Sister-chromatid cohesion mediated by the alternative RF-C^{Ctf18/Dcc1/Ctf8}, the helicase Chl1 and the polymerase-α-associated protein Ctf4 is essential for chromatid disjunction during meiosis II

The propagation of our genomes from one generation to the next requires the accurate segregation of chromosomes during both mitotic and meiotic divisions. Errors during mitosis contribute to oncogenesis and those during meiosis give rise to aneuploid gametes. In mitotic cells, sister chromatids are held together from their generation during S-phase until the metaphase-to-anaphase transition (Guacci et al., 1994). This phenomenon, called sister-chromatid cohesion, enables cells to determine which chromatids are sisters and which must therefore be pulled towards opposite poles at the onset of anaphase.

Work over recent years has identified a conserved multiprotein complex called cohesin that holds sister chromatids together until the onset of anaphase (Haering and Nasmyth, 2003; Hirano, 2002). In budding yeast, cohesin consists out of four core subunits, Smc1, Smc3, Scc1 and Scc3. It is tightly associated with chromosomes from S-phase until metaphase, particularly around centromeres but also at discrete locations along chromosome arms (Guacci et al., 1997; Michaelis et al., 1997; Orr-Weaver, 1999). Biochemical and structural analyses have shown that cohesin forms a ring-like structure in which Scc1 bridges ABC-like ATPase heads situated at the termini of a V-shaped Smc1-Smc3 heterodimer (Anderson et al., 2002; Gruber et al., 2003; Haering et al., 2002). This ring-like structure has led to the hypothesis that cohesin could hold sister chromatids together by trapping sister DNA molecules within its ring (Haering et al., 2002).

The segregation of sister chromatids to opposite poles at anaphase requires that sister kinetochores be attached to microtubules emanating from opposite spindle poles (a state called biorientation or amphitelic attachment). Cohesin promotes biorientation by ensuring that tension is generated when microtubules attach in an amphitelic manner by resisting the tendency of microtubules to pull sisters apart (He et al., 2000; Tanaka et al., 2000). This tension stabilizes kinetochore-microtubule attachments, which are otherwise disrupted through the action of the Ipl1/Aurora-B protein kinase (Tanaka et al., 2002). In the absence of sister chromatid cohesion (as in cohesin mutants) chromosomes are unable to biorient and segregate at random (Guacci et al., 1997; Michaelis et al., 1997; Tanaka et al., 2000).

Cleavage of cohesin's Scc1 subunit by a cysteine protease called separase triggers the metaphase-to-anaphase transition and the segregation of sister chromatids to opposite poles (Uhlmann et al., 1999; Uhlmann et al., 2000). Separase is kept inactive for most of the cell cycle by binding to securin (Pds1 in budding yeast) (Ciosk et al., 1998; Uhlmann et al., 1999). Destruction of securin at the hands of a ubiquitin protein ligase called the anaphase-promoting complex or cyclosome (APC/C) (Cohen-Fix et al., 1996; Funabiki et al., 1996) at the metaphase-to-anaphase transition occurs when every single chromosome has bioriented. Unaligned chromosomes inhibit the APC/C through activating Mad2 (Musacchio and Hardwick, 2002).

A major variation of this theme enables the production of haploid gametes from diploid somatic cells during meiosis (Petronczki et al., 2003), which involves two successive rounds of chromosome segregation, termed meiosis I and II, preceded by a single round of DNA replication. The two meiotic divisions differ in important but distinct respects from mitosis. Homologous and not sister centromeres are pulled in opposite directions during meiosis I. This is made possible by reciprocal recombination between homologous chromatids that lead to the interconnection of maternal and paternal centromeres through chiasmata and by the action of `monopolin' proteins like Mam1, which suppress biorientation and promote monoorientation of sister kinetochores (Rabitsch et al., 2003; Toth et al., 2000). Chiasmata interconnect homologous centromeres by virtue of sister-chromatid cohesion distal to crossovers and they are therefore resolved by the destruction of this cohesion, an event that triggers the first meiotic division. Unlike mitosis, sister-chromatid cohesion in the vicinity of centromeres is preserved at this point and so it can be used to biorient sister centromeres during meiosis II. Meiosis II is more similar to mitosis but differs from it by the absence of any DNA replication. Meiosis II chromosome segregation is triggered by destruction of the cohesion between sister centromeres inherited from meiosis I.

In many eukaryotes, cohesin's Scc1 subunit is largely replaced during meiosis by a meiosis-specific variant called Rec8 (Klein et al., 1999; Watanabe and Nurse, 1999). In yeast, the first meiotic division is triggered by separase-mediated cleavage of Rec8 along chromosome arms, an event caused by the destruction of securin by the APC/C (Buonomo et al., 2000; Kitajima et al., 2003; Salah and Nasmyth, 2000). Protection of Rec8 in the vicinity of centromeres from separase at this point requires a kinetochore protein called Sgo1, which is related to Drosophila Mei-S332 (Katis et al., 2004; Kitajima et al., 2004; Marston et al., 2004; Rabitsch et al., 2004). Destruction of securin that had reaccumulated after the first meiotic division leads to a second round of separase activation (Salah and Nasmyth, 2000), causing cleavage of centromeric Rec8 and thereby the onset of anaphase II (Kitajima et al., 2003).

