Cell cycle control of spindle pole body duplication and splitting by Sfi1 and Cdc31 in fission yeast

The mitotic spindle is a complex microtubule‐based structure responsible for the accurate segregation of chromosomes. Its assembly and function are therefore under strict and robust regulatory mechanisms. Spindle assembly starts during prophase when duplicated centrosomes move apart to establish a bipolar structure of antiparallel overlapping interpolar microtubules. Although spindle assembly can proceed in the complete absence of centrosomes in animal cells (Khodjakov et al., 2000; Basto et al., 2006), centrosomes, which nucleate and anchor microtubules, contribute to spindle assembly efficiency and robustness. They are, for instance, essential during fast cycles of early development (Stevens et al., 2007). They are generated by conservative duplication of pre‐existing single‐copy centrosomes to limit their number to exactly two per cell after duplication and prevent deleterious effects of supernumerary centrosomes in spindle assembly, organ morphogenesis and tumor formation (Basto et al., 2008; D'Assoro et al., 2002; Marthiens et al., 2013).

Spindle pole bodies (SPBs) are yeast plaque‐like organelles attached to the nuclear envelope that are functionally similar to centrosomes of animal cells. Like centrosomes, they are generated by conservative duplication of pre‐existing SPBs. In budding as well as fission yeast, the daughter SPB assembles at the tip of a mother SPB appendage called the bridge, which maintains the association of the duplicated SPBs until mitotic onset. Duplicated SPBs inserted in the nuclear envelope nucleate intranuclear microtubules to initiate bipolar spindle assembly (Adams and Kilmartin, 2000; Ding et al., 1997; Jaspersen and Ghosh, 2012; Jaspersen and Winey, 2004; Lim et al., 2009; Uzawa et al., 2004).

In Saccharomyces cerevisiae, the SPB bridge was shown to contain two major evolutionarily conserved proteins, Sfi1 and Cdc31 (yeast centrin), that are essential for SPB duplication and bipolar spindle formation (Baum et al., 1986; Kilmartin, 2003; Li et al., 2006; Paoletti et al., 2003; Spang et al., 1993). Their orthologs play important roles at centrosomes and cilia in a variety of eukaryotes (Avasthi et al., 2013; Azimzadeh et al., 2009; Azimzadeh et al., 2012; Balestra et al., 2013; Dantas et al., 2012; Delaval et al., 2011; Gogendeau et al., 2007; Gogendeau et al., 2008; Jerka‐Dziadosz et al., 2013; Stemm‐Wolf et al., 2013; Wei et al., 2014; Zhang and He, 2012).

Crystallographic studies performed on Sfi1 and Cdc31 have established that Sfi1 molecules form an extended α‐helix containing repeats on which the Ca²⁺‐binding calmodulin‐like Cdc31 molecules bind. Moreover, based on electron microscopy studies, it has been proposed that, to form a half‐bridge, these molecules assemble into a parallel array with Sfi1 N‐termini attached to the mother SPB. In this model, daughter SPB assembly is initiated by half‐bridge duplication, achieved by assembly of a second array of Sfi1 molecules, creating a new half‐bridge, anti‐parallel to the parental half‐bridge and connected to it by interactions between Sfi1 C‐termini (Anderson et al., 2007; Li et al., 2006).

Very recent studies further show that the coordination of half‐bridge duplication and bridge splitting with cell cycle progression is controlled by the cyclin‐dependent kinase Cdk1 (encoded by cdc2). Cdk1 phosphorylates the Sfi1 C‐terminus to promote the segregation of the two half‐bridges at mitotic entry (Anderson et al., 2007; Avena et al., 2014; Elserafy et al., 2014). This event is also controlled to a lesser extent by the polo‐like kinase Cdc5 that targets additional sites on the Sfi1 C‐terminus (Elserafy et al., 2014). Importantly, Cdk1‐dependent phosphorylation of Sfi1 C‐terminus was also shown to prevent SPB reduplication after half‐bridge splitting (Avena et al., 2014; Elserafy et al., 2014). This inhibition is relieved by the Cdc14 phosphatase that dephosphorylates the Sfi1 C‐terminus at mitotic exit (Avena et al., 2014; Elserafy et al., 2014). Finally, SPB duplication is also regulated by Mps1 kinase, the activity of which is necessary for half‐bridge duplication (Elserafy et al., 2014). Interestingly, one G1 target of Mps1 is Cdc31. Phosphorylation of Cdc31 by Mps1 was shown to be necessary for SPB duplication. Whether this phosphorylation is sufficient to promote half‐bridge duplication remains unknown (Araki et al., 2010).

Like budding yeast Cdc31, fission yeast Cdc31 is a component of the SPB half‐bridge and bridge, and is essential for SPB duplication and bipolar spindle assembly (Paoletti et al., 2003). Sfi1 is also conserved in fission yeast (Kilmartin et al., 2003), and a very recent study showed that Sfi1 is required for SPB duplication (Lee et al., 2014). Nevertheless, when half‐bridge duplication takes place or how half‐bridge duplication and bridge splitting are coordinated with cell cycle progression in fission yeast remains largely unknown. In this study, we have analyzed how Sfi1 and Cdc31 function is regulated in fission yeast. We show that Cdc31 is necessary for the assembly of Sfi1 parallel arrays and that it functions at mitotic exit when Sfi1 accumulates quickly at the SPB to promote SPB duplication. In addition, Cdc31 is phosphorylated on an N‐terminal Cdk1 consensus site to promote bridge splitting and SPB separation at mitotic entry.

Fission yeast Sfi1 localizes to SPBs like its budding yeast counterpart (Kilmartin, 2003; Lee et al., 2014). Because we could not determine Sfi1 ultrastructural localization by immunoelectron microscopy (immuno‐EM), we developed a fluorescent light microscopy method to verify whether Sfi1 associates with the SPB half‐bridge and bridge like Cdc31 (Paoletti et al., 2003); cells expressing Sfi1 tagged C‐terminally with mRFP, a fully functional fusion that can replace Sfi1 at its endogenous locus, and another SPB component tagged with GFP were fixed with methanol to prevent SPB movements. Sample stability was assessed by performing two successive acquisitions of red and green signal and checking the perfect superimposition of the two sets of red or green images. We finally used dual color fluorescent beads to check the registration of red and green signals in the microscopy setup used. Spatial shifts between red and green channels were less than 1 pixel in the field of view.

