Epithelial re-organization and dynamics of progression through mitosis in Drosophila separase complex mutants

Separase is a thiol protease that cleaves α-kleisin subunits (Scc1/Mcd1/Rad21/Rec8 family members) of cohesin complexes and thereby contributes to sister chromatid separation during mitotic and meiotic divisions (Buonomo et al., 2000; Uhlmann et al., 1999; Uhlmann et al., 2000). Before anaphase onset, protease activity is inhibited by securin, a protein that binds to separase during interphase. However, at the metaphase-to-anaphase transition securin is degraded after ubiquitination by the anaphase-promoting complex/cyclosome (APC/C) (Ciosk et al., 1998; Cohen-Fix et al., 1996; Funabiki et al., 1996; Uhlmann et al., 2000; Zou et al., 1999).

Apart from α-kleisin cleavage, budding yeast separase has been shown to function in an additional pathway during exit from mitosis (Stegmeier et al., 2002; Tinker-Kulberg and Morgan, 1999). Interestingly, protease activity does not appear to be required in this FEAR (Cdc fourteen early anaphase release) pathway. Catalytically inactive separase versions are sufficient to trigger Cdc14 release from the nucleolus during early anaphase (Sullivan and Uhlmann, 2003). The released active Cdc14 phosphatase promotes exit from mitosis (Jaspersen et al., 1999; Shou et al., 1999; Stegmeier et al., 2002; Visintin et al., 1998). The inner centromere protein Sli15 is dephosphorylated by Cdc14, resulting in the transfer of Sli15-Ipl1(Aurora) kinase complexes from a centromeric chromosomal localization to the central spindle (Pereira and Schiebel, 2003). Slk19, another yeast protein, which transfers to the central spindle during exit from mitosis, is also regulated by separase. Slk19 associates with separase following securin degradation. Slk19 and its cleavage by separase enhances the stability of anaphase spindles (Sullivan et al., 2001; Ross and Cohen-Fix, 2004). Moreover, Slk19 co-operates with separase in the early anaphase release of Cdc14. In addition to regulating spindle and mitotic exit dynamics, FEAR has recently been shown to promote sister chromatid separation in the nucleolus by stimulating the binding of condensin complexes to the rDNA region (D'Amours et al., 2004; Sullivan et al., 2004).

The FEAR pathway, however, does not appear to be absolutely essential. Completion of mitosis is only transiently delayed by about 30 minutes in separase mutants, but eventually cells exit from mitosis because Cdc14 release in budding yeast can also be triggered by an additional regulatory system, the mitotic exit network (MEN). In wild-type cells, MEN activation is controlled by the mitotic spindle position and it maintains Cdc14 activity after the initial release by FEAR (Bardin et al., 2000; Stegmeier et al., 2002).

It is not yet clear whether pathways comparable to budding yeast FEAR and MEN operate in higher eukaryotes that also express Cdc14 homologues. Some, but not all of the other pathway components can also be identified in higher eukaryotes, but it remains to be determined whether Cdc14 and the other apparent homologues provide the same function as in yeast. In particular, it is not known whether separase provides FEAR-like functions in higher eukaryotes. Cytokinesis and exit from mitosis are not delayed after RNAi-mediated separase elimination in C. elegans (Siomos et al., 2001). In contrast, completion of cytokinesis appears to depend on separase function in human cells (Hauf et al., 2001; Waizenegger et al., 2000).

To elucidate the role of separase during completion of mitosis in further detail, we describe phenotypic analyses in Drosophila. During the evolution of Drosophila, the separase gene has apparently become two genes, three rows (thr) and Separase (Sse) which encode the N-terminal regulatory domain and the C-terminal protease domain, respectively (Jäger et al., 2001; Jäger et al., 2004). Drosophila securin is encoded by the pimples (pim) gene (Leismann et al., 2000; Stratmann and Lehner, 1996). SSE, THR and PIM form a trimeric complex during interphase. After degradation of PIM at the metaphase-to-anaphase transition, SSE promotes sister chromatid separation and also cleaves the associated THR subunit, a process that contributes to timely SSE inactivation (Herzig et al., 2002). We have previously shown that a loss of PIM, THR or SSE function results in a failure of sister chromatid separation during mitosis (D'Andrea et al., 1993; Jäger et al., 2001; Stratmann and Lehner, 1996). In addition, non-cleavable THR versions were shown to interfere with cellularization (Herzig et al., 2002), which corresponds to a modified form of cytokinesis converting the syncytial into the cellularized blastoderm during interphase of cycle 14.

