Loss of Drosophila borealin causes polyploidy, delayed apoptosis and abnormal tissue development

The chromosomal passenger complex (CPC) is conserved from yeast to humans,and consists of at least three components that regulate multiple mitotic events. Its name stems from the observation that CPC proteins colocalise on condensing chromosomes during prophase, and are carried along to centromeres and to the equator of the mitotic spindle during metaphase(Earnshaw and Bernat, 1991). After metaphase, the CPC components re-localise to the midzone and midbody of the spindle, where they remain until the completion of cytokinesis(Andrews et al., 2003; Carmena and Earnshaw, 2003). The CPC components include Aurora B kinase, inner centromere protein (Incenp)and Survivin, an inhibitor of apoptosis-like protein(Bolton et al., 2002), as well as the recently discovered Borealin/Dasra protein(Gassmann et al., 2004; Sampath et al., 2004).

Borealin/Dasra was identified in human cell lines and in Xenopusextracts, respectively, and found to colocalise with other CPC proteins throughout mitosis (Gassmann et al.,2004; Sampath et al.,2004). The correct localisation of human Borealin in mitotic cells depends on the function of the other CPC components; conversely, RNAi-mediated depletion of Borealin in HeLa cells causes mislocalisation of Aurora B, Incenp and Survivin (Gassmann et al.,2004). Human Borealin binds directly to Incenp and Survivin in vitro, and forms a complex with the other CPC components in vivo. Its loss of function, like that of other CPC components, causes multiple mitotic defects,including failures in chromosome attachment to the spindle, multifocal spindles and uneven chromosome segregation. This typically results in multinucleate cells, aneuploidy and polyploidy, as well as, ultimately,apoptosis (Gassmann et al.,2004; Sampath et al.,2004). However, cells that lack CPC function can also occasionally escape apoptosis as they appear to be defective for their spindle attachment checkpoint (Lens and Medema,2003; Yang et al.,2004).

Little is known about the role of the CPC during development, except for its function in the early C. elegans embryo(Kaitna et al., 2000; Kaitna et al., 2002; Romano et al., 2003). Here, we present the first detailed characterisation of a CPC mutation in Drosophila, using a loss-of-function allele of borealin. This gene was identified independently in a recent RNAi screen for cytokinesis defects in cultured Drosophila cells, and was named borr(borealin-related) (Eggert et al., 2004). We provide evidence, based on its subcellular localisation and function during the cell cycle, that Borr is the functional counterpart of vertebrate Borealin/Dasra. We show that borr is an essential gene, and that loss of borr function causes mitotic defects, including multipolar spindles that result in large polyploid cells and often in delayed apoptosis. The developmental consequences of these defects include striking cell-autonomous and non-autonomous defects in cell-type specification and tissue architecture.

E133 was isolated fortuitously in an EMS screen for genes interacting with activated Armadillo (Thompson et al.,2002). E133 turned out to be a non-interacting `passenger' hit,and its lethality was mapped to 30A on chromosome 2L. RNA interference experiments (see below) identified CG4454 as the gene affected by E133. Double-stranded sequencing of genomic DNA identified a single base-pair deletion in the CG4454 coding region.

Borr was tagged N-terminally with green fluorescent protein (GFP) by Gateway cloning (Invitrogen), using the borr cDNA from clone LD36125 and the pAGW vector (Terence Murphy, Carnegie Institution of Washington). The resulting construct was confirmed by sequencing.

Kc167 cells were cultured at 25°C in Schneider's medium supplemented with 10% heat-inactivated foetal bovine serum and antibiotics. DmD8 cells were obtained from the Drosophila Genome Resource Center, and cultured similarly,with 10 μg/ml insulin (Sigma) added to the medium. They were transfected with the FuGene transfection reagent (Roche) according to the manufacturers instructions, with a ratio of 4 μg DNA:1 μl FuGene. Cells were processed for analysis 24 hours after transfection.

To identify the gene responsible for the embryonic phenotype of E133, an RNA interference screen of candidate open reading frames within the genomic region 30A was performed, as follows. Genomic DNA was isolated from yw flies, and amplified by PCR with primer pairs containing a T7 promoter sequence at the 5′ end designed to amplify a large uninterrupted stretch of coding DNA. PCR products were used as templates in transcription reactions using the MegaScript RNAi kit (Ambion), which resulted, in the case of CG4454, in a dsRNA of 264 bp. The predicted size of the dsRNA products was verified by agarose gel electrophoresis, and their concentrations were determined by comparison with a known standard.