Using this method, we could observe that in short G2 or mitotic cells, Sfi1–mRFP staining was always closely apposed to but never perfectly colocalized with calmodulin (Cam1), which binds the nuclear side SPB component Pcp1 (Flory et al., 2002; Fong et al., 2010; Moser et al., 1997) or other SPB core components such as Sid2, which associates with the external surface of the SPB (Sparks et al., 1999), or the transmembrane SUN domain protein Sad1 (Hagan and Yanagida, 1995) (Fig. 1A,C,D,G). We then measured fluorescence intensity across SPBs using the linescan tool of Metamorph software. In early G2, peak intensities of Sfi1 never coincided with the peak intensities of core SPB markers and were generally separated by 2 pixels representing ∼130 nm theoretically (supplementary material Fig. S1A–D).

At later stages of G2, as judged by cell length, two dots corresponding to duplicated SPBs could be discriminated with the three SPB core markers. These dots corresponded to two peaks generally separated by 4 pixels on fluorescence intensity profiles measured by using the Metamorph linescan tool, representing ∼260 nm theoretically. In contrast, a single peak was observed with Sfi1–mRFP, between the peaks of the SPB core markers, consistent with Sfi1 localization at the bridge connecting duplicated SPBs (Fig. 1A,C,D,G).

Immunodetection of Cdc31 further showed that Sfi1 colocalized perfectly with this half‐bridge component throughout the cell cycle and that its peak intensity coincided perfectly with that of Cdc31 on fluorescence intensity profiles across SPBs (Fig. 1B,G; supplementary material Fig. S1B).

To confirm these results, we finally performed colocalization studies between Sfi1–mCherry and Sid4–GFP or between Sfi1–GFP and Sid4–mCherry, inverting the color channels used to visualize the two proteins (Fig. 1E,F; supplementary material Fig. S1E,F). This experiment confirmed that Sfi1 does not perfectly colocalize with SPB core components and sits between duplicated SPBs after duplication, independent of fluorescence wavelength. We therefore conclude that Sfi1 is a genuine half‐bridge and/or bridge component in S. pombe.

We next determined how Sfi1 targeting to the half‐bridge is controlled. To do so, we produced a series of Sfi1 constructs fused to GFP at the C‐terminus that we expressed in wild‐type cells in addition to endogenous Sfi1 (Fig. 2A). We found that deleting the Sfi1 N‐terminus (amino acids 1–188), but not Sfi1 C‐terminus (amino acids 766–840) strongly reduced Sfi1 targeting to the half‐bridge but did not abolish it completely. When expressed alone, the central region of Sfi1 containing Cdc31‐binding repeats (amino acids 189–765) was targeted to the half‐bridge at similar low levels. Finally, the N‐terminal or C‐terminal domains did not associate with the half‐bridge on their own. Thus, efficient targeting of Sfi1 to the half‐bridge depends on the Cdc31‐binding central domain combined with the Sfi1 N‐terminus.

This result suggested that Cdc31 might be required for Sfi1 association with the half‐bridge. To test this further, we used the pnmt*‐E147K‐cdc31 mutant in which E147K‐Cdc31 can be depleted upon addition of thiamine at 36°C, blocking SPB duplication and inducing the assembly of monopolar spindles (Paoletti et al., 2003). We found that Sfi1–GFP, now expressed from the endogenous Sfi1 locus, was lost from half‐bridges upon Cdc31 depletion, whereas control cells maintained strong Sfi1–GFP staining (Fig. 2B). This demonstrates that Cdc31 is essential for the stability of Sfi1 half‐bridge arrays.

We next wanted to test whether Cdc31 localization to the half‐bridge reciprocally requires Sfi1. Given that sfi1 is essential for cell survival, we produced a thermo‐sensitive mutant of sfi1 by random mutagenesis of the Sfi1 N‐terminus. This mutant carries a frame shift at valine 10 and a point mutation changing asparagine 171 to aspartic acid. Why this mutant is thermosensitive is unclear. One possibility is that it might initiate translation at methionine 17 more efficiently at 25°C than at 36°C.

Immunostaining of sfi1‐1 cells grown for 4 hours at 36°C revealed some G2 cells lacking Cdc31 staining as well as mitotic cells with monopolar spindles lacking Cdc31 on the single spindle pole (Fig. 3A). We conclude that there is a co‐requirement for Cdc31 and Sfi1 to assemble into stable half‐bridge arrays.

To better characterize the sfi1‐1 mutant, we performed time‐lapse imaging in sfi1‐1 cells expressing GFP‐tagged tubulin. This confirmed that ∼90% of total spindles were monopolar after 4 hours at 36°C (Fig. 3B,C). These monopolar spindles led to a prolonged mitosis, likely as a result of spindle checkpoint alert (Weiss and Winey, 1996), but eventually disassembled (Fig. 3C). Septation started before spindle disassembly and gave rise to cut and multi‐septation phenotypes (supplementary material Fig. S2A), similar to the phenotype of the pnmt*‐cdc31‐E147K mutant (Paoletti et al., 2003).

We next used Ppc89–GFP, a key structural component of the SPB core (Rosenberg et al., 2006) to verify the duplication status of the SPB. With this marker, we could observe the gradual accumulation of duplicated SPB pairs in the population of wild‐type cells of increasing length that progress through G2 (supplementary material Fig. S2B). SPB pairs also accumulated gradually with cell size in the cdc10‐V50 mutant blocked at the G1/S transition for 3 or 5 hours at 36°C, or in cells blocked in S phase by treatment with hydroxyurea for 5 hours at 25°C (supplementary material Fig. S2B). These results are in agreement with recent electron microscopy observations showing that single and duplicated SPBs can be found in short G2 cells (Höög et al., 2013) but also confirm that SPB duplication does not require S phase entry (Uzawa et al., 2004). They show that Ppc89–GFP constitutes a good tool to monitor SPB duplication.

