Sister chromatid cohesion, which is established during the S phase of the eukaryotic cell cycle and persists until the onset of anaphase, is essential for the maintenance of genomic integrity. Cohesion requires the multi-protein complex cohesin, as well as a number of accessory proteins including Pds5/BIMD/Spo76. In the budding yeast Saccharomyces cerevisiae Pds5 is an essential protein that localises to chromosomes in a cohesin-dependent manner. Here we describe the characterisation in the fission yeast Schizosaccharomyces pombe of pds5+, a novel, non-essential orthologue of S. cerevisiae PDS5. The S. pombe Pds5 protein was localised to punctate nuclear foci in a manner that was dependent on the Rad21 cohesin component. This, together with additional genetic evidence, points towards an involvement of S. pombe Pds5 in sister chromatid cohesion. S. pombe pds5 mutants were hypersensitive to DNA damage and to mitotic metaphase delay, but this sensitivity was apparently not due to precocious loss of sister chromatid cohesion. These cells also suffered increased spontaneous chromosome loss and meiotic defects and their viability was dependent on the spindle checkpoint protein Bub1. Thus, while S. pombe Pds5 has an important cohesin-related role, this differs significantly from that of the equivalent budding yeast protein.
During DNA replication in eukaryotes, the newly replicated sister chromatids become stably associated along their length. Some level of sister chromatid cohesion is maintained until the onset of mitotic anaphase, and is required for the establishment of bipolar attachments to spindle microtubules, and hence for accurate chromosome segregation. Dissolution of cohesion is thought to be fundamental to the initiation of poleward chromosome movement in anaphase. Cohesion is maintained by the conserved multi-protein complex cohesin, which contains Smc1, Smc3, Scc1/Mcd1 and Scc3 (in the protein nomenclature of the budding yeast Saccharomyces cerevisiae), although it is not yet clear if this complex constitutes a literal protein `bridge' between sister chromatids ( Guacci et al., 1997; Losada et al., 1998; Michaelis et al., 1997). Dissolution of cohesion in S. cerevisiae is brought about by the Esp1-catalysed proteolysis of Scc1 ( Uhlmann et al., 2000). DNA polymerase σ/Trf4 was recently shown to be required for the establishment of cohesion during S phase in S. cerevisiae ( Wang et al., 2000b). This novel polymerase may be recruited by a replication factor C (RFC)-like complex containing Ctf18, Ctf8 and Dcc1 to replication forks that have encountered sites of cohesion ( Mayer et al., 2001). Additional factors (Eco1/Ctf7/Eso1, Scc2/Mis4 and Scc4) are required for the establishment of cohesion, but not for its maintenance ( Ciosk et al., 2000; Furuya et al., 1998; Tanaka et al., 2000; Tomonaga et al., 2000; Toth et al., 1999). The cohesin-associated protein Pds5 is also required for the establishment and maintenance of cohesion in S. cerevisiae ( Hartman et al., 2000; Panizza et al., 2000), and Pds5-like proteins appear to play related roles, both in other fungi and in metazoans ( Denison et al., 1993; Holt and May, 1996; Huynh et al., 1986; Sumara et al., 2000). Detailed sequence analysis has identified tandem HEAT (huntingtin, eelongation factor 3, alpha subunit of protein phosphatase 2A, Tor1) repeat units in both S. cerevisiae Pds5 and Scc2, suggesting that in the context of these proteins this structural element might interact with cohesin ( Neuwald and Hirano, 2000; Panizza et al., 2000).
S. cerevisiae Pds5 is essential for maintaining viability during the mitotic cell cycle. Temperature sensitive S. cerevisiae pds5 mutants were identified in a screen for selective loss of viability during a brief mitotic arrest in comparison with a G1 arrest ( Hartman et al., 2000). On shifting such pds5 mutants to the restrictive temperature, most cells fail to complete mitosis, being delayed in anaphase or telophase. This terminal phenotype suggests that the cohesion-promoting activity of Pds5 is essential for cell viability. This interpretation is strengthened by the finding that the bimD6 mutation in A. nidulans can be suppressed by mutation of sudA, which encodes an Smc3-like protein ( Holt and May, 1996). The bimD6 terminal phenotype also indicates an essential role for BIMD in the successful completion of mitosis ( Denison et al., 1993). However, the bimD6 allele contains a nonsense mutation that would be predicted to lead to the production of a short truncated peptide ( van Heemst et al., 2001). Thus, the BIMD protein may be non-essential for mitotic growth at low temperatures. Spo76, the BIMD functional homologue in Sordaria macrospora ( van Heemst et al., 1999), is dispensable for vegetative growth, in contrast to the essential nature of S. cerevisiae Pds5. Although spo76 mutants exhibit clear defects in meiotic and mitotic chromosome cohesion, in the mitotic cycle they show only a comparatively subtle delay in the transition from prometaphase to metaphase ( van Heemst et al., 1999).
