Asymmetric Segregation of the Tumor Suppressor Brat Regulates Self-Renewal in Drosophila Neural Stem Cells

Asymmetric Segregation of the Tumor Suppressor Brat Regulates Self-Renewal in Drosophila Neural Stem Cells Joerg Betschinger,1,2 Karl Mechtler,1,2 and Juergen A. Knoblich1,* 1

Institute of Molecular Biotechnology of the Austrian Academy of Sciences (IMBA), Dr Bohr Gasse 3-5, 1030 Vienna, Austria Research Institute of Molecular Pathology (IMP), Dr Bohr Gasse 7, 1030 Vienna, Austria *Contact: [email protected] DOI 10.1016/j.cell.2006.01.038 2

SUMMARY

How stem cells generate both differentiating and self-renewing daughter cells is unclear. Here, we show that Drosophila larval neuroblasts—stem cell-like precursors of the adult brain—regulate proliferation by segregating the growth inhibitor Brat and the transcription factor Prospero into only one daughter cell. Like Prospero, Brat binds and cosegregates with the adaptor protein Miranda. In larval neuroblasts, both Brat and Prospero are required to inhibit self-renewal in one of the two daughter cells. While Prospero regulates cell-cycle gene transcription, Brat acts as a posttranscriptional inhibitor of dMyc. In brat or prospero mutants, both daughter cells grow and behave like neuroblasts leading to the formation of larval brain tumors. Similar defects are seen in lethal giant larvae (lgl) mutants where Brat and Prospero are not asymmetric. We have identified a molecular mechanism that may control self-renewal and prevent tumor formation in other stem cells as well.

INTRODUCTION The recent identification of tumor stem cells has challenged our view of tumorigenesis (Al-Hajj and Clarke, 2004). It is now thought that individual tumor cells vary considerably in their ability to generate the other cell types and only very few stem cells exist that can recreate the whole tumor mass. Conventional therapies targeted at destroying rapidly proliferating cells may not always hit these stem cells, and this could explain why tumors often relapse even after a dramatic reduction of tumor mass (Michor et al., 2005). Understanding how proliferation and self-renewal are regulated in stem cells is therefore an important prerequisite for developing new therapeutic strategies.

The development of the Drosophila central nervous system (CNS) has been well described during embryogenesis (Betschinger and Knoblich, 2004) and can serve as a model for certain aspects of stem cell biology. CNS neurons arise from neuroblasts that divide asymmetrically into a larger cell that retains neuroblast characteristics and continues to divide in a stem cell-like fashion and a smaller so-called ganglion mother cell (GMC) that divides only once more into two differentiating neurons. During each neuroblast division, the endocytic protein Numb (Knoblich et al., 1995), the transcription factor Prospero (Hirata et al., 1995; Knoblich et al., 1995), and the RNA binding protein Staufen (Broadus et al., 1998; Schuldt et al., 1998; Shen et al., 1998) localize asymmetrically to the basal cell cortex and segregate into the GMC. While Staufen is dispensible for neuronal specification, Numb was shown to act as a cell-fate determinant in other tissues (Rhyu et al., 1994), and Prospero controls GMC-specific genes in some neuroblast lineages (Doe et al., 1991; Vaessin et al., 1991). However, neither numb nor prospero mutants show a transformation of GMCs into neuroblasts, suggesting that other determinants exist. Asymmetric segregation of Staufen and Prospero requires the protein Miranda (Broadus et al., 1998; Ikeshima-Kataoka et al., 1997; Matsuzaki et al., 1998; Schuldt et al., 1998; Shen et al., 1997, 1998). Miranda associates and colocalizes with Prospero and Staufen and acts as an adaptor that is essential for their localization to the plasma membrane and their asymmetric segregation into only one daughter cell. Asymmetric protein segregation into the GMC requires a conserved protein complex consisting of Par-3 (Schober et al., 1999; Wodarz et al., 1999), Par-6 (Petronczki and Knoblich, 2001), and the protein kinase aPKC (Rolls et al., 2003; Wodarz et al., 2000). These proteins localize apically before and during neuroblast division and are essential for determinant localization in neuroblasts. The key substrate for aPKC is the cytoskeletal protein Lgl (Lethal (2) giant larvae) (Betschinger et al., 2003). Lgl is required for the cortical localization of Miranda and its binding partners (Ohshiro et al., 2000; Peng et al., 2000), but phosphorylation inactivates the protein by inducing an intramolecular association (Betschinger et al., 2005). It is thought that Miranda localizes asymmetrically because aPKC phosphorylation

Cell 124, 1241–1253, March 24, 2006 ª2006 Elsevier Inc. 1241

on one side of the neuroblast restricts Lgl activity—and consequently cortical recruitment of Miranda—to the opposite side. Drosophila embryonic neuroblasts divide no more than 12 times (Bossing et al., 1996). Since there is no significant cell growth throughout embryogenesis, they shrink with each division and their proliferation is limited by their decreasing size (Fuse et al., 2003). This is different during larval development where a small number of neuroblasts gives rise to thousands of neurons in the adult fly brain. Larval neuroblasts seem to use the same molecular machinery that regulates asymmetric cell division during embryogenesis (Akong et al., 2002; Ceron et al., 2001; Rolls et al., 2003). Larval brain development is accompanied by a dramatic increase in tissue size, suggesting that neuroblasts are able to grow between each asymmetric division. How cell growth in the nervous system is regulated is completely unclear. Several mutants have been identified which lead to overproliferation and formation of enlarged tumor-like brains (Gateff, 1978). Remarkably, several of these mutants affect genes like lgl, which have been implicated in asymmetric cell division, suggesting that growth regulation might involve the asymmetric segregation of proliferation control genes. We have identified the growth regulator Brat as an asymmetrically segregating determinant that cooperates with the transcription factor Prospero to establish GMC fate in the embryonic nervous system and controls proliferation in the larval brain. Like lgl mutants, brat mutants show overgrowth and neoplastic transformations in the larval brain. Our data suggest that this is due to a transformation of GMCs into additional neuroblasts resulting in exponential overproliferation of neural precursor cells. We show that Brat might be a posttranscriptional regulator of dMyc and thereby control ribosome biogenesis and protein translation. We propose that the asymmetric segregation of growth regulators might be a general mechanism by which stem cells regulate self-renewal and control the balance between proliferating and differentiating daughter cells. RESULTS Brat Is a Binding Partner of Miranda To identify additional proteins that segregate into the GMC during neuroblast division, we searched biochemically for binding partners of Miranda. Miranda acts as an adaptor that binds to Staufen and Prospero via a C-terminal domain (Fuerstenberg et al., 1998; Matsuzaki et al., 1998; Schuldt et al., 1998; Shen et al., 1998) and is required for their segregation into the GMC. TAP (tandem affinity purification)-tagged Miranda deletion constructs (Figure 1A) were expressed in transgenic flies and isolated from protein extracts on an IgG column via the protein A part of the tag. Specific and nonspecific interaction partners were eluted by low pH and analyzed by LC-MS/MS (see Experimental Procedures). While hundreds of interaction partners were identified with each construct, only four proteins could reproducibly be isolated with the