Injection of dsRNA into embryos was carried out as described(Desbordes and Sanson, 2003),except that the dsRNA was delivered in water. All preparation and injection steps were carried out at room temperature, and the embryos were aged for∼24 hours at 18°C before fixation.

RNAi of Kc167 cells was carried out basically as described(Clemens et al., 2000), except that 500 μl of cells were plated at a concentration of 10⁶ per well of a 24-well plate. Control cells were treated identically, but without dsRNA.

To estimate nuclear volumes, individual wild-type and borr mutant ventral nerve cord (VNC) nuclei stained with Hoechst were outlined using ImageJ, and their maximal circumference was measured. From these measurements,the volumes of the corresponding spheres were calculated, providing estimates of nuclear volumes. This modelling of nuclear volumes by spheres was validated as a best approximation by 3D reconstructions of individual nuclei. To estimate mitotic indices, the mitotic cells were identified on the basis of chromatin morphology, Hoechst and serine-10 phosphorylated histone H3 (P-H3)staining, and their numbers were determined per hemineuromere for abdominal segments 4, 5 and 6 (see also Results and Fig. 4 legend).

FRT/FLP mediated recombination (Xu and Rubin, 1993) was used to induce homozygous mutant borrclones. Flies of the genotypes borr^E133 FRT40A/SM6a-TM6band yw hsflp; Ub-NLS-GFP FRT40A/CyO or f yw hsflp; ck,f+FRT40A/Cyo (kindly provided by K. Basler) were crossed. Embryos were collected for 24 hours, aged at 25°C, and heat-shocked after a further 36 or 84 hours. Mutant phenotypes were analysed in dissected larval imaginal discs, dissected pupal wings or in adult tissues.

Embryos were immunostained as previously described(Cliffe et al., 2003). Imaginal discs and pupal wings were stained using standard methods. Briefly, tissues were fixed with 4% formaldehyde (30 minutes at room temperature for imaginal discs, overnight at 4°C for pupal wings), washed, blocked and incubated overnight at 4°C with primary antibodies in PBS+0.1% Triton-X-100+1% BSA(BBT). Tissues were then washed several times in BBT and incubated with secondary antibodies (Molecular Probes) for 2-3 hours at room temperature. The following primary antibodies were used: mouse E7 anti-β-tubulin (1:100;Developmental Studies Hybridoma Bank, DHSB); rabbit anti-P-H3 (1:500; Abcam);rabbit anti-activated human caspase 3 (1:700; BD Biosciences), which has been shown to cross-react with the Drosophila ortholog(Yu et al., 2002); mouse anti-Wg (1:100; DHSB); mouse anti-Cut (1:100; DHSB); rabbit anti-GFP (1:2000;gift from R. Arkowitz); guinea pig anti-Senseless (1:1000)(Barbosa et al., 2000); rabbit anti-Aurora B (Giet and Glover,2001) (1:200); and rabbit anti-Incenp(Adams et al., 2001) (1:500). DNA was stained with Hoechst dye or DAPI(Fig. 5). Images were collected on a BioRad 1024 confocal microscope or a Zeiss Axiovert 200M(Fig. 5).

E133 is a loss-of-function allele of CG4454, with a single base pair deletion at position 290 in the first exon of its coding region. The resulting frameshift introduces a stop codon immediately after this deletion into the predicted protein, truncating it after serine 98(Fig. 1A). The CG4454 locus consists of three exons that encode a protein of 319 amino acids, without recognisable domains or known sequence motifs. Stringent Psi-Blast searches revealed a significant similarity between CG4454 and Borealin/Dasra, the only protein with any detectable sequence relationship to CG4454 (see also Gassmann et al., 2004). This suggests that CG4454 may be the Drosophila ortholog of Borealin/Dasra.

Zygotic homozygosity for the borr mutation results in late embryonic lethality, but the mutant embryos lack overt morphological defects,probably owing to rescue by maternal gene product. Consistent with this, borr is ubiquitously expressed in the early Drosophilaembryo, although it appears to be restricted to the VNC and brain during later embryonic stages (Berkeley Drosophila Genome Project in situ data).