We do not currently understand how sister-chromatid cohesion is established. It is nevertheless clear that cohesin must be present during S-phase. When expressed artificially after DNA replication, cohesin complexes tightly associate with chromosomes but fail to establish functional connections between sister chromatids (Uhlmann and Nasmyth, 1998; Watanabe et al., 2001). This has led to the notion that cohesion arises either from the passage of DNA replication forks through cohesin rings or from a hitherto obscure process facilitated by the physical proximity of the newly synthesized DNA molecules. Establishment of sister-chromatid cohesion during S-phase but neither its maintenance later nor the association of cohesin with chromosomes requires a protein called Eco1/Ctf7 (Skibbens et al., 1999; Toth et al., 1999). The notion that an Eco1-dependent event distinct from mere cohesin loading occurs during S-phase is supported by the finding that overexpression of the DNA polymerase processivity factor and sliding clamp proliferating cell nuclear antigen (PCNA) partially suppresses the lethality of eco1/ctf7 mutants (Skibbens et al., 1999). Furthermore, deletion of the non-essential CTF4 gene, whose product binds DNA polymerase α in vitro (Miles and Formosa, 1992), causes detectable but not lethal defects in sister-chromatid cohesion (Hanna et al., 2001), as do deletions of CTF18, DCC1 and CTF8, whose gene products along with Rfc2, Rfc3, Rfc4 and Rfc5 form an alternative replication factor C (RF-C^{Ctf18/Dcc1/Ctf8}) (Mayer et al., 2001; Naiki et al., 2001). The canonical RF-C^Rfc1 (containing Rfc1 instead of Ctf18/Dcc1/Ctf8) is an integral part of the replication fork and loads PCNA onto DNA, thereby facilitating the switch from the primosome to the replicative DNA polymerase δ (Diffley and Labib, 2002).

Despite elevated chromosome loss and a Mad2-dependent anaphase onset delay (Hanna et al., 2001; Kouprina et al., 1993; Mayer et al., 2001; Spencer et al., 1990), the physiological significance of sister-chromatid cohesion defects detected in mutants lacking Ctf4 or RF-C^{Ctf18/Dcc1/Ctf8} is unclear for two reasons. First and foremost, neither CTF4 nor CTF18/DCC1/CTF8 are essential genes and deletion mutants clearly possess sufficient sister-chromatid cohesion for chromosome biorientation. Second, their cohesion defects have only been measured in cells arrested in M-phase for long periods by treatment with spindle poisons (Hanna et al., 2001; Mayer et al., 2001). It has never been established whether sister-chromatid cohesion in cycling cells is also compromised by the mutations.

We describe here the isolation of CHL1 as a high-copy-number suppressor of a ctf8Δ cells. We show that CHL1, CTF8 and CTF4 are all required for efficient sister-chromatid cohesion in cycling mitotic cells and provide evidence that the helicase encoded by CHL1 is required for the establishment of cohesion during S-phase. We demonstrate that Chl1, Ctf4 and RF-C^{Ctf18/Dcc1/Ctf8} are essential for chromosome segregation during meiosis and, using a functional assay, show that they make a major contribution to the establishment of sister-chromatid cohesion in the vicinity of centromeres. Finally, we show that fission-yeast mutants ctf18 and dcc1 have similar meiotic defects to those observed in budding yeast, suggesting a conserved role for the alternative RF-C^{Ctf18/Dcc1/Ctf8} in sister-chromatid cohesion.

The high proportion of cells with two URA3-GFP signals we observed in mutant strains for the experiments shown in Figs 1, 3 are not a consequence of hyperploidy of the strains for the green-fluorescent-protein (GFP) labeled chromosome V but rather of defects in sister-chromatid cohesion. We determined the degree of hyperploidy of the GFP-labeled chromosome V by counting the number of GFP signals in securin-negative unbudded mitotic cells for Fig. 1 and in unbudded G1 cells prior to entry into meiosis for Fig. 3. Consistent with data from other publications, the proportion of mutant cells hyperploid for URA3-GFP never exceeded 4% and therefore does not influence the interpretation of the results.

The genotypes of all yeast strains used in this study are provided in supplementary Table S1. Tagging, promoter replacement and deletion of genes were performed using the one-step PCR method (Knop et al., 1999; Longtine et al., 1998; Wach et al., 1994). The URA3-GFP system was described previously (Michaelis et al., 1997).

A diploid ctf8Δ yeast (K12359) strain was transformed with a 2μ high-copy-number genomic library constructed in YEp13 (Nasmyth and Tatchell, 1980). 50,000 transformants were grown on plates lacking leucine for 72 hours, replica plated for 48 hours onto plates lacking leucine and regenerated on YEPD for 24 hours at 30°C. Colonies were transferred onto sporulation medium, incubated for 72 hours at 30°C, replica plated onto YEPD glass plates and immediately treated with diethyl ether (Rockmill et al., 1991) to kill all non-sporulated cells. Colonies were germinated and grown for 48 hours at 30°C. Clones showing increased colony growth compared with the parental strain were picked. The library plasmids were recovered and retested for suppression of spore lethality by transformation into K12359. Sequencing of both ends of the insert in the library plasmids revealed the genes responsible for suppression.

Synchronization of strain K12131 by α factor was performed as described previously (Amon, 2002). The experiment was carried out at 25°C in YEP containing 2% raffinose, with or without the addition of 2% galactose at the indicated time points. Samples for flow cytometry and protein extracts, taken every 10 minutes, and samples for in situ immunofluorescence, taken 90 minutes after the release from α factor, were processed as described below.

For the list of genes selected for deletion, see supplementary Table S2. Genes were deleted in the diploid strain K8409, processed and analysed as described previously for a different screen (Rabitsch et al., 2001).