Analysis of mitotic cells expressing Ppc89–GFP and mCherry‐tagged tubulin (Fig. 3D) showed that sfi1‐1 mutant cells failed to duplicate their SPBs at 36°C; more than 90% of mitotic sfi1‐1 cells with monopolar spindles presented a single SPB labeled with Ppc89–GFP. In contrast, in the sad1‐1 mutant that also makes monopolar spindles (Hagan and Yanagida, 1995), we observed two Ppc89–GFP‐stained SPBs at the single pole at high frequency. This confirmed that duplicated but not segregated SPBs can be distinguished easily using this marker in mitosis and that SPB duplication fails in sfi1‐deficient cells. It also indicates that Sad1, in contrast to Sfi1, might contribute to the SPB cycle after its duplication phase. Finally, wild‐type cells expressing Ppc89–GFP showed two segregated single SPBs at the poles of bipolar spindles. We also observed some cells with more than two Ppc89–GFP stained structures and multipolar spindles at low frequency (<4% mitotic cells). This suggests that tagging Ppc89 with GFP might not be fully innocuous at 36°C. The incidence of this type of defect was increased in sfi1‐1 mutants (∼9% mitotic cells) suggesting that Ppc89–GFP might be synthetic lethal with the sfi1‐1 mutant.

SPB duplication defects in the sfi1‐1 mutant were confirmed by analyzing SPB status in late G2, as judged by the presence of interphase microtubules and a cell length ranging between 10 and 12 µm. Whereas side‐by‐side duplicated SPBs could be distinguished in the vast majority of wild‐type cells (∼75%), they could be identified in a much reduced proportion of sfi1‐1 cells (∼27%), confirming the results obtained in mitotic cells (supplementary material Fig. S2C).

We finally analyzed the SPB status in the sfi1‐1 mutant in post‐mitotic cells using Sid2–GFP as a marker for the SPB (Fig. 3E; supplementary material Fig. S2D). Indeed, Sid2 also localizes to the division site in post‐mitotic cells, allowing their identification. We found that 80% of post‐mitotic sfi1‐1 mutant cells grown at 36°C for at least 3 hours presented a single SPB, in contrast to control cells that showed duplicated and segregated SPBs. The phenotype of the sfi1‐1 mutant was again distinct from that of the sad1‐1 mutant, where two juxtaposed Sid2–GFP‐labeled SPBs could be easily discriminated (Fig. 3E; supplementary material Fig. S2D).

We conclude that Sfi1 is essential for SPB duplication in fission yeast, in agreement with its Cdc31‐dependent localization to the half‐bridge. This result is in agreement with the new SPB assembly defects identified in the sfi1‐M46 mutant when cells inherit insufficient Sfi1 (Lee et al., 2014).

We next wondered when half‐bridge duplication takes place during cell cycle progression to allow the assembly of the daughter SPB. To address this question, we first used a quantitative approach to detect variations in Sfi1 amounts at the SPB that might signal half‐bridge duplication. Live‐cell imaging of cells expressing Sfi1–GFP from the sfi1 endogenous locus by spinning‐disc confocal microscopy with low laser power, combined with SPB automatic tracking, allowed us to record Sfi1 intensity at SPBs for more than a complete cell cycle (Fig. 4A–C).

In agreement with observations made recently by Lee et al. (Lee et al., 2014), two phases of Sfi1–GFP intensity increase were detected after SPB segregation at mitotic entry. The first phase started just a few minutes after mitotic entry, in early anaphase, and lasted until mid‐septation. Accordingly, we could measure that by the end of anaphase, Sfi1–GFP intensity had already increased by ∼1.5‐fold compared to mitotic entry. In contrast, GFP–Cam1 intensity, used as a control, remained constant during the same period (Fig. 4D). Similarly, Sid4–GFP or Ppc89–GFP, two central components of the SPB core, displayed a limited increase during this period, showing that the fast intensity increase observed at mitotic exit is specific to Sfi1. The second phase started after septation and lasted throughout G2 phase until the next round of mitosis.

Interestingly, Sfi1–GFP kept accumulating at the SPB when G2 phase lasted longer in a cdc25‐22 mutant grown at 36°C (Fig. 4E). Sfi1–GFP accumulation was also maintained upon overexpression of Cdc18 (Fig. 4F), which blocks cells in an abortive S phase and prevents SPB early maturation, i.e. extension of SPB plaques after duplication (Uzawa et al., 2004). This indicates that Sfi1–GFP accumulates at SPBs independently of SPB growth in size.

To determine which phase of Sfi1 accumulation corresponds to half‐bridge duplication, we defined when Sfi1–Cdc31 function is necessary for SPB duplication. Instead of the sfi1‐1 mutant, which still exhibits monopolar spindles at low frequency under non‐restrictive conditions and precludes determination of when Sfi1 function is accomplished, we used the cdc31‐8 thermo‐sensitive mutant isolated as described previously (Ohta et al., 2012), in which the cdc31 open reading frame (ORF) carries a substitution changing arginine 134 to glycine and the 3′UTR is replaced by the ADH terminator.

Fast temperature shift of the cdc31‐8 mutant was performed under the microscope (Velve Casquillas et al., 2011). The cell cycle stage at the time of temperature shift was defined according to microtubule organization, using the fluorescent tubulin marker GFP–Atb2 (Atb2 is also known as Tub1) (Fig. 5C). This experimental setup allowed us to define at which point of the cell cycle cdc31‐8 inactivation led to a monopolar spindle phenotype during the next round of mitosis (Fig. 5A,B).

We found that cells shifted during late septation or G2 (i.e. during the second phase of Sfi1 accumulation) always produced bipolar spindles in the following mitosis, indicating that the half‐bridge was already duplicated at these stages (Fig. 5D,E). Cells shifted in early septation with a post‐anaphase array of microtubules showed only a low percentage of monopolar spindles during the next mitosis. In contrast, cells shifted during mitosis (i.e. before or concomitant to the first phase of Sfi1 accumulation) showed high levels of monopolar spindles during the next mitosis (Fig. 5D,E).

In a control experiment, we used the cdc25‐22 wee1‐as8 cdc31‐8 strain (Tay et al., 2013). This strain can be blocked in G2 for several hours at the restrictive temperature of 36°C, due to cdc25 inactivation, and released synchronously into mitosis by addition of the ATP analog 3Br‐PP1 that inactivates the wee1‐as8 allele and allows the cells to overcome the cdc25‐22 block. We observed that >90% of cells blocked in G2 at 36°C for up to 4 hours showed bipolar spindles when released into mitosis at 36°C (Fig. 5F). This shows that the differential outcomes in mitotic spindle organization observed upon shift of the cdc31‐8 mutant to 36°C are not due to long term incubation at the restrictive temperature but are truly specific to the cell cycle stage at which the shift was performed. This data strongly supports the hypothesis that the first phase of Sfi1 accumulation in late mitosis/early septation represents half‐bridge duplication.