Here, we describe time-lapse analyses of progression through mitosis in live pim and thr mutant embryos. These analyses indicate that Drosophila separase is primarily required for sister chromatid separation. Cytokinesis and exit from mitosis appear to be surprisingly normal in the mutants. Cleavage furrows do not revert because of the non-separated chromatin mass present in these mutants. Interestingly, however, we find that the epithelial organization is severely affected in these mutants.

Fly stocks with the mutant alleles pim¹, thr¹, dup^a1, pbl^7O and pbl^11D have been described previously (Lehner, 1992; Stratmann and Lehner, 1996; Whittaker et al., 2000). A pim¹ dup^a1 double mutant chromosome was obtained by meiotic recombination. The pim¹ pbl double mutant embryos were obtained by crossing males heterozygous for pim¹ and pbl^7O with females heterozygous for pim¹ and pbl^11D. We used blue balancer chromosomes, which allowed the distinction of homozygous mutant from sibling progeny derived from heterozygous parents by anti-β-galactosidase labeling. G147 carries a gene trap insertion in CG31363, resulting in expression of a microtubule-binding GFP fusion protein (Morin et al., 2001). Spider carries a gene trap insertion in gilgamesh, resulting in expression of a GFP fusion protein marking the cell cortex (Morin et al., 2001). The His2AvD-mRFP transgene driving expression of histone H2AvD fused to mRFP1 (Campbell et al., 2002) was constructed analogously to the His2AvD-GFP transgene (Clarkson and Saint, 1999) and will be described in detail elsewhere. A His2AvD-mRFP insertion on chromosome III was recombined meiotically with either the G147 or the Spider gene trap insertion and crossed into the pim¹, thr¹ or pim¹ dup^a1 mutant background.

Eggs were collected on apple juice agar plates and aged to the desired stages. Embryos were dechorionated, aligned and immobilized on coverslips according to standard procedures. After covering embryos with halocarbon oil, confocal laser scanning microscopy on an inverted Leica DM IRBE microscope equipped with a TCS SP1 system was used for time-lapse imaging of GFP and mRFP1 fluorescence signals at 22-24°C in a temperature-controlled room. Light damage was prevented by minimizing the laser intensity and opening the pinhole. Frames were acquired at intervals of 10 or 30 seconds. Representative examples of time-lapse movies are provided as supplementary material.

Embryos aged to the desired developmental stages were fixed according to standard procedures. For immunofluorescent staining we used mouse anti-γ-tubulin GTU-88 (Sigma) at 1:500, rabbit anti-Bazooka (kindly provided by E. Knust, University of Düsseldorf, Germany) at 1:600, rabbit anti-phospho-histone H3 (Upstate) at 1:800, mouse anti-α-spectrin 9A (Dubreuil et al., 1987) at 1:50, rabbit anti-cyclin B (Jacobs et al., 1998) at 1:2000 and mouse (Promega) and rabbit (Cappel) anti-β-galactosidase at 1:250 and 1:1000, respectively. In addition, secondary goat antibodies conjugated to Alexa488 (Molecular Probes), Cy3 or Cy5 (Jackson Immuno Research Laboratories) were applied. DNA was labeled with either Hoechst 33258 (1 μg/ml) or propidium iodide (1 μg/ml) for confocal microscopy. Terminal transferase dUTP nicked-end labeling (TUNEL) assays of apoptotic cells was performed essentially as described previously (Wang et al., 1999). For comparisons of the density of nucleated cells and of the numbers of mitotic cells, we analyzed three and ten pairs of fixed mutant and sibling wild-type embryos, respectively, which were carefully matched with regard to developmental stage and orientation. Fields of identical size from identical regions of the epidermal cell layer were imaged with a 63× objective and used for the determination of cell and nuclear counts.

