The Drosophila Lkb1 kinase is required for spindle formation and asymmetric neuroblast division

Drosophila neuroblasts (NBs) are one of the best model systems for the study of the control of cell polarity and asymmetric cell division. During Drosophila embryogenesis, NBs delaminate basally from the neuroectodermal epithelium and divide asymmetrically along the apical/basal axis to produce another NB and a smaller ganglion mother cell (GMC). The newly generated apical NB divides repeatedly in an asymmetric fashion, whereas the basal GMC divides symmetrically just once to generate equally sized daughter cells that differentiate into neurons or glia (reviewed by Betschinger and Knoblich, 2004; Wodarz, 2005).

The asymmetric division of Drosophila NBs is regulated by several proteins that concentrate at the cell cortex. The basal cortex is enriched in the cell fate determinants Prospero (Pros) and Numb, as well as in their respective adaptor proteins Miranda (Mira) and Partner of Numb (Pon). These proteins are preferentially segregated into the GMC following NB cytokinesis. Localization of Pros-Mira and Numb-Pon at the basal cortex is mediated by a large multiprotein complex that concentrates at the apical cortex. This complex includes two functionally distinct subcomplexes. One of them contains Bazooka (Baz; PAR-3 in C. elegans), aPKC (Drosophilaatypical protein kinase C, DaPKC) and Par-6; this assembly is hereafter referred to as the Baz/Par-6 subcomplex. The other subcomplex includes the Gαi subunit of the heterotrimeric G-protein complex and Partner of inscuteable (Pins; Raps - Flybase), and is hereafter referred to as the Pins/Gαi subcomplex. The Baz/Par-6 and Pins/Gαi subcomplexes are integrated in a larger apical complex by the Inscuteable (Insc) protein that binds both Pins and Baz (reviewed by Betschinger and Knoblich, 2004; Wodarz, 2005).

Recent genetic analyses have shown that the Baz/Par-6 subcomplex is mainly involved in the control of proper basal localization of Pros/Mira and Numb/Pon. The Pins/Gαi subcomplex is instead required for spindle orientation during NB divisions. Both complexes, however, cooperate in controlling cleavage furrow positioning during asymmetric NB divisions. Mutations that disrupt either the Baz/Par-6 or the Pins/Gαi pathway have little or no effect on asymmetric cytokinesis. However, mutations that disrupt both pathways completely abrogate spindle displacement during telophase,leading to symmetric cytokinesis (Cai et al., 2003; Izumi et al.,2004; Shaefer et al., 2000; Yu et al., 2000; Yu et al.,2003).

In this study, we have addressed the role of Drosophila Lkb1(Dlkb1), the homolog of LKB1 kinase (STK11 - Human Gene Nomenclature Database)in NB division. LKB1 kinase is mutated in the Peutz-Jeghers syndrome, an autosomal dominantly inherited disorder characterized by the formation of intestinal polyps and a high incidence of various cancer types. Somatic mutations in the LKB1 gene have also been detected in sporadic adenocarcinomas (reviewed by Alessi et al.,2006; Baas et al.,2004b). There is evidence that LKB1 plays a conserved role in the control of cell polarity. Recent work has unambiguously shown that activation of LKB1 leads to rapid and complete polarization of human intestinal epithelial cells (Baas et al.,2004a). Similarly, PAR-4, the C. elegans homolog of LKB1,is required for correct polarity and asymmetric division of one-cell embryos(Watts et al., 2000). Furthermore, Dlkb1 mediates determination of anterior/posterior polarity of egg chambers and embryos, as well as the proper polarity of follicle cells(Martin and St Johnston,2003). Here, we demonstrate that Dlkb1 controls many asymmetries that characterize the mitotic division of larval NBs. dlkb1 mutations also disrupt mitotic spindle assembly, leading to the frequent formation of polyploid cells. Thus, in addition to cell polarity and the geometry of cell division, Dlkb1 directly or indirectly regulates the stability of spindle microtubules (MTs).

The dlkb1³¹⁵ mutant allele was isolated from a collection of 1600 third chromosome late lethals induced by ethylmethanesulfonate (EMS) in C. Zuker's laboratory(Koundakjian et al., 2004). The dlkb1⁷ allele is associated with the chromosome carrying Df(3R)su(Hw)7. This and all the deficiencies used for mapping were obtained from the Bloomington Stock Center. The pins^P62 null allele and the asl²mutation have been described previously(Bonaccorsi et al., 1998; Yu et al., 2000); dlkb1³¹⁵ pins^P62 and dlkb1³¹⁵asl² double mutants were generated by recombination. All mutations were maintained over the TM6B balancer, and mutant larvae were identified based on their non-tubby phenotype. Genetic markers and special chromosomes are described in FlyBase(http://www.flybase.org/). Germline transformation was performed as previously described(Vernì et al.,2004).