1242 Cell 124, 1241–1253, March 24, 2006 ª2006 Elsevier Inc.

C-terminal domain but were never present in any of the other isolates. Besides Staufen and Prospero, we identified Brat and CG17593 as novel specific Miranda interaction partners. We generated transgenic flies expressing N-terminally GFP-tagged versions of both proteins. While GFP-CG17593 is cytoplasmic (Figure 1B), GFP-Brat colocalizes with Prospero at the basal cell cortex of neuroblasts (Figure 1C) and was therefore chosen for further analysis. Brat Is a Segregating Cell-Fate Determinant To test the expression and subcellular localization of Brat, we raised antibodies against an N-terminal peptide of the protein. Brat is a uniformly expressed cytoplasmic protein. In dividing embryonic neuroblasts, Brat colocalizes with Miranda throughout mitosis: it accumulates at the apical cell cortex in prophase (Figure 1D), forms a basal crescent in metaphase (Figure 1E), and segregates into the GMC. It is cortical in telophase (Figure 1F) but becomes cytoplasmic in interphase after Miranda is degraded (Figure 1F, open arrowhead). In prospero mutants, Brat localization is unaffected (Figure 1J) but in miranda mutant neuroblasts, Brat is cytoplasmic throughout mitosis (Figure 1K). Conversely, neither Prospero nor Miranda localization is affected in embryos lacking both maternal and zygotic brat function (data not shown). Asymmetric segregation of Brat is also observed in dividing sensory organ precursor (SOP) cells where Miranda is expressed but without a described function (Figures 1G–1I). Thus, Brat is the third protein besides Prospero and Staufen that uses Miranda as an adaptor to segregate into one of the two daughter cells during asymmetric cell division. To characterize Brat function, we studied nervous system development in brat mutant embryos. We chose to analyze a subset of neurons that can be identified by even-skipped (Eve) expression and that has been studied in mutants affecting asymmetric cell division before (Doe et al., 1991; Ikeshima-Kataoka et al., 1997). In wild-type embryos, the EL (Figure 2A, arrowhead), RP2 (Figure 2A, open arrowhead), aCC/pCC, and U/CQ neurons express Eve. In prospero mutants, most of the Eve-expressing RP2, aCC/pCC, and U/CQ neurons are missing but the EL cluster is largely unaffected (Doe et al., 1991 and Figure 2C). Zygotic brat mutant embryos have no obvious CNS phenotype (Figure 2B and data not shown), but in embryos lacking both maternal and zygotic brat function, RP2 neurons are frequently missing (Figure 2E, open arrowhead). This analysis is however complicated by the strong morphological defects that are observed in embryos lacking maternal and zygotic brat due to a role in translation of the gap gene hunchback (Sonoda and Wharton, 2001). Since overexpression of GFP-Brat causes an increase in the numbers of RP2 neurons (marked by the expression of Eve and Zfh-1 in Figure 2F), we believe that Brat acts as a cell-fate determinant in the embryonic CNS. This is further supported by a strong genetic interaction between brat and prospero: In embryos lacking zygotic prospero and zygotic brat function, most

Figure 1. Identification of Miranda Binding Proteins (A) Schematic representation of Miranda deletion constructs used for the identification of interacting protein (green: asymmetry domain; red: cargo domain). (B and C) GFP-Brat (C), but not GFP-CG17593 (B) colocalizes asymmetrically with Prospero in metaphase embryonic neuroblasts. (D–I) Brat localizes to the apical cell cortex of embryonic neuroblasts in prophase (D, arrowhead), localizes basally in metaphase (E, arrowhead), segregates into the basal GMC in telophase (F, arrowhead), and is enriched in the cytoplasm of GMCs (F, open arrowhead). In metaphase SOP cells, Brat is localized to the anterior cell cortex (G), segregates asymmetrically during telophase (H), and is enriched in the anterior pIIb daughter cell after cytokinesis (I). (J and K) Brat localizes asymmetrically in pros17 (J), but not miraL44 (K) mutant neuroblasts. Scale bars are 10 mm.

Eve-expressing neurons are missing, leading to dramatic defects and a severe reduction of axonal marker staining in the embryonic CNS (Figure 2D). Thus, brat and prospero are partially redundant in specifying cell fates in embryonic neurons. Brat Controls Proliferation and GMC Fate in the Larval CNS Brat is a member of a family of evolutionarily conserved tumor suppressor proteins (Arama et al., 2000). Besides Brat, this family contains the Drosophila proteins Dappled and Mei-P26 and the mouse proteins TRIM2, TRIM3, and TRIM32. While dappled mutants develop tumors in the fat body (Rodriguez et al., 1996) and mei-P26 mutants have ovarian tumors (Page et al., 2000), brat mutants show

brain overgrowth (Gateff, 1978). We therefore characterized brat mutant larval brains by staining with neuroblast and cell-cycle markers. Drosophila larval neuroblasts can be identified by staining for Miranda (Mira) and by the expression of membrane bound CD8-GFP in neuroblasts and all their progeny from the Gal4 line 1407 inserted in the inscuteable promotor. Neuroblasts are present in the ventral nerve cord, the optic lobes, and in the medial areas of the two brain lobes where they are called central brain (CB) neuroblasts (Figure 3A). In the anterior half of the brain (Figure 3B), CB neuroblasts seem to be continuous with neuroblasts in the ventral nerve cord and generate less than a hundred neurons with some lineages projecting into the ventral ganglion. On the posterior side (Figure 3C), CB neuroblasts can generate several

Cell 124, 1241–1253, March 24, 2006 ª2006 Elsevier Inc. 1243

Figure 2. Brat Acts as a Cell-Fate Determinant in the Embryonic Central Nervous System Merged confocal z stacks of stage 15 embryos. (A–D) Eve (arrowheads mark EL neuron cluster, open arrowheads RP2 neurons) and BP102 expression is normal in brat150 embryos (B). Although CNS morphology is severly impaired in pros17 mutants, the EL neuron cluster is unaffected (C). In pros17,brat150 embryos, the EL neurons are mostly missing and axon formation is greatly alleviated (D). (E) brat150 maternal/zygotic mutant embryos show loss of RP2 neurons (open arrowhead), while overexpression of GFP-Brat using maternalGal4 increases the number of Eve- and Zfh1-expressing RP2 neurons (open arrowhead) from two per segment to three (29.7% of n = 164 segments) or four (7.1%). Scale bars are 20 mm.

hundred neurons which project ipsi- and contralaterally within the brain. Like in embryonic neuroblasts, Brat localizes asymmetrically and segregates into the GMC during larval neuroblast division (see Figures 5A and 5U). During early third instar, the number of Miranda-positive neuroblasts in brat mutant brains is comparable to wild-type (Figures 3D and 3E). During mid-third instar, however, the number of CB neuroblasts in the posterior half of the brain is dramatically increased (Figures 3F and 3G). By late-third instar, BrdU-incorporating neuroblasts overgrow essentially the whole brain (Figures 3H–3M). We reproducibly detect a region that is free of Miranda-expressing cells and seems to correspond to the position of the proliferation centers of the optic lobes (Figure 3J, arrowhead). Double staining with Prospero (Pros), a marker for differentiating cells in the larval CNS, reveals that neuroblast overproliferation is at the expense of GMCs and neurons (Figures 3H–3K). The excess neuroblasts observed in brat mutants are positive for Cyclin E (CycE) (Figures 3N and 3O) and Worniu (Wor) (Figures 3P and 3Q). Individual Neuroblasts Overproliferate in brat Mutant Brains The brat mutant phenotype could be due to specification of more neuroblasts or a defect in the neuroblast lineage. To distinguish these possibilities, we labeled individual wild-type or brat mutant neuroblasts using the MARCM system (Lee and Luo, 1999). This method allows the generation of individual homozygous mutant neuroblast clones which express membrane bound CD8-GFP in an otherwise heterozygous, GFP-negative background. We focused our analysis on the posterior CB neuroblasts which seem to be responsible for the overgrowth pheno-

