Budding Yeast Bub2 Is Localized at Spindle Pole Bodies and Activates the Mitotic Checkpoint via a Different Pathway from Mad2

All yeast strains were derivatives of or were backcrossed at least three times to W303 (ade2-1, trp1-1, leu2-3,112, his3-11,15, ura3, and ssd1). Strains used for this work are listed in Table I. Cells were grown in YEP medium (1% yeast extract, 2% bactopeptone, and 50 mg/liter adenine) supplemented with 2% glucose (YEPD). α factor was used at 2 μg/ml, unless otherwise stated, nocodazole at 15 μg/ml, benomyl at 12.5 μg/ml and hydroxyurea at 150 mM. All the experiments were performed at 25°C, with the exception of those involving the ts mutants mps1-1 and ndc10-1, which were performed at 37°C.

Standard genetic techniques were used to manipulate yeast strains (Sherman, 1991) and standard protocols were used for genetic manipulations (Maniatis et al., 1992). For tagging BUB2 at the COOH terminus, a NotI cassette containing nine tandem repeats of the myc epitope (myc9; Piatti et al., 1996) was inserted before the termination codon of the HindIII-EcoRI COOH-terminal fragment of BUB2 subcloned in Yiplac204. The resulting plasmid (pSP49) was cut with BclI for integration at the BUB2 locus of W303, thus generating a full-length tagged version of BUB2 flanked by a truncated untagged version of the gene (strain ySP710). For tagging NDC10 at the COOH terminus, a NotI cassette containing six copies of the myc epitope (myc6; Piatti et al., 1996) was inserted before the termination codon of the NdeI/XbaI COOH-terminal fragment of NDC10 subcloned in pRS304. The resulting plasmid (pSP18) was cut with BclI for integration at the NDC10 locus of W303, thus generating a full-length tagged version of NDC10 flanked by a truncated untagged version of the gene (strain ySP460).

For Western blot analysis, protein extracts were prepared by TCA precipitation (see Fig. 3) as previously described (Piatti et al., 1996) or as described in Surana et al., 1993 (see Fig. 6 B). Proteins were transferred to Protran membranes (Schleicher and Schuell). myc-tagged Bub2 and Pds1 were detected with 9E10 mAb, whereas polyclonal antibodies (Amon et al., 1992) were used to detect Clb2. Anti-actin antibodies were purchased from Sigma Chemical Co. Secondary antibodies were purchased from Amersham and proteins were detected by an enhanced chemiluminescence system according to the manufacturer.

Flow cytometric DNA quantitation was determined according to Epstein and Cross (1992) on a Becton Dickinson FACScan^®. In situ immunofluorescence was performed according to Nasmyth et al. (1990). Visualization of Tet operators using GFP and chromosome spreading were performed as described in Michaelis et al., 1997. Immunostaining of Bub2myc9 and Ndc10myc6 were detected by incubation with the 9E10 mAb followed by indirect immunofluorescence using CY3-conjugated goat anti–mouse Ab (1:500; Amersham); signals were amplified using a CY3-conjugated donkey anti–goat Ab (1:500; Amersham). Spindle pole bodies were visualized with polyclonal anti-Spc72 (Knop and Schiebel, 1998) followed by incubation with FITC-conjugated anti–rabbit Ab (1:50, Jackson ImmunoResearch Labs.).

Since the mitotic checkpoint is thought to play an important role while chromosomes try to reach a bipolar attachment to the spindle, and inactivation of proteins involved in the mitotic checkpoint increases the frequency of chromosome loss (Li and Murray, 1991; Bernard et al., 1998), we asked whether the checkpoint proteins Mad2 and Bub2 had any role in the timing of sister chromatid separation during a normal cell cycle. For this purpose, we constructed wild-type, bub2Δ, and mad2Δ strains carrying the tetR-GFP/tetO constructs which allow to monitor sisters separation at the centromeric regions of chromosome V (Michaelis et al., 1997). Cell cultures of the above strains were synchronized in G1 by α factor treatment and released into fresh medium either lacking or containing the microtubule depolymerizing drug nocodazole, which activates the mitotic checkpoint. As previously reported (Hoyt et al., 1991; Li and Murray, 1991; Straight et al., 1996), in the presence of nocodazole both bub2Δ and mad2Δ cells lost the cohesion between sister chromatids, rebudded, and rereplicated, accumulating with DNA contents higher than 2C, whereas wild-type cells arrested in G2 as dumbbells with duplicated but unseparated chromosomes (Fig. 1 B). Interestingly, we reproducibly found that in these conditions bub2Δ mutants separated sister chromatids and started rereplicating later than mad2Δ, suggesting that, in the absence of Bub2, entry into anaphase can be still somewhat delayed by other mechanisms. Without nocodazole, all three strains underwent anaphase and proceeded through the cell cycle with similar kinetics (Fig. 1 A), thus indicating that either Mad2 and Bub2 have no role in the timing of sister chromatid separation during a normal cell cycle or bipolar chromosome attachment to spindle fibers in budding yeast might be a process too fast and efficient in order to allow detection of subtle differences in the timing of anaphase entry in the absence of surveillance mechanisms.

