Direct evaluation of cohesin-mediated sister kinetochore associations at meiosis I in fission yeast

During sexual reproduction, germ cells undergo two rounds of chromosome segregation, resulting in the formation of gametes containing half the original number of chromosomes. In the first round (meiosis I), homologous chromosomes, which are physically linked by a recombination product called chiasma, are attached to opposite spindle poles (bi-oriented attachment), whereas sister chromatids are attached to the same pole (mono-oriented attachment); consequently, homologous chromosomes segregate apart from each other (reductional segregation) (Duro and Marston, 2015; McIntosh and Hays, 2016; Miller et al., 2013; Ohkura, 2015). This chromosome segregation is strikingly different from that seen in mitosis, in which sister chromatids attach to the opposite poles and segregate apart from each other (equational segregation).

Chromosome attachment to the spindle depends on protein complexes assembled on centromeres, known as kinetochores, and it is generally thought that meiosis I chromosome attachment requires meiosis-specific positioning of the kinetochores. Studies of multicellular organisms have revealed that kinetochores are positioned separately on the opposite sides of sister centromeres during mitosis, whereas they are positioned side by side and associate with each other in meiosis I (Fig. 1A) (Goldstein, 1981; Lee et al., 2000; Li and Dawe, 2009; Parra et al., 2004). Sister kinetochore associations are thought to facilitate the mono-oriented attachments of sister chromatids, and establish bi-oriented attachments of homologous chromosomes, with the aid of chiasma-mediated homologous chromosome association, which eliminates erroneous chromosome attachments by generating tension at kinetochores and/or coordinating the oscillation of homologous chromosomes between poles (Duro and Marston, 2015; Hauf and Watanabe, 2004; Miller et al., 2013; Nasmyth, 2015; Nicklas, 1997; Wakiya et al., 2021; Watanabe, 2012; Yamamoto, 2021). However, the mechanisms underlying sister kinetochore associations are not fully understood.

Cohesin, a protein complex that mediates sister chromatid cohesion and forms topologically associating domains (Davidson and Peters, 2021; Onn et al., 2008; Peters and Nishiyama, 2012; Uhlmann, 2016), is thought to play a crucial role in establishing sister kinetochore associations (Duro and Marston, 2015; Hauf and Watanabe, 2004; Ishiguro, 2019; Miller et al., 2013; Nasmyth, 2015; Watanabe, 2012). The role of cohesin in meiotic chromosome segregation has been studied extensively in the fission yeast Schizosaccharomyces pombe. In S. pombe, Rec8-containing meiotic cohesin is essential for co-segregation of sister chromatids, and loss of Rec8 results in equational segregation of sister chromatids (Watanabe and Nurse, 1999; Watanabe et al., 2001). Consistently, cohesin regulators are required for sister chromatid co-segregation at meiosis I. DNA replication-related factors, including the replication fork protection complex (FPC) proteins, regulate sister chromatid cohesion (Ansbach et al., 2008; Petronczki et al., 2004), whereas the meiosis-specific centromere protein Moa1 regulates meiotic centromere cohesion by recruiting a Polo kinase homolog (Plo1) to centromeres (Kim et al., 2015; Ma et al., 2021; Sakuno et al., 2009; Yokobayashi and Watanabe, 2005). Elimination of DNA replication-related cohesin regulators, such as the FPC component Mrc1 (Shimmoto et al., 2009), or Moa1, results in equational segregation of sister chromatids in chiasma-lacking cells (Hirose et al., 2011; Yokobayashi and Watanabe, 2005). Whereas Rad21-containing cohesin is localized at pericentromeric regions and establishes centromere cohesion in mitosis (Tomonaga et al., 2000), Rec8-containing cohesin is additionally localized at the centromere core region and establishes association of the centromere cores in addition to pericentromeric cohesion (Sakuno et al., 2009; Watanabe et al., 2001). From these results, it has been proposed that cohesin-mediated centromere core association induces association of sister kinetochores, enabling mono-oriented attachment of sister chromatids (Fig. 1A) (Sakuno et al., 2009; Watanabe et al., 2001). In support of this view, loss of the Moa1 homolog meikin, which probably regulates centromeric cohesin, results in separation of sister kinetochores on metaphase sister chromatids in mammalian oocytes (Kim et al., 2015), and age-related loss of centromere cohesin results in sister kinetochore separation in mammalian or fly oocytes (Chiang et al., 2010; Wang et al., 2019; Zielinska et al., 2019).

