The Drosophila Bub3 protein is required for the mitotic checkpoint and for normal accumulation of cyclins during G2 and early stages of mitosis

During mitosis, a checkpoint mechanism delays entry into anaphase until all chromosomes are properly attached and aligned at the metaphase plate, thus preventing the unequal segregation of genetic material (Skibbens and Hieter, 1998). Genetic screens in budding yeast for mutants that do not arrest in mitosis after the induction of spindle damage allowed the identification of several components of this checkpoint. These include the Mad1, Mad2 and Mad3 proteins, the Bub1, Bub2 and Bub3 proteins and the kinase Mps1 (Hoyt et al., 1991; Li and Murray, 1991; Weiss and Winey, 1996). Soon after, homologues of most of these proteins were identified in higher eukaryotes, including Bub1, Bub3, Mad1, Mad2 and the human/mouse homologue of Mad3, BubR1; these proteins were further shown to localize preferentially to unattached kinetochores (Musacchio and Hardwick, 2002).

A biochemical link between the checkpoint and known regulators of mitotic progression first emerged from studies showing that the Xenopus and human homologues of Mad2 are able to bind and sequester Cdc20/Fizzy, an activator of the APC/C (Amon, 1999; Musacchio and Hardwick, 2002; Shah and Cleveland, 2000). The APC/C is a multi-subunit E3 ubiquitin ligase that targets several mitotic regulators, including securin and mitotic cyclins, for degradation by the proteasome, thus triggering mitotic exit. The APC/C subunits and many of its target proteins are present throughout the cell cycle, but APC/C activity and specificity towards the substrates is modulated by its association with co-factors such as Cdc20/Fizzy and Cdh1/Fizzy-related. APC/C association with Cdc20 occurs upon entry into mitosis and requires the phosphorylation of APC/C subunits. These phosphorylation events are thought to be mediated by cdc2/cyclin B and Polo kinases, and enhance the activity of the APC/C^Cdc20 towards its substrates. The activity of APC/C^Cdc20 triggers the metaphase-anaphase transition both by inducing the ubiquitination of securin and by targeting cyclin B for degradation. Cdh1, another co-factor, mediates the ability of the APC/C to degrade mitotic regulators, like polo and aurora kinases, and to degrade cyclin B completely, thus promoting mitotic exit. The interaction of the APC/C with Cdh1 is inhibited by Cdh1 phosphorylation, which is mediated by Cdk1 and Cdk2. As a result of cyclin B destruction, Cdk activity drops, ensuring that Cdh1 remains dephosphorylated and active, thus preventing the accumulation of mitotic cyclins during the subsequent G1 (Harper et al., 2002; Peters, 2002; Zachariae and Nasmyth, 1999). However, the G1/S transition and the G2 stage of the cell cycle require accumulation of cyclins A and B and therefore APC/C^Cdh1 inactivation. The mechanism by which the APC/C is regulated during these stages of cell cycle is still poorly understood. Recently it was found that APC/C inactivation during the G1/S transition is achieved by Emi1, a newly identified inhibitor of the APC/C, as well as by phosphorylation of Cdh1 by Cdks (Hsu et al., 2002; Reimann et al., 2001a; Reimann et al., 2001b).

The downstream target of the mitotic checkpoint is APC/C^Cdc20, whose inhibition prevents sister chromatid separation. However, the role of the various checkpoint components in APC/C inhibition has been a matter of some controversy. Not only Mad2, but also Mad3/BubR1 can interact directly with Cdc20 (Hardwick et al., 2000; Tang et al., 2001; Wu et al., 2000). Sudakin and colleagues purified an APC/C^Cdc20 inhibitory complex from interphase cells, called the mitotic checkpoint complex (MCC), and showed that it contains Mad2, BubR1, Bub3 and Cdc20 (Sudakin et al., 2001). A similar complex was found in budding yeast and shown to be independent of kinetochore assembly (Fraschini et al., 2001; Hardwick et al., 2000). Tang and colleagues obtained similar results showing that a BubR1-containing complex was a stronger inhibitor of the APC/C^Cdc20 than was Mad2, although this complex contained only BubR1 and Bub3 (Tang et al., 2001). Despite the discrepancies relative to the constitution of the complexes, these results suggest that checkpoint proteins may exist as APC/C inhibitory complexes already in interphase, however, the function of these complexes before mitosis is not yet known.

Even though most checkpoint components are strongly conserved through evolution, the role of some checkpoint proteins, in the checkpoint response is still unknown, particularly in the case of Bub3. Besides its association with BubR1 and Mad2 during interphase, Bub3 was also found in two independent complexes with Bub1 and BubR1 in mitotic mammalian and Xenopus cells (Campbell and Hardwick, 2003; Taylor et al., 1998); and was shown to be required for the localization of the mammalian proteins to the kinetochores (Taylor et al., 1998). In yeast, Bub3 was also found in a complex with Mad1 and Bub1; the formation of this complex is dependent on Mad2 and seems essential for the checkpoint response (Brady and Hardwick, 2000).