The Drosophila securin PIM not only functions as an inhibitor of SSE, it also provides a positive function which is absolutely required for sister chromatid separation during mitosis (Stratmann and Lehner, 1996). Nevertheless, initial development of embryos homozygous for pim¹ is normal as long as the maternal pim⁺ contribution to the egg provided by the heterozygous mothers is sufficient. However, starting with mitosis 15, this maternal contribution is no longer sufficient for successful divisions. Sister chromatids are therefore not separated during mitosis 15 in pim¹ mutants (Stratmann and Lehner, 1996). The mutation present in pim¹ affects a splice junction (Stratmann and Lehner, 1996) and we have failed to detect protein products expressed from this allele in immunoblotting experiments (data not shown). To complement our previous characterization of fixed embryos by live analyses, we collected eggs from a pim¹/CyO stock carrying two transgenes that result in expression of a red fluorescent histone H2AvD variant (His2AvD-mRFP) and a green fluorescent microtubule binding protein. Progression through mitosis 15 in both pim¹ and pim⁺ sibling embryos was followed by time-lapse imaging of live embryos. Representative movies from these and the following analyses are available as supplementary material.

As expected from the previous findings with fixed pim embryos, the first mitotic defects became apparent at the metaphase-to-anaphase transition (Fig. 1). Sister chromatids did not separate in pim¹ mutants. However, the dynamic re-organization of the mitotic spindle that accompanies wild-type anaphase and telophase clearly occurred in pim¹ mutants. The observed spindle behavior in pim¹ mutants during exit from mitosis suggested that absence of separase activity does not have a prominent effect on microtubule stability, in contrast to the findings in budding yeast.

Since budding yeast separase accelerates the exit from mitosis significantly, we compared the speed of chromosome decondensation and spindle disassembly in pim¹ and pim⁺ sibling embryos. These comparisons revealed only a minor delay during exit from mitosis in pim¹ mutants (Fig. 1). The time from anaphase onset until the end of mitosis was determined as 4.8 minutes in pim¹ mutant cells (n=10), a value that was only 1.2-fold higher than in the pim⁺ sibling embryos (n=7). In vivo imaging using His2AvD-mRFP in combination with a GFP fusion protein marking the cell cortex confirmed that progression through mitosis 15 occurs at normal speed in pim¹ mutants except for the surprisingly minor delay during exit from mitosis (Table 1).

After exit from mitosis 15, pim¹ mutants progressed through interphase 16 and entered mitosis 16 without apparent delays. However, during mitosis 16, we observed an extensive delay in the mutants both by in vivo imaging (Fig. 2, Table 1) and in fixed embryos (Fig. 3) where the mitotic index was found to be drastically increased. In pim¹ mutants, mitosis 16 lasts on average at least four times longer than in pim⁺ siblings. Therefore in pim¹ mutants mitosis 16 is substantially prolonged in contrast to mitosis 15, which is only slightly delayed.

We considered two potential explanations for the pronounced difference in the dynamics of mitosis 15 and 16 in pim mutants. In principle, the difference might reflect the dynamics of the exhaustion of the maternal pim⁺ contribution. The levels of residual maternally derived pim⁺ function during mitosis 15 might be higher than during mitosis 16. Alternatively, the unusual diplochromosomes, which are present during mitosis 16 in pim mutants as a consequence of the sister chromatid separation failure during the preceding mitosis 15 (Stratmann and Lehner, 1996), might trigger the mitotic spindle checkpoint because of difficulties in integrating these abnormal chromosomes in a regular bipolar manner into the mitotic spindle. In this latter case, the delay is expected to occur before the onset of cyclin B degradation which is known to be blocked by the mitotic spindle checkpoint (Minshull et al., 1989; Whitfield et al., 1990; Minshull et al., 1994). In the former case, however, the delay is not expected to occur before cyclin B degradation because separase is thought to exert its functions only after cyclin B and securin degradation. Immunolabeling of pim mutant embryos fixed at the stage of mitosis 16 clearly revealed that the great majority of mitotic cells still contained high levels of cyclin B (Fig. 3F). Moreover, in vivo imaging confirmed that the delay occurred during metaphase, clearly before the onset of anaphase (Fig. 2).