S phase takes place during septation in wild‐type fission yeast, such that daughter cells are already in G2 when the septum is cleaved. To determine whether half‐bridge duplication occurs before S phase onset, we combined cdc31‐8 with the wee1‐as8 strain to delay S phase onset by addition of 3Br‐PP1. Indeed, wee1 inactivation generates very short cells following division that need to grow before committing to a new cell cycle and entering S phase (Fantes and Nurse, 1978; Sveiczer et al., 1996). In this experiment, owing to the incompatibility between 3BrPP1 and polydimethylsiloxane (PDMS) used to build microfluidic devices for fast temperature shift, cells were instead transferred to lectin‐coated glass‐bottomed dishes, treated with 3Br‐PP1 for 6 hours at 25°C to inactivate Wee1 and reduce cell size, then transferred to the microscope incubator heated at 36°C where imaging was started within a few minutes, followed by addition of hot medium to ensure fast completion of the temperature shift. Cell cycle stage was recorded at this point according to microtubule organization. With this experimental setup, cells might have experienced temperature shift slightly earlier than recorded. Nevertheless, >90% of cells shifted at 36°C after completion of septation displayed bipolar spindles during the next mitosis (supplementary material Fig. S2E,F), indicating that cdc31 function is already accomplished in short G1 cells. We conclude that half‐bridge duplication takes place immediately at mitotic exit when Sfi1–GFP accumulates quickly at the SPB.

Another striking observation from the quantitative analysis of Sfi1–GFP intensity, also reported by Lee et al. (Lee et al., 2014), is that upon segregation at mitotic onset, the two SPBs exhibit ∼30% fluorescence intensity of the pre‐mitotic SPB, instead of the expected 50% (Fig. 6A,B). This feature is specific for Sfi1 because we did not observe it with GFP–Cam1, for which the fluorescence intensity of segregated SPBs after mitotic entry was ∼50% of the pre‐mitotic SPB, nor for Sid4–GFP or Ppc89–GFP (Fig. 6B). This indicates that approximately a third of Sfi1 molecules dissociate from the bridge at mitotic onset. Whether Sfi1 molecules are dispersed in the cytoplasm or are degraded is unclear.

Interestingly, we noticed that Sfi1–GFP intensity started to decrease a few minutes before bridge splitting (Fig. 4B,C). This suggested that the partial dissociation of Sfi1 from the half‐bridge might be triggered by a mitotic kinase and could destabilize the bridge to favor bridge splitting. Searching for potential Cdk1 sites, we found that Cdc31 contains two Cdk1 consensus phosphorylation sites in its N‐terminal part (serine 15 and threonine 17). To determine whether these sites were phosphorylated in vivo during mitosis, Cdc31 was co‐immunoprecipitated from a mitotic cell extract with a soluble N‐terminal fragment of Sfi1 that contains a single Cdc31 binding repeat and is more favorable for quantitative immunoprecipitation than full‐length Sfi1. Mass spectrometric analysis showed that the serine 15 in a Cdk1 consensus site was phosphorylated in vivo during mitosis (Fig. 6C; supplementary material Fig. S3).

A serine‐to‐alanine mutation was introduced at position 15 of Cdc31 at the cdc31 locus to inhibit phosphorylation. The cdc31‐S15A strain was viable, indicating that this mutation does not abolish Cdc31 function. Phosphorylation of the Cdc31 peptide containing Cdk1 consensus sites was abolished (Fig. 6D), confirming that serine 15 is phosphorylated in vivo but threonine 17 is not. Furthermore, Cdc31 purified from bacteria was a substrate of mammalian cyclin‐B–CDK1 in vitro (Fig. 6E). Finally, Cdc31 phosphorylation was fully abolished in vivo upon mutation of serine 15 to alanine (Fig. 6F). These results indicate that Cdc31 might be a target of mitotic Cdk1. We noted, however, that serine 15 remains phosphorylated 1 hour after release of the mitotic block, indicating that Cdc31 phosphorylation is not strictly restricted to mitosis.

Interestingly, analysis of microtubules using the fluorescent marker GFP–Atb2 revealed that cdc31‐S15A cells assembled transient monopolar spindles (Fig. 7A,B). The monopolar spindle phase lasted for variable amounts of time (Fig. 7B) but resolved into a bipolar spindle. Accordingly, >50% of monopolar spindles were resolved into bipolar spindles in 1‐hour‐long movies (supplementary material Fig. S4C). This result was confirmed using Plo1–GFP that strongly associates with spindle poles during mitosis (Mulvihill et al., 1999). Plo1–GFP staining showed two juxtaposed dots in early mitosis, eventually segregating after a variable delay when the spindle became bipolar (supplementary material Fig. S4A,B).

The phenotype of the cdc31‐S15A mutant is clearly distinct from that of cdc31‐8 or sfi1‐1 mutant phenotypes where SPB duplication is compromised. Accordingly, analysis of Sid2–GFP revealed two segregated SPBs in 97% of post‐mitotic cdc31‐S15A cells (supplementary material Fig. S4D,E). We thus conclude that the cdc31‐S15A mutant is not defective for SPB duplication but displays a delay in separation of duplicated SPBs at mitotic onset.

Finally, quantitative analysis of Sfi1–GFP in the cdc31‐S15A mutant revealed that close to 100% of the Sfi1 fluorescence intensity remained associated with SPBs at mitotic entry compared to only around two thirds in the wild‐type strain (Fig. 7C,D). Surprisingly, an unequal distribution of Sfi1–GFP between the two SPBs was also evident after delayed SPB segregation (Fig. 7E). This asymmetry could result from an asymmetrical rupture of the bridge when proper regulation of bridge splitting is compromised or from an asymmetry in bridge biogenesis.

Analysis of Sfi1–GFP fluorescence intensity in the cdc31‐S15A mutant showed a first increase at mitotic exit that might correspond to half‐bridge duplication. Remarkably, although GFP intensity also increased during G2 on low‐intensity SPBs, GFP intensity remained stable on high‐intensity SPBs (Fig. 7F). Thus, the asymmetry in half‐bridges inherited by the two daughter cells is compensated for at least in part by a differential accumulation of Sfi1 molecules in the bridge during the next G2 phase. Taken together, our data reveals that Cdc31 phosphorylation on serine 15 regulates Sfi1 dissociation from the bridge at mitotic onset to promote SPB separation and allow immediate assembly of a bipolar spindle.