1244 Cell 124, 1241–1253, March 24, 2006 ª2006 Elsevier Inc.

type in brat mutants. Control clones always contain one large Miranda (Figure 4A0 ), Worniu (Figure 4C00 ), and Cyclin E (Figure 4C0 ) positive neuroblast and small Prosperoexpressing daughter cells that extend one axon bundle (Figure 4A00 , open arrowheads and data not shown). Miranda, Worniu, and Cyclin E expression is downregulated in the small daughter cells (Figures 4A0 , 4C0 , and 4C00 ), although some remnants can be found in the most recently generated GMCs (open arrowheads). In comparison to wild-type clones (172 cells per clone of n = 14 clones), brat mutant clones are dramatically enlarged (707 cells per clone of n = 6 clones) (Figure 4E). Most cells in brat clones coexpress Miranda, Worniu, and Cyclin E (Figures 4B and 4D), and the mitotic index is significantly increased compared to wild-type clones (0.96 of n = 14 wild-type clones compared to 3.44 of n = 6 brat clones) (Figure 4E). Prospero is not expressed and no axons extend from any of the daughter cells (Figure 4B00 and data not shown). While control clones always have one large (>8 mm) and many small cells, cell size is more variable in brat mutant clones and more cells are larger than 8 mm in diameter (Figure 4F). Thus, in wild-type neuroblasts, the daughter cell that inherits Brat downregulates Miranda and Worniu, exits the cell cycle, stops cell growth, and differentiates into a neuron. In the absence of Brat, however, this cell expresses neuroblast markers, grows, and continues the cell cycle leading to exponential proliferation and tumor formation. Aberrant Growth Control in Mutants Affecting Brat Asymmetry To test whether Brat acts as a segregating cell growth inhibitor, we analyzed brain development in mutants affecting the asymmetric localization of Brat. Like Miranda, Brat

Figure 3. Loss of Brat Leads to Overgrowth of Neuroblasts (A) Schematic view of the larval central nervous system. (B and C) Confocal slice of the anterior (B; green slice in A) and posterior (C; red slice in A) region of a 1407-Gal4 larval brain lobe expressing CD8-GFP. Insets show magnified regions of the anterior (B0 and B0 0 ) and posterior (C0 and C0 0 ) brain (neuroblasts marked by arrowheads). (D–G) Miranda-expressing neuroblasts overgrow in the posterior brain region in brat mutants during mid (G) but not early (E) third instar stages. (H–M) In contrast to control late third instar larval brains (I and L), BrdU-incorporating (M) and Miranda-expressing neuroblasts (K) overgrow the posterior region in brat mutant brains and invade into the anterior brain cortex (H and J). Neuroblast overgrowth is to the expense of Prospero-expressing neurons (red in H–K). (N–Q) Overgrowing cells express Cyclin E (O) and Worniu (Q). Genotypes are brat150/brat192 (E, G, J, K, M, O, and Q). Scale bars are 50 mm (B–M) and 10 mm (N–Q).

localization requires the cytoskeletal protein Lgl (Figures 5A and 5B). lgl was originally identified as a mutation that leads to neoplastic transformation and overgrowth of larval tissues (Hadorn, 1938). In lgl mutants, a visible increase in brain size can be detected at the end of larval development. It has been proposed that this is due to an overproliferation of larval neuroblasts (Gateff, 1978). Indeed, while neuroblast number and position are unchanged in lgl mutants during early third instar larval stages (data not shown), we detect a significant increase in the number of posterior CB neuroblasts during mid-third instar (Figures 5C and 5D). During late-third instar, the number of posterior CB neuroblasts is dramatically increased (Figures 5G and 5H) while anterior CB neuro-

blasts still appear approximately normal in number and spacing (Figures 5E and 5F). Overgrowing cells incorporate BrdU (Figures 5I and 5J) and almost all cells in the posterior brain express the neuroblast markers Miranda and Worniu (Figures 5O and 5P) and are positive for the cell-cycle marker Cyclin E (Figures 5M and 5N). Neuroblast overproliferation in lgl mutants is at the expense of differentiating cells since the number of Prosperoexpressing cells is strongly reduced (Figures 5K and 5L). Thus, lgl mutants are remarkably similar to brat mutants, suggesting that Brat needs to segregate into one of the two daughter cells to act as a growth inhibitor. Consistent with this, we also detect neuroblast overgrowth upon inactivation of Lgl by neuroblast-specific overexpression

Cell 124, 1241–1253, March 24, 2006 ª2006 Elsevier Inc. 1245

Figure 4. Brat Inhibits Neuroblast Fate, Cell Cycle, and Cell Growth in Daughter Cells (A–D) CD8-GFP-expressing single-cell neuroblast MARCM clones. In control clones, the neuroblast (arrowhead) and only some recently generated small daughter cells (open arrowheads) express Miranda (A0 ), Cyclin E (C0 ), and Worniu (C0 0 ) while all daughter cells show nuclear Prospero (A0 0 ). In brat clones, all daughter cells express Miranda (B0 ), Cyclin E (D0 ), and Worniu (D0 0 ) but are negative for Prospero (B0 0 ). (E) Four days after induction, brat192 clones contain more cells (left) and the percentage of mitotic cells scored by expression of phosphoH3 is increased (right) as compared to control clones. (F) brat150 cells grow to neuroblast-like sizes 3 days after clone induction. Genotypes are hsFlp, C155-Gal4, UAS-CD8-GFP; FRT40A/FRT40A (A and C); hsFlp, C155-Gal4, UAS-CD8-GFP; FRT40A, brat150/FRT40A, brat150 (B); hsFlp, C155-Gal4, UAS-CD8-GFP; FRT40A, brat192/FRT40A, brat192 (D). Scale bars are 10 mm. **: p < 0.05 (Student’s t test); error bars are standard deviation of the mean.

of constitutively active aPKC, a kinase that phosphorylates and inactivates Lgl (Figures 5Q–5T). Furthermore, depletion of Miranda by transgenic RNAi in neuroblasts

1246 Cell 124, 1241–1253, March 24, 2006 ª2006 Elsevier Inc.

causes an increase in proliferating neuroblasts (Figures 5U–5X and data not shown). However, no brain tumors were detected in lgl mutant clones. Since asymmetric

Figure 5. Loss of Lgl Leads to Overgrowth of Neuroblasts (A and B) Brat and Miranda are localized to the cytoplasm of mitotic lgl mutant neuroblasts. (C–J) Loss of Lgl leads to overgrowth of Miranda-expressing cells in the posterior (H) but not the anterior (F) brain region of late third instar larvae. Overgrowth becomes already apparent during mid-third instar stages (D). (I–P) Overgrowing cells incorporate BrdU (J), fail to generate differentiating Prospero-positive daughter cells (L), and express Cyclin E (N) and Worniu (P). (Q–T) Overexpression of aPKCDN using 1407-Gal4 delocalizes Miranda into the cytoplasm of mitotic neuroblasts (R) and leads to a mild overgrowth of neuroblasts in the posterior brain cortex (T). (U–X) Expression of Mira RNAi delocalizes Brat into the cytoplasm of mitotic neuroblasts (V) and leads to overgrowth of Cyclin E-positive cells in the posterior brain region (X). Genotypes are lgl1/lgl1 (F, H, and L) and lgl1/lgl4 (B, D, J, N, and P). Scale bars are 50 mm (C–J) and 10 mm (A, B, K–X).