A. nidulans bimD mutants are hypersensitive to DNA damage ( Denison et al., 1993), and this sensitivity could be attributable to a role of BIMD in chromatid cohesion. Furthermore, mutation of Schizosaccharomyces pombe rad21 (the SCC1 orthologue) confers defects in the repair of double-strand DNA breaks and mis4 mutants (which are defective in cohesin loading) are sensitive to ultraviolet (UV) irradiation ( Birkenbihl and Subramani, 1992; Furuya et al., 1998). Similarly, a requirement for cohesion in postreplicative DNA repair has recently been demonstrated in S. cerevisiae ( Sjögren and Nasmyth, 2001). Cohesion therefore appears to promote DNA repair from the undamaged sister chromatid and hence affords resistance to a variety of forms of DNA damage. This interpretation is supported by the observation that, during the selection of templates for recombinational DNA repair in vivo, sister chromatids are preferred to homologous chromosomes ( Kadyk and Hartwell, 1992).
Despite these studies on Pds5 and its orthologues in diverse species, there is no clear consensus regarding the role played by this protein in sister chromatid cohesion. The recent completion of the genome sequence of S. pombe, which is only distantly related to S. cerevisiae, A. nidulans and S. macrospora has allowed the identification of the only S. pombe gene significantly related to PDS5. Here, we describe the functional characterisation of this fission yeast gene, which we designate pds5+, and its involvement in the maintenance of genome stability.
Materials and Methods
Searches to identify Pds5-related sequences in S. pombe were performed using the Sanger Centre BLAST server (http://www.sanger.ac.uk/Projects/S_pombe/blast_server.shtml). ψ-BLAST (http://www.ncbi.nlm.nih.gov/cgi-bin/BLAST/nph-psi_blast) searches were used to identify similarities between Pds5 and proteins in the Swissprot database. Megalign (DNASTAR Inc., Madison, WI) was used to generate multiple protein sequence alignments, and the sequence relatedness data shown in Fig. 1A were derived using BLAST2 (BLASTP 2.1.2; http://www.ncbi.nlm.nih.gov/gorf/b12.html) with the values BLOSUM 62, GAP OPEN 11 and GAP EXTENSION 1. Protein sequence motifs were identified using ProfileScan (http://hits.isb-sib.ch/cgi-bin/hits_motifscan).
Fission yeast strains and methods
Conditions for growth, maintenance and genetic manipulation of fission yeast were as described previously ( Moreno et al., 1991). A complete list of the strains used in this study is given in Table 1. Except where otherwise stated, strains were grown at 30°C in YE or EMM2 medium with appropriate supplements. Where necessary, gene expression from the nmt1 promoter was repressed by the addition of 5 μM thiamine to the growth medium.
Minichromosome loss assays were performed using strains as shown in Table 1 containing the non-essential ade6-M216 marked Ch16 minichromosome derivative of chromosome 3 ( Niwa et al., 1986). The ade6-M216 allele complements an unlinked ade6-M210 marker in these strains such that they remain ade+ as long as the minichromosome is maintained. Chromosome loss was measured in the progeny from a single ade+ cell after a known number of generations during which selection for adenine prototrophy had been relaxed by growth on YE agar. Rates of chromosome loss per generation were calculated exactly as described elsewhere ( Murakami et al., 1995; Stewart et al., 1997), according to the formula: where Rn is the proportion of ade+ cells n generations after removal of selection. For each strain tested, mean rates were calculated from five independent measurements.
Gene disruption and related techniques
The one-step disruption method was used, following PCR-mediated generation of the entire ura4+ gene flanked by 80 bp segments from the 5′ and 3′ regions of pds5+, using oligonucleotides PDS5A and B ( Table 2). Following transformation of a diploid strain 428h/429h, ura+ progeny were screened for the desired integration pattern by diagnostic PCR reactions using primer pairs spanning the presumptive recombination sites (details of the additional primers used for this purpose are available from the authors on request). Meiosis and sporulation were induced by plating onto malt extract agar, and tetrad dissection was performed with an MSM micromanipulator (Singer Instruments, UK) as described elsewhere ( Moreno et al., 1991). Construction of the chromosomally HA- and GFP-tagged pds5 strains (pds5-HA and pds5-GFP) was accomplished by an analogous method using primers TAGA and B ( Table 2). The pds5::LEU2 allele was generated by a secondary one-step disruption of the pds5::ura4+ allele using primers as described previously ( Wang et al., 2000a).
