The Drosophila phosphatidylinositol transfer protein encoded by vibrator is essential to maintain cleavage-furrow ingression in cytokinesis

The process of cytokinesis is the final step of cell division that requires first the constriction of the cell in the equatorial region of the spindle and then abscission of the two daughters. Considerable attention has been focused upon both the establishment of the contractile ring and its constriction. The initiation of constriction leading to formation of the cleavage furrow is mediated by a signalling process that requires the so-called centralspindlin complex. This complex is minimally composed of a conserved kinesin-like motor protein (Pavarotti-KLP, Zen4, MKLP) and a GTPase-activating protein (RacGAP) [Tumbleweed (RacGAP50C) (Goldstein et al., 2005; Zavortink et al., 2005), Cyk4, MgcRacGAP] in Drosophila, C. elegans and human, respectively (D'Avino et al., 2005; Mishima and Glotzer, 2003). The microtubule cytoskeleton participates in this process not only in the delivery of the signal but also in ensuring the formation and maintenance of the contractile structures. This is achieved through a mutually dependent interaction of the central spindle, an overlapping array of spindle microtubules that forms in late anaphase, with the contractile ring itself (Gatti et al., 2000). Thus mutants that disrupt the central spindle such as klp3A, orbit and fascetto (Inoue et al., 2004; Verni et al., 2004; Williams et al., 1995) lead to the collapse of the contractile ring, and mutants that disrupt the contractile ring such as spaghetti squash, diaphanous and chickadee lead to central spindle defects (Giansanti et al., 1998; Somma et al., 2002).

Recently, several studies have emphasised a role for membrane trafficking in the late stages of cytokinesis (Albertson et al., 2005; Strickland and Burgess, 2004). Membrane-fusion-inducing SNARE components, syntaxin-2 and endobrevin/VAMP8, are required during cleavage in mammalian cells (Low et al., 2003). The exocyst, a multiprotein complex that targets secretory vesicles to distinct sites on the plasma membrane, is also involved in cell cleavage in yeast (Dobbelaere and Barral, 2004; VerPlank and Li, 2005), in Drosophila (Echard et al., 2004), and in mammalian cells (Skop et al., 2004). Furthermore, in mammalian cells SNARE complexes and the exocyst appear to interact with the centrosomal component centriolin to facilitate abscission (Gromley et al., 2003; Gromley et al., 2005).

Phosphatidylinositides play roles in both vesicle trafficking and in regulating actin dynamics, and so have the potential to play key roles linking these processes in cytokinesis. Phosphatidylinositol (PtdIns) is the core lipid that can be phosphorylated at single or several positions to give seven different phosphorylated forms. Of these, PtdIns(4,5)P₂ is of particular importance for membrane trafficking and also as the source of the two second messengers, inositol-1,4,5-trisphosphate (IP3) and diacylglycerol (DAG) (reviewed by De Matteis and Godi, 2004). A requirement for multiple steps of the PtdIns-cycling pathway has been described for cytokinesis of crane fly spermatocytes in a study using chemical inhibitors (Saul et al., 2004). Moreover, genetic studies have implicated the two kinases that successively phosphorylate PtdIns: in Drosophila the phosphatidylinositol 4-kinase (PI-4 kinase) four wheel drive (fwd) (Brill et al., 2000), and in S. pombe the phosphatidylinositol-4-phosphate 5-kinase [PI(4)P-5 kinase] (Cullen et al., 2000; Zhang et al., 2000), to produce PtdIns(4,5)P₂ as being required for cytokinesis. Three recent studies have confirmed this in cultured mammalian cells, in Chinese hamster ovary (CHO)-K1 fibroblasts and in Drosophila primary spermatocytes (Emoto et al., 2005; Field et al., 2005b; Wong et al., 2005). Each of these groups showed that PtdIns(4,5)P₂ localises to the cleavage furrow - as assessed by its specific ability to bind and localise the pleckstrin homology (PH) domain of phospholipase C (PLC) tagged with EGFP - although it was also found to be more widely distributed in the plasma membrane of fly cells. Consistently cytokinesis defects were seen after over-expression of PH-domain proteins; the PtdIns(4,5)P₂ phosphatases synaptojanin or SigD, or a dominant-negative PI(4)P-5 kinase. The study of Drosophila spermatocytes also showed that PtdIns(4,5)P₂ must be hydolysed by PLC for the furrow to remain stably connected to the contractile ring.