Given its high expression levels in the embryonic nervous system, we scrutinised this tissue more carefully after staining embryos with Hoechst dye. Indeed, by stage 12, we detected cells in the VNC and brain with abnormally large nuclei (Fig. 1B,C). We estimate that the volumes of the borr mutant VNC nuclei are on average ∼3 times larger than those of wild-type VNC nuclei (Fig. 1D). This implies an increased DNA content (>2N) of the mutant cells, and suggests that borr loss affects the divisions of VNC cells. We also observed similarly oversized nuclei in other tissues (in addition to severe morphological defects such as failure of germ band retraction), after injection of borr dsRNA into wild-type embryos, which potentially also depletes maternal gene product (not shown). Thus, borr loss appears to affect many, if not all, dividing cells in the embryo.

To monitor the mitotic events that are affected in the borr mutant embryos, we stained these embryos with an antibody against serine 10 phosphorylated histone H3 (P-H3), a histone modification specifically found in mitotic cells that has been ascribed to Aurora B kinase activity in several organisms, including Drosophila (see below)(Giet and Glover, 2001; Hsu et al., 2000). Counting the mitotic cells per hemi-neuromere in wild-type and borr mutant embryos, we found that these numbers were reduced significantly in the mutants, to ∼50% of the wild type at stage 12, and to ∼20% at stage 14(Fig. 2A; see also Fig. 4). Our estimates suggest that, in mutant embryos, the overall number of cells per hemi-neuromere is also lower than normal (although it is technically difficult to obtain accurate counts of total cell numbers). Nevertheless, these counts suggest that the fraction of mitotic cells (i.e. the mitotic index) in the VNC of borr mutant embryos may be reduced compared with the wild type.

To see whether the borr mutation affects a specific mitotic stage,we classified each P-H3-positive cell as one of four different mitotic stages(based on the shapes of their chromatin masses; see below), and we determined the frequencies of these stages as a percentage of the total of mitotic cells. This revealed that the percentages of prophase and prometaphase cells were higher in borr mutants compared with the wild type, whereas anaphases and telophases were underrepresented in the mutants(Fig. 2B,C). This profile shift of the mitotic stages appears to be progressive during embryonic development,and becomes more pronounced by stage 14 when telophases have become exceedingly rare (Fig. 2C),maybe as a result of cumulative defects during consecutive abnormal cell divisions. This profile shift suggests that borr loss causes a severe attenuation, or block, prior to metaphase.

Two further features were noticeable in the P-H3 staining patterns of the borr mutant VNC cells. First, many of the rare anaphases detected at stage 12 appeared abnormal, showing evidence of uneven segregation of chromatin (Fig. 2D; see also Figs 3, 5). Second, the P-H3 staining intensity was reduced markedly, which is particularly noticeable during metaphase, but also during telophase when P-H3 staining normally fades away(Fig. 2D; see also Fig. 4). These observations are consistent with the profile shift of the mitotic stages in borrmutant embryos (Fig. 2B,C), and they underscore the notion that the first major defect during the mutant cell cycle occurs prior to metaphase. A similar prometaphase block has been reported for human Borealin (Gassmann et al., 2004) and for other CPC components in Drosophilacells (Adams et al., 2001; Giet and Glover, 2001).

In order to observe the subcellular localisation of Borr, Drosophila DmD8 cells were transfected with a construct encoding GFP-tagged full-length Borr. As expected, GFP-Borr is associated with chromatin during prometaphase (Eggert et al., 2004) (not shown), and is subsequently concentrated at the central spindle midbody and at the cell cortex in the cleavage furrow during telophase and cytokinesis (Fig. 3A-C,E,F). We shall refer to this pattern as `localisation to the mitotic spindle'. Significantly, GFP-Borr colocalises with both endogenous Aurora B and Incenp (Fig. 3B-G), in agreement with the results by Eggert et al.(Eggert et al., 2004), who also observed co-localisation of Borr and Aurora B throughout mitosis. These results are consistent with Borr being a CPC component, like its vertebrate counterparts.