Flow cytometry of DNA content and the preparation of protein extracts were performed as described previously (Klein et al., 1999; Piatti et al., 1996). For immunoblotting, hemagglutinin (HA)-tagged proteins were detected using a monoclonal rat anti-HA 3F10 antibody (Eurogentec) at 1:2000 and the loading control Swi6 was detected using a polyclonal rabbit anti-Swi6 antibody at 1:5000 (Taba et al., 1991). Sporulation time courses were performed as described previously (Buonomo et al., 2000). In situ immunofluorescence was performed as described previously (Pringle et al., 1991) on cells that were fixed overnight on ice in 3.7% formaldehyde. The following primary antibodies were used: monoclonal rat anti-α-tubulin YOL1/34 1:100 (Serotec), monoclonal mouse anti-Myc 9E10 1:5 (Evan et al., 1985), polyclonal rabbit anti-Myc 1:500 (Gramsch) and monoclonal mouse anti-HA 16B12 1:500 (Eurogentec). Chromosome spreads were performed as described previously (Nairz and Klein, 1997) and stained with a monoclonal rat anti-HA 3F10 antibody (Eurogentec) at 1:500 and a rabbit anti-Zip1 serum at 1:500 (kindly provided K. Schmekel). Highly cross-adsorbed secondary antibodies conjugated to Alexa488 (Molecular Probes) (1:500), Cy3 (1:500) or Cy5 (Chemicon) (1:100) were used. DNA was visualized by staining with DAPI. Cells sporulated on plates at 30°C were fixed in ethanol at –20°C, stained with DAPI and scored for the segregation of URA3-GFP under the fluorescence microscope. Image acquisition was performed on an Axioplan 2 Imaging microscope (Zeiss) equipped with a CoolSNAP fx camera (Roper Scientific, Photometrics). Spore viability was determined by microdissection of tetrads sporulated on plates at 30°C for 48 hours and subsequent germination and growth at 30°C for 4 days on YEPD.

The genotypes of all fission yeast (Schizosaccharomyces pombe) strains used in this study are listed in supplementary Table S1. The ura4+ cassette was inserted near the N-terminal end of the dcc1 open reading frame (ORF) (180 bp of 1050 bp, SPAC31A2.15c) and in the middle region of the ctf18 ORF (1603 bp of 2883 bp, SPBC902.02c). Sporulation and the analysis of chromosome segregation using cen1-GFP was performed as described previously (Watanabe and Nurse, 1999).

Deletion of genes encoding subunits specific to RF-C^{Ctf18/Dcc1/Ctf8} causes massive chromosome mis-segregation during meiosis, leading to low spore viability (see below). To identify factors that ameliorate this phenotype, we screened for genes that, when present at high copy number on 2μ plasmids, increased the viability of ctf8Δ spores. In addition to plasmids encoding CTF8 and CTF18, we recovered multiple clones containing the entire CHL1 gene. High-copy-number vectors containing CHL1 were able to increase ctf8Δ spore viability from 5% to 20%. CHL1 belongs to an evolutionary conserved family of DEAH-box helicases and has been shown to be important for the fidelity of chromosome transmission during mitosis in budding yeast (Gerring et al., 1990; Kouprina et al., 1993; Shiratori et al., 1999; Spencer et al., 1990). The defect responsible for chromosome mis-segregation in chl1Δ mutants was nevertheless unknown. Our results raise the possibility that the CHL1 helicase might be involved in generating sister-chromatid cohesion.

RF-C^{Ctf18/Dcc1/Ctf8} and CTF4 are required to maintain efficient sister-chromatid cohesion during a prolonged cell-cycle arrest induced by nocodazole treatment (Hanna et al., 2001; Mayer et al., 2001). It has, however, never been established whether these proteins or Chl1 have such a role in cycling cells in the presence of microtubule pulling forces. To address this, we analysed wild-type, chl1Δ, ctf8Δ and ctf4Δ strains expressing a Myc-tagged version of securin (Pds1-Myc18) and carrying a tandem array of tetracycline operators integrated at the URA3 locus 35 kbp away from the centromere of chromosome V (Michaelis et al., 1997). Expression of a tet-repressor–GFP fusion protein enabled us to visualize sister-chromatid cohesion at the URA3 locus (URA3-GFP). Cycling cells were fixed and stained with DAPI and antibodies against tubulin and the Myc epitope. We identified metaphase cells on the basis of their bipolar spindles, large buds and high levels of nuclear securin, and determined the proportion of these cells containing one or two URA3-GFP signals (Fig. 1A). In contrast to wild-type cells, which never split their replicated URA3 loci into two GFP signals before the onset of anaphase, precocious separation of URA3 loci along the spindle axis was observed in 43% of chl1Δ, 56% of ctf4Δ and 45% of ctf8Δ metaphase cells (Fig. 1A). We noted that the proportion of cells in metaphase was 13% in wild-type, but 31% in chl1Δ and 28% in ctf8Δ, asynchronous cultures. The simplest explanation for the increase in metaphase cells in chl1Δ and ctf8Δ cells is that their defective sister-chromatid cohesion causes a transient metaphase arrest induced by the Mad2-dependent mitotic checkpoint, consistent with previous findings of a Mad2-dependent increase of mutant cells with a 2C DNA contents (Li and Murray, 1991; Mayer et al., 2001). We conclude that Chl1, Ctf8 and Ctf4 are necessary for efficient sister-chromatid cohesion in cycling cells as well in cells arrested for long periods in a metaphase-like state by nocodazole.

Despite their striking cohesion defects, the frequency of mis-segregation of sister chromatids during anaphase in ctf8Δ, ctf4Δ and chl1Δ mutant cells is very low (<2%, data not shown), which is consistent with their high viability. The mutant cells clearly generate sufficient cohesion along chromosome arms to biorient most sister kinetochores despite splitting of sequences at least up to 35 kbp away from the centromere.

Membership of Ctf18, Dcc1 and Ctf8 as part of an alternative RF-C and Chl1's homology to helicases suggest that this class of proteins might have a role in the establishment of sister-chromatid cohesion during S phase. To determine whether Chl1 must act during S phase or whether it can also act during G2 or M phase, the endogenous CHL1 ORF was placed under control of the galactose-inducible GAL1-10 promoter and its N-terminus was fused to three HA epitopes to monitor protein expression. Yeast cells grown in medium containing raffinose (GAL1-10 off) were first arrested in G1 phase with α factor. After release from this arrest, the culture was split into three aliquots. To one culture, no galactose was added (Fig. 1B); to the second, galactose was added immediately upon release (0 minutes; Fig. 1C); to the third, galactose was added 45 minutes after release (Fig. 1D). DNA replication was monitored by flow cytometry and expression of HA3-Chl1 by immunoblotting. DNA replication occurred with similar kinetics in all three cultures. After 90 minutes, when most cells had entered metaphase, we determined the proportion of metaphase cells (with bipolar spindles and high securin levels) with split URA3-GFP signals (Fig. 1E). In the absence of galactose, 44% of metaphase cells had two distinct URA3-GFP loci (Fig. 1E), which is comparable to the proportion in chl1Δ cells (Fig. 1A). Addition of galactose at 0 minutes led to the accumulation of HA3-Chl1 protein to normal levels at 40-50 minutes and before bulk DNA replication. This expression was sufficient for efficient sister-chromatid cohesion because only 7% of metaphase cells had split URA3-GFP signals (Fig. 1E). Addition of galactose at 45 minutes resulted in significant accumulation of HA3-Chl1 by 80 minutes, by which time all cells had completed DNA replication. Unlike early expression, this was not sufficient for efficient sister chromatid cohesion and sister URA3-GFP signals were split in 40% of metaphase cells (Fig. 1E). These data are consistent with the helicase Chl1 being required during S phase.