In order to study the function of Bub3 further, we searched for mutant alleles in Drosophila and carried out depletion of the protein in S2 cells. Our data show that Bub3 has an additional role besides its involvement in a checkpoint dependent mitotic arrest upon spindle damage. We report here that Bub3 is necessary to prevent APC/C-dependent degradation of mitotic cyclins during G2, thereby regulating both entry and transit through the initial stages of mitosis. Furthermore, our data suggest the existence of differentially activated APC/C complexes, which are inhibited by Bub3 to ensure accumulation of mitotic cyclins.

Df (3R)Dr-rv1 (breakpoints 99A01-02; 99B06-11) and cdc27^l(3)L7123 were obtained from the Bloomington (IN) Stock Center. To rescue the bub3¹ mutation, a full-length bub3 cDNA was cloned into the pP{UAST} vector (Brand and Perrimon, 1993), injected into w¹¹¹⁸ embryos and a number of stable transformants were obtained. To activate transcription of the transgene, p{UAST-bub3}; bub3¹/TSTL individuals were crossed with a strain carrying Mz1061 and the bub3¹ mutation. The resulting ({pUAST-bub3}/+; bub3¹/bub3¹) flies were viable and no mitotic phenotype was observed. A non-degradable form of cyclin B was expressed from the transgene p{UAS-CBTPM-GFP} (Wakefield et al., 2000). bubR1¹ was described previously (Basu et al., 1999) (see also Logarinho et al., 2004).

Third instar larval brains were dissected, fixed and stained as previously described (Llamazares et al., 1991). Whenever required, brains were dissected in PBS and incubated in 10 μM colchicine (Sigma) for 1 hour prior to fixation. Mitotic index was defined as the number of mitotic cells per optical field, with every optical field in the brain being scored. Immunostaining of neuroblasts from third instar larvae was performed as previously described (Platero et al., 1995). Anti-Bub3 antibody (Logarinho et al., 2004) was diluted 1:500, anti-Polo antibody (Llamazares et al., 1991) diluted 1:50 and anti-cyclin B antibody (Jacobs et al., 1998) was used at 1:3000 dilution. For analysis of cyclin B levels, all images were acquired using the same parameters in a Zeiss Axioscop microscope with a SPOT 2 camera (Diagnostic Instruments, USA). Quantification of cyclin levels per cell was carried out using ImageJ 1.30v software (http://rsb.info.nih.gov). The mean pixel intensity per cell area for different genotypes at different stages of the cell cycle was determined and used for further statistical analysis (Student's t-test). Cyclin levels in at least five brains were scored for each genotype.

To deplete Bub3 from S2 cells, a fragment of 600 bp spanning the 5′ UTR of bub3 and including the ATG initiation codon was cloned into the vectors pSPT18 and pSPT19 (Roche). RNA was synthesized using the T7 Megascript kit (Ambion). 15 μg dsRNA were added to 10⁶Drosophila S2 cells in Schneider's medium (Gibco, BRL) and incubated for 1 hour, after which cells were supplemented with 2 ml media with 10% FBS (Gibco, BRL). Experiments were performed at least three times, in six-well plates and at each time point cells were collected and processed for immunofluorescence and immunoblotting. For immunoblotting, cells were collected by centrifugation, washed in PBS supplemented with protease inhibitors (Roche) and resuspended in 2× SDS-sample buffer before loading on a 12% SDS-PAGE. When required, cells were incubated with 30 μM colchicine (Sigma). Mitotic index was determined as the percentage of anti-phosphorylated histone H3 (PH3: Upstate Biotechnology)-positive cells.

Cells were centrifuged onto coverslips and fixed in 3.7% formaldehyde, 0.5% Triton X-100 in 1× PBS, for 8 minutes, followed by a washing step in 1× PBS, 0.1% Triton X-100 (PBST). Rabbit anti-PH3 (Upstate Biotechnology) was diluted to 1:1000, γ-tubulin (GTU88, Sigma) was diluted at 1:1500 and anti-cyclin A was diluted to 1:4000 (Whitfield et al., 1990). Anti-rabbit Alexa 488 and anti-mouse Alexa 568 (Molecular Probes), were used as secondary antibodies. DNA was counterstained with DAPI in Vectashield (Vector, UK). All images were collected, deconvolved and projected using Cell Observer System (Zeiss, Germany). Adobe Photoshop 7.0 (Adobe Systems) was used to process all images.

Third instar larvae brains were dissected and homogenized at 4°C in 1× PBS supplemented with protease inhibitors (Roche Diagnostics). 10-40 μg of total protein extracts were resolved in 10-12% SDS-PAGE and transferred to nitrocellulose membrane (Schleicher and Schuell). Immunopurified anti-Bub3 was diluted 1:50 in PBST containing 1% BSA. α-tubulin was detected using mAb B512 (Sigma) diluted 1:5000. Anti-cyclin B antibody was diluted at 1:10,000. Secondary antibodies conjugated to HRP (Vector, UK) were used according to the manufacturer's instructions.

