Genes-encoding members of the Ctf18 replication factor C (RFC)-like complex were among the first nonessential genes reported to have roles in sister-chromatid cohesion. Ctf18, Dcc1, and Ctf8 physically interact with the Rfc2-Rfc3-Rfc4-Rfc5 core, but do not interact with the large subunit of the canonical RFC, Rfc1 (Hanna et al. 2001; Mayer et al. 2001; Naiki et al. 2001). Removal of any of the nonessential Ctf18-RFC components results in precocious dissolution of cohesion, and double and triple deletions fail to display any additive effects, suggesting that the three subunits work together as a complex to promote cohesion (Mayer et al. 2001). Ctf18-RFC catalyzes both loading and unloading of the sliding clamp PCNA around DNA in vitro (Bermudez et al. 2003; Shiomi et al. 2004; Bylund and Burgers 2005) and deletion of CTF18 results in a reduction in the levels of PCNA at arrested replication forks in early S phase (Lengronne et al. 2006). Ctf18-RFC is also involved in S-phase checkpoint activation. Although ctf18Δ has no detectable checkpoint defect, the double mutants ctf18Δ rad24Δ and ctf18Δ rad9Δ are checkpoint defective (Naiki et al. 2001; Bellaoui et al. 2003; Pan et al. 2006). In addition, mutants in genes encoding Ctf18 complex members are sensitive to the DNA-damaging agents camptothecin and methyl methanesulfonate (MMS) and the replication inhibitor HU (Naiki et al. 2001; Chang et al. 2002; Bellaoui et al. 2003; Kanellis et al. 2003; Parsons et al. 2004), further suggesting a role in the DNA damage response.

In addition to its role in sister-chromatid cohesion (Hanna et al. 2001), Ctf4 has close links to DNA replication. It binds to DNA polymerase α with high affinity (Miles and Formosa 1992; Zhou and Wang 2004) and is found in a high-molecular-weight complex with the GINS and MCM complexes, Cdc45, Mrc1, Tof1, and Csm3, and the histone chaperone FACT (Gambus et al. 2006; Lengronne et al. 2006). Recent data indicate that Ctf4 is a component of elongating replication forks (Lengronne et al. 2006). Although not essential in budding yeast, its homolog in fission yeast, mcl1⁺, is essential for viability, and overexpression of mcl1⁺ causes an S-phase delay (Williams and Mcintosh 2002). The ctf4 deletion mutant displays genetic interactions with mutants in DNA replication genes, DNA damage checkpoint response genes, spindle assembly checkpoint genes, and genes involved in sister-chromatid cohesion, microtubule function, and chromosome structure (Tong et al. 2004), further suggesting that Ctf4 functions in diverse chromosome metabolism pathways.

Like Ctf4, the nonessential cohesion genes Mrc1, Tof1, and Csm3 also have intimate connections to DNA replication. All three proteins are present at replication forks during normal S-phase progression (Katou et al. 2003; Osborn and Elledge 2003; Gambus et al. 2006). Mrc1 and Tof1 are thought to be important in stabilizing stalled replication forks, as deletion of MRC1 or TOF1 leads to the uncoupling of replication fork proteins from sites of DNA synthesis when replication is arrested by hydroxyurea (Katou et al. 2003). Although Mrc1, Tof1, and Csm3 have common cellular roles, there are important differences between Mrc1 and Tof1/Csm3. Deletion of MRC1 has a more dramatic effect on activation of the replication checkpoint than deletion of TOF1, whereas Tof1 and Csm3, but not Mrc1, are critical for replication fork pausing at replication fork barriers (Calzada et al. 2005; Tourriere et al. 2005).

Chl1 is involved in the establishment, but not the maintenance, stage of sister-chromatid cohesion (Petronczki et al. 2004). It is a putative DNA helicase, having an essential ATP-binding site, and is localized to the nucleus (Skibbens 2004). In addition to being required for the establishment of sister-chromatid cohesion (Petronczki et al. 2004; Skibbens 2004), Chl1 has a role in transcriptional silencing, rDNA recombination, and aging (Das and Sinha 2005). Unlike other nonessential cohesion genes (CTF4, CTF18, CTF8, and DCC1), which are required for transcriptional silencing at both HMR and telomeres (Suter et al. 2004), the absence of CHL1 results in increased silencing at HMR and, conversely, in decreased silencing at the telomere (Das and Sinha 2005). The different phenotypes of these cohesion factors in transcriptional silencing could reflect the different mechanisms by which the corresponding proteins establish cohesion and affect chromatin structure.

A number of other nonessential cohesion factors have been discovered by genetic screens: BIM1, KAR3, CHL4, RAD27, RRM3, RAD50, SRS2, SGS1, RSC2, and RAD61 (Mayer et al. 2004; Warren et al. 2004). In addition to their roles in sister-chromatid cohesion, they all have other cellular roles. BIM1 and KAR3 encode microtubule-binding proteins involved in spindle assembly and spindle orientation (Endow et al. 1994; Page et al. 1994; Schwartz et al. 1997; Bloom 2000). The gene product of CHL4 is an outer kinetochore protein (Mythreye and Bloom 2003; Pot et al. 2003). RAD50, XRS2, and MRE11 are involved in DNA damage repair (Nakada et al. 2003) while RAD27 plays roles in both DNA replication and repair (Gary et al. 1999; McHugh et al. 2000; Boiteux and Guillet 2004; Torres et al. 2004). RRM3, SRS2, and SGS1 are all DNA helicases involved in DNA replication, repair, and recombination (Ivessa et al. 2002; Barbour and Xiao 2003; Macris and Sung 2005). RAD61 is involved in resistance to ionizing radiation (Game et al. 2003) and RSC2 encodes a nonessential subunit of the RSC chromatin-remodeling complex (Baetz et al. 2004). These connections between sister-chromatid cohesion and diverse cellular functions suggest that sister-chromatid cohesion impinges on multiple cellular processes and that diverse functions are important for sister-chromatid cohesion.