The anti-Dlkb1 antibody was generated in guinea pig using a maltose-binding protein (MBP)/Dlkb1 fusion protein. Expression of the fused protein in Escherichia coli and the production and purification of antibodies against this fusion were according to Vernì et al.(Vernì et al., 2004). Immunoblotting was performed as previously described(Vernì et al., 2004);the anti-Dlkb1, anti-Pins and anti-Giotto(Giansanti et al., 2006)antibodies were diluted 1:2000, 1:1000 and 1:5000, respectively.

Brains from third instar larvae were dissected and fixed according to Bonaccorsi et al. (Bonaccorsi et al.,2000). After several rinses in PBS, brain preparations were incubated overnight at 4°C with a monoclonal anti-α-tubulin antibody(Sigma-Aldrich), diluted 1:1000 in PBS-T (PBS with 0.1% Triton X-100), and any of the following rabbit antibodies, also diluted in PBS-T: anti-Centrosomin(1:300; gift of T. Kaufman, Indiana University, Bloomington, IN, USA),anti-Deadpan (1:400; gift of Y. Jan, Howard Hughes Medical Institute,University of California, CA, USA), anti-Mira (1:500; gift of Y. Jan),anti-Bazooka (1:50; gift of F. Matsuzaki, Japan Science and Technology Corporation, Kobe, Japan); anti-Gαi (1:200; gift of J. Knoblich,Austrian Academy of Sciences, Vienna, Austria); anti-Par-6 (1:1000; gift of J. Knoblich); anti-DaPKC (1:100; Santa Cruz Biotechnology) and anti-Mud (1:200;gift of F. Matsuzaki). After two rinses in PBS, primary antibodies were detected by a 1-hour incubation at room temperature with FITC-conjugated anti-mouse IgG+IgM (1:20; Jackson Laboratories) and Alexa 555-conjugated anti-rabbit IgG (1:300; Molecular Probes), diluted in PBS.

For double Centrosomin/Pins immunostaining, brains were incubated overnight at 4°C with the rabbit anti-Centrosomin antibody (1:300) and a rat anti-Pins antibody (1:100; gift of W. Chia, The National University of Singapore, Singapore) diluted in PBS-T. Detection was performed by 1-hour incubation at room temperature with Alexa 555-conjugated anti-rabbit IgG(Molecular Probes) and FITC-conjugated anti-rat IgG (Jackson Laboratories)diluted 1:300 and 1:20 in PBS, respectively.

For Dlkb1 immunostaining, brain preparations were incubated overnight at 4°C with the anti-Dlkb1 antibody (1:100 in PBS-T) and, after rinsing in PBS, were incubated 1 hour at room temperature with Alexa 555-conjugated anti-guinea pig IgG diluted 1:500 in PBS.

In all cases, immunostained preparations were mounted in Vectashield medium H-1200 (Vector Laboratories) containing the DNA-dye DAPI(4′,6-diamidino-2-phenylindole). Preparations were examined with a Zeiss Axioplan microscope, equipped with an HBO100W mercury lamp and a cooled charged-coupled device (CCD camera; Photometrics CoolSnap HQ). Grayscale images were collected separately, converted to Photoshop (Adobe Systems),pseudocolored and merged.

Spindle measurements were taken on enlarged digital images and scaled down to their size in μm. In preparations stained for Centrosomin, measurements were taken from centrosome-to-centrosome. In the absence of Centrosomin staining, measurements were taken from pole-to-pole in anastral spindles; in the presence of asters, measurements were taken from the center of the astral MT array.

In the course of a screen aimed at the isolation of mitotic mutants (see Materials and methods), we identified a lethal mutation that causes frequent polyploid cells in larval brains (see below). Animals homozygous for this mutation die at the larval/pupal transition, as do most mitotic mutants; most probably, they exploit maternally supplied products to survive until late larval stages (Gatti and Baker,1989). Deficiency mapping showed that this mutation is uncovered by both Df(3R)urd and Df(3R)26c, which define a map interval that contains only 16 annotated genes. During these mapping studies, we also identified another mutant allele of this same mitotic gene. This allele is associated with the chromosome that carries Df(3R)su(Hw)7, but is independent of the deficiency. We next sequenced the candidate genes and found that both mutant stocks carry lesions in the Drosophila lkb1 gene(dlkb1, also known as CG9374). This gene encodes a 567 amino acid serine/threonine kinase homologous to the PAR-4 kinase of C. elegansand the human LKB1 kinase mutated in Peutz-Jeghers syndrome(Martin and St Johnston,2003). The dlkb1³¹⁵ mutant allele isolated in our screen carries a frameshift mutation resulting in a truncated Dlkb1 protein of 234 amino acids; the dlkb1⁷ allele, associated with Df(3R)su(Hw)7, has a stop codon that truncates Dlkb1 to a 346 amino acid polypeptide (Fig. 1A). A genomic fragment including sequences that extend roughly 1 kb on either side of dlkb1 (Fig. 1A) rescued both the lethality and the mitotic phenotypes of dlkb1³¹⁵/dlkb1³¹⁵ and dlkb1³¹⁵/dlkb⁷ mutants.