To test whether the abnormal diplochromosomes are responsible for the delay during metaphase of mitosis 16 in pim mutants, we analyzed embryos with mutations in both the pim and the double-parked (dup) gene. dup encodes the Drosophila Cdt1 homolog and is required for DNA replication (Whittaker et al., 2000). As in pim mutants, a dup⁺ maternal contribution supports normal initial development. However, dup^a1 mutant embryos are unable to replicate chromosomes during cycle 16 (Garner et al., 2001; Whittaker et al., 2000). pim¹ dup^a1 double mutant embryos therefore should not contain diplochromosomes at mitosis 16 and thus, if the mitotic delay in pim single mutants results from the presence of these abnormal chromosomes, it should not occur in the double mutants. Therefore, we analyzed the number of mitotic cells present at the stage of mitosis 16 in pim¹ and dup^a1 single and pim¹ dup^a1 double mutants. Compared to sibling control embryos (Fig. 3A,E), the number of mitotic cells was not only increased in pim¹ mutants (Fig. 3B,E) but also in dup^a1 mutants (Fig. 3C,E). The increased mitotic index in dup^a1 mutants has previously been described and shown to result from the presence of single chromatid chromosomes, which cannot be stably integrated in a bipolar fashion into the mitotic spindle (Parry et al., 2003; Whittaker et al., 2000). The chromosomes therefore fail to congress into a metaphase plate during mitosis 16 in dup^a1 mutants (Fig. 3C, inset) (Parry et al., 2003). Moreover, chromosomes are not exposed to tension within the centromeric region and therefore the mitotic spindle checkpoint remains active (Garner et al., 2001; Whittaker et al., 2000). In contrast to pim¹ and dup^a1 single mutants, we did not observe an increased number of mitotic cells in the pim¹ dup^a1 double mutants (Fig. 3D,E). In addition, chromosomes were found to congress into a metaphase plate (Fig. 3D, inset). These findings, therefore, indicate that the abnormal diplochromosomes are largely responsible for the delay observed during mitosis 16 in the pim¹ single mutants. This conclusion was fully confirmed by in vivo imaging of the progression through mitosis 16 in pim¹ dup^a1 double mutants (Table 1).

The protein encoded by the thr gene corresponds to the N-terminal regulatory domain of separase proteins of other eukaryotes (Jäger et al., 2001; Jäger et al., 2004). The maternal thr⁺ contribution present in thr¹ mutants is sufficient for normal initial development until mitosis 15. However, just as in pim mutants, sister chromatid separation during mitosis 15 fails in thr¹ mutants (D'Andrea et al., 1993). With our antibodies raised against the C-terminal third of THR, we were unable to detect protein products expressed from the thr¹ allele, indicating that the mutant protein is either unstable or C-terminally truncated (data not shown). By in vivo imaging we analyzed whether the thr¹ mutation affects the dynamics of exit from mitosis. However, as observed in the pim¹ mutants, we detected only a slight delay during exit from mitosis 15 and 16 (Table 1). In addition, we also observed an extensive delay during metaphase 16 (Table 1). To date, neither in vivo imaging nor other analyses have revealed phenotypic differences in pim¹ and thr¹ mutants.

Our previous phenotypic characterizations of fixed pim and thr mutant embryos had suggested that cytokinesis during mitosis 15 is not completed in these mutants (D'Andrea et al., 1993; Stratmann and Lehner, 1996). Cleavage furrows pinching the undivided chromosomes were often observed in fixed mutant embryos at the stage of mitosis 15. However, at a later stage, during the following interphase 16, the nuclear density in the epidermal cell layer of fixed mutant embryos was found to be 1.6-fold lower than in sibling control embryos (D'Andrea et al., 1993; Stratmann and Lehner, 1996) (see below), suggesting that the cleavage furrows had failed to cut completely through the chromosomes during mitosis 15. The cleavage furrows were thus thought to have had aborted in most cells.

After such an inferred cytokinesis failure during mitosis 15 and subsequent duplication of centrosomes during cycle 16, mitotic spindles would be predicted to be tetrapolar during mitosis 16 in pim¹ or thr¹ mutant cells. In contrast to this expectation, we observed apparently bipolar mitotic spindles and normal metaphase plates during mitosis 16 in pim¹ and thr¹ mutants (Fig. 3F, and data not shown). Immunolabeling of centrosomal γ-tubulin demonstrated that the majority of the epidermal cells in pim¹ and thr¹ mutant embryos have only two and not four centrosomes at the stage of mitosis 16 (Fig. 4).