Miranda localization is unaffected in these clones (data not shown), we assume that the lack of a clonal lgl phenotype is due to perdurance of Lgl protein. Brat Regulates Nucleolar Size in Neuroblasts How does Brat suppress cell growth in GMCs and neurons? The best studied signaling pathway that controls growth and proliferation in Drosophila involves the insulin receptor (InsR), its downstream target PI3K (phosphatidylinositol-3 kinase), and the PI3K inhibitor PTEN (Goberd-

han et al., 1999; Leevers et al., 1996). Since activation of this pathway leads to cytoplasmic retention of the transcription factor dFOXO (Puig et al., 2003), we used antidFOXO staining to test the InsR pathway in wild-type and brat mutant brains. Both in neuroblasts and GMCs, dFOXO is found mostly in the cytoplasm and its subcellular localization is unchanged in brat mutant neuroblasts (Figures 6A and 6B). Furthermore, overexpression of wild-type or dominant-negative Dp110, the catalytic subunit of PI3K (Leevers et al., 1996), did not result in

Cell 124, 1241–1253, March 24, 2006 ª2006 Elsevier Inc. 1247

Figure 6. Brat Inhibits dMyc Expression Posttranscriptionally (A and B) dFOXO localizes to the cytoplasm of neuroblasts (arrowheads) and Prospero-expressing daughter cells (open arrowheads) in control (A) and brat brains (B). (C) dMyc is detected in neuroblasts (arrowhead) but not daughter cells (open arrowhead) in control tissue while it is ubiquitously expressed in brat mutant cells. (D) Clonal induction of dMyc and lacZ using actinGal4 reveals expression of dMyc in neuroblasts (arrowheads) but not daughter cells (open arrowheads). (E) Overexpressed dMyc is detected in brat mutant cells. (F and G) Nucleolar size in control (1407-Gal4 expressing CD8-GFP) and brat brains. Neuroblasts are marked by arrowheads and daughter cells by open arrowheads. Note that the CD8-GFP panel used for cell size measurement is not shown in (F). (H) Quantification of the ratio between nucleolar and cellular diameter. Genotypes are brat150/brat192 (B and G); hsFlp, C155-Gal4, UAS-CD8-GFP; FRT40A, brat150/FRT40A, brat150 (C); actinGal4, UAS-dMyc, UASlacZ (D); hsFlp, C155-Gal4, UAS-CD8-GFP; FRT40A, brat150/FRT40A, brat150; UAS-dMyc (E); and 1407-Gal4, UAS-CD8-GFP (F). Scale bars are 10 mm. **: p < 0.05 (Student’s t test); error bars are standard deviation of the mean.

neuroblast overgrowth or change the size ratio between neuroblasts and GMCs significantly (Figure S1). Thus, the brat mutant phenotype does not seem to be caused by altered activity of the InsR pathway. Besides the InsR pathway, the transcription factor dMyc has been described as an important regulator of cell-cycle progression and cell growth. In control brains (Figure 6C and data not shown), dMyc is expressed in neuroblasts (Figure 6C0 , arrowhead) but not in the differen-

1248 Cell 124, 1241–1253, March 24, 2006 ª2006 Elsevier Inc.

tiating daughter cells (Figure 6C0 , open arrowheads). In brat mutant clones, however, dMyc is found in all cells (Figure 6C0 ), suggesting that Brat directly or indirectly inhibits dMyc expression. To test whether dMyc is regulated transcriptionally or posttranscriptionally, we overexpressed dMyc in the larval neuroblast lineage using the UAS-Gal4 system. We used a line in which the actin promoter and the Gal4 coding region are separated by a stop cassette which can be removed by Flp-mediated

recombination. When recombined in a single neuroblast, this line drives lacZ expression in all daughter cells (Figure 6D). Overexpressed dMyc protein, however, is detected specifically in the neuroblast (Figure 6D0 , arrowhead), indicating that it is inhibited posttranscriptionally in the differentiating daughter cells. When overexpressed in a brat mutant clone, however, dMyc is detected in all daughter cells (Figure 6E0 ), indicating that Brat is required for the posttranscriptional inhibition of dMyc in one of the two daughter cells of larval neuroblasts. dMyc regulates cell growth by controlling the synthesis of ribosomal RNA in the nucleolus, thereby influencing ribosome biogenesis (Grewal et al., 2005). Interestingly, brat as well as its C. elegans homolog ncl-1 regulate nucleolar growth (Frank et al., 2002; Frank and Roth, 1998). To test whether Brat could regulate cell growth by affecting nucleolar biogenesis in neuroblasts, we analyzed the nucleolus using the marker Fibrillarin. In wild-type neuroblasts, the nucleolus is consistently larger than in GMCs (Figure 6F). In brat mutants, however, all cells have an enlarged nucleolus (Figure 6G). This is not an indirect consequence of the increase in cell size since the ratio between nucleolar and cellular diameter is significantly increased (Figure 6H). We conclude that Brat regulates the size of the nucleolus in the Drosophila brain, presumably by downregulating dMyc expression. Since dMyc expression and nucleolar size have been shown to correlate with ribosomal RNA synthesis and cell growth in other tissues (Frank et al., 2002), it is conceivable that Brat inhibits growth of the GMC by reducing its rate of protein synthesis. Prospero Acts as Tumor Suppressor in the Larval Brain Our data show a strong genetic interaction between Brat and Prospero in the embryonic CNS. To test whether Prospero also acts as a tumor suppressor in the larval CNS, we analyzed prospero mutant MARCM clones. As in brat mutant clones, we observe dramatic overproliferation when such clones are generated in the posterior half of the brain lobes. In contrast to Brat, however, loss of Prospero can also generate tumors in the ventral nerve cord (data not shown). Cells in prospero mutant tumors are enlarged and express the neuroblast markers Miranda and Worniu as well as the proliferation marker Cyclin E (Figures 7A and 7B and data not shown). Surprisingly, however, these clones downregulate the expression of Brat (Figure 7B0 ). Thus, in contrast to embryos where Brat and Prospero act redundantly to specify GMC fate, their expression is co-dependent in larval brain neuroblasts, and this may explain why mutations in either gene can induce the formation of neuroblast tumors. DISCUSSION How stem cells regulate their proliferation rate and how they control the balance between self-renewing and differentiating daughter cells is not understood. Our data reveal a molecular mechanism that controls self-renewal in Dro-