In this study, we have shown that Sfi1 and its partner protein Cdc31 together control two key steps of the SPB duplication cycle in fission yeast: half‐bridge duplication, which permits in turn the assembly of the daughter SPB, as well as SPB segregation at mitotic onset, necessary for bipolar spindle assembly (see Fig. 8). Sfi1–Cdc31 function is therefore largely conserved between budding and fission yeast (Anderson et al., 2007; Avena et al., 2014; Elserafy et al., 2014). Whereas the SPB duplication function was reported previously (Lee et al., 2014; Paoletti et al., 2003), our data provides the first evidence for a role of Cdc31 in the temporal control of SPB segregation and spindle bipolarity establishment.

We found that Cdc31 and Sfi1 are both required for the assembly of half‐bridge arrays. Furthermore, our analysis of a new cdc31 mutant, cdc31‐8, using a fast microfluidic‐based temperature shift system, combined with quantitative analysis of Sfi1 amounts at the SPB along cell cycle progression, indicates that half‐bridge duplication is initiated during anaphase when the first phase of Sfi1 accumulation starts. Whether an SPB precursor or satellite assembles immediately upon half‐bridge duplication remains to be determined.

Using Ppc89–GFP as a marker to follow daughter SPB biogenesis, we could also confirm that SPBs are committed to duplication before S phase entry because duplication was not blocked in the cdc10‐V50 mutant, as reported previously (Uzawa et al., 2004). Nevertheless, side‐by‐side duplicated SPBs were rarely detected in early G2 using this tool, and they gradually accumulated in the cell population during G2 phase. Although some duplicated SPBs might not be recognized using light microscopy owing to their orientation or when the daughter SPB size is too small, this result is in good agreement with electron microscopy data showing the presence of duplicated and nonduplicated SPBs in short G2 cells (Höög et al., 2013). It might also explain why duplication figures were initially only identified in long G2 cells (Ding et al., 1997).

A second phase of Sfi1 accumulation lasts all along G2 phase. We found that it persists longer if progression into mitosis is blocked. This accumulation does not require S phase exit and is therefore independent of SPB growth in size or maturation because it persisted upon overexpression of Cdc18 (Uzawa et al., 2004). Thus, it seems that the bridge and the SPB core are submitted to distinct regulations during cell cycle progression.

We have also shown that SPB separation at mitotic entry is regulated by Cdc31 phosphorylation on a Cdk1 consensus site on serine 15. This phosphorylation is necessary for the partial dissociation of Sfi1 from the bridge and immediate SPB separation at mitotic entry. Although we could show that Cdc31 is a substrate for CDK1–cyclin‐B in vitro, serine 15 phosphorylation does not appear to be restricted to mitosis in vivo because we could observe this phosphorylation in post‐mitotic extracts. New tools will be necessary to establish the precise pattern of phosphorylation of serine 15 according to cell cycle progression. At this stage, we can only speculate that serine 15 could either be protected from dephosphorylation by the Cdc14‐like phosphatase Clp1 (also known as Flp1) that antagonizes CDK1 phosphorylation at mitotic exit (Cueille et al., 2001; Trautmann et al., 2001) or that it could be the target of additional kinases.

Another point that remains to be elucidated is the exact effect of serine 15 phosphorylation on Sfi1–Cdc31 arrays. Because Sfi1 association with the half‐bridge requires Cdc31, phosphorylation of Cdc31 on serine 15 could modulate its interaction with Sfi1 to destabilize the half‐bridge at mitotic entry. However, the Cdc31 N‐terminus might not participate in the Sfi1–Cdc31 interaction in budding yeast, as judged by the crystal structure obtained for a fragment of budding yeast Sfi1 in complex with budding yeast Cdc31, where the Cdc31 N‐terminus was not visible and hence not engaged in interactions with the Sfi1 helix (Li et al., 2006). In addition, a Cdc31 mutant lacking N‐terminal residues can replace wild‐type Cdc31 in S. cerevisiae (Li et al., 2006), similar to the cdc31S15A mutant in S. pombe. However, the timing of SPB separation was not reported in this budding yeast mutant.

An alternative role for the phosphorylation of fission yeast Cdc31 N‐terminus is to modulate the oligomerization of Sfi1–Cdc31 complexes in parallel arrays. Additional biochemical work might solve this question if fission yeast Sfi1 can be purified in a soluble form. Interestingly, in contrast to the budding yeast half‐bridge, the fission yeast half‐bridge is a thick three‐dimensional (3D) appendage (Ding et al., 1997; Paoletti et al., 2003; Uzawa et al., 2004). Moreover, the Cdc31 N‐terminus is longer in S. pombe than in S. cerevisiae, where it lacks a Cdk1 consensus site. It is therefore possible that interactions between Sfi1–Cdc31 complexes involve additional interactions and regulations in fission yeast compared to S. cerevisiae.

It is also worth noting that SPB separation is delayed but not fully blocked when Cdc31 phosphorylation on serine 15 is inhibited. Thus, other regulatory mechanisms might function in parallel to Cdc31 phosphorylation to promote bridge splitting and release the two duplicated SPBs. One possibility is that it might involve the phosphorylation of the Sfi1 C‐terminus as demonstrated recently in budding yeast (Avena et al., 2014; Elserafy et al., 2014). Sfi1 C‐terminus contains a Cdk1 consensus site but its mutation to alanine led to SPB duplication defects (A.M., A.P., unpublished data). Because the Sfi1 C‐terminus is very short in fission yeast (74 amino acids), these defects could possibly stem from global folding defects that might prevent the establishment of interactions between Sfi1 C‐termini. In any case, this precludes drawing conclusions on the possible role of this consensus Cdk1 site in controlling SPB separation.

Alternatively, bridge splitting in the cdc31‐S15A mutant could occur by non‐regulated rupture when the kinesin Cut7 (a homolog of mammalian Eg5, also known as KIF11), associated with SPBs (Hagan and Yanagida, 1992), starts exerting the forces necessary to slide antiparallel microtubules apart to separate the two spindle poles. Such a mechanism would be conceptually similar to the process of centriole disengagement during the development of Caenorhabditis elegans embryos, where separase‐driven centriole separation becomes dispensable when the forces exerted by microtubule asters are strong enough to induce disengagement (Cabral et al., 2013).

