Centrosome migration into the Drosophila oocyte is independent of BicD and egl, and of the organisation of the microtubule cytoskeleton

At the anterior tip of the Drosophila ovary lies the germarium, a structure where new 16-cell cysts are formed after four divisions of a cystoblast. These 16 cystocytes are connected through intercellular bridges called ring canals. Two cells of the cyst have four canals and are known as the pro-oocytes. Initially, both pro-oocytes behave similarly in several respects; as the cyst matures, however, one of them becomes the oocyte, whereas the other pro-oocyte and the rest of the cystocytes adopt a nurse cell fate (Spradling, 1993). The selection of the oocyte fate can be followed with at least three different classes of markers. The first class is the formation of the synaptonemal complex (SC), a tripartite structure that holds the homologous chromosomes together during meiosis. The SC initially forms in the nuclei of the pro-oocytes, and is later restricted to the oocyte (Carpenter, 1975; Huynh and St Johnston, 2000). The second class are the centrioles of the 15 nurse cells, which migrate through the ring canals into the oocyte during cyst development in the germarium (Grieder et al., 2000; Mahowald and Strassheim, 1970). The third class are cytoplasmic markers, such as Bicaudal-D (BicD), Cup or Orb proteins, which initially accumulate in the two pro-oocytes but are later localised to the oocyte (González-Reyes and St Johnston, 1998; Keyes and Spradling, 1997; Lantz et al., 1994; Suter et al., 1989; Wharton and Struhl, 1989).

The mechanism by which one of the pro-oocytes is selected as the oocyte is not completely understood, but it does not seem to be a stochastic process in which both pro-oocytes have the same chance to become the oocyte. Rather, the cyst develops an asymmetry that labels one of the future pro-oocytes since the very first division of the cystoblast. This asymmetry is present in the fusome, a large cytoplasmic organelle that connects all the cystocytes of a cyst through the ring canals (de Cuevas and Spradling, 1998; McKearin, 1997; Telfer, 1975). Since it has been shown that the future oocyte contains more fusome material than its 15 sibling nurse cells, and since the polarisation of the fusome is the first visible asymmetry in the cyst, it is possible that the fusome provides the initial clue for choosing the oocyte (de Cuevas and Spradling, 1998; Grieder et al., 2000; Lin and Spradling, 1995). How the generation of an asymmetric fusome directs the selection of one oocyte remains to be unravelled. It has been recently proposed that the inherent polarity of the fusome is used to organise the microtubule (MT) cytoskeleton of developing cysts (Grieder et al., 2000). Since the correct polarisation of the MTs is necessary for choosing the oocyte (Koch and Spitzer, 1983; Theurkauf et al., 1993), the fusome may play a direct role in oocyte selection. However, this is probably not the only function of the fusome, as meiosis is restricted to one cell even in the presence of very high concentrations of colcemid (Huynh and St Johnston, 2000).

egalitarian (egl) and BicD mutants produce cysts with 16 nurse cells and no oocyte that are unable to organise their MTs into a polarised, stable lattice. Lack of egl and BicD also abolishes the restriction of meiosis to one cell and prevents the localisation of all known oocyte markers (Huynh and St Johnston, 2000; Mach, 1997; Ran et al., 1994; Schüpbach and Wieschaus, 1991; Theurkauf et al., 1993). However, egl and BicD mutants develop asymmetric fusomes. This finding demonstrates that the presence of a polarised fusome is not sufficient to select the oocyte and that this selection requires at least the activity of BicD and egl (de Cuevas and Spradling, 1998).

To understand the role of the fusome in the polarisation of the germline cyst, we have examined the behaviour of the centrosomes in wild-type cysts, and in three mutants that block oocyte selection. We have also analysed the role that MTs may have in centrosome migration. Our results reveal that the migration of the centrosomes into the prospective oocyte differs from the localisation of either cytoplasmic or nuclear markers for oocyte selection, leading us to propose the existence of a functional asymmetry in the cyst independent of the organisation of the MT cytoskeleton and of the activity of BicD and egl.

Mutant alleles used in this study are BicD^r5, egl^WU50, egl^RC12 and Dhc64C^greco. Dhc64C^greco is a new amorphic allele of Dhc64C. To generate hemizygous combinations we used Df(2R) bw, S46 for egl (Li et al., 1994; Mach, 1997; Ran et al., 1994; Schüpbach and Wieschaus, 1991).