Although the list of nonessential genes that are involved in sister-chromatid cohesion continues to grow, little is known about how these genes contribute to sister-chromatid cohesion. To understand the cohesion role of nonessential cohesion genes, and their relationship to each other, we created a series of double mutants and performed cohesion assays on these double mutants. Taking advantage of the data from large-scale synthetic-lethal screens (Tong et al. 2004), we were able to identify genetic backgrounds in which normally nonessential cohesion genes are required for viability or fitness and to thereby examine the cohesion defect of targeted double-mutant cells. The results of this analysis support a model in which at least two parallel pathways, one containing Tof1, Csm3, Ctf4, and Chl1 and the other containing Mrc1 and Ctf18-RFC, contribute to the establishment of sister-chromatid cohesion.

Strains: Strains used in this study are listed in supplemental Table S1 at http://www.genetics.org/supplemental/. Standard yeast media and growth conditions were used (Sherman 1991).

Assessing sister-chromatid cohesion: Strains used for cohesion assays are in the W303 background. The wild-type strain was YPH1477 (MATa ade2-1 trp1-1 can1-100 his3-11, 15 leu2LEU2tetR-GFP ura33xURA3tetO112 PDS1-13myc-TRP1), which contains the tet operator array inserted 35 kb from CEN5. Mutants used in cohesion assays were all created in YPH1477.

For cohesion assays, strains were grown to midexponential phase in YPD, collected by centrifugation, resuspended in pH 3.9 YPD medium containing 2 μg/ml α-factor and grown for 2–3 hr at 30°. The G₁ block was monitored by the formation of shmoos. Cells were then collected by centrifugation and washed with normal-pH YPD medium once before releasing into YPD medium containing 15 μg/ml nocodazole for 1.5 hr at 30° (or other temperatures as indicated). Cell pellets were collected by centrifugation and fixed in 4% paraformaldehyde for 5 min, washed once with SK buffer (1 m sorbitol, 0.05 m K₂PO₄), and resuspended in SK for cohesion assessment. Cells were briefly sonicated prior to microscopic examination. One GFP spot per cell indicates closely linked sister chromatids. Two GFP spots within one cell in the presence of nocodazole indicates premature separation of sister chromatids. For temperature-sensitive (ts) alleles, the G₁ block was carried out at 26° for 3 hr, followed by release into YPD + nocodazole at 37° for 1.5 hr before fixation.

To quantify sister-chromatid cohesion, the percentage of cells with two GFP dots from at least 200 cells was calculated. Each experiment was repeated two to three times with different transformants for each genotype, and the average and standard deviation was plotted. To rule out effects of aneuploidy, G₁ cells (blocked in α-factor) were tested in every cohesion assay. Statistical analysis (t-test) was performed for some cohesion assays to identify significant differences in sister-chromatid separation.

Pds1 assay: Cells were arrested in G₁ with α-factor, released in the presence of nocodazole, and fixed with paraformaldehyde after 90 min. GFP and PDS1-13myc were detected by indirect immunofluorescence in nocodazole-arrested, paraformaldehyde-fixed cells using anti-GFP (Abcam) and anti-myc (9E10; Roche) antibodies, respectively, as described (Mayer et al. 2004).

Immunoprecipitation and immunoblotting: Immunoprecipitation was performed essentially as described (Naiki et al. 2001). Briefly, 25 OD₆₀₀ of midlog phase cells were lysed in 50 mm HEPES, pH 7.5, 100 mm NaCl, 10% glycerol, 1 mm EDTA, 1 mm DTT containing protease inhibitors (5 μg/ml leupeptin, 2 μg/ml pepstatin A, 1 mm PMSF, 5 μg/ml 1-chloro-3-tosylamido-7-amino-2-heptanone, 2.5 μg/ml aprotinin, 1 mm DTT). Equal amounts of cell lysate were immunoprecipitated using anti-myc antibody (9E10; Roche) followed by incubation with protein G agarose. Ten percent of the input extract and 50% of the immunoprecipitate were fractionated by 7.5% SDS–PAGE and tagged proteins were detected on Western blots with anti-myc antibodies or peroxidase–antiperoxidase (Sigma, St. Louis). Immunoblots were developed using Supersignal ECL (Pierce, Rockford, IL). Protein samples for Rad53 activation assays were prepared by trichloroacetic acid fixation and analyzed on immunoblots as described (Bellaoui et al. 2003).

Synchronization and flow cytometry: Cells were arrested in G₁ with 2 μg/ml α-factor for 2–3 hr at 30° (or 26° for ts strains) in pH 3.9 YPD. Cells were released into the cell cycle by harvesting, washing, and resuspending in normal pH YPD. Flow cytometry was performed as described (Chang et al. 2002).

