Essential functions of Sds22p in chromosome stability and nuclear localization of PP1

Type 1 protein phosphatase (PP1) is a highly conserved member of the PPP family of protein serine-threonine phosphatases that plays important roles in a wide range of cellular functions. In the budding yeast Saccharomyces cerevisiae the essential gene GLC7 encodes the catalytic subunit of PP1. A variety of studies using glc7 mutants have revealed that yeast PP1 is required for many key cellular functions including glucose repression (Tu and Carlson, 1994; Tu and Carlson, 1995), kinetochore function (Bloecher and Tatchell, 1999; Sassoon et al., 1999), vesicle fusion during protein secretion (Peters et al., 1999), glycogen synthesis (Feng et al., 1991), meiosis (Bailis and Roeder, 2000; Tu et al., 1996) and cell wall integrity (Andrews and Stark, 2000). Rather than acting on its own, the catalytic subunit of PP1 (PP1_C) has been found to interact with proteins that regulate specific PP1 dephosphorylation events (Hubbard and Cohen, 1993; Bollen, 2001). For example, mammalian PP1 has been shown to bind tissue-specific glycogen-binding subunits such as G_L, and G_M, which target the enzyme to glycogen particles where it can dephosphorylate key enzymes of glycogen metabolism such as glycogen synthase (Doherty et al., 1995; Hubbard and Cohen, 1989). In S. cerevisiae, a related protein (Gac1p) performs a similar function and is required for PP1 to activate glycogen synthase by dephosphorylation (François et al., 1992; Stuart et al., 1994), while Reg1p is involved in the roles of PP1 in glucose repression and glucose signalling (Frederick and Tatchell, 1996; Jiang et al., 2000; Tu and Carlson, 1995). In budding yeast, over a dozen proteins have been demonstrated or proposed to interact with PP1_C (Stark, 1996; Uetz et al., 2000), while other PP1_C-interacting proteins known in higher eukaryotes include regulatory subunits that target PP1 to myosin (Alessi et al., 1992) or that form complexes with PP1_C in the nucleus (Ajuh et al., 2000; Allen et al., 1998; Kreivi et al., 1997; Van Eynde et al., 1995). A conserved motif of the form -V/IXF- has been identified that mediates the interaction between PP1_C and many of its regulatory subunits (Egloff et al., 1997; Zhao and Lee, 1997).

SDS22 is the budding yeast homologue of Schizosaccharomyces pombe sds22⁺ (Hisamoto et al., 1995; MacKelvie et al., 1995). S. pombe Sds22 is a nuclear protein that binds directly to the catalytic subunit of PP1 and which largely consists of 11 tandem leucine-rich repeats (Stone et al., 1993), although it lacks the -V/IXF- motif found in many PP1_C-interacting proteins. The corresponding gene (sds22⁺) was isolated as a high-copy suppressor of a conditional mutation (dis2-11) affecting the major fission yeast PP1 catalytic subunit Dis2 (Ohkura and Yanagida, 1991). A temperature-sensitive sds22 allele led to a metaphase-like arrest of fission yeast cells at the restrictive temperature, with high histone H1 kinase activity, a short spindle and condensed chromosomes (Stone et al., 1993). Like fission yeast Sds22, S. cerevisiae Sds22p interacts with PP1_C (Glc7p) as established by multiple criteria (Hisamoto et al., 1995; Hong et al., 2000; MacKelvie et al., 1995). Extra copies of SDS22 suppress the temperature-sensitivity of glc7-12, a GLC7 allele that confers a mitotic arrest phenotype (MacKelvie et al., 1995). Thus, work in both S. cerevisiae and S. pombe supports a model whereby Sds22p activates mitosis-specific functions of PP1_C, and evidence from the latter yeast suggests that, like other PP1 regulatory subunits, Sds22p might act by modifying the substrate specificity of PP1_C (Stone et al., 1993). A human homologue of Sds22p has also been identified (Renouf et al., 1995), but like several of the other nuclear proteins that bind PP1_C, human Sds22 appears to act as an inhibitor of PP1_C using the specific substrates tested (Dinischiotu et al., 1997).

In this study, we have generated conditional sds22 alleles in order to develop a better understanding of how Sds22p regulates Glc7p function. Surprisingly, we found no evidence of a mitotic arrest phenotype when strains carrying two distinct, temperature-sensitive sds22 alleles were shifted to the restrictive temperature. However, the mutant strains showed clear evidence of chromosome instability, suggestive of a chromosome segregation defect. Strikingly, the sds22 mutations led to a rapid loss of nuclear Glc7p localization under restrictive conditions, showing that Sds22p plays a role in maintaining the normal nuclear localization of Glc7p.