sophila larval neuroblasts. We show that the growth regulator Brat segregates asymmetrically during neuroblast division and inhibits self-renewal in one of the two daughter cells. Together with the asymmetrically segregating transcription factor Prospero, Brat ensures that this daughter cell will stop growing, exit the cell cycle, and differentiate into neurons. In brat or prospero mutants, or in lgl mutants, where Brat and Prospero are not asymmetrically segregated, both daughter cells proliferate leading to the formation of a brain tumor and death of the animal. These tumors are neoplastic and can be transplanted into the abdomen of wild-type flies where they overgrow, invade other tissues, and eventually kill the host (Caussinus and Gonzalez, 2005; Gateff, 1978). Brat and Prospero Regulate Stem Cell Self-Renewal Asymmetric cell division has been studied in the Drosophila central and peripheral nervous systems. In the peripheral nervous system, the determinants Numb and Neuralized segregate into one of the two daughter cells, and in their absence, this cell is transformed into its sister cell (Le Borgne and Schweisguth, 2003; Rhyu et al., 1994). In the embryonic central nervous system, Prospero acts as a segregating determinant, but in prospero mutants, many GMCs are still correctly specified. Our data suggest that this is because Prospero acts partially redundant with Brat. In embryos double mutant for prospero and brat, most GMCs expressing the marker Eve are missing and neuronal differentiation in the embryonic CNS is greatly impaired. These observations suggest that Brat and Prospero act together to specify GMC fate in Drosophila embryos (Figure 7C). Although cell-cycle markers are expressed longer and stronger in prospero (Li and Vaessin, 2000) and brat, prospero mutant (data not shown) embryos, uncontrolled overproliferation has not been described in Drosophila embryos so far. In larvae, however, both brat and prospero mutant neuroblasts can initiate tumor formation. We propose that this difference is due to distinct mechanisms of cell growth during the two stages. During embryogenesis, cell number increases dramatically but the total volume of the embryo remains constant. Embryonic neuroblasts therefore shrink with each division and they might exit the cell cycle simply because they become too small. Support for this model comes from mutations affecting cell size asymmetry during neuroblast divisions, like Gb13F (Fuse et al., 2003) or Ric-8 (Hampoelz et al., 2005): in these mutants, GMCs are larger, neuroblasts shrink faster and— as a consequence—divide less often. In larval neuroblasts, the situation is quite different. Several results indicate that larval neuroblasts grow significantly while cell growth is inhibited in GMCs. First, the total volume of clones generated from individual neuroblasts is several times more than the initial volume of the neuroblast (see for example Figure 4C). Second, the size of ‘‘old’’ and ‘‘young’’ (earlier and more recently generated) GMCs is approximately the same, indicating that GMCs do not grow significantly during clone formation. Third, larval

Cell 124, 1241–1253, March 24, 2006 ª2006 Elsevier Inc. 1249

Figure 7. Prospero Acts as a Tumor Suppressor in Larval Brains (A and B) Single neuroblast pros17 clones marked by the expression of CD8-GFP show expression of Miranda (A0 ) and Worniu (B0 0 ) and downregulation of Brat (B0 ) in daughter cells. Scale bars are 10 mm. (C) Model for the function of Brat and Prospero in embryonic and larval brain neuroblasts.

neuroblasts do not become progressively smaller during development indicating that the loss of cytoplasm from each division must be compensated for by growth. Taken together, these results suggest that larval neuroblasts might be more appropriate as a model for the control of self-renewal in stem cells. Our experiments show that the restriction of cell growth in the GMC requires the genes lgl, brat, and prospero (Figure 7C). While lgl seems to be required indirectly due to its role in asymmetric protein segregation, Prospero and Brat act in the GMC to regulate several important events: They repress neuroblast fate, inhibit cell-cycle progression, and prevent cell growth. Prospero is a homeodomain transcription factor, and the cell-cycle genes Cyclin A, Cyclin E, and Dacapo (the fly homolog of the CDK inhibitor p21) were shown to be among its transcriptional targets (Li and Vaessin, 2000; Liu et al., 2002). Similar to Drosophila Prospero, its vertebrate homolog Prox-1 has been shown to regulate cell-cycle genes, and loss of prox-1 leads to increased proliferation of retinal progenitor cells (Dyer et al., 2003).

1250 Cell 124, 1241–1253, March 24, 2006 ª2006 Elsevier Inc.

For Brat, two different functions have been described: First, it acts as a translational regulator of the gap-gene hunchback (Sonoda and Wharton, 2001). Hunchback is expressed in the embryonic nervous system but is not present in wild-type or brat mutant larval neuroblasts (data not shown) and is unlikely to be relevant for the growth regulatory activity of Brat. More likely, Brat acts through its well-described inhibitory activity on ribosomal RNA synthesis. Cells mutant for brat or its C. elegans homolog ncl-1 have larger nucleoli, more ribosomal RNA, and higher rates of protein synthesis, and these activities have been made responsible for the cell size increase that is observed in C. elegans and Drosophila brat mutant cells (Frank et al., 2002; Frank and Roth, 1998). Our data suggest that this second function of Brat is also linked to posttranscriptional gene regulation. We propose that Brat downregulates dMyc in one of the two daughter cells and thereby inhibits protein synthesis and cell growth. Whether Brat controls dMyc translation, protein stability, or RNA stability is unclear. Interestingly, the C. elegans Brat homolog ncl-1 has been identified as one of the

genes required for RNAi (Kim et al., 2005). Since the microRNA pathway was shown to be involved in regulation of Drosophila stem cell proliferation (Hatfield et al., 2005), differential regulation of this pathway in neuroblasts and GMCs by Brat could provide another explanation for its mutant phenotype. Brat Homologs and Cancer Stem Cells Brat is part of a protein family that is characterized by a Cterminal NHL domain, several zinc-finger like B boxes, and a coiled-coil region. While the vertebrate members of this family (TRIM-2, TRIM-3, and TRIM-32) (Arama et al., 2000) are not well characterized, the mutant phenotype of the two other Drosophila members (Dappled and Mei-P26) suggests a common function as tumor suppressors. Mutations in dappled cause melanomic tumors of the fat body (Rodriguez et al., 1996), and mei-P26 mutations lead to ovarian tumors (Page et al., 2000). While dappled tumors have not been well characterized, the mei-P26 phenotype has been attributed to overproliferation of undifferentiated germ cells. It is similar to—and genetically interacts with—bag of marbles (Page et al., 2000), a well-characterized repressor of proliferation in the daughter cells of germline stem cells (McKearin and Ohlstein, 1995). Thus, it is conceivable that proliferation control in stem cells is a common activity of NHL domain proteins. Recent evidence suggests that some human brain tumors contain stem cell-like neural progenitors that are responsible for tumor formation (Galli et al., 2004; Singh et al., 2003, 2004). Together with the identification of stem cell-like subpopulations in leukaemia, multiple myeloma, and breast cancer, this has led to the so-called cancer stem cell hypothesis which proposes that only a small population of cells in a tumor have the ability to proliferate and self-renew (Al-Hajj and Clarke, 2004). This discovery suggests mechanisms for tumorigenesis other than the simple loss of proliferation control, in particular dedifferentiation of cells into additional stem cells and symmetric division of stem cells. Animal models for tumor stem cells are essential for developing new therapeutic approaches that target these mechanisms. Although Drosophila can only mimic some aspects of tumorigenesis, it might contribute to the identification of the molecular pathways operating in tumor stem cells. Human Lgl has already been implicated in tumor progression (Grifoni et al., 2004; Schimanski et al., 2005), and the characterization of Brat homologs will verify the relevance of Drosophila as a tumor stem cell model.