Live analysis of mitosis was done on S2 cells stably expressing GFP-tubulin. Control or Bub3 RNAi-treated cells were incubated for 96 hours and plated on glass coverslips treated with 30 μg/ml concanavalin A (Sigma). Time-lapse images were collected at 2 minute intervals, starting from the time asters could be visualized, using a Cell Observer System (Zeiss, Germany). Image sequence analysis and movie assembly was done with ImageJ Software (NIH, USA) and Quicktime 6 (Apple Computer, USA). The time between the appearance of the asters and the nuclear envelope breakdown, as well as the beginning of anaphase was determined and used for further statistical analysis (Student's t-test).

During a search for Drosophila genes involved in the regulation of mitotic progression, we isolated a recessive mutant line in which the lethality mapped to the bub3 locus. Genomic sequencing of the bub3 gene in this line showed that it contained a single point mutation in the coding region, leading to the substitution of a conserved glycine at position 193 by aspartic acid (data not shown). This mutation is responsible for the lethality of the line, as expression of wild-type bub3 cDNA during larval development using a specific GAL4 driver rescues the mitotic phenotype and gives rise to viable adults. Thus, in Drosophila, bub3 is an essential gene and we have designated this mutant allele bub3¹. Western blot analysis of total protein extracts from bub3¹ homozygous or bub3¹/Df(3R)Dr-rv1 hemizygous individuals indicates that the mutant protein is expressed (Fig. 1A). To evaluate whether the mutant protein is able to localize properly, antibodies were used to determine its intracellular localization in bub3¹ third instar larvae neuroblasts. Although in wild-type cells Bub3 always localizes to kinetochores during prometaphase, in approximately half bub3¹ cells the mutant protein is not detected at kinetochores (Fig. 1B). However, some bub3¹ cells (45%) show localization of the mutant protein at one or two kinetochore pairs, whereas only very few (9%) show localization to all kinetochores, suggesting that this allele might behave as a hypomorph.

To determine whether bub3¹ cells have an abnormal mitotic checkpoint response, we measured the mitotic index of both bub3¹ homozygous and hemizygous neuroblasts in the presence and absence of spindle damage (Fig. 1C and supplementary material Table S1). The results show that in the absence of spindle damage, the mitotic index of mutant larvae is not significantly different from wild-type controls. However, if the spindle is disrupted by colchicine, neither bub3¹ homozygous nor hemizygous mutant cells are able to arrest in mitosis and undergo precocious sister chromatid separation (PSCS) (Fig. 1C,D and supplementary material Table S1) indicating loss of mitotic checkpoint response.

We next examined mitotic progression in both homozygous and hemizygous mutant individuals. As the phenotype associated with both genotypes is nearly identical, only observations on homozygous mutant cells are shown (Fig. 1E). bub3¹ mutant cells undergo cytologically normal prophase but, a significant proportion of cells with a prometaphase-like configuration (19%) display PSCS. Furthermore, most of the anaphases (70%) observed are disorganized and contain lagging chromatids, and of the total mitotic figures a high proportion (38-42%) are aneuploid, probably reflecting the occurrence of PSCS in prior rounds of cell division.

Although our observations of late mitotic stages indicated that bub3¹ homozygous and hemizygous mutant cells had nearly identical phenotypes, the two genotypes behave differently during the early stages of mitosis (Fig. 1F and supplementary material Table S1). Although the frequency of prophase and prometaphase cells in bub3¹ homozygous mutants does not differ significantly from the wild type, hemizygous individuals have a higher frequency of cells in prophase and a correspondingly lower frequency of cells in prometaphase. This classical genetic test verifies our previous conclusion from cytological analysis that the bub3¹ allele behaves as a hypomorph. Importantly, these results further suggest that a more severe depletion of Bub3 function causes a slow progression through early stages of mitosis, a phenotype that to our knowledge has not been described before for any checkpoint protein.

The results above suggest an unexpected role for Bub3 during early mitotic stages that only became apparent in bub3¹ hemizygous individuals. We therefore proceeded to analyse the consequences of depleting Bub3 in Drosophila S2 cells by double-stranded RNA interference (RNAi). Treatment of S2 cells with bub3 RNAi (Fig. 2) caused a significant reduction of the Bub3 protein levels within 48 hours and by 96 hours the protein was barely detectable (Fig. 2A). Immunofluorescence analysis on RNAi-treated cells failed to detect Bub3 at kinetochores in all mitotic cells examined (Fig. 2B and data not shown). Accordingly, Bub3-depleted cells are unable to arrest in mitosis when the spindle is damaged (Fig. 2C), and show elevated levels of PSCS (Fig. 2D,E). Similar to the results presented above for bub3¹ mutant neuroblasts, depletion of Bub3 by RNAi demonstrates that in Drosophila S2 cells, Bub3 is required for an efficient mitotic checkpoint response.

