Modifying sister chromatid cohesion for meiosis

In proliferating cells, the sister chromatids segregate to opposite sides of the cell during mitosis, thereby transferring copies of chromosomes into each of the two daughter cells. This process is called equational chromosome segregation. For it to occur properly, corresponding sister chromatids must attach to microtubules emanating from opposite poles (bipolar attachment). The cohesion at kinetochores counteracts the pulling force of the microtubules and plays an essential role in bipolar attachment, which is stabilized by the tension generated across the kinetochores. Only when all kinetochores come under tension is the spindle checkpoint inactivated. This provokes activation of the anaphase-promoting complex (APC), leading to the ubiquitin-dependent destruction of securin, a guardian of the specific endopeptidase separase. In turn, the released separase cleaves the cohesin complexes that hold sister chromatids together, thereby allowing the sisters to be segregated to opposite poles by the forces pulling on the spindle (Figs 1, 2) (Nasmyth, 2002; Uhlmann, 2003).

Meiosis is a specialized cell cycle that reduces the chromosome number by half to allow fertilization during sexual reproduction to generate one normal nucleus. It consists of two rounds of chromosome segregation following a single round of DNA replication. At the first meiotic division (meiosis I), homologous chromosomes, rather than sister chromatids, are pulled to opposite poles. Only at the second meiotic division (meiosis II) do sister chromatids segregate as they do in mitosis (Fig. 1) (Lee and Orr-Weaver, 2001; Petronczki et al., 2003). Thus, meiosis I has a unique form of chromosome segregation, and understanding its molecular mechanism is a major challenge. Three crucial events at chromosomes must take place correctly for meiosis I to proceed. First, homologous chromosomes (homologues) pair to recombine, forming chiasmata in which one sister chromatid from one homologue is covalently attached to a sister chromatid from the other homologue. Second, sister kinetochores must attach to microtubules from the same spindle pole (monopolar attachment). Therefore, spindles create tension only when homologues are pulled in opposite directions because regions distal to chiasmata physically link them. Third, only arm cohesion must be destroyed during anaphase I, whereas centromeric cohesion must persist; this results in the separation of homologues but not sister chromatids. The residual centromeric cohesion is important during meiosis II for bipolar attachment of sister kinetochores (Miyazaki and Orr-Weaver, 1994). Here, I discuss the specific modifications of sister chromatid cohesion in meiosis that underpin these events and the molecular mechanisms behind them.

During mitosis, sister chromatid cohesion is carried out by cohesin complexes comprising two SMC (for `structural maintenance of chromosomes') -family proteins - Smc1 and Smc3 - and two non-SMC subunits - Rad21/Scc1/Mcd1 and Scc3 (Nasmyth, 2002). Smc1-Smc3 heterodimers are proposed to interact through their long coiled-coil stretches and embrace sister chromatids. Scc1 is thought to lock their ends, presumably with the aid of Scc3 (Fig. 2) (Haering and Nasmyth, 2003). The disassembly of the cohesin complex triggers sister chromatid separation, following cleavage of Rad21/Scc1 by separase (Fig. 2).

During meiosis, the Rad21/Scc1 subunit is mostly replaced by a meiosis-specific version, Rec8. If Rec8 is depleted in meiosis, recombination, synaptonemal complex formation, monopolar attachment (at least in fission yeast) and persisting centromeric cohesion all fail to occur (DeVeaux and Smith, 1994; Parisi et al., 1999; Klein et al., 1999; Watanabe and Nurse, 1999). Remarkably, these meiosis-specific properties of chromosomes are not restored in rec8Δ cells when Rad21/Scc1 is expressed ectopically (Toth et al., 2000; Yokobayashi et al., 2003). Meiosis-specific cohesin thus appears not merely to be a link holding sister chromatids together but also to engage in other aspects of chromosome behaviour. Studies in mammals have suggested that meiotic cohesin core proteins recruit the recombination apparatus (Eijpe et al., 2003; Pelttari et al., 2001) and promote synapsis between homologous chromosomes together with associated synaptonemal-complex-specific proteins (Eijpe et al., 2000). Cohesins establish sister chromatid cohesion during DNA replication (Uhlmann and Nasmyth, 1998). If meiosis is initiated artificially after mitotic DNA replication by ectopically activating the meiosis-inducing network, meiotic cohesin does not appear to function, even though Rec8 complexes associate with chromatin (Watanabe et al., 2001). Therefore, the exchange of mitotic cohesins for meiotic cohesins must occur prior to DNA replication. Studies of budding yeast suggest that recombination occurs only after the passage of the replication fork (Borde et al., 2000; Cha et al., 2000). Thus, the regulatory processes involved in meiotic chromosome cohesion must initiate at the latest during DNA replication.