Creating temperature-sensitive mutants: Temperature-sensitive mutants were generated by random PCR mutagenesis with the PCR random mutagenesis kit from BD Biosciences. CSM3 was amplified with the primers 5′-TGATATACTGGATTAAAATGCCATGAAAACGTGAACAGAAACTTTTATTGAGGTC3′ and 5′-GGGACGAGGCAAGCTAAACAGATCTCTGCCGGTGCTGATAATACGG-3′. The URA3 cassette was amplified by regular PCR with the primers 5′-TAGATGCCCACACGCACGTTTGGATTATTACCTTCAATGACATTGGAATTCGAGCTCGTTTAAACTGGA-3′ and 5′-AGATCTGTTTAGCTTGCCTCGT-3′. Homologous sequences were designed in the CSM3 and URA3 PCR primers to ensure that these two PCR products could recombine by in vivo homologous recombination in yeast. Mutagenized CSM3 and regular amplified URA3 were cotransformed into yeast strain Y4681 (MATα ctf18ΔnatR can1 ΔMFA1pr-HIS3-MFα1pr-LEU2 his3Δ0 leu2Δ0 ura3Δ0 lys2Δ0) and selected on SD–URA plates. Temperature-sensitive transformants were screened by replica plating to 37° and further confirmed by PCR and tetrad analysis.

The CTF8 ts allele was created with the same method as CSM3 in the strain Y4413 (tof1 ΔnatR can1 ΔMFA1pr-HIS3-MFα1pr-LEU2 his3 Δ0 leu2 Δ0 ura3Δ0 lys2Δ0) with the primers 5′-AGTGATAGAAAAAAGAATTATCACTATCATTCAGCCCAATAAACAGCTGAAAAGA-3′ and 5′-GGGACGAGGCAAGCTAAACAGATCTACAACTCCGAACAATAACTAAGTAC-3′ for PCR mutagenesis. The URA3 cassette was amplified by regular PCR with the primers 5′-ACACTTTACACAGAGCGTGAAGTCTGCGCCAAATAACATAAACAAGAATTCGAGCTCGTTTAAACTGGA-3′ and 5′-AGATCTGTTTAGCTTGCCTCGT-3′. When ts alleles were moved between different strain backgrounds, transformants with the ts phenotype were selected and further confirmed by PCR and sequence analysis.

Physical interactions among Mrc1, Tof1, and Csm3: Mrc1, Tof1, and Csm3 share common roles in DNA replication, S-phase checkpoint, and sister-chromatid cohesion. Genetic interactions with SRS2 suggest that these three genes may also be important for homologous recombination (Xu et al. 2004). Physical interaction between Tof1 and Csm3 has been described (Mayer et al. 2004), suggesting that these proteins function in a complex. To explore a physical link between Mrc1 and Tof1/Csm3, we tagged the C terminus of Mrc1p with 13 myc epitopes in a Csm3-TAP-tagged strain. Immunoprecipitation of Mrc1-13myc specifically co-immunoprecipitated Csm3-TAP from extracts prepared from Mrc1-13myc Csm3-TAP cells (Figure 1A). The association of Mrc1 with Csm3 was also observed when the immunoprecipitation was carried out with rabbit IgG to pull down Csm3-TAP (data not shown). These results are in agreement with mass spectrometric analysis of affinity-purified Csm3 and Mrc1 (Nedelcheva et al. 2005), which also identified a physical interaction between Mrc1 and Csm3. We next tested whether the interaction between Csm3 and Tof1 depends on Mrc1. Immunoprecipitates of Csm3-13myc from an mrc1Δ strain contained Tof1-TAP (Figure 1B), indicating that the interaction between Csm3 and Tof1 is not mediated by Mrc1. Thus, Mrc1, Csm3, and Tof1 all share physical interactions, and although Csm3 interacts with Mrc1, this interaction is not required for Csm3 and Tof1 to interact.

Figure 1.— Mrc1 physically interacts with Csm3. (A) Extracts from yeast strains expressing the indicated epitope-tagged proteins were immunoprecipitated with anti-myc antibodies followed by protein G agarose. Ten percent of the input extract and 50% of the immunoprecipitate (more ...)

Mrc1 promotes sister-chromatid cohesion in a pathway that is parallel to Tof1 and Csm3: Despite the physical interactions among Mrc1, Tof1, and Csm3 and the common roles shared by these three proteins, synthetic genetic data suggested that they function in parallel pathways: there are synthetic-sick interactions between mrc1Δ and csm3Δ and between mrc1Δ and tof1Δ, suggesting that they may function distinctly, but not between tof1Δ and csm3Δ, further suggesting that these genes may function within the same pathway (Tong et al. 2004). These genetic interactions could reflect redundancy between MRC1 and TOF1/CSM3 in the establishment of sister-chromatid cohesion. To test this possibility, all pairwise double mutants were created with mrc1Δ, tof1Δ, and csm3Δ and the strains were assayed for defects in sister-chromatid cohesion. To perform the cohesion analysis, single and double deletions were constructed in YPH1477, which contains a Tet repressor-GFP fusion as well as Tet operator repeats located 35 kb from the centromere of chromosome V. Strains containing these deletions were grown logarithmically, arrested in G₁ for 2–3 hr, and then released into nocodazole for 1.5 hr. Cells were then fixed and the number of GFP foci per cell was scored (Figure 2, A and B). In the wild-type control, two GFP foci were evident in ~5% of cells. Deletion of MRC1, TOF1, or CSM3 caused mild cohesion defects with ~20% of cells exhibiting two GFP foci (Figure 2B). This level of precocious sister-chromatid separation was similar to previous reports (Mayer et al. 2004; Xu et al. 2004). Both mrc1Δ tof1Δ and mrc1Δ csm3Δ double mutants exhibited an additive cohesion defect relative to the relevant single mutants (Figure 2B), with ~40% of cells displaying two GFP foci. By contrast, the tof1Δ csm3Δ double mutant had only 20% of cells with two GFP foci, a level similar to that seen in the single mutants. Although high levels of cohesion loss were scored for mrc1Δ tof1Δ and mrc1Δ csm3Δ, these double mutants had not entered anaphase, as evidenced by the high percentage of cells (>85%) that displayed Pds1 staining, as assessed by indirect immunofluorescence (Figure 2C). Moreover, we found that >90% of the cells with separated sister chromatids had high levels of Pds1, indicating that these cells had not initiated anaphase (Figure 2C).