We next addressed the role of Bub3 in mitotic progression. Depletion of Bub3 causes a reduction in cell proliferation (Fig. 2F) that is not due to loss of cell viability (data not shown), but instead to an increase in the culture doubling time. Quantification of the mitotic index over the course of the experiment shows that in comparison with control cells, the number of cells in mitosis increases after depletion of Bub3 (Fig. 2G). These results contradict a role for Bub3 as a negative regulator of mitotic exit and suggest that Bub3-depleted cells spent more time in mitosis, leading to a slower proliferation rate. To better understand this unexpected behaviour, we carefully analysed mitotic progression. The results show that depletion of Bub3 leads to an increase in the number of cells with PSCS (Fig. 2H). Compared to control cells, there is also a twofold increase in the total number of cells in anaphase and most anaphases (68%) are disorganized or have lagging chromatids (data not shown). Notably, after depletion of Bub3 there is a significant reduction in the number of cells in prometaphase and metaphase, whereas approximately 57% of the prometaphase cells exhibit PSCS (Fig. 2I). These data are all consistent with a premature mitotic exit. If these cells exit mitosis earlier than normal, then the higher mitotic index observed after depletion of Bub3 (Fig. 2G) must result from a delay at an earlier stage of mitosis. Indeed, there is a twofold increase in the percentage of cells in prophase 72 hours after treatment with RNAi (Fig. 2J). Therefore, a severe depletion of Bub3 in S2 cells results in mitotic alterations similar to those observed in bub3¹ hemizygous neuroblasts. Both sets of observations indicate that besides its function in the mitotic checkpoint response, Bub3 has a role in promoting normal transit through early stages of mitosis, particularly through prophase.

To clarify further a possible role of Bub3 in mitotic progression, we performed in vivo time-lapse analysis of S2 cells stably expressing GFP-tubulin after depletion of Bub3 by RNAi. Control and Bub3-depleted cells were recorded every 2 minutes from the time asters first appeared to the reformation of the daughter nuclei. GFP-tubulin is an excellent marker to follow mitotic progression, as we can time various events. The visualization of well-defined asters was taken to indicate initiation of prophase, nuclear envelope breakdown (NEBD) can be seen by rapid entry of GFP-tubulin into the nuclear area, retraction of kinetochore bundles shows the beginning of anaphase A, spindle elongation marks anaphase B and nuclear re-formation indicates telophase. Thus, we can determine the duration of all major mitotic phases including prophase (from appearance of the asters until NEBD), prometaphase-metaphase (from NEBD until anaphase initiation), anaphase A (from microtubule bundle retraction to spindle elongation) and anaphase B (from spindle elongation to nuclear reformation) allowing a direct correlation between the in vivo data and results from fixed material. The results show that depletion of Bub3 causes significant changes in mitotic progression during both prophase and prometaphase (Fig. 3 and supplementary material Movies 1 and 2). The appearance of the asters was set as time zero, and the duration of each mitotic stage was plotted (Fig. 3A,B). In vivo analysis of untreated S2 cells shows that prophase takes on average 16 minutes, however, after Bub3 depletion, prophase is significantly delayed lasting on average 21 minutes (Fig. 3A,B and supplementary material Movies 1 and 2). Furthermore, although in control cells the timing between NEBD and anaphase onset takes on average 34 minutes (Fig. 3A,B and supplementary material Movie 1), after Bub3 depletion it is significantly reduced to 21 minutes on average, indicating an accelerated progression through prometaphase-metaphase. These data are in agreement with the proposed function of Bub3 in the mitotic checkpoint response.

Thus, depletion of Bub3 causes a significant delay in prophase, strongly supporting our observations in fixed cells (see above). However, we were surprised not to find a more significant prophase delay as depletion of Bub3 causes a twofold increase in the percentage of cells in prophase (Fig. 2J). To clarify this discrepancy, the number cells in prophase, in S2 cells expressing GFP-tubulin after Bub3 RNAi was quantified using PH3 as a mitotic marker. Two hallmark events of prophase were analysed: the appearance of PH3 at Ser10, which in Drosophila correlates with the beginning of prophase (Hendzel et al., 1997) and separation of the centrosomal asters. This analysis also showed a twofold increase in the number of cells in prophase after Bub3 depletion as described previously (data not shown). However, unlike control cells, in Bub3 depleted cells, two types of PH3 positive cells were clearly observed: cells in which formation of the asters was markedly visible and cells with an interphase arrangement of microtubules without visible asters (Fig. 3C). The latter phenotype accounted for 53% of the prophase cells observed after Bub3 depletion and was never observed in control cells. In the control population, the appearance of the asters always correlated with PH3 staining. These observations suggest that the in vivo analysis is likely to underestimate the prophase delay observed after depletion of Bub3.