Drosophila brains contain mostly two types of dividing cells: NBs and GMCs (Goodman and Doe,1993). Wild-type larval NBs are characterized by many asymmetries that develop during the course of mitosis. Prometaphases and metaphases of larval NBs exhibit centrosomes and asters of similar sizes at the two cell poles. However, as NBs progress through anaphase and telophase, the MTs of the basal aster shorten dramatically, whereas those of the apical asters elongate slightly (Fig. 1B). Concomitantly, the basal centrosome becomes smaller than the apical one(Bonaccorsi et al., 2000) (see Fig. 4A below). These changes in aster and centrosome morphology are accompanied by a progressive displacement of the central spindle towards the basal pole, resulting in unequal cytokinesis (Giansanti et al.,2001). GMCs display equally sized centrosomes and very small asters throughout mitosis, and divide symmetrically(Fig. 1B)(Bonaccorsi et al., 2000).

To determine the mitotic defect leading to polyploid cell formation in dlkb1 mutants, we examined larval brain preparations from dlkb1³¹⁵/dlkb1³¹⁵,dlkb1³¹⁵/Df(3R)urd and dlkb1³¹⁵/dlkb1⁷ larvae stained for both tubulin and DNA. These mutant combinations showed identical mitotic aberrations. Most strikingly, mutant spindles showed an overall MT density substantially lower than that seen in wild-type spindles (Fig. 1B). In approximately 80% of mutant spindles, asters were absent or severely reduced; in control brains, the frequency of spindles without asters, or with very small asters, was 49%(Fig. 1B; Table 1). In addition, most mutant prometaphase and metaphase figures were characterized by low densities of both kinetochore and interpolar MTs, and ana-telophases displayed central spindles thinner than their wild-type counterparts(Fig. 1B and Fig. 3D below). Mutant brains also showed an increase in the relative frequency of metaphase figures with respect to wild type, suggesting that dlkb1 mutations lengthen metaphase duration (Table 1). Finally, mutant brains displayed approximately 20% polyploid cells (not shown); in wild-type brains, the frequency of polyploid cells was virtually zero (Table 1). The phenotype of dlkb1³¹⁵ homozygotes was qualitatively and quantitatively similar to that observed in dlkb1³¹⁵/Df(3R)urd hemizygotes, indicating that dlkb1³¹⁵ is a null mutation(Table 1).

The spindle phenotypes observed in dlkb1 mutants could be due either to a defect in MT elongation and/or stability, or to a defect in centrosome function. To distinguish between these possibilities, we sought to eliminate centrosome function in dlkb1 mutants. We have previously shown that brain cells of asterless (asl) mutants fail to assemble functional centrosomes and nucleate astral MTs. Nonetheless, asl NBs and GMCs manage to form robust anastral spindles that are able to mediate chromosome segregation(Bonaccorsi et al., 2000; Giansanti et al., 2001). We thus constructed dlkb1 asl double mutants and compared their phenotype with those exhibited by dlkb1 and asl single mutants. The anastral spindles from asl single mutants displayed a MT density comparable to wild type (Fig. 1C). By contrast, in dlkb1 asl double mutants, the density of spindle MTs was substantially reduced with respect to wild type and similar to that observed in dlkb1 single mutants(Fig. 1C). These results strongly suggest that the low density of spindle MTs observed in dlkb1 mutants does not depend on centrosome dysfunction. Thus, the spindle phenotype of dlkb1 mutants is likely to be attributable to either a decreased rate of MT growth or an increased MT instability.

Observation of mitotic divisions stained for tubulin and DNA revealed that the spindles of dlkb1 mutant cells are generally smaller than in wild type (Fig. 1). In addition, the frequency of asymmetric telophases in mutant brains (35-37%) was significantly lower than in wild-type brains (65%) (Table 1). These phenotypes could reflect a partial loss of morphological asymmetry during NB division, resulting in smaller than normal daughter NBs. We thus examined in greater detail the pattern of cell division in dlkb1³¹⁵/Df(3R)urd brains, and compared this pattern with those observed in wild-type pins and asl brains. Comparison between dlkb1 and pins mutants was prompted by two previous findings. First, mutations in pins partially suppress the asymmetry of NB divisions, leading to a progressive reduction in NB size(Cai et al., 2003; Parmentier et al., 2000). Second, the Drosophila and the human LKB1 kinases interact with the orthologous proteins Pins and AGS3 (GPSM1 - Human Gene Nomenclature Database),respectively (Blumer et al.,2003). In addition, more detailed comparisons between dlkb1 and asl mutants would allow us to assess more precisely the role of astral MTs in asymmetric NB divisions.