Figure 2.— mrc1Δ tof1Δ and mrc1Δ csm3Δ exhibit additive cohesion defects relative to single mutants. (A) Micrographs of mrc1Δ tof1Δ in which one sister-chromatid locus was visualized with TetR-GFP. (Top) DIC image. (more ...)

To examine the effect of nocodazole in the cohesion assays, the timing of sister-chromatid separation in wild-type and mutant cells was scored after first arresting cells in G₁ phase with α-factor and then releasing them into rich medium without nocadazole. In mrc1Δ tof1Δ and mrc1Δ csm3Δ mutant strains, cells with two GFP dots accumulated 40 min after α-factor release, which was at least 20 min earlier than wild type (Figure 2, D–F). Additionally, Pds1 degradation in the double mutants was delayed compared to wild-type cells. Therefore, in mrc1Δ tof1Δ and mrc1Δcsm3Δ, a high percentage of cells displayed separated sister chromatids in the presence of Pds1 (i.e., prior to the initiation of anaphase). Thus, the high level of precocious separation of sister chromatids that we observed in the mrc1Δ tof1Δ and mrc1Δ csm3Δ double mutants was not dependent on the use of a nocodazole arrest.

In addition to promoting sister-chromatid cohesion, MRC1 and TOF1 also play roles in DNA replication and in S-phase checkpoint activation. We tested mrc1Δ tof1Δ double mutants for defects in S-phase progression and checkpoint activation (Figure 3). Flow cytometric analysis of DNA contents during progression through S phase indicated that mrc1Δ tof1Δ completed DNA replication with kinetics that were almost identical to tof1Δ and slightly accelerated relative to mrc1Δ (Figure 3A). In all strains, DNA replication was completed 60 min after release from the G₁ arrest. Although the checkpoint function of MRC1 is not required for MRC1 to promote sister-chromatid cohesion (Xu et al. 2004), we assessed S-phase checkpoint activation in the mrc1Δ tof1Δ double mutant by detecting the characteristic mobility shift of phosphorylated Rad53 on immunoblots following treatment with MMS (Figure 3B). The mrc1Δ tof1Δ double mutant displayed a Rad53 mobility shift similar to that seen in mrc1Δ. Therefore, in contrast to the additive cohesion defect seen in the double mutant, we did not observe either an additive defect in S-phase progression or a defect in S-phase checkpoint activation in mrc1Δ tof1Δ double mutants.

Figure 3.— mrc1Δ tof1Δ does not exhibit additive replication or checkpoint defects relative to the single mutants. (A) Flow cytometric analysis of DNA contents during synchronous progression through S phase. Log-phase cultures were synchronized in (more ...)

Together, these data support two conclusions: First, nonessential cohesion mutants, when combined, can result in additive cohesion defects. Second, Mrc1 and Tof1/Csm3 appear to promote sister-chromatid cohesion via distinct pathways, as deletion of MRC1 and TOF1 or of MRC1 and CSM3 resulted in an additive cohesion defect, whereas deletion of TOF1 and CSM3 did not.

Examining a temperature-sensitive conditional allele of CSM3 in ctf18Δ cells: Cohesion epistasis analysis with mrc1Δ, tof1Δ, and csm3Δ indicated that Mrc1 had a role in cohesion distinct from that of Tof1 and Csm3. This suggested that there were at least two pathways to promote sister-chromatid cohesion, one involving the Tof1/Csm3 complex and the other involving Mrc1. We next sought to understand the relationship of other nonessential cohesion genes with respect to these two putative pathways. In particular, we were interested in examining the relative contributions of Tof1/Csm3 and the Ctf18 complex to sister-chromatid cohesion.