To analyse the localization of Chl1, we performed in situ immunofluorescence on mitotic cells expressing Chl1-Myc18 from its endogenous locus. Consistent with a function in sister-chromatid cohesion, we found that Chl1-Myc18 localizes to the nucleus (Fig. 1F).

RF-C^{Ctf18/Dcc1/Ctf8}, CTF4 (Hanna et al., 2001; Mayer et al., 2001) and CHL1 (this work) are required for sister-chromatid cohesion in mitotic cells (Fig. 1A). Nevertheless, ctf8Δ, chl1Δ and ctf4Δ mutants are viable and usually disjoin sister chromatids to opposite poles in anaphase (see above). Whether these genes have a more important function in meiosis is unknown. Using a candidate-gene approach based on functional genomic data (see below), we identified two subunits of the alternative RF-C, DCC1 and CTF8 as being essential for meiotic chromosome segregation.

Parallel phenotypic analysis of sporulation in budding yeast has identified 158 genes that are not required for spore formation but appear to be required for efficient spore germination (Deutschbauer et al., 2002). We reasoned that deletion of genes required for meiotic chromosome segregation would result in aneuploidy of spores, low spore viability and hence inefficient germination. We therefore selected 55 of the 158 germination-defective genes [see supplementary Table S2], excluding genes with known functions during meiosis (e.g. SPO11) or clearly assigned roles other than chromosome segregation (e.g. FAR1), and deleted them in a sporulation-proficient diploid yeast strain in which both homologs of chromosome V were tagged with GFP (homozygous URA3-GFP, 35 kbp away from CEN5). This allowed us to monitor the segregation of chromosome V in tetrads following meiosis. Out of the 55 selected genes, deletion of three genes CTF8, DCC1 and YGL214W resulted in severe chromosome segregation defects during meiosis. In wild-type cells, each spore contains one GFP signal corresponding to one chromatid of chromosome V (Fig. 2). More than half of ctf8Δ and dcc1Δ tetrads carry at least one spore without a copy of chromosome V and one with two copies (Fig. 2). YGL214W overlaps with C-terminus of the SKI8 ORF located on the complementary strand. SKI8 is known to be involved in meiotic recombination (Gardiner et al., 1997), which would account for the mis-segregation phenotype of ygl214wΔ. In addition to Ctf8 and Dcc1, Ctf18 is the third subunit specific to RF-C^{Ctf18/Dcc1/Ctf8} that has been shown to be required for sister-chromatid cohesion in mitosis (Hanna et al., 2001; Mayer et al., 2001). Deletion of CTF18 resulted in severe chromosome mis-segregation during meiosis similar to ctf8Δ and dcc1Δ cells (Fig. 2).

We next analysed chromosome segregation in strains containing only a single homolog of chromosome V tagged with GFP (heterozygous URA3-GFP). In the wild type, the two GFP-labeled chromatids always segregated into opposite spores during meiosis II (Fig. 2). In all three deletion strains, sister GFP signals segregated into the same spore in about 30% of tetrads (Fig. 2). This high incidence of meiosis II non-disjunction (Fig. 2) compared with the low frequency of meiosis I homolog non-disjunction (see below, Fig. 3E,F) suggests that meiosis II errors are largely responsible for the chromosome mis-segregation observed in the mutant cells. Consistent with severe aneuploidy, spores from ctf18Δ, dcc1Δ and ctf8Δ diploids have a low viability (Fig. 2).

The identification of CHL1 as a high-copy-number suppressor of ctf8Δ spore lethality (see above) together with the fact that chl1Δ cells have sister-chromatid cohesion defects comparable to ctf8Δ cells during mitosis led us to investigate meiosis in chl1Δ diploids. We found that deletion of CHL1 caused massive chromosome mis-segregation, with a high frequency of meiosis II non-disjunction and low spore viability (Fig. 2). Deletion of CTF4 had a very similar effect (Fig. 2).

The severe meiosis II chromosome mis-segregation observed in ctf8Δ, chl1Δ and ctf4Δ strains cannot be a consequence of a failure to repair DNA double-strand breaks induced during meiotic recombination because deletion of the recombination-initiating endonuclease SPO11 (Keeney et al., 1997) did not alter the meiosis II non-disjunction (data not shown). We conclude that, in contrast to the situation in mitosis, Chl1, Ctf4 and RF-C^{Ctf18/Dcc1/Ctf8} are essential for chromosome segregation during meiosis II and for the production of viable gametes.