et al., 1999); UAS-mycDp110 and UAS-mycDp110D954A (Leevers et al., 1996); UAS-dMyc (from B. Edgar); UAS-CD8-GFP (Bloomington stock center); UAS-aPKCDN (Betschinger et al., 2003); MARCM stocks using C155Gal4 (Lee and Luo, 1999); yw, hsFlp; Act-FRT-yFRT-Gal4, UAS-lacZ(nls) for clonal overexpression. Transgenes were expressed in embryos using maternalGal4 (from D. St. Johnston) and in brains using 1407-Gal4 (Luo et al., 1994) inserted in the inscuteable promotor. Maternal zygotic null mutants were generated using ovoD and hsFlp. Constructs and Antibodies Miranda fragments and Brat and CG17593 coding sequences were PCR-amplified with primers containing attB recombination sites and recombined into pDONR221 (Invitrogen). TAP-tagged constructs were generated by recombining pDONR221-constructs into a hspCasper destination vector containing the TAP tag (Rigaut et al., 1999) and attR recombination sites. GFP-fusion constructs were generated by recombining inserts into a pUAST-vector with a N-terminal EGFP and attR recombination sites. Transgenic Mira RNAi was generated by cloning inverted repeats (nucleotides 940-1209) of the Miranda coding sequence into pMF3 (G. Dietzl and B. Dickson, personal communication). Brat antibodies (affinity-purified, 1:200) were raised in rabbits against a peptide containing amino acids 4–20 (CSPTPSLDSMR GGANSIE). Miranda antibodies (affinity-purified, 1:100) were generated in rabbits using a peptide containing amino acids 96–118 (CSLPQRLRFRPTPSHTDTATGSGS). Other antibodies are mouse anti-Pros (1:10, MR1A, Developmental Studies Hybridoma Bank), rabbit anti-Eve (1:2000, from M. Frasch), mouse anti-CNS axons (1:10, BP102, Developmental Studies Hybridoma Bank), rabbit anti-Zfh1 (1:500, from Z. Lai), mouse anti-GFP (1:100, Roche), rabbit anti-GFP (1:100, Abcam), rabbit anti-Mira (Shen et al., 1997), rat anti-Brat (Sonoda and Wharton, 2001), mouse anti-Wor (Cai et al., 2001), rat anti-CycE (1:500, from H. Richardson), mouse anti-CycE (1:10, from H. Richardson), mouse anti-BrdU (1:5, BD Pharmigen), mouse antiphosphoH3 (1:1000, Cell Signaling Technology), rat anti-CD8 (1:100, Caltag Laboratories), rabbit anti-dFOXO (Puig et al., 2003), chicken anti-lacZ (1:1000, Abcam), mouse anti-c-myc (1:100, Santa Cruz Biotechnology), mouse anti-dMyc (1:5, from B. Edgar), and mouse antiFibrillarin (1:10, Abcam). Immunohistochemistry Immunofluorescence experiments in embryos were carried out essentially as described (Betschinger et al., 2003). For larval brain stainings, larvae were dissected in PBS, and brains with attached ventral nerve cords were fixed for 20 min in 5% PFA; 0.2% Triton X-100 and processed like embryos. For hsFlp experiments, larvae were heatshocked for 1 hr at 37ºC on two successive days and dissected 3 or 4 days later. BrdU incorporation was essentially performed as described (Ceron et al., 2001). Briefly, larvae were dissected in Schneider’s medium (Gibco) and brains were incubated for 30 min with 37.5 mg BrdU/ml (BD Pharmigen) in Schneider’s medium. Brains were fixed for 3 min in modified Carnoy’s fixative (58% ethanol; 30% chloroform; 10% acetic acid; 2% formaldehyde) and treated with 75% EtOH for 30 min and 2N HCl for 40 min. After extensive washing, brains were processed by standard immunochemistry protocols.

EXPERIMENTAL PROCEDURES Fly Strains brat mutants have been described previously (Luschnig et al., 2004). Mutations were identified by sequencing homozygous mutant genomic DNA: in brat2L-150-11-2 (called brat150) and brat2L-192-9-1 (called brat192), Gln (926) and Gln (417) are changed into stop codons, respectively. Other fly strains are miraL44 (Matsuzaki et al., 1998); pros17 (Doe et al., 1991); FRT82B, pros17 (from V. Rodrigues); FRT40A, lgl1 and FRT40A, lgl4 (from F. Matsuzaki); FRT40A, dPTEN1 (Goberdhan

Biochemistry and Mass Spectrometry TAP-tagged constructs were expressed in 4–7 hr embryos by a 30 min heat shock at 37ºC. After 45 min recovery at room temperature, embryonic extracts were prepared in lysis buffer (20 mM HEPES pH 8.0; 150 mM NaCl; 0.5% NP40; 10% glycerol; 0.5 mM EDTA; 1 mM DTT; 1 mM PMSF; protease inhibitor cocktail [Roche]) and centrifuged for 15 min at 14000 rpm. The supernatant was incubated with chemically crosslinked (see below) IgG-Sepharose for 1 hr. After extensive washing with lysis buffer, beads were equilibrated in lysis buffer without

Cell 124, 1241–1253, March 24, 2006 ª2006 Elsevier Inc. 1251

detergent. Beads were eluted three times with 1 bed volume of 50 mM HCl and eluates were pooled. To avoid elution of the IgG light chain by HCl treatment, light chains were crosslinked to heavy chains. For this, IgG-Sepharose (Amersham Biosciences) was equilibrated in 0.2 M sodium borate pH 9.0 and treated with 40 mM dimethyl pimelimidate (Sigma) for 1 hr. Excess binding sites were blocked with 1 M Tris pH 8.0 and beads were pre-eluted with 50 mM HCl. Eluates were neutralized with Tris pH 8.0, reduced with DTT, carboxymethylated using iodoacetamide, and digested with trypsin overnight. After acidifying with TFA, solutions were separated by nano HPLC chromatography (LC-Packings, Netherland) on a PepMap C 18 column. The eluate of the column was applied online to a LTQ linear ion trap mass spectrometer (Thermo Finnigan). Mass data on all peptides and their fragmentation patterns were analyzed using MASCOT software (Matrix Science). Identified proteins with a MASCOT probability score above 40 obtained in two independent experiments were related in a database (Access, Microsoft) and scored for proteins present in the Miranda cargo domain but in none of the control construct purifications (scores are: Brat, 704/418.6; Prospero, 182/264.6; Staufen, 113.4/73.7; CG17593, 67.7/100.5). Supplemental Data Supplemental data include one figure and can be found with this article online at http://www.cell.com/cgi/content/full/124/6/1241/DC1/. ACKNOWLEDGMENTS We wish to thank J. Hutchins, R. Neumueller, and F. Wirtz-Peitz for comments on the manuscript; D. Berdnik, C. Berger, C. Doe, B. Edgar, J. Krauss, S. Leevers, S. Luschnig, F. Matsuzaki, K. Moberg, B. Moussian, O. Puig, H. Richardson, V. Rodrigues, B. Seraphin, R. Wharton, C. Wilson, X. Yhang, the Developmental Studies Hybridoma Bank (DSHB), and the Bloomington Drosophila Stock Center for antibodies, constructs, and flystocks; A. Hutterer for contributing SOP cell stainings; I. Steinmacher and R. Imre for help with mass spectrometry; P. Steinlein for programming the relational database; M. Madalinski for peptide synthesis and affinity purification; G. Dietzl and B. Dickson for providing the Miranda RNAi construct; members of the Knoblich lab for discussions; K. Paiha for bio-optics support; and E. Kleiner and Y.N. Zhang for technical assistance. Work in J.A.K.’s lab is supported by the Austrian Academy of Sciences and the FWF. Received: September 29, 2005 Revised: December 1, 2005 Accepted: January 3, 2006 Published: March 23, 2006 REFERENCES Akong, K., McCartney, B.M., and Peifer, M. (2002). Drosophila APC2 and APC1 have overlapping roles in the larval brain despite their distinct intracellular localizations. Dev. Biol. 250, 71–90.