Additional meiosis-specific cohesin subunits have been found in various organisms (Table 1). Mammalian SMC1β is reportedly expressed only in testis (Revenkova et al., 2001) and is probably a meiosis-specific version of SMC1. The role of SMC1b and its specific requirements in meiotic sister chromatid cohesion and recombination have recently been determined (Revenkova et al., 2004). Mammals also have at least three Scc3-like proteins, SA1-SA3, and SA3 is meiosis specific (Prieto et al., 2001). Fission yeast have two: the ubiquitous Psc3 and the meiosis-specific Rec11. Interestingly, meiotic Rec8 complexes along chromosome arms largely contain Rec11, whereas those in the vicinity of centromeres have Psc3 (Kitajima et al., 2003b). The interaction between Psc3 and heterochromatin protein Swi6 [a fission yeast homologue of heterochromatin protein 1 (HP1), which forms pericentromeric heterochromatin structures] could partly account for this. Two Scc3-like proteins thus define distinct organizations of meiotic cohesins at centromeres and in the chromosome arm regions of fission yeast. Mammalian SA3 disappears from chromosomes during anaphase I and is not detected at centromeres (Prieto et al., 2001), whereas centromeric Rec8 is preserved (Eijpe et al., 2003; Lee et al., 2003). SA3 might play a role similar to that of Rec11.

During mitosis, enrichment of cohesin at centromeres is important because centromeric cohesion plays a central role in counteracting the pulling force of the spindle and thus in the proper alignment of chromosomes at metaphase (Bernard and Allshire, 2002). By contrast, to facilitate the resolution of sister chromatids, arm cohesion is mostly lost during prophase (Losada et al., 2002). However, this system might not be appropriate for meiosis, because cohesion along chromosome arms plays a crucial role promoting recombination and holding homologues tightly together to ensure disjunction at meiosis I. This may be why a specific arm cohesin component is produced in meiosis. Budding yeast are exceptional in lacking the centromeric heterochromatin HP1-cohesin enrichment system and therefore may have relatively stronger arm cohesion, which potentially explains why this organism lacks a Rec11 homologue.

Meiotic recombination is catalysed by Spo11, which initiates formation of double-strand breaks and the ensuing repair reaction. The recombinational strand exchange contributes to the generation of genomic diversity of gametes. More importantly, perhaps, it generates chiasmata, which are physical links between homologues that secure two homologues being pulled from opposite poles at meiosis I (Carpenter, 1994). If recombination is abolished (e.g. by the spo11 mutation), homologues fail to align properly on the spindle and so segregate randomly at meiosis I (Klein et al., 1999).

In fission yeast, the selective inactivation of Rec11 or Psc3 at meiosis can distinguish cohesin functions at centromeres and along the arm regions (Kitajima et al., 2003b). The inactivation of Psc3 does not influence arm cohesion at all, but both monopolar attachment and persistence of centromeric cohesion (centromeric protection) are impaired at meiosis I. By contrast, rec11Δ cells preserve centromeric cohesion while exhibiting a precocious separation of chromosome arms. Consequently, rec11Δ causes a high incidence of non-disjunction of homologues at meiosis I, illuminating the importance of arm cohesion for ensuring homologue segregation.

