscope (Philips Electronic Instruments, Inc., Mahwah, NJ).

Cells were prepared for flow cytometry as described (47). Propidium iodide (Sigma Chemical Co.) was used to stain the DNA. A Becton Dickinson FACS^®can flow cytometer (San Jose, CA) was used together with the CELLFIT and LYSYS software packages.

For Western blot analysis, whole cell lysates were prepared from yeast spheroplasts as follows. Spheroplasts derived from 5-ml cultures at a density of 3 × 10⁷ cells per ml were resuspended in 100 μl lysis buffer (50 mM Tris-Cl, pH 7.5, 1 mM EDTA, 1 mM DTT, 1 mM PMSF, 0.5 μg/ml leupeptin, 2 μg/ml aprotinin) and lysed using 20 strokes in a Dounce homogenizer. This solution was centrifuged for 15 min at top speed in a microfuge at 4°C, and the supernatant was transferred to a fresh tube. Total protein concentration in the clarified lysates was determined using the Bio-Rad protein assay (Bio Rad Laboratories, Hercules, CA) and BSA as a standard. Equivalent amounts of total protein were subjected to SDS-PAGE (58). Proteins were transferred to nitrocellulose in a transphor electrophoresis unit (Hoefer Scientific Instruments, San Francisco, CA) for 1 h at 0.3 A. The nitrocellulose membrane was blocked with 5% nonfat dry milk in TBS/Tw (TBS, 0.05% Tween-20) for 1 h, and then incubated with antiGFP antiserum (Clontech, Palo Alto, CA) at a 1:4,000 dilution in TBS/Tw for 1 h. The membrane was washed in TBS/Tw and incubated with goat anti–rabbit IgG-HRP (Accurate Chemical and Scientific Corp.) at a 1:5,000 dilution for 1 h, followed by washing in TBS/Tw. All incubations were done at room temperature. The signal was developed using the enhanced chemiluminescence system (Amersham Corp.).

We identified PCS1 while sequencing near CDC31 (4), a gene located on the right arm of S. cerevisiae chromosome XV (see Materials and Methods for experimental details and rationale). Briefly, one end of a genomic fragment of DNA containing CDC31 internally was sequenced and found to encode a portion of a novel gene encoding a new member of a diverse family of ATPases (86). Since members of this family are known to play roles in the regulation of mitosis, we pursued its characterization further. We identified a plasmid (YL1-19) containing an 11-kb insert that includes the full-length version of the new gene. It is located ∼1 kb from CDC31 and faces in the opposite orientation. A partial restriction map of the relevant portion of the YL1-19 insert is shown in Fig. 2,A; the relative positions of PCS1, CDC31, and a third open reading frame are indicated by arrows beneath the map. Sequence analysis revealed that a 1.3-kb open reading frame (Fig. 2 B), designated PCS1, represents the novel ATPase-encoding gene.

To examine the relationship between Pcs1p and other protein sequences in the databases, we used the BLAST alignment program (2). This analysis confirmed the identity of Pcs1p as a member of the “AAA” (ATPases Associated with diverse cellular Activities [56]) superfamily. Members of this superfamily share a conserved region of ∼200 amino acids that contains motifs A and B characteristic of ATP binding, underlined in Fig. 2,B (99). Pcs1p is most similar to a subclass of this superfamily composed of proteasomal 19S cap subunits. Fig. 3 shows an alignment of Pcs1p with three of the highest scoring proteins in the BLAST search. p42 (25) and S4 (23) are subunits of the human 26S proteasome, and their sequences are respectively 67% and 38% identical to the Pcs1p sequence as determined by the pairwise alignment program BESTFIT (36). SUG1 (92)/CIM3 (31) (and CIM5 [31]) are S. cerevisiae genes; the sequences of their predicted products are 43% and 39% identical to the Pcs1p sequence. Homologs of the SUG1/CIM3 and CIM5 gene products were identified in purified preparations of 26S proteasomes from Drosophila (31), and the Sug1/Cim3 protein has been directly shown to be a subunit of the yeast 26S proteasome (82). Pcs1p is also closely related to the S. pombe mts2⁺ gene product, a 26S proteasome component necessary for completion of mitosis (35). The very high degree of sequence similarity between these known proteasome cap subunits and Pcs1p provides strong evidence that Pcs1p is itself a 19S cap subunit. This identity has been confirmed by the recent work of Russell et al. (83), who demonstrated that the protein encoded by SUG2 (which is identical to PCS1) is a component of the 26S proteasome, while our work was in progress.

