Mitotic Regulators Govern Progress through Steps in the Centrosome Duplication Cycle

Embryos were fixed (37% formaldehyde) and stained with purified polyclonal anti–γ-tubulin (made against a peptide present in the maternal isoform of γ-tubulin [QIDYPQESPAVEASKAG]), anti–β-tubulin antibody (1:100 dilution; Amersham). Fluorescent secondaries were used at 1:300 (Jackson ImmunoResearch Laboratories, Inc.). DNA was stained with 10 μg/ml of bisbenzamid (Hoechst 33258). Fluorescence was observed on a Leica DMRD or an Olympus 1X70 microscope. Photographs were taken on the Leica using a Photometrics CH250 CCD and 35-mm cameras. Three-dimensional data sets of centrosomes were collected using Deltavision software (Applied Precision, Inc.), based on an Olympus 1X-70 microscope with a Photometrics PXL-cooled CCD camera. Data sets were projected onto a single plane and exported to Adobe Photoshop for further processing.

Embryos were collected for 2 h from a stock carrying transgenes encoding nondegradable versions of the Drosophila mitotic cyclins, cyclin A and cyclin B (homozygous for both transgenes) (Sprenger et al. 1997). After aging the collection for 2 h at 25°C, stable cyclins were induced by floating the agar collection plates on 37°C water for 30 min (heat shock). The embryos were allowed to recover for 1–2 h and fixed.

string7B (Edgar and O'Farrell 1989) and fizzy1 (Dawson et al. 1993; Sigrist et al. 1995) are amorphic alleles.

Embryos were prepared for EM by high pressure freezing, followed by freeze substitution (McDonald 1994), or by slight modifications of a glutaraldehyde fixation protocol (Callaini and Riparbelli 1990; Callaini et al. 1997) (string and fizzy mutants). Thin serial sections (80–90 nm) cut on a Leica ultramicrotome were stained and examined on a Philips EM 400 at 80 kV, and the images were recorded on film. Though only one section is shown, all centrioles were reconstructed from thin serial sections to assess separation and orientation of centrioles.

Embryos were stained with Hoechst and homozygous fizzy mutant embryos were identified based on their abnormally high mitotic index during late stage 11 and stage 12. string embryos were identified by a low cell density and the absence of mitotic figures.

Cell cycle stages in wild-type embryos, prepared for EM, were identified based on the morphology of the associated nuclei. Cells with condensing/condensed chromatin not fully aligned at the metaphase plate were identified as prophase; cells with a mitotic spindle and aligned chromosomes as metaphase; cells containing decondensed chromatin in nuclei that are separated away from each other as telophase; and cells with intact nuclear membranes containing decondensed chromatin as interphase cells.

Despite differences in lengths of G2, cells enter mitosis with only two centrosomes, suggesting that centrosome replication is coupled to cell cycle progress, or that centrosome replication is independently, yet appropriately, regulated (e.g., by a developmental timer) (Edgar et al. 1994a). To distinguish between these possibilities, we monitored centrosome duplication in string mutant embryos that arrest in G2 but continue to develop. Arrested cells can be driven into mitosis by expression of Cdc25^string transcribed from a heat shock–inducible promoter on a transgene (Edgar and O'Farrell 1990).

After delaying cells in G2 beyond their normal time of mitosis, we induced expression of Cdc25^string. β-Tubulin staining of the resultant mitotic cells (see Fig. 2 C) revealed no significant spindle abnormalities (99.4% bipolar, n = 1,350) (data not shown). Immunostaining for γ-tubulin revealed two foci of staining, one at each pole of 42 preanaphase spindles (see Fig. 2 D). In the absence of coupling, we would expect to see four centrosomes in cells that had remained in G2 beyond their normal time of entry into mitosis. We conclude that centrosome duplication does not continue during a G2 arrest.

To assess the ultrastructure of the centrioles during the G2 arrest, we examined string mutant embryos by EM. In 21 out of 22 centriole pairs (Table), the daughters were significantly shorter than the mothers and their tubular fine structure was not obvious (e.g., compare Fig. 2 E to wild-type mitotic centrioles in Fig. 1). We conclude that Cdc25^string is required, directly or indirectly, for the completion of daughter centriole assembly.

We investigated for a role of the APC in the centrosome cycle by examining centrioles in embryos mutant for the APC activator, Cdc20^fizzy. fizzy embryos exhibit mitotically arrested cells in the ventral epidermis (VE) during cell cycle 16 (Dawson et al. 1993; Sigrist et al. 1995). However, morphogenesis continues, providing a timer to estimate the duration of the mitotic arrest.