Sister kinetochores attach to microtubules emanating from opposite spindle poles during mitosis or meiosis II - proper attachment being ensured by mechanisms that monitor the tension generated at kinetochores. A recent study in budding yeast has indicated that a circular unreplicated chromosome carrying two kinetochores widely separated from each other preferentially establishes bipolar attachment at mitosis (Dewar et al., 2004). This elegant experiment stressed the importance of tension sensing rather than the geometry of kinetochores for establishing bipolar attachment. Since homologues are connected by chiasmata at meiosis I, tension is also produced when sisters are pulled to the same pole but homologues are pulled in opposite directions. However, tension might not be the primary determinant of sister kinetochore orientation at meiosis I, because monopolar attachment occurs even in the absence of interaction between homologues (e.g. in recombination-minus meiosis) (Klein et al., 1999; Yamamoto and Hiraoka, 2003), in which case tension can be produced only by the bipolar attachment of sister kinetochores.

Thus, the geometry of sister kinetochores might be a major determinant in monopolar attachment at meiosis I. Electron microscopy (EM) has shown that both sister kinetochores indeed attach to microtubules (Goldstein, 1981; Parra et al., 2004), excluding the possibility that inactivation of one sister kinetochore results in monopolar attachment. An important clue linking cohesin complexes to kinetochore orientation came from the finding that the fission yeast rec8 mutant undergoes equational, rather than reductional, division at meiosis I (Watanabe and Nurse, 1999). Such bipolar attachment of sister kinetochores at meiosis I also occurs in the maize meiotic mutant absence of first division 1 (afd1), which lacks the rec8 homolog Afd1 (W. Z. Cande, personal communication). Subsequent analysis revealed that the equational segregation at meiosis I depends on Rad21, which relocates to the centromeres and establishes centromeric cohesion if Rec8 is absent (Yokobayashi et al., 2003). This suggests that monopolar attachment depends on Rec8, and that Rad21 cannot substitute for it. `Equational meiosis I' proceeds with normal kinetics (Toth et al., 2000; Yokobayashi et al., 2003), but there is no centromeric cohesion at anaphase I when Rad21 is present instead of Rec8. Thus, centromeric protection depends on the character of Rec8. Analysis of the localization of cohesins around centromeres in fission yeast has revealed that Rad21 complexes localize mainly to pericentromeric heterochromatin, whereas Rec8 complexes localize to central core regions as well as pericentromeric regions. The enrichment of cohesins at the pericentromeric region depends on the formation of heterochromatin at mitosis, providing robust cohesion at centromeres that might be important for faithful chromosome segregation (Bernard et al., 2001b; Nonaka et al., 2002). Likewise, heterochromatin is also required at meiosis for the enrichment of Rec8 at pericentromeric regions, but not at the central core regions. Interestingly, a lack of heterochromatin specifically abolishes the centromeric cohesion that persists during anaphase of meiosis I but does not affect monopolar attachment (Kitajima et al., 2003b). A corollary is that the persisting cohesion at meiosis I depends on pericentromeric Rec8 cohesins, whereas the mono-orientation of sister kinetochores is carried out by the central core rather than pericentromeric Rec8 complexes (Fig. 3). From the DNA sequences of the centromere, Yanagida and co-workers have predicted that the central core region (∼10 kb) is likely to form structures that protrude from other regions of the centromere (Fig. 3) (Takahashi et al., 1992). Moreover, kinetochore proteins that interact with spindle microtubules usually localize to the central core region, which suggests that the spindle attachment interface lies in this region (Nakaseko et al., 2001). One attractive model is that Rec8 creates monooriented protrusions at centromeres by establishing cohesion of sister strands at the central core during DNA replication (Fig. 3) (Watanabe et al., 2001; Yokobayashi et al., 2003).

Unlike the centromeres of fission yeast and other eukaryotes, the budding yeast centromere is unique in comprising a short stretch of DNA (∼125 bp) that is attached to a single microtubule and lacks pericentromeric heterochromatin. Knockout screening in budding yeast for genes that are upregulated during meiosis has identified the meiosis-specific kinetochore factor Mam1 (Toth et al., 2000). In mam1Δ cells, the meiosis I spindle fails to separate homologous chromosomes but the meiosis II spindle can separate sister chromatids. This suggests that Mam1 plays a specialized role establishing monopolar attachment at meiosis I (Fig. 3). Because centromeric protection is largely preserved in mam1Δ cells, this analysis suggests that monopolar attachment is independent of the protection process.