Large-scale genetic analysis identified similarities between the genetic interaction profiles of Csm3/Tof1 and those of genes encoding components of the Ctf18 complex (Tong et al. 2004), suggesting that these two complexes function in similar cellular processes. However, double mutants carrying mutations in CSM3 or TOF1 and mutations in genes encoding members of the Ctf18 complex generally show a synthetic-lethal phenotype, implying a parallel relationship between these two complexes. Due to this synthetic lethality, we were unable to perform cohesion epistasis with deletion alleles. To examine the relative contributions of Tof1/Csm3 and the Ctf18 complex to sister-chromatid cohesion, random mutagenic PCR was used to generate a csm3 ts allele in a ctf18Δ deletion mutant. By screening ~5000 transformants, one allele (csm3-9) was isolated that displayed a ts conditional-lethal phenotype in the ctf18Δ deletion strain background. Sequence analysis revealed four mutational alterations in the csm3-9 mutant: Tyr95 changed to His, Lys105 to Thr, Ile275 to Ser, and Asp307 to Gly. To assess the cellular morphology of the ts strain csm3-9 ctf18Δ, log-phase cultures of wild-type and single and double mutants were shifted from the permissive temperature (26°) to the restrictive temperature (37°) for 4 hr. At 37°, csm3-9 ctf18Δ displayed an accumulation of large-budded cells when compared to wild type (Figure 4A). The relative fractions of unbudded cells, cells with a small bud, cells with a large bud and a single nucleus at or near the bud neck, large-budded cells with an elongated nucleus spanning the bud neck, or large-budded cells with divided nuclei were measured following fixation and DAPI staining of wild-type, single-, and double-mutant cells grown at 37° for 4 hr (Figure 4B). The csm3-9 ctf18Δ double mutant accumulated a high fraction of large-budded cells with a single nucleus close to or across the bud neck (Figure 4B), indicating a mitotic delay, most likely pre-anaphase. Flow cytometric analysis of asynchronously growing csm3-9 ctf18Δ revealed accumulation of cells with 2C DNA content even at the permissive temperature (26°) (Figure 4C).

Figure 4.— csm3-9 ctf18Δ accumulates large-budded cells at the restrictive temperature. (A) DIC images of wild type (left) and csm3-9 ctf18Δ (right) after growth at the restrictive temperature (37°) for 4 hr. (B) Nuclear morphology of wild-type, (more ...)

This mitotic delay could have arisen by activation of the DNA damage checkpoint pathway or the spindle assembly checkpoint pathway. To determine the mechanism of this mitotic delay, triple mutants were constructed in combination with either rad9Δ to inactivate the DNA damage checkpoint pathway or mad2Δ to inactivate the spindle assembly checkpoint pathway. Deletion of RAD9 had no effect on the mitotic delay of csm3-9 ctf18Δ at the restrictive temperature (Figure 4D). By contrast, there was a reduction in the fraction of large-budded cells in the csm3-9 ctf18Δ mad2Δ triple mutant (Figure 4D). These results indicated that the mitotic delay in the csm3-9 ctf18Δ double mutant requires a functional spindle assembly checkpoint.

Csm3/Tof1 and Ctf18-RFC complexes function in parallel cohesion pathways: We examined the relative contributions of Csm3 and the Ctf18-RFC complex to sister-chromatid cohesion. We transferred csm3-9 to the wild-type cohesion testing strain and to the cohesion testing strain carrying ctf18Δ. At the permissive temperature, the csm3-9 ctf18Δ double mutant displayed a level of separated sister chromatids that was similar to that seen in the ctf18Δ single mutant (Figure 5A). By contrast, at the restrictive temperature, there was a substantial increase in the frequency of separated GFP dots in csm3-9 ctf18Δ. This increase was additive when compared to the single mutants and therefore suggested that Csm3 and Ctf18 function in parallel cohesion pathways. It is worth noting that the high fraction of cells displaying separated sister chromatids (almost 60%) was similar to that seen in mutants in essential cohesion genes (Guacci et al. 1997; Michaelis et al. 1997; Toth et al. 1999; Ciosk et al. 2000). To further confirm this result, csm3-9 was transferred to a cohesion testing strain carrying ctf8Δ. As in csm3-9 ctf18Δ, there was a much higher premature separation of sister chromatids in csm3-9 ctf8Δ than in csm3-9 or ctf8Δ single mutants at the restrictive temperature (Figure 5A). Similarly, when csm3-9 dcc1Δ was constructed, this double mutant also showed an additive cohesion defect relative to the single mutants (Figure 5A). These data indicate that Csm3 and the three nonessential members of the Ctf18-RFC complex function in parallel pathways to promote sister-chromatid cohesion.

Figure 5.— (A) csm3-9 exhibits additive cohesion defects when combined with mutants in ctf8, ctf18, or dcc1. Sister-chromatid cohesion assays were performed at both the permissive (26°) and the restrictive temperature (37°) in nocodazole-arrested (more ...)

To define the relationship between Tof1 and Ctf18-RFC in promoting cohesion, we again required a conditional allele because of the synthetic-lethal genetic interactions between tof1Δ and deletion mutants of genes encoding Ctf18-RFC members (Mayer et al. 2004). As was done with Csm3, mutant forms of CTF8 were created by random PCR mutagenesis and transformed into a tof1Δ strain. Temperature-sensitive alleles were isolated by replica plating to high temperature. Using this approach, nine ctf8 mutant alleles were generated, and ctf8-17 was selected for further analysis because of its tight ts phenotype in the tof1Δ background. Sequence analysis identified four point mutations in ctf8-17: Asp7 changed to Glu, Gln62 to Arg, Arg69 to Gln, and Ile103 to Thr.

We then transferred ctf8-17 into cohesion testing strains carrying tof1Δ or csm3Δ. When cohesion analysis was performed at the restrictive temperature, both tof1Δ ctf8-17 and csm3Δ ctf8-17 showed higher levels of separated sister chromatids than did the single mutants (Figure 5B). Taken together with the results from the cohesion analysis of csm3-9 double mutants, these data indicate that the Ctf18-RFC complex functions in a cohesion pathway that is distinct from the Tof1/Csm3 complex.