We scored centriole pairs as engaged, if they were at right angles to each other and closely spaced (48–56 nm). In late stage 11 fizzy embryos, >80% of cells in the VE are arrested at metaphase (Fig. 3, A–D, and Table) and 62% of the metaphase cells (n = 21, from two embryos) contained engaged centrioles (Fig. 3 G). If centriole disengagement was unperturbed in fizzy mutants, only cells traversing metaphase should have engaged centrioles. Since fewer than 5% of the VE cells are normally in metaphase at any time (in wild type), <5% of the cells in fizzy mutant embryos would be traversing metaphase. Since ∼50% of cells in the VE contained orthogonal centrioles (62% of the metaphase cells), we conclude that centriole disengagement is delayed in metaphase cells arrested as a result of Cdc20^fizzy depletion.

Older fizzy embryos exhibited disengaged centrioles. In early stage 12 embryos, corresponding to ∼30 min of arrest, all observed mother and daughter centrioles appeared disengaged but not widely separated (Fig. 3E and Fig. H). During midstage 12, corresponding to ∼1 h of arrest, all observed centriole pairs had disengaged and separated (Fig. 3F and Fig. I). Thus, disengagement and separation of centrioles, events that normally occur within a couple of minutes after metaphase, occur more slowly in the fizzy mutant. We conclude that Cdc20^fizzy is required, directly or indirectly, for timely centriole disengagement.

We monitored the centrosome cycle upon expression of stable cyclins, which blocks some aspects of mitosis. Destruction box and NH₂ terminally deleted forms of both cyclins A and B (Sprenger et al. 1997) were ectopically expressed when transgenic embryos were exposed to heat shock. Entry into mitosis occurred at the normal time in these embryos, but cells failed to exit mitosis (Sigrist et al. 1995) (Fig. 4A and Fig. B). Immunostaining for γ-tubulin revealed no more than two centrosomes at each pole (50% had two centrosomes, 23% had one, and 27% had one elongated centrosome, n = 51 from two embryos) (Fig. 4 C).

To explore the status of centrioles, we examined stable cyclin-expressing embryos by EM. Centrioles at 17 spindle poles had disengaged with varying degrees of separation (between 80 nm to 1 μm), consistent with the distribution of γ-tubulin foci. Importantly, all observed centrioles (n = 39; Table), were present as singlets, i.e., were not associated with daughter centrioles (Fig. 4 D). Possible heat shock effects independent of transgene induction (Debec et al. 1990) were investigated by monitoring events in the centrosome cycle upon heat shock of nontransgenic embryos. Both cell cycle progression and procentriole formation continued after heat shock (data not shown). We conclude that stabilization of mitotic cyclins can, directly or indirectly, inhibit production of procentrioles.

Cell doubling requires the coordination of many cellular events including DNA replication and centrosome duplication. We show that the activities of mitotic regulators influence progress of the centrosome cycle in addition to the nuclear cycle (Fig. 5). Cdc25^string, which induces entry into mitosis, is required for the completion of daughter centriole assembly. Cdc20^fizzy triggers chromosome separation by APC-dependent degradation of inhibitors and is required for timely centriole disengagement. Destruction of cyclins at mitosis is required for synthesis of new daughter DNA strands and elaboration of a new daughter centriole. By genetically arresting cell cycle progress at specific points and identifying a correspondingly specific arrest of the centrosome cycle, we have resolved multiple points of coupling between the cell cycle and the centrosome cycle. Each point of coupling might represent direct action of the cell cycle regulator on the centrosome cycle, or a checkpoint induced as a consequence of the cell cycle arrest. Ultimately, available genetic and molecular tools ought to allow identification of the specific events responsible for the observed coupling.

In Drosophila embryos, a procentriole appears during S phase, but its assembly is completed during mitosis (Callaini et al. 1997). We found that Cdc25^string function, the expression of which drives cells into mitosis, is required for the completion of daughter centriole assembly. To our knowledge, this is the first indication that a G2 arrest does not support daughter centriole maturation. However, it is possible that centriole maturation continues at a slow rate in the absence of Cdc25^string function and is accelerated upon its expression. The centriole maturation induced by Cdc25^string might be secondary to the activation of mitotic cyclin/Cdk1 and subsequent mitotic events. For example, since several centrosomal proteins are recruited to the nucleus during interphase, nuclear membrane breakdown at mitosis might stimulate centriolar growth by releasing previously sequestered components.