To address whether chromosome mis-segregation during meiosis in ctf8Δ and chl1Δ cells might be caused by a loss of sister-chromatid cohesion, we used in situ immunofluorescence to analyse diploids in which one copy of chromosome V was tagged with GFP (heterozygous URA3-GFP) and that expressed securin tagged with Myc18 (Pds1-Myc18) and Rec8 tagged with HA3 (Rec8-HA3) (Fig. 3). chl1Δ cells proceeded through meiosis with kinetics similar to the wild type. Premeiotic DNA replication, accumulation of securin and expression of Rec8 in chl1Δ cells were very similar to wild-type cells (data not shown). ctf8Δ cells proceeded through meiosis more slowly and less efficiently but, nevertheless, half of them underwent both meiotic divisions (data not shown). In the wild type, centromere-linked URA3 sequences marked by GFP remain closely associated until the onset of anaphase II. They invariably move to the same pole at anaphase I, remain bound together during metaphase II and move to opposite poles only at anaphase II, ending up in separate spores (Fig. 3A,B). Precocious separation of sister URA3 sequences occurred in a large proportion of ctf8Δ and chl1Δ cells. During metaphase I, sisters were split in 15% of ctf8Δ and 36% of chl1Δ cells (Fig. 3C). At this stage, most split GFP signals were found close to one of the two spindle poles. Despite the precocious splitting, most sister URA3 sequences segregated to the same pole in ctf8Δ (93%) and chl1Δ (88%) anaphase I cells. Thus, URA3-GFP signals were segregated to opposite poles in only 7% of ctf8Δ and 12% of chl1Δ cells. When sister URA3 sequences had co-segregated at the first division, the URA3-GFP signals were found as two distinct spots in 10% of ctf8Δ and 28% of chl1Δ anaphase I cells (Fig. 3C), a phenotype never observed in the wild type. In metaphase II, sister URA3-GFP loci were split along the same spindle axis in 56% of ctf8Δ and in 50% of chl1Δ cells in which sister centromeres had segregated to the same pole at meiosis I (Fig. 3C). During the ensuing anaphase II, sister chromatids mis-segregated to the same pole in 30% of ctf8Δ and 31% of chl1Δ cells (Fig. 3C), and, as a consequence, were found in a single spore of tetrads (Fig. 3D). These frequencies of meiosis II non-disjunction are consistent with the mis-segregation of heterozygous URA3-GFP observed in tetrads (Fig. 2).

The above findings demonstrate that deletion of CTF8 and CHL1 leads to defects in sister-chromatid cohesion throughout meiosis, but this appears to be accompanied by massive chromosome mis-segregation only during meiosis II, when cohesion is confined to centromeric regions. As might be expected of cells with reduced cohesion between sister centromeres, metaphase II spindles were elongated and nuclei stretched in 52% of ctf8Δ and 64% of chl1Δ cells that had not yet destroyed securin (Fig. 3A,C). In some cases, chromatids had already segregated into two DNA masses despite the presence of securin (Fig. 6D). Cohesion between sister chromatids cannot be entirely absent in ctf8Δ and chl1Δ cells during meiosis II because we detected only 30% and not 50% sister-chromatid non-disjunction at anaphase II. Nevertheless, up to 60% of chromatids segregated at random during meiosis II in the mutant cells. Our results not only confirm that RF-C^{Ctf18/Dcc1/Ctf8} is involved in sister-chromatid cohesion but also demonstrate that this complex and the Chl1 helicase have a crucial role in this process during meiosis.

To test whether removal of CTF8 or CHL1 affects the segregation of homologous chromosomes to opposite poles during meiosis I, we analysed diploid wild-type, ctf8Δ and chl1Δ strains carrying both copies of chromosome V tagged with GFP (homozygous URA3-GFP) by in situ immunofluorescence in a meiotic time-course experiment (Fig. 3E,F). Wild-type cells invariably segregated homologous chromosomes to opposite poles during anaphase I (Fig. 3E). Meiosis I homolog non-disjunction to the same pole was observed in 4% of chl1Δ and 5% of ctf8Δ cells in anaphase I. The vast majority of mutant cells (96% for chl1Δ and 95% for ctf8Δ) segregated the URA3-GFP signals to opposite poles (Fig. 3F). This finding supports our conclusion that frequent meiosis II errors rather than meiosis I homolog non-disjunction predominantly contribute to the meiotic chromosome mis-segregation phenotype observed in the mutant cells. However, the sister-chromatid cohesion defect in ctf8Δ and chl1Δ cells appears to result in a significant proportion of meiosis I mis-segregation owing to the equational segregation of sister chromatids and non-disjunction of homologous chromosomes.

In a random insertion mutagenesis screen for meiotic chromosome mis-segregation mutants in S. pombe (S. Yokobayashi and Y. Watanabe, unpublished), we identified insertions in the fission yeast orthologs of DCC1 and CTF18 that disrupted their ORFs. To analyse chromosome segregation during meiosis, we used strains in which cen1 was marked by GFP on one of the two homologs. Although there was a low but significant increase in segregation to opposite poles, sister cen1-GFP loci segregated to the same pole during meiosis I in most mutant cells as in the wild type (data not shown). As in budding yeast (Fig. 2, Fig. 3C,D), the predominant phenotype caused by the dcc1 and ctf18 insertions was meiosis II sister-chromatid non-disjunction (Fig. 4). These results suggest that RF-C^{Ctf18/Dcc1/Ctf8} might have an important role in sister-chromatid cohesion during meiosis in a wide range of eukaryotes.

Although the precocious separation of sister URA3-GFP sequences in budding yeast chl1Δ and ctf8Δ mutants is consistent with the notion that defective cohesion between sister centromeres is responsible for their high rate of meiosis II non-disjunction (Fig. 2, Fig. 3C,D), we cannot be certain merely from observing precocious separation of URA3 sequences that defective sister-centromere cohesion is responsible for the meiosis II non-disjunction. For example, similar cohesion defects during mitosis are not accompanied by high levels of non-disjunction (Fig. 1A).