Since Mad1, Mad2, Bub1, and Bub3 have been found associated to unattached kinetochores during prophase and prometaphase in higher eukaryotic cells (Chen et al., 1996, 1998; Li and Benezra, 1996; Taylor and McKeon, 1997; Taylor et al., 1998), while the intracellular localization of Bub2 had not been established and little was known about its relationships with other mitotic checkpoint proteins, we analyzed the subcellular localization of Bub2. We tagged Bub2 at the COOH terminus with nine copies of the myc epitope, thus generating the Bub2myc9 variant, which was fully functional, as judged by the checkpoint proficiency of BUB2myc9 cells in nocodazole (see Fig. 3 A and not shown). Since Bub2myc9 staining in whole cells was fairly homogeneous and it was difficult to distinguish specific staining from the background observed with the same antibody on cells lacking tagged Bub2, we studied the protein localization by indirect immunofluorescence on chromosome spreads (Klein et al., 1992; Tanaka et al., 1997). By this technique, squashed nuclei stick to the slide and nucleoplasmic proteins are washed away, allowing detection of proteins bound to subnuclear insoluble structures, such as chromatin and nuclear cytoskeleton. We found that in all nuclei, Bub2myc9 staining was concentrated in one or two dots, very reminiscent of spindle pole bodies (SPBs). We, therefore, repeated the experiment by double staining the spreads with both anti-myc and anti-Spc72 antibodies, the latter of which recognize a constitutive SPB component (Knop and Schiebel, 1998). As shown in Fig. 2 A, the Bub2myc9 and Spc72 stainings completely overlapped, indicating that Bub2 resides at SPBs. As a control, we performed a double staining of the kinetochore protein Ndc10 and Spc72. To this purpose, we used a strain where the only functional copy of NDC10 was expressing a protein tagged with six copies of the myc epitope at the COOH terminus (Ndc10myc6). As it was previously shown for yeast centromeres (Funabiki et al., 1993; Guacci et al., 1997; Jin et al., 1998b), we found that Ndc10myc6 formed clusters near the SPBs, both in cycling and in nocodazole-arrested cells (Fig. 2 B). However, not only Ndc10 clusters had a different shape compared with that of SPBs and Bub2, but not all Ndc10myc6 clusters colocalized with Spc72, while Bub2myc9 always did. Although we cannot fully prove that the Ndc10myc6 clusters correspond to kinetochores, these data strongly suggest that Bub2 does not localize at kinetochores.

We then investigated further whether Bub2 protein levels and localization varied during the cell cycle and after nocodazole treatment. To this purpose, the BUB2myc9 strains were released from an α factor G1 arrest, and the Bub2myc9 protein levels and localization were analyzed at different time points. An aliquot of the synchronized culture was also incubated for 3 h in nocodazole. Bub2myc9 protein levels remained constant throughout the cell cycle and when the mitotic checkpoint was activated (Fig. 3 B). Furthermore, the protein is constitutively localized at SPBs: G1 cells showed a single dot of Bub2myc9 staining, whereas two bright dots started appearing at the time of bud emergence and entry into S phase, which coincides with the time of SPB duplication (data not shown). About 30–40% of G1 nuclei displayed a fainter staining of Bub2 at SPBs than during the other cell cycle phases, regardless of growth conditions. About 55% of nuclei from nocodazole arrested cells showed two bright dots of Bub2myc9 staining side by side (Fig. 2 A). This is, in these conditions, the typical spatial arrangement observed by electron microscopy for SPBs, which are duplicated but do not migrate apart in the absence of microtubules (Goetsch, L., and B. Byers, personal communication). Therefore, the nuclei showing single SPB signals by immunofluorescence (∼40%) under the same conditions likely had duplicated SPBs, which could not be resolved by this technique. Finally, we found that the localization of Bub2myc9 at SPBs was unaffected in the mitotic checkpoint mutants mad1Δ, mad2Δ, mad3, bub1-1, bub3Δ, mps1-1, and ndc10-1 (data not shown). Therefore, Bub2 is constitutively present at SPBs during the cell cycle and its localization does not require any of the analyzed mitotic checkpoint proteins.