By contrast, little is known how ordered addition of new lipids into membranes is achieved during cytokinesis. The phosphatidylinositol transfer proteins (PITPs) were originally described for their properties of binding and delivering either PtdIns or phosphatidylcholine to lipid-deficient membranes but have now been involved in a variety of aspects of PtdIns biology (reviewed by Cockcroft, 2001). In addition to stimulating the activity of several of the PtdIns biosynthetic kinases, PITPs have also been proposed to promote PLC activity and might thus be required for PtdIns(4,5)P₂ hydrolysis as well as its synthesis. They are also involved in the trafficking of secretory vesicles from the Golgi (Ohashi et al., 1995). Here, we show that PITP is required for maximal contraction of the actin ring, for progression of the cleavage furrow and to maintain attachment of the furrow to the underlying cytoskeleton.

We originally isolated a vibrator allele in a screen of a collection of third chromosome P-element insertion mutants (Deak et al., 1997) for mutants that showed defects in cytokinesis, in mitotic larval neuroblasts or in male meiosis (Fig. 1). Meiosis in wild-type males produces cysts of 64 spermatids, each containing a nucleus and mitochondrial derivative (Fig. 1A, white and dark spheres, respectively), the nebenkern. By contrast, individual spermatids of vibrator males have multiple nuclei associated with an enlarged nebenkern that is characteristic of cytokinesis defects (Fig. 1B). These multinucleate spermatids undergo elongation, in which multiple nuclei can be seen associated with a bundle of axonemal fibres (Fig. 1C).

Rescue of plasmid containing P-element DNA from the mutant and sequencing of the flanking DNA (Deak et al., 1997) identified the insertion to be in the vibrator gene. vibrator was previously described by Spana and Perrimon (Spana and Perrimon, 1999) as the Drosophila PITP, and was suggested to have a number of roles throughout development in actin-based and signal-transduction processes. The encoded protein is the single Drosophila counterpart of the vertebrate PITPα and PITP β, and was named after the mouse PITPα mutant vibrator (Hsuan and Cockcroft, 2001). Drosophila vibrator maps to 91F10-11 on chromosome 3R and is uncovered by the deficiency Df(3R)Dl-BX12. In addition to our original allele, five other transposon insertions were subsequently identified (Table 1; Fig. 2A). We found that some of the chromosomes carrying these insertions into vibrator have additional mutations at other sites (see legend to Fig. 2). Thus, we have carried out most of our studies on hemizygous mutants in which the chromosome carrying the transposon insertion mutant was placed over the deficiency chromosome. This also allowed us to rank the mutant alleles in an allelic series depending upon the lethal stage when hemizygous (Table 1). The strongest allele, vib^j7A3 is a second-instar larval lethal, vib^S110416 an early pupal lethal, vib^j5A6 and vib^EP513 are pharate adult lethals, and the weakest allele to show a mutant phenotype, vib^S045002, is viable, female fertile but male sterile with non-motile sperm (not shown). A final P-insertion vib^EP651, positioned outside of the transcribed region and oriented in a way that should promote expression of vibrator (Bidet et al., 2003), was phenotypically normal (Fig. 2A).

To relate the lethal stage to expression levels, we raised an antibody against the full-length vibrator protein expressed in E. coli (Materials and Methods) (Fig. 2B). This antibody recognised a protein of 35 kDa that was barely detectable in the strongest allele vib^j7A3 when either hemizygous or homozygous. When other members of the allelic series were placed against vib^S110416, the amount of vibrator protein detected was broadly consistent with the allelic strength assessed by the lethal phase of the mutation.