PCS1 is predicted to encode a protein 437 amino acids in length, with an expected molecular mass of 49 kD and isoelectric point of 5.5. A notable characteristic of Pcs1p is the very hydrophilic region extending between amino acids 13 and 96, shown graphically in Fig. 2,C. This region also corresponds to the most divergent part of Pcs1p compared with the sequences most closely related to it (Fig. 3). Additionally, the boxed region indicated in Fig. 2,B is predicted to be a potential nuclear localization signal by the program PSORT (69). To gain further information about the conformation of Pcs1p, we used the method of Robson-Garnier (29) to predict the secondary structural features of the Pcs1p sequence. This analysis, also depicted graphically in Fig. 2,C, suggests that the protein may be largely globular, since the three conformational features of α helix, turn, and β sheet generally alternate. One region largely predicted to be α-helical, between amino acid residues 74 and 101, displays a limited heptad repeat motif suggestive of the ability to form a short α-helical coiledcoil. This motif consists of a repeating heptad with hydrophobic residues enriched at positions 1 and 4 (1) (residues at position 1 are indicated with a triangle beneath them in Fig. 2 B). The program COILS (62) gives this region a probability near 1.0 of adopting a coiled-coil conformation. Other members of the 19S ATPase family also contain heptad motifs near their NH₂ termini (78).

To determine whether PCS1 gene function is essential for cell viability, we constructed haploid strains disrupted for PCS1 (pcs1::TRP1). We reasoned that the highly conserved ATP-binding motif A (P-loop; underlined in Fig. 2,B) would almost certainly be essential for Pcs1p function. Therefore, we constructed a plasmid, pHM47, that when integrated into the genome results in the insertion of the S. cerevisiae TRP1 gene into the P-loop–encoding region (see Materials and Methods); the insertion site is indicated by the arrow in Fig. 2 B. The diploid strain D8bx5cA was transformed with this plasmid, which was cut at EcoRI to direct integration at the PCS1 locus. Southern blotting was used to confirm that the plasmid had integrated into one copy of PCS1 as expected in each of eight Trp⁺ transformants (data not shown). The transformant chosen for further experiments (designated YHM9.1) was sporulated, and the resulting tetrads were dissected. In all cases (20/20), the four spores segregated in a ratio of two viable to two inviable; this result held true at both 23° and 37°C. The viable spores were always auxotrophic for tryptophan, indicating that they did not carry the disruption marker. We examined the phenotype of the inviable spore microcolonies after 3 d at either 23° or 37°C and observed a small effect of temperature on the cell cycle at which division arrested. At 23°C, the spore germinated and divided twice, arresting as four large-budded cells. At 37°C, the spore germinated and divided once, arresting in the second division cycle as two large-budded cells. Often the bud was misshapen and could not be pulled away from the mother cell with a dissecting needle. These results indicate that PCS1 function is necessary for cell viability and are consistent with an essential role for this protein in mitosis.

To determine whether the sequenced length of DNA shown in Fig. 2 B is sufficient to supply PCS1 function, we constructed a plasmid (pHM49; see Materials and Methods) that contains the PCS1 sequence carried on a CENbased vector marked with URA3. pHM49 was transformed into strain YHM9.1, which is heterozygous for the pcs1:: TRP1 disruption. Sporulation and tetrad analysis revealed that more than two spores per tetrad formed colonies. All Trp⁺ colonies were also Ura⁺, indicating that a haploid strain disrupted for chromosomal PCS1 can be maintained by the sequence cloned in pHM49.