To further study the function of borr during mitosis, we used dsRNA interference in Drosophila Kc167 tissue culture cells. Indeed,72 hours after addition of borr-specific dsRNA, Kc167 cells displayed a range of mitotic defects when compared with their controls(Fig. 3H-M). Most notably,highly abnormal multipolar spindles were observed in mitotic cells(Fig. 3I,J), and interphase cells often showed single large nuclei – reminiscent of the VNC nuclei in borr mutant embryos (Fig. 1C) – or became multi-nucleate(Fig. 3K-M). Some of these cells appear to have up to eight distinct nuclei, in addition to DNA fragments strewn around the cytoplasm (Fig. 3M, arrows). Similar phenotypes were observed in HeLa cells after RNAi-mediated depletion of Borealin, and also after RNAi-mediated depletion of CPC components in Drosophila cells(Adams et al., 2001; Eggert et al., 2004; Gassmann et al., 2004; Giet and Glover, 2001; Sampath et al., 2004). They support the notion that Borr is a functional ortholog of human Borealin. Furthermore, the multi-nucleate cells and the multipolar spindles suggest that Borr is required for faithful segregation of chromosomes during mitosis, and that its loss can cause polyploidy and/or aneuploidy (for simplicity, we shall refer to this as `polyploidy').

One crucial role of the CPC during mitosis is to mediate the H3 phosphorylation of serine 10 (P-H3) by Aurora B, as has been demonstrated in budding yeast, C. elegans and Drosophila(Adams et al., 2001; Giet and Glover, 2001; Hsu et al., 2000). As already mentioned (Fig. 2), the numbers of P-H3-positive (dividing) cells are reduced in the VNC of borrmutant embryos (Fig. 4A-D). Furthermore, the P-H3 levels of individual borr mitotic nuclei are typically reduced compared with those of wild-type nuclei(Fig. 4E-J; see also Fig. 2D). Often, they exhibit blotchy P-H3 staining (Fig. 4H,J) rather than the more `structured' staining outlining condensed chromosomes as observed in the wild type(Fig. 4E,G). A similar loss of P-H3 staining has also been observed in borr RNAi-depleted Kc167 cells (Eggert et al., 2004). This reduction of the P-H3 levels in borr mutant cells is consistent with a loss of Aurora B kinase activity and, thus, with a disruption of CPC function.

Despite the strong reduction of the P-H3 levels in mitotic VNC cells of borr mutant embryos, these cells display only a slight undercondensation of their chromatin (Fig. 4I, arrow, compare with mitotic cell in F), although the degree of undercondensation is somewhat variable from cell to cell(Fig. 4I, and not shown). These results suggest that borr may not be essential for chromatin condensation.

To examine the effects of borr loss on actively dividing epithelial cells, we used FRT-FLP-mediated recombination(Xu and Rubin, 1993) to generate borr mutant clones in imaginal discs whose cells undergo cell divisions throughout larval development. If borr mutant clones are induced during early larval stages and examined in fully grown larval discs, these clones are rare and are much smaller than the corresponding wild-type twin spots, suggesting that a large fraction of the mutant cells die(see below). Hoechst staining revealed that many of the surviving borr mutant cells are large, with giant but well-formed nuclei that appear healthy, and well integrated into the epithelial tissue (see Movie 1 in the supplementary material).

We stained imaginal discs bearing borr mutant clones with antibodies against Incenp and Aurora B, to assess the effect of borrloss on these CPC components during mitosis. Wild-type cells in metaphase show characteristic well-ordered mitotic spindles, with distinct staining of Aurora B and Incenp at specific sites along condensed chromatin(Fig. 5A,F,B′-J′). By contrast, borr mutant cells invariably show abnormal mitotic spindles, including multipolar ones (Fig. 5A-J). Most of these mutant spindles do not show any chromatin-associated Incenp or Aurora B staining(Fig. 5C,H), although occasionally patches of Incenp staining can still be observed, but they do not seem to be associated with any of the spindle components (not shown). These staining patterns suggest that these CPC components fail to localise properly to mitotic spindles in the absence of borr (and their levels may also be reduced, though the low frequency of surviving borr mutant cells does not allow us to assess this quantitatively). Therefore, as in mammalian cells, the correct localisation of Incenp and Aurora B to mitotic spindles of dividing imaginal disc cells depends on Borr. This is further evidence that Borr is a CPC protein, and that it interacts functionally with other known CPC components.