We then compared the extent of defective cytokinesis in two representative alleles. vib^S110416, the strongest allele, permitting examination of both abnormal mitosis and male meiosis, and vib^S045002, the weakest allele, for which a phenotype could be detected. Phase-contrast imaging of onion-stage spermatids showed that 99.9% of wild type (Canton S) and 100% of the deficiency stock heterozygous with a balancer chromosome (Df(3R)Dl-BX12/TM6B) had the expected 1:1 ratio of nuclei to nebenkern (Fig. 1A; Table 2). By contrast, only 19.2% and 68.3% of spermatids in vib^S110416 / Df(3R)Dl-BX12 and vib^S045002 / Df(3R)Dl-BX12, respectively, had this 1:1 ratio (Table 2). Unlike wild-type spermatids, the remaining mutant cells showed multiple nuclei in proportion to the strength of the allelic combination. We also analysed larvae of these same genotypes for defects in cytokinesis in the developing central nervous system. Although we were able to detect polyploid cells in vib^S110416 - whether homozygous or hemizygous - these were at a much lower frequency than in mutant testes (Table 3). Furthermore, we were unable to detect any polyploid cells in vib^S045002 larval brains. Consistently, we observed a dramatic reduction in the level of vibrator protein in extracts of testes from vib^S110416 / Df(3R)Dl-BX12 animals in comparison with wild type, but not in extracts of brains from homozygous vib^S110416 animals (compare Fig. 2C and D). We conclude that a reduction in the levels of vibrator protein results in cytokinesis defects. The persistence of protein in the larval brain could represent perdurance of maternally provided protein, as has been described for other cell-cycle gene products in Drosophila (Carmena et al., 1991). The more extensive depletion of protein seen in the testes reflects most probably the extensive requirement for membrane biosynthesis in the male germ line that has to be built from a small number of germ cells laid down during embryogenesis.

We then assessed the sub-cellular localisation of the Vibrator PITP throughout the meiotic cycles. In the late part of the extended G2 preceding meiosis, we found Vibrator PITP appeared to accumulate in regions known to be membrane rich, including nuclear and plasma membranes (Fig. 3A). An intense concentration of Vibrator protein appeared to be associated with central spindle microtubules at anaphase (Fig. 3B). This particular distribution suggests an association of PITP with the parafusorial and mitochondrial membranes. Vibrator PITP then appeared to concentrate within the mitochondrial aggregate of the nebenkern in spermatids (Fig. 3C) and associated with sperm tails in a manner consistent with its continued association with membranes (not shown). Thus, the distribution of Vibrator PITP suggested it to be present in a variety of membranous structures in the cell. This is very similar to the reported localisation of PtdIns(4,5)P₂ in the plasma membrane and cleavage furrows (Wong et al., 2005) and reflects the large numbers of membranous structures that contain phosphoinositol lipids in dividing Drosophila spermatocytes.

The presence of onion-stage cells with more than the expected 1:1 ratio of nuclei to nebenkern indicated that vibrator mutants are defective in cytokinesis in one or both meiotic divisions. In Drosophila, the central spindle is fundamental for successful cytokinesis. In wild-type primary spermatocytes, the central spindle is formed after the release of microtubules from the centrosomes and their subsequent bundling (Inoue et al., 2004) in a process that requires Klp3A (Williams et al., 1995), Fascetto (Verni et al., 2004), the Pavarotti KLP of the centralspindlin complex (Adams et al., 1998) and a number of other microtubule-associated proteins (Gatt et al., 2005; Inoue et al., 2004). Rings containing anillin, the septin peanut, and actin encircle the central spindle microtubules (see Fig. 4B,C and Fig. 5A, respectively). Contraction of the acto-myosin filaments leads to ingression of the furrow and compaction of the central spindle to the tight microtubule structure observed in cleaving cells. In vibrator primary spermatocytes the central spindle region appeared to be properly organised in most cells, suggesting there is no problem with its formation. Pavarotti was similarly associated as a tight band on such spindles like in wild-type (Fig. 4A). However, within some cells the central spindle was not compact but broad or displaced (Fig. 4A yellow arrowhead), possibly suggesting that the ring had not fully constricted or had constricted and then partially relaxed. In these cells, Pavarotti was present in a wider, more discontinuous band. We made similar observations in secondary spermatocytes, although here, the majority of cells contained two meiosis-II spindles. Nevertheless, these spindles joined to give a predominantly bipolar structure with a single contractile ring of varying degrees of integrity (Fig. 4D). We interpret the relatively high proportion of spindles with a broad central region as an indication that progression through this stage of meiosis was being delayed as a consequence of the mutation. Alternatively, the central spindle might fail to form correctly or might become unstable late in cytokinesis.