The finding that Bub2 is localized at SPBs, and SPBs are mostly unseparated in nocodazole, prompted us to investigate whether Bub2 could be required to delay cell cycle progression only before SPB separation. For this purpose, bub2Δ, mad2Δ, and wild-type cells were synchronized by hydroxyurea (HU) treatment, which blocks DNA synthesis but allows duplication and separation of SPBs, thus arresting cells in S phase with short bipolar spindles. Cell cultures were then released from the HU block into nocodazole containing medium, in order to disassemble the previously formed spindles, and cell cycle progression of the three strains was analyzed. Formation and disassembly of the spindle in HU and in nocodazole, respectively, was confirmed by immunostaining of tubulin (data not shown). As shown in Fig. 4, under these conditions wild-type cells arrested in G2, with 2C DNA contents and unseparated sister chromatids, while mad2Δ and bub2Δ cells attempted to undergo anaphase and entered a new cell cycle, as indicated by their ability to separate sister chromatids and rereplicate DNA. Therefore, both Mad2 and Bub2 checkpoint functions are required also after SPBs separation and bipolar spindle formation.

Since we found Bub2 concentrated at SPBs, whereas Mad1 and Mad2 are localized, at least in higher eukaryotes, at unattached kinetochores, we asked whether Bub2 and Mad1, 2 might activate the mitotic checkpoint via different pathways.

To address this question, we first constructed mad1Δ bub2Δ and mad2Δ bub2Δ double mutants, and compared their sensitivity to the microtubule depolymerizing drug benomyl to that of the single mutants. In fact, hypersensitivity to benomyl of mad and bub mutants correlates with their mitotic checkpoint defect (Hoyt et al., 1991; Li and Murray, 1991), and inactivation of two proteins that lead to checkpoint arrest by different routes should result in a more severe checkpoint defect, and therefore in an increased benomyl sensitivity, compared with inactivation of either one. As shown in Fig. 5, mad1Δ bub2Δ and mad2Δ bub2Δ cells were much more sensitive to benomyl than mad1Δ, mad2Δ, and bub2Δ cells, suggesting that BUB2 belongs to an epistasis group different from that of MAD1 and MAD2.