Betschinger, J., Eisenhaber, F., and Knoblich, J.A. (2005). Phosphorylation-induced autoinhibition regulates the cytoskeletal protein lethal (2) giant larvae. Curr. Biol. 15, 276–282. Bossing, T., Udolph, G., Doe, C.Q., and Technau, G.M. (1996). The embryonic central nervous system lineages of Drosophila melanogaster. I. Neuroblast lineages derived from the ventral half of the neuroectoderm. Dev. Biol. 179, 41–64. Broadus, J., Fuerstenberg, S., and Doe, C.Q. (1998). Staufen-dependent localization of prospero mRNA contributes to neuroblast daughter-cell fate. Nature 391, 792–795. Cai, Y., Chia, W., and Yang, X. (2001). A family of snail-related zinc finger proteins regulates two distinct and parallel mechanisms that mediate Drosophila neuroblast asymmetric divisions. EMBO J. 20, 1704–1714. Caussinus, E., and Gonzalez, C. (2005). Induction of tumor growth by altered stem-cell asymmetric division in Drosophila melanogaster. Nat. Genet. 37, 1125–1129. Ceron, J., Gonzalez, C., and Tejedor, F.J. (2001). Patterns of cell division and expression of asymmetric cell fate determinants in postembryonic neuroblast lineages of Drosophila. Dev. Biol. 230, 125–138. Doe, C.Q., Chu-LaGraff, Q., Wright, D.M., and Scott, M.P. (1991). The prospero gene specifies cell fates in the Drosophila central nervous system. Cell 65, 451–464. Dyer, M.A., Livesey, F.J., Cepko, C.L., and Oliver, G. (2003). Prox1 function controls progenitor cell proliferation and horizontal cell genesis in the mammalian retina. Nat. Genet. 34, 53–58. Frank, D.J., and Roth, M.B. (1998). ncl-1 is required for the regulation of cell size and ribosomal RNA synthesis in Caenorhabditis elegans. J. Cell Biol. 140, 1321–1329. Frank, D.J., Edgar, B.A., and Roth, M.B. (2002). The Drosophila melanogaster gene brain tumor negatively regulates cell growth and ribosomal RNA synthesis. Development 129, 399–407. Fuerstenberg, S., Peng, C.Y., Alvarez-Ortiz, P., Hor, T., and Doe, C.Q. (1998). Identification of Miranda protein domains regulating asymmetric cortical localization, cargo binding, and cortical release. Mol. Cell. Neurosci. 12, 325–339. Fuse, N., Hisata, K., Katzen, A.L., and Matsuzaki, F. (2003). Heterotrimeric g proteins regulate daughter cell size asymmetry in Drosophila neuroblast divisions. Curr. Biol. 13, 947–954. Galli, R., Binda, E., Orfanelli, U., Cipelletti, B., Gritti, A., De Vitis, S., Fiocco, R., Foroni, C., Dimeco, F., and Vescovi, A. (2004). Isolation and characterization of tumorigenic, stem-like neural precursors from human glioblastoma. Cancer Res. 64, 7011–7021. Gateff, E. (1978). Malignant neoplasms of genetic origin in Drosophila melanogaster. Science 200, 1448–1459. Goberdhan, D.C., Paricio, N., Goodman, E.C., Mlodzik, M., and Wilson, C. (1999). Drosophila tumor suppressor PTEN controls cell size and number by antagonizing the Chico/PI3-kinase signaling pathway. Genes Dev. 13, 3244–3258.

Al-Hajj, M., and Clarke, M.F. (2004). Self-renewal and solid tumor stem cells. Oncogene 23, 7274–7282.

Grewal, S.S., Li, L., Orian, A., Eisenman, R.N., and Edgar, B.A. (2005). Myc-dependent regulation of ribosomal RNA synthesis during Drosophila development. In Nat Cell Biol, pp. 295–302.

Arama, E., Dickman, D., Kimchie, Z., Shearn, A., and Lev, Z. (2000). Mutations in the beta-propeller domain of the Drosophila brain tumor (brat) protein induce neoplasm in the larval brain. Oncogene 19, 3706–3716.

Grifoni, D., Garoia, F., Schimanski, C.C., Schmitz, G., Laurenti, E., Galle, P.R., Pession, A., Cavicchi, S., and Strand, D. (2004). The human protein Hugl-1 substitutes for Drosophila Lethal giant larvae tumour suppressor function in vivo. Oncogene 23, 8688–8694.

Betschinger, J., and Knoblich, J.A. (2004). Dare to be different: asymmetric cell division in Drosophila, C. elegans and vertebrates. Curr. Biol. 14, R674–R685.

Hadorn, E. (1938). Die Degeneration der Imaginalscheiben bei letalen Drosophila-Larven der Mutation ‘Lethal-giant’. Rev. Suisse Zool. 46, 425–429.

Betschinger, J., Mechtler, K., and Knoblich, J.A. (2003). The Par complex directs asymmetric cell division by phosphorylating the cytoskeletal protein Lgl. Nature 422, 326–330.

Hampoelz, B., Hoeller, O., Bowman, S.K., Dunican, D., and Knoblich, J.A. (2005). Drosophila Ric-8 is essential for plasma-membrane localization of heterotrimeric G proteins. Nat. Cell Biol. 7, 1099–1105.

1252 Cell 124, 1241–1253, March 24, 2006 ª2006 Elsevier Inc.

Hatfield, S.D., Shcherbata, H.R., Fischer, K.A., Nakahara, K., Carthew, R.W., and Ruohola-Baker, H. (2005). Stem cell division is regulated by the microRNA pathway. Nature 435, 974–978.

Petronczki, M., and Knoblich, J.A. (2001). DmPAR-6 directs epithelial polarity and asymmetric cell division of neuroblasts in Drosophila. Nat. Cell Biol. 3, 43–49.

Hirata, J., Nakagoshi, H., Nabeshima, Y., and Matsuzaki, F. (1995). Asymmetric segregation of the homeodomain protein Prospero during Drosophila development. Nature 377, 627–630.

Puig, O., Marr, M.T., Ruhf, M.L., and Tjian, R. (2003). Control of cell number by Drosophila FOXO: downstream and feedback regulation of the insulin receptor pathway. Genes Dev. 17, 2006–2020.

Ikeshima-Kataoka, H., Skeath, J.B., Nabeshima, Y., Doe, C.Q., and Matsuzaki, F. (1997). Miranda directs Prospero to a daughter cell during Drosophila asymmetric divisions. Nature 390, 625–629.