Basic yeast methods and growth media were as described (Kaiser et al., 1994) and yeast transformation was carried out as described (Gietz et al., 1992). Routine recombinant DNA methodology was performed as described elsewhere (Sambrook et al., 1989). All restriction enzymes were from New England Biolabs (Beverly, MA). Oligonucleotide primers used are summarized in Table 1. When testing for suppression of Ts^– mutations by genes present on high-copy plasmids, cultures of strains grown under selection for the plasmid were diluted to 5×10⁵, 5×10⁴ and 5×103 cells/ml in YPD medium and samples (approximately 5 μl) spotted on YPD agar using a multipronged inoculating manifold (Dan-Kar Corp.). All yeast strains and plasmids used in this study are summarized in Table 2 and Table 3, respectively. To generate SAY100, AY927 was transformed with an sds22::TRP1 construct and a derivative in which one genomic copy of SDS22 had been replaced by the knockout was generated and verified as previously described (MacKelvie et al., 1995). After transformation of this strain with pLMY-SDS22, SAY100 was generated following sporulation and tetrad dissection. Chromosome loss was measured using a method developed previously (Spencer et al., 1990) in which loss of a chromosome III fragment carrying the SUP11 gene was monitored by the red pigment formed upon loss of suppression of the ade2-101 mutation. Chromosome loss rate per cell division was calculated as described previously (Yoon and Carbon, 1999). FACS analysis was carried out as reported elsewhere (Andrews and Stark, 2000).

SDS22 was first inserted as a 1.35 kb BscI-XhoI fragment from pLMY-SDS22 between the EcoRI and PstI sites of YCplac111 after both vector and insert had been treated with the Klenow fragment of T4 DNA polymerase to generate blunt ends. The resulting plasmid (YCpLEU-SDS22) was digested with PstI and NdeI to remove codons 33-273 of SDS22. A PCR fragment encompassing SDS22 was amplified with primers SDS22Δ KS^– with pBS351 as template, using mutagenic conditions as described previously (Stark, 1998) but with a Mn²⁺ concentration of 0.3 mM. The former primer corresponds to a region ∼300 bp downstream of SDS22, while the latter primer corresponds to vector sequence upstream of SDS22 in the template construct. The PCR fragment was digested with EcoRI and BclI, the 1.01 kb central portion was gel purified and then co-transformed with gapped YCpLEU-SDS22 into SAY102. After eviction of pLMY-SDS22 by growth on 5-fluoroorotic acid (5-FOA), 600 transformants were screened for temperature-sensitive growth at 37°C and the sds22-5 and sds22-6 alleles thus isolated. The sites of mutational alteration in these alleles were identified by DNA sequence analysis. Strains in which the mutant alleles were integrated at either the LEU2 or the HIS3 loci were generated by transformation of SAY100 with derivatives of YIplac128 (Gietz and Sugino, 1988) or pRS303 (Sikorski and Hieter, 1989), respectively, carrying either wild-type SDS22, sds22-5 or sds22-6 as a 2.2 kb SpeI-XhoI fragment inserted into the XbaI-SalI interval of the vector. Integration was verified by Southern blot analysis and then strains dependent solely on the integrated copy were obtained following 5-FOA selection (Sikorski and Boeke, 1991).

A triple myc epitope tag was inserted immediately after the last sense codon of SDS22 as follows. First, equimolar amounts of primers 180 and 181, which share 25 bp of complementarity at their 3′ ends, were annealed together by heating at 100°C for 5 minutes followed by slow cooling to room temperature. The 3′ ends were extended at 72°C by addition of dNTPs and Expand™ High-Fidelity Polymerase (Roche), forming a DNA fragment encoding a triple myc epitope (myc₃). After addition of primers 183 and 184, the ∼130 bp myc₃ fragment was amplified by 30 cycles of 1 minute at 95°C, 1 minute at 50°C and 2 minutes at 72°C such that it was flanked by sequences just preceding and just following the SDS22 stop codon. In a second PCR reaction, the SDS22 open reading frame was similarly amplified using primers SDS22-3 and 182. The two products were recovered using the High Pure PR product purification kit (Roche) and then fused by mixing equimolar amounts, annealing as described above and amplifying the fusion product using primers Sds22-3 and 184. The product was recovered as above and cleaved with NdeI and BclI to generate a 306-bp fragment corresponding to the C-terminal, myc-tagged region of SDS22. This fragment was inserted into psds22-6 cleaved with the same enzymes, generating a myc₃-tagged sds22-6 allele. The tagged gene was verified by DNA sequencing and moved as an SpeI-HindIII fragment into YCplac111 (cut with the XbaI-HindIII) to generate YCp3Msds22-6. A wild-type tagged SDS22 construct was then generated by replacing the ∼800 bp PstI-NdeI interval of YCp3Msds22-6 with the corresponding wild-type sequence, generating YCp3MSDS22. Strains dependent on the tagged wild-type or mutant alleles were generated by transformation of SAY100 followed by selection for loss of pLMY-SDS22 using 5-fluoroorotic acid.

A derivative of strain AYS927 in which the genomic copy of SDS22 was tagged at the 3′ end with protein A was generated as described (Rayner and Munro, 1998), using primers SDS22-PrA-5′ and SDS22-PrA-3′ with pZZ-His5 as template. Following verification by PCR and detection of the tagged protein by western blot analysis, haploid strains in which protein A-tagged Sds22p (Sds22p-PrA) was the sole source of Sds22p function were generated by tetrad analysis.