We then verified whether the increased benomyl sensitivity of mad2Δ bub2Δ and mad1Δ bub2Δ double mutants correlated with an enhanced checkpoint defect. Cultures of mad1Δ, mad2Δ, bub2Δ, mad1Δ bub2Δ, mad2Δ bub2Δ, mad1Δ mad2Δ and wild-type strains were arrested with α factor and released in either the presence or the absence of nocodazole. Progression through the cell cycle in the absence of nocodazole was very similar in all strains (data not shown); however, in the presence of the drug mad1Δ bub2Δ and mad2Δ bub2Δ cells started rereplicating much faster and more efficiently than the single mutants and the mad1Δ mad2Δ double mutant (Fig. 6 A). Since the timing and efficiency of rereplication likely reflect the kinetics by which cyclin B–dependent kinases are inactivated, these data suggest that mad1Δ bub2Δ and mad2Δ bub2Δ double mutants advance inactivation of Clb-kinases with respect to the other mutants. Conversely, deletion of BUB2 in mad1Δ or mad2Δ cells did not accelerate their timing of sister chromatids separation (Fig. 6 A), indicating that Bub2 might play only a minor role in controlling degradation of the anaphase inhibitor Pds1. We then measured directly by Western blot analysis the levels of Pds1 and Clb2 during a similar experiment, where mad2Δ, bub2Δ, mad2Δ bub2Δ and wild-type strains expressed a myc-tagged Pds1 protein (Shirayama et al., 1998). As shown in Fig. 6 B, Clb2 was degraded in both mad2Δ and bub2Δ cells in nocodazole, suggesting that rereplication in these mutants is likely to result from their inability to prevent cyclin proteolysis. In addition, Clb2 disappeared more rapidly in the mad2Δ bub2Δ double mutant than in either single mutant, thus confirming that acceleration in the kinetics of rereplication in the double mutant correlates with an advanced inactivation of Clb/Cdk1 kinases. Although bub2Δ cells did also degrade Pds1, they did it very slowly, whereas Pds1 disappeared very rapidly and with nearly identical kinetics in both mad2Δ and mad2Δ bub2Δ cells. Thus, Pds1 degradation and the onset of anaphase appear to be mainly controlled by a Mad2-dependent pathway that does not seem to require Bub2. Altogether these data, while confirming that Mad1 and Mad2 are involved in the same pathway, indicate that Bub2 and Mad2 prevent B-type cyclin proteolysis by different routes, since the concomitant lack of Mad2 (or Mad1) and Bub2 has an additive effect on Clb2 degradation, exit from mitosis and entry into a new cell cycle. Conversely, Bub2 seems to play only a minor role, if any, in controlling the onset of anaphase.

Since mitotic checkpoint arrest involves APC inactivation (He et al., 1997; Li et al., 1997; Hwang et al., 1998; Kim et al., 1998; Schott and Hoyt, 1998), it was possible that different APC subunits might be alternatively involved in the Mad2- and the Bub2-dependent checkpoint response. To address this question, we asked whether various APC subunits were required for cell cycle progression of mad2Δ or bub2Δ mutants in the presence of microtubule depolymerizing drugs. Unfortunately, most apc mutations are temperature sensitive and cause a cell cycle arrest in metaphase already during an otherwise unperturbed cell cycle (Irniger et al., 1995; Zachariae et al., 1996; Hwang and Murray, 1997). Furthermore, bub2Δ mutants show a poor checkpoint defect at 37°C (data not shown). However, we were able to test whether CDC26, which is totally dispensable in a normal cell cycle at 25°C (Zachariae et al., 1996), might become indispensable for cell cycle progression of either mad2Δ or bub2Δ cells in the presence of nocodazole. The role of Cdc26 in the mitotic checkpoint was also of particular interest as we found that cdc26Δ mutants are hypersensitive to benomyl at the permissive temperature (Amato, S., and S. Piatti, unpublished results). Cell cycle progression of mad2Δ cdc26Δ and bub2Δ cdc26Δ double mutants was compared with that of each single mutant after release from α factor in the presence of nocodazole at 25°C. As shown in Fig. 7, bub2Δ cdc26Δ cells failed to proceed in the cell cycle and arrested mostly in G2 with duplicated and unseparated chromosomes. Conversely, deletion of CDC26 did neither affect the time of anaphase onset nor cell cycle progression of mad2Δ cells. These results indicate that under these conditions Cdc26 becomes strictly required for both the onset of anaphase and exit from mitosis in the absence of Bub2 but not of Mad2, again consistently with the involvement of these two proteins in different mitotic checkpoint pathways.