As already mentioned, early-induced borr mutant clones are rare,and are much smaller than their twin spots(Fig. 6A, blue arrow). Indeed,many twin spots do not appear to have mutant cells associated with them(Fig. 6A, red arrow),indicating that the mutant cells have all died. The frequency of surviving borr mutant clones is increased if they are induced in a Minute background, which provides the mutant cells with a proliferative advantage. They can thus occupy a significant fraction of imaginal disc territories in third instar larvae(Fig. 6B). All discs are equally affected, and they tend to be smaller than wild-type discs of an equivalent stage. Larvae with these clones do not survive pupariation.

Closer examination of the borr mutant cells revealed essentially two distinct phenotypes: large cells with giant well-formed nuclei, as described above (Fig. 6B,C,grey arrow), and cells that appear to be undergoing apoptosis. The clearest examples of the latter show compacted almost perfectly spherical nuclei that are found at the basal-most level of the disc epithelium, well separated from the healthy nuclei of the wing pouch (Fig. 6C, white arrow). We also observed borr mutant cells that may be at an earlier step in the apoptotic process: their nuclei are less compacted, and they are just beginning to drop basally within the epithelium(Fig. 6D, red arrow). Antibody staining against active caspase 3 confirmed that the borr mutant cells with compacted DNA are indeed undergoing apoptosis(Fig. 6E, white arrows), in contrast to the borr mutant cells that are well-integrated into the epithelium and display only background levels of active caspase 3 staining(Fig. 6E, grey arrow). Cells with low caspase staining can also be observed(Fig. 6E, red arrow): these show apparently fragmented but not yet compacted DNA, and may thus represent an intermediate stage similar to that shown in Fig. 6C.

These results, together with our observations in Borr-depleted embryos and tissue culture cells, suggest that borr mutant cells can undergo several consecutive abnormal mitoses, which results in large polyploid cells that eventually undergo apoptosis. Apoptotic cells appear to be cleared by basal extrusion from the epithelium.

To assess the consequences of Borr loss on the development of the imaginal discs, we induced borr mutant clones in first or early second instar larvae, and we examined the resulting adult flies. The most common defects in these flies are abnormal legs and rough eyes (see Fig. S1 in the supplementary material). In addition, they often show other striking defects in tissue architecture, e.g. large wing nicks (Fig. 7A,B). In all these cases, a twin spot is apparent (e.g. Fig. 7C, outlined in white),but no mutant tissue is detectable. This indicates that, by the adult stage,each of these early-induced borr mutant cells has undergone apoptosis. The nature and extent of the adult defects also suggests that they may be due partly to non-autonomous effects of the borr mutant clones on their neighbouring wild-type tissue.

To gain more direct evidence for these putative non-autonomous effects, we examined the expression of Wingless (Wg) in wing discs bearing borrmutant clones, a secreted morphogen that is expressed in a thin stripe along the developing margin of the wild-type disc(Fig. 7D) and controls its formation (Couso et al., 1994; Neumann and Cohen, 1996). As expected from the adult phenotypes, Wg expression is perturbed in various ways by borr mutant clones. Some of the surviving giant borrmutant cells within the Wg-expressing territory cause a significant lateral expansion of Wg staining by virtue of their sheer size(Fig. 7E-G, red arrows). Other cases of expanded Wg staining are not detectably associated with mutant cells(Fig. 7E,G, blue arrows), and thus appear to be cell non-autonomous consequences of borr loss.

We also observe clear non-autonomous effects of borr mutant cells if we examine the expression of cut and senseless, two of the ultimate target genes responding to the Wg morphogen in the marginal region (Neumann and Cohen,1996; Parker et al.,2002). For example, a single surviving giant borr mutant cell expressing high levels of Cut can cause suppression of Cut and Senseless expression in neighbouring wild-type cells(Fig. 7H-J). A similarly striking example is the introduction of a V shape into the patterns of Cut and Senseless expression caused by a borr mutant clone(Fig. 7K-M). The presence of a large twin spot associated with this abnormality indicates that the causative borr mutant clone arose early when the disc contained only a small number of cells. Again, the borr mutant cells have disappeared in this case, most likely through apoptosis (see above). The kink introduced into the expression domains of both proteins appears to coincide with a rearrangement of cells in this region (Fig. 7L, arrow). Indeed, it appears that a single giant borrmutant cell (see Movie 1 in the supplementary material), in the process of basal displacement, might drag along normal epithelial cells. Thus, apoptosis and basal extrusion of a giant cell may exert sufficient disruption of the epithelium to induce compensatory cell rearrangements aimed at repairing epithelial integrity, which in the event compromise patterning.