To conclude, we have shown that fission yeast Cdc31 is an important regulator of Sfi1 arrays and is the target of modulation by phosphorylation with an important impact on spindle bipolarity and cell division. Cdc31 orthologs, the centrins, do not contain an N‐terminal CDK1 consensus site but are subject to phosphorylation in physiological and pathological situations such as cancer (Araki et al., 2010; Lutz et al., 2001; Yang et al., 2010). It will be of great interest to determine whether centrin phosphorylation by CDKs or alternative kinases can modulate the assembly of arrays of Sfi1 orthologs to control their functions in centrosomes in other eukaryotes.

All S. pombe strains are isogenic to 972 and are listed in supplementary material Table S1. Standard S. pombe molecular genetics techniques and media were used (http://research.stowers.org/baumannlab/documents/Nurselab_fissionyeasthandbook.pdf). Sfi1 deletion and C‐terminal tagging with GFP or mRFP at the endogenous locus was performed as described previously (Bähler et al., 1998). Sid4 tagging with mCherry was performed similarly.

The S65T‐GFP‐Cam1 strain is a generous gift from Trisha Davis (University of Washington). The Sid2‐GFP and Sid4‐GFP strains were obtained from Dannel Mc Collum, University of Massachusetts. Ppc89‐GFP and Plo1‐GFP strains were given by Kathleen Gould (Vanderbilt University). The nmt1‐cdc18 strain was obtained from Paul Nurse (Cancer Research UK, London Research Institute). The sad1‐1, wee1‐as8 and wee1‐as8 cdc25‐22 mutants are generous gifts from Iain Hagan, Cancer Research UK, Manchester Institute. GFP–Atb2‐ and mCherry–Atb2‐expressing strains have been described previously (Sato et al., 2009). These strains were crossed using the random spore method to obtain the strains listed in supplementary material Table S1.

Most plasmids used in this study derive from the integrative vector pJK148 (Keeney and Boeke, 1994). pAP244 (pJK148‐psfi1‐sfi1‐tnmt1, leu1+) was obtained by insertion into pJK148 of a KpnI‐SmaI PCR fragment encoding the sfi1 promoter and ORF amplified from genomic DNA and a SmaI‐SacI fragment encoding the nmt1 terminator digested from pREP3X.

pAP246 [pJK148‐psfi1‐sfi1‐GFP‐Tnmt1, leu1+], pAP248 [pJK148‐psfi1‐sfi1(1‐765)‐GFP‐Tnmt1, leu1+], pAP249 [pJK148‐psfi1‐sfi1(1‐188)‐GFP‐Tnmt1, leu1+], pAP250 [pJK148‐psfi1‐sfi1(766‐840)‐GFP‐Tnmt1, leu1+], pAP251 [pJK148‐psfi1‐sfi1(189‐765)‐GFP‐Tnmt1, leu1+], pAP254 [pJK148‐psfi1‐sfi1(189‐840)‐GFP‐Tnmt1, leu1+] were obtained by insertion of SalI‐NotI fragments encoding part of the sfi1 ORF obtained by PCR amplification and a NotI‐SmaI fragment from pSGP572a encoding S65T–GFP in pAP244 digested by SalI and SmaI, except for pAP246 and pAP248, where NsiI and NcoI were used, respectively, instead of SalI. These plasmids were integrated at the leu1 locus into a leu1‐32 strain using the lithium acetate‐DMSO method (Bähler et al., 1998) after digestion by NruI.

To produce Cdc31 in bacteria, the Cdc31 ORF was subcloned between BamHI and XmaI sites in pEGEX‐6P1 (Amersham; pIB14). To isolate the sfi1‐1 mutant, a SalI‐NarI fragment encoding the Sfi1 N‐terminus was amplified by PCR under mutagenic conditions and ligated into a pJK148‐psfi1‐sfi1‐tnmt1, leu1+ plasmid (pAP244) digested by similar enzymes. Mutagenized pAP244 was integrated at the leu1 locus in a leu1‐32/leu1‐32 sfi1+/sfi1Δ::kanMX6 diploid strain by homologous recombination after digestion by NruI. Leu1+ colonies grown at 33°C on EMM medium without leucine were sporulated on ME and the sfi1‐1 temperature‐sensitive mutant selected at 36°C on plates containing phloxin B after spore germination. Sequencing revealed a frame shift in the codon of valine 10, a silent mutation in the codon of alanine 63 and a substitution of asparagine 171 to aspartic acid. The Cdc31‐ts8 mutant was isolated as described previously (Ohta et al., 2012).

To produce the cdc31‐S15A mutant, cdc31‐Tadh‐KanMX6 sequences, flanked by 0.5‐kb sequences up‐ and down‐stream of the cdc31 ORF were cloned into pCR2.1‐TOPO (Invitrogen) and subjected to site‐directed mutagenesis. The mutant sequence was introduced at the cdc31 locus in a diploid cdc31Δ::ura4+/cdc31+ strain. Haploid G418‐resistant cdc31S15A colonies were recovered after sporulation.

Immunolabeling of Cdc31 with an anti‐Cen3 rabbit antibody, a generous gift of Michel Bornens (Institut Curie) (Paoletti et al., 2003) of tubulin with TAT1 monoclonal antibody (mAb) obtained from Keith Gull (University of Oxford), and DNA with DAPI was performed on cells fixed with methanol at −20°C for 2 minutes as described previously (Paoletti et al., 2003).

Epi‐fluorescence images shown in Fig. 1A–D, Fig. 2, Fig. 3A,D,E, supplementary material Fig. S1A–D, Fig. S2B–D; and Fig. 1E,F, supplementary material Fig. S1E,F were acquired, respectively, on a DMRXA2 and DM 5000 B upright microscope (Leica Microsystems), equipped with a 100×/1.4NA oil‐immersion PlanApo objective, a PIFOC objective stepper and a Coolsnap HQ CCD camera (Photometrics) on cells fixed with −20°C methanol for 2 minutes to stop SPB movement and rehydrated in PEM buffer (100 mM PIPES pH 6.9, 1 mM EGTA, 1 mM MgCl₂) and mounted between slide and coverslip. Stacks of nine planes spaced by 0.5 µm were acquired, and maximum projections were performed using Metamorph software.