To obtain a conditional allele that could be used for more detailed phenotypic analysis, we constructed a haploid strain, designated YHM11.2, in which the chromosomal PCS1 gene was replaced with an allele (pcs1^td) that encodes a degron–Pcs1p fusion (see Materials and Methods). The degron unit is a heat-inducible degradation signal (21) that at high temperature (37°C) is degraded, together with sequences fused downstream of it, by the N-end rule pathway of protein degradation (3). Strain YHM11.2 was viable at 23° and 30°C (Fig. 4,A, patch b), but was inviable at 36° and 37°C. Strain Wx257-5c, the wild-type strain from which YHM11.2 is derived, was viable at all temperatures tested (Fig. 4 A, patch a). Thus, pcs1^td is a temperaturesensitive allele of PCS1. We also confirmed that the pcs1^td mutation is recessive, as expected for an allele designed to encode a protein that is degraded rapidly: pcs1^td/PCS1 diploid cells grew normally at 37°C (data not shown), and the mutant phenotype caused by pcs1^td can be complemented by pHM49 (see above).

To test whether this temperature sensitivity is a result of degradation via the N-end pathway as predicted, we deleted UBR1 from strains YHM11.2 and Wx257-5c to create strains YHM12.1 and YHM13.1, respectively. The UBR1 gene is required for the N-end rule pathway and for the temperature sensitivity of several alleles encoding degron fusions (21). We found that deleting UBR1 partially rescued the temperature sensitivity caused by pcs1^td. Although deletion of UBR1 does not restore growth of the pcs1^td strain at 37°C (Fig. 4 A, patch d), growth did occur at 36°C. These results confirm an important role for the N-end rule pathway in the loss of function for pcs1^td.

To determine whether depletion of PCS1 function causes arrest at a specific stage in the cell cycle, we shifted asynchronous liquid cultures of strains Wx257-5c (PCS1) and YHM11.2 (pcs1^td) from 23° to 37°C and observed their behavior after various times. Fig. 4,B shows the relative proportions of unbudded, small-budded, and large-budded cells after 9 h at the restrictive temperature. About 90% of the pcs1^td cells were found to be large budded at the restrictive temperature, compared with ∼30% for the culture continuing to grow at the permissive temperature. To investigate the kinetics of entry into this state of arrest, we used synchronous cultures of PCS1 and pcs1^td obtained by arresting growth in G1 with α-factor and then shifting them to 37°C in fresh media without α-factor. Fig. 4, C and D, shows the proportions of cell types at time points after this shift. From this experiment it is clear that pcs1^td cells underwent at least two cell divisions before becoming arrested as large-budded cells, since a second cycle of bud formation was evident at ∼5 h. We also noted a partial defect in cell separation, for pairs of budded cells remained attached to each other as the second cell cycle progressed. By 9 h after release from α-factor, ∼25% of the cells were detected as attached pairs of large-budded cells.

We also followed individual cells after shift to the restrictive temperature. Cells were arrested with α-factor, transferred to a plate, and observed after 27 h at 37°C. About half of the cells had arrested in the second cycle as two large-budded cells, while the other half had completed this cycle and undergone the same type of arrest in the third cycle. It has been observed that the effectiveness of the degron system improves if cells are held in stationary phase before the temperature shift (Dirick, L., personal communication). Accordingly, we permitted pcs1^td cells to remain at stationary phase for 3 d at 23°C, at which time ∼90% were unbudded, transferred them to a plate, and again observed them after incubation for 27 h at 37°C. 50% of the cells reached the final stage of arrest as large-budded cells within the first cycle, 30% in the second cycle, and 20% in the third, suggesting that degron–Pcs1p was more effectively depleted in this experiment.

Flow cytometry experiments indicate that pcs1^td cells arrest with a 2C (G2) content of DNA. Fig. 5,A shows the cytometry profiles of asynchronous PCS1 and pcs1^td cultures at 23°C, where it can be seen that about equal proportions of pcs1^td cells had apparent 1C and 2C DNA contents. After 9 h at 37°C, however, 87% of the pcs1^td cells had accumulated in a state with a 2C content of DNA, whereas a higher proportion of PCS1 cells had a 1C content of DNA at 37°C than at 23°C (Fig. 5 B). These findings demonstrate that cells lacking PCS1 function undergo arrest in the division cycle after DNA synthesis, and that this function therefore must be required for progression from G2 into mitosis.