In contrast to pim¹ and thr¹ mutant embryos, in pebble (pbl) mutants the centrosome number was found to increase to four during the cycle following a cytokinesis failure (Fig. 4, Table 2). pbl encodes a Rho-GEF required for cytokinesis (Hime and Saint, 1992; Lehner, 1992; Prokopenko et al., 1999). This apparent difference in centrosome behavior in pim¹ and thr¹ mutants, on the one hand, and pbl mutants, on the other, raised the question of whether the Drosophila separase complex might be required during mitosis for centrosome duplication in the following cell cycle. However, in pim pbl double mutants, centrosome number per cell was found to increase with each cycle (Fig. 4), indicating that centrosome duplication is not dependent on the function of the separase complex.

The observed behavior of centrosomes in pim¹ and thr¹ mutants would be readily explained, if these mutations did not interfere with completion of cytokinesis during mitosis 15. Therefore, we re-evaluated cytokinesis in pim mutants, this time by in vivo imaging of embryos expressing a green fluorescent cell cortex marker in addition to His2AvD-mRFP. Time-lapse recordings clearly demonstrated that cytokinesis is completed successfully during mitosis 15 in the great majority of the epidermal cells in both pim¹ and thr¹ mutants (Fig. 5, and data not shown). The dynamics of cytokinesis was only slightly slower in the mutants compared to sibling controls (Fig. 5, Table 1). The mass of non-separated chromosomes therefore is not a significant obstacle for cytokinesis. In about half of the cells, the cleavage furrow was observed to cut through the chromatin mass, resulting in two daughter cells, which frequently contained unequal amounts of chromatin. In the other half of the cells, the chromatin mass moved towards one pole and cleavage resulted in an enucleate and a nucleate cell pair. Cleavage furrows were not observed to revert after mitosis 15 even during extended observation periods. During the following mitosis 16 in pim¹ and thr¹ mutants, cytokinesis was considerably more abnormal and variable from cell to cell (data not shown). Cleavage furrows were rarely observed to cut through the chromatin mass and they appeared to revert occasionally.

Our finding that cytokinesis is completed successfully in pim¹ and thr¹ mutants, at least during mitosis 15, was a surprise in the light of the significantly lower densities of cells and nuclei within the superficial epidermal layer in these mutants during cycle 16. Based on our analysis of cytokinesis during mitosis 15, cell density before mitosis 16 should be comparable in pim¹ and pim⁺ sibling embryos and the nuclear density should be 25% lower in the mutants. However, cell and nuclear densities were significantly below these expectations. Quantification of the nuclear densities revealed a 38% reduction in pim¹ before mitosis 16. A comparable extensive decrease in the density of nucleated cells was also observed in thr¹ mutant embryos.

In principle, this more extensive reduction might result from apoptosis of the aneuploid cells generated by cytokinesis despite sister chromatid separation failure in pim¹ and thr¹ mutants. To compare the rate of apoptosis in pim¹ and pim⁺ sibling embryos we used the TUNEL assay, which detects nucleate apoptotic cells. We did not observe any difference in apoptosis between pim¹ and pim⁺ sibling embryos until after the stage of mitosis 16. Up to mitosis 16, we observed only very few apoptotic cells, in the characteristic developmental pattern of programmed cell death in both pim¹ and pim⁺ sibling embryos (Fig. 6A,B). After mitosis 16, during germband retraction, pim¹ embryos had slightly more apoptotic cells in regions where developmentally programmed cell death in wild-type is very rare (Fig. 6C,D). After full germband retraction, the number of TUNEL-positive cells was clearly, but still not dramatically, increased in pim¹ mutants (data not shown). Aneuploid cells in pim mutants therefore appear to undergo apoptosis eventually, but only after mitosis 16.

To quantify the cellular and nuclear densities in pim mutants after mitosis 15, our initial attempts involved double labeling with a DNA stain and an antibody against the apical junction protein Bazooka (BAZ). In pim⁺ sibling embryos, each cell displayed an apical ring of anti-BAZ above a centrally located nucleus, as expected. However, in pim¹ mutants, nuclei were very frequently displaced relative to the apical ring of anti-BAZ staining, indicating that the cylindrical organization of the wild-type epithelial cells had been lost (data not shown). Double labeling of DNA and the cell cortex protein α-spectrin further confirmed that the epithelial organization in pim¹ mutants had an abnormal pseudostratified appearance after mitosis 15 but not before (Fig. 7). While single confocal sections through the nucleus of every epidermal cell could readily be obtained in pim⁺ sibling embryos before mitosis 16 (Fig. 7C), there were striking irregularities in the sections of pim¹ mutants and they never included all of the nuclei of the nucleated epidermal cells present in other planes of the z stacks (Fig. 7D).

