Asymmetric cell division: microtubule dynamics and spindle asymmetry

Asymmetric partitioning of cell fate determinants is employed to generate cell type diversity in a number of different organisms^†(Horvitz and Herskowitz,1992). However, asymmetric division does not necessarily result in unequal cell size. While asymmetrically segregated determinants can give cells a distinct developmental potential^§, cell size can correlate with how often a cell divides and can thereby fulfil an important function during cell lineage determination. For example, the C. elegans zygote divides to give two differently sized daughter cells: AB is about 25% larger than P1 and most of the cells of the hatching larva, 389 of 558, are derived from the AB lineage(Sulston et al., 1983). During neurogenesis in Drosophila, neural precursor cells (neuroblasts)divide asymmetrically. The newly born neuroblast is approximately 75% larger than its sister ganglion mother cell (GMC) and divides between 1 and 30 times further (Bossing et al., 1996). In contrast, the GMC divides only once to give rise to two glia or neurons. As a result, tissues are generated in the correct position and at the correct time in the developing embryo.

The direction of division and the ability of a cell to divide symmetrically or asymmetrically in size is brought about by rearrangement of the cytoskeleton. We know little about the molecular mechanisms that regulate cell size asymmetry; however, in the last few years much has been learned about the targets of these controls. One such substrate is the mitotic spindle and there is good evidence that its orientation and position in the cell determines the site of cell cleavage (reviewed by Strome and Wood, 1983). Asymmetric division can be likened to slicing a piece of cake (Fig. 1) A vertical slice divides the cake into two pieces of equal size and content,both with the same amount of chocolate cake and strawberry icing(Fig. 1A). However, a horizontal slice gives two pieces of unequal size and content, a large piece of chocolate cake and a small piece of cake with all of the icing(Fig. 1B). If instead of icing we consider cell fate determinants, it then becomes clear how the orientation and position of the mitotic spindle and the cleavage furrow direct symmetric or asymmetric cell division. For example, Drosophila epithelial cells divide symmetrically along the planar axis of the embryo to produce two daughters of equal size and mitotic potential. Factors localised at the basolateral cortex are segregated equally to both daughter cells(Fig. 1C)(Matsuzaki et al., 1998). In contrast, during neuroblast division localised cell fate determinants such as Prospero, a homeodomain-containing transcription factor that contributes to the identity of the GMC, are segregated asymmetrically into the basal GMC(Fig. 1D)(Doe et al., 1991;Matsuzaki et al., 1992;Matsuzaki et al., 1998;Vaessin et al., 1991). Therefore, despite having the same ectodermal origin as epithelial cells, Drosophila neuroblasts divide asymmetrically and the resulting daughter cells have distinct cell sizes, mitotic potential and cell fate(reviewed by Lu et al.,2000).

Changing from a symmetric to an asymmetric division requires a reorientation of the division axis. In Drosophila embryonic neuroblasts, this involves a 90° rotation of the pro/metaphase mitotic spindle (Kaltschmidt et al.,2000). In the early C. elegans embryo a 90° rotation of the centrosome-nucleus complex positions the cleavage plane such that localised P-granules, which are thought to play a role in germ line determination in the later embryo, are asymmetrically partitioned to the germ line precursor daughter (Hyman and White,1987; Kemphues and Strome,1997; Strome and Wood,1982). Additional control of spindle dynamics is necessary when,as in both Drosophila neuroblasts and the C. elegans zygote,the generation of daughter cells of distinct cell fates is accompanied by a difference in cell size. The mitotic spindle plays a key role in setting up the eccentrically placed cleavage furrow. Until recently the cleavage furrow in animal cells was thought to form equidistant between the two spindle poles;however, a growing number of observations show that asymmetry can be achieved in different ways, both between organisms and within a single organism. Here we present a short survey of asymmetric cell division, highlighting the recent findings on the control of spindle positioning and specification of the cleavage plane in the two-cell nematode C. elegans embryo and comparing them with results that have been obtained in other systems,especially the fruit fly, Drosophila melanogaster.