The mitotic checkpoint is likely to play a crucial role also during an unperturbed cell cycle, delaying the onset of anaphase until all chromosomes are properly aligned on the bipolar spindle and thereby preventing the occurrence of aneuploidies. Mistakes made during mitosis may be responsible for the abnormal karyotype of many human tumour cells and have an important role in oncogenesis. Loss of function of the p53 tumour suppressor gene impairs the mitotic checkpoint and the normal regulation of centrosome duplication (Fukasawa et al., 1996). In addition, a number of human tumour cell lines have been recently found to be defective in the activation of the mitotic checkpoint in response to microtubule depolymerizing drugs (Li and Benezra, 1996; Cahill et al., 1998). In addition, the chromosomal instability observed in two colorectal cancers is associated to mutational inactivation of BUB1 (Cahill et al., 1998). Finally, expression of dominant negative mutant versions of either murine Bub1 or human Mad1 accelerates exit from mitosis and causes accumulation of multinucleate cells, respectively, even in the absence of microtubule depolymerizing drugs, whereas microinjection of anti-Mad2 antibodies into mammalian cells induces the onset of anaphase before chromosomes have congressed to the metaphase plate (Gorbsky et al., 1998), suggesting that these proteins play an important role in regulating the normal timing of mitosis (Taylor and McKeon, 1997; Jin et al., 1998). We tested directly whether loss of function of MAD2, BUB2, or both, might accelerate progression through mitosis of yeast cells. We found that the kinetics of anaphase and exit from mitosis were unaffected in mad2Δ, bub2Δ (Fig. 1) and mad2Δ bub2Δ cells (data not shown) compared with wild-type. Since mad mutants were shown to increase the frequency of chromosome loss (Li and Murray, 1991), our data suggest that either Bub and Mad proteins are not required for the correct timing of mitosis during an unperturbed cell cycle, or bipolar attachment of yeast chromosomes to spindle fibers is so fast and efficient that does not enable to detect subtle differences in the absence of surveillance mechanisms. In agreement with the latter hypothesis is the finding that in the absence of fission yeast Bub1 cells do not prematurely enter mitosis despite showing massive chromosome missegregation (Bernard et al., 1998).

The mechanism by which Mad1, 2, 3 block cell cycle progression in response to spindle depolymerization or monopolar attachment has been recently uncovered. Mad2 has been shown by several groups to bind and inactivate Cdc20/APC, thus causing stabilization of Pds1 and cyclins B with subsequent block of anaphase and mitotic exit (Fang et al., 1998b; Hwang et al., 1998; Kallio et al., 1998; Kim et al., 1998; Wassman and Benezra, 1998). This physical interaction might occur directly at the kinetochore; consistently, the mammalian homologue of budding yeast APC1, Tsg24, was constitutively localized at centromeres (Jörgensen et al., 1998) and p55CDC, homologous to Cdc20, was also found at kinetochores during mitosis (Kallio et al., 1998). However, APC components, p55CDC and HsMad1 have also been localized at centrosomes (Tugendreich et al., 1995; Jin et al., 1998a; Kallio et al., 1998). Our finding that BUB2 and MAD1, 2 belong to different epistasis groups, as indicated by the accelerated degradation of Clb2, exit from mitosis and rereplication of mad1 bub2 and mad2 bub2 compared with that of the corresponding single mutants, suggests that Bub2 might block cell cycle progression in case of monopolar attachment of chromosomes or spindle damage by a pathway distinct from the one dependent on Mad1, 2. This conclusion is further strengthened by the observation that cell cycle progression of bub2Δ, but not mad2Δ, cells in nocodazole depends on the unessential APC subunit Cdc26. Although Cdc26 seems to be constitutively part of the APC, it is required for anaphase and ubiquitination of B-type cyclins only at 37°C; furthermore, it is induced after heat shock (Zachariae et al., 1996). In S. pombe, overexpression of the Cdc26 homologue Hcn1 suppresses a cut9 mutation which causes defects in APC assembly (Yamada et al., 1997). Thus, Cdc26/Hcn1 is a dispensable subunit that might be only required to modulate APC activity under stress conditions or when other APC subunits are defective (Townsley and Ruderman, 1998). The fact that cell cycle progression of mad2, but not bub2, mutants tolerates loss of Cdc26 function suggests that either Bub2 acts by finally inhibiting a special form of APC which requires Cdc26, or that the function of a Cdc26-less APC is somewhat compromised already at 25°C and the lack of Bub2 is not sufficient to circumvent the defect.