Mrc1 functions in the same cohesion pathway as the Ctf18-RFC complex: Cohesion analysis with double mutants indicated that Csm3/Tof1 and the Ctf18-RFC complex were in different cohesion pathways and that Csm3 and Mrc1 were in different cohesion pathways. These data raised two possibilities: either Mrc1 acts in the same pathway as the Ctf18-RFC complex members or there are more than two nonessential pathways that promote sister-chromatid cohesion. To distinguish between these possibilities, we created an mrc1Δ ctf8-17 cohesion testing strain. Strikingly, when cohesion was analyzed at the restrictive temperature, mrc1Δ ctf8-17 exhibited a level of cohesion defect similar to that seen in the mrc1Δ single mutant (Figure 6A). This suggested that Mrc1 was in the same cohesion pathway as the Ctf18-RFC complex. To verify this result, ctf8-54, an allele of ctf8 that displayed a slow-growth phenotype, was tested in combination with mrc1Δ. As was seen with ctf8-17 mrc1Δ, ctf8-54 mrc1Δ also showed a level of separated sister chromatids that was not significantly different from that displayed by the single mutants (Figure 6A). Given that mrc1Δ displayed an additive cohesion defect with csm3Δ and tof1Δ, but not with ctf8-17, we concluded that Mrc1 and the Ctf18-RFC complex were in the same sister-chromatid cohesion pathway, in parallel with the Csm3/Tof1 cohesion pathway.

Figure 6.— CTF8 is in the same cohesion pathway as MRC1. (A) Cohesion assays were performed at the restrictive temperature (37°) following treatment with nocodazole. The relevant genotypes are indicated, and the number of GFP signals in each cell was scored (more ...)

One possibility raised by these data was that a defect in sister-chromatid cohesion could have been masked by the inability of the mrc1Δ ctf8-17 double mutant to complete S phase and produce sister-chromatid DNA. However, flow cytometric analysis of DNA contents during progression through S phase indicated that mrc1Δ ctf8-17 completed DNA replication with kinetics that were almost identical to mrc1Δ (Figure 6B). In all strains, DNA replication was completed by 50 min after α-factor release; therefore, at the time that the cohesion assay was performed (90 min after α-factor release), the sister chromatids were fully replicated. Thus, failure to complete DNA replication was not likely to be the cause of the absence of an additive cohesion defect in mrc1Δ ctf8-17.

Ctf4 and Chl1 function in the Csm3/Tof1 cohesion pathway: Our results indicated that there are two sister-chromatid cohesion pathways: one pathway mediated by Csm3 and Tof1 and another pathway containing the Ctf18-RFC complex and Mrc1. Genes encoding members of both the Csm3 complex and the Ctf18 complex display genetic interactions with additional genes involved in sister-chromatid cohesion (Mayer et al. 2001, 2004; Tong et al. 2004). We tested whether these genes define additional cohesion pathways.

In particular, CTF4 has genetic interactions with both the CSM3 and the CTF18 pathway: ctf4Δ is synthetically sick with csm3Δ and tof1Δ and synthetically lethal with ctf8Δ and mrc1Δ (Tong et al. 2004). To understand the relationship between CTF4 and the two cohesion pathways, double mutants were constructed between ctf4Δ and mutants in each pathway. The cohesion analysis demonstrated that ctf4Δ had no additive cohesion defect with csm3Δ, but did display an additive cohesion defect with ctf8-17 (Figure 7A). This result placed CTF4 in the CSM3/TOF1 cohesion pathway, in parallel with the CTF18-RFC/MRC1 pathway.

Figure 7.— Ctf4 and Chl1 are additional members of the Csm3/Tof1 cohesion pathway. (A) Ctf4 is in the same cohesion pathway as Csm3/Tof1. Cohesion assays were performed on the indicated strains, and the number of GFP signals in each cell was scored for at least (more ...)

CHL1, which encodes a DNA helicase, is another nonessential gene that promotes sister-chromatid cohesion (Mayer et al. 2004). chl1Δ is synthetically lethal with deletion mutants in genes encoding Ctf18-RFC members and is synthetically sick with mrc1Δ, but has no genetic interaction with csm3Δ or tof1Δ, implying that CHL1 may be in the same cohesion pathway as TOF1 and CSM3. To test this hypothesis, a series of double mutants were created in cohesion testing strains: chl1Δ csm3Δ, chl1Δ ctf8-17, ctf4Δ chl1Δ, and chl1Δ mrc1Δ. Cohesion analysis with these strains demonstrated that chl1Δ had no additive cohesion defect with csm3Δ or ctf4Δ, but displayed additive cohesion defects with both ctf8-17 and mrc1Δ (Figure 7B). Since the difference in cohesion loss between mrc1Δ chl1Δ and each single mutant was small, a statistical analysis (t-test) was performed. The difference in cohesion loss was statistically significant for mrc1Δ chl1Δ and chl1Δ (P < 0.05) and for mrc1Δ chl1Δ and mrc1Δ (P < 0.01). These data placed CHL1 in the CSM3 cohesion pathway, together with TOF1 and CTF4.