Antibodies and immunoblotting
Immunoblotting was performed essentially as described elsewhere ( Ausubel et al., 1995) using Mini-Protean electrophoresis equipment (Bio-Rad, Hercules, CA) and a semi-dry transfer apparatus (Hoefer) in conjunction with Hybond ECL membranes (Amersham Pharmacia, UK). Proteins were detected using enhanced chemiluminescence (ECL, Amersham Pharmacia, UK) following one hour incubations at room temperature with the respective primary and horseradish peroxidase-conjugated anti-mouse antibodies (Sigma, Poole, UK). The mouse anti-influenza hemagglutinin (HA) monoclonal HA-11 (Covance Research Products, Berkeley, CA) was used at 1μ g/ml for detection of HA-tagged Pds5. Cdc2 was detected using the mouse monoclonal antibody Y100 (generated by J. Gannon and kindly provided by H. Yamano).
Gel filtration chromatography
Chromatographic separation of S. pombe lysates was carried out using a superose-6™ column attached to an FPLC workstation (Amersham Pharmacia, UK). Cells from a 100 ml culture of the pds5-HA strain in mid-exponential growth were washed and resuspended in 300 μl FPLC buffer (20% glycerol, 20 mM Tris-Cl pH 7.5, 100 mM NaCl, 0.1 mM EDTA, 1 mMβ -mercaptoethanol, 60 mM β-glycerophosphate, 0.1 mM sodium orthovanadate, 1 mM PMSF) with additional protease inhibitors (Complete™, Roche) and lysed by bead beating (3×30 seconds). After clearing by centrifugation (16,000 g, 15 minutes) the lysate was loaded onto the column that had been preequilibrated with 50 ml FPLC buffer. The eluate was collected in 0.5 ml fractions and these were analysed by SDS-PAGE and immunoblotting. The column was calibrated using the standards: blue dextran (∼2000 kDa), thyroglobin (670 kDa), apoferritin (443 kDa),β -amylase (200 kDa), BSA (66 kDa) and carbonic anhydrase (29 kDa).
Cells fixed in 3.8% formaldehyde were washed in phosphate-buffered saline and stained with 4′,6-diamidino-2-phenylindole (DAPI) before examination by fluorescence microscopy. Images were acquired using a Zeiss Axioskop microscope equipped with a Planapochromat 100× objective, an Axiocam cooled CCD camera and Axiovision software (Carl Zeiss Ltd, Welwyn Garden City, UK), and were assembled using Adobe PhotoShop. In some experiments living cells growing in EMM2 medium were stained by the addition of 5 μg/ml bis-benzimide (Hoechst 33342, Sigma) before examination by fluorescence microscopy. Visualisation of GFP-Swi6 in living cells embedded in 0.6% LMP agarose was performed at room temperature (approximately 22°C) as described elsewhere ( Pidoux et al., 2000).
S. pombe pds5 is not essential for mitotic growth
BLAST (basic local alignment search tool) searches of the complete S. pombe genome using the S. cerevisiae Pds5 amino acid sequence as a query identified a single S. pombe gene significantly similar to PDS5. ProfileScan analysis of the sequence of the corresponding predicted 1205 amino acid residue S. pombe protein (SPAC110.02; accession number Q9HFF5) showed a good match to the bipartite nuclear localisation sequence (NLS) consensus at residues 1129-1146. Similar NLS motifs are present in the C-terminal regions of BIMD, Spo76 and S. cerevisiae Pds5 ( van Heemst et al., 1999). Multiple sequence alignments showed that, over its entire length, the predicted S. pombe protein sequence displays 23% identity (45% similarity) to S. cerevisiae Pds5 ( Fig. 1A), and is related to similar extents to S. macrospora Spo76, A. nidulans BIMD and the human Pds5 orthologue AS3 ( Geck et al., 1999). Among these proteins, the sequence conservation is concentrated in four main blocks, with the intervening spacers being somewhat longer in Spo76 and BIMD than in the other members of the family ( Fig. 1A). In recognition of this level of sequence conservation, and in deference to the pre-existing S. cerevisiae nomenclature, we refer to this S. pombe gene as pds5+. We identified several partial matches to the HEAT motif in the S. pombe Pds5 protein, including one between residues 396 and 423. This sequence lies within a more extensive region of significant similarity between Pds5 (residues 287 to 423) and Mis4 (residues 784 to 918; data not shown).
The one-step gene disruption method was used in a ura4-/ura4- diploid S. pombe strain to replace one copy of the entire pds5 open reading frame with the ura4+ selectable marker. After induction of meiosis and sporulation, microdissection of tetrads showed that the pds5::ura4+ alleles segregated 2:2, indicating that, in contrast to its S. cerevisiae orthologue, S. pombe pds5+ is not essential for mitotic growth. Microscopic examination of exponentially growing cells with the pds5::ura4+ genotype (pds5Δ) showed that there were no gross abnormalities in cell growth or division associated with pds5 deletion ( Fig. 1B).