While previous studies in S. pombe have suggested that cohesin-mediated centromere core cohesion generates sister kinetochore associations, core cohesion-dependent sister kinetochore association on intact chromosomes has yet to be demonstrated directly. Examination of association states of the centromere cores by excising the core region provided supportive evidence (Sakuno et al., 2009); however, in this previous study, the core region was removed from chromosomes in cells arrested before entry into meiosis I by mei4 mutation, and, therefore, the association state of these centromere cores on chromosomes and at metaphase I remains elusive. In addition, confusingly, even with wild-type meiotic kinetochore organization, bi-oriented attachment of sister kinetochores occurs frequently in chiasma-lacking cells (Hirose et al., 2011; Wakiya et al., 2021), and loss of anaphase cohesin protection alone can cause equational segregation of sister chromatids (Dudas et al., 2011; Hirose et al., 2011; Sakuno et al., 2011), which has been thought to be the evidence of impaired sister kinetochore association. As all analyzed cohesin-related mutants are defective in anaphase centromeric cohesion (Hirose et al., 2011; Ma et al., 2021; Yokobayashi and Watanabe, 2005), we cannot exclude the possibility that impaired anaphase cohesion, rather than impaired sister kinetochore association, causes equational sister chromatid segregation in cohesin-related mutants.

In the present study, we describe a novel method for the direct evaluation of the kinetochore association state in S. pombe, and examine kinetochore associations in mitosis and meiosis I. Using this method, we show that sister kinetochores are indeed separately positioned in mitosis but are associated with each other at meiosis I on intact chromosomes, as observed previously in higher eukaryotes, and that kinetochore association depends on a meiotic cohesin and its regulators. We also show that in addition to cohesin-related factors, mating-pheromone signaling is required for sister kinetochore association, as previously suggested (Asakawa et al., 2005; Yamamoto and Hiraoka, 2003).

Kinetochore association state can be evaluated on hyper-condensed chromosomes in metaphase-arrested vertebrate cells. S. pombe chromosomes are relatively small; however, we reasoned that chromosome hyper-condensation would increase spatial separation of bi-oriented sister kinetochores, enabling us to distinguish separated and associated sister kinetochore pairs by optical microscopy. We arrested cells at metaphase I using an nda3-KM311 β-tubulin mutation, which induces microtubule depolymerization and activates the spindle assembly checkpoint (SAC) at low temperatures (Fig. 1B; Fig. S1A) (Hiraoka et al., 1984). To efficiently arrest cells in metaphase I, and to prevent chiasma-dependent homologous chromosome association that could hinder the identification of individual sister chromatid pairs, we used haploid nda3-KM311 cells that contain both mating type genes, which lack homologous chromosomes and can enter meiosis synchronously in nitrogen-free medium (Yamamoto and Hiraoka, 2003; Yoshida et al., 2013). Indeed, when these cells were induced to enter meiosis in nitrogen-free medium and subsequently incubated at 18°C, three condensed individual sister chromatid pairs were observed, as expected (Fig. S1B). However, we encountered two problems with this strategy. First, SAC activation does not completely block cell cycle progression at metaphase, and a fraction of SAC-activated cells proceed into anaphase (Rieder and Maiato, 2004), obscuring evaluation of sister kinetochore association state, because sister kinetochores separate during anaphase I (Parra et al., 2004). Second, synchronous induction of meiosis is incomplete in haploid cells, and upon shifting temperature to 18°C, a fraction of cells is in mitosis or meiosis II, hindering the evaluation of meiosis I kinetochore association. To overcome these problems, we marked metaphase I chromosomes using a GFP-tagged cohesin subunit, Rec8, which localizes at chromosomes specifically in meiosis I and is mostly removed from the chromosomes at anaphase I onset. Likewise, we marked mitotic metaphase chromosomes using a GFP-tagged mitotic cohesin subunit, Rad21.

Using this method, we evaluated the kinetochore association state in mitosis and meiosis I. For mitotic kinetochore evaluation, nda3-KM311 cells were grown at 30°C and subsequently incubated at 18°C (Fig. 1C, Mitosis). For meiotic kinetochore evaluation, nda3-KM311 haploid cells containing both mating type genes were induced to enter meiosis at 30°C in nitrogen-free medium, and after a single round of mitotic division, which occurs upon nitrogen depletion (Yamamoto and Hiraoka, 2003), the temperature was shifted to 18°C (Fig. 1C, Meiosis I). Both methods yielded condensed chromosomes in ∼30–40% of the cells. We examined the association state of sister kinetochores by visualizing Nuf2, a subunit of the outer kinetochore Ndc80 complex (Musacchio and Desai, 2017). We also examined centromere core association state by visualizing the fission yeast CENP-A homolog, Cnp1 (Takahashi et al., 2000). Visualization of kinetochores and centromere cores yielded similar results: in both cases, two separated GFP signals were observed on >25% of mitotic condensed chromosomes but were never observed on condensed metaphase I chromosomes (Fig. 1D–F).

Signals separated by a distance shorter than the resolution limit of the fluorescence microscope determined by the fluorescence wavelength (∼600 nm for mCherry and ∼500 nm for GFP) cannot be recognized as separated signals (the Rayleigh criterion) (Inoué and Spring, 1997). Indeed, the average distances between separated signals was close to or greater than the resolution limit (∼0.5 µm; Fig. S2A), and we predicted that this would result in the underestimation of separation. To overcome this problem, we examined the shape of non-separated signals, reasoning that the shape of unresolved signals of separated sister kinetochores or centromere cores would be wider than that of associated kinetochores or the centromere cores. We analyzed the width of non-separated signals by measuring the ratio of the major-to-minor axis of an ellipse fitted to the signal (Fig. 1G). We found that the mean ratio of the major-to-minor axis of non-separated signals of Nuf2 or Cnp1 on mitotic chromosomes was significantly greater than that of the non-separated signals on meiosis I chromosomes, indicating that the shape of the signals was wider in mitosis than in meiosis I (Fig. 1G). Considering all these results, we concluded that sister kinetochores and centromere cores are separated in mitosis but are associated with each other in meiosis I.