To unambiguously distinguish between NB and GMC spindles, we immunostained preparations from control and mutant brains for both tubulin and the NB marker Deadpan (Dpn) (Bier et al.,1992) (Fig. 3Abelow). The analysis of Dpn-positive cells showed that the dlkb1 and pins NBs are indeed defective in aster formation. However, the two mutants displayed different patterns of spindle defects. In brains homozygous for the pins^P62 null mutation(Yu et al., 2000), most NB prometaphases and metaphases showed normal asters but most ana-telophases were characterized by an abnormally small apical aster(Fig. 2A-D). Despite this defect in astral MTs, the density of the spindle MTs in pins NBs was comparable to that observed in their wild-type counterparts (compare Fig. 2A-D with Fig. 1B). By contrast, the spindles of dlkb1 NBs not only showed a reduction in MT density, but also displayed small asters in both metaphase and ana-telophase figures(Fig. 1B and Fig. 2E).

In pins and asl mutants, the spindles of Dpn-negative GMCs displayed a normal morphology and were indistinguishable from their wild-type counterparts (data not shown). However, in dlkb1 mutants,GMC spindles were characterized by low MT density just as were those of the NBs (Fig. 1B). Thus, the wild-type function of dlkb1 is required for proper spindle formation in both NBs and GMCs.

We next measured the spindle length of metaphase and ana-telophase figures in both NBs (Dpn-positive) and GMCs (Dpn-negative). In dlkb1 and pins mutant brains, the average sizes of NB spindles were substantially smaller than those measured in either asl or wild-type brains. This is mainly due to the absence of large NBs, as both dlkb1and pins mutants lacked NB metaphases and telophases longer than 19 and 26 μm, respectively; these large NBs represented approximately 20% of the NBs found in wild-type or asl mutant brains. By contrast, the average sizes of the GMC spindles observed in dlkb1, asl and pins mutants were very similar and comparable to those of wild-type controls (Fig. 3B,C). An explanation for these results is that in both dlkb1 and pinsmutants, NBs divide more symmetrically than in either asl or wild type. To test this possibility, we directly evaluated the degree of asymmetry of NB telophases showing strong Dpn staining. The asymmetry index was determined using the formula a-b/a+b, in which a is the long axis of the spindle and b its short axis(Fig. 3D). This analysis(Fig. 3E) clearly shows that the NBs from dlkb1 and pins mutants divide more symmetrically than those of either asl or wild type. Collectively,these results indicate that mutations in either dlkb1 or pins partially suppress the asymmetry of NB division, leading to a reduction in the NB size at each cell division cycle.

To ask whether mutations in dlkb1 and pins affect centrosome size, brain preparations were stained for Centrosomin (Cnn), an integral component of Drosophila centrosomes(Megraw et al., 2001). Observations were restricted only to those cells that, according to our analysis of spindle size distribution (Fig. 3B,C), were likely to be NBs (wild-type, dlkb1 and pins metaphases longer than 12 μm, and ana-telophases longer than 16 μm). This analysis (Fig. 4A,B) revealed that 88% (n=76) of wild-type NBs display centrosomes of different sizes at their poles. By contrast, only in 34%(n=180) of dlkb1 NBs and 39% (n=100) of pins NBs was the centrosome at the apical pole larger than that at the basal pole. These results indicate that dlkb1 and pinscontrol asymmetry in centrosome size during NB division.

We examined whether dlkb1 and pins mutations affect the distribution of Mira in dividing NBs. Larval brain preparations were simultaneously stained for both tubulin and Mira and analyzed for Mira localization (Fig. 5A). We again restricted our analysis to large mitotic figures that are likely to be NBs by the criteria employed above. Both dlkb1 and pinsmutant NBs revealed abnormal Mira distribution, but the patterns of Mira mislocalization were different (Fig. 5A,B). In wild type, 93% of NB metaphases and ana-telophases displayed a clear Mira crescent at the basal pole, whereas the remaining 7%showed diffuse Mira staining. By contrast, in dlkb1 and pinsmutants the frequencies of NBs with a basal Mira crescent were 47% and 26%,respectively. Most (97%) of the dlkb1 mutant NBs lacking a Mira crescent displayed a diffuse cytoplasmic localization of Mira. However,although the majority (67%) of pins mutant NBs without a Mira crescent had this same pattern, a substantial minority (33%) of these cells showed a diffuse cortical distribution of Mira(Fig. 5B).

We next determined whether dlkb1 mutations affect the localization of Pins and Gαi at the apical cortex of dividing NBs (NBs were again identified by their size). A regular Pins signal was observed in 95%(n=60) of the wild-type NBs and in 84% (n=145) of the dlkb1 NBs (Fig. 6A,B). Gαi formed a crescent at the apical pole of 96% of wild-type NBs(n=61) and 74% of dlkb1 mutant NBs (n=75)(Fig. 6A,B). Consistent with previous results (Cai et al.,2003; Schaefer et al.,2001), we observed a Gαi apical crescent only in 4%(n=50) of pins mutant NBs (data not shown). Thus, although dlkb1 mutations affect Mira localization at the basal cortex, they have little or no effect on Pins and Gαi localization at the apical cortex.