Deletion of pds5 causes hypersensitivity to DNA damage
Since A. nidulans bimD6 mutant cells display increased sensitivity to DNA damaging agents ( Denison et al., 1993), we were interested to know if the same might be true of S. pombe pds5Δ cells. Exposure of pds5Δ cells and those of a pds5+ control strain to UV showed that the pds5Δ strain is UV-hypersensitive, with a tenfold reduction in cell viability after exposure to 200 J/m2 ( Fig. 2A). Examination of the cells 12 hours after UV irradiation at 150 J/m2 (a dose which only marginally reduced viability of the pds5+ control) showed that most of the pds5Δ cells were highly elongated and had fragmented and/or misshapen nuclei. The sensitivity of pds5Δ cells was not restricted to UV-induced DNA damage, as they were also mildly hypersensitive to the alkylating agent methyl methanesulphonate (MMS) and the radiomimetic agent bleomycin ( Fig. 2B). While this sensitivity was significant, it was not as extreme as that seen in a strain lacking the checkpoint signalling kinase Rad3 (rad3Δ; Fig. 2B).
Mutants defective in the establishment of sister chromatid cohesion or in elements of checkpoint signalling have been shown in many cases to lose viability when DNA replication is inhibited. However, in contrast to their sensitivity to DNA damaging agents, pds5Δ cells were not sensitive to hydroxyurea (which inhibits ribonucleotide reductase and hence DNA replication), in comparison with those of a rad3Δ strain ( Fig. 2C).
Pds5 is required for maintenance of viability when mitosis is arrested
S. cerevisiae pds5 mutants show a characteristic loss of viability during arrest in mitosis, and it was therefore of interest to determine if the same is true of the S. pombe pds5Δ strain. In comparison with a pds5+ strain, pds5Δ cells were hypersensitive to the spindle poison thiabendazole (TBZ; Fig. 3A) and deletion of pds5 showed synthetic lethality with the nda3-KM311 β tubulin mutation ( Hiraoka et al., 1984) at 23°C, at which temperature the cold-sensitive nda3-KM311 single mutant is still able to grow ( Fig. 3B). These data suggest that pds5 is involved in maintenance of viability during delayed progression through mitotic metaphase. In order to define the pds5Δ defect more precisely, release from an nda3-KM311-induced metaphase block was monitored by microscopy of pds5+ and pds5Δ cells. By 9 minutes after release most of the pds5+ cells had entered anaphase with two apparently equal daughter nuclei ( Fig. 3C). Under the same conditions the pds5Δ cells appeared to enter anaphase on schedule, but showed an unusually high proportion of cells with lagging chromosomes ( Fig. 3C). At 2 hours after release these cells showed a variety of apparently lethal abnormalities, including total failure of chromosome segregation and the `cut' (cell untimely torn) phenotype (data not shown). We conclude that failure to complete mitosis probably accounts for the observed genetic interaction between pds5 deletion and nda3-KM311, and for the sensitivity of the pds5Δ strain to TBZ.
Sensitivity of S. cerevisiae pds5 mutants to anti-microtubule agents has been ascribed to their failure to maintain sister chromatid cohesion during mid-mitotic arrest ( Hartman et al., 2000; Panizza et al., 2000). Sister centromere separation in living S. pombe cells can be monitored conveniently by the use of a strain expressing a green fluorescent protein (GFP)-tagged version of the kinetochore component Ndc80 ( Wigge and Kilmartin, 2001). Ndc80-GFP appears as a single fluorescent spot in interphase cells, while in anaphase the clustered kinetochores are seen as two spots, one in each daughter nucleus. This pattern of Ndc80-GFP distribution was seen in nda3-KM311 pds5Δ cells grown at the permissive temperature of 30°C ( Fig. 3D). When this strain was shifted to the restrictive temperature of 18°C to induce arrest in a metaphase, most cells (95%) still showed a single Ndc80-GFP spot, indicating that the sister centromeres were clustered. The remaining cells had three well-defined masses of condensed chromatin, presumably corresponding to the three S. pombe chromosomes, with an Ndc80-GFP spot associated with each ( Fig. 3D). Of a total of 500 cells examined, none with more than three Ndc80-GFP spots was seen. Thus, in contrast to the corresponding S. cerevisiae mutant, S. pombe pds5Δ cells showed no indication of premature sister centromere separation during metaphase arrest.
Elevated rate of chromosome loss on deletion of pds5
Given the chromosome segregation defects observed on release of pds5Δ cells from metaphase arrest, we reasoned that Pds5 might also be required to ensure the normal fidelity of chromosome segregation during unperturbed mitosis. To address this point, pds5+ and pds5Δ strains containing a non-essential minichromosome (Ch16) carrying an adenine biosynthetic marker were used to measure rates of chromosome loss ( Fig. 4A). This assay showed that spontaneous loss of the minichromosome was 30-fold more frequent in the pds5Δ strain in comparison with the pds5+ control (mean loss rates were 0.00714 and 0.00024 per generation, respectively). We conclude that, while not essential, pds5+ makes an important contribution to maintenance of genome stability in S. pombe.