Next, we examined the association states of sister kinetochores and centromere cores at meiosis I in cells containing mutations in cohesin-related genes that cause frequent equational segregation of sister chromatids in chiasma-lacking, recombination-deficient rec12Δ cells (Fig. S2B) (Hirose et al., 2011; Yokobayashi and Watanabe, 2005). As Rec8-GFP could not be used for chromosome visualization in rec8Δ cells, we developed a chromosome fluorescent marker that specifically identifies metaphase I chromosomes, utilizing a strategy developed for separase sensors (Shindo et al., 2012; Yaakov et al., 2012). We made a construct encoding a fusion of histone H3 (Hht1) and mCherry, which were connected by a Rec8 fragment linker containing separase cleavage sites (Rec8F) (Fig. 2A). We expressed the fusion construct in a meiosis-specific manner by placing it under the spo5 meiosis-specific promoter (Mata et al., 2002) and introducing a DNA element called the determinant of selective removal (DSR; Fig. 2A), which causes degradation of the transcript during mitosis (Harigaya et al., 2006). Fluorescence of the fusion successfully marked chromosomes upon entering meiosis, whereas fluorescence disappeared at the onset of anaphase I (Fig. 2B), which enabled the identification of metaphase I chromosomes. Consistently, when haploid cells were synchronously induced to enter meiosis, as described previously (Hirose et al., 2011), the fusion protein appeared after the mitotic division that precedes meiosis and decreased as cells underwent meiotic divisions (Fig. 2C). When the association states of sister kinetochores and centromere cores were analyzed in wild-type cells using the fusion, no separated Nuf2 or Cnp1 signals were observed (Fig. 3A–D), confirming that the chromosome marker did not affect the association states of sister kinetochores or centromere cores at meiosis I.

Using the developed chromosome marker, we first examined the effect of loss of meiotic cohesin on the association state of sister kinetochores and centromere cores. In rec8Δ cells, signals separated by a distance close to the resolution limit were observed for both kinetochores and centromere cores at similar frequencies to mitotic cells (Figs 1F and 3A–D; Figs S2C and S3A,B), and non-separated signals were significantly wider in shape than in wild-type cells (Fig. 3E,F; Fig. S3C,D). These results indicate that meiotic cohesin is required for the association of sister kinetochores and centromere cores at meiosis I.

Next, we examined the association state of sister kinetochores and centromere cores in mutants defective in cohesin regulation. In cells lacking Sgo1 (sgo1Δ), which protects centromeric cohesin during anaphase, no separated kinetochore signals were observed (Fig. 3A,C), although sister chromatids frequently underwent equational segregation in the absence of chiasmata (Fig. S2B). Likewise, no separated core signals were observed in cells lacking Swi6 (the heterochromatin protein 1 homolog in S. pombe), which is required for the recruitment of Sgo1 to centromeres (Fig. 3D) (Yamagishi et al., 2008). In both cases, the shape of non-separated signals was not significantly different from wild-type cells (Fig. 3E,F), indicating that sister kinetochore association is intact in these cells. Thus, loss of anaphase cohesin protection does not affect the association state of sister kinetochores or centromere cores.

By contrast, separation of sister kinetochores or centromere cores was observed in cells lacking the cohesin regulator Moa1, although separation was not statistically significant (Fig. 3A–D; Figs S2C and S3A,B). Non-separated signals were found to be significantly or nearly significantly wider than in wild-type cells (Fig. 3E,F; Fig. S3C,D). These results indicate that association of sister kinetochores and probably centromere cores was impaired in moa1Δ cells, although this impairment was milder than in rec8Δ cells. Moreover, the separation frequencies of sister kinetochores and centromere cores and the width of non-separated signals in moa1Δ rec8Δ cells were not significantly different from rec8Δ single mutant cells (Fig. 3E,F; Fig. S3A–D), supporting the idea that a major task of Moa1 is cohesin regulation (Yokobayashi and Watanabe, 2005).

In mrc1Δ cells, sister kinetochores and centromere cores separated at a low, but significant, level (Fig. 3A–D; Figs S2C and S3A,B), and non-separated signals were significantly wider in shape (Fig. 3E,F; Fig. S3C,D). Thus, as in moa1Δ cells, sister kinetochore association is impaired but at milder levels than in rec8Δ cells. In addition, the separation frequencies of sister kinetochores and centromere cores were not significantly different in mrc1Δ rec8Δ cells compared with rec8Δ cells (Fig. 3E,F; Fig. S3A–D), supporting the idea that Mrc1 establishes sister kinetochore association by regulating cohesin. However, Rec8 centromere localization was not significantly different in mrc1Δ cells (Fig. 3G), indicating that Mrc1 does not regulate the centromere localization of Rec8. Additionally, the DNA replication checkpoint function of Mrc1 is not required for sister kinetochore association, because deletion of cds1, which encodes an effector kinase functioning downstream of Mrc1 in the DNA replication checkpoint pathway (Alcasabas et al., 2001; Murakami and Okayama, 1995; Tanaka and Russell, 2001), did not affect the kinetochore association state or sister chromatid segregation (Fig. 3C,E; Fig. S2C).