If borr mutant clones are induced late (from the early third larval instar onwards), the resulting flies are viable and display no gross patterning defects. Indeed, analysis of marked clones and twin spots in adult wings suggests that all borr mutant clones are fully viable, given that they occupy roughly the same amount of territory as their twin spots(Fig. 8A). This is somewhat unexpected in the light of our results with earlier-induced clones whose survival was severely compromised (Figs 6, 7) owing to abnormal mitoses(Fig. 5). Indeed, the size of the late-induced borr mutant clone in Fig. 8A indicates that the mutant cells have survived three or four consecutive (abnormal) mitoses without entering the apoptotic pathway.

Closer examination of the flies bearing late-induced borr mutant clones revealed that their wing blades contain clusters of hairs (trichomes)surrounded by large clearings, rather than the usual regularly spaced single hairs (Fig. 8A). The number of hairs per cluster varies, with the largest cluster observed consisting of 12 hairs. All these hair clusters are produced by borr mutant cells (as judged by their trichome marker), so this phenotype is strictly cell-autonomous. The borr mutant clones do not significantly affect the planar polarity in the wing blade as mutant and surrounding wild-type hairs appear normally oriented (Fig. 8A).

Examination of borr mutant clones in pupal wing discs supports our notion that all late-induced borr mutant clones occupy roughly the same amount of territory as their twin spots, confirming that the mutant cells are fully viable at this stage (Fig. 8B-D). In support of this, we did not observe any nuclei with compacted DNA (that would indicate imminent apoptosis; see Fig. 6E). As in the larval discs, the surviving borr mutant cells in the pupal discs are much larger than their neighbours, often with giant nuclei(Fig. 8B, arrows), indicating a high degree of ploidy. These giant borr mutant cells appear healthy and are well integrated within the epithelial tissue(Fig. 8B-D). Their large size provides an explanation for the observed adult phenotype, and are consistent with a single borr mutant cell producing multiple hairs: other conditions that produce large cells – for example, cdc2, UltAor UltB mutant clones, or wounding – result in similar cell-autonomous clusters of trichomes, albeit in some cases with fewer hairs per cluster (Adler et al.,2000; Weigmann et al.,1997) (data not shown).

We also observe abnormal giant bristles in the wing margins of flies bearing late-induced borr mutant clones; these giant bristles invariably lack sockets (Fig. 8E). As we could not determine whether these abnormal bristles are derived from mutant cells (owing to the weak phenotype of their bristle marker), we visualised incipient bristles in the pupal wing by β-tubulin antibody staining. This revealed large borr mutant trichogen cells(identifiable by their lack of GFP) that generate bristles twice the normal size (Fig. 8F,G). In addition,unlike wild-type bristles, these giant bristles do not exhibit anyβ-tubulin accumulation at their bases(Fig. 8H), confirming that the developing socket is absent around the borr mutant bristles.

Bristles are part of sensory organs, which are composed of four cells– the trichogen (bristle-producing cell), tormogen (socket-producing cell), neuron and thecogen (sheath cell); these are the progeny of a single sensory organ precursor cell produced by consecutive invariant lineage divisions (Lai and Orgogozo,2004). Evidently, loss of borr compromises the lineage-generating divisions, and the single polyploid mutant cell seems to develop invariably as a trichogen at the expense of the tormogen and,possibly, of the other two sensory organ cells.