A subsequent study revealed that a Mam1 protein complex called monopolin includes at least two additional proteins, Csm1 and Lrs4, both of which localize to the nucleolus in vegetative cells but travel to kinetochores shortly before meiosis I, in a complex with Mam1 (Rabitsch et al., 2003). The localization of monopolin at kinetochores depends on the activity of the polo-like kinase Cdc5; thus, budding yeast Cdc5 is required for establishing monopolar attachment (Clyne et al., 2003; Lee and Amon, 2003). Monopolin does not appear to be conserved in other organisms. Fission yeast contains a protein related to Csm1, Pcs1. However, Pcs1 is required for chromosome segregation during mitosis and meiosis II but not for meiosis I (Rabitsch et al., 2003). Moreover, monopolar attachment can be established in budding yeast rec8Δ cells forced to express Scc1 during meiosis (Toth et al., 2000), which contrasts with the observations in fission yeast (Yokobayashi et al., 2003). The mechanisms for the establishment of monopolar attachment thus seem to have diverged between budding yeast and fission yeast. Further studies are required if we are to formulate a generalized view of the regulation of monopolar attachment in meiosis.

During meiosis I, sister chromatid cohesion is disrupted along chromosome arms, whereas centromeric cohesion is protected. Rec8 along the chromosome arms mostly disappears during meiosis I, but centromeric Rec8 persists until meiosis II (Fig. 1). Several lines of evidence suggest that Rec8 along chromosome arms is cleaved by separase, the same enzyme that cleaves the mitotic cohesin Scc1/Rad21. In various organisms, the inactivation of separase blocks homologue segregation at meiosis I (Herbert et al., 2003; Siomos et al., 2001; Terret et al., 2003). Similarly, in both budding yeast and fission yeast, expression of a non-cleavable Rec8 mutant lacking the separase target sequences also blocks their segregation; indeed, abolishing recombination alleviates this blockage (Buonomo et al., 2000; Kitajima et al., 2003a). These results strongly suggest that the cleavage of Rec8 by separase and the subsequent resolution of the chiasmata could be a trigger for the segregation of homologues at meiosis I. When the formation of arm cohesin complexes is disrupted by depleting the arm-specific cohesin subunit Rec11 in fission yeast, the hindrance of homologue separation by non-cleavable Rec8 is alleviated whereas sister separation remains tightly blocked. This blockage is released by inactivation of Psc3, a major centromeric partner of Rec8 (Kitajima et al., 2003a). Thus, these observations suggest that the cleavage of Rec8 in the vicinity of centromeres is required for the initiation of anaphase II. The securin accumulation-degradation cycle operates not only at meiosis I but also at meiosis II, indicating that separase is indeed activated at both divisions in meiosis (Kitajima et al., 2003a; Salah and Nasmyth, 2000).

An obvious question concerns which molecules protect centromeric Rec8 from degradation during meiosis I but not during meiosis II. Budding yeast spo13 mutants show significant defects in centromeric protection as well as in monopolar orientation, largely undergoing equational division at meiosis I (Klapholz and Esposito, 1980; Klein et al., 1999). Moreover, ectopic expression of Spo13 together with Rec8 blocks sister chromatid separation at mitotic anaphase (Lee et al., 2002; Shonn et al., 2002). These results suggest that meiosis-specific Spo13 is involved in the protection of centromeric Rec8. But Spo13 might not be a centromeric protein and its homologues are not identified in other organisms.

Recently, functional screening in fission yeast identified shugoshin (Sgo1; Japanese for `guardian spirit') as a Rec8 protector. The co-expression of Sgo1 and Rec8 in mitotic cells blocks the degradation of centromeric Rec8 at anaphase, resulting in inhibition of sister chromatid separation and subsequent cell death. This protection is specific to Rec8 because, if Rad21 is predominant in cohesin complexes in the cell, the forced expression of Sgo1 has little effect on mitotic growth (Kitajima et al., 2004). Sgo1 is a meiosis-specific protein that is usually expressed during meiosis I. Deletion of sgo1 leads to a defect in the persistence of centromeric Rec8 after anaphase I and subsequently results in random segregation at meiosis II (Fig. 4). Independent knockout screens in fission yeast and budding yeast also identified the sgo1/SGO1 gene (Katis et al., 2004; Marston et al., 2004; Rabitsch et al., 2004). Fission yeast Sgo1 localizes to kinetochores, especially to the pericentromeric heterochromatin regions, and associates with Rec8 complexes in vivo. Because the heterochromatin-dependent enrichment of Rec8 was independently shown to be important for persisting centromeric cohesion, these results suggest that Sgo1 protects pericentromeric Rec8 complexes from separase attack at the onset of anaphase I. Sgo1 is degraded completely by meiosis II, which is consistent with the notion that centromeric Rec8 is protected only during meiosis I and not during meiosis II (Fig. 4) (Kitajima et al., 2004; Rabitsch et al., 2004).