To generate an integrative plasmid encoding protein A-tagged Glc7p, the HA-tagged GLC7 construct from YCpHA-GLC7 was excised as a HindIII-MscI fragment and cloned between the HindIII and SmaI sites of YIplac204. A small SpeI-ClaI fragment at the 5′ end of GLC7 (carrying the HA epitope tag) was removed and replaced with a larger SpeI-ClaI fragment carrying a protein A tag and TEV protease cleavage site, generating YIplac204-PrA-GLC7. This fragment was made by PCR using primers PrA-GLC7-5′ and PrA-GLC7-3′ with pZZ-His5 as template, and was cleaved by SpeI and ClaI prior to subcloning. The sequence of this fragment was verified by DNA sequencing of the relevant region of the final construct. To generate a strain (LKY150) solely dependent on PrA-Glc7p for Glc7p function, YIplac204-PrA-GLC7was integrated into SBY-SSa at the trp1-1 locus followed by selection on 5-FOA to remove YCp-GLC7(URA3). LKY118 was similarly generated usingYIplac204-HA-GLC7. A control strain expressing unfused protein A (from the NOP1 promoter) was generated by transformation of AY925 with pNOPPATA-1L.

To identify proteins complexed with Sds22p, cultures (1.5 l) of strain SAY1228 were grown in YPD medium at 26°C until they reached ∼10⁷ cells/ml. The cells were harvested by centrifugation and then washed twice with water and finally with 10 ml ice-cold extraction buffer (50 mM Hepes-KOH (pH 7.5), 150 mM KCl, 0.1% Triton X-100, 0.1 mM EDTA, 10% (v/v) glycerol). The cell pellet was resuspended in an equal volume of extraction buffer containing 1× complete protease inhibitors (Roche) and supplemented with an equal volume of acid-washed glass beads (0.4 mm diameter). After 20 cycles of vortexing (30 seconds) and cooling on ice (30 seconds) the cell debris was removed by centrifugation for 10 minutes at 3500 rpm in a Jouan CR/312 centrifuge. The cell pellet was subject to a further 10 cycles of disruption after addition of the same volume of extraction buffer with protease inhibitors and the two resulting supernatants were pooled and spun at 20,000 g for 20 minutes. Protein A-tagged Sds22p (Sds22-PrA) was recovered from the extract by elution over a column containing 0.75 ml IgG-Sepharose (Amersham Pharmacia Biotech) which had been pre-washed with 20 ml extraction buffer. After washing the column with 10 bed volumes of extraction buffer, the column material was resuspended in 3 ml of extraction buffer without Triton X-100 and glycerol but containing 30 units TEV protease (Life Technologies) and incubated for 1 hour at room temperature. The released material was recovered by elution with extraction buffer (lacking Triton X-100 and glycerol) and examined by SDS-PAGE, with silver staining to visualize the protein bands. For identification of protein bands by mass spectrometry, appropriate fractions of eluate were desalted using CentriPlus (Millipore), concentrated using Microcone protein concentrators (Millipore), alkylated with 4-vinylpyridine and separated by SDS-PAGE using 4-12% Bis-Tris gels (Novex). Gels were stained using Coomassie Brilliant Blue and mass fingerprint analysis carried out as described (Morrice and Powis, 1998). Proteins interacting with PrA-Glc7p were similarly identified using strain LKY150. In each case, strain AY925 transformed with pNOPPATA-1L (expressing the unfused protein A tag) (Lau et al., 2000) was used to control for nonspecific protein binding.

The behaviour of Sds22p-PrA on gel filtration of yeast cell extract (prepared as above from strain SAY1228) was examined using a Superose-200 column calibrated with the following markers (Sigma): apoferritin, α-amylase, alcohol dehydrogenase, bovine serum albumin (BSA). SAY1228 also expressed HA-Glc7p as the sole source of PP1_C. Sds22p-PrA was located by western blot analysis and its apparent Stokes radius calculated using a plot of (–log K_AV)^1/2 against published values for the Stokes radii of the marker proteins. The sedimentation of Sds22p-PrA in a 10-40% (v/v) glycerol gradient centrifuged at 60,000 rpm in a Beckman SW60 rotor for 8 hours at 4°C was examined using a similar sample of cell extract from LKY168, with thyroglobulin, catalase, aldolase and BSA as markers (Amersham Pharmacia Biotech). The molecular size of Sds22p-PrA was calculated from its Stokes radius and sedimentation coefficient as previously described (Siegel and Monty, 1966).