In higher eukaryotes, the heterochromatic centromere domain is located underneath the centromere core and mediates sister centromere cohesion. In S. pombe, two heterochromatic DNA regions consisting of outer centromere repeats, which are present on both sides of the centromere central core region, are responsible for sister centromere cohesion. However, it remains unclear whether these two regions form a single heterochromatic centromere domain underneath the centromere core, as in higher eukaryotes. In addition, Rad21-containing mitotic cohesin is localized at the heterochromatic DNA regions in rec8Δ cells and mediates centromere cohesion in place of Rec8-containing meiotic cohesin (Yokobayashi et al., 2003); however, it remains unclear whether heterochromatic centromere domains of sister centromeres remain associated with each other, as in mitotic cells. We addressed these points by visualizing heterochromatin domains using GFP-tagged Swi6.

A single, strong Swi6 signal was mostly associated with both two separated and a single non-separated Cnp1 signal in mitotic cells; however, Cnp1-associated Swi6 signals were occasionally unclear because of weak signal intensities (Fig. 4A,B). Similarly, in meiotic cells, a single Swi6 signal was associated with a single non-separated Cnp1 signal (Fig. 4B,C). These observations support the idea that two stretches of outer centromere repeats form a single centromere domain, serving as the foundation for the centromere core, and mediating centromere cohesion in mitosis and meiosis I. We also found that a single Swi6 signal was located between the separated Cnp1 signals in rec8Δ cells, as in mitotic cells, although the separation frequency of centromere cores was lower in this particular rec8Δ strain (Fig. 4B,D). This observation indicates that heterochromatin-dependent centromere cohesion is intact in rec8Δ cells, despite centromere core association being impaired, and supports the hypothesis that Rad21-containing mitotic cohesin mediates sister centromere cohesion, as in mitotic cells (Yokobayashi et al., 2003). These observations are consistent with the current hypotheses of centromere organization (Hauf and Watanabe, 2004; Musacchio and Desai, 2017; Schalch and Steiner, 2017).

Mrc1 and other DNA replication-related cohesin regulators are produced in mitosis and meiosis, whereas Rec8 and Moa1 are exclusively produced in meiosis. Next, we investigated whether the production of Rec8 and Moa1 induces sister kinetochore association in mitosis. We enabled mitotic cells to produce a C-terminally tagged Rec8, which can form cohesin complex in mitosis (Kitajima et al., 2003). Despite the use of the native promoter of rec8, the tagged rec8 gene was mitotically expressed because of the lack of a DSR in the 3′ untranslated region (Fig. S4A) (Harigaya et al., 2006), and the Rec8 product accumulated on condensed chromosomes with enrichment at the centromere (Fig. 5A,B). We also enabled mitotic cells to produce an N-terminally tagged Moa1 using the nda3 promoter, which was shown to be functional in meiosis (Yokobayashi and Watanabe, 2005). We eliminated all DSR core and DSR-like sequences from the Moa1-coding nucleotide sequence, without altering the Moa1 amino acid sequence (Moa1-7dsr; Fig. S4B–D) (Harigaya et al., 2006; Yamashita et al., 2012), because the wild-type moa1⁺ gene failed to produce Moa1 in mitosis. N-terminally tagged Moa1-7dsr was successfully produced (Fig. S4A), localizing at the centromere core (Fig. 5A,C), and reducing the amount of core-localized Rec8 in mitosis (Fig. 5B), as Moa1 does in meiosis (Yokobayashi and Watanabe, 2005). These results suggest that Moa1 is functional in mitosis. In addition, the expression of rec8 and/or moa1-7dsr slightly or hardly reduced cell viability (Fig. S4E), demonstrating that these factors do not severely compromise chromosome segregation in mitosis.

The expression of rec8 reduced separation of sister kinetochores and centromere cores at non-significant levels (Fig. 5D; Fig. S5A.B). By contrast, moa1-7dsr expression alone also reduced kinetochore and centromere core separation at significant levels (Fig. 5D; Fig. S5A.B). This result indicates that Moa1 is able to affect kinetochore and centromere organization in the absence of Rec8. Interestingly, the separation frequencies were reduced in cells simultaneously producing Rec8 and Moa1-7dsr to a similar level to that of Rec8 alone (Fig. 5D; Fig. S5A,B), suggesting a functional relationship between Rec8 and Moa1. Despite an alteration in separation of sister kinetochores and centromere cores, separation was not eliminated (Fig. 5D), and the shape of non-separated signals was not significantly altered in Rec8- and/or Moa1-7dsr-producing cells (Fig. 5E; Fig. S5C,D). Therefore, although Rec8 and Moa1-7dsr have an impact on sister kinetochore organization, they are not sufficient for the induction of sister kinetochore association in mitosis.