The results described above suggest that after Bub3 depletion, nuclear and cytoplasmic events, namely chromosome condensation and spindle formation, appear to uncouple. It is well established that accumulation of mitotic cyclins and the activation of the cdk/cyclin complexes is essential for entry and progression through mitosis (Zachariae and Nasmyth, 1999). Besides being important for chromosome condensation (Kimura et al., 1998) cdk activity is responsible for the dramatic changes in microtubule dynamics that occur at the onset of mitosis, resulting in microtubule nucleation from centrosomes and assembly of the mitotic spindle (Buendia et al., 1992; Verde et al., 1992). Accordingly, we next tested whether the slower progression through early mitotic stages observed after Bub3 depletion could result from abnormal accumulation of mitotic cyclins. Cyclin B levels were measured by immunofluorescence and western blot analysis in cells that had been depleted of Bub3 by RNAi (Fig. 4). Control and RNAi-treated cells were incubated in colchicine, to promote high levels of cyclin accumulation and increase the number of cells in mitosis, and cyclin B levels were measured by immunofluorescence (Fig. 4A,B). The centrosomal marker γ-tubulin was used to distinguish between different cell cycle stages. Interphase cells without γ-tubulin staining were classified as G1/S, whereas those with centrosomal staining but without centrosome separation were classified as G2. Cells in mitosis (prophase and prometaphase) had γ-tubulin staining, well-separated centrosomes and clear chromosome condensation (Fig. 4A). The fluorescence intensity values obtained for G1/S cells did not differ between control and RNAi-treated cells (Fig. 4A,B and supplementary material Table S2) allowing a direct comparison between the two samples. In agreement with the expected pattern for cyclin B accumulation, S2 control cells start to accumulate cyclin B during G2 and attain their highest levels of cyclin B during mitosis (Fig. 4A,B and supplementary material Table S2). However, cyclin B accumulation in Bub3-depleted cells is surprisingly different. These cells already show significantly lower levels of cyclin B in G2 and also during mitosis. The failure to maintain high levels of cyclin B during mitosis is in agreement with the predicted function of Bub3 in the mitotic checkpoint response (similar results were also observed in bub3¹ mutant neuroblasts: see supplementary material Fig. S1 and Table S2). However, these results also indicate that Bub3 might be required during G2 to ensure accumulation of cyclin B. To eliminate the possibility that colchicine treatment could indirectly affect cyclin B accumulation during G2, cyclin B levels were measured in control and Bub3-depleted cells in the absence of the drug (Fig. 4C,D). Quantification of cyclin B levels shows that in contrast to control cells, Bub3-depleted cells fail to accumulate cyclin B during G2 (Fig. 4D and supplementary material Table S3). In agreement with the immunofluorescence data, western blot analysis of total protein extracts shows that after Bub3 depletion, accumulation of cyclin B is significantly reduced reaching only 50% of the levels observed in untreated cells (Fig. 4E,F). In addition, although treatment with colchicine leads to accumulation of cyclin B in control cells, the same is not observed after Bub3 depletion, where accumulation of cyclin B is severely compromized. These results show that Bub3 does indeed appear to be required during G2 to promote accumulation of cyclin B, strongly suggesting that Bub3 might have a function in G2 before its well-established role in the mitotic checkpoint during prometaphase.

Accumulation of cyclins during G2 in somatic cells depends primarily upon transcriptional activation and the fact that the anaphase promoting complex/cyclosome (APC/C) is inhibited at this stage by early mitotic regulators (Reimann et al., 2001a). To test whether APC/C activity might be required to mediate the reduction in cyclin B levels after depletion of Bub3, we analysed the effect of removing the APC/C subunit cdc27 in bub3 mutant cells. It has recently been shown that in Drosophila, the APC/C subunit cdc27 is required for the degradation of cyclin B but not of securin (another APC/C substrate), as mutations in cdc27 mutant neuroblasts or depletion of cdc27 by RNAi in S2 cells, results in cells with high levels of cyclin B and separated sister chromatids (Deak et al., 2003; Huang and Raff, 2002). Accordingly, mutations in the bub3 and cdc27 genes would be predicted to have opposite effects on the accumulation of cyclin B and the combination of the two mutations might even lead to normal mitotic entry. We therefore measured cyclin B levels in single (bub3¹ or cdc27) and double (bub3¹;cdc27) mutant cells after incubation in colchicine (Fig. 5A,B and supplementary material Table S4). The results show that cdc27 mutant cells have significantly higher levels of cyclin B both in G2 and mitosis than do control wild-type cells, in agreement with the phenotype previously described for this mutant (Deak et al., 2003). As expected, accumulation of cyclin B in double mutant cells (bub3¹; cdc27) during G2 is not significantly different from wild-type controls and it is higher than in bub3¹ cells (Fig. 5B). These results support the hypothesis that the inability to accumulate normal levels of cyclin B after depletion of Bub3 is mediated by the APC/C.