Cultures of yeast cells (200 ml) were grown at 26°C to a density of 1×10⁷ cells/ml in YPD, harvested by centrifugation and washed in an equal volume of water. Cells were collected in a 15 ml centrifuge tube and the pellet supplemented with an equal volume of lysis buffer containing 50 mM Tris-HCL (pH 7.5), 100 mM NaCl, 10 mM MgCl₂, 5 mM EDTA, 1 mM DTT, 1% (v/v) Triton X-100, 1× complete protease inhibitors (Roche). Acid-washed glass beads (0.4 mm diameter; 0.7 g per ml) were added and the cells lysed by 20 cycles of vortexing for 30 seconds followed by 30 seconds on ice. Extracts were centrifuged for 5 minutes at 21,000 g to pellet cell debris and the supernatant removed. The glass beads were washed with one pellet volume of lysis buffer and the supernatants pooled. Protein concentrations were determined using the Bio-Rad protein assay. To make Sds22p-myc₃ immunoprecipitates, Protein G-Sepharose beads (Amersham Pharmacia Biotech) were first equilibrated in lysis buffer and 80 μl of a 50% suspension used to pre-clear lysates (containing 2.5 mg protein) at 4°C for 2 hours on a rotary mixer. Purified 9E10 monoclonal antibody (1 μg) was added to the pre-cleared lysate and incubated at 4°C for 2 hours with gentle mixing. The antibody-lysate mix was divided into two equal parts and 20 μl 50% Protein G-Sepharose bead suspension added to each. One was incubated at 4°C for 2 hours while the other was incubated at 30°C, both with gentle mixing as above. The beads were recovered by centrifugation, washed three times in lysis buffer and then resuspended in 30 μl 2× SDS-PAGE sample buffer and boiled for 2 minutes. Recovered proteins were separated by SDS-PAGE on 10% polyacrylamide gels and Sds22p-myc₃ and HA-tagged Glc7p visualized by western blotting with ECL detection using either 9E10 with sheep anti-mouse IgG-HRP conjugate (Amersham Pharmacia Biotech) or mouse anti-HA HRP conjugate (Roche), respectively. Similar procedures were used to compare Sds22p-myc₃ and HA-tagged Glc7p levels in total cell extracts using anti-calmodulin antibodies (Stirling et al., 1994) to confirm equivalent protein loading.

Cells were prepared for immunofluorescence microscopy as described (Ayscough and Drubin, 1998) using the general immunofluorescence protocol, but omitting the methanol and acetone fixing step. High-affinity 3F10 (anti-HA) and 9E10 (anti-myc) antibodies were used at dilutions of 1:100 while both sheep anti-mouse IgG-Cy3 conjugate (Sigma) and goat anti-rat IgG-FITC conjugate (Cappel) secondary antibodies were used at 1:200 dilutions. To visualize DNA, the mounting medium contained 1 mg/ml 4′, 6-diamidino-2-phenylindole dihydrochloride (DAPI). Images were acquired using a Deltavision Restoration microscope (Applied Precision Inc., USA) fitted with a Nikon PlanApo 100× (1.4NA) objective and a Photometrics series 350 cooled CCD camera, taking a Z series encompassing the whole cell. Images were deconvolved and processed using the Deltavision Softworx application on a Silicon Graphics Octane Workstation (Silicon Graphics Inc., USA) and a single optical section presented.

To monitor the localization of Glc7p in wild-type and sds22 mutant strains, a functional GFP-Glc7p construct (Bloecher and Tatchell, 2000) was integrated into the ura3 locus of strains KT2070 (SDS22/SDS22), KT2067 (SDS22/sds22-6) and KT2066 (sds22-6/sds22-6). The sds22-6 and sds22::TRP1 alleles were introduced into KT2066 and KT2067 after backcrossing SAY304 seven times to KT1357. For imaging, cells were grown to mid logarithmic phase in synthetic complete medium, collected by centrifugation and then placed on a microscope slide over a thin agarose slab and under a cover slip, as described elsewhere (Waddle et al., 1996). Cells were observed through a 100× N.A. 1.25 Olympus objective equipped with a Bioptechs objective heater (Butler, PA) and images were collected with a 12-bit Princeton Instruments Micro Max CCD camera, capturing images at 30 minute intervals using 2 second exposures at 6% full intensity. All images for a given strain were normalized in the same way but the normalization was different for each strain. To calculate the ratio of nuclear/cytoplasmic fluorescence at each time point, a minimum of 102 cells in which the focal plane went through the nucleus were quantitated using the IPLab Spectrum software. Individual cells were imaged only once to avoid photobleaching.

We modified SDS22 by addition of a triple myc epitope inserted immediately preceding the stop codon. Triple myc-tagged Sds22p (Sds22p-myc₃) was fully functional as judged from its ability to support normal growth when present as the sole source of Sds22p. Using such a strain, Sds22p-myc₃ was found to be mainly localized to the nucleus (Fig. 1). Although Sds22p largely co-localized with nuclear Glc7p, in some cells Sds22p appeared to be located within a slightly more restricted region of the nucleus than Glc7p, and most cells contained at least some regions where Sds22p did not overlap with Glc7p. We also tagged Sds22p with green fluorescent protein (GFP) and obtained similar results, although the signal obtained was rather weak (not shown). Thus like its fission yeast homologue, Sds22p is largely confined to the nucleus. In view of its potential role in targeting Glc7p to other cellular components, we adopted an affinity isolation approach to identify other polypeptides with which Sds22p could interact. Sds22p was tagged with Protein A at its C-terminus (Sds22p-PrA) and a strain in which Sds22p-PrA was the sole source of Sds22p was generated. Sds22p-PrA was affinity isolated on IgG Sepharose and proteins that co-purified were recovered following cleavage of the fusion protein by TEV protease at a site immediately preceding the Protein A tag. However, comparison with a control strain expressing unfused Protein A showed that only two specific bands were present in the Sds22p-Protein A TEV eluate, and mass fingerprint analysis revealed that these were Sds22p and Glc7p (Fig. 2A). By comparison, when a similar approach was used to examine Glc7p-associated proteins, several polypeptides could be specifically isolated in complexes with Protein A-tagged Glc7p, although Sds22p represented one of the more abundant Glc7p-associated proteins (Fig. 2B). Catalytic activity of the Glc7p polypeptide is not required for its interaction with Sds22p since Sds22p could still form a complex with Glc7p containing an inactivating mutation (H124A) at the catalytic site (not shown).