It was previously shown that a lack of mating-pheromone signaling results in equational segregation of sister chromatids at meiosis I in chiasma-lacking cells, although Rec8 and Moa1 are localized at centromeres (Hayashi et al., 2006; Yamamoto and Hiraoka, 2003). This suggests that in addition to Rec8- and Moa1-dependent regulation, mating-pheromone signaling is additionally required for the establishment of sister kinetochore association. To explore the requirement of the mating-pheromone signaling, we induced haploid cells to enter meiosis without mating-pheromone signaling by inactivating Pat1 kinase (also known as Ran1), a key negative regulator of meiosis (Iino and Yamamoto, 1985), and examined kinetochore and centromere core association. The pat1-114 temperature sensitive mutation is often used for Pat1 inactivation, but construction of a pat1-114 nda3 double mutant was unsuccessful, possibly because of incompatibility of the pat1-114 mutation with the nda3 cold-sensitive mutation. To overcome this problem, we created an analog-sensitive allele of the pat1 gene (pat1-as2) placed under the mitosis-specific rad21 promoter; the pat1-as2 gene product can be inhibited in the presence of an ATP analog (Cipak et al., 2012; Guerra-Moreno et al., 2012) and the rad21 promoter represses the pat1 gene expression upon nitrogen starvation (https://www.pombase.org/gene/SPCC338.17c), reducing Pat1 activity.

We synchronized haploid pat1-as2 nda3 cells in G1 phase in nitrogen-free medium and incubated them in the presence of the ATP analog 3MB-PP1 at 32°C to induce meiosis. Approximately 80% of cells entered meiosis and formed multiple nuclei (Fig. S6), indicating efficient induction of meiosis. To induce chromosome condensation, the temperature was shifted to 18°C 4 h after induction of meiosis. We found that ∼30% of cells contained condensed metaphase I chromosomes 24 h after the temperature shift (Fig. 6A). Sister kinetochores and centromere cores were separated on ∼15% of these condensed chromosomes (Fig. 6B,C), and non-separated signals were significantly wider (Fig. 6D). These results show that sister kinetochore and centromere core associations are not established in Pat1-induced meiosis. Thus, mating-pheromone signaling is required for sister kinetochore association.

In the present study, we have developed a novel method to directly evaluate the association state of sister kinetochores and centromere cores on metaphase chromosomes in live haploid cells of S. pombe. Using this method, we have shown for the first time that sister kinetochores and centromere cores are positioned separately on associated sister centromeres in mitosis, whereas they associate with each other on the same side of sister centromeres in meiosis I, as in higher eukaryotes. The association state of meiotic kinetochores or centromere cores may differ between haploid and diploid cells. However, given similar sister chromatid segregation patterns in haploid and diploid meiotic cells (Yamamoto and Hiraoka, 2003), it is probable that the association state observed in haploid cells is similar to that observed in diploid cells. Although only about a quarter of sister kinetochores and centromere cores showed clear separation, the increased width of non-separated signals in mitosis, compared with meiosis I, indicates that the remaining sister kinetochores and centromere cores were likely to be separated by a distance shorter than the resolution limit of our microscope system. Variability in the level of separation may result from flexibility in the centromere structure, or non-uniform folding of centromeric chromatin fibers. Although we cannot completely rule out the possibility that hyper-condensation of chromosomes caused separation of sister kinetochores and centromere cores and/or alterations in signal shape, which would otherwise not occur, given the differences observed between mitosis and meiosis I, it is likely that our data accurately reflect the association states of sister kinetochores and centromere cores.

We have shown that the single heterochromatic centromere domain lies underneath centromere cores, irrespective of the centromere core association state. This observation indicates that two pericentromeric, heterochromatic DNA regions, flanking the core DNA region on both sides, form a single heterochromatin domain, on which the centromere core forms, and that sister centromeres associate through these heterochromatin domains. Recent studies have shown that HP1 proteins including Swi6 can gather DNA strands to form condensates by phase separation (Keenen et al., 2021; Larson et al., 2017; Sanulli et al., 2019; Strom et al., 2017). Swi6-dependent phase separation may enable two pericentromeric heterochromatic regions to form the single heterochromatic centromere domain. Taken together, these data suggest that both centromere and kinetochore organization in S. pombe is similar to that of higher eukaryotes.

We have shown that both sister kinetochores and centromere cores are separated in rec8Δ cells. This supports the idea that Rec8-containing cohesin generates sister kinetochore associations by mediating centromere core cohesion (Watanabe et al., 2001). In rec8Δ cells, Rad21-containing cohesin becomes localized at the pericentromeric regions instead of Rec8-containing cohesin and mediates centromere cohesion in meiosis (Yokobayashi et al., 2003). Consistently, we found that in meiotic rec8Δ cells, pericentromeric heterochromatin domains remained associated, whereas sister kinetochores and centromere cores were separated at similar frequencies to those in mitosis. In addition, we have shown that an anaphase cohesin protector, Sgo1, is required solely for cohesin protection and not for sister kinetochore association.