The EM analysis of fizzy embryos reveals that disengagement and separation of mother and daughter centrioles are delayed: disengagement for ∼10 times the length of metaphase. Requirement for Cdc20^fizzy in timely centriole disengagement could reflect a corresponding requirement for degradation of an APC substrate that inhibits centriole disengagement. The block to centriole disengagement might be incomplete because residual maternal Cdc20^fizzy could slowly promote centriole disengagement in mitotically arrested cells. Alternatively, centriole disengagement might be timed by activation of Cdc20^fizzy, while not absolutely requiring its function. For example, slow Cdc20^fizzy-independent separation might reflect APC-independent turnover of an inhibitor of centriole disengagement.

The mitotic cyclins are degraded in a fizzy-dependent fashion at the mitotic exit (Sigrist et al. 1995), and could be candidates for the inhibitors of centriole disengagement. However, cyclins A and B persist in fizzy mutant embryos at times when we observe disengagement (Sigrist et al. 1995). Additionally, ectopic expression of stabilized cyclins A and B does not block centriole disengagement but inhibits assembly of new procentrioles. Similar results have been obtained upon injection of stable cyclin B into sea urchin embryos (Hinchcliffe et al. 1998). Inhibition of procentriole formation during mitosis may reflect a requirement for nuclear envelope formation (say for nuclear sequestration of inhibitors) or dephosphorylation of certain key proteins upon downregulation of mitotic cyclin/Cdk1. We infer that in an unperturbed cell, downregulation of mitotic cyclin/Cdk1 is essential for procentriole formation.

How might these observed points of coupling be enforced? Cell cycle regulators could contribute rather indirectly to progression through the centrosome cycle. In the yeast Saccharomyces cerevisiae, spindle pole body duplication is coordinated with the nuclear cycle by cell cycle–specific transcription of spindle pole body components (Kilmartin et al. 1993). Two of the three points of coupling that we have uncovered in Drosophila operate within mitosis, during which there is no transcription (Shermoen and O'Farrell 1991). Hence, mitotic maturation of daughter centrioles and disengagement of centriole pairs are apparently regulated posttranscriptionally. In contrast, procentriole formation might require new gene expression and would be deferred until transcription resumes upon mitotic exit. While such a mechanism is possible, it would be short circuited by the persistence of maternal gene products, which appears to play a significant role in the progression of the postblastoderm cell cycles.

The present work demonstrates coupling of three steps in the centrosome duplication cycle to transitions in the cell cycle. However, uncoupled centrosome duplication has been observed in early embryonic systems and tissue culture cell lines in the absence of visible cell cycle progression (Sluder and Lewis 1987; Gard et al. 1990; Balczon et al. 1995; Hinchcliffe et al. 1998; Lacey et al. 1999). These disparate observations could be explained if coupling is conserved, but unseen local oscillations of activities of cell cycle regulators (maybe even at the centrosome) drive the centrosome cycle. Alternatively, perhaps the coupling itself is differentially regulated in various species and at different stages of development. Thus, the syncytial cell cycles (that precede the postblastoderm cell cycles) may exhibit regulation of the centrosome cycle similar to sea urchin and Xenopus. One mechanism that could account for regulated coupling is regulated expression of inhibitors that block specific steps in the centrosome cycle which, in turn, are then inhibited by cell cycle regulators.

Cyclin E has been shown recently to be required for centrosome duplication (Hinchcliffe et al. 1999; Lacey et al. 1999; Matsumoto et al. 1999). During the postblastoderm cell cycles, cyclin E/Cdk2 activity is continually high (Sauer et al. 1995). Our data do not address roles for cyclin E in Drosophila, but indicate that its continuous activity is not sufficient for continued centrosome duplication in the absence of the requirements that we have uncovered.

We thank Kent McDonald (University of California, Berkeley, Berkeley, CA) for help with high pressure freezing and sectioning; Michelle Moritz for help making Drosophila anti–γ-tubulin; and the members of the O'Farrell lab, Tin Tin Su (University of Colorado, Boulder, Boulder, CO), Anita Sil, and Yuh Nung Jan for critical reading.

This work was supported by an HHMI Predoctoral Fellowship to S.J. Vidwans and a National Institutes of Health grant (GM37193) to P.H. O'Farrell.