Although the gross nuclear morphology of most pds5Δ cells appeared normal ( Fig. 1B), we were interested to see if the elevated rate of mini-chromosome loss would be reflected in chromosome segregation defects as scored by microscopy. To address this point, pds5Δ and pds5+ strains were constructed containing an integrated GFP-tagged swi6 gene. Like the endogenous chromodomain protein Swi6, GFP-Swi6 localizes to the nucleus and is concentrated at the heterochromatic centromeres and telomeres ( Pidoux et al., 2000). Anaphase chromosome separation, monitored in living cells by GFP-Swi6 fluorescence microscopy, appeared normal in the pds5+ background, as well as in the majority of pds5Δ cells. Abnormal anaphase progression was nonetheless detectable in approximately 5% of the pds5Δ cells, but not in the pds5+ control ( Fig. 4B,C). In some cells this took the form of segregation, with apparently normal kinetics, of two unequal masses of GFP-Swi6 ( Fig. 4B). In other pds5Δ cells an extended bridge of GFP-Swi6 persisted for some time between the presumptive daughter nuclei ( Fig. 4C, upper panels), while still others contained discrete centers of GFP-Swi6 fluorescence, presumably corresponding to entire lagging chromosomes ( Fig. 4C, lower panels). Chromosome segregation defects at the frequencies seen would be sufficient to account for the elevated rate of minichromosome loss in pds5Δ cells.
One possible explanation for the observed mitotic defects in pds5Δ cells would be that, in the absence of Pds5, some aspect of centromere function or spindle microtubule attachment is defective in a significant proportion of cells. To address this possibility, the effect of combining the pds5 deletion with deletion of the bub1 spindle checkpoint gene ( Bernard et al., 1998) was investigated. In the absence of bub1, cells with centromeric defects fail to delay the onset of anaphase appropriately and suffer catastrophic chromosome missegregation. Tetrad dissection following sporulation of a pds5Δ/pds5+ bub1Δ/bub1+ diploid strain showed that pds5Δ bub1Δ double mutants grew much more slowly than the equivalent single mutants ( Fig. 4D). Examination of DAPI stained samples from liquid cultures showed that bub1Δ cells suffered chromosome mis-segregation events ( Fig. 4E), as reported previously ( Bernard et al., 1998). These were more frequent than those seen in pds5Δ cells, but were not sufficient to cause a significant growth disadvantage such as that seen in pds5Δ bub1Δ mutants ( Fig. 4D). The growth disadvantage in the latter was associated with a variety of chromosome missegregation events ( Fig. 4E). Interestingly, no such genetic interaction was seen between pds5Δ and deletion of mad2, another spindle checkpoint gene (data not shown).
Meiosis is frequently abnormal in the absence of pds5
In S. macrospora, Spo76 is required for sister chromatid cohesion both in mitosis and in meiosis. Having established that Pds5, the only protein in S. pombe significantly related to Spo76, has a role in the mitotic cell cycle, we therefore made a preliminary investigation of the possibility that Pds5 might also have a role in meiosis. Induction of meiosis and sporulation in a homozygous pds5Δ diploid strain gave rise to asci, 30% of which had fewer than four spores, which were in most cases misshapen and abnormally sized and had substantially reduced viability ( Fig. 5). The DNA content of these abnormal spores was frequently either greater or less than that of normal haploid spores ( Fig. 5A, lower panels). By contrast, atypical asci of this sort represented <1% of those formed in parallel by a pds5+/pds5+ diploid strain. Completion of meiosis and/or sporulation at the normal frequency thus depends at some level on pds5+ function.
Characterisation of Pds5 protein in S. pombe
Targeted recombination was used to add an HA epitope tag sequence to the 3′ end of the pds5 open reading frame in its normal chromosomal context, generating the pds5-HA strain (see Materials and Methods). Immunoblotting of pds5-HA whole cell lysates with an anti-HA antibody showed a single protein band of approximately 140 kDa, consistent with the predicted relative molecular mass for Pds5 of 138874 ( Fig. 6; and data not shown). Immunoblotting of anti-PDS5 immunoprecipitates from human cells has been used to demonstrate an interaction between PDS5 and cohesin subunits, although most PDS5 was not found to be stably associated with cohesin as judged by sucrose gradient fractionation of cell extracts ( Sumara et al., 2000). No size fractionation data have been published to date for S. cerevisiae Pds5 or any of its fungal orthologues, however. We therefore used gel filtration chromatography to estimate the apparent size of S. pombe Pds5 in a cell lysate prepared from the pds5-HA strain ( Fig. 6). Under the conditions used, the HA-tagged Pds5 migrated as a single peak with an apparent size between 670 and 2000 kDa. Parallel fractionation of an extract from a rad21-HA strain showed that the Rad21-HA protein (and, by extension, other previously characterized cohesin components) was also present in a single peak with a similar size distribution to that seen for Pds5-HA.