Rhyu, M.S., Jan, L.Y., and Jan, Y.N. (1994). Asymmetric distribution of numb protein during division of the sensory organ precursor cell confers distinct fates to daughter cells. Cell 76, 477–491.

Kim, J.K., Gabel, H.W., Kamath, R.S., Tewari, M., Pasquinelli, A., Rual, J.F., Kennedy, S., Dybbs, M., Bertin, N., Kaplan, J.M., et al. (2005). Functional genomic analysis of RNA interference in C. elegans. Science 308, 1164–1167.

Rigaut, G., Shevchenko, A., Rutz, B., Wilm, M., Mann, M., and Seraphin, B. (1999). A generic protein purification method for protein complex characterization and proteome exploration. Nat. Biotechnol. 17, 1030–1032.

Knoblich, J.A., Jan, L.Y., and Jan, Y.N. (1995). Asymmetric segregation of Numb and Prospero during cell division. Nature 377, 624–627.

Rodriguez, A., Zhou, Z., Tang, M.L., Meller, S., Chen, J., Bellen, H., and Kimbrell, D.A. (1996). Identification of immune system and response genes, and novel mutations causing melanotic tumor formation in Drosophila melanogaster. Genetics 143, 929–940.

Le Borgne, R., and Schweisguth, F. (2003). Unequal segregation of Neuralized biases Notch activation during asymmetric cell division. Dev. Cell 5, 139–148. Lee, T., and Luo, L. (1999). Mosaic analysis with a repressible cell marker for studies of gene function in neuronal morphogenesis. Neuron 22, 451–461. Leevers, S.J., Weinkove, D., MacDougall, L.K., Hafen, E., and Waterfield, M.D. (1996). The Drosophila phosphoinositide 3-kinase Dp110 promotes cell growth. EMBO J. 15, 6584–6594. Li, L., and Vaessin, H. (2000). Pan-neural Prospero terminates cell proliferation during Drosophila neurogenesis. Genes Dev. 14, 147–151. Liu, T.H., Li, L., and Vaessin, H. (2002). Transcription of the Drosophila CKI gene dacapo is regulated by a modular array of cis-regulatory sequences. Mech. Dev. 112, 25–36. Luo, L., Liao, Y.J., Jan, L.Y., and Jan, Y.N. (1994). Distinct morphogenetic functions of similar small GTPases: Drosophila Drac1 is involved in axonal outgrowth and myoblast fusion. Genes Dev. 8, 1787–1802. Luschnig, S., Moussian, B., Krauss, J., Desjeux, I., Perkovic, J., and Nusslein-Volhard, C. (2004). An F1 genetic screen for maternal-effect mutations affecting embryonic pattern formation in Drosophila melanogaster. Genetics 167, 325–342. Matsuzaki, F., Ohshiro, T., Ikeshima-Kataoka, H., and Izumi, H. (1998). miranda localizes staufen and prospero asymmetrically in mitotic neuroblasts and epithelial cells in early Drosophila embryogenesis. Development 125, 4089–4098. McKearin, D., and Ohlstein, B. (1995). A role for the Drosophila bag-ofmarbles protein in the differentiation of cystoblasts from germline stem cells. Development 121, 2937–2947. Michor, F., Hughes, T.P., Iwasa, Y., Branford, S., Shah, N.P., Sawyers, C.L., and Nowak, M.A. (2005). Dynamics of chronic myeloid leukaemia. Nature 435, 1267–1270. Ohshiro, T., Yagami, T., Zhang, C., and Matsuzaki, F. (2000). Role of cortical tumour-suppressor proteins in asymmetric division of Drosophila neuroblast. Nature 408, 593–596. Page, S.L., McKim, K.S., Deneen, B., Van Hook, T.L., and Hawley, R.S. (2000). Genetic studies of mei-P26 reveal a link between the processes that control germ cell proliferation in both sexes and those that control meiotic exchange in Drosophila. Genetics 155, 1757–1772. Peng, C.Y., Manning, L., Albertson, R., and Doe, C.Q. (2000). The tumour-suppressor genes lgl and dlg regulate basal protein targeting in Drosophila neuroblasts. Nature 408, 596–600.

Rolls, M.M., Albertson, R., Shih, H.P., Lee, C.Y., and Doe, C.Q. (2003). Drosophila aPKC regulates cell polarity and cell proliferation in neuroblasts and epithelia. J. Cell Biol. 163, 1089–1098. Schimanski, C.C., Schmitz, G., Kashyap, A., Bosserhoff, A.K., Bataille, F., Schafer, S.C., Lehr, H.A., Berger, M.R., Galle, P.R., Strand, S., and Strand, D. (2005). Reduced expression of Hugl-1, the human homologue of Drosophila tumour suppressor gene lgl, contributes to progression of colorectal cancer. Oncogene 24, 3100–3109. Schober, M., Schaefer, M., and Knoblich, J.A. (1999). Bazooka recruits Inscuteable to orient asymmetric cell divisions in Drosophila neuroblasts. Nature 402, 548–551. Schuldt, A.J., Adams, J.H.J., Davidson, C.M., Micklem, D.R., Haseloff, J., Johnston, D.S., and Brand, A.H. (1998). Miranda mediates asymmetric protein and RNA localization in the developing nervous system. Genes Dev. 12, 1847–1857. Shen, C.P., Jan, L.Y., and Jan, Y.N. (1997). Miranda is required for the asymmetric localization of Prospero during mitosis in Drosophila. Cell 90, 449–458. Shen, C.P., Knoblich, J.A., Chan, Y.M., Jiang, M.M., Jan, L.Y., and Jan, Y.N. (1998). Miranda as a multidomain adapter linking apically localized Inscuteable and basally localized Staufen and Prospero during asymmetric cell division in Drosophila. Genes Dev. 12, 1837–1846. Singh, S.K., Clarke, I.D., Terasaki, M., Bonn, V.E., Hawkins, C., Squire, J., and Dirks, P.B. (2003). Identification of a cancer stem cell in human brain tumors. Cancer Res. 63, 5821–5828. Singh, S.K., Hawkins, C., Clarke, I.D., Squire, J.A., Bayani, J., Hide, T., Henkelman, R.M., Cusimano, M.D., and Dirks, P.B. (2004). Identification of human brain tumour initiating cells. Nature 432, 396–401. Sonoda, J., and Wharton, R.P. (2001). Drosophila brain tumor is a translational repressor. Genes Dev. 15, 762–773. Vaessin, H., Grell, E., Wolff, E., Bier, E., Jan, L.Y., and Jan, Y.N. (1991). prospero is expressed in neuronal precursors and encodes a nuclear protein that is involved in the control of axonal outgrowth in Drosophila. Cell 67, 941–953. Wodarz, A., Ramrath, A., Kuchinke, U., and Knust, E. (1999). Bazooka provides an apical cue for Inscuteable localization in Drosophila neuroblasts. Nature 402, 544–547. Wodarz, A., Ramrath, A., Grimm, A., and Knust, E. (2000). Drosophila atypical protein kinase C associates with Bazooka and controls polarity of epithelia and neuroblasts. J. Cell Biol. 150, 1361–1374.

Cell 124, 1241–1253, March 24, 2006 ª2006 Elsevier Inc. 1253