To determine what proportion of Sds22p was complexed with Glc7p, HA-tagged Glc7p (HA-Glc7p) was immunodepleted from extracts and the behaviour of Sds22p-PrA examined. Fig. 2C shows that near complete immunodepletion of HA-Glc7p concurrently removed most of the Sds22p-PrA from the extract, demonstrating that the majority of Sds22p-PrA present must be complexed with Glc7p. When Sds22p-PrA was examined by gel filtration, it eluted as a single peak with a Stokes radius of ∼45 Å, which overlapped with fractions containing HA-Glc7p, although HA-Glc7p was also present in the void volume and in fractions that eluted later than Sds22p-PrA, demonstrating that not all HA-Glc7p was bound to Sds22p-PrA (not shown). When Sds22p-PrA was analysed by glycerol density gradient ultracentrifugation, it sedimented as a single peak of ∼5S which overlapped with the sedimentation profile of HA-Glc7p (not shown). The method of Siegel and Monty (Siegel and Monty, 1966) yielded an estimated M_r of ∼93,000 based on these data. This estimate is consistent both with the expected size of a 1:1 Sds22p-PrA–HA-Glc7p complex (M_r 91,759) and with the failure to identify any other Sds22p-interacting proteins at stoichiometric levels. Thus, like Hong et al. (Hong et al., 2000), we conclude that Sds22p forms a stable complex with Glc7p and that this complex is unlikely to contain any other components. This differs from results obtained in HeLa cell extracts, where human Sds22p was found in a much larger complex of approximately 260 kDa (Dinischiotu et al., 1997).

To gain greater insight into the role of Sds22p, we next generated temperature-sensitive sds22 alleles by random PCR-mediated mutagenesis. Of several such alleles isolated we chose to characterize two in particular. sds22-5 was found to carry four point mutations leading to four amino acid replacements in the protein (T25A, Q28R, L73P, F285S), while sds22-6 encodes two missense mutations (I54T, I193T) together with a silent mutation in codon 51. The L73P and F285S mutations in sds22-5 and the I193T mutation in sds22-6 represent nonconservative changes to key hydrophobic residues located in leucine-rich repeats (LRRs) 1 and 10 (sds22-5) and in LRR 6 (sds22-6). Both alleles were found to be recessive (not shown). Strains dependent on either plasmid-borne or integrated copies of sds22-5 were unable to grow at 32°C or higher, while the equivalent sds22-6 strains grew up to 35°C. In the presence of nocodazole, strains containing either the sds22-5 or sds22-6 allele arrested with a large-budded morphology at the same concentration required to provoke an equivalent arrest in the SDS22 control strain, and this arrest was maintained for 3 hours when the nocodazole-arrested cultures were shifted to 37°C (not shown). Thus, continued Sds22p function is not required for maintenance of a mitotic checkpoint arrest. Although in W303 strains high-copy GLC7 is unable to compensate for complete loss of SDS22 function (not shown), the restrictive temperature of each sds22 mutant was raised significantly by high-copy GLC7 (Fig. 3; and data not shown), suggesting that elevated levels of Glc7p can in part compensate for loss of Sds22p function. High-copy SDS22 was similarly able to raise the restrictive temperature of the glc7 alleles glc7-5 and glc7-12 (Fig. 3). Taken together, these genetic data support a model whereby Sds22p and Glc7p function in an interdependent manner.

Since Glc7p is required for kinetochore function and chromosome stability (Sassoon et al., 1999), we examined chromosome loss rates in the sds22 mutant strains using a standard colony-sectoring assay. Both sds22-5 and sds22-6 conferred a marked chromosome loss phenotype at higher growth temperatures, with the sds22-6 allele showing the greatest defect in chromosome stability (Table 4): at 32°C, which is more than 3°C below the temperature at which sds22-6 strains fail to grow, the chromosome loss frequency per cell division was elevated around 30-fold compared to wild-type strains. Thus like the glc7-10 mutant (Sassoon et al., 1999), the Ts^–sds22 mutants show a profound chromosome loss phenotype at higher growth temperatures. However, unlike glc7-10 and glc7-129 PP1_C mutations (Bloecher and Tatchell, 1999; Sassoon et al., 1999), neither sds22-5 nor sds22-6 conferred a distinct mitotic arrest phenotype at 37°C: after shifting asynchronous cultures of the mutants to the nonpermissive conditions, little change in the proportion of budded cells could be detected but proliferation ceased within 1-2 generations (not shown). Fig. 4 shows that when cells synchronised in G1 with mating pheromone were released at the restrictive temperature, both sds22 mutant and wild-type cells initiated budding and DNA replication, although both events were slightly delayed in the mutants. Again, however, no homogeneous arrest phenotype could be observed. In contrast to the glc7-10 and glc7-129 PP1_C mutations, the defect in sds22 mutants that confers chromosome instability does not apparently activate the mitotic checkpoint. This is in contrast to S. pombe cells carrying a temperature-sensitive sds22 allele that arrested in mid-mitosis (Stone et al., 1993).