We have also shown that cohesin regulators, Moa1 and Mrc1, are required for the association of sister kinetochores and centromere cores. Epistasis analysis of rec8Δ with moa1Δ or mrc1Δ suggested that Moa1 and Mrc1 regulate kinetochore and core association through cohesin. Perhaps Moa1 and Mrc1 establish cohesin-mediated centromere core cohesion by contributing to DNA-replication-dependent cohesion establishment, as suggested previously (Mayer et al., 2001; Xu et al., 2007; Yokobayashi and Watanabe, 2005). However, Moa1 might also have a cohesin-independent function, as Moa1 alone significantly reduced sister kinetochore and centromere core separation in mitosis (Fig. 5D,E). In addition, regulation by Moa1 or Mrc1 is distinct, because centromere core localization of Rec8 increases in moa1Δ cells but is unchanged in mrc1Δ cells (Fig. 3G) (Yokobayashi and Watanabe, 2005). Moreover, the centromere cohesion state differs in moa1Δ and mrc1Δ cells. In chiasma-lacking cells, sister centromeres only occasionally undergo transient separation in moa1Δ cells during metaphase I, but this is frequent in mrc1Δ cells (Hirose et al., 2011). We also found that significant centromere core separation could be detected in mrc1Δ cells but not in moa1Δ cells (Fig. 3D). These observations could mean that cohesion impairment is milder in moa1Δ cells than in mrc1Δ cells. Alternatively, moa1Δ cells might fail to form proper centromere structures, unlike mrc1Δ cells. Recent studies have shown that cohesin generates DNA loops (Davidson et al., 2019) and clusters (Xiang and Koshland, 2021). Core-localized cohesin might contribute to the folding of centromere core DNA through DNA loop formation and cohesin–cohesin interactions. In moa1Δ cells, the increased localization of cohesin at the centromere core (Yokobayashi and Watanabe, 2005) could lead to abnormal DNA folding, compromising the core structure and impairing sister kinetochore association.

We have shown that production of Rec8 and Moa1 was insufficient to establish sister kinetochore association in mitotic cells (Fig. 5D). Mitotically produced Rec8 leads to the formation of the meiosis-type of cohesin complex (Kitajima et al., 2003), and Moa1 was able to function during mitosis, as shown by its centromere core localization and induction of a reduction in the centromere core localization of Rec8 (Fig. 5A–C). Furthermore, DNA-replication-related cohesin regulators, including Mrc1, are produced and functional in mitosis. Therefore, defective establishment of sister kinetochore association cannot be attributed to a lack of activity in these factors.

We have shown that mating-pheromone signaling is required for the establishment of sister kinetochore association. It was previously shown that mating-pheromone signaling induces telomere clustering-dependent kinetochore delocalization from the centromere, and detachment of centromeres from the spindle pole body (SPB) (Asakawa et al., 2005; Hayashi et al., 2006; Katsumata et al., 2016). Delocalization and subsequent reconstruction of kinetochores might be required for the establishment of sister kinetochore association. As-yet-unknown SPB-associated factors might also inhibit the establishment of sister kinetochore association, resulting in a requirement for centromere detachment. Alternatively, mating-pheromone signaling might cause activation and/or production of as-yet-unidentified factors required for sister kinetochore association. In budding yeast, the monopolin complex, which is able to crosslink kinetochore proteins, is thought to induce sister kinetochore association (Corbett and Harrison, 2012; Corbett et al., 2010; Sarangapani et al., 2014; Tóth et al., 2000). Although the current available evidence does not support a role for monopolin components in sister kinetochore association in S. pombe (our unpublished observation) (Gregan et al., 2007; Rabitsch et al., 2003), unidentified monopolin-like, kinetochore-regulating factors might contribute to sister kinetochore association, and the mating-pheromone signaling might regulate such factors.

In the present paper, we have described a novel method that can be used to evaluate association state of sister kinetochores or centromere cores more accurately than chromosome segregation analysis in live S. pombe cells. In combination with visualization of various loci including centromeres on individual chromosomes, this method is perhaps also useful to evaluate impairments in individual centromere association or locus-specific chromosome arm association in various cohesin-related mutants. Indeed, the chromosomal locus can be visualized on condensed chromosomes using the lacI/lacO system (our unpublished observation) (Nabeshima et al., 1998; Yamamoto and Hiraoka, 2003). Accumulating evidence indicates that age-related maternal chromosome missegregation is associated with separation of sister kinetochores (Chiang et al., 2010; Patel et al., 2016; Sakakibara et al., 2015; Zielinska et al., 2015, 2019). Therefore, understanding the molecular mechanism underlying sister kinetochore association is clinically important. Our novel evaluation method will contribute significantly to our understanding of sister kinetochore association.

Strains used in this study are shown in Table S1. Media and genetic methods used in this study were described previously (Moreno et al., 1991). Plasmid construction was carried out using a PCR-based method described elsewhere (Jacobus and Gross, 2015). Sequences of DNA oligonucleotides used as PCR primers are listed in Table S2.