To address whether Drosophila separase complex subunits provide functions beyond sister chromatid separation during mitosis, we have further characterized the phenotypic consequences of mutations in pim and thr, which encode the Drosophila securin and the equivalent of the N-terminal regulatory domain of non-dipteran separases, respectively.

Our time-lapse analyses of live embryos demonstrate that progression through mitosis 15, i.e. the first mitosis during which the maternal contribution is no longer sufficient for sister chromatid separation in the pim and thr mutants, is only marginally delayed. The duration of pro- and metaphase 15 did not appear to be affected and the time from anaphase onset until the start of cytokinesis was also not prolonged in the mutants. In contrast, completion of cytokinesis was slightly but significantly extended. This extension by about 1 minute in the Drosophila mutants is less than the 20-30 minutes delay during exit from mitosis that results from mutations in the ESP1 separase gene in budding yeast (Stegmeier et al., 2002; Sullivan and Uhlmann, 2003). The difference in the extent of mitotic delays apparent in Drosophila and budding yeast is reduced but not eliminated when expressed relative to the overall duration of exit from mitosis, which lasts about eight times longer in budding yeast.

While sister chromatid is largely inhibited during mitosis 15 in pim and thr mutants, one exceptional, apparently normal division could be observed in the hundreds of analyzed cells. Therefore, the maternal pim⁺ and thr⁺ contributions are unlikely to be completely eliminated in the mutants by the time of mitosis 15. As PIM and THR are degraded during mitosis (Herzig et al., 2002; Leismann and Lehner, 2003; Stratmann and Lehner, 1996) any residual maternal contribution, which is still present during mitosis 15, is expected to be further reduced at the stage of mitosis 16. The slightly more extensive delay during cytokinesis and exit from mitosis that is observed in the mutants during mitosis 16 might therefore be taken as an indication that a complete elimination of PIM and THR might result in extensive delay. However, alternative explanations are not excluded. Exit from mitosis 16 might be kinetically abnormal not because of a further depletion of the maternal contribution but because of some indirect consequences of the mitosis 15 defects. Mutant cells appear to exit mitosis 16 after adaptation to a spindle checkpoint arrest resulting from mitosis 15 defects (see below). The adaptation mechanisms are unknown and they might have effects on mitosis exit dynamics, which are absent during normal mitosis 16. Such alternative, potential explanations for the slightly stronger delay during exit from mitosis 16 compared to mitosis 15 underscore the technical difficulties of experimental analysis after complete elimination of PIM and THR. In vertebrate systems, the effect of separase on mitotic exit kinetics has not yet been addressed. In C. elegans, elimination of separase to a level insufficient for sister chromatid separation does not significantly delay exit from mitosis in controlled osmotic conditions (Siomos et al., 2001).

Arguably the most striking effect on the dynamics of progression through mitosis observed in pim and thr mutants, i.e. the extensive delay during metaphase of mitosis 16, results indirectly from the failure of sister chromatid separation during mitosis 15. After replication of these non-separated chromatid pairs during S phase 16, pim and thr mutant cells enter mitosis 16 with abnormal diplochromosomes (Stratmann and Lehner, 1996). Here we demonstrate, with the help of mutations in Drosophila Cdt1/Dup, that inhibition of S phase 16 in pim mutants prevents most of the delay during metaphase 16. The metaphase extension observed in pim and thr mutants might therefore be caused by spindle checkpoint activation predominantly resulting from difficulties with bipolar integration of abnormal diplochromosomes into mitotic spindles. We do not understand why the delay during metaphase 16 is not completely reversed in the pim dup double mutants, but the residual delay might again reflect adaptation effects, in this case after incomplete inhibition of progression through the preceding S phase 16 (Garner et el., 2001).