The polyubiquitin-D-TACC-GFP stock is a gift from Jordan Raff (Gegerly et al., 2000). To express the tau-GFP fusion in the germline, we cloned an EcoRI-XbaI fragment from UASt-tau-GFP (Kaltschmit et al., 2000) into the UASp vector (Rorth, 1998). The nod-GFP DNA was kindly donated by Peter Kolodziej (HHMI, Vanderbilt University Medical Center, Nashville, Tennessee, USA). A NotI fragment containing the nod-GFP DNA was cloned in the UASp vector. To express UASp-nod-GFP and UASp-tau-GFP in the germ line we used the nanos-Gal4:VP16 driver (Van Doren et al., 1998).

Antibody and rhodamine-phalloidin stainings were performed according to standard procedures. Detailed protocols are available upon request. Antibodies were used at the following concentrations: rabbit anti-cnn (Li and Kaufman, 1996), 1/500; rabbit anti-Dhc64C (Li et al., 1994), 1/250; mouse anti-orb (Lantz et al., 1994) monoclonal antibodies 4H8 and 6H4 from the Iowa Hybridoma Center, 1/200 each; rat anti-cup (Keyes and Spradling, 1997), 1/2000; rabbit anti-spectrin (Byers et al., 1987), 1/100; mouse anti-α-tubulin (clone DM1A) and mouse anti-γ-tubulin (clone GTU-88), from Sigma, 1/500; rabbit anti-inscuteable (Huynh and St Johnston, 2000; Kraut et al., 1996) (to label the SC), 1/1000; rabbit anti-anillin (de Cuevas and Spradling, 1998; Field and Alberts, 1984), 1/2500. FITC-, Cy2-, Cy3- and Cy5-conjugated secondary antibodies (Jackson Laboratories) were used at a final concentration of 1/200. Images were collected using either Bio-Rad 1024 or Bio-Rad Radiance scanning Confocal microscopes. Images were assembled using Adobe Photoshop and labelled in Adobe Illustrator.

We believe that the staining shown by the anti-γ-tubulin antibody corresponds to centrosomes for the following reasons. Firstly, EM studies have shown that germline centrioles behave like the γ-tubulin positive dots (see Results) (Mahowald and Strassheim, 1970). Secondly, this antibody marks a γ-tubulin positive structure at each pole of the mitotic spindles of germline and follicle cells (data not shown). Thirdly, the antigen recognised by this antibody colocalises in region 1 with Cnn, another centrosomal protein, indicating that the GTU-88 antibody recognises centrosomes more than centrioles only, at least in region 1 (see Results).

Females were starved for 4 hours and then fed yeast mixed with a 200 μg/ml colcemid solution for 24 or 48 hours (Sigma cat. n° D-6165). The ovaries were then processed as for standard antibody staining using the anti-α-spectrin and anti-γ-tubulin antibodies.

Germline clones were made using the FLP/FRT technique (Chou, 1992). BicD^r5 and Dhc64C^greco clones were identified by the absence of nuclear GFP (gift from Stefan Luschnig, Tübingen). The FRTs used were FRT-40A and FRT-2A. Clones were induced by heat-shocking third instar larvae at 37°C for 2 hours, during 2 consecutive days. Adult flies were dissected up to 12 days after heat shock to avoid the protein perdurance factor.

Centrosomes are the primary MT organising centres in most cell types and are composed of two perpendicular centrioles surrounded by a cloud of electron-dense pericentriolar material (Stearns and Winey, 1997). In order to visualise the subcellular distribution of centrosomes in the germarium, we have used an antibody against the integral component of centrosomes, γ- tubulin (Joshi, 1994). In the germarium, this antibody recognises ‘dots’ in somatic follicle cells and in germline cells, and several lines of evidence suggest that the structures labelled by this antibody are indeed centrosomes (see Materials and Methods).

The centrosomes of germline cells are initially present in all cells of early regions 1 and 2a cysts, but most of them accumulate in the oocyte by late region 2b (Fig. 1A; see Fig. 3A for a scheme of germarial stages). This centrosome migration takes place along the fusome, as centrosomes remain associated with the fusome until region 2b (Fig. 1B-D). By late region 2b, most of the centrosomes are located inside the oocyte, although some can still be observed at the ends of the fusome (Fig. 1E). The association of the centrosomes with the fusome during their migration suggests that centrosome movement depends on fusome polarity. Centrosomes are often seen on the ‘distal’ side of the ring canals from the oocyte, indicating that the movement of centrosomes towards the oocyte is not uniform and that it is slowed down prior to their translocation across the canal (Fig. 1F). A more detailed description of centrosome migration in the germarium has been recently published (Grieder et al., 2000).