To gain a better understanding of how a lack of PCS1 function perturbs mitosis, we visualized DNA and microtubules using fluorescence microscopy in pcs1^td cells that had accumulated in the terminal arrest described by flow cytometry (see above). After 9 h at 37°C, the DNA remained unsegregated and was localized in a region immediately adjacent to the bud neck in ∼80% of these cells (Fig. 6, B–E). Nuclei occupying this position in PCS1 cells grown at 37°C were shown by tubulin antibody staining to contain a short intranuclear spindle (Fig. 6,A), as is characteristic of yeast cells in G2 or mitotic metaphase. However, the arrested pcs1^td cells did not contain discrete spindles and generally had fewer detectable intranuclear microtubules. The most common phenotype is shown in Fig. 6,B, in which a single bundle of cytoplasmic microtubules extended away from the nucleus and through the bud neck. Additionally, some cells contained a relatively bright focus of microtubule staining at the nuclear periphery, as well as cytoplasmic microtubules emanating away from this focus (Fig. 6,C). Less frequently, we observed cells in which microtubules appeared to lie along the surface of the nucleus (Fig. 6,D), and other cells in which two bundles of cytoplasmic microtubules extended away from the nucleus in an antennae-like array (Fig. 6 E). (Significant numbers of cells lacking spindle structures were not observed at earlier time points in this experiment, presumably because the first cycle proceeded normally, albeit slowly.) These findings suggest that cells deprived of PCS1 function fail to form a mitotic spindle while retaining the ability to assemble cytoplasmic microtubule arrays.

The attachment between pairs of large-budded cells seen after transfer of a pcs1^td culture to the restrictive temperature (as described earlier) provided us with a method of identifying mother and daughter cells unambiguously. Fig. 7,A shows cells in this configuration. Since the cells had been arrested with α-factor before the shift to high temperature, the original mother is identifiable as the schmooshaped cell 1. This cell presumably had budded once to produce cell 2 and a second time to produce bud 3. Cell 2 had also budded to produce bud 4. In Fig. 7,B, it can be seen that the unsegregated DNA lies largely within the buds (3 and 4) that are the products of the second division. Cytoplasmic microtubule staining is evident in bud 4 (Fig. 7 C); faint microtubule staining was also observed in bud 3 in a different focal plane. Although we could not distinguish the mother cell and bud unambiguously among all cells examined in this manner, the DNA localized to the mother ∼50% of the time and bud 50% of the time when both could be identified (data not shown). A mutant that is specifically defective in spindle pole body duplication, ndc1-1, has previously been shown to display random localization of the single functional spindle pole and the bulk of the DNA between the mother cell and the bud (93, 97). Similarly, certain mutations that cause cells to arrest with a bipolar spindle also display such random localization of DNA (72).

Given our finding that the pcs1^td cells arrest more quickly if held in stationary phase before a shift to the restrictive temperature, we also examined the microtubule arrays of cells treated in this manner at various times after the temperature shift. Since mutations in the closely related SUG1/CIM3 and CIM5 genes have previously been reported to cause cells to arrest with short spindles (31), we also wished to determine whether the phenotypes of pcs1^td, cim3-1, and cim5-1 mutants were indeed distinct when the same conditions were used for the experiment. Accordingly, strains YHM11.2 (pcs1^td), CMY763 (cim3-1), and CMY765 (cim5-1) (each congenic with the S288C background) were held at stationary phase for 3 d at 23°C, and then diluted into fresh media at 37°C. At the time of the temperature shift, no abnormal spindles were observed in any of the strains. Fig. 8 shows the percentage of largebudded cells containing short spindles, long spindles, or no spindles in each of the strains at different times after the temperature shift. While most large-budded pcs1^td cells were observed to contain no spindles at 4 and 6 h after the shift, most large-budded cim3-1 and cim5-1 cells contained short spindles at these time points. Incidentally, we also observed cytoplasmic microtubule arrays in the cim3-1 cells containing short spindles that were broader and more elongated than those seen in wild-type or cim5-1 cells; this characteristic of cim3-1 is also apparent in the figure from the original paper describing these mutants (31). More strikingly, these data demonstrate that loss of PCS1 function causes an arrest that differs distinctly from that caused by loss of CIM3 or CIM5 function.