The first evidence to suggest a role for the mitotic spindle in positioning the cleavage plane came from experiments carried out over 80 years ago(Conklin, 1917). By exposing Crepidula eggs to centrifugal force and thereby moving the mitotic spindle, Conklin showed that its position determines the cleavage plane and that different regions of the cell cortex can respond to form a cleavage furrow. Conklin's work, together with later experiments by Hiramoto, in which he aspirated the spindle from sea urchin embryos(Hiramoto, 1956), further established the timing of cleavage site determination by the spindle and showed that it is complete by mid-anaphase^¶.

To initiate the cleavage furrow the mitotic spindle must dictate local changes of the cell surface in the form of either relaxation or contraction of the cell membrane. Which element of the mitotic spindle, the asters or the midzone microtubules^¶,provides that stimulus, remains an important question that has yet to be fully resolved. Nor is it necessary that all organisms, or even all cells within,use the same mechanism. Three mechanisms have been proposed by which different elements of the mitotic spindle could signal to the cell cortex to position the cleavage furrow (reviewed by Field et al., 1999; Gatti et al.,2000; Oegema and Mitchison,1997). According to the astral relaxation model(White and Borisy, 1983), the asters signal to the cell cortex near the poles, inducing it to relax(Fig. 2A). Alternatively, the equatorial region of the cell could be induced to contract, either by a signal from the asters (Fig. 2B)(Devore et al., 1989;Rappaport, 1986) or from the overlapping microtubules of the spindle midzone(Fig. 2C). It is also possible that both signals act together.

Several landmark experiments have addressed whether the spindle asters or the midzone produce the signal that positions the division plane. Historical evidence that the spindle asters determine the site of cleavage comes from micro-manipulation studies on sea urchin eggs(Rappaport, 1961). Microsurgical removal of the centre of the egg during the first division results in a horseshoe-shaped cell with two nuclei. At the next division, the two spindles produce three cleavage planes: two that bisect each of the spindles and one extra plane between the two adjacent spindles poles. As a result, four daughter cells are formed. The outcome of this experiment suggests that interacting spindle asters control the position of the cleavage plane, possibly by where they touch the cell cortex. However, this hypothesis is difficult to verify since there is, as yet, no direct experimental evidence for such causality. Nevertheless, one possible explanation in support of this model has been put forward (Foe et al.,2000). Foe and co-workers have studied the interactions of spindle microtubules and the actomyosin cytoskeleton in syncytial Drosophilablastoderm embryos and find that filamentous actin and cytoplasmic myosin II are transported towards microtubule plus ends(Foe et al., 2000). From these findings they suggest that cleavage plane induction occurs at sites where actin filaments attach both to the cortex and to microtubules.

A number of studies in grasshopper neuroblasts(Kawamura, 1977), newt kidney epithelial cells and echinoderm eggs(Rappaport and Rappaport,1974) support the model that the midzone microtubules specify the site of cleavage furrow formation. For example, inserting a small block between the midzone microtubules and the cell cortex in flattened echinoderm eggs results in inhibition of cell division(Rappaport, 1986;Rappaport and Rappaport,1983). By comparing cleavage activity with the position of midzone microtubules in cultured epithelial cells(Cao and Wang, 1996;Wheatley and Wang, 1996), it was concluded that the signal triggering furrow formation is emitted by the midzone microtubules. A more exacting experiment would be to eliminate the astral microtubules specifically and observe whether the midzone on its own is able to induce a cleavage plane. Bonaccorsi et al. did just this and showed that spermatocytes and larval neuroblasts from a Drosophila asterless(asl) mutant could still undergo anaphase and telophase, thus implying that astral microtubules are not necessary to induce cytokinesis(Bonaccorsi et al., 2000;Bonaccorsi et al., 1998). It is difficult in these experiments to be certain that all astral microtubules are lacking, and it is possible that only a few astral microtubules are sufficient to induce cytokinesis. Furthermore, it has been shown that cells with acentrosomal spindles (which resemble anastral spindles in that they also lack astral microtubules) form bipolar spindles and enter anaphase, but cytokinesis often fails (Khodjakov and Rieder,2001).