Once in the oocyte, the centrosomes first localise anterior to the oocyte nucleus but in region 3 they detach from the fusome and move towards the posterior of the oocyte (Fig. 1A). At the same time, all of the oocyte markers analysed (Orb, Cup and Bic-D) are relocalised from the anterior to the posterior of the oocyte, and form a crescent with the highest concentration near the posterior cortex at stage 2 (Fig. 1A′ and data not shown).

To investigate further how centrosome migration is regulated, we analysed the behaviour of the centrosomes in egl and BicD mutant cysts. The centrosomes of egl cysts migrate along the fusome and accumulate in a single cell in late region- 2b and region-3 cysts (Fig. 2A). The centrosomes do not move to the posterior of this cell, however, and remain associated with the remnants of the fusome (Fig. 2A′,A′′). The cell in which the centrosomes accumulate is normally placed at the posterior of the cyst, and possesses four ring canals (not shown). In order to analyse the distribution of the centrosomes in BicD mutant cysts, we generated germline clones of a BicD null allele, and compared centrosome migration in these mutant cysts to that of BicD/+ cysts. We find that, like in egl⁻ cysts, centrosomes in BicD null cysts migrate along the fusome to a single cell and remain attached to the fusome at the anterior of this cell (Fig. 2B). These data demonstrate that, in striking contrast to the uniform distribution of oocyte markers, egl and BicD mutant cysts still possess an asymmetry that directs the accumulation of the centrosomes into a single cell.

It has been described that BicD mutants never form a polarised MT network (Theurkauf et al., 1993). Since centrosome migration still takes place in BicD cysts, this argues against a role for MTs in the accumulation of centrosomes in the oocyte. However, egl mutants have been reported to form transiently a polarised MT network in the germarium, and the migration of centrosomes in these mutant cysts may be a consequence of this fact (Theurkauf et al., 1993). In order to assess the role of MTs in centrosome movement in egl mutants, we have studied the distribution of MTs in wild-type and egl cysts (using a stronger mutant combination to the one previously reported), and with the help of several molecular markers for MT distribution and polarity.

To visualise MTs directly, we analysed the pattern of expression in live germaria of the MT binding protein tau-GFP (Kaltschmit et al., 2000). In early region-2a cysts, all the cells show similar levels of tau-GFP staining (not shown). Later in region 2a, a local concentration of tau-GFP can be seen in the two pro-oocytes. By region 2b, there is a restriction of tau-GFP accumulation to the oocyte (Fig. 3B). The identity of the cells with high levels of tau-GFP was confirmed by double staining with the oocyte marker Cup (Keyes and Spradling, 1997) (not shown). The localisation of tau-GFP and Cup to two and one cells takes place simultaneously, demonstrating that the organisation of the MT cytoskeleton in germarial cysts follows the same steps as the localisation of oocyte markers, and suggests that both are connected events.

The polarity of a MT network can be inferred by the behaviour of proteins, such as MT motors, that recognise the inherent polarity of MT polymers (Hirokawa, 1998; Vallee and Gee, 1998; Zhang et al., 1990). To determine the polarity of the MT cytoskeleton in germarial cysts, we have used three different MT polarity markers that are thought to localise to MT minus ends: the heavy chain of cytoplasmic dynein (Dhc64C) and two GFP chimaeras, D-TACC-GFP (Drosophila Transforming, Acidic Coiled-Coil-containing family of proteins homologue) and nod-GFP (no distributive disjunction) (Clark et al., 1997; Gegerly et al., 2000; Li et al., 1994; Zhang et al., 1990). The three markers give similar results and show a uniform distribution in all the cells of region-2a cysts, then an accumulation in the cytoplasm of the pro-oocytes and finally a localisation to the oocyte (Deng and Lin, 1997; Li et al., 1994) (Fig. 3C-E′). Since the minus ends of MTs often mark the place of their nucleation, these data suggest that MTs in germarial regions 2 and 3 cysts are primarily nucleated within the pro-oocytes and oocyte. Furthermore, the MT nucleating ability of these cells does not seem to be concentrated in a defined site; rather the microtubules’ minus ends appear to be present in the entire cell. These data, together with the distribution of tau-GFP, demonstrate that the single MT organising centre (MTOC) described for region-2 and 3-cysts (Theurkauf et al., 1993) lies initially within the two pro- oocytes, and is later restricted to the oocyte.