To learn more about how loss of PCS1 function perturbs spindle formation, we used serial section EM to examine cells from strain YHM11.2 (pcs1^td) held at 37°C for 9 h. We found that the majority of cells (21/25 examined) had failed to duplicate their SPBs and therefore possessed only a monopolar spindle. The remaining cells (4/25) had duplicated their SPBs and each contained a bipolar spindle. Fig. 9,A shows one pole of a bipolar spindle from one of these latter cells. The SPB seen here is indistinguishable from that of wild type: its half-bridge (arrow), the darkly staining segment of the nuclear envelope immediately adjacent to the SPB, has the same appearance as that which is normally seen in wild-type cells (7). Fig. 9, B, B′ and C, C′, show (in adjacent sections) the single SPBs from two cells containing monopolar spindles. These SPBs are larger than normal, as has been observed for other mutants that fail to duplicate their SPBs (e.g., 8, 101). Normal half-bridge structures were not seen in any of the cells (n = 21) containing unduplicated SPBs; instead, most cells examined contained an unusual type of structure in one of the adjacent sections in the series. This structure, shown in Fig. 9, B′ and C′, appears to represent an aberrant half-bridge that is unusually curved in profile and is wrapped partly around the SPB.

To gain more insight into Pcs1p function, we determined the localization of a GFP–Pcs1p fusion. Plasmid pHM54 was constructed to contain the sequence encoding this fusion in a CEN-based vector marked with LEU2 (see Materials and Methods). This plasmid was transformed into haploid yeast strain YHM10.1.49, which contains a disrupted chromosomal PCS1 gene and is rendered viable by the presence of PCS1 on a URA3-marked CEN plasmid. Leu⁺ transformants were transferred to growth on 5-FOA, which kills cells that are Ura3⁺ (5), effectively selecting for loss of the URA3-marked plasmid. The generation of 5-FOA–resistant colonies indicated that cells can use the GFP–Pcs1p fusion as their only functional source of Pcs1p. One of these isolates was retained for further analysis and designated YHM10.1.54. In this strain, GFP–Pcs1p is probably not being expressed to a significantly higher level than Pcs1p in a wild-type strain, both because its expression is controlled by the PCS1 promoter and because it is borne on a CEN vector, which should be present at only one or two copies per cell (95).

The GFP–Pcs1p fusion is predicted to have a molecular mass of 76 kD. To confirm the presence of this product in strain YHM10.1.54, equivalent amounts of whole-cell extracts from strains YHM10.1.54 and Wx257-5c (wild-type control) were analyzed by Western blot (Fig. 10,A). AntiGFP antiserum recognized a band of approximately the correct size in YHM10.1.54 extract (Fig. 10 A, lane 2) but not in the wild-type control (lane 1).

An asynchronous YHM10.1.54 culture was stained with DAPI to visualize DNA and examined by fluorescence microscopy. The GFP signal colocalized with the nuclear DNA, indicating that the GFP–Pcs1p fusion resides principally within the nucleus (Fig. 10, B and C). (A similar GFP signal was observed in cells that were not stained with DAPI, confirming that this signal was not due to bleedthrough of the DAPI signal.) This colocalization was consistently seen in all stages of the cell cycle, including the unbudded (1), small-budded (2), and large-budded (3 and 4) cells indicated in the figure. The GFP signal in cell 3 is observed to colocalize with a thin finger of DNA that extends through the bud neck, as well as with the bulk of the DNA. In ∼5% of the cells examined, an additional patch of GFP signal was observed at a position separate from the site of DNA staining, as indicated (*) in Fig. 10 B. The cytological basis for this occasional patch of apparent cytoplasmic localization is unknown, and it remains possible that it represents an extension of the nucleus. Nevertheless, it is clear that the great majority of GFP–Pcs1p fusion colocalizes with the nuclear DNA and must therefore reside within the nucleus.