One way in which sds22 mutations might exert their effect would be destabilization of the mutant protein. We therefore generated a strain in which a triple myc-tagged copy of sds22-6 was the sole source of Sds22p and examined Sds22-myc₃ protein levels by western blot analysis following transfer of the strain to 37°C. This showed that the steady-state protein level remained essentially unchanged for at least 6 hours at the restrictive conditions (Fig. 5). Similarly, we found that levels of Glc7p were essentially unchanged in the sds22-6 mutant strain at 37°C over a similar time course (Fig. 5). Thus loss of Sds22p function conferred by the sds22-6 allele cannot be explained by instability of either Sds22p or Glc7p at the restrictive temperature. We therefore next tested by co-immunoprecipitation whether the formation of the Sds22p-Glc7p complex was affected by sds22 mutations. Extracts were made from wild-type or sds22-6 mutant strains grown under permissive conditions at 26°C and Sds22p immunoprecipitations carried out either at 4°C or 30°C. Although the recovery of Sds22p-myc₃ by immunoprecipitation was lower from the sds22-6 mutant despite similar overall levels of the protein in the extracts, HA-Glc7p co-precipitated with both wild-type and mutant Sds22p-myc₃ at 4°C (Fig. 6, lanes 3,5). By comparison, when immunoprecipitation was carried out at 30°C, no detectable HA-Glc7p was present in the precipitates (Fig. 6, compare lanes 3 and 5). We also examined whether the glc7-12 mutation affected formation of the Sds22p-Glc7p complex, but since the level of the HA-tagged mutant Glc7p in the extracts was lower it is difficult to assess whether its reduced recovery in Sds22p immunoprecipitates indicates reduced interaction with Sds22p (Fig. 6, lanes 3,4). However, the Sds22p-Glc7p interaction was not rendered noticeably temperature-sensitive in vitro by the glc7-12 mutation (Fig. 6, lane 4), in contrast to the effect of sds22-6. We conclude therefore that the sds22-6 mutations affect the interaction of the mutant protein with Glc7p at higher temperatures, consistent with the in vivo temperature-sensitivity of the sds22-6 mutant strain. The amount of Sds22p-Glc7p complex recovered at 4°C when both proteins were mutated was lower than when extracts contained just one mutant polypeptide (Fig. 6, lane 6). However, since the glc7-12 mutation reduced the overall level of HA-Glc7p and the immune precipitates from the sds22-6 strain contained lower levels of Sds22p, the two mutations probably do not confer an additive defect in Sds22p-Glc7p complex formation. This is consistent with our finding that there is no additional growth defect seen when sds22-6 and glc7-12 are combined (not shown).

Recent work has revealed that both histone H3 and the kinetochore protein Ndc10p are likely to be substrates of nuclear PP1. The Ipl1p protein kinase can phosphorylate each of these proteins and both genetic and biochemical approaches have shown that PP1 opposes the action of the Ipl1p kinase in regulating their phosphorylation state (Biggins et al., 1999; Hsu et al., 2000; Sassoon et al., 1999). For example, certain glc7 loss-of-function mutants show allele-specific hyperphosphorylation of histone H3 on serine-10, while the ipl1-2 mutant shows greatly reduced histone H3 serine-10 phosphorylation in vivo. Ipl1p kinase can also phosphorylate histone H3 at this position in vitro. In keeping with the model that Ipl1p kinase and PP1 oppose each other in regulating the phosphorylation state of proteins, glc7-127 suppressed the temperature-sensitivity of the ipl1-2 mutation in an allele-specific manner (Hsu et al., 2000). Similarly, glc7-10, a PP1 loss of function mutant which is defective in Ndc10p dephosphorylation (Sassoon et al., 1999), and ipl1-321, a mutant which lacks a protein kinase activity required for inactivation of kinetochore function in an in vitro assay (Biggins et al., 1999), show mutual suppression (N. Rachidi and M.J.R.S., unpublished). We therefore examined the effect of combining sds22-6 with ipl1-2. If sds22-6 leads to reduced PP1 activity in vivo then a clear prediction would be that it should relieve the Ts^– defect associated with ipl1-2. Fig. 7 shows that while the ipl1-2 strain was unable to grow at 34°C, when combined with sds22-6 it grew just as well as the sds22-6 IPL1 control strain and almost as well as a wild-type strain. Thus, like glc7-127 (Hsu et al., 2000), sds22-6 suppresses the temperature-sensitivity of the ipl2-1 mutant. This result is consistent with the notion that the sds22-6 mutation causes reduced nuclear PP1 function, thereby compensating for the reduced kinase activity in the ipl1-2 mutant. However, the double ipl1-2 sds22-6 strain remained unable to grow at 37°C, perhaps either because suppression requires some residual Ipl1p function or because there is incomplete overlap between the cellular targets of the two proteins.