The relationship between Pds5 and cohesin
To monitor the localisation of Pds5 in living cells, the one step gene replacement method was used to generate a strain (pds5-GFP) encoding a C-terminally GFP tagged version of Pds5. The GFP-tagged protein appeared to be functional as judged by the lack of hypersensitivity to UV of the pds5-GFP strain compared with pds5Δ ( Fig. 2A,B). Examination of living pds5-GFP cells by fluorescence microscopy showed that Pds5-GFP (and, by inference Pds5) was predominantly localised to the nucleus, within which it gave a diffusely speckled signal ( Fig. 7A). There were no obvious differences in this pattern among cells at different cell cycle stages, and in nda3-KM311 cells arrested in a metaphase-like state by incubation at the restrictive temperature of 18°C for 12 hours the Pds5-GFP signal co-localised with condensed chromosomes.
Localisation of Pds5 to chromatin in S. cerevisiae is dependent upon the integrity of the Scc1/Mcd1 cohesin component ( Hartman et al., 2000; Panizza et al., 2000). To see if the same might be true of the orthologous S. pombe proteins, the localisation of Pds5-GFP was monitored in a temperature-sensitive rad21-K1 strain ( Fig. 7B). At the permissive temperature of 25°C, Pds5-GFP was again predominantly nuclear, but after inactivation of the Rad21 cohesin component by shifting to the restrictive temperature of 36°C the Pds5-GFP signal was diffusely localised throughout the cell. No such change in localisation was seen on shifting a pds5-GFP (rad21+) strain from 25°C to 36°C ( Fig. 7B). Thus Pds5 localisation to the nucleus is cohesin-dependent in S. pombe, as it is in S. cerevisiae. By contrast, Rad21-GFP remained in the nuclear compartment on deletion of pds5, although in pds5Δ rad21-GFP cells the Rad21-GFP appeared more punctate than it did in pds5+ cells ( Fig. 7C).
To investigate further this apparent relationship between pds5 and cohesin, the meiotic progeny of a diploid rad21-K1 pds5Δ strain ( Table 1) were characterised following tetrad microdissection ( Fig. 8A). From 102 tetrads, the numbers of pds5+ rad21+, pds5+ rad21-K1 and pds5::LEU2 rad21+ segregant colonies visible after 5 days growth at 25°C were 77, 45 and 78, respectively. These numbers are lower than the 102 that would be expected for independently segregating markers, and indicate that there was an overall loss of spore viability attributable to heterozygosity at the pds5 and rad21 loci in the diploid. Nonetheless, haploid segregants bearing either the rad21-K1 or the pds5::LEU2 allele were able to grow and form colonies reasonably efficiently. By contrast, no rad21-K1 pds5::LEU2 segregant colonies were visible after 7 days growth. In many cases the positions occupied by spores of this genotype could be deduced from the genotypes of the other segregants. Microscopic examination showed that these rad21-K1 pds5::LEU2 cells were highly elongated, suggesting that partial loss of rad21 function in the absence of pds5 leads to a failure of cell cycle progression ( Fig. 8A).
Further genetic interactions between the pds5 deletion and genes encoding components of the sister chromatid cohesion pathway were sought. S. pombe Mis4 is a cohesin loading factor/adherin orthologous to S. cerevisiae Scc2. Introduction of the temperature sensitive mis4-242 mutation into a pds5Δ background revealed a synthetic lethality phenotype at 28°C, at which temperature each of the corresponding single mutants grew relatively normally ( Fig. 8B). Reducing the temperature to 25°C allowed the pds5Δ mis4-242 cells to grow, albeit slowly. On microscopic examination these cells showed a variety of aberrant morphologies indicative of cell cycle defects, including extensive elongation and chromosome missegregation ( Fig. 8B, lower panel). No such defects were seen in the single mis4-242 mutant grown at the same temperature. Thus pds5 can be linked genetically both to cohesin itself and to a cohesin loading factor.
Pds5-related proteins have previously been implicated in sister chromatid cohesion in organisms as diverse as fungi and vertebrates. Loss of function of PDS5/SPO76/bimD in distantly related simple eukaryotes is nonetheless associated with a marked variety of phenotypes, encompassing complete failure of mitotic chromosome segregation as well as much more subtle mitotic and meiotic defects. In order to clarify common functions of these proteins, we have investigated the localisation and function of the product of pds5, the sole representative of the PDS5/SPO76/bimD family in S. pombe. The primary amino acid sequences of Pds5 and its orthologues are related to each other to similar degrees throughout their length, with particularly strong conservation in the N-terminal ∼150 amino acid residues ( Fig. 1A).