The best-studied example of asymmetric cell division is the C. elegans embryo, which is characterised by a cascade of asymmetric divisions throughout its early development. The large cell size, simplicity and transparency of the animal, together with the strictly defined hierarchy of cell divisions, makes it well suited to study the mechanics of asymmetric cell division and cytokinesis (reviewed byPlasterk, 1999). Forward and reverse genetic approaches, together with mechanical manipulation and sophisticated imaging techniques, have been used to address spindle positioning and specification of the cleavage plane during early C. elegans development.

The anterior-posterior polarity of the one-cell C. elegans embryo is established between the time of fertilisation and the first mitotic division in response to an external cue provided by the sperm(Wallenfang and Seydoux,2000). The next phase in generating polarity is marked by a number of characteristic changes, which result in the production of two cells distinct in cell fate and cell size. After the oocyte and sperm pronuclei meet in the posterior hemisphere, they move to the centre of the embryo(Albertson, 1984;Hyman and White, 1987). Following pronuclear migration, the mitotic spindle in the C. eleganszygote is initially positioned symmetrically along the anterior-posterior axis. As the spindle grows, one centrosome moves towards the posterior cell cortex while the other remains relatively stationary, generating a spindle that is off-centre. Cleavage occurs midway between the two spindle poles giving rise to a large anterior AB cell and a smaller posterior P1 cell(Albertson, 1984).

What are the molecular forces that act on the mitotic spindle to cause this asymmetry? Two types of force that are dependent on microtubules play a role in spindle positioning and spindle pole separation in different model organisms. First a sliding force generated between the overlapping midzone microtubules, possibly mediated by plus-end-directed motor proteins of the kinesin family, could separate spindle poles(Fig. 3A). Second, the astral microtubules on each pole could act, perhaps via the minus-end directed microtubule motor dynein, to drag the spindle poles to opposite sides of the cell (Fig. 3B). In addition,cortical cues could cause microtubules to be selectively destabilised(Fig. 3C) or stabilised(Fig. 3D) in one region of the cell, which would result in an overall imbalance of the microtubule polymerisation forces.

A possible mechanism by which such cell polarity cues translate to the asymmetric spindle positioning in C. elegans has been proposed(Grill et al., 2001). Time-lapse analysis of the one-cell C. elegans embryo had previously shown that the anterior centrosome remains fixed in position, while the posterior centrosome oscillates and becomes smaller as it moves closer to the cell cortex (Albertson, 1984;Hyman and White, 1987;Keating and White, 1998). The consequence of such unequal centrosomal movement is an asymmetrically positioned spindle with the posterior centrosome closer to the cell wall. To reveal the forces that act on each spindle pole, Grill et al. removed the central spindle by laser ablation while leaving both spindle poles intact(Grill et al., 2001). In irradiated wild-type embryos, the posterior spindle pole moved faster and further than the anterior pole. This elegant experiment reveals that pulling forces act on the spindle poles, and that the posterior shift of the spindle in wild-type C. elegans zygotes results from a larger pulling force acting on the posterior pole than on the anterior pole.

The asymmetry of the net forces acting on the two spindle poles is under control of the par genes (Grill et al., 2001). In wild-type one-cell C. elegans embryos,PAR-3 localises to the anterior cortex(Etemad-Moghadam et al.,1995), while PAR-2 localises to the posterior(Boyd et al., 1996). The spindle in both, par-2 and par-3, is centrally positioned(Kemphues et al., 1988). After removal of the central spindle in par-3 mutants, the velocity of both spindle poles resembles that of the posterior spindle pole in wild-type zygotes. In contrast, after removal of the central spindle of par-2mutants both spindle poles move apart with a velocity equal to that of the anterior spindle pole in wild-type zygotes. This shows that the microtubule dynamics in the C. elegans zygote are under the control of polarity factors asymmetrically localised in the cell(Grill et al., 2001).