The effect on Glc7p localization of shifting sds22-6 mutant cells to 37°C was next examined using a GFP-Glc7p fusion. In control strains, the nuclear localization of GFP-Glc7p was essentially unaffected by shifting cells to 37°C. However, in the sds22-6 mutant the relatively uniform nuclear GFP-Glc7p fluorescence in the majority of cells relocalized to a small number of bright foci after as little as 30 minutes (Fig. 8), some but not all of which appeared to be located within the nucleus. To quantitate nuclear levels of GFP-Glc7p, the fluorescence intensity of nuclear and cytoplasmic GFP-Glc7p was measured in over 100 cells at each time point and the nuclear/cytoplasmic ratio was estimated. Table 5 shows that in the sds22-6 mutant the relative nuclear fluorescence was significantly decreased at 37°C compared with the wild-type. Given that the overall level of Glc7p remained essentially constant after shift to 37°C (Fig. 5B), these data suggest that the sds22-6 mutation affects the localization of Glc7p under restrictive conditions: overall levels within the nucleus are reduced and, rather than being present throughout the nucleus, it is relocalized into a small number of foci.

In this work we have used a variety of biochemical and genetic approaches to address the cellular function of S. cerevisiae Sds22p, a putative regulator of protein phosphatase 1 (PP1). One aim was to identify novel Sds22p-interacting proteins as a means of understanding Sds22p function. However, Glc7p was the only protein that could be specifically affinity isolated using protein A-tagged Sds22p. By contrast, Sds22p and a number of as yet uncharacterized proteins could be specifically isolated in complexes with protein A-tagged Glc7p. Furthermore, the majority of Sds22p appears to be bound to Glc7p (Fig. 2C), consistent with their similar patterns of localization within the nucleus. While we cannot rule out the existence of transient interactions between Sds22p and other proteins that would not have been detected by this approach, for example any potential substrates of an Sds22p-Glc7p complex, the evidence strongly points to the absence of any other proteins stably and stoichiometrically associated with the Sds22p-Glc7p complex. This is fully consistent with the apparent molecular size of Sds22p in cell extracts determined from two different biophysical parameters and which strongly suggests a 1:1 stoichiometry. In vitro, the mutant sds22-6 product is profoundly defective in its ability to bind Glc7p at high temperatures. This suggests that the ability to form the Sds22p-Glc7p complex is critical and is consistent with the reciprocal high-copy suppression data shown in Fig. 3. PP1_C interacts with many of its regulatory subunits via a -V/IXF- PP1_C binding motif in the latter, which interacts with a cleft on the opposite face of the phosphatase to the catalytic site (Egloff et al., 1997). Since Sds22p lacks such a sequence, its mode of interaction with PP1_C must be distinct, consistent with the finding that Glc7p carrying mutations in the -V/IXF- binding cleft interacts normally with Sds22p (Wu and Tatchell, 2001). It would therefore be interesting to determine the structure of the Sds22-PP1_C complex to uncover the molecular basis of the Sds22p-Glc7p interaction.

As previously found for its human and fission yeast homologues, S. cerevisiae Sds22p is a nuclear protein despite the absence of any clearly defined monopartite (Kalderon et al., 1984) or bipartite (Robbins et al., 1991) nuclear localization sequence (NLS). Although a 39 kDa polypeptide such as Sds22p could potentially gain entry to the nucleus without the aid of a specific import signal, we think this is unlikely in the case of Sds22p, which is largely composed of LRRs and is therefore expected to be non-globular and rather asymmetric (Kobe and Deisenhofer, 1994). In this case, Sds22p may either contain a novel signal for nuclear import, or alternatively it could be imported as a complex with another protein such as Glc7p. Although Glc7p has an excellent candidate monopartite NLS (RKKK) at its extreme C-terminus, recent work suggests that this region is not essential either for Glc7p nuclear localization or function (Hong et al., 2000). Furthermore, mutant Sds22 proteins in fission yeast that failed to bind PP1_C could nonetheless localize to the nucleus, while other variants which were excluded from the nucleus could still bind PP1_C (Stone et al., 1993), strengthening the idea that PP1_C binding and nuclear import are not obligatorily linked. Additional examples of nuclear LRR proteins that also apparently lack a classical NLS include Drosophila LRR47 (Buchanan et al., 1998) and the human splicing factor U2A’ (Sillekens et al., 1989), although other LRR proteins such as CIITA (a transactivator of human MHC class II genes) (Hake et al., 2000), contain potential NLSs as well as LRRs. Mutations in the LRR region of CIITA specifically affect its nuclear localization (Hake et al., 2000). It is therefore conceivable that LRRs constitute a novel determinant of nuclear import, although this could be due to their role as mediators of protein-protein interaction rather than because they contain an intrinsic nuclear import signal. In fission yeast Sds22, mutational analysis revealed that C-terminal truncation or point mutations in LRRs 5 or 9 led to nuclear exclusion of the mutant proteins and loss of function, although mutations in some of the other LRRs failed to affect nuclear localization (Stone et al., 1993).