In some respects, S. pombe pds5+ more closely resembles S. macrospora SPO76 than S. cerevisiae PDS5 or A. nidulans bimD. Specifically, the S. pombe gene is dispensable for mitotic growth but is required for the efficient completion of meiosis and sporulation (Figs 1, 5). Pds5 was found previously to be required for chromosome condensation as well as cohesion in S. cerevisiae ( Hartman et al., 2000), but we find no indication that S. pombe Pds5 is required for the chromosome condensation seen in metaphase arrested cells ( Fig. 3D). If the link between cohesion and condensation that has been described in S. cerevisiae ( Guacci et al., 1997) also operates in the fission yeast, the competence of pds5Δ cells for chromosome condensation would be consistent with their lack of overt cohesion defects ( Fig. 3). However, Mis4, which is required for sister chromatid cohesion in S. pombe, is not required for condensation ( Furuya et al., 1998), suggesting that any such link may not be straightforward. Indeed, it has been suggested that, in Xenopus, cohesin may not be involved in chromosome condensation at all ( Losada et al., 1998). An involvement in chromosome condensation in S. cerevisiae but not in S. pombe might explain why Pds5 is essential for viability in the former but not the latter.
What then might be the mitotic role of S. pombe Pds5? Despite the lack of precocious sister chromatid separation in pds5Δ cells under the conditions tested here, a number of lines of evidence point towards an intimate connection between Pds5 and the cohesion process. First, like its budding yeast counterpart, S. pombe Pds5 localises to nuclear foci in a manner that is dependent on cohesin function ( Fig. 7A-C). The constitutive localisation of Pds5 to this nuclear compartment throughout the S. pombe cell cycle is reminiscent of the behaviour of the Rad21 cohesin ( Tomonaga et al., 2000) and Mis4 adherin ( Furuya et al., 1998). The majority of cohesin remains associated with chromatin throughout the mitotic cell cycle in S. pombe, in contrast to the situation in budding yeast, where Scc1/Mcd1 cleavage at the onset of anaphase is associated with dissociation of cohesin and Pds5 from the chromosomes ( Guacci et al., 1997; Hartman et al., 2000; Michaelis et al., 1997; Panizza et al., 2000; Toth et al., 1999). The persistence of Pds5-GFP in cells at all stages of mitosis ( Fig. 7A) suggests that, unlike Scc1/Rad21, Pds5 may not be regulated by proteolysis. Disappearance of Pds5-GFP from the nucleus on inactivation of Rad21 ( Fig. 7C) suggests that the putative bipartite NLS in Pds5 ( Fig. 1A) is insufficient to confer constitutive nuclear localisation. In the absence of functional cohesin, Pds5 may be subject to nuclear export or degradation; distinction between these possibilities awaits further investigation.
A role for S. pombe Pds5 in chromatid cohesion would also be consistent with the genetic interactions between pds5, rad21 and mis4 ( Fig. 8). The observed lack of hypersensitivity of pds5Δ cells to HU ( Fig. 2C) suggests that pds5 may not be required for the establishment of cohesion in S phase, in contrast to mis4, which when mutated confers HU hypersensitivity ( Furuya et al., 1998). A role in cohesion could also be suggested by the sensitivity of pds5Δ cells to DNA damage ( Fig. 2). Mutation of other genes involved in the establishment or maintenance of cohesion has previously been shown to lead to DNA damage sensitivity ( Birkenbihl and Subramani, 1992; Denison et al., 1993; Furuya et al., 1998; Sjögren and Nasmyth, 2001). The elongation of pds5Δ cells seen after UV irradiation ( Fig. 2A) suggests that cell cycle checkpoint responses to UV-induced DNA damage are established in the absence of pds5, but that some aspect of DNA repair or cell cycle resumption is defective, accounting for the observed increase in UV sensitivity. A general role for Pds5 in cell cycle resumption seems unlikely, however, as G2 arrest for up to 6 hours in a cdc25-22 pds5::ura4+ strain was not accompanied by any significant loss of viability (data not shown). Instead, a role for Pds5 in promoting repair of a variety of DNA lesions seems more likely ( Fig. 2). This function could be an indirect consequence of a primary involvement of Pds5 in cohesion.