What is required is a method of assaying the functionality of cohesion between sister centromeres. Mutants defective in the monopolin complex provide such an assay. Deletion of MAM1, which encodes monopolin's meiosis-specific subunit, causes the attachment of sister kinetochores to opposite poles at meiosis I. As a consequence, mam1Δ cells attempt to split and segregate sister centromeres to opposite poles following the destruction of securin and the removal of Rec8 cohesin complexes from chromosome arms in anaphase I but are prevented from doing so by the persistence of cohesion between sister centromeres (Toth et al., 2000). As a consequence, the spindle fails to elongate and nuclear division fails to take place (`abortive' anaphase I; Fig. 5C). If cohesion between sister centromeres were severely compromised in the absence of CTF8, CHL1 and CTF4 then deletion of these genes should permit mam1Δ cells to undergo anaphase I, albeit separating sister centromeres to opposite poles (anaphase I; Fig. 5F). To test this, we used in situ immunofluorescence to compare meiotic chromosome segregation in wild-type, mam1Δ, ctf8Δ, chl1Δ, ctf4Δ and double mutant combinations with mam1Δ (Fig. 5). All strains expressed a tagged version of securin (Pds1-Myc18) and had one of the two homologs of chromosome V marked by GFP (heterozygous URA3-GFP).

We selected meiosis I cells with bipolar spindles that had clearly degraded securin and scored whether they had segregated their chromosomes and extended their spindles (Fig. 5G). Whereas wild-type, ctf8Δ, chl1Δ and ctf4Δ cells readily elongated their spindles and segregated their chromosomes after securin destruction in meiosis I (Fig. 5G), the vast majority of mam1Δ cells failed to segregate chromosomes and accumulated as mononucleated cells with short metaphase-I-like spindles (`abortive' anaphase; Fig. 5A-C,G). Deletion of CTF8, CHL1 or CTF4 almost completely restored meiosis I chromosome segregation to mam1Δ cells (Fig. 5D-G). The vast majority of mam1Δ ctf8Δ, mam1Δ chl1Δ and mam1Δ ctf4Δ double mutant cells that had destroyed securin possessed symmetrical DNA masses at either end of an elongated anaphase I spindle (anaphase; Fig. 5D-G). Whereas, in wild-type and single ctf8Δ, chl1Δ or ctf4Δ mutants, most sister URA3-GFP sequences segregated to the same pole (Fig. 3A,C), they segregated to opposite poles in 84% of mam1Δ ctf8Δ (Fig. 5E), in 86% of mam1Δ chl1Δ and in 82% of mam1Δ ctf4Δ double mutant cells (n=100). These observations imply that the cohesion between sister centromeres responsible for blocking the attempted equational division at meiosis I in mam1Δ mutants is largely dependent on CTF8, CHL1 and CTF4. The corollary is that these proteins are crucially important for creating the cohesion between sister centromeres needed for meiosis II chromosome segregation as already suggested by the high incidence of meiosis II non-disjunction in mutant cells (Fig. 2, Fig. 3C,D). Our data do not imply that CTF8, CHL1 and CTF4 are more important for cohesion around centromeres than along chromosome arms. The exquisite sensitivity of meiosis II to the elimination of these proteins more probably stems from the fact that sister-chromatid cohesion is confined to a much smaller chromosomal region at meiosis II than at mitosis or meiosis I.

The restoration of chromosome segregation to mam1Δ mutants could be due either to a failure of ctf8Δ, chl1Δ or ctf4Δ mutants to generate cohesion at premeiotic S phase or to a failure to protect centromeric Rec8 from separase at the onset of anaphase I. For example, Scc1 expressed ectopically instead of Rec8 during meiosis is not protected from separase cleavage at the onset of anaphase I. It is entirely removed from chromosomes at this stage, and this is sufficient to rescue the mam1Δ nuclear division block (Toth et al., 2000). To address whether Rec8 is first loaded onto chromosomes normally and subsequently protected from separase in the vicinity of centromeres, we used chromosome spreads and in situ immunofluorescence to analyse the distribution of Rec8 (tagged with three HA epitopes) during meiosis in wild-type, ctf8Δ and chl1Δ mutant cells.

During pachytene in meiotic prophase, Rec8-HA3 was associated along the entire length of synapsed chromosomes in wild-type as well as in ctf8Δ and chl1Δ cells (Fig. 6A,B, data not shown). Deletion of CTF8 and CHL1 had little or no effect on chromosome synapsis and condensation at that stage (data not shown). Following cleavage of chromosome arm Rec8 at anaphase I, centromeric Rec8-HA3 staining was observed in the center of the metaphase II spindles in 100% of wild-type cells (Fig. 6C). The fraction of ctf8Δ and chl1Δ cells staining positive for Rec8-HA3 in metaphase II was very similar, namely 92% and 94%, respectively (Fig. 6D). The metaphase II spindles in ctf8Δ and chl1Δ mutants were much longer than those in the wild type, and centromeric Rec8-HA3 was distributed along the spindle axes in mutant cells indicating that kinetochores had attached to microtubules but had not properly bioriented (Fig. 6D). These data imply that CTF8 and CHL1 are neither required for the stable association of Rec8 with chromosomes nor for the protection of centromeric Rec8 from separase cleavage at the onset of anaphase I. The meiosis II non-disjunction observed in the mutants cannot therefore be caused merely by an absence of centromeric cohesin but rather by a failure of cohesin complexes to generate cohesion between sister DNA molecules.

The budding yeast DNA polymerase σ, Trf4, has been reported to be involved in sister-chromatid cohesion during mitosis (Wang et al., 2000) and proposed to be loaded onto DNA by RF-C^{Ctf18/Dcc1/Ctf8} to regulate cohesion (Carson and Christman, 2001). If this hypothesis were correct and the major function of RF-C^{Ctf18/Dcc1/Ctf8} in mediating sister-chromatid cohesion is to load Trf4 onto DNA, deletion of TRF4 should result in mitotic sister-chromatid cohesion defects and severe chromosome mis-segregation during meiosis comparable to mutations in the alternative RF-C subunits. Contrary to this prediction, we found that deletion of TRF4 had a very minor if any effect on sister-chromatid cohesion in mitotic metaphase cells and on meiotic chromosome segregation (Fig. 1A, Fig. 2). Splitting of URA3-GFP signals occurred in only 1% of trf4Δ cells with a bipolar spindle and high levels of nuclear securin (Fig. 1A). Furthermore, deletion of TRF4 did not lead to the characteristic increase in metaphase cells in an asynchronous mitotic culture that we observed for mutations in CTF8 and CHL1 (data not shown). Consistent with the lack of a cohesion defect in the absence of Trf4, this result indicates that trf4Δ cells do not delay anaphase onset through the activation of the spindle checkpoint. During meiosis, only 4% and 3% of trf4Δ tetrads showed mis-segregation of homozygous and heterozygous URA3-GFP, respectively (compared with 54% and ∼30% or more in alternative RF-C mutants) and the vast majority of trf4Δ tetrads displayed wild-type segregation (Fig. 2). Therefore, our data in both mitosis and meiosis do not support the notion that the alternative RF-C^{Ctf18/Dcc1/Ctf8} loads the DNA polymerase σ, Trf4, onto DNA and thereby facilitates the generation of sister-chromatid cohesion.