We investigated next the organisation of MTs in egl cysts. These mutant cysts do not show a visible MTOC in regions 2 or 3 and MTs are uniformly distributed in all the cells of the cyst (Fig. 3F). Furthermore, neither Dhc64C nor nod-GFP localise to one cell in mutant cysts (Fig. 3G,H; see also Li et al., 1994, for Dhc64C distribution in egl cysts). Although it has been previously reported that a polarised focus of MTs forms in one cell of egl mutant cysts, but is not maintained (Theurkauf et al., 1993), our data show that egl is required for all steps in the polarisation of the MT cytoskeleton of the cyst. Thus, egl mutants have the same effect on MT organisation as null alleles of BicD. Furthermore, the distribution of MTs in egl and BicD mutant cysts demonstrates that a polarised MT cytoskeleton is not necessary for the localisation of centrosomes to one cell of the cyst.

We tested this possibility further by treating wild-type ovaries with high concentrations of the microtubule-assembly inhibitor colcemid. The treatment affected the centrosomes themselves, as centrosomes of mitotically active germline cysts and follicle cells appeared enlarged (Fig. 4 and not shown). On the contrary, in a third of region-2b cysts and in most region- 3 and older cysts we cannot detect anti-γ-tubulin staining. In those late region-2b cysts that show a detectable staining with the anti-γ-tubulin antibody, we find that the vast majority of the centrosomes can still be detected in a single cell (Fig. 4). Although it is possible that the drug treatment has not broken down all of the microtubules, this result strongly supports the idea that centrosome migration does not depend on MTs.

Since BicD and egl abolish the polarisation of the MT cytoskeleton in the cyst, the only candidate for a polarised structure that could direct centrosome migration in these mutants is the fusome (de Cuevas and Spradling, 1998). In fact, the cells that accumulate the centrosomes of egl and BicD cysts possess the largest portion of the degenerating fusome (Fig. 2A′,A′′ and data not shown). This observation demonstrates that, like in wild type, the asymmetry established during fusome morphogenesis persists until the stages when the centrosomes migrate in BicD and egl cysts.

In order to test a direct role for the fusome in centrosome movement we analysed the behaviour of the centrosomes in a mutant that affects the integrity of the fusome. Germline clones of null alleles of Dynein heavy chain 64C divide correctly and produce cysts that show a very similar phenotype to BicD and egl. These mutant cysts contain 16 nurse cells in which oocyte cytoplasmic markers do not accumulate in a single cell (McGrail and Hays, 1997). In addition to this phenotype, we have discovered that a null allele of Dhc64C isolated in one of our laboratories, Dhc64C^greco, affects the integrity of the fusome. Dhc64C^greco mutant cysts possess a normal-looking fusome in region 1 and early in region 2a. However, the fusome of older region-2a cysts shows a fragmented appearance. Interestingly, the centrosomes of these cysts fail to migrate to a single cell, strongly suggesting that centrosome migration requires an intact fusome (Fig. 5A-E).

As presented above, Dhc64C is necessary for the localisation of centrosomes and cytoplasmic markers to the oocyte. We then investigated if Dhc64C was also required for the restriction of meiosis to the oocyte and analysed the distribution of the synaptonemal complex in Dhc64C^greco germline clones. We find that, like BicD mutants, Dhc64C is required for the formation of the SC (Fig. 5F). Thus, lack of function of dynein blocks the three asymmetries present in region-3 oocytes, suggesting that the restriction of meiosis to the oocyte, the organisation of a polarised MT centred in this cell, and the migration of centrosomes to the oocyte, depend upon the polarisation of the fusome.

Since the MT cytoskeleton seems to be polarised toward the oocyte prior to the migration of the centrosomes, this suggests that most of the centrosomes of region-2 cysts might have lost their MT nucleating properties. These post-mitotic centrosomes thus would act differently to their region-1 counterparts, which retain the ability to grow microtubules, at least during the mitotic divisions of the cyst (Grieder et al., 2000). We tested whether the molecular composition of post- mitotic centrosomes was different to mitotic ones. We analysed the distribution of Centrosomin (Cnn), a marker for the active centrosomes of mitotic cells (Li and Kaufman, 1996). Cnn, like γ-tubulin, is present in region-1 centrosomes (Fig. 6A). In contrast, Cnn is absent or barely detectable in region-2 and −3 cysts (Fig. 6B). This change in composition of centrosomes depends upon the activity of egl, as in egl mutant cysts Cnn reappears in post-mitotic region-2b centrosomes and by region 3 they possess a noticeable staining with the α-cnn antibody (Fig. 6C). Although we cannot offer an explanation for this difference, it suggests that the correct determination of the oocyte among the cells of the cyst affects the composition of the germline centrosomes.