PAR-3 may stabilise or anchor microtubules (reviewed byRose and Kemphues, 1998;Cheng et al., 1995;Etemad-Moghadam et al., 1995)and there is good reason to think that asymmetrically localised PAR-3 could act to regulate aster movement by controlling microtubule stability. Only microtubules on the anterior side of the zygote are stabilised, leaving those of the posterior aster free to depolymerise. It is intriguing that the mammalian homologues of the serinethreonine kinase PAR-1, which is localised to the posterior cortex of the C. elegans zygote, have been shown to destabilise microtubules (Drewes et al.,1997; Ebneth et al.,1999). Although no direct effect of PAR-1 on microtubule dynamics has yet been observed, these findings would support the working model proposed by Grill et al.: all microtubules of both asters generate equal forces, but the interactions of the microtubules and the posterior cortex are weaker than those of the anterior cortex (Grill et al., 2001).

Until recently the cleavage furrow in animal cells was thought to form equidistant between the two poles of the mitotic spindle. If so, then repositioning of the mitotic spindle or migration of the nucleus would be sufficient to give rise to an eccentrically placed cleavage plane. As described above, the asymmetric division of the C. elegans zygote provides an example for repositioning the mitotic spindle(Fig. 4A). Similarly, in the clam Spisula zygote, one aster is stationary while the other oscillates to a position closer to the cell cortex(Fig. 4A)(Dan and Inoué, 1987). In vegetal cells of sea urchin embryos, lateral migration of the nucleus led by a centrosome prior to spindle formation has been shown to give rise to an asymmetrically placed spindle (Fig. 4B) (Dan, 1979;Schroeder, 1987). However, the generation of asymmetry during Drosophila neuroblast division follows a different scheme altogether, since the cleavage furrow does not form equidistant between the two spindle poles(Fig. 4C).

During neurogenesis in Drosophila, neuroblasts delaminate from the neuroectoderm and undergo asymmetric stem-cell like divisions, generating another neuroblast and a GMC. In vivo imaging of Drosophila embryos expressing a GFP (green fluorescent protein) fusion to the microtubule binding protein tau (Brand, 1995)reveals that, in embryonic neuroblasts, the mitotic spindle is symmetric and centrally placed through metaphase. However, at the onset of anaphase, the microtubules appear to shorten on the basal side of the cell and elongate on the apical side (Kaltschmidt et al.,2000). The overlapping apical and basal astral microtubules, which are distinctly different in length in Drosopila embryonic neuroblasts, could specify the eccentric position of the cleavage furrow. The elongation of the apical astral microtubules towards the emerging GMC occurs before the cell membrane starts to pucker, and membrane invagination occurs before the midbody moves towards the cleavage furrow(Kaltschmidt et al., 2000). The eccentric placement of the cleavage plane in Drosophila embryonic neuroblasts might, therefore, support the postulated role of astral microtubules in specifying the site of the cleavage furrow as described above(reviewed by Oegema and Mitchison,1997). In contrast, in Drosophila asl and centrosomin(cnn) mutants, larval neuroblast divisions are still asymmetric, in spite of partially defective mitotic centrosomes and the absence of detectable astral microtubules (Bonaccorsi et al.,2000; Megraw et al.,2001; Megraw et al.,1999; Vaizel-Ohayon and Schejter, 1999). Giansanti et al. postulate that the eccentric position of the cleavage plane in Drosophila neuroblasts is defined by signals originating from the midbody, and it is the asymmetric position of the midbody that leads to the overall spindle asymmetry(Giansanti et al., 2001). It is interesting to note that these two types of asymmetric division, an asymmetric division reminiscent of that seen in the C. elegans zygote and an asymmetric spindle similar to that described for Drosophilaneuroblasts, can exist within the same lineage, such as that of the Drosophila sensory organ precursor (SOP) cells(Roegiers et al., 2001).