We also analyzed Baz, DaPKC and Par-6 localization in both wild-type and dlkb1 mutant NBs. In wild-type larval brains, the Baz signal was rather weak and only 48% (n=63) of the dividing NBs displayed a clear Baz crescent at the apical pole. However, in dlkb1 mutant brains,only five of the 106 NBs scored showed a discernible Baz crescent(Fig. 6A,B). The DaPKC and Par-6 apical crescents were observed in 96% (n=24) and 71%(n=35) of wild-type NBs, respectively, but most dlkb1 NBs did not show apical accumulations of these proteins: DaPKC and Par-6 crescents were detected only in 7% (n=30) and 12% (n=33) of dlkb1 mutant NBs. These results suggest that the wild-type function of dlkb1 is required for the localization of Baz, DaPKC and Par-6 at the NB apical pole.

Recent work has suggested that DaPKC delocalization from the apical cortex can result in NB overproliferation (Lee et al., 2006a). Consistent with this idea, the brains from third instar larvae of dlkb1-null mutants exhibit a dramatic hyperplasia of both the hemispheres and the ventral ganglion; this phenotype has been attributed to a reduction in developmental apoptosis during embryogenesis(Lee et al., 2006b). We observed a clear brain overgrowth in all our dlkb1 mutant alleles,confirming that Dlkb1 regulates Drosophila brain size (data not shown). It is likely that the brain hyperplasia elicited by dlkb1mutations results from both defective apoptosis and DaPKC-related NB overproliferation.

We examined 128 metaphases of dlkb1 mutant NBs stained for Mira;53 of them displayed a Mira crescent, but only in one case was this crescent incorrectly oriented with respect to the spindle axis(Fig. 5B). By contrast, this crescent was misoriented with respect to the spindle axis in nine of the 29 pins NB metaphases with a Mira crescent(Fig. 5B). These results confirm that Pins is required for proper spindle rotation during NB division and indicate that the Dlkb1 kinase is not involved in this process.

Recent work has shown that that spindle rotation is regulated by Mud(Mushroom body defect), a protein related to vertebrate NuMA (also known as Numa1) that interacts with both Pins and the spindle MTs. In embryonic NBs,Mud forms an apical crescent and accumulates at the spindle poles; in larval NBs, the cortical localization of Mud is weak or undetectable but the protein remains enriched at the spindle poles(Bowman et al., 2006; Izumi et al., 2006; Siller et al., 2006). Immunostaining for Mud revealed that the protein is enriched at the centrosomes and the astral MTs in 93% (n=45) of prophase and early prometaphase NBs (Fig. 7A). With progression through mitosis, Mud localization became more diffuse and 62%(n=45) of NB metaphase figures did not exhibit clear Mud accumulations at the spindle poles (Fig. 7A); however, Mud relocalized at the pericentrosomal regions of most anaphases and telophases (83%, n=30; data not shown). In dlkb1 mutant NBs, Mud accumulated at the centrosomes/asters in 91%(n=35) of prophase and early prometaphase NBs(Fig. 7C), and remained associated with the spindle poles in 78% (n=37) of the metaphases(Fig. 7D) and 85%(n=20) of the ana-telophases (not shown). Thus, mutations in dlkb1 do not affect Mud localization during metaphase and ana-telophase, but appear to increase Mud concentration at the spindle poles during metaphase.

To determine the subcellular localization of Dlkb1, we raised a guinea pig antibody against the entirety of Dlkb1. Western blot analysis showed that this antibody recognizes a band of the expected size (∼63 kDa) in larval,embryonic and S2 cell extracts. This band was absent from both dlkb1³¹⁵/Df(3R)urd and dlkb1³¹⁵/dlkb1⁷ larvae(Fig. 8A), demonstrating that it corresponds with Dlkb1. Since the truncated forms of Dlkb1 encoded by the dlkb1³¹⁵ and dlkb1⁷ mutant alleles were not observed in mutant animals, either the mutant transcripts or the truncated proteins are unstable. These findings provide strong support for the genetic data (Table 1),indicating that the dlkb1³¹⁵ mutant allele is functionally null.

Immunolocalization experiments revealed that Dlkb1 is dispersed in both the nucleus and the cytoplasm of interphase larval brain cells, and in the cytoplasm of both NBs and GMCs undergoing mitotic division. Immunostaining of dlkb1³¹⁵/Df(3R)urd mutant cells did not reveal any clear cytoplasmic signal, confirming the specificity of the antibody(Fig. 8B and data not shown). The diffuse localization of Dlkb1 in brain cells is not consistent with its cortical localization in Drosophila oocytes(Martin and St Johnston,2003). However, the Dlkb1 localization pattern in brain cells does not reflect the quality of our antibody, as the same antibody revealed a cortical accumulation of Dlkb1 in oocytes (data not shown).