Fission yeast has a paralogue of shugoshin, Sgo2, that is ubiquitously expressed throughout its life cycle and specifically localizes to kinetochores at metaphase during mitosis and meiosis. sgo2Δ cells show chromosome instability during proliferation and a substantial increase in the non-disjunction of homologues at meiosis I, indicating that Sgo2 plays a crucial role at kinetochores in chromosome segregation in general (Kitajima et al., 2004; Rabitsch et al., 2004). Although the precise molecular function of Sgo2 remains to be addressed, it is likely that Sgo2 is a prototype of shugoshin, from which Sgo1 diverged to become a specific protector of Rec8. Budding yeast have a single shugoshin, Sgo1, which behaves similarly to fission yeast Sgo1 and Sgo2 (Katis et al., 2004; Kitajima et al., 2004; Marston et al., 2004). The conservation of shugoshin in both yeasts is noteworthy. Careful inspection of the databases, using conserved short stretches of fungi shugoshins, revealed proteins with reasonably similar regions in several multicellular organisms (Kitajima et al., 2004). Strikingly, one such protein is Drosophila MEI-S332, a previously identified kinetochore protein, mutations in which lead to premature sister centromere separation in meiosis (Kerrebrock et al., 1995; Lee and Orr-Weaver, 2001). Consequently, MEI-S332 is now accepted to act as a Rec8 protector. MEI-S332 also has some role in mitosis, like budding yeast Sgo1. Curiously, budding yeast, flies and worms seem to have single shugoshin, whereas fission yeast, plants and other animals have two (Table 1) (Kitajima et al., 2004). In the latter organisms, shugoshin functions might be divided. It remains to be seen exactly how shugoshin protects Rec8 from cleavage by separase, a problem that will be solved by further biochemical and crystallographic analyses.

It would be interesting to know the precise localization of budding yeast Sgo1, because the centromere structure in this organism has diverged from that in fission yeast. The forced expression of Rec8 in budding yeast does not block sister separation during mitosis (despite the presence of Sgo1), whereas co-expression of Rec8 with Spo13 does. Therefore, budding yeast Sgo1 may be specifically activated or modified to protect Rec8 during meiosis I, and this presumably depends on Spo13. In fission yeast, the Bub1 checkpoint protein is required for shugoshin (both Sgo1 and Sgo2) to localize to kinetochores (Kitajima et al., 2004), which explains the observation that centromeric Rec8 fails to persist throughout meiosis I in bub1Δ cells (Bernard et al., 2001a).

We are now starting to understand the molecular details of chromosome segregation during meiosis. Strikingly, the main conserved players, cohesins and shugoshin, have ancestral molecules that act during mitosis. Indeed, one would intuitively think that meiosis evolved from mitosis not by the abrupt emergence of a set of novel proteins but by a subtle divergence of the mitotic apparatus, for example by gene duplication. One of the mysteries of biology is why sexual reproduction is so predominant in the eukaryotic world. A likely explanation is that organisms must continuously evolve to survive the continually changing environment on earth, for which the mixing of genomes from two organisms has obvious advantages. Whatever the answer, the essence of meiotic devices should be found even in yeast, one of the most primitive eukaryotic organisms.

Work in my lab is funded in part by grants from the Ministry of Education, Science and Culture of Japan. I am grateful to members of my lab for many discussions and in particular to Silke Hauf for critical reading of the manuscript. I thank Richard Sever for extensive editing of the text.