To induce chromosome condensation in mitotic cells, nda3 cells were grown in YES rich medium to exponential phase at 30°C and subsequently incubated at 18°C. To induce chromosome condensation in mat gene-induced haploid meiosis, nda3 cells containing both mating type genes were grown to stationary phase (∼1×10⁸ cells/ml) in YES rich medium at 30°C and then transferred to EMM medium lacking a nitrogen source (EMM-N). After cells entered G1 through one mitotic division (Yamamoto and Hiraoka, 2003), the culture was incubated at 18°C to induce chromosome condensation. To induce chromosome condensation in pat1-induced haploid meiosis, pat1-as2 nda3 haploid cells were grown in YES rich medium at 30°C. When the cell culture reached a cell density of 2–6×10⁶ cells/ml, cells were synchronized in G1 by incubating the cells in EMM-N medium for 14–16 h at 30°C. G1 cells were induced to enter meiosis at 32°C by incubating cells in fresh EMM-N medium containing 5 µM 3-MB-PP1. After 4 h of incubation at 32°C, the temperature was shifted to 18°C to induce chromosome condensation.

Sister chromatid segregation at meiosis I in diploid zygotes was examined as described previously (Hirose et al., 2011).

To visualize the nuclear membrane, we used a GFP-tagged N-terminal portion of NADPH-cytochrome p450 reductase (Ding et al., 2000). We cloned amino acids 1–275 of the NADPH-cytochrome p450 reductase gene (ccr1-N) and its promoter into a lys1 integration plasmid bearing the GFP gene and the ADH1 terminator, yielding pMH3. pMH3 was introduced into lys1-131, and integrants were selected by lys1⁺ phenotype.

Cells of opposite mating types containing Hht1-Rec8F-mCherry and GFP-tagged Ccr1-N were grown on YES solid medium at 30°C and mixed on ME solid medium. They were induced to enter meiosis by incubation at 25°C for 16–18 h. The cells were suspended in EMM-N liquid medium, and a drop of the suspension was placed on the bottom of 35 mm glass-bottom dishes (Matsunami Glass Ind., Ltd., Osaka, Japan) coated with 5 mg/ml lectin (Sigma-Aldrich Japan, Tokyo, Japan). Cells were observed on a DeltaVision microscope system (GE Healthcare Life Sciences Inc., Tokyo, Japan) equipped with a 60×/1.42 NA Plan Apo oil-immersion objective lens (Olympus, Tokyo, Japan) operated by SoftWoRx software. Time-lapse images of cells were collected on 10 focal planes spaced at 0.5 µm intervals every 5 min using a cooled charge-coupled device (CCD) camera. During image collection, the cells were kept at 25°C in a microscope chamber. The resultant images were processed by deconvolution using SoftWoRx and analyzed using Priism/IVE software (Chen et al., 1996).

Before harvesting cells, they were incubated in YES medium containing 2 mM phenylmethylsulfonyl fluoride (PMSF) for 5 min at 30°C. A total of 1×10⁸ Hht1-Rec8F-mCherry cells, or 4–5×10⁶ Rec8-FLAG or FLAG-Moa1-7dsr cells were harvested and resuspended in 100 µl ice-cold distilled water and immediately heated at 95°C for 5 min. 100 µl of urea buffer [8 M urea, 4% sodium dodecyl sulfate (SDS), 20% glycerol, 0.02% Bromophenol Blue, 1 mM dithiothreitol, 120 mM Tris-HCl, pH 7.4] was added, and the cells were disrupted using glass beads and a cell disrupter. Proteins were extracted by heating at 95°C for 5 min, and cell debris was removed by centrifugation. After the protein concentration of each sample was determined, the cell lysates were stored at −80°C until use. A 50 µg sample of protein was loaded on to a 10% SDS polyacrylamide gel. After electrophoresis, the proteins were detected by western blotting using the following antibodies: rabbit polyclonal anti-DsRed antibody (1:3000; Takara Bio, Inc.), mouse monoclonal anti-PSTAIR antibody (1:200,000; Sigma-Aldrich Corp.), mouse monoclonal anti-FLAG antibody (clone M12; 1:1000; Sigma-Aldrich Corp.), and mouse anti-Cdc13 antibody 6F11/2 (1:4000, Abcam plc).

To analyze the association state of sister kinetochores and centromere cores in mitosis, cells containing condensed chromosomes were resuspended in ice-cold EMM medium. A 1.8 µl volume of the cell suspension was placed on a 40×50 mm cover glass and covered with an 18×18 mm coverslip. Images of cells were collected on 11 focal planes spaced at 0.3 µm intervals using an IX71 inverted microscope equipped with a cooled CCD camera (CoolSNAP-HQ2; Nippon Roper Co. Ltd., Tokyo, Japan) and a 100×/1.40 NA Plan Apo oil-immersion objective lens (Olympus, Tokyo, Japan). During observation, cells were kept below 18°C. The resultant images were processed by deconvolution and analyzed using MetaMorph (version 7; Molecular Devices Japan, Tokyo, Japan) or ImageJ software (Schneider et al., 2012). Distances between separated signals were obtained from three-dimensional coordinates, which were determined using a multi-dimensional motion analysis module in the MetaMorph software.