Mutations in budding yeast ESP1, which abolish protease activity and prevent sister chromatid separation, do not necessarily eliminate activity in the FEAR pathway (Sullivan and Uhlmann, 2003). Therefore the Drosophila separase protein which is still present in pim and thr mutants (Herzig et al., 2002) might be sufficient, in principle, to activate a putative FEAR pathway. Thereby Drosophila separase might prevent more extensive mitotic delays in pim and thr mutants. However, experiments in budding yeast have demonstrated that relatively minor N-terminal truncations abolish FEAR activity of Esp1 (M. Sullivan and F. Uhlmann, personal communication). It appears unlikely, therefore, that Drosophila separase, which corresponds to the C-terminal protease domain of Esp1, should be able to function in a putative homologous FEAR pathway in a monomeric form without its usual complex partner THR, which corresponds to the N-terminal domain of Esp1.

In budding yeast, both the proteolytic and the FEAR activity of Esp1 contribute to spindle stability during anaphase, in part by recruiting the budding yeast INCENP-aurora B complex to the central spindle (Pereira and Schiebel, 2003; Ross and Cohen-Fix, 2004; Sullivan et al., 2001). In Drosophila, INCENP and aurora B still transfer to the central spindle in pim and thr mutants during mitosis 15 (A. Herzig and C.F.L., unpublished). Moreover, in vivo imaging has not revealed severe abnormalities in microtubule organization during exit from mitosis 15. Cytokinesis also proceeded surprisingly normally in pim and thr mutants. Despite an equatorial mass of undivided chromosomes, cleavage furrows contracted rapidly and completely in the mutants during mitosis 15. Therefore, non-separated equatorial chromosomes appear to affect completion of cytokinesis to a variable extent in different cell types. In Drosophila and fission yeast, cleavage furrows are definitely able to cut readily through the chromosomes (Uzawa et al., 1990) (Fig. 5). In contrast, in cultured human cells, eventual regression of the cleavage furrow has been observed after expression of mutant SCC1 versions, which cannot be processed by separase (Hauf et al., 2001).

The effects on cytokinesis and spindle organization in the mutants were more prominent during mitosis 16 compared to mitosis 15. Again, as discussed for the mitotic exit kinetics above, this increase in the severity of observed defects might reflect the progressive depletion of the maternal contributions and/or indirect consequences of earlier defects. A rather stochastic segregation of undivided chromosomes during mitosis 15 might generate phenotypic variability during mitosis 16. The observation that some cells exit mitosis 16 in a manner comparable to mitosis 15, while others are more severely affected, therefore argues for indirect consequences.

An interesting aspect of the pim and thr mutant phenotype revealed by our analyses concerns epithelial organization. Cellularization during wild-type embryogenesis results in a regular monolayer of cylindrical cells, which is maintained in the epidermis throughout the three postblastoderm division cycles, 14-16. In the mutants, this regular epithelial organization is lost after mitosis 15. The abnormal pseudostratified appearance of the epidermis develops rapidly and long before apoptotic responses. The loss of epithelial organization is therefore unlikely to be a consequence of altered gene expression in the aneuploid cells. Rather the loss of separase complex function might affect epithelial organization via effects on the cytoskeleton. Since budding yeast separase has been shown to regulate microtubule stability, a similar role of the Drosophila separase complex remains an attractive explanation. Moreover, after mitosis 15, pim and thr mutant cells contain very variable amounts of chromatin. They contain either no nuclei or nuclei of variable sizes. Since the nucleus might function as a mechanical element within the cells, its variability might contribute to the epithelial disorganization in the mutants. In addition, even though cleavage furrows are able to cut through the undivided chromosomes quite effectively and central spindles as well as midbodies are formed, it is possible that these latter structures are not fully functional, resulting in an inhibition of the establishment of effective junctional contacts between the newly formed daughter cells. The rounding up of cells on the apical side, which occurs upon entry into mitosis, might provide a force that, in particular, displaces some of the already divided cells towards the basal side. Our phenotypic analyses in Drosophila embryos emphasize that a loss of separase function can have consequences that might be difficult to observe with cultured cells. Loss of separase regulation, as for instance after overexpression of the oncogenic human securin PTTG (Melmed, 2003), might promote tumors, not only by increasing genetic instability but also by effects on tissue organization.

We thank A. Herzig for his help with experiments, X. Morin and A. Debec for providing Drosophila strains and E. Knust for antibodies. This work was supported by the Fonds der Deutschen Chemie and by grants from the Deutsche Forschungsgemeinschaft (DFG Le 987/3-1, 3-2, 3-3).