The source of information to position an eccentric cleavage plane is unclear. One possibility is a mechanical connection between the mitotic spindle and the cell cortex (Dan and Ito,1984). Indeed, a specific mechanical linkage between the spindle and the plasma membrane has been demonstrated in a number of unequally dividing cell types (Conklin,1917; Harvey,1935). The most convincing evidence for a specific site comes from micro-manipulation studies in Chaeopterus oocytes: after being displaced by a needle, the spindle returns to its original position(Lutz et al., 1988).

Giansanti et al. suggest that in Drosophila neuroblasts the shift of the midbody towards the GMC occurs via a mechanical link between the cortex, nucleus and midbody (Giansanti et al., 2001). Alternatively, one could imagine a mechanism whereby astral microtubules on the basal spindle pole of Drosophilaneuroblasts are induced to depolymerise while those of the apical aster are stabilised, resulting in a larger apical and smaller basal aster that together constitute an asymmetric spindle. This model agrees with a role of astral microtubules in the placement of the cleavage plane as previously suggested(Rappaport, 1961). Certainly,both this model and that of mechanical linkage predict a specialised site on the GMC cortex to facilitate local interaction between the spindle microtubules and the cell cortex. If such an interaction actively destabilises microtubules locally, then polymerisation forces would become unbalanced,resulting in either asymmetrically positioned asters and/or differently sized asters.

Bearing in mind the requirement for asymmetrically stabilised spindle poles during asymmetric cell division, it is intriguing that all the above mentioned examples of eccentrically placed spindles share an asymmetry in the morphology of their centrosomes and the microtubules they produce. For example, the anterior aster of the asymmetrically-positioned spindle of the C. elegans zygote is large and has many microtubules, while the posterior aster appears flattened and smaller and has fewer astral microtubules(Keating and White, 1998). A difference in aster size and morphology has also been described for the divisions of the four-cell sea urchin embryo. The micromere centrosomes contain less centrosomal material than the macromere poles(Holy and Schatten, 1991) and the macromere aster is spherical, whereas the micromere aster undergoes elongation during late anaphase and telophase and is flattened perpendicular to the spindle axis (Holy and Schatten,1991). Note, however, that it is possible that micromere centrosomes contain the same material but are merely more condensed than the macromere centrosomes. The unique morphology of the micromere aster in sea urchin embryos has been suggested to be due to proximity to the plasma membrane (Dan and Nakajima,1956). As the two centrosomes have already begun to become distinct in metaphase (Holy and Schatten,1991), this must also reflect an intrinsic difference between the two microtubule organising centres. In Drosophila embryonic neuroblasts, as the astral microtubules become longer and more abundant at the beginning of anaphase, the apical aster enlarges. The basal aster is concomitantly reduced in size and the basal centrosome has reduced levels of the centrosomal proteins γ-tubulin, CP60 and CP190(Kaltschmidt et al., 2000). This was found also to be the case for larval Drosophila neuroblasts(Bonaccorsi et al., 2000;Ceron et al., 2001).

Microfilaments may be part of the spindle-positioning machinery in several organisms. First, actin has been reported to localise temporarily to the anterior region of the C. elegans zygote(Hill and Strome, 1988;Strome, 1986) and asymmetric positioning of the mitotic spindle is inhibited by disrupting microfilaments with cytochalasin D during a narrow time interval in the first cell cycle(Strome and Wood, 1983). Second, an enrichment of actin has also been postulated to play a role in establishing cortical polarity in the mouse egg(Longo and Chen, 1985). In immature mouse oocytes actin is cortical, while in mature eggs it is asymmetrically localised. When induced to undergo maturation, the meiotic spindle forms in the centre of the oocyte and then moves towards the actin-rich periphery, where it becomes anchored to the plasma membrane(Chambers, 1917;Conklin, 1917;Longo and Chen, 1984). Disruption of microfilaments with cytochalasin B inhibits this movement. Third, actin has also been shown to localise asymmetrically in Drosophila larval neuroblasts(McCartney et al., 1999) and it is possible that localised actin functions (possibly only briefly) to set up or respond to the spatial cues that are needed to establish the asymmetry in the spindle of Drosophila larval neuroblasts.