To visualize kinetochores, we used endogenously GFP- or mCherry-tagged Nuf2 strains (Nabetani et al., 2001). To visualize the centromere core, we used a previously reported endogenously GFP-tagged Cnp1 strain (Takayama et al., 2008), or an mCherry-tagged Cnp1 strain in which we introduced mCherry-cnp1⁺ at an ectopic locus. We generated an mCherry-cnp1⁺ fusion under the cnp1⁺ promoter and introduced it into an integration vector, which contains the nmt1 terminator and a partial lys1 gene as a selection marker (Chikashige et al., 2004), yielding integration plasmid pUK6. Similarly, we cloned the mCherry-cnp1⁺ fusion gene and cnp1⁺ promoter into integration vector pTO2 (Yoshida et al., 2013) and pDUAL-HFG1 (Matsuyama et al., 2004), yielding pAK16 and pAK22, respectively. pUK6, pAK16 or pAK22 was introduced into a lys1-131, aur1⁺ or leu1-32 strain, respectively, and integrants were selected by lys⁺, leu⁺ or aureobasidine A-resistant phenotype.

To visualize metaphase I chromosomes, we cloned a histone H3 gene, hht1⁺, and the mCherry tag into integration vector pTO2 harboring the spo5 promoter and DSR (Harigaya et al., 2006). Subsequently, we inserted the rec8 gene portion between hht1⁺ and mCherry, yielding pMN3. pMN3 was introduced into an aur1⁺ strain, and integrants were selected by aureobasidine A resistance. For CFP visualization of metaphase I chromosomes, we inserted a DNA fragment encoding the hht1⁺-rec8F-mCherry fusion into pDUAL-HFG1, together with the spo5 promoter and the DSR fragments, and replaced mCherry with CFP, yielding pAK19. For CFP visualization of mitotic metaphase chromosomes, we cloned a DNA fragment encoding the rad21⁺ gene and its promoter, and a CFP gene fragment, into pDUAL-HFG1, yielding pAK11. pAK19 and pAK11 were introduced into leu1-32 strains, and integrants were selected by leu⁺ phenotype.

For mitotic production of GFP-tagged Rec8, we introduced a DNA fragment encoding the rec8⁺ gene and its promoter, GFP and the ADH1 terminator (Wach et al., 1997) into a lys1 integration vector, yielding pYI35. For mitotic production of FLAG-tagged Rec8, we introduced a 5FLAG fragment together with rec8 and its promoter into an aur1^r integration vector, pTO2, yielding pKI10. pYI35 and pKI10 were introduced into lys1-131 and aur1⁺ strains, respectively, and integrants were selected by lys1⁺ or resistance to aureobasidine A.

For mitotic production of Moa1, we first cloned the moa1⁺ gene by replacing the α₂-tubulin (atb2⁺) gene with moa1⁺ in pMY53, a plasmid encoding the mCherry-tagged atb2⁺ gene under the nda3 promoter (Yoshida et al., 2013). We then eliminated the core and putative DSR sequences from moa1⁺ (moa1-7dsr) using the KOD-Plus-Mutagenesis kit (TOYOBO Co., Ltd, Osaka, Japan), yielding pAK17. We further replaced the mCherry gene with 5FLAG, yielding pAK18. pAK17 and pAK18 were transformed into a lys1-131 strain, and integrants were selected by lys⁺ phenotype.

To construct the pat1-as2 strain, we cloned the pat1 gene into the lys1 integration vector pYC36 (Chikashige et al., 2004). We then introduced a L95A mutation into the pat1 using the KOD-Plus-Mutagenesis kit, generating the pat1-as2 gene. We then fused the rad21 promoter (∼700 bp upstream region of the rad21 gene) and the pat1-as2 gene by cloning both into pYC36, yielding pYI14. pYI14 was transformed into a lys1-131 strain, and integrants were selected by lys1⁺ phenotype. Subsequently, we replaced the endogenous pat1⁺ gene in the integrant with the kan^r resistance marker (Bähler et al., 1998), by transforming a PCR cassette encoding the kan^r gene flanked by the pat1 promoter and terminator into a pat1⁺ strain. The DNA fragment was amplified using a plasmid generated by inserting the pat1 promoter and terminator into pFA6a-kanMX (Bähler et al., 1998), as a template.

Centromere localization of Rec8 and Moa1 was analyzed by chromatin immunoprecipitation using an M2 anti-FLAG antibody (1:600; Sigma-Aldrich Japan K. K., Tokyo, Japan), as described previously (Yamada et al., 2013).

We thank Shigeaki Saitoh for a CFP plasmid; Mikiaki Hiraoka for Ccr1 cloning; Yumi Hinohara for her contribution to the very early stage of this study; former lab members for constructing various plasmid vectors; the Yeast Genetic Resource Center for strains; and Akira Shinohara for the use of the DeltaVision microscope system.

The peer review history is available online at https://journals.biologists.com/jcs/article-lookup/doi/10.1242/jcs.259102.