The cellular role of Sds22p is clearly a critical question and at least two types of model can be proposed for how the Sds22p-Glc7p complex might influence PP1_C activity. One possibility is that Sds22p activates Glc7p function towards key nuclear PP1_C substrates that are required for chromosome stability and other functions. This is supported by the reciprocal, high-copy suppression by each gene of recessive mutations in the other, and by the finding that sds22-6 mimics glc7 mutations in partially suppressing ipl1-2 temperature-sensitivity (Hsu et al., 2000). This suggests that Sds22p and Glc7p function together in a positive sense to create a nuclear PP1 activity. Furthermore, although fission yeast Sds22 immunoprecipitates lacked the phosphorylase phosphatase activity shown by PP1_C alone, they contained a histone H1 phosphatase activity (Stone et al., 1993), supporting a model whereby the Sds22-PP1_C complex can be active as a phosphatase, at least against certain substrates. However, when Sds22p function is lost, the normally uniform nuclear localization of Glc7p is dramatically changed such that the overall level of nuclear PP1 is reduced and the remaining PP1 becomes localized in a small number of foci. Since the interaction of Glc7p with the mutant sds22-6 polypeptide is itself temperature-sensitive, this suggests that Sds22p binding is required to maintain proper nuclear localization of Glc7p. A second possibility is therefore that Sds22p plays a chaperone-like role for nuclear Glc7p, preventing aggregation of the free PP1_C subunit and/or helping to retain it in the nucleus, but not necessarily directing it towards specific substrates. This model would predict that the sds22 mutations might affect a wider range of nuclear PP1_C functions.

In addition to the dramatic effect of sds22-6 on nuclear localization of Glc7p, both sds22 alleles described here confer a profound chromosome loss phenotype at higher growth temperatures. If reflected uniformly across all 16 chromosomes, the high rates of loss seen in sds22-6 at 32°C would suggest that only ∼75% of cells would inherit a complete genome at each cell division. High-copy GLC7 also promotes chromosome instability (Francisco et al., 1994) and exacerbates the chromosome loss defect in the sds22 mutants despite partially suppressing their growth defect (M.W.P. and M.J.R.S., unpublished). Thus the lethality of the sds22 mutants is unlikely to result from chromosome loss per se. A number of glc7 alleles including glc7-10 (Sassoon et al., 1999), glc7-129 (Bloecher and Tatchell, 1999) and glc7-12 (MacKelvie et al., 1995) A. Engles and M.J.R.S., unpublished) confer a mitotic arrest phenotype due to mitotic checkpoint activation. Such mutants show an in vitro defect in microtubule binding by kinetochores (Sassoon et al., 1999) (I. Sassoon et al., unpublished) that if representative of the in vivo situation would be sufficient to account for the observed checkpoint activation This is in contrast to the budding yeast sds22 mutants described in this work that do not arrest in mitosis despite showing severe chromosome instability at higher growth temperatures. Conversely, the conditional fission yeast sds22 mutant described by Stone et al. (Stone et al., 1993) arrested homogeneously in mitosis, although it is not known whether this synchronous arrest is checkpoint-dependent. If Sds22p is required generally for nuclear PP1_C function then it is perhaps surprising that our sds22 mutants do not also activate the checkpoint. However, Glc7p most likely has multiple nuclear roles that do not all result in checkpoint activation. For example, some glc7 mutants that clearly suppress the temperature-sensitivity of ipl1-2 and raise the phosphorylation level of histone H3 on ser-10 (glc7-127), don’t by themselves activate the mitotic checkpoint. Conversely, glc7-129 mutants lacking the Mad/Bub checkpoint are still delayed at the end of the cell cycle, pointing to an additional defect late in mitosis or during cytokinesis. We have also observed synthetic lethality between glc7-129 and several genes encoding microtubule motors (A.B. and K.T., unpublished) that point to a cell cycle role separate from microtubule binding at the kinetochore. Perhaps either the sds22 mutants show chromosome loss for some reason other than faulty kinetochore regulation that does not lead to checkpoint activation or, alternatively, the checkpoint-dependent arrest expected to ensue from the type of kinetochore defect seen in the above glc7 mutants is masked due to defects in multiple Glc7p nuclear functions.

We thank Kim Nasmyth for providing various yeast strains, Kim Arndt for the HA-tagged GLC7 construct and Paul Andrews for useful comments and suggestions. We gratefully acknowledge the support of the Wellcome Trust (project grant 046956 to M.J.R.S.) and the National Institutes of Health (NIH grant GM47789 to K.T.).