To gain a better understanding of how Pcs1p distribution compares with that of other proteasome subunits, we examined the localization of a different 19S subunit. The Sen3 protein from S. cerevisiae is a component of the proteasome cap, unrelated in sequence to the ATPase subunits (16). To study the localization of this protein, we used MHY851 cells, which express a functional Sen3 protein tagged with three copies of the HA epitope (16). We used anti-HA antibodies to visualize Sen3 protein distribution in these cells and found that it also colocalized with the nuclear DNA throughout the cell cycle (Fig. 11, A and B). No staining was observed in cells expressing untagged Sen3 protein (strain MHY849; data not shown). Similarly, the Sug1/Cim3 and Cim5 proteins have been localized primarily to the nucleus using affinity-purified antibodies, although more significant cytoplasmic staining was observed for Cim5 than for Cim3 (Mann, C., personal communication). Thus, a number of different proteasome cap subunits appear to localize largely within the nucleus.

We have described the identification and characterization of PCS1, an essential yeast gene whose predicted product shows a high level of sequence identity with a family of ATPases that are components of the 19S proteasome cap. The 67% identity between Pcs1p and p42, a known human 26S proteasome regulatory subunit (25), provides strong evidence that Pcs1p is itself a proteasome subunit. This idea has recently been confirmed biochemically (83). The p42 subunit has been identified not only as a component of the proteasome cap (also known as PA700) but also of the proteasome “modulator,” a second complex that enhances cap activation of proteolytic activity (18). Whether a modulator of this sort also exists in S. cerevisiae is unknown.

It has been suggested that different ATPases within the 19S cap might serve to recognize distinct degradation substrates; a structural model for how this recognition specificity might be mediated has been proposed by Rechsteiner et al. (78). These authors note that members of this family are highly conserved in their COOH-terminal regions and more divergent in their NH₂-terminal segments (see Fig. 3), suggesting that the latter portion of the molecule might be involved in specific substrate selection. As is also the case for Pcs1p, these NH₂-terminal domains contain stretches of sequence that are predicted to form leucine zippers or α-helical coiled-coils. Rechsteiner et al. propose a “helix shuffle” hypothesis, in which these regions of the ATPase subunits interact with similar regions within target proteins. ATP hydrolysis would then serve to reverse this binding and to permit transfer of the substrates to the 20S proteasome core. In partial confirmation of this model, it has been shown that the leucine zipper in the murine homolog of Sug1/Cim3 protein interacts with the leucine zipper in c-Fos; both were detected in a 26S proteasome preparation (100), suggesting that murine Sug1/Cim3 may be involved in the degradation of c-Fos.

Genetic analysis supports the concept that different cap subunits might serve to recognize different substrates that have been targeted for degradation. At the restrictive temperature, pcs1^td strains are arrested in the cell division cycle as large-budded cells with replicated, unsegregated DNA that is localized to the bud neck. Surprisingly, these cells do not form bipolar spindles, although cytoplasmic microtubule staining is clearly seen. The same phenotype has been seen for a temperature-sensitive allele of pcs1 generated by hydroxylamine mutagenesis (McDonald, H.B., and B. Byers, unpublished observations). Using EM, we determined that pcs1^td cells fail to duplicate their SPBs. This phenotype differs strikingly both from any stage seen in normal cell cycle progression and from the arrests caused by cim3 and cim5. The SUG1/CIM3 and CIM5 gene products from S. cerevisiae are 19S cap subunits and are closely related to Pcs1p (Fig. 3). Conditional cim3 and cim5 mutations cause a cell to arrest in G2/M with a large bud and a short spindle (31), a state that is quite similar to that seen at the G2/M transition in a normal ongoing cell cycle. This phenotype suggests that the wild-type products of these genes play an active role in the degradation of APC-targeted substrates. Compatible with this view, the levels of B-type cyclins are increased in cim3 and cim5 mutants relative to wild type, even if the mutants are maintained in exponential growth at the permissive temperature (31). Since cyclin destruction is required for exit from mitosis, but not for initiation of anaphase (45, 91), the Sug1/Cim3 and Cim5 gene products are presumed to act in the degradation of other substrate(s) besides the cyclins. The distinctive pcs1^td mutant phenotype suggests, on the other hand, that Pcs1p is required for the destruction of yet another class of substrate(s).