The finding that Dlkb1 and Pins co-precipitate(Blumer et al., 2003), and that dlkb1 and pins mutations cause similar (but not identical)phenotypes, prompted us to perform an epistasis analysis. We thus compared the phenotype of the dlkb1³¹⁵ pins^P62 double mutant with those of the single mutants by examining brain preparations stained for tubulin, Dpn and DNA. In dlkb1 pins mutant brains, the spindles of both NBs and GMCs were much more defective than those observed in either of the single mutants (compare Fig. 9 with Figs 1 and 2). In addition to cells with severely defective spindles (Fig. 9A,C,F,G,I), we also observed many (50%, n=300) mitotic figures in which the spindle morphology was barely recognizable(Fig. 9B,D,E,H); the frequency of the latter type of cells was only 4% (n=201) in the dlkb1single mutant. In dlkb1 pins metaphases, the MT density was extremely low, the spindle poles had a characteristic pointed appearance and the asters were completely absent (Fig. 9A). The ana-telophases were also devoid of asters and displayed few and sparse central spindle MTs, which were never pinched in the middle(Fig. 9C,F). These results indicate that dlkb1 and pins function in parallel pathways to control spindle formation.

The absence of central spindle pinching, which suggests an accompanying failure of cytokinesis, prevented a reliable assessment of the degree of asymmetry of NB divisions. However, NB spindles of the double mutant were smaller than in wild type (Fig. 9J). In addition, the analysis of centrosome size in large metaphase figures (longer than 14 μm), most of which are likely to be NBs,revealed that 90% (n=106) of them had equally sized centrosomes. In wild type, dlkb1 and pins, the frequencies of NB metaphases with centrosomes of equal size were 12%, 72% and 61%, respectively(Fig. 4B). Finally, only 3%(n=120) of the dlkb1 pins NB metaphases were characterized by a Mira crescent; in the remaining cells, Mira was either diffuse in the cytoplasm (85%) or associated with the entire cell cortex (12%). Thus, the Mira mislocalization phenotype observed in dlkb1 pins double mutants is stronger than that seen in the single mutants (see Fig. 5B).

Although the dlkb1³¹⁵ and pins^P62alleles are both functionally null, it cannot be excluded that the brains of the dlkb1³¹⁵ pins^P62 double mutants retain residual amounts of the maternally supplied Dlkb1 and Pins proteins. We thus performed a western blotting analysis of extracts from third instar larval brains of dlkb1³¹⁵, pins^P62 and dlkb1³¹⁵ pins^P62 mutants. As shown in Fig. 9K, dlkb1³¹⁵ and dlkb1³¹⁵pins^P62 brains did not exhibit detectable amounts of the Dlkb1 kinase, consistent with the results shown in Fig. 8A. Similarly, the Pins protein appeared to be completely absent from pins^P62 and dlkb1³¹⁵ pins^P62 brains. Thus, the phenotypes observed in the single and double mutants reflect a complete loss of the wild-type function of either Dlkb1 or Pins, or both.

Collectively, our results suggest that dlkb1 and pins act in different pathways to control the asymmetry of NB division. These genes also function in parallel pathways involved in MT stability and spindle formation. Whether the latter pathways are the same as those that control the asymmetry of NB mitosis remains to be determined.

Our results indicate that mutations in the dlkb1 gene disrupt spindle formation in both NBs and GMCs. In addition, the finding that the imaginal discs of dlkb1 mutant larvae are small and misshapen suggests a defect in imaginal cell mitosis. Previous studies have shown that late larval lethality and small imaginal discs are diagnostic of abnormalities in mitotic divisions (Gatti and Baker,1989). Thus, the dlkb1 phenotype strongly suggests that the Dlkb1 kinase plays an important mitotic role not only in NBs, but also in other somatic cell types. Despite the low density of spindle MTs, most mutant metaphases enter anaphase (the frequency of anaphases in dlkb1mutants and in wild-type controls was 10-13% and 16%, respectively; see Table 1), suggesting that in a substantial fraction of mutant cells the spindle checkpoint is either not induced or only transiently activated. However, the defects in NB spindles are likely to lead to the formation of polyploid cells. These cells could arise through two different mechanisms. Cells blocked in metaphase owing to either reduced MT density or activation of the spindle checkpoint could revert to interphase and become polyploid after an additional round of DNA replication. Alternatively, cells that enter anaphase but assemble an abnormally thin central spindle might be unable to undergo cytokinesis and in consequence produce polyploid cells (Vernì et al., 2004).

The precise function of Dlkb1 in spindle formation and/or maintenance is currently unclear. However, the finding that the spindles of dlkb1 asl double mutants display a lower MT density than asl single mutants argues for a defect in MT stability and not in centrosome function. Studies in mammalian cells have shown that LKB1 is a master kinase that phosphorylates at least 14 kinases, all of which are related to AMP-activated kinases (AMPK). Kinases of the AMPK family include regulators of cellular energy levels, as well as four Microtubule affinity regulating kinases (MARKs)(reviewed by Alessi et al.,2006; Baas et al.,2004b). The MARK enzymes are the mammalian homologs of C. elegans and Drosophila Par-1. However, DrosophilaPar-1, which controls MT stability in oocytes(Shulman et al., 2000),appears to act upstream of Dlkb1 (Martin and St Johnston, 2003). It is therefore unlikely that the Dlkb1 substrate required for the stability of spindle MTs is Par-1. Further evidence that Dlkb1 does not act via Par-1 phosphorylation comes from RNAi experiments showing that Dlkb1 depletion, but not Par-1 depletion, causes defects in spindle morphology (Bettencourt-Dias et al., 2004). Thus, current data indicate that the Dlkb1 kinase regulates the activity of an unknown factor required for the stability of the spindle MTs; this factor could either be a direct substrate of Dlkb1 or a substrate for one of the kinases acting downstream of Dlkb1.