The isolation of mutants with defective chromosome segregation has been a rich source of proteins involved directly or indirectly in sister-chromatid cohesion (Guacci et al., 1993; Kouprina et al., 1993; Michaelis et al., 1997; Spencer et al., 1990; Toth et al., 1999). However, the defective sister-chromatid cohesion of some mutants could be due to secondary consequences of inactivating physiological processes that have little direct bearing on this process. Of all the proteins required for sister-chromatid cohesion, the cohesin complex is the one most directly implicated in holding sister chromatids together. Cohesin is found along interchromatid axes and, crucially, cleavage of cohesin's Scc1 or Rec8 subunit appears to trigger the onset of anaphase in mitotic and meiotic cells (for a review, see Petronczki et al., 2003). No other cohesion protein has been implicated in the actual process of cohesion in such a direct manner, but some have been found either to be associated with cohesin or to be required for its association with chromosomes. Thus, a sizeable proportion of the Scc2-Scc4 complex has been found to be bound to cohesin and to be essential for its association with chromosomes (Arumugam et al., 2003; Ciosk et al., 2000). Likewise, the Caenorhabditis elegans TIM1 protein, which is a homolog of the protein encoded by the Drosophila and mammalian clock gene TIMELESS, is found associated with cohesin and is required for the stable association with meiotic chromosomes of cohesin's Rec8 subunit (Chan et al., 2003). We noted that the budding yeast TOF1 gene is the TIMELESS ortholog of Saccharomyces cerevisiae. Consistent with very recent studies (Mayer et al., 2004; Warren et al., 2004), we find that deletion of TOF1 results in a modest defect in mitotic sister-chromatid cohesion and meiotic chromosome segregation (B.C. and M.P., unpublished).

Most cohesion proteins have, by contrast, not yet been directly implicated in cohesin function. Some, like Eco1/Ctf7, are essential proteins that have potentially indispensable functions in sister-chromatid cohesion (Skibbens et al., 1999; Toth et al., 1999), whereas others, like Ctf4 and RF-C^{Ctf18/Dcc1/Ctf8}, are not essential for mitosis (Hanna et al., 2001; Mayer et al., 2001). Sister-chromatid cohesion is compromised but not eliminated in their absence. Another problem with the interpretation of the cohesion defects in this class of mutants is that measurements have largely been confined to mutant cells induced to arrest for prolonged periods in a metaphase-like state by incubation in spindle poisons.

It is therefore desirable to apply other more physiological assays to assess sister-chromatid cohesion in these types of mutant strain and, in particular, to find circumstances in which their cohesion defects have clear biological consequences. In this study, we describe three new assays to evaluate the roles of Ctf4 and RF-C^{Ctf18/Dcc1/Ctf8}. We show that both are required for efficient sister-chromatid cohesion in unperturbed mitotic cells and that both are essential for orderly sister-chromatid disjunction during meiosis II and for the cohesion between sister centromeres that prevents the first meiotic division in monopolin mutants. Our work demonstrates that meiosis is particularly suitable for detecting cohesion defects because, during meiosis II, sister-chromatid cohesion is confined to the vicinity of centromeres. We suggest that cohesion defects have a disproportionate effect at this stage because the reduced cohesion cannot be compensated for by cohesion along chromosome arms.

Sister chromatids segregate to the same pole at meiosis II in 30% of ctf8Δ mutant cells, which suggests that segregation might be random for 60% of chromosomes, a level of non-disjunction that causes 95% of spores to be inviable. This enabled us to screen for genes that can, when overexpressed, partially suppress the inviability of spores derived from ctf8Δ diploids. Our aim was to isolate genes that either facilitated the maintenance of sister-centromere cohesion or helped to improve the establishment of cohesion. By this means, we identified a gene called CHL1 whose overexpression increased the viability of spores derived from ctf8Δ diploids. CHL1 had previously been identified in two independent screens for mitotic chromosome transmission fidelity mutants (Kouprina et al., 1993; Spencer et al., 1990). We showed that CHL1 has a function in sister-chromatid cohesion that is of similar importance to those of Ctf4 and RF-C^{Ctf18/Dcc1/Ctf8}: a non-essential role in establishing cohesion during mitosis but a crucial role in avoiding a disastrous level of sister-chromatid non-disjunction during meiosis II. Our findings are consistent with the very recent finding that chl1Δ mutations cause sister-chromatid cohesion defects at centromeres and chromosome arms in nocodazole-arrested cells (Mayer et al., 2004). The previous observation that chl1Δ cells mis-segregate small artificial chromosomes during mitosis at a much higher frequency than larger ones (Holloway, 2000) supports the notion that meiosis II is more sensitive to cohesion defects than mitosis because this division relies on sister-chromatid cohesion within a more limited chromosomal interval.

Our finding that deletion of CTF8, CHL1 and CTF4 suppresses the centromeric-cohesion-dependent nuclear division block of mam1Δ cells in meiosis I demonstrates for the first time in a functional assay that all three genes are essential for centromeric sister-chromatid cohesion. Despite the strong cohesion defects and high incidence of non-disjunction in meiosis II, ctf8Δ and chl1Δ cells retain association of the meiotic cohesin subunit Rec8 with centromeres during meiosis II. The mutants also appear to have little or no defect in the loading of Rec8 onto chromosome arms before or during pre-meiotic DNA replication. This indicates that these genes are required for neither the association of cohesin with centromeres nor the protection of centromeric Rec8 from cleavage by separase in meiosis I.