There is precedent for mutations in different proteasomal cap subunits differing in their effects on cell cycle regulation. The NIN1 gene of S. cerevisiae encodes a regulatory subunit of the 26S proteasome, but this protein is structurally unrelated to the family of ATPases that includes the SUG1/CIM3 and CIM5 products (55). Although the majority of nin1 cells arrest at the G2/M stage of the cell cycle, the nature of this arrest is distinct from that caused by cim3 or cim5. Arrested cim3 and cim5 mutants show activation of Cdc28p kinase, the S. cerevisiae CDK (31), consistent with the increased levels of mitotic cyclins observed in these mutants. In contrast, Cdc28p kinase is not activated in nin1 mutant cells after a shift to the restrictive temperature. The basis for this difference is unknown, but an intriguing suggestion offered is that NIN1 function is specifically required for the degradation of a CDK inhibitor (55).

The effectiveness of the pcs1^td temperature-sensitive allele supports the idea that PCS1 function is dispensable for the degradation one particular substrate, the degron– Pcs1p fusion. Initially we were hesitant to attempt the construction of a temperature-sensitive PCS1 allele using the degron-based method, because degron fusions are targeted for proteolysis at high temperature by the N-end rule/ubiquitin/proteasome system. We therefore feared it might not be possible to deplete Pcs1p levels sufficiently by this technique, as Pcs1p might be required to play a role in its own destruction. However, we have identified conditions in which fully half the population of pcs1^td cells will arrest during the first division cycle after shift to high temperature, suggesting that the degron–Pcs1p fusion probably is not a component of the proteolytic machinery that is needed for its destruction.

If different cap subunits are indeed responsible for the degradation of distinct substrates, several possible models could account for this phenomenon. For example, there might exist within the cell multiple subpopulations of 26S proteasomes, each containing different cap subunits responsible for the degradation of a specific set of substrates. Distinct subpopulations of proteasomes were suggested by recent experiments in Manduca sexta, where levels of 19S ATPases were found to vary differentially during development (15). An alternative model would be that all 26S proteasomes in yeast are identical, with different subunits within the proteasome cap being responsible for the recognition and import of different classes of substrates. Regardless of which model proves true, the largely nuclear localization of this and certain other 19S cap subunits supports the idea that a substantial fraction of fully assembled 26S proteasomes reside within the yeast nucleus. These findings suggest, in turn, that the majority of substrates targeted for degradation may also be localized to the yeast nucleus. More specifically, we propose that Pcs1p is required for the degradation of yet to be identified nuclear substrate(s), and that this activity is required for SPB duplication to occur.

Strikingly, this work has added a new member to the small collection of yeast genes for which conditional mutations specifically perturb the process of SPB duplication while permitting other early cell cycle functions to proceed unabated. Several distinct patterns of failure in SPB duplication are seen among these mutants. In the pattern displayed by mps2 (101) and ndc1 (102) mutants, a nascent SPB is assembled on the half-bridge, but it fails to be inserted into the nuclear envelope. In contrast, mutations in CDC31 (8), KAR1 (80), and MPS1 (101) prevent formation of the satellite, which serves as the precursor for the nascent SPB. The half-bridge in mps1 cells is dramatically enlarged relative to the more normal-appearing half-bridge found in cdc31-1 or kar1 cells at temperature-sensitive arrest. pcs1^td cells likewise lack a normal half-bridge structure, but in this case the aberrant half-bridge is distinctively curvilinear in profile. Thus, mutations in these genes appear to alter the nature of half-bridge formation in different ways, each of which leads to failure of satellite formation and SPB duplication. In the case of pcs1^td, we hypothesize that some nuclear factor(s) must be degraded for half-bridge formation to proceed correctly. Further experiments would be needed to identify the relevant substrate(s) and to test other possible roles for this cap component.

We thank Drs. J. Dohmen, A. Varshavsky, M. Moser, R. Tsien, R. Heim, and F. Russo for gifts of plasmids, and Drs. M. Hochstrasser and C. Mann for yeast strains. We also thank Dr. C. Mann for communicating results prior to publication, L. Goetsch for her generous help with EM, Drs. L. Dirick and R. Coyne for useful discussions, and two anonymous reviewers for their useful comments.

This work was supported by a National Institutes of Health (NIH) postdoctoral fellowship (GM16027) to H.B. McDonald and an NIH research grant (GM18541) to B. Byers.