The conclusion that Mad1, 2 and Bub2 both contribute to the activation of the mitotic checkpoint via distinct routes can only be drawn for what concerns mitotic exit, which is a consequence of the inactivation of cyclin B–dependent kinases. Conversely, loss of cohesion between sister chromatids, which instead depends on Pds1 degradation, is mostly driven by the inactivation of the Mad1, 2 pathway, as the kinetics of sister chromatids separation of mad1Δ bub2Δ and mad2Δ bub2Δ double mutants is very similar to those displayed by mad1Δ or mad2Δ cells. Obviously, bub2Δ cells are able to degrade Pds1 and attempt to undergo anaphase to a certain extent in the absence of a functional spindle, but this might be a consequence, rather than a primary effect, of the inactivation of mitotic CDKs. Indeed, we found that degradation of Pds1 is very slow and inefficient in bub2Δ cells in the presence of nocodazole, suggesting that Bub2 plays only a minor role in controlling the onset of anaphase. The loss of sister chromatids cohesion we observe in bub2Δ cells treated with nocodazole might just reflect a pathological situation rather than shedding light on the actual role of Bub2. We therefore propose that Bub2's major, if not sole, role in the activation of the checkpoint is to prevent, by inhibiting B-type cyclins proteolysis, inactivation of mitotic CDKs and, therefore, cytokinesis and entry into a new cell cycle. The finding that the fission yeast homologue of Bub2, Cdc16, in addition to having a role in the mitotic checkpoint, is implicated in regulating septation and cytokinesis during a normal cell cycle (Fankhauser et al., 1993; Furge et al., 1998) supports this hypothesis. We currently do not know the molecular mechanism by which the Bub2-dependent pathway contributes to maintaining high levels of these CDKs in nocodazole. In fact, Bub2 might inhibit directly the APC-dependent degradation of cyclins B, by inhibiting Hct1 or Cdc5 or even by interfering with the integrity of the APC subunit composition; alternatively, it could prevent Sic1 accumulation or Cdc28 inhibitory phosphorylations on Thr18 and Tyr19, which have been previously shown to play a role in the mitotic checkpoint (Minshull et al., 1996; Wang and Burke, 1997). Whatever the mechanism is, the absence of Bub2 in cells treated with nocodazole might cause a drop in Clb1-4/Cdc28 kinase activity, which would in turn lead to dephosphorylation and unscheduled activation of Hct1; Hct1/APC, possibly specifically involving the Cdc26 subunit, would then promote degradation of both mitotic cyclins and Pds1, despite Cdc20 is kept inactive by the Mad1, 2 pathway. Several lines of evidence strongly support this model: overexpression of Hct1 in nocodazole arrested cells is able to promote degradation of Pds1, albeit inefficiently (Visintin et al., 1997), whereas deletion of BUB2 is sufficient to drive a slow degradation of Pds1 and continuous cell cycle progression in cdc20 mutants (Shirayama et al., 1998; Tavormina and Burke, 1998).

Many mitotic checkpoint proteins, like Mad1-3, Bub1, and Bub3, have been localized at kinetochores not attached to spindle fibers, such as in prophase and prometaphase of a normal mitosis or in the presence of microtubule disrupting agents (Chen et al., 1996, 1997; Li and Benezra, 1996; Taylor and McKeon, 1997; Taylor et al., 1998). Their localization is fully consistent with previous lines of evidence implicating the kinetochore in transmitting an inhibitory signal to the cell cycle machinery in case of monopolar attachment (Rieder et al., 1994, 1995). Other observations suggested that the mitotic checkpoint also monitors the lack of tension at kinetochores (Li and Nicklas, 1995; Nicklas et al., 1995). Phosphorylation of the kinetochore-associated 3F3/2 epitope correlates with the activation of the mitotic checkpoint (Campbell and Gorbsky, 1995) and is activated by tension (Nicklas et al., 1995). Conversely, localization of Mad2 to kinetochores depends only on microtubule attachment but not on tension (Waters et al., 1998), suggesting that different proteins might be involved in sensing and transducing different signals. Localization of Bub2 at SPBs does not correlate with the activation of the mitotic checkpoint, but is fairly constitutive throughout the cell cycle, raising the questions of how Bub2 function is regulated, whether it plays any role during a normal cell cycle, and which signals it monitors. Whatever the signals are, we have shown that Bub2, as well as Mad2, is required to arrest cell cycle progression also after SPB separation. One possibility is that signals coming from single unattached kinetochores are integrated at SPBs, thus activating the Bub2 dependent checkpoint control. It is worth noting that, unlike in other organisms (Chen et al., 1998), human Mad1 is localized at kinetochores during interphase but at centrosomes during mitosis (Jin et al., 1998a), suggesting that microtubule organizing centers are likely to play a central role in the mitotic checkpoint. Another alternative is that the signal detected by Bub2 might be different from that sensed by the other Mad and Bub proteins; for instance, it has been proposed that Bub2 might be involved in detecting alterations in the spindle structure, since it is not required to arrest the cell cycle in response to impaired kinetochore function (Wang and Burke, 1995). Unfortunately, this hypothesis cannot be tested directly because spindle defects would necessarily also compromise the kinetochore-microtubule attachment. Finally, since we propose that Bub2 only plays a minor role, if any, in regulating the degradation of Pds1 and separation of sister chromatids, we speculate that Bub2 might be involved in the activation of the mitotic checkpoint only later than the other Mad and Bub proteins, by preventing cytokinesis and exit from mitosis until anaphase has taken place. An intriguing hypothesis is that SPBs can somehow monitor when anaphase has been completed, and the inhibitory signal generated by Bub2 is rapidly extinguished when sister kinetochores have reached the poles.