Spinning‐disc confocal images shown in Fig. 3C, Fig. 4A, Fig. 5C,D, supplementary material Fig. S2A,E and Fig. 7A,E, supplementary material Fig. S4A,D were acquired, respectively, on a Nikon Eclipse TE2000‐U microscope equipped with a 100× 1.45NA, oil‐immersion objective, a PIFOC objective stepper, a Yokogawa CSU22 confocal unit and a Roper HQ2 CCD camera and a laser bench (Errol) with 491–561‐nm diode laser, 50 mW each (Cobolt) and a Nikon Eclipse Ti‐E microscope equipped with the Perfect Focus System, a 100×/1.45NA PlanApo oil‐immersion objective, a Mad City Lab piezo stage, a Yokogawa CSUX1 confocal unit a Photometrics HQ2 CCD camera and a laser bench (Errol) with 491–561‐nm diode laser, 100 mW each (Cobolt).

Cells were mounted between slides or coverslips or, for movies, mounted on YE5S agarose pads (Tran et al., 2004) (Fig. 3C; Fig. 4A) or in microfluidic devices (Velve‐Casquillas et al., 2010) (Fig. 5C,D; Fig. 7A–E; supplementary material Fig. S2A; Fig. S4A–D). Stacks of 9–11 planes spaced by 0.5 µm were acquired and maximum projections were performed using Metamorph software. For time‐lapse movies of sfi1‐1, GFP–Atb2 images were acquired every 10 minutes for 5 hours (binning 2 gain 3, 300 ms at 30% GFP laser power). For Sfi1–GFP time‐lapse movies shown in Fig. 4A, images were acquired every 2 minutes for 5 hours (binning 2 gain 3, 300 ms at 10% GFP laser power). For time‐lapse movies of Cam1–GFP (Fig. 4D; Fig. 6B) images were acquired every 2 minutes for 5 hours (binning 2 gain 3; 300 ms at 20% GFP laser power). The Cdc31‐S15A mutant expressing GFP–Atb2 was imaged every minute for 1 hour (Fig. 5D, binning 2 gain 3; 200 ms at 10% GFP laser power). Plo1–GFP‐ and Sid4–mCherry‐expressing strains were imaged every 2 minutes for 2 hours (supplementary material Fig. S3C, binning 2 gain 3; 200 ms at 10% GFP laser power and 300 ms at 15% mCherry laser power). Strains expressing Sid2–GFP and mCherry–Atb2 were imaged for 1 hour at 1‐minute intervals (supplementary material Fig. S3E, binning 2 gain 3; 200 ms at 10% GFP laser power and 300 ms at 12% mCherry laser power). Strains expressing Sfi1–GFP and mCherry–Atb2 were imaged for five hours with 2‐minute intervals for GFP and 4‐minute intervals for mCherry (supplementary material Fig. S3F,G, binning 2 gain 3; 400 ms at 15% mCherry laser power and 200 ms at 10% GFP laser power).

The cdc10‐V50 mutant expressing Ppc89–GFP was grown exponentially at 25°C in liquid YE5S then transferred at 36°C for 3.5 hours or 5 hours (Uzawa et al., 2004). Control cells expressing only Ppc89–GFP were grown under the same conditions. To block cells in S phase, cells expressing Ppc89–GFP were grown exponentially in YE5S at 25°C and treated for 5 hours with 11 mM hydroxyurea (HU). Cells were fixed in methanol at −20°C for 2 minutes as described previously (Paoletti et al., 2003) prior to imaging. Cell size was measured in Metamorph from DIC images.

For intensity measurements shown in Fig. 4B,C, SPBs were tracked automatically on 3D spinning‐disc confocal stacks using MIA software developed on site (Racine et al., 2006; Théry et al., 2006) or using the tracking object tool in Metamorph. Sfi1–GFP fluorescence intensity was corrected for background and computed over time in Matlab. In other figures, SPBs were either tracked manually or using the Tracking Objects tool of Metamorph (Molecular Devices) using regions of 6×6 pixels. Fluorescence ratios shown in Fig. 6A,B are derived from intensities measured immediately before and after SPB segregation at mitotic entry. In Fig. 7C,D, mean SPB fluorescence intensity was recorded on regions of 5×5 or 6×6 pixels using Metamorph on maximum projections of spinning‐disk confocal z‐stacks and corrected for background inside the same cell.

Sfi1–GFP fluorescence analyses were performed similarly in cdc25‐22 cells grown in YE5S and blocked in G2 by shift to 36°C (Fig. 4E) or in pnmt1‐cdc18 cells grown overnight in EMM medium containing 5 µg/ml thiamine and then blocked in S phase by overexpression of Cdc18 by growth for 17 hours at 30°C in the absence of thiamine (Fig. 4F).

Fast temperature shift experiments shown in Fig. 5C–E were performed using an in‐house built fast temperature control system (Velve Casquillas et al., 2011), a prototype of Tempocell (Elvesys). Briefly, cells were set at 25°C under a spinning‐disc confocal microscopy setup described previously (Almonacid et al., 2011) and shifted to 36°C after recording a first image to determine the cell cycle stage. Cells were then imaged every 10 minutes for 5 hours.

We crossed the cdc31‐8 mutant expressing GFP–Atb2 with cdc25‐22 wee1‐as8 (Tay et al., 2013). The cdc31‐8 cdc25‐22 wee1‐as8 mutant expressing GFP–Atb2 (AP4831) was grown exponentially in liquid YE5S and transferred to glass‐bottomed Petri dishes (WPI) coated with soybean lectin (0.1 mg/ml, Sigma) for 4 hours at 36°C to block cells in G2 and inactivate Cdc31. Imaging was then started at 36°C and the G2 block was released at frame 2 by the addition of 10 µM 3BrPP1. Spindle morphology was recorded during the next mitosis.

We also isolated a cdc31‐8 wee1‐as8 strain expressing GFP–Atb2 (AP4833). These cells were grown exponentially in liquid YE5S and transferred to glass‐bottomed Petri dishes coated with lectin (0.5 mg/ml, Sigma) for 6 hours at 25°C in the presence of 10 µM 3BrPP1 to produce short cells following division. Imaging was then started within a few minutes at 36°C and preheated medium containing 10 µM 3BrPP1 was added immediately after to ensure complete temperature shift to 36°C. Cell cycle stage was recorded immediately before the addition of preheated medium and spindle morphology was recorded at the next round of mitosis.