Our cytological analyses have shown that the spindles of dlkb1 pins double mutants display a MT density that is substantially lower than that observed in either single mutant. Here again, highly defective spindles were observed in both NBs and GMCs. In addition, double-mutant larvae showed extremely reduced imaginal discs, suggesting an underlying mitotic defect. Given that the dlkb1³¹⁵ and pins^P62alleles used in the analysis are both functionally null and that the corresponding proteins were undetectable in mutant brains(Yu et al., 2000) (this study), these results indicate that Pins and Dlkb1 function in different pathways for the control of MT stability. The observation that the spindles of pins mutants display a normal MT density further suggests that Pins plays a redundant role in the maintenance of MT stability. A role for Pins in spindle formation and/or stability has never been demonstrated in Drosophila. However, the mammalian homolog of Pins binds NuMA and regulates mitotic spindle organization and positioning(Du et al., 2001).

We have analyzed the phenotypic consequences of dlkb1 mutations in larval brain NBs. In contrast to embryonic NBs that display small, regularly sized spindles (their metaphase spindles are approximately 5 μm long),brain NBs exhibit spindles of very different sizes (ranging from 5 to 32 μm for metaphase spindles). Nonetheless, dividing brain NBs exhibit the same asymmetries as their embryonic counterparts, including asymmetries in aster and centrosome size, localization of specialized protein complexes and positioning of the cleavage furrow (Bowman et al., 2006; Giansanti et al., 2001; Lee et al.,2006a; Parmentier et al.,2000; Rolls et al.,2003; Siller et al.,2006) (this study). However, the degree of asymmetry of brain NB division is directly related to the cell size, so that large NBs divide more asymmetrically than small NBs (Fig. 3E). This is likely to render large brain NBs particularly sensitive to mutations that affect cleavage furrow positioning. Consistent with this hypothesis, mutations in pins have mild effects on the asymmetry of embryonic NB divisions (Cai et al., 2003), but disrupt unequal cytokinesis in most larval brain NBs (Parmentier et al.,2000) (Fig. 3E).

dlkb1 larval NBs also divide more symmetrically than their wild-type counterparts, leading to larval brains devoid of large NBs. In addition, most dlkb1 NBs display centrosomes of equal size and very small asters at both poles. However, the symmetric cytokinesis of dlkb1 NBs cannot result from their short astral MTs, as aslNBs divide asymmetrically in the complete absence of asters(Fig. 3E). dlkb1mutant NBs are also characterized by the abnormal distribution of several components of the apical and basal complexes. In dlkb1 mutant brains,most NBs display normal Pins and Gαi crescents at their apical pole but fail to accumulate Baz, DaPKC and Par-6 at the same pole. In addition, most dlkb1 mutant NBs fail to exhibit a normal Mira crescent at the basal pole cortex. A normal localization of Pins and Gαi has been observed in most embryonic NBs defective in the Baz/Par-6 pathway(Cai et al., 2003; Izumi et al., 2004; Schaefer et al., 2000; Yu et al., 2000; Yu et al., 2003). Moreover,studies on embryonic NBs have suggested that Baz, Par-6 and DaPKC function as a complex, are interdependent for their localization at the NB apical pole,and are required for the formation of the Mira crescent at the basal pole(Petronczki and Knoblich,2000; Wodarz et al.,2000). However, subsequent work on second instar larval NBs has shown that these proteins are not mutually dependent for the formation of the Baz/Par-6/DaPKC apical crescent; they accumulate at the apical cortex in a hierarchical fashion, with Baz and Par-6 mediating proper DaPKC localization(Rolls et al., 2003). Mutations that disrupt the Pins/Gαi pathway prevent asymmetrical localization of either Pins or Gαi in embryonic NBs but do not substantially affect Mira accumulation at the basal pole(Cai et al., 2003). However,it should be noted that mutations in pins partially disrupt asymmetric Mira localization in larval brain NBs(Parmentier et al., 2000)(this study), suggesting that larval NBs differ from embryonic NBs in some aspects of the control of Mira localization. Thus, taking into account the differences between embryonic and larval NBs, our results indicate that mutations in the dlkb1 gene and those that disrupt the Baz/Par-6 pathway affect similar aspects of NB mitotic division.