What are the factors known to be necessary for induction of spindle asymmetry in embryonic Drosophila neuroblasts, and how are they regulated during the embryonic neuroblast cell cycle? First, Inscuteable [a protein of 859 amino acids encoding a putative SH3 target site, ankyrin repeats and a PDZ-binding domain (Kraut and Campos-Ortega, 1996)] localises as an apical crescent in neuroblasts from late interphase until anaphase(Kraut et al., 1996) and is both necessary and sufficient to direct apical-basal cell division in neuroblasts (Knoblich et al.,1999; Kraut et al.,1996; Tio et al.,1999). In inscuteable^P72 null embryos, the mitotic spindle fails to rotate and the direction of neuroblast division is no longer strictly apical-basal (Kaltschmidt et al., 2000; Kraut et al.,1996). Several proteins are required to localise Inscuteable. Pins(Partner of Inscuteable), a tetratrico-peptide (TPR) repeat protein, binds to Inscuteable and shows an almost identical subcellular localisation(Parmentier et al., 2000;Schaefer et al., 2000;Yu et al., 2000). Inscuteable localisation is established but not maintained in pins mutants and as a consequence the mitotic spindle in embryonic neuroblasts is misoriented.

Pins encodes three `GoLoco' motifs, which are present in proteins that bind the α subunits of heterotrimeric G-proteins, Gα₀ and Gα_i (Schaefer et al.,2000). Gα₀/Gα_i, together with Gβγ, comprise the G-protein complex and are involved in the organisation of the actin cytoskeleton and asymmetric localisation of cortical proteins in several different organisms (reviewed byChant, 1999;Jin et al., 2000). Schaefer et al. have recently shown that, in Drosophila embryonic neuroblasts,Inscuteable functions via Pins as an apical adaptor for Gα_i,which in turn sets up a polarity cue at the apical neuroblast cortex(Schaefer et al., 2001). In addition, overexpression of Gα_i in neuroblasts produces two equal-sized daughter cells (Schaefer et al., 2001). The heterotrimeric G-protein cascade, which is confined to the apical cell cortex, thereby mediates asymmetric neuroblast division, possibly via reorganisation of the actin cytoskeleton (reviewed bySchweisguth, 2000). Gαis also required for correct positioning and morphology of the mitotic spindle in the C. elegans zygote (reviewed byGotta and Ahringer, 2001a;Gotta and Ahringer, 2001b). Thus, G-protein signalling during spindle orientation may be a process conserved between Drosophila and C. elegans.

Gotta and Ahringer suggest that G-protein signalling may function to connect spindle position and polarity in C. elegans(Gotta and Ahringer, 2001a). The mammalian homologue of Pins, Ags-3, functions as a receptor-independent activator of G-protein signalling(Takesono et al., 1999). In C. elegans simulatenous inhibition of two genes with weak homology to Pins (ags-3.2 and ags-3.3) recapitulates the phenotype of embryos lacking Gα activity: asymmetric spindle positioning is affected as is the generation of different-sized daughter cells [M. Gotta and J. Ahringer, personal communication; S. Grill, P. Gonczy and A. Hyman, personal communication (Gotta and Ahringer,2001a)]. In Drosophila pins mutants a large number of larval neuroblasts divide symmetrically(Parmentier et al., 2000),although this has not been observed in embryonic neuroblasts(Schaefer et al., 2000;Yu et al., 2000), possibly due to maternal Pins protein being present and sufficient for embryonic divisions.

Lu et al. have shown that adherens junctions are responsible for the default, planar orientation of the mitotic spindle in epithelial cell(Lu et al., 2001). RNAi against the epithelial-cell-enriched (E)-adenomatous polyposis coli (APC)tumor suppressor protein and microtubule-associated protein EB1 (both of which are adherens-junction-associated proteins), causes epithelial cells to switch from a symmetric to an asymmetric division pattern(Lu et al., 2001). This implies a possible function of adherens junctions in preventing asymmetric cell division. It is noteworthy, however, that cells in the procephalic neurogenic region (PNR) have adherens junctions but nonetheless divide asymmetrically.