Next, we analysed several mitotic parameters in single (bub3¹ or cdc27) and double (bub3¹; cdc27) mutant cells after spindle damage induced by incubating the cells in colchicine (Fig. 5C,D and supplementary material Table S5). Under these conditions, mutant cdc27 neuroblasts arrest in mitosis with well-condensed chromosomes and unseparated sister chromatids resulting in an increased mitotic index when compared to the wild-type control. This behaviour is due to the additive effect of the mutation in the APC/C subunit and the checkpoint-dependent arrest induced by spindle damage (Fig. 5C). In neuroblasts mutant for both bub3¹ and cdc27, the percentage of cells in mitosis is increased relative to bub3¹ alone. This result shows that the premature mitotic exit characteristic of bub3¹ mutant cells is dependent on APC/C activity. Nevertheless, the mitotic index of double mutant cells (bub3¹ and cdc27) is significantly lower than that seen in wild-type cells. This difference relative to control cells is most likely explained by premature mitotic exit due to the absence of an effective checkpoint response when Bub3 activity is absent, coupled with the fact that cdc27 does not seem to be required for APC/C-mediated securin degradation (data not shown). Finally, we observed that the frequency of cells in prophase in double mutant cells is indistinguishable from wild-type controls, suggesting that the double mutant transits normally through this early stage of mitosis (Fig. 5D). To rule out the possibility that cyclin B levels could be regulated by a mechanism other than APC/C-driven proteolysis, we overexpressed a non-degradable form of cyclin B (Wakefield et al., 2000) in bub3¹ mutant neuroblasts using the GAL4/UAS system. The results show that stabilization of cyclin B in bub3¹ mutant cells results in an increase in the mitotic index to values similar to those observed in control cells (Fig. 5E). Taken together, these results indicate that Bub3 appears to negatively regulate APC/C activity during the G2-M transition, allowing the accumulation of sufficient cyclin B for timely progression through the early stages of mitosis.

The results described above suggest that Bub3 is required for normal accumulation of cyclin B before and during mitosis. However, cyclin A is also required to promote mitotic entry in Drosophila (Parry and O'Farrell, 2001; Su, 2001). Cyclin A accumulates during S phase and G2 and it is degraded by the APC/C prior to the degradation of cyclin B as cells progress through early stages of mitosis. However, in contrast to cyclin B, cyclin A levels are not stabilized by the spindle damage-associated checkpoint response (Geley et al., 2001; Kaspar et al., 2001; Whitfield et al., 1990). In order to determine if Bub3 is also required to promote accumulation of cyclin A, we analysed its levels during cell cycle progression in Bub3-depleted cells. The centrosomal marker γ-tubulin was used to distinguish cells in G1/S or G2. The results show that although cyclin A accumulates from G1 to G2 in control cells (Fig. 6A,B and supplementary material Table S6) and can still be detected in some early mitotic cells (data not shown), cyclin A fails to accumulate and can hardly be detected in cells depleted for Bub3. Analysis of cyclin A accumulation in bub3¹ homozygous mutant neuroblasts gave very similar results (data not shown). These observations further support the role of Bub3 as a negative regulator of the APC/C during G2 and mitosis.

Unlike other checkpoint proteins like Mad2 or BubR1, Bub3 has never been found to interact directly with the APC/C or with its activator Cdc20 (Fang, 2002; Fang et al., 1998; Wassmann and Benezra, 1998; Wu et al., 2000). Thus, we were interested to determine if other proteins that interact simultaneously with Bub3 and Cdc20 or APC/C subunits could mediate Bub3-dependent APC/C inhibition. BubR1 is, at first glance, a good candidate to perform this function as it has been found in an interphase high molecular weight complex (together with Bub3) that is able to inhibit the APC/C (Sudakin et al., 2001; Tang et al., 2001). Accordingly, we tested whether mutations in bubR1 could affect cyclin B accumulation in G2 or mitosis (Fig. 7). Wild-type or bubR1¹ third larval neuroblasts were incubated in colchicine and immunostained to reveal the level of chromatin condensation and cyclin B. Analysis of control cells shows that we could identify cells with no cyclin B and no chromosome condensation (classified as G1/S); cells with cyclin B levels but no visible chromatin condensation (classified as G2); and cells with high levels of cyclin B and well-condensed chromosomes (classified as mitotic), as expected for normal cyclin B accumulation. Similarly, we could detect bubR1¹ mutant neuroblasts in G1 and also in G2 (Fig. 7A) with normal levels of cyclin B (Fig.7B). However, mitotic cells showed significantly lower levels of cyclin B (Fig. 7A,B) consistent with its known function in the mitotic checkpoint response. These results show that the pattern of cyclin B accumulation in the absence of bubR1¹ is significantly different from that of bub3¹ mutant cells (see supplementary material Fig. S1). Therefore, these data suggest that accumulation of cyclin B during G2 and early mitosis requires Bub3 independently of its interaction with BubR1.