Four independent lines of evidence argue that Borr is the functional ortholog of vertebrate Borealin/Dasra. First, based on stringent database searches, we found that borr is the only gene in the Drosophila genome with significant sequence similarity to Borealin/Dasra, and vice versa. Second, like vertebrate Borealin/Dasra and other CPC components(Andrews et al., 2003; Carmena and Earnshaw, 2003; Gassmann et al., 2004; Sampath et al., 2004), Borr colocalises with endogenous Incenp and Aurora B in transfected mitotic Drosophila DmD8 cells. Third, like its vertebrate counterpart(Gassmann et al., 2004), borr is required for the correct subcellular localisation of Incenp and Aurora B in dividing cells. Fourth, Borr loss causes similar mutant phenotypes in mitotic Kc cells and in developing embryonic and larval cells as does depletion of other Drosophila CPC components in tissue culture,or depletion of Borealin/Dasra and other CPC components in mammalian cell lines. These phenotypes include abnormal spindles and uneven chromosome segregation, leading to giant multi-nucleate and/or polyploid cells and,usually, to apoptosis (see also below). A noticeable molecular consequence of Borr loss is also the reduction in the P-H3 levels – given that this phosphorylation event is mediated by Aurora B(Adams et al., 2001; Giet and Glover, 2001; Hsu et al., 2000), this links Borr function specifically to the activity of this CPC component. We note that the C. elegans protein CSC-1 appears to be another functional ortholog of Borealin/Dasra, despite showing very limited sequence similarity to these proteins, based on its mutant phenotypes in the embryo and on its functional interactions with other CPC components(Romano et al., 2003).

One striking mutant phenotype of mitotic borr mutant VNC cells is a significant reduction of their P-H3 levels(Fig. 2D; Fig. 4). Normally, this phosphorylation appears during prophase and spreads throughout the chromosomes, with peak levels during metaphase, followed by dephosphorylation during anaphase and telophase (Hans and Dimitrov, 2001; Nowak and Corces, 2004; Wei et al.,1998). Although the function of P-H3 is not known, correlations have been noted in many species between the P-H3 levels and the degree of DNA condensation, and a T. thermophila strain with a non-phosphorylatable version of H3 showed perturbed chromosome condensation and abnormal chromosome segregation (Wei et al.,1998). This led to the hypothesis that S10 phosphorylation of H3 may be necessary for chromosome condensation.

However, in borr mutant embryos, the condensation of the chromosomes in mitotic VNC cells is barely affected, yet their P-H3 staining is often strongly reduced (Fig. 2D; Fig. 4D,J). This argues that H3 phosphorylation occurs in parallel or subsequent to chromosome condensation, rather than driving it. Consistent with this, others have also reported a lack of correlation between chromosome condensation and P-H3 levels (Adams et al.,2001), including Yu et al. (Yu et al., 2004) who have observed normal levels of P-H3 on undercondensed chromosomes in greatwallmutants of Drosophila. Indeed, it has been suggested that P-H3 may be a sort of licensing factor, namely a mark placed on mitotic chromosomes to indicate their readiness to undergo separation during the subsequent stages of the cell cycle (Hans and Dimitrov,2001).

The striking reduction of the P-H3 levels in borr mutant embryonic cells, and in Borr-depleted cultured Drosophila cells(Eggert et al., 2004), is in contrast to the situation in HeLa cells in which RNAi-mediated depletion of Borealin did not affect their P-H3 levels(Gassmann et al., 2004). These authors suggested that, in these cells, H3 phosphorylation may be mediated by a Borealin-independent subcomplex of Aurora B and Incenp(Gassmann et al., 2004). More work is required to determine whether this apparent discrepancy between human and Drosophila Borealin function in mediating phosphorylation of H3 is genuine and cell type- or species-specific, or whether it is simply due to methodological differences in the analyses.

We have shown that borr is an essential gene in Drosophila, and that borr loss results in multiple successive defects during mitosis, including a reduction of P-H3, a severe attenuation prior to metaphase, multipolar spindles and uneven chromosome segregation. These defects may all reflect a function of Borr in the attachment of kinetochores to the mitotic spindle, given that this process often fails in Borealin-depleted HeLa cells(Gassmann et al., 2004). However, it is also possible that they reflect additional underlying activities of the CPC during the progression of mitosis. However, all of these mitotic defects are probably due, ultimately, to the observed failure of other CPC components such as Aurora B to localise correctly to the mitotic spindle(Fig. 5).