We next examined the localisation of anillin, a protein known to physically interact with several cleavage-furrow components including F-actin (Oegema et al., 2000) and myosin II (Straight et al., 2005). Anillin also contains a PH domain in its C-terminus, suggesting an ability to interact with membranes through inositol lipids (Field et al., 2005a). In wild-type primary spermatocytes, anillin localises to the cleavage furrow from mid-anaphase through to telophase and is maintained in the ring canals, the persistent circular products of cytokinesis that connect cells within the cyst (Fig. 4B, inset, arrowhead) (Giansanti et al., 1999). In vibrator primary spermatocytes, many anillin rings appeared normal but were actually broader or thinner (Fig. 4B). The ring canal resulting from the first meiotic division often appeared larger than that of wild type, and could be included within the cell rather than attached to the cell membrane (Fig. 4E, arrow). Anillin has been shown to be required for the ingression of furrow canals at normal rates in embryos (Field et al., 2005a). Thus, its rather discontinuous distribution in vibrator cells could reflect defects in its interaction with phospholipids.

We found peanut, a Drosophila septin (Fig. 4C, inset), to be localised in the contractile rings of wild-type spermatocytes (Hime et al., 1996) but in ring canals it was less easily detected; also reported previously by Carmena et al. (Carmena et al., 1998). In vibrator primary spermatocytes, peanut was conspicuous in cells having a well-formed central spindle, but not easily seen in those cells with diminished central spindle microtubules (arrowhead Fig. 4C). Strong bands of peanut staining were also seen in secondary spermatocytes with well-formed central spindles often encircling two joined spindles (Fig. 4F). Such bands could be discontinuous in cells with disorganised central spindles (Fig. 4F, right arrow).

Actin showed significant differences in its localisation in wild-type and vibrator mutant spermatocytes at cytokinesis. As described above for the other ring components, actin forms a dense band, constricting the central spindle microtubules at telophase in meiosis I (Fig. 5A), in wild-type cells. In vibrator primary spermatocytes of a similar stage, the actin ring was broader and, again, discontinuous (Fig. 5B). Whereas at the onset of meiosis II, wild-type secondary spermatocytes have tightly constricted contractile rings associated with the plasma membrane (Fig. 5C), in the binucleate vibrator secondary spermatocytes, the contractile rings appeared disorganised (Fig. 5D) and similar to those reported in spermatocytes of the fwd mutant (Brill et al., 2000). Actin polymerisation has been reported to be regulated in multiple steps by phospholipids (see Discussion). Thus, our observations indicate that abnormal phospholipid metabolism in the vibrator mutant could affect actin polymerisation and its sub-cellular distribution.

Since it was difficult to assess from studies of fixed preparations how the observed defects in the organisation of the central spindle and contractile ring might arise, we turned to studies of live spermatocytes. We followed the behaviour of fluorescently labelled microtubules as a result of a GFP-tagged β-tubulin transgene introduced into wild-type and vibrator mutant flies. We could simultaneously follow the major membrane features of the cell revealed by DIC microscopy. Although the dynamics of microtubules have been well studied in relation to cytokinesis in spermatocytes of living Drosophila (Gatt et al., 2005; Inoue et al., 2004), their relationship to the dynamics of the cleavage furrow are less well documented. We, therefore, first examined the timing of furrow formation in wild-type spermatocytes and found it was initiated 9.8±0.5 minutes after onset of anaphase (n=6). The furrow then ingressed at a rate of 1.9±0.15 μm/minute (n=9) and proceeded until the central spindle microtubules had been compacted to a diameter of 2.3±0.05 μm (n=8) by the contractile ring (Table 4; Fig. 6, 19 minutes). The end point of constriction in the cytokinesis of a spermatocyte is formation of the ring canal, the stabilised residue of the contractile ring that does not close, leading to a persistent cytoplasmic connection between cells. The ring canal can be seen by DIC optics towards the end of furrow ingression (Fig. 6, arrow, 19 minutes; and supplementary material Movie 1).