Mitotic checkpoint proteins are essential inhibitors of the metaphase-anaphase transition as they ensure that errors in the interactions between the spindle and the kinetochores can be corrected prior to exit from mitosis. Here we show that apart from its essential role in the checkpoint, Bub3 is also required during the G2/M transition and prophase to allow normal accumulation of mitotic cyclins presumably by regulating the activity of the APC/C. In the absence of Bub3, cells are unable to arrest in response to spindle damage and also progression through early stages of mitosis is delayed owing to significantly lower levels of cyclins A and B.

In order to study the role of Bub3 during mitosis we identified and characterized bub3¹, the first mutant allele of the gene in Drosophila. Analysis of bub3¹ mutant cells indicates that as in other systems (Kalitsis et al., 2000), bub3 is an essential gene. The bub3¹ mutant allele appears to be hypomorphic as hemizygous neuroblasts and S2 cells depleted of Bub3 by RNAi show a more severe phenotype than bub3¹ homozygous mutant cells. Also, as previously shown for other organisms (Babu et al., 2003; Campbell and Hardwick, 2003; Kalitsis et al., 2000), we show that in Drosophila, bub3 is required for checkpoint-dependent mitotic arrest as its loss either by mutation or by RNAi causes PSCS, abnormal anaphase organization, significant aneuploidy and inability to arrest in mitosis after spindle damage.

In vivo analysis of cell division in Bub3-depleted cells reveals that this checkpoint protein could be required for the normal timing of mitosis. Using S2 cells stably expressing GFP-tubulin we find that after Bub3 depletion, mitosis (from NEBD to anaphase onset) is significantly faster than in control cells. This is at odds with recently published results where it was shown that in HeLa cells stably expressing H2B-GFP, Bub3-depleted cells enter anaphase with misaligned chromosomes but the timing between NEBD and anaphase onset is not altered (Meraldi et al., 2004). The reason for this discrepancy is not clear and we can only attribute it to a species-specific requirement for Bub3 during mitosis.

More significantly, our data revealed that cell cycle progression after bub3 mutation or depletion is characterized by a high frequency of cells in prophase, suggesting a slower progression through the early stages of mitosis. Live analysis of Bub3-depleted cells confirmed these results. This delay in prophase appears to result from a defective accumulation of cyclins in both interphase and mitotic cells. Indeed, if we stabilize cyclin B in bub3¹ mutant cells by a mutation in the gene for the APC/C subunit cdc27 or through expression of a stable form of cyclin B, the mitotic index is increased and bub3¹ mutant cells transit normally through early stages of mitosis. These results also suggest that the defective accumulation of cyclins in bub3¹ mutant cells is likely to be APC/C dependent, suggesting that Bub3 is able to regulate APC/C activity well before its established role in the mitotic checkpoint response during prometaphase.

Bub3 has never been shown to bind either the APC/C or cdc20 directly. Therefore, it is possible that Bub3 affects APC/C activity through one of its binding partners, for example BubR1. However, the analysis of mitotic progression of bub3¹ mutant cells causes a very different mitotic phenotype from that caused by mutations in bubR1 (Basu et al., 1999) (see also Logarinho et al., 2004). Cells mutant for bubR1 enter prophase normally but progress into anaphase as soon as the nuclear envelope breaks down, even before the completion of chromosome condensation, resulting in chromosome breakage and apoptosis. None of these phenotypes is observed after mutation of bub3 or depletion of the Bub3 protein in S2 cells. These cells show significant aneuploidy and undergo premature mitotic exit, but with properly condensed chromosomes and showing no signs of apoptosis. Furthermore, the analysis of Bub3-depleted cells showed that the majority of the cells are delayed in prophase because they display mitosis-specific phosphorylation of histone H3 and retain an intact nuclear envelope. Although classified as prophase, only half of these cells show matured asters, a cytoplasmic event that should have taken place as cells enter mitosis. These results suggest that after Bub3 depletion, nuclear and cytoplasmic events can uncouple, an observation that is in agreement with a reduced activity of the cyclin/cdk complexes in the cytoplasm. A correlation between cyclin A and cyclin B dependent kinases and formation of the mitotic spindle is well established. It has been shown that cyclin A-dependent kinase activity increases the microtubule nucleating activity of centrosomes (Buendia et al., 1992). On the other hand, the reorganization of microtubules that ultimately leads to mitotic spindle assembly seems to involve both cyclin A- and cyclin B-dependent kinase activities (Verde et al., 1992). Furthermore, it has been recently proposed that centrosome nucleated microtubules, at the prophase-prometaphase transition, promote tension when attached to the nuclear envelope and induce tearing of the nuclear lamin, thus promoting nuclear envelope invagination, permeabilization and eventually NEBD (Beaudouin et al., 2002). This model could explain how mitotic spindle formation and nuclear disassembly are two highly coordinated processes. Furthermore, in vivo studies in Drosophila embryos have shown that APC/C subunits, cdc27 and cdc16, accumulate at the nuclear envelope region during interphase and are only enriched in the nuclear area as cells enter prophase and NEBD takes place (Huang and Raff, 2002). Thus, the integrity of the nuclear envelope appears to establish a barrier between the nucleus and the cytoplasm during early stages of mitosis. Uncoupling of mitotic processes and delayed prophase may explain why bub3¹ mutant cells undergo mitosis with properly condensed chromosomes.