Multifocal spindles as observed in borr mutant cells(Fig. 2D; Fig. 3J; Fig. 5D,I) are expected to cause aneuploidy, and may trigger checkpoint function. They should thus be cleared from the developing tissue by apoptosis. Our observation of apoptotic borr mutant cells in larval imaginal discs(Fig. 6E) provide direct support that cell death is often the ultimate consequence of borrloss at the cellular level. However, borr mutant cells can also clearly evade apoptosis, and can undergo several consecutive abnormal divisions, given that the surviving (and dying) borr mutant cells in imaginal disc epithelia are typically large, with giant nuclei and greatly increased ploidy. Consistent with this, mammalian cells lacking CPC function appear to be defective for their spindle attachment checkpoint and can thus escape apoptosis (Lens and Medema,2003; Yang et al.,2004). A similar defect in the checkpoint function of borr mutant epithelial cells would explain why these cells can survive multiple abnormal mitoses, instead of entering apoptosis in response to the uneven chromosome segregation of a single abnormal mitosis. However,the survival capacity of the mutant cells is clearly limited, and most of them die ultimately – except in late larval and pupal discs in which they survive, possibly because of the slowing down of mitotic activity and/or growth at these stages, which perhaps provides a more permissive environment for the abnormally dividing borr mutant cells.

We found that borr mutant epithelial cells can cause major non-autonomous disruptions of the patterning of adjacent wild-type cells. This is unusual as imaginal discs can tolerate considerable cell death without compromising the development of normal adult tissues (see Perez-Garijo et al., 2004; Ryoo et al., 2004). The reason for this appears to be that apoptotic imaginal disc cells activate transient bursts of extracellular signalling by Wg and Dpp, to induce compensatory cell divisions in their wild-type neighbours. However, if apoptosis is suppressed through inhibition of caspase activity (which creates `undead' cells)(Perez-Garijo et al., 2004; Ryoo et al., 2004), this produces more sustained signalling, which in turn causes gross pattern abnormalities in the resulting adult tissue. It thus appears that interfering with, or suspending, the apoptotic pathway leads to over-compensatory responses.

We propose that a similar situation arises in the case of borrmutant imaginal disc cells: given that these can survive multiple abnormal divisions, they may be doomed – i.e. on a suspended apoptosis path– for an extended period of time and thus mimic some characteristics of`undead' cells. Like the latter(Perez-Garijo et al., 2004; Ryoo et al., 2004), doomed borr mutant cells may induce a burst of compensatory responses in their neighbours by stimulating the expression of extracellular signals such as Wg. This is suggested by our analysis of larval discs bearing early-induced mutant clones (Fig. 7), which revealed examples of overexpressed Wg in giant borr mutant cells, and also lateral expansion of Wg in twin spot areas whose associated borrmutant cells have died. Doomed borr mutant cells may also affect signalling by other pathways, e.g. the Notch pathway, given some of the borr mutant phenotypes (Fig. 7H-J; see Fig. S1 in the supplementary material) (e.g. Neumann and Cohen, 1996), but we have not examined this directly.

Our analysis further suggests that cell rearrangements can take place as a result of dying, or dead, borr mutant cells(Fig. 7L). These could be a consequence of compensatory signalling, and they may be aimed at repairing the substantial gaps in epithelial integrity expected to arise after the death of a giant borr mutant cell.

We have shown that borr loss also affects the lineage divisions of the external sensory organs: our evidence from late-induced borrmutant clones indicates that surviving giant borr mutant cells develop large bristles without sockets(Fig. 8). This phenotype suggests a defect or block in the division of the pIIa precursor cell that normally gives rise to the trichogen and tormogen(Lai and Orgogozo, 2004). It is less likely that the division of pI (the initial sensory organ precursor cell) is blocked by borr loss in these instances, as evidence from the analysis of embryonic sensory organs suggests that blockage of the first lineage division should result in the precursor cell adopting a neural fate(Hartenstein and Posakony,1990).

Why a borr mutant cell should adopt the bristle fate at the expense of the socket fate is not immediately obvious. One possibility is that the determining factor is its increased DNA content and large size. Notably,the trichogen cells that produce the stout bristles of the wing margin undergo at least one round of endoreplication during their differentiation(Hartenstein and Posakony,1989; Hartenstein and Posakony, 1990) (though in other external sensory organs the tormogen does as well) (Lai and Orgogozo,2004). Thus, borr loss could mimic an aspect of normal trichogen development, and could actively promote the acquisition of the bristle fate. It is thus conceivable that endoreplication is instructive during the process of sensory organ development.

We thank Mar Carmena, Bill Earnshaw, David Glover and Sarah Bray for providing antibodies; Konrad Basler for fly strains; and Fiona Townsley,Antonio Baonza and Monica Bettencourt-Diaz for technical advice. K.K.H. is supported by a MRC Laboratory of Molecular Biology/Cambridge Overseas Trust PhD studentship.