The minus-end directed microtubule motors and their activators may also play a role in generating asymmetry. Several models have been proposed for the function of dynein and dynactin during nuclear migration/spindle positioning in S. cerevisiae. For example, dynein may alter microtubule dynamics(Carminati and Stearns, 1997). In the absence of dynein, the rates of microtubule polymerisation and depolymerisation are significantly slower, and catastrophe frequencies are reduced by half. One explanation could be the presence of dynein at the spindle pole and on cytoplasmic microtubules(Yeh et al., 1995). Dynein may induce conformational changes in microtubules and thereby affect microtubule growth dynamics (e.g. reduce growth and shrinking rates as well as the frequency of catastrophe). Alternatively, the presence of dynein on a microtubule might exclude binding of microtubule-associated proteins that would normally act to stabilise the microtubules.

In the C. elegans zygote, cytoplasmic dynein and dynactin are required during pronuclear migration, centrosome positioning and pronuclear rotation (Gönczy et al.,1999; Skop and White,1998). Cytoplasmic dynein is further involved in maintaining the tight association between the centrosomes and the male pronucleus. Gönczy et al. suggest a mechanism by which cytoplasmic dynein, anchored to the pronucleus, drives centrosome separation. This model predicts that the pulling forces required during centrosome separation are provided by interactions between cytoplasmic dynein anchored on the nuclear membrane and astral microtubules (Gönczy et al.,1999).

The dynein-dynactin complex has also been implicated in the rotation of the centrosome-nucleus-complex in the C. elegans P1 cell(Skop and White, 1998;Waddle et al., 1994). The original model suggested a cortical capture mechanism(Hyman and White, 1987). Laser microsurgery experiments identified a cortical site rich in actin,actin-capping proteins and dynactin(Hyman, 1989;Waddle et al., 1994). Reducing the levels of two orthologues of the C. elegans dynactin complex results in misalignment of the spindle in the P1 cell(Skop and White, 1998). By localising to the cell cortex, dynactin may both tether microtubule ends and bind to the minus-end directed, microtubule-associated dynein, thereby activating its motor activity. While tethered to the cell cortex, dynein could reel in one aster by moving along astral microtubules, depolymerising and shortening them (Skop and White,1998; Waddle et al.,1994). An alternative interpretation of P1 spindle misalignment is that the spindle is displaced to an eccentrically localised cortical site as a result of asynchronous ingression of the first cleavage furrow(Gönczy et al., 1999). This would imply that the dynein-dynactin complex controls the position of spindle attachment.

In Drosophila embryonic neuroblasts, a subunit of dynactin,p150^Glued, is localised in a basal cortical crescent before it is asymmetrically segregated to the GMC cortex (J.A.K. and A.H.B., unpublished). It is possible that, by binding to dynactin at the GMC cortex, dynein mediates both the rotation of the pro/metaphase spindle and the difference in length of astral microtubules in the neuroblast.

Asymmetric cell division relies on the position of the mitotic spindle,which is regulated by several different mechanisms, including the Par proteins, G-protein signalling and the dynein-dynactin complex. It remains to be seen how these mechanisms are coordinated and regulated. Technical advances, such as laser ablation (reviewed byKhodjakov et al., 1997) and speckle imaging analysis of microtubule dynamics(Waterman-Storer and Salmon,1999) should help to further our understanding of these questions.

We thank Julie Ahringer, Monica Gotta, Jordan Raff, Adelaide Carpenter,Cayetano González, Alfonso Martinez Arias and Markus Schober for helpful discussions and critical reading of the manuscript. J.A.K. was supported by a Wellcome Prize Studentship and King's College Cambridge. This work is funded by the Wellcome Trust.