We saw no variation between wild-type and vibrator spermatocytes in the formation and initial compaction of the central spindle at anaphase, revealed by fluorescence of GFP-tagged tubulin. When cytokinesis was successful in time-lapse imaged vibrator mutant cells (3/7; a frequency comparable to fixed preparations, Table 2), the cleavage furrow formed at a similar time to that observed in wild-type cells (Table 4). When cytokinesis was unsuccessful in the mutant, furrow formation was delayed by approximately 4 minutes (Table 4). The dynamics of maturation of the late telophase spindle and of the progression of cleavage were dramatically different between vibrator and wild-type cells. Furrow ingression occurred more slowly in the mutant [1.4±0.07 μm/minute (n=7) in vibrator compared with wild type 1.9±0.15 μm/minute (n=9)]. Even in vibrator cells with successful cytokinesis, the time from furrow initiation to completion of cleavage was twice that of wild-type (Table 4).

In those vibrator spermatocytes that cleaved successfully, the central spindle became compacted to a similar degree as observed in wild-type cells [Table 4, compare 2.4±0.06 μm (n=3) in vibrator mutant spermatocytes with 2.3±0.05 μm (n=8) in wild-type cells]. By contrast, in those vibrator mutant spermatocytes that failed to divide, the central spindle did not compact beyond 4.5±1.0 μm (n=4), about twice the diameter of wild type (Fig. 7 and Fig. 8B,C). This suggests that the actin ring was no longer able to contract further and compressed the central spindle microtubules only to this point, accounting for the enlarged rings of actin observed in fixed preparations of vibrator spermatocytes. Cytokinesis then appeared to progress no further for a period of time (Fig. 8C; 18.5±6.8 minutes, n=4). We consistently observed that the furrow, which could ingress no further, was no longer stably maintained and regressed rapidly (Figs 7, 8). The rapidity with which the furrow snapped back indicates that the membrane is highly elastic and suggests that its bond with the underlying cytoskeletal structure is suddenly lost (see supplementary material, Movie 2).

Here, we have documented a sequential series of defects as vibrator mutant spermatocytes progress through the two meiotic divisions in spermatogenesis. Studies with fixed and living cells indicate that, in the strongest allelic combination studied, approximately 70% of cells fail in cytokinesis. In these cases, the central spindles and associated rings appear to assemble normally but furrow formation is delayed. There is also a reduced rate of furrow ingression, and compression of the central spindle microtubules becomes stalled before its completion. In cells that fail cytokinesis, the membranes of the furrow become abruptly disconnected from the underlying cleavage ring and central spindle. The timing of these events has not previously been determined in wild type, or when cytokinesis is affected by other mutations or drug treatments that affect phospholipid metabolism. Nevertheless, aspects of these phenotypes appear qualitatively similar to those described following loss of function in several steps in polyphosphoinositide synthesis. Reduced levels of PI-4 kinase in fwd mutants permit initiation of furrow formation, however, this is followed by subsequent furrow regression (Brill et al., 2000). PI-4 kinase is required for synthesis of PtdIns(4)P that in turn is converted to PtdIns(4,5)P₂ by PI(4)P-5 kinase. The hydrolysis of PtdIns(4,5)P₂ to inositol (1,4,5)-triphosphate [Ins(1,4,5)P₃] by PLC is similarly required to maintain the furrow (Emoto et al., 2005; Field et al., 2005b; Wong et al., 2005). Together, these findings are consistent with inhibitor studies in crane fly spermatocytes that suggest the requirement of continuous PtdIns metabolism for cytokinesis (Saul et al., 2004). A precise assessment of comparative functions of different phosphorylated forms of PtdIns in cytokinesis will require a more detailed analysis of furrow behaviour following the downregulation of other steps in the pathway, by using comparable technical approaches. We noticed that another PITP, Nir2, the human homologue of the fly protein retinal degeneration B (RdgB) has also been described as being required for cytokinesis (Litvak et al., 2002). It will be of future interest to determine whether Drosophila RdgB is also involved in cytokinesis and to what extent it may be redundant with the Vibrator protein.