Nonessential cohesion pathways are required before M phase: Establishment of sister-chromatid cohesion occurs during S phase (Uhlmann and Nasmyth 1998) and must be maintained during G₂/M (Nasmyth 1999). We tested whether the function of the nonessential cohesion pathways was required at G₂/M by introducing the temperature-sensitive mutants ctf8-17 tof1Δ and csm3-9 ctf8Δ into a cohesion test strain in which Cdc20 was expressed under the control of the GAL1-10 promoter. Repressing the expression of Cdc20 by growth in glucose medium leads to the accumulation of cells at the metaphase-to-anaphase transition (Visintin et al. 1997). When cells were incubated at the permissive temperature (25°) before and after arrest, both ctf8-17 tof1Δ and csm3-9 ctf8Δ displayed a background level of cohesion loss due to the single gene deletion (Figure 8, open bars). When cells were arrested at 37°, high levels of cohesion loss were observed (Figure 8, shaded bars). By contrast, when cells were arrested at the metaphase-to-anaphase transition at the permissive temperature and then shifted to the restrictive temperature, cohesion loss remained at background levels (Figure 8, solid bars). The scc1-73 strain displayed increased cohesion loss when shifted to the restrictive temperature either before or after arrest, as would be expected for a gene that functions in both the establishment of cohesion in S phase and the maintenance of cohesion in G₂ phase. The scc2-4 strain exhibited loss of cohesion when Scc2 was inactivated before, but not after, arrest at the metaphase-to-anaphase transition, consistent with the role of Scc2 in the loading of cohesin but not in the maintenance of sister-chromatid cohesion (Ciosk et al. 2000). Since the pattern of cohesion loss for both ctf8-17 tof1Δ and csm3-9 ctf8Δ most closely resembled that of scc2-4, we conclude that the nonessential cohesion pathways described by Ctf8 and Csm3 were not required for the maintenance of cohesion during G₂/M.

Figure 8.— The nonessential cohesion genes are not required for maintenance of cohesion. The indicated strains, with Cdc20 expressed exclusively under the control of the GAL1-10 promoter, were grown in the presence of galactose (Cdc20 on) at 26° to mid-logarithmic (more ...)

In recent years, the number of nonessential cohesion genes found to have important roles in sister-chromatid cohesion has steadily increased (Stead et al. 2003; Baetz et al. 2004; Mayer et al. 2004; Warren et al. 2004; Xu et al. 2004). In contrast to essential cohesion components, mutants in nonessential cohesion genes show only mild precocious sister-chromatid separation. By analyzing double mutants, we found that nonessential cohesion mutants, when combined, lead to increased levels of sister-chromatid separation and that the levels of sister-chromatid separation in the double mutants are similar to those seen in essential cohesion gene mutants. This indicates that nonessential cohesion genes make critical and redundant contributions to sister-chromatid cohesion.

Although some double cohesion mutants had additive cohesion defects, this was not the case for every possible pairwise combination of double mutants, which indicated that there are different pathways or categories for the genes in sister-chromatid cohesion. By classifying the genes without additive cohesion defect into the same pathway and the genes with additive cohesion defect into different cohesion pathways, we found that there were two pathways by which nonessential genes promote sister-chromatid cohesion. The classification of these two pathways by cohesion analysis was largely correlated with the genetic interaction data: there were mostly synthetic-lethal interactions between members of the different pathways, while there were either no interactions or only synthetic-sick interactions between members of the same pathway (Figure 9).

Figure 9.— Cohesion analysis correlates with genetic interactions between two cohesion pathways. Results of cohesion analysis are represented by straight blue lines. Solid blue lines indicate additive cohesion defects in double mutants; dashed blue lines indicate (more ...)

The two cohesion pathways identified in this study were mediated through two complexes: the Tof1/Csm3 complex and the Ctf18-RFC complex. In addition to these two complexes, each pathway consists of additional members, with Ctf4 and Chl1 in the same pathway as the Tof1/Csm3 complex and Mrc1 in the same pathway as the Ctf18-RFC complex. One reasonable possibility is that the additional pathway members have physical interactions with the known Tof1/Csm3 or Ctf18-RFC complexes. Indeed, Ctf4 was found to be in a high-molecular-weight complex with Tof1/Csm3 (Gambus et al. 2006). However, we were unable to detect physical interactions between Tof1/Csm3 and Chl1 or between Mrc1 and Ctf18-RFC (data not shown). It is possible that physical links among components in the same cohesion pathway are indirect or transient.

Curiously, although Mrc1 was classified into a cohesion pathway parallel to Tof1/Csm3, we detected a physical interaction between Mrc1 and Csm3. Our cohesion analysis indicates that the function of Mrc1 in promoting sister-chromatid cohesion is unlikely to require complex formation with Tof1/Csm3 as Mrc1 and Tof1/Csm3 make independent contributions to cohesion. Consistent with this, analysis of Tof1/Csm3/Mrc1 complexes indicates that Mrc1 is a substoichiometric component of a stoichiometric Tof1/Csm3 complex (Nedelcheva et al. 2005). Additional evidence suggests that Mrc1 and Tof1/Csm3 reside in separate pathways. The Tof1/Csm3 complex, but not Mrc1, is specifically required for fork pausing in the rDNA and at other natural pausing sites (Tourriere et al. 2005). Furthermore, Tof1/Csm3 and Mrc1 function in redundant pathways in the response to DNA damage induced by topoisomerase I poisons (Redon et al. 2006). It will be of great interest to determine the architecture of the protein complexes that underlie our genetic observations.