It is widely accepted that during mitosis APC/C activity is suppressed by the checkpoint proteins as long as there is one kinetochore that is unattached to the spindle or that is not otherwise under tension (Rieder et al., 1995; Rieder et al., 1994). According to this model, checkpoint proteins including Bub1, BubR1, Bub3 and Mad2 are recruited to unattached kinetochores where specific protein complexes are produced that directly or indirectly inhibit APC/C activity (Chen et al., 1998; Martinez-Exposito et al., 1999; Shah and Cleveland, 2000; Skoufias et al., 2001). However, our results show that Bub3 appears to be required for inhibition of the APC/C independently of its localization to kinetochores as it only localizes to kinetochores during mitosis. Recently published results also suggest that multiprotein complexes, including Bub3-BubR1, Bub3-Mad2 or Bub3-BubR1-Mad2, that inhibit the APC/C can form in the absence of kinetochores (Sudakin et al., 2001; Tang et al., 2001). These results establish a possible mechanistic basis for the action of checkpoint proteins as regulators of APC/C activity independent of their kinetochore localization. Tight regulation of the APC/C activity ensures sequential destruction of APC/C substrates and the correct timing of mitotic events. During interphase, APC/C activity is regulated by Emi1, most probably by preventing binding of substrates to the APC/C (Reimann et al., 2001b). During G1/S, Emi1 blocks APC/C^Cdh1 activity, allowing cyclin A accumulation and thus promoting G1/S transition (Reimann et al., 2001b). During G2 and early prophase, Emi1 is able to inhibit Cdc20, thereby promoting cyclin B accumulation and mitotic progression (Reimann et al., 2001a). Degradation of Emi1 occurs during prophase releasing APC/C inhibition (Margottin-Goguet et al., 2003). In this context, our data suggest that Bub3 could mediate APC/C inhibition before NEBD.

Recent results have suggested that more than one APC/C complex may be responsible for either sister chromatid separation or cyclin B destruction during mitosis (Huang and Raff, 2002). Our results on the mitotic behaviour and cyclin B accumulation in single (bub3¹ and cdc27) and double (bub3¹; cdc27) mutant cells support this data and suggest that Bub3 might affect the activity of the different APC/C complexes. First, we show that incubation of cdc27 mutant neuroblasts with colchicine leads to a mitotic arrest with unseparated sister chromatids and normal cyclin B levels, revealing a classical checkpoint response. Second, sister chromatid cohesion in cdc27 mutants can be abolished by a mutation in bub3, showing that the APC/C-dependent separation of sister chromatids does not require cdc27. Third, cyclin B can be degraded during G2 or mitosis in the absence of cdc27 when Bub3 is depleted or mutated. These observations are fully in accordance with previously published results showing that the APC/C subunits cdc16 and cdc27 have distinct locations during mitosis and that individual depletion of cdc16 or cdc27 proteins by RNAi leads to distinct mitotic phenotypes (Huang and Raff, 2002). Similarly, in yeast it has been shown that the APC/C subunit cdc27 is not required for the degradation of securin as overexpression of the cdk inhibitor Sic1 is sufficient to rescue the viability of cdc27 mutants (Thornton and Toczyski, 2003). Furthermore, mutations in the Drosophila homologue of APC5, another APC/C subunit, results in a phenotype similar in all respects to mutations in cdc27 including high levels of cyclin B and sister chromatid separation (Bentley et al., 2002). These data suggest that the activity of the different APC/C subunits may be required at different times during mitosis and at different locations within a cell, and may help to determine the specificity of the APC/C towards the substrates, thus reflecting differentially activated APC/C complexes.

Overall, our results suggest that checkpoint proteins might be required to restrain APC/C activity at multiple times during entry and progression through mitosis, revealing that what has been previously called the spindle assembly checkpoint is indeed a much broader regulatory mechanism that monitors events both before and during mitosis.

We thank C. F. Lehner for anti-cyclin B antibody, D. Glover for anti-cyclin A antibody, Ron Vale for the GFP-tubulin S2 cell line and J. Raff and J. Urban for fly stocks. Madalena Costa provided excellent technical assistance. C.S.L. is supported by a grant from FCT of Portugal. The laboratory of C.E.S. is supported by grants from FCT of Portugal and the TMR program of the EU. Work in the laboratory of M.L.G. is supported by NIH grant R01GM48430.