The cytokinesis defects we observed appear to have arisen from both a failure of the central spindle microtubules to become fully compressed, and from defective connections between the plasma membrane and underlying cytoskeletal structures. The former could be explained by an arrest of further constriction of the contractile ring, such that the central spindle does not become compacted beyond this point. This is also evident from the broader, less defined appearance of contractile rings revealed by immunostaining at late stages of cytokinesis. Failure of the ring to fully constrict is, in turn, also reflected in the abnormal ring canals formed in these cells. The discontinuity of these rings might reflect abnormalities of actin polymerisation brought about by a variety of proteins that have been proposed to be regulated by phospholipids, principally PtdIns(4,5)P₂. These include profilin, which promotes actin assembly, the actin-related Arp2-Arp3 complex and WASP family proteins, and the actin-severing protein cofilin. Each of those proteins have been shown to have a role in cytokinesis (Giansanti et al., 1998; Gunsalus et al., 1995; Pelham and Chang, 2002; Withee et al., 2004), and all have been proposed to be regulated by binding to PtdIns(4,5)P₂ (reviewed in Yin and Janmey, 2003). Consistently, of all components of the contractile ring, actin is the most affected in vibrator mutants; actin rings appear discontinuous and in many cells are not maintained late in cytokinesis. Actin undergoes re-distribution at cytokinesis. It has to be depolymerised at the cell poles, following which, polymerised actin is recruited to the region of the furrow (Cao and Wang, 1990; Fishkind and Wang, 1993). Since this is a highly dynamic process it will be of future interest to follow actin dynamics in time-lapse studies of these mutants, using approaches analogous to those we describe here.

The sudden elastic regression of the furrow in vibrator mutant cells suggests that the plasma membrane remains bonded with the underlying cytoskeleton until a point at which it is catastrophically overcome. There are at least two possible explanations of this observation. First, that Vibrator PITP is required for vesicle transport and insertion of new lipid into the plasma membrane at the cleavage furrow. A failure to do this could lead to abnormally high tension at the leading edge of the furrow and its sudden regression. Alternatively, these observations could point to a function for Vibrator PITP in regulating the dynamics of structures at the interface between the plasma membrane and the cytoskeleton. One important role played by the phospholipids, particularly PtdIns(4,5)P₂, is to mediate cohesion between the plasma membrane and the underlying cytoskeleton (Raucher et al., 2000). This could occur directly through interactions between PtdIns(4,5)P₂ and cytoskeletal anchoring proteins, such as spectrin, α-actinin and other cortical proteins. Alternatively, anillin has been proposed to interact directly with the lipids in the plasma membrane through its PH motif (Field et al., 2005a). Potential defects in the synthesis of PtdIns(4,5)P₂ resulting from the loss of Vibrator PITP could thus become critical at this crucial point of contraction of the ring leading to the sudden disconnection of the membrane.

It also appeared that parafusorial membranes were not fully dispersed as a consequence of reduced Vibrator PITP (Fig. 7). The sub-cellular localisation of Vibrator PITP suggests that it is to be found within a variety of membranous structures in addition to the plasma membrane (Fig. 3). This is perhaps not surprising given the well-documented involvement of phosphoinositides in multiple stages of membrane trafficking (reviewed by De Matteis and Godi, 2004). Indeed, some of the machinery for trafficking may well be `hijacked' for a role in cytokinesis. This idea is supported by the requirement for several components of the exocyst for the late stages of cytokinesis (see Introduction). Indeed, PITP has been shown to be required for vesicle budding at the Golgi and for regulated exocytosis (reviewed by Cockcroft, 2001). The process of cell division requires that all intracellular membranes and organelles are equitably divided between daughter cells. The apparent failure of intracellular membranes to become dispersed in vibrator mutant cells, like it normally occurs in wild-type cells, points towards this process being regulated through phospholipids. It will be of future interest to examine this potential function of phospholipids during cytokinesis in greater detail.