Our analyses have shown that in dlkb1 pins double mutants, the NBs divide more symmetrically than in the corresponding single mutants. This indicates that the dlkb1 and pins genes act in different pathways that mediate unequal cytokinesis. Previous studies have shown that the asymmetry of NB cytokinesis depends on the Baz/Par-6 and Pins/Gαi redundant pathways. When only one of these pathways is impaired, NBs still divide asymmetrically, but they divide symmetrically when both are disrupted(Cai et al., 2003). The simplest interpretation of our findings is that dlkb1 acts in the Baz/Par-6 pathway. In addition, the observation that Dlkb1 is required for proper localization of Baz, Par-6 and DaPKC suggest that this kinase acts at the top of the hierarchical mechanism that mediates accumulation of the Baz/Par-6 complex at the apical cortex. However, although we favor the hypothesis that Dlkb1 acts in the Baz/Par-6 pathway, we cannot exclude the possibility that this kinase functions in both the Baz/Par-6 and Pins/Gαi pathways, or in a third pathway different from either.

In this context, it is important to note that our results exclude the possibility that dlkb1 acts via Pins phosphorylation. Previous studies have shown that mammalian LKB1 co-precipitates and phosphorylates AGS3, the mammalian ortholog of Pins(Blumer et al., 2003). Dlkb1 and Pins coimmunoprecipitate as well, but it is currently unclear whether Pins is phosphorylated by Dlkb1 (Blumer et al.,2003). Regardless of whether Pins is a substrate of Dlkb1, the phenotypes elicited by dlkb1 mutations cannot be the consequence of an impairment of Pins function. dlkb1 and pins mutant NBs do in fact differ in a number of phenotypic traits, including spindle organization and the pattern of Mira localization, and do not belong to the same epistasis group.

In vivo imaging has shown that the spindles of embryonic NBs rotate during metaphase to become aligned with the center of the Pins apical crescent(Kaltschmidt et al., 2000). By contrast, the spindles of larval NBs align with the Pins crescent at prophase(Siller et al., 2006). Failure of proper rotation of larval NB spindles results in spindles that are misoriented with respect to the apical (Pins) and basal (Mira) crescents(Giansanti et al., 2001; Siller et al., 2006). There is also evidence that proper positioning of larval NB spindles depends on astral MTs, because in approximately 50% of asl NB metaphases the Mira crescent is misoriented with respect to the spindle axis(Giansanti et al., 2001).

Our results indicate that spindle rotation occurs normally in dlkb1 mutant NBs. In addition, we have shown that prophase/prometaphase larval NBs of dlkb1 mutants normally accumulate the Mud protein, which mediates proper spindle alignment in both embryonic and larval NBs (Bowman et al.,2006; Izumi et al.,2006; Siller et al.,2006). Together, these results indicate that the Dlkb1 kinase is not required for spindle rotation and that the short astral MTs of dlkb1 mutant NBs can mediate proper spindle positioning. These results are consistent with the idea that the Pins/Gαi, but not the Baz/Par-6, pathway is involved in spindle rotation(Izumi et al., 2004; Siegrist and Doe, 2005) and provide further support for the hypothesis that Dlkb1 functions in the latter pathway.

Recent work has shown that in the absence of the Baz/Par-6 pathway, astral MTs can mediate the localization of Pins/Gαi at the apical cortex(Siegrist and Doe, 2005). Assuming that Dlkb1 acts in the Baz/Par-6 pathway, the finding that this kinase is not required for the formation of Pins/Gαi crescents indicates that the short astral MTs of dlkb1 NBs retain the ability to mediate Pins/Gαi cortical localization.

Our results indicate that Dlkb1 and Pins function in partially redundant pathways controlling the stability of spindle MTs. These proteins are also required for the asymmetry of NB divisions and, here again, they appear to function in different pathways. Pins acts in a common pathway with Gαi,whereas Dlkb1 is likely to function in the Baz/Par-6 pathway. Intriguingly,recent work has shown that simultaneous loss of pins and bazfunctions results in the formation of abnormally small embryonic NB spindles that lack astral MTs at both poles (Fuse et al., 2003). Thus, the embryonic NBs of baz pins double mutants have a spindle phenotype reminiscent of that observed in dlkb1 larval NBs. These findings raise the question of whether the Pins/Gαi and Baz/Par-6 pathways redundantly control spindle organization as they do for the asymmetry of NB divisions. The extant data do not provide a clear answer to this question. The analysis of the roles of the two pathways in spindle formation and their precise relationships with the Dlkb1 kinase are interesting issues to be addressed in future studies.

Previous studies in Drosophila and mammalian cells have led to the suggestion that loss of epithelial cell polarity is ultimately responsible for the Peutz-Jeghers cancer syndrome (Martin and St Johnston, 2003; Baas et al., 2004b). Here, we have shown that Dlkb1 plays an essential mitotic role and is required for the asymmetry of NB division. These results lead us to propose that tumor development in Peutz-Jeghers patients depends on the impairment of multiple processes, including cell polarity, the asymmetry of stem cell division and the fidelity of chromosome segregation during mitosis.

We thank W. Chia, Y. Jan, T. Kaufman, J. Knoblich and F. Matsuzaki for antibodies and fly stocks. This work was supported by grants from Centro di Eccellenza di Biologia e Medicina Molecolare (BEMM) to M.G. and by NIH grant GM48430 to M.L.G.