Several lines of evidence support a direct link between DNA replication and sister-chromatid cohesion. The recruitment of the cohesin loading factor Scc2 onto chromatin requires DNA replication licensing in Xenopus egg extracts (Gillespie and Hirano 2004). In mammalian cells, stably bound cohesin, which is speculated to form the topological links between sister chromatids, can be established only co-replicationally (Gerlich et al. 2006). In Schizosaccharomyces pombe, mutant forms of hsk1⁺ and dfp1⁺, which are essential for the initiation of replication initiation, have cohesion defects (Takeda et al. 2001; Bailis et al. 2003). In S. cerevisiae, the origin recognition complex components Orc5 and Orc2, which are essential for replication, promote cohesion in a pathway parallel to cohesin (Suter et al. 2004; Shimada and Gasser 2007). Additionally, direct physical links have been found between the cohesin subunit Smc1 and DNA replication proteins in S. cerevisiae (Ryu et al. 2006) and also between Eco1 and PCNA, a cofactor of DNA polymerase (Moldovan et al. 2006). Like the essential genes, a number of nonessential cohesion genes also have links with DNA replication. Mrc1, Tof1, Csm3, Ctf4, and Ctf18 are all found at DNA replication forks (Katou et al. 2003; Osborn and Elledge 2003; Gambus et al. 2006; Lengronne et al. 2006). Mrc1 functions constitutively to promote normal replication fork progression (Szyjka et al. 2005; Tourriere et al. 2005) and Tof1/Csm3 is critical for replication fork pausing (Tourriere et al. 2005; Mohanty et al. 2006). These data support models in which the establishment of sister-chromatid cohesion is coupled with DNA replication during S phase and is intimately connected to the progression of DNA replication forks.

Two models of cohesion establishment have been described, one in which cohesion is established passively by passage of the replication fork through a cohesin ring that encircles the DNA (Haering et al. 2002; Gruber et al. 2003) and one in which interactions between cohesin complexes on sister chromatids establish cohesion (Campbell and Cohen-Fix 2002; Milutinovich and Koshland 2003; Huang et al. 2005). Roles for the nonessential cohesion pathways that we have described can readily be envisaged in the context of either model. In the first model, a role for appropriate replisome architecture has recently been proposed (Lengronne et al. 2006), alluding to the possibility that the replisome might have to be the correct shape to pass through the cohesin ring without disrupting it. Replisomes lacking certain fork components, such as the nonessential cohesion proteins studied here, might cause a partial phenotype in which not every cohesin ring is disrupted, which could account for those components being nonessential and for mutants in single components not resulting in detectable disruption or redistribution of bound cohesin (Lengronne et al. 2006). Since we have found that two nonessential cohesion pathways make redundant contributions to sister-chromatid cohesion, it will be of particular interest to determine whether the cohesin redistribution or disruption that this model predicts can be detected in double mutants in which both nonessential cohesion pathways are disrupted. Another possible function of the nonessential cohesion pathways that is relevant to this first model is that they are important primarily as determinants of replisome integrity (Lengronne et al. 2006). When replisome integrity is compromised sufficiently, for example, when genes in both cohesion pathways are mutant, then the primary effect could be loss of replisome function, which then leads indirectly to cohesion defects. We disfavor this view for two reasons: First, we would expect that loss of replisome function would result in persistence of unreplicated regions, and unreplicated regions should prevent the separation of sister chromatids. However, we observed a clear increase in sister-chromatid separation in double mutants at or very near the end of S phase. Second, if replication through cohesin rings establishes cohesion in a truly passive manner, then completion of DNA replication should be sufficient to establish cohesion. In the case of the mrc1Δ tof1Δ mutant, flow cytometric analysis indicated that replication was complete by the time that separated sisters were observed. It remains possible, however, that subtle defects in replication that are not detected by flow cytometry contribute to the cohesion defects that we observed. An additional possibility, which could be relevant to either model, is that the nonessential cohesion pathways take part in a more sophisticated reaction that moves cohesin from in front of the replication fork to sites behind the fork, where it then mediates cohesion either by encircling both sister chromatids or by oligomerization of cohesin bound to each sister chromatid. In this view, the nonessential cohesion proteins could function as redundant factors that facilitate loading and unloading of cohesin during passage of the replication fork.

Recent work has identified additional unique pathways that contribute to sister-chromatid cohesion: one at chromosome arm sites that is mediated by the condensin chromosome condensation complex (Lam et al. 2006) and one that is mediated by the origin recognition complex in a pathway parallel to the cohesin pathway (Shimada and Gasser 2007). Here we have classified eight nonessential cohesion genes into two pathways that contribute to the establishment of sister-chromatid cohesion. As additional nonessential sister-chromatid cohesion genes are identified, it will be of interest to determine if they define additional pathways to promote cohesion or if they can be placed in either of the two pathways that we have identified or into the condensin- or ORC-dependent pathways. This determination will be greatly facilitated by the availability of the conditional alleles of csm3 and ctf8.

We thank Phil Hieter and Kim Nasmyth for strains and Bri Lavoie for valuable discussions and advice on cohesion assays and Pds1 staining. This work was supported by grants from the Canadian Institutes of Health Research to G.W.B. and to C.B., Genome Canada to C.B., and Genome Ontario to C.B. G.W.B. is a Research Scientist of the National Cancer Institute of Canada.