Six P-element-generated mutant lines were used in this study: vib^j7A3, vib^j5A6, vib^EP513 (Rorth, 1996), vib^EP651 (Rorth, 1996), vib^S110416 and vib^S045002. The deficiency Df(3R)Dl-BX12 uncovered the gene vibrator. Flies were maintained using standard culture methods.

For phase-contrast imaging of onion-stage cysts, testes were dissected in testis buffer (183 mM KCl, 47 mM NaCl, 10 mM Tris-HCl pH 6.8, 1 mM EDTA) (Gonzalez and Glover, 1993) and gently squashed under an 18×18 mm coverslip until the appropriate degree of flattening was attained. Specimens were screened for intact cysts of primary spermatocytes using phase-contrast on a Nikon Microphot-FX microscope at low magnification (25×), and the morphology and number of cells in those cysts were analysed. Images were acquired with a Spot RT camera (Diagnostic Instruments) running the included software package on a PC.

Brain squashes were carried out according to Gonzalez and Glover (Gonzalez and Glover, 1993). Briefly, brains were dissected in PBS, incubated in 45% acetic acid for 30 seconds, transferred to 60% acetic acid for 3 minutes, covered with an 18×18mm siliconized coverslip and squashed for 1 minute between two sheets of blotting paper using mechanical force. Slides were frozen in liquid nitrogen, coverslips removed and the slide immediately immersed in PBS for 5 minutes, rinsed in water and left to air-dry. Squashed preparations were mounted in Vectashield with DAPI (Vector Laboratories, Inc. H-1200) and sealed with nail polish.

To confirm that the phenotype we observed was due to mutations in the gene vibrator (vib), we generated two types of rescue construct: a vib-cDNA driven by the ubiquitin promoter (UB) and a genomic rescue construct.

To generate the vib-cDNA driven by the ubiquitin promoter, vibrator cDNA was amplified and cloned into pCasPeR4 (UB) as follows: a full-length vibrator cDNA (SD01527) was used as a template for PCR that introduced artificial restriction sites and a 5′-ribosome-binding site. The following primers were used: vib_cDNA3, 5′-GTC TAG AAC AGC CAC CGG CAA AGA TGC AGA TCA AAG-3′ and vib_cDNA5, 5′-TGC TAG CTT AAT CGG CAT CCG CGC GCA TAC C-3′.

The forward primer vib_cDNA3 included an XbaI site and the β-1-globin ribosome-binding site (ACAGCCACC) separated from the start codon by seven base pairs. The reverse primer vib_cDNA5 contained the stop codon and an NheI site. The fragment was inserted non-directionally into the XbaI site of pCasPeR4 (UB) and colonies were screened to identify those that could be driven by the UB promoter. The total size of this construct was approximately 11 Kb.

To generate the genomic rescue construct in pCasPeR4, a genomic fragment was generated from PCR using wild-type DNA and cloned non-directionally into the KpnI site of pCasPeR4. The following primers were used: vib_genomic1, 5′-AGG TAC CGC TAT AGC AGA AGA GTG CGG-3′ and vib_genomic3, 5′-TGG TAC CCC AAC CAG AAT CGA TCC GTG-3′.

These primers generated a genomic fragment of 10950 bp. This included 1779 bp upstream the start codon (including some of the ORF of CG11703) and 1033 bp downstream the 3′ UTR. The complete rescue construct was 18.8 Kb.

Transformation of both constructs was achieved following injection into w¹¹¹⁸ flies according to standard techniques. Transformed recombinant lines were selected based on eye colour. Both rescue constructs rescued the lethality and fertility of vib^S110416 / Df(3R)Dl-BX12.

A full-length vibrator cDNA (SD01527) was amplified by PCR and cloned into pET23b to express a C-terminally His-tagged recombinant protein. The resultant recombinant protein was expressed in E. coli and purified on a Ni⁺ column under denaturing conditions. Two rabbits were immunised (AbCam Ltd) and serum of both (dilution 1:1000) recognised a single band with a molecular mass of 35 kDa on western blots of Drosophila extracts. Western blotting was carried out according to standard procedure. Actin (1:2000, Sigma A2066) and γ-tubulin (1:5000, Sigma T6557) were used as loading controls for western blotting.