RF-C^{Ctf18/Dcc1/Ctf8}, Chl1 and Ctf4 are all evolutionarily conserved proteins, and orthologs can be found in most eukaryotic genomes (Amann et al., 1997; Hanna et al., 2001; Mayer et al., 2001; Merkle et al., 2003; Shiratori et al., 1999; Williams and McIntosh, 2002). We show here that mutations in fission yeast orthologs of CTF18 and DCC1 cause meiotic chromosome mis-segregation phenotypes similar to those observed in budding yeast mutants. Therefore, the role of RF-C^{Ctf18/Dcc1/Ctf8} in regulating sister-chromatid cohesion is presumably an evolutionarily conserved feature of this protein complex. Data on human orthologs are consistent with this notion. The CTF18, DCC1 and CTF8 proteins from humans interact with each other and with canonical RF-C subunits (Merkle et al., 2003). In addition, human RF-C^{Ctf18/Dcc1/Ctf8} is found in the nucleus and hCTF18 appears to bind chromatin preferentially during S phase (Merkle et al., 2003).

CHL1 encodes a conserved helicase belonging to the DEAH family (Shiratori et al., 1999). The human homolog hCHLR1 has DNA-dependent ATPase and DNA-helicase activities (Hirota and Lahti, 2000). Mutations in budding yeast CHL1 predicted to inhibit the protein's enzymatic function abolish its ability to support accurate chromosome segregation (Holloway, 2000), suggesting that Chl1's helicase activity is important for the establishment of efficient sister-chromatid cohesion. hCHLR1 accumulates within nuclei in proliferating cells upon their entry into S phase but not in quiescent cells (Amann et al., 1997), which is consistent with a role in establishing sister-chromatid cohesion. It is conceivable that mutations in human RF-C^{Ctf18/Dcc1/Ctf8} or hCHLR1 could lead to chromosomal instability in mitotic cells, a hallmark of most solid cancers (Jallepalli and Lengauer, 2001), or to infertility owing to the production of aneuploid gametes during meiosis.

The underlying molecular functions of RF-C^{Ctf18/Dcc1/Ctf8}, the Chl1 helicase and Ctf4 in regulating sister-chromatid cohesion remain unclear. All three factors are either similar to or interact with replication-related proteins, suggesting a function in the establishment of cohesion at the replication fork during S phase. We provide evidence that Chl1 is indeed required during this period of the cell cycle. Whether Ctf4 and RF-C^{Ctf18/Dcc1/Ctf8} also act during S phase has not been addressed. It is nevertheless worth noting that mutations that eliminate RF-C^{Ctf18/Dcc1/Ctf8}, Chl1 or Ctf4 do not greatly if at all affect replication of bulk DNA during S phase (Hanna et al., 2001; Mayer et al., 2001) (this study). We used affinity purification and mass spectrometry to investigate potential protein partners, and confirmed the composition of RF-C^{Ctf18/Dcc1/Ctf8} but failed to find any stable partners of either Chl1 or Ctf4 (B.C. and M.P., unpublished). It therefore remains unclear whether Ctf4 really does interact with DNA polymerase α in vivo.

Neither we nor others have been able to detect any stable physical association between RF-C^{Ctf18/Dcc1/Ctf8} and cohesin (Mayer et al., 2001; Merkle et al., 2003). Human RF-C^{Ctf18/Dcc1/Ctf8}, by contrast, has been shown to interact with PCNA (Merkle et al., 2003; Ohta et al., 2002) and is capable of loading this polymerase clamp onto DNA in vitro, albeit with a lower efficiency than the canonical RF-C^Rfc1 (Bermudez et al., 2003). It has been suggested that RF-C^{Ctf18/Dcc1/Ctf8} promotes efficient cohesion by loading a specialized DNA polymerase onto cohesin-rich regions of DNA and coordinates their replication with the establishment of cohesion (Carson and Christman, 2001). One candidate for such an enzyme is the putative DNA polymerase σ Trf4, because trf4 mutants have been reported to be defective in sister-chromatid cohesion in nocodazole-arrested cells (Wang et al., 2000). Our observations that deletion of TRF4 had little or no effect on sister-chromatid cohesion in an unperturbed cell cycle or on chromosome segregation during meiosis do not support this model. In addition, our TRF4 results stress the importance of analysing cohesion phenotypes under more physiological conditions rather than in a prolonged mitotic arrest. Although our data strongly question whether Trf4 plays a role in sister-chromatid cohesion, it is still possible that defects in trf4Δ cells are masked by redundancy between TRF4 and its paralog TRF5 (Edwards et al., 2003; Wang et al., 2000). Nevertheless, the main function of RF-C^{Ctf18/Dcc1/Ctf8} in mediating sister-chromatid cohesion is unlikely to be only the recruitment of Trf4 via PCNA. PCNA can act as a platform for the interaction with many different proteins in addition to serving as a sliding clamp for replicative DNA polymerases (Maga and Hubscher, 2003). It remains to be determined whether RF-C^{Ctf18/Dcc1/Ctf8} loads PCNA or a related clamp onto DNA in vivo and, if so, which factors are subsequently recruited by the clamp.

In conclusion, our work now leaves little doubt that RF-C^{Ctf18/Dcc1/Ctf8}, the helicase Chl1 and Ctf4 have crucial roles in sister-chromatid cohesion. Future studies will have to focus on the function of these proteins at the replication fork and how they influence the generation of cohesive structures that bind sister chromatids together.

We thank A. Schleiffer for bioinformatic analysis, K. Mechtler and I. Steinmacher for mass spectrometry, and the Nasmyth lab for helpful discussions and advice. Work in the Nasmyth lab is supported by Boehringer Ingelheim International and by the Austrian Industrial Research Promotion Fund (FFF).