The determination of a single oocyte within 16-cell cysts is a stepwise process, as revealed by the distribution of oocyte markers and the formation of synaptonemal complexes (Carpenter, 1975; González- Reyes and St Johnston, 1998; Huynh and St Johnston, 2000). We have shown that the organisation of MTs in post-mitotic cysts also follows these steps, since it is possible to distinguish a gradual, differential concentration of MTs in the pro-oocytes concomitant with that of oocyte markers. Later, when the oocyte is chosen, MTs largely originate from the oocyte. It has been proposed that this arrangement of the MT cytoskeleton is essential for the selection of a single cell within the developing cyst to become the oocyte (Gutzeit, 1986; Theurkauf, 1994; Theurkauf et al., 1993). However, meiosis can be restricted to a single cell in the absence of microtubules, implying that the cyst possesses an inherent asymmetry that is independent of MTs. In addition, the two pro-oocytes of wild-type cysts form synaptonemal complexes early in region 2a of the germarium, long before we can observe any signs of a polarised MT cytoskeleton (Huynh and St Johnston, 2000). Thus, the pro-oocytes and the oocyte behave differently to the rest of the cells of the cyst, independent of the organisation of the microtubule cytoskeleton.

In addition to the restriction of meiosis and to the localisation of specific markers, the oocyte can also be distinguished by the accumulation of the centrosomes. The nurse cell centrosomes migrate to the oocyte as the cyst matures in the germarium, and by region 3 most of them localise inside the oocyte (Fig. 1; see also Grieder et al., 2000). This migration is independent of the organisation of the MTs, and is most likely independent of the MTs themselves, since (1) mutants that do not possess a polarised MT cytoskeleton do not block centrosome migration, and (2) we observe that centrosomes accumulate in one cell located at the posterior of the egg chamber in the presence of high concentrations of colcemid (Fig. 2). In fact, the centrosomes appear to recognise the polarity of the fusome and migrate along it toward the oocyte. Our data on egl, BicD and Dhc64C offer strong support for this hypothesis. egl and BicD are believed to act close to the determination of the oocyte, as mutant cysts block most of the known asymmetries of germarial cysts (Huynh and St Johnston, 2000; Mach, 1997; Ran et al., 1994; Schüpbach and Wieschaus, 1991; Suter and Steward, 1991; Theurkauf et al., 1993). In striking contrast, however, these cysts localise centrosomes to one of the four-ring-canal cells, demonstrating that one of these cells can be selected in the absence of the activity of egl and BicD. Since it has been previously noted that the fusome of egl and BicD cysts is asymmetric (de Cuevas and Spradling, 1998), and since the disrupted fusome of Dhc64C cysts blocks centrosome movement to a single cell, it is reasonable to propose that the polarised fusome of egl, BicD and wild-type cysts is responsible for the migration of the centrosomes. This proposition is supported by our finding that the fusome asymmetry still can be detected in region-3 wild- type, egl or BicD cysts, when most of the centrosomes localise to one cell.