In line with the proposed connection between Pds5 and cohesion in S. pombe, soluble Pds5 was present in a high molecular weight complex similar in size to that containing the Rad21 cohesin component ( Fig. 6). This is consistent with the behaviour of human PDS5, which could be co-immunoprecipitated with cohesin components from crude cell lysates, although it was exclusively localised to size fractions smaller than those containing cohesin when lysates were separated by sucrose density gradient centrifugation ( Sumara et al., 2000). Interestingly, size fractionation of pds5-HA lysates in the absence of phosphatase inhibitors yielded a Pds5-HA peak with a much smaller average size, consistent with that expected for the monomeric protein (data not shown). This suggests that retention of Pds5 in a high molecular weight complex, perhaps including cohesin itself, depends on the maintenance of phosphorylation of one or more proteins. Despite this apparent co-migration, we were unable to co-immunoprecipitate known cohesin components reproducibly with Pds5-HA, or to precipitate Pds5-HA with an anti-Rad21 antibody (data not shown). Any interaction between Pds5 and cohesin therefore appears quite labile in soluble extracts. However, our data do not rule out the possibility that such an interaction might be more stable in the context of chromatin-associated cohesin.
In the mitotic cycle, pds5Δ cells had an elevated rate of chromosome loss and lagging chromosomes were readily detected ( Fig. 4). These cells were also unusually vulnerable to arrest at metaphase, as judged by TBZ sensitivity and synthetic lethality with nda3-KM311 at 23°C ( Fig. 3). Microscopic examination of pds5Δ cells released from an nda3-KM311 arrest suggested that chromosome segregation was frequently grossly abnormal under these circumstances. In the absence of any evidence for precocious sister separation, it is still possible that these observations reflect altered cohesion. For example, Pds5 may be required for the correct localisation of cohesin to specific chromosomal sites, or for the inhibition of excessive cohesin loading. These possibilities would be consistent with our observation that Rad21-GFP remains associated with chromatin in the absence of Pds5, although the precise pattern of its chromatin localisation appears subtly different ( Fig. 7C). In this case the frequent failure of pds5Δ cells to complete mitosis properly after metaphase delay ( Fig. 3C) could be the result of anaphase progression without complete loss of sister chromatid cohesion. Similar defects, occurring at a lower frequency, could explain the 30-fold elevation in minichromosome loss and the presence of lagging chromosomes in pds5Δ cells ( Fig. 4). As this aspect of the pds5Δ phenotype is particularly marked after a metaphase delay, it would appear that the putative regulatory role of Pds5 is more important during mitosis than it is in interphase. Analogous defects in meiosis I and/or II could explain the observed defects in spore formation ( Fig. 5).
The sensitivity of pds5Δ cells to metaphase arrest does not appear to reflect a loss of spindle checkpoint integrity, since nda3-KM311 pds5Δ cells were able to arrest in a metaphase-like state on being shifted to the restrictive temperature ( Fig. 3C,D). The potent genetic interaction observed between pds5 and bub1 ( Fig. 4D,E) could suggest that the spindle checkpoint is at least partially activated in pds5Δ cells in the absence of any additional perturbation. If this is the case Pds5, through its putative role in `fine tuning' sister chromatid cohesion, may be required for the efficient establishment of bipolar attachments of chromosomes to spindle microtubules. Interestingly, we found no equivalent genetic interaction between pds5 and mad2, while the mitotic delay in S. pombe cohesin mutants was reported to be mad2-dependent ( Tomonaga et al., 2000). A recent study demonstrated that, in addition to its spindle checkpoint function, Bub1 is required for centromeric cohesion during meiosis ( Bernard et al., 2001). The genetic interaction between pds5 and bub1 ( Fig. 4) could therefore indicate that both are also involved in mitotic cohesion.
Investigation of cohesin-related components in S. pombe should prove complementary to studies in S. cerevisiae with respect to understanding chromatid cohesion and its regulation in other organisms, as the overall organisation of mitosis in the two yeasts differs in several respects ( Russell and Nurse, 1986). Indeed, significant differences between these species in terms of cohesin composition and proteolysis have been described recently ( Tomonaga et al., 2000). Specifically, Psc3 (the S. pombe orthologue of Scc3) is not stably associated with cohesin, and only a minor fraction of Rad21 is subject to proteolysis at the onset of anaphase. Genetic approaches in distantly related simple eukaryotes, combined with biochemical investigations in these and other systems, should eventually provide a comprehensive understanding of sister chromatid cohesion and its regulation.
We are grateful to Mitsuhiro Yanagida and Takashi Toda for valuable advice and provision of strains, to John Kilmartin, Tony Carr, Robin Allshire, Jean-Paul Javerzat and Hideo Ikeda for additional strains, and to Ian Hickson and other members of the ICRF Molecular Oncology Laboratory for their advice and comments on the manuscript. This work was supported by the Imperial Cancer Research Fund, the Association of Commonwealth Universities (scholarship to R.L.R.) and the Association for International Cancer Research.
- Accepted October 26, 2001.
- © The Company of Biologists Limited 2002