A volume of 1.5 l of cells was grown to 4×10⁷ cells at 30°C in YE5S medium concentrated two times as compared to regular YE5S medium (YE5S 2X). AP4100 and AP4110 cells were shifted for 7 hours at 18°C to block cells in mitosis before extract preparation. The cells were first washed with 50 ml of STOP buffer (150 mM NaCl, 50 mM NaF, 10 mM NaEDTA, 1 mM NaN₃), then resuspended in 6 ml of 1D buffer (50 mM HEPES pH 7.5, 100 mM NaCl, 1 mM EDTA, 1% NP40, 20 mM β‐glycerophosphate, 50 mM NaF, 0.1 mM Na₃VO₄, 1 mM PMSF complemented with complete EDTA‐free anti‐protease tablets from Roche). Cell aliquots of 600 µl were added to 600 µl of glass beads and lysed using a FastPrep FP120A instrument (Qbiogene; two cycles of 40 s at maximum speed). Lysates were then spun at 10,000 g for 10 minutes at 4°C, and supernatants were recovered. Soluble extracts were incubated with anti‐mouse‐IgG magnetic beads (M‐280 DYNAL, Invitrogen) coupled to 18 µg of anti‐GFP mAb (Roche) for 2 hours at 4°C; then, the beads were washed seven times with 1D buffer, and were resuspended in SDS‐PAGE sample buffer and heated at 95°C for 5 minutes. The immunoprecipitation samples were subjected to SDS‐PAGE, and the Cdc31 band stained with Colloidal Blue was excised, reduced with 10 mM DTT and alkylated with 55 mM iodoacetamide. After washing and shrinking of the gel pieces with 100% acetonitrile, in‐gel digestion was performed using chymotrypsin (Sequencing Grade, Promega) overnight in 25 mM ammonium bicarbonate at 30°C. The extracted peptides were analyzed by nano‐LC‐MS/MS using an Ultimate 3000 system (Dionex S.A.) coupled to an LTQ‐Orbitrap XL mass spectrometer (Thermo Fisher Scientific, Bremen, Germany). Samples were loaded on a C18 precolumn (300 µm inner diameter×5 mm; Dionex) at 20 µl/min in 5% acetonitrile, 0.1% trifluoroacetic acid (TFA). After 3 minutes of desalting, the precolumn was switched on line with the analytical C18 column (75 µm inner diameter×50 cm; C18 PepMapTM, Dionex) equilibrated in solvent A (2% acetonitrile, 0.1% formic acid). Bound peptides were eluted using a 160‐minute linear gradient [from 0 to 30% (v/v)] of solvent B (80% acetonitrile, 0.085% formic acid) at a 150 nl/min flow rate and an oven temperature of 40°C. Data‐dependent acquisition was performed on the LTQ‐Orbitrap mass spectrometer in the positive‐ion mode. Survey MS scans were acquired in the Orbitrap on the 350–1000 m/z range with the resolution set to a value of 100,000. Each scan was recalibrated in real time by co‐injecting an internal standard from ambient air into the C‐trap (‘lock mass option’). The five most intense ions per survey scan were selected for collision‐induced dissociation (CID) fragmentation and the resulting fragments were analyzed in the linear trap (LTQ). Target ions already selected for MS/MS were dynamically excluded for 30 seconds. Data were recorded using the Xcalibur software (version 2.2) and the resulting spectra were then analyzed using the MascotTM Software created with Proteome Discoverer (version 1.4, Thermo Scientific) using the SwissProt Schizosaccharomyces pombe (fission yeast) database, containing 5089 protein sequences. Carbamidomethylation of cysteines, oxidation of methionine, protein N‐terminal acetylation and phosphorylated serine, threonine and tyrosine were set as variable modifications for all Mascot searches. Specificity of chymotrypsin and trypsin digestion was set, and four missed cleavage sites were allowed. The mass tolerances in MS and MS/MS were set to 2 ppm and 0.8 Da, respectively. Phosphorylated peptides that have their non‐phosphorylated counterparts were manually validated.

Cdc31 was produced in fusion with GST from pIB14 as described above in BL21/DE3 bacteria. Expression was induced by addition of 0.5 mM IPTG for 16 hours at 20°C. Bacteria were lysed in lysis buffer [PBS containing 0.5% Triton X‐100, 10 mM MgCl₂, 1 mg/ml lysozyme (Euromedex), 1 mg/ml DNase I (Roche), 1 mM PMSF (Sigma), 5 mg/ml Pepstatin A (Sigma), Complete Mini‐EDTA‐free Protease Inhibitor Cocktail Tablet (Roche)]. GST–Cdc31 was purified on glutathione beads and eluted from beads by cleavage with PreScission protease between GST and Cdc31. Purified Cdc31 was subjected to an in vitro kinase assay with cyclin‐B–CDK1 (BioLabs), in parallel to histone H1 (positive control, BioLabs), according to the manufacturer's instructions. Proteins were then subjected to SDS‐PAGE and the gel was stained with ProQ Diamond (Life Technologies) to detect phospho‐proteins, according to the manufacturer's instructions, and Coomassie Blue to detect all proteins.

Similarly, Cdc31 was purified from nda3‐KM311 cells expressing Sfi1 tagged at the N‐terminus with GFP and blocked in mitosis for 7 hours at 16°C or released from the block for 1 hour at 32°C using an anti‐GFP mAb (Roche) as described above. Immunoprecipitates were subjected to SDS‐PAGE and the gel was stained with ProQ Diamond to detect phospho‐proteins and Sypro Ruby to detect all proteins according to the manufacturer's instructions.

We thank Ahmed El Marjou for technical help with Cdc31 purification; T. Davis, K. Gould, I. Hagan, D. McCollum and P. Nurse for strains; M. Bornens and K. Gull for antibodies; and M. Bornens, R. Basto, S. Rincon, M. Guzman‐Vendrell and K. Scheffler for critical reading of the manuscript. The authors also wish to thank Vincent Fraisier from the PICT‐IBiSA Lhomond Imaging facility of Institut Curie were imaging was performed.