The selection of one oocyte among the cells of the cyst involves the establishment of at least three asymmetries, which are revealed by the localisation of different types of markers: cytoplasmic markers (such as BicD, Orb or Cup), nuclear markers (synaptonemal complex) and the centrosomes (Grieder et al., 2000; Huynh and St Johnston, 2000; Keyes and Spradling, 1997; Lantz et al., 1994; Suter and Steward, 1991) (Fig. 2). The localisation of cytoplasmic markers is MT-dependent, and requires the activity of the egl, BicD and Dhc64C genes (Koch and Spitzer, 1983; Mach, 1997; McGrail and Hays, 1997; Ran et al., 1994; Theurkauf et al., 1993). The restriction of meiosis to the oocyte still requires egl, BicD and Dhc64C, but not MTs (Huynh and St Johnston, 2000) (Fig. 4). Finally, the localisation of the centrosomes to the oocyte is independent of egl, BicD and a polarised MT network but depends on an intact fusome (Figs 2, 5). Thus, the three asymmetries present in region-3 oocytes represent three different ways to localise markers to this cell. In addition, since that of the fusome is the first visible asymmetry in the cyst, we propose that the migration of centrosomes, the restriction of meiosis to a single cell and the nucleation of a polarised cytoskeleton responsible for the localisation of oocyte cytoplasmic markers, are all consequences of the polarisation of the fusome in region 1 (Fig. 7). This view is reinforced by the phenotype of Dhc64C mutant cysts. Dhc64C is the only mutant known to affect the fusome after it has fulfilled its role during the divisions of the cyst, thus providing a way to analyse the importance of the fusome in oocyte determination. Although Dhc64C is a component of the microtubule motor dynein (Hirokawa, 1998; Li et al., 1994) and thus its mutant phenotypes might reflect a role for MTs in centrosome migration and SC formation, we favour a model in which Dhc64C disrupts the three asymmetries present in region-3 oocytes by affecting the fusome, especially since all our other data argue against a direct role for MTs in the movement of the centrosomes and in the restriction of meiosis to the oocyte. The fragmented appearence of the fusomes of Dhc64C mutant cysts may reflect a role of this molecular motor in the transport of vesicles necessary for the maintenance of the fusome’s integrity.

Microtubules of region-1 cysts are organised from their centrosomes, at least during mitosis. However, as the cyst enters region 2, most of the microtubules do not seem to be nucleated from centrosomes, as seen with the distribution of microtubule polarity markers and EM sections (Fig. 3) (Grieder et al., 2000; Mahowald and Strassheim, 1970). Region-2a cysts nucleate microtubules mainly in the pro-oocytes. Since at this stage each cell only contains one centrosome, the main microtubule cytoskeleton of the cyst is therefore organised from a maximum, if any, of two centrosomes. In region 2b the only MT organising centre (MTOC) of the cyst resides in the oocyte (Fig. 3B) (Theurkauf et al., 1993). Although we cannot rule out the possibility that the centrosomes of the pro-oocytes and later of the oocyte act as MTOCs, we favour the idea that these cells possess a nucleating activity independent of centrosomes. Two arguments support this view: (1) the distribution of three different minus-end reporter proteins (Dhc64C, nod-GFP and D-TACC-GFP) in region-2 oocytes is suggestive of a wide distribution of the minus ends of microtubules in these cells; (2) the fact that BicD and egl cysts do not have a visible MTOC but still localise the centrosomes to a single cell indicates that these localised centrosomes do not act as a nucleation centre. There are precedents for centrosome-independent nucleation of microtubules in differentiated cells such as the epithelial cells of the pupal wing (Mogensen et al., 1989; Tucker et al., 1986) and the oocyte itself, which is able to nucleate microtubules from its anterior cortex in stage-7 egg chambers in absence of any γ-tubulin- positive structures (data not shown) (González et al., 1998; Theurkauf et al., 1992). Indeed, the meiotic spindle of the oocyte is also organised in a centrosome-independent manner (Huettner, 1933).

This centrosome-independent nucleation of MTs implies that a process of centrosome inactivation, which allows the generation of a single MTOC in region-2 cysts, takes place in the germarium (Theurkauf, 1994). Several observations support this hypothesis. First, we have shown that the accumulation of centrosomes in one cell in egl and BicD mutants is not accompanied by an increase in the number of MTs nucleated by this cell. Second, the effect of colcemid treatment on active centrosomes (region-1 or follicle-cell centrosomes) versus region-2 centrosomes is different, as the latter stain more weakly with anti-γ-tubulin than the former ones (not shown). Third, centrosomes have to move through ring canals. Since the lumen of these canals in region 2b is 1- 2 μm and is mostly filled with fusome material (Mahowald and Strassheim, 1970), it seems difficult for an active centrosome with a large number of microtubules attached to it to be transported across the canal. In this respect, the positive γ- tubulin signal present in post-mitotic centrosomes cannot be regarded as an unequivocal indication of centrosome activity, as γ-tubulin is a structural component of centrosomes present in the pericentriolar material and centrioles.

We would like to thank Tom Hays, Ron Dubreuil, Thomas Kaufman, Peter Kolodziej, Jordan Raff, Bill Chia, Christine Field, John Kilmartin, Yuh Nung Jan, Allan Spradling, the Developmental Studies Hybridoma Bank (University of Iowa) and the Bloomington Stock Centre for fly stocks and reagents. María Martín-Bermudo, Jordi Casanova, Jordan Raff and Josh Shulman provided useful comments on the manuscript.