Which proteins does Bub2 interact with at SPBs in order to prevent cytokinesis? Recent work in fission yeast has started to elucidate the mechanism by which the Bub2 homologue Cdc16 might function. In fact, Cdc16 has been found to physically interact with Byr4, which was previously implicated in the control of septation (Song et al., 1996; Furge et al., 1998). The Cdc16/Byr4 complex displays GAP activity towards Spg1 (Furge et al., 1998), a GTPase homologous to S. cerevisiae Tem1 (Shirayama et al., 1994; Schmidt et al., 1997). Remarkably, Spg1 is constitutively bound to SPBs and in the GTP-bound form seems to recruit at SPBs the Cdc7 protein kinase, which is involved in promoting cytokinesis (Fankhauser and Simanis, 1994; Sohrmann et al., 1998). Mutations affecting budding yeast TEM1 cause cells to arrest in telophase and have been recently found to impair the APC-dependent ubiquitination of mitotic cyclins (Shirayama et al., 1994; Jaspersen et al., 1998), thus providing a link between SPBs, APC function, and cytokinesis. The phenotype of tem1 mutants is very similar to that of several mutants defective in late mitotic events, like cdc15, dbf2, and cdc14. Their characterization has uncovered a complex network of genetic interactions which led to the hypothesis that all the corresponding genes might be involved in the same process. In addition, Cdc14, which directly dephosphorylates both Sic1 and Hct1, has been postulated to be the downstream effector of the whole cascade (Visintin et al., 1998).

In conclusion, our current model of how the mitotic checkpoint could work is depicted in Fig. 8. During prophase, while chromosomes start to attach to spindle fibers, Mad1, 2, 3, Bub1, and Bub3 present at unattached kinetochores prevent activation of Cdc20 and, therefore, degradation of Pds1 and inactivation of mitotic CDKs. At the same time Bub2, localized at SPBs, also prevents inhibition of Clb1-4/Cdc28 kinases and cytokinesis by an independent mechanism, which might involve inactivation of late mitotic proteins, like Tem1, Cdc15, Dbf2, and Cdc14. Exit from mitosis and cytokinesis are therefore blocked until anaphase has taken place, a process which might switch Bub2 function off. In the absence of Bub2 the ectopic activation of Hct1 triggered by Cdc14 would bring about Pds1 degradation and sister chromatid separation even if Cdc20 is kept inactive.

We are grateful to Rafal Ciosk and Kim Nasmyth for providing the tetR-GFP system and other strains; Michael Knop and Elmar Schiebel for antibodies against Spc72; Loretta Goetsch and Breck Byers for communicating unpublished results; Etienne Schwob for anti-Clb2 antiserum; Pier Carlo Marchisio and Simona Tognin for help with photomicroscopy; Léon Dirick, Marco Muzi Falconi, and Maria Pia Longhese for critical reading of the manuscript.

This work was supported by grants from Associazione Italiana Ricerca sul Cancro to S. Piatti, Cofinanziamento 1997 by Ministero Università e Ricerca Scientifica e Tecnologica, Università di Milano to G. Lucchini, Training and Mobility of Researcher contract ERBFMRXCT98-0212 to S. Piatti and CNR Target Project on Biotechnology grant CT.97.01180.PF49(F).

APC

anaphase-promoting complex

BUB

budding uninhibited by benzimidazole

cyc

cycling

exp

exponential phase

HU

hydroxyurea

MAD

mitotic arrest-deficient

SPB

spindle pole body