Testes, from third instar larvae or pAdult males, were prepared for immunostaining using standard methods (methanol-acetone fixation) (Cenci et al., 1994). The following antibodies were used in this study: anti-tyrosinated α-tubulin (1:10, YL1/2, Harlan Sera-Labs); anti-pavarotti (1:750, GM2, this laboratory); anti-anillin (1:1000, a kind gift of C. Field) (Field and Alberts, 1995); mouse monoclonal anti-peanut (1:4, 4C9H4, Developmental Studies Hybridoma Bank maintained by The University of Iowa, Department of Biological Sciences, Iowa City, IA) (Neufeld and Rubin, 1994); anti-vibrator (1:100, this study). Toto-3-iodide (T-3604, Molecular Probes) was used to counterstain DNA. To visualize the distribution of actin, testes were fixed with 3.7% formaldehyde (Gunsalus et al., 1995) and probed with Rhodamine-labelled phalloidin (Molecular Probes). To visualize the distribution of vibrator protein in spermatocytes, testes were fixed with 3.7% formaldehyde using the following protocol, which omits Triton X-100 from all buffers to preserve membranous structures. Testes were dissected and rinsed briefly in testes buffer before being transferred to a 0.75 μl drop of the same buffer on a Poly-Prep™ slide (Sigma P0425). The testes were opened with forceps to release the cysts of spermatocytes. As soon as the buffer evaporated, 10 μl of 3.7% formaldehyde in testes buffer was gently pipetted onto the released spermatocytes and left for 10 minutes. Slides were washed twice for 5 minutes in PBS; cells were permeabilized in 0.5% saponin (Sigma S-4521) in PBS for 30 minutes and then blocked in PBS plus 1% BSA for 1 hour. The spermatocytes were incubated overnight with the vibrator antibody (diluted 1:100 in PBS plus 1% BSA) at 4°C, washed three times in PBS plus 1% BSA, incubated with the secondary antibody for 2 hours at room temperature, washed twice in PBS plus 1% BSA then once in PBS. DNA was counterstained with Toto-3-iodide. Unless otherwise stated, secondary antibodies were obtained from Jackson Immunochemicals and used according to the supplier's instructions.

Images were acquired on a Nikon Microphot microscope fitted with a MRC1024 scanning confocal head (Biorad) using a ×63 NA1.4 objective lens. Figures shown are the maximum-intensity projection of optical sections acquired at 0.5-1 μm steps.

Live cell imaging was carried out according to the method described by Inoue et al. (Inoue et al., 2004). In brief, testes, isolated from third instar larvae, were dissected under 10S Voltalef oil (Elf Atochem) onto coverslips (No. 1 1/2) attached to an open chamber. Near simultaneous images of both DIC and fluorescence were obtained, each cell was sectioned six times with a 1 μm z-step and images were captured at 1-minute intervals. Time-lapse images were obtained on a Zeiss Axiovert 200 (Carl Zeiss Microimaging) microscope fitted with a 100× (N.A. 1.4) differential interference contrast (DIC)-lens and -condenser (N.A. 0.55) using appropriate filters and were acquired with a CoolSnap HQ camera (Roper Scientific). Metamorph (Universal Imaging) was used for the analysis of images. The fluorescent images shown were generated from the maximum-intensity projection of all six sections whereas the DIC image is from a single z-section.

The ingressing furrow was measured in the central-most DIC section. This distance was plotted in Excel (Microsoft) and the average rate of furrow ingression determined from the slope of the line of best fit from eight or more continuous points from the steepest part of the graph.

While this manuscript was under review, another group also reported a role in cytokinesis for a gene termed giotto, that corresponds to vibrator (Giansanti et al., 2006).

This work was supported by the Medical Research Council and Cancer Research UK. We would like to thank Monica Bettencourt-Dias, Paolo D'Avino, Matthew Savoian and Tetsuya Takeda for advice and comments on the manuscript, and Ed Jones, a part II student, for his involvement during the early stage of this project. We also wish to thank Chris Field for anti-Anillin antibodies, the Bloomington and